sign in sign up
<<
Home , 2016 San Antonio Open, Bill Woodcock, Ello (social network), Esther (Handel), Herd behavior, Internet in Estonia

102725 Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized world development report

DIGITAL DIVIDENDS

A Group Flagship Report

world development report

DIGITAL DIVIDENDS © 2016 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; : www.worldbank.org Some rights reserved 1 2 3 4 19 18 17 16 This work is a product of the staff of The World Bank with external contributions. The fi ndings, interpretations, and conclusions expressed in this work do not necessarily refl ect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifi cally reserved.

Rights and Permissions

This work is available under the Creative Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons.org /licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: World Bank. 2016. World Development Report 2016: Digital Dividends. Washington, DC: World Bank. doi:10.1596/978-1-4648-0671-1. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an offi cial World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, fi gures, or images.

All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: [email protected].

ISSN, ISBN, e-ISBN, and DOI:

Softcover ISSN: 0163-5085 ISBN: 978-1-4648-0671-1 e-ISBN: 978-1-4648-0672-8 DOI: 10.1596/978-1-4648-0671-1

Hardcover ISSN: 0163-5085 ISBN: 978-1-4648-0728-2 DOI: 10.1596/978-1-4648-0728-2

Interior design: Reboot (www.reboot.org), New York, New York, and George Kokkinidis, Design Language, Brooklyn, New York

Cover photo: This 2013 World Press Photo of the Year shows migrants crowding the night shore of in an attempt to capture inexpensive cellphone signals from neighboring Somalia. © John Stanmeyer/National Geographic Creative. Used with the permission of John Stanmeyer/National Geographic Creative. Further permission required for reuse.

Library of Congress Cataloging-in-Publication data has been requested. Contents

xiii Foreword xv Acknowledgments xix Abbreviations

1 Overview: Strengthening the analog foundation of the digital revolution 5 Digital transformations—digital divides 8 How the internet promotes development 11 The dividends: Growth, jobs, and delivery 18 The risks: Concentration, inequality, and control 25 Making the internet universal, affordable, open, and safe 29 Analog complements for a digital economy 36 Global cooperation to solve global problems 38 Reaping digital dividends for everyone 38 Notes 39 References 42 Spotlight 1: How the internet promotes development

49 Part 1: Facts and analysis 50 Chapter 1: Accelerating growth 51 Connected businesses 55 More , higher productivity, and greater 70 Digital technologies can lead fi rms and to diverge 73 The nexus of technology and regulation 80 The future of markets 82 Notes 85 References 90 Sector focus 1: Agriculture 94 Spotlight 2: Digital fi nance

100 Chapter 2: Expanding opportunities 101 Connected people 104 Creating jobs, boosting labor productivity, and benefi ting consumers 118 Labor polarization can lead to greater inequality 120 The race between skills and technology v vi CONTENTS

130 The future of jobs 135 Notes 138 References 146 Sector focus 2: Education 148 Spotlight 3:

152 Chapter 3: Delivering services 153 Connected governments 155 Greater state capability and citizen participation 171 Digital technologies too often fail to empower citizens 177 The gap between technology and institutions 181 The future of public services 181 Notes 183 References 190 Sector focus 3: e-health 194 Spotlight 4: Digital identity

199 Part 2: Policies 200 Chapter 4: Sectoral policies 200 Making the internet universal, affordable, open, and safe 203 Shaping the digital economy 204 Supply-side policies: Availability, accessibility, and affordability 221 Demand-side policies: Open and safe internet use 228 Promoting the digital economy 232 Notes 235 References 240 Sector focus 4: Smart 244 Spotlight 5: The data revolution

248 Chapter 5: National priorities 248 Analog foundations for a digital economy 249 The interdependence between technology and complements 253 Regulations: Helping businesses connect and compete 258 Skills: Making the internet work for everyone 272 Institutions: Connecting for a capable and accountable government 279 Digital safeguards 281 Notes 282 References 288 Sector focus 5: Energy

292 Chapter 6: Global cooperation 292 Internet 297 Toward a global digital market 303 Leveraging information for 317 Notes 318 References 322 Sector focus 6: Environmental management 326 Spotlight 6: Six digital technologies to watch CONTENTS vii

Boxes O.1 5 Frequently asked questions: The Report 2.2 110 The economics of online outsourcing at a glance 2.3 112 Expanding opportunities through O.2 10 e-commerce with Chinese online music characteristics: Inclusion, effi ciency, 2.4 114 Bridging the disability divide through and innovation in Taobao villages digital technologies O.3 15 Bridging the disability divide through 2.5 115 Using digital technologies to match digital technologies workers with jobs: Souktel in West O.4 16 Digital dividends and the bottom billion Bank and Gaza O.5 20 What “Likes” reveal—the 2.6 116 The impact of digital technologies on convenience-privacy trade-off O.6 26 Nailing Jell-O to the wall—restrictions 2.7 123 Skills wanted: Key concepts on the fl ow of information 2.8 129 Concerns about technological O.7 27 Is the internet a ? are not new O.8 27 The four digital enablers 2.9 133 The challenge of keeping up with new technologies in O.9 29 Technology and complements: Lessons from academic research 2.10 134 Digital technologies and economic opportunities: A gender lens O.10 32 Opening the M-Pesa mobile platform to competition 3.1 158 Digital technology and crisis management O.11 33 Mobilizing technology in teaching in Rio’s Educopedia 3.2 159 Empowering women through digitally enabled social programs O.12 35 Can continuous monitoring and small sanctions improve provider 3.3 160 Targeted public transit benefi ts in performance? Bogotá O.13 37 European Union: A fragmented market 3.4 161 Streamlining services through one-stop for digital trade service centers S1.1 45 Three ways in which the internet 3.5 165 The high failure rate of e-government promotes development projects 1.1 56 Tracing back growth to a single, 3.6 169 Digitally enabled teacher management new technology suffers from severe in private schools measurement problems 3.7 173 Improving the integrity of elections 1.2 57 Is this time different? Predicting through crowdsourcing and labor productivity growth at the collaboration technological frontier based on lessons 3.8 178 Digital technologies can strengthen from past industrial revolutions control 1.3 59 Is the internet reshaping economic 4.1 201 Policy challenges for digital geography? Not yet. development 1.4 61 Successful online platforms account for 4.2 204 Is the internet a public good? local context and institutions 4.3 207 Fragile states, resilient digital 1.5 63 The growth impact is largest when economies fi rms in traditional sectors use digital 4.4 209 How public-private partnership helped technologies to modernize their build the internet backbone in the business Republic of Korea 1.6 65 Do digital technologies embed 4.5 212 The last (1,000) mile(s) productivity ? 4.6 215 : An early pioneer of 1.7 69 Much of the benefi t from the internet is spectrum auctions unmeasured 4.7 219 How better ICT data can lead to cheaper S2.1 95 Innovations in digital payments services S2.2 98 Technology can help unveil illicit 4.8 223 The costs of cybercrime money fl ows 4.9 229 Tech hubs in 2.1 109 Business process outsourcing and jobs in the Philippines: Opportunities and 4.10 231 as a startup nation challenges from technological change S5.1 245 “Big data” and open data in action viii CONTENTS

5.1 250 Three ideas about the interaction O.4 6 Digital transformation in action between technology and its O.5 8 The internet remains unavailable, complements inaccessible, and unaffordable to a 5.2 257 Mobile money: A success story and yet majority of the world’s population a regulatory minefi eld O.6 9 The in access is high in 5.3 261 The impact of digital technologies on Africa, and the divide in capability is cognitive capacities and high in the European Union 5.4 261 One Laptop per Child: Strengthening O.7 9 The internet promotes development analog foundations and careful through three main mechanisms evaluation O.8 11 Many digital transactions involve all 5.5 263 Khan Academy: A supplemental three mechanisms and a two-sided educational resource in and outside the market classroom O.9 12 How the three mechanisms apply to 5.6 264 Using digital technologies to foster businesses, people, and governments collaboration and learning: Rio de Janeiro’s Educopedia O.10 13 The size of the ICT sector and its contribution to GDP growth is still 5.7 265 Emerging lessons from digital relatively modest programs O.11 13 Vietnamese fi rms using e-commerce 5.8 267 Building new economy skills: Escuela have higher TFP growth, 2007–12 Nueva in and O.12 18 More complaints were resolved more 5.9 269 Building modern skills: Game-based quickly in the water after learning and “gamifying” education the introduction of digital customer 5.10 271 Massive open online courses (MOOCs): feedback A promising tool for lifelong learning O.13 18 Without strong analog complements, 5.11 276 Increasing the impact of e-government opportunities may turn into risks systems O.14 19 Factors explaining the lower adoption 5.12 277 Regular, small-stakes monitoring of digital technologies by businesses 5.13 279 ’s X-Road O.15 21 Labor shares of national income are 6.1 294 Categories of stakeholders in internet falling in many countries, including governance some developing countries 6.2 297 European Union: A fragmented market O.16 21 Falling labor shares in national income for digital products are associated with rising inequality 6.3 307 The Social Observatory and P-tracking O.17 22 The labor market is becoming more polarized in many developing countries 6.4 309 ICTs and the Sustainable Development Goals O.18 23 From a technological standpoint, 6.5 310 Digital Green: “How-to” for two-thirds of all jobs are susceptible to agriculture and health automation in the developing world, but the effects are moderated by lower 6.6 312 Disaster risk management and slower technology adoption 6.7 314 Can “big data” provide real-time, O.19 25 Internet voting can increase voter geographically detailed information on participation but can be biased toward ? more privileged groups 6.8 316 Disruptive approaches to development BO.6.1 26 Autocratic governments have promoted S6.1 329 Using cellphones for medical diagnosis, e-government while censoring the thanks to 3D printing internet O.20 28 A policy framework for improving Figures connectivity O.1 3 Digital technologies have spread rapidly O.21 30 The quality of complements and in much of the world technology rises with incomes O.2 3 The pessimism concerning the global O.22 31 Policy priorities for countries that outlook is not because of digital are emerging, transitioning, or technologies, but in spite of them transforming O.3 4 Why digital dividends are not spreading O.23 33 The types of skills needed in a modern rapidly—and what can be done economy CONTENTS ix

O.24 35 Digital safeguards in the WDR’s platforms varies substantially among framework sectors and countries in , 2014 S1.1 44 Internet users trade personal data for 1.17 73 Mobile money markets are often useful services controlled by one or two operators, 2014 SB1.1.1 45 A graphic representation of how the 1.18 75 Firms’ ICT as a internet promotes development share of GDP in several countries are comparable to those of U.S. 1.1 51 A framework for the internet and fi rms, but they invest much less in complementary skills and 1.2 52 More fi rms are using broadband reorganization, 2006 internet 1.19 76 Many countries still have poor postal 1.3 53 Many advanced digital technologies delivery systems have not yet diffused across fi rms in 1.20 78 Young fi rms use the internet more high-income countries, 2014 intensively in high-income countries— 1.4 54 Higher-productivity fi rms are more old fi rms, in low-income countries, likely to use the internet, 2010–14 2010–14 1.5 54 African fi rms using the internet are 1.21 78 Domestic fi rms use the internet more more productive, 2014 intensively when they face foreign competition, 2010–14 1.6 55 Larger fi rms use the internet more intensively across all income groups, 1.22 79 Firms in Mexico facing higher import 2006–14 competition from use more ICTs more productively 1.7 57 ICT capital accounted for nearly one- fi fth of global growth, 1995–2014 1.23 80 Restrictive product market regulations in services and higher nontariff B1.2.1 58 U.S. labor productivity during the technical barriers to trade in electrifi cation era (1890–1940) shares are associated with remarkably common patterns with the lower ICT use, 2010–14 ICT era (1970–2012) 1.24 81 The dominance of a few politically 1.8 60 The internet enables more fi rms to connected fi rms stifl es competition and reach new markets, 2001–12 innovation in Morocco, 2004 and 2007 B1.5.1 63 The ICT sector accounts for 4–7 percent F1.1 91 Introducing mobile phone service of GDP in most OECD countries, 2011 reduces dispersion in local 1.9 65 Vietnamese fi rms using e-commerce markets have higher TFP growth, 2007–12 S2.1 95 ’s M-Pesa payment system 1.10 67 Firm entry rates rose after countries reached 80 percent of introduced online registration systems, within four years 2006–12 2.1 101 A framework for the internet and 1.11 67 Two out of three fi rms report economic opportunities competitive pressure from digital 2.2 103 All regions are converging in mobile innovations, 2014 phone access, but South and Sub-Saharan Africa are falling behind in 1.12 68 of taxi medallions have started to decline following the entry of on-demand services and reduced 2.3 103 How people use mobile phones and the demand for traditional taxis 1.13 71 Firms’ use of online banking varies 2.4 104 The digital divide within countries substantially across countries at remains wide, especially in internet use comparable incomes, 2003–06 and 2.5 107 Employment in the ICT sector and in 2008–13 ICT occupations remains small 1.14 71 Firms’ use of the internet varies among 2.6 108 In Brazil, internet and software use six African countries, 2014 by fi rms throughout the economy is 1.15 72 The share of fi rms in the retail sector associated with higher earnings that sell their products online varies B2.2.1 110 Online labor markets provide work substantially among Latin American and fairly good pay for workers in countries, 2010 developing countries 1.16 72 The share of fi rms using integrated 2.7 111 Online work expands women’s access to customer relationship management work x CONTENTS

2.8 111 Flexibility in hours worked and the 2.27 133 The less educated and the bottom 40 ability to work from home are the percent of the main advantages of online work, but are most vulnerable to technological relatively poor pay and lack of career changes in the labor market prospects are concerns 3.1 153 A framework for digital technologies 2.9 113 Returns to education remain high and government service delivery despite signifi cant expansion in the 3.2 153 Low-income countries have invested supply of educated workers, especially heavily in e-government for tertiary education 3.3 154 The priority in low-income countries 2.10 113 Returns to education are particularly has been core e-government systems, high in ICT-intensive occupations 2014 B2.5.1 115 Online platforms improve female labor 3.4 155 Governments use digital technologies force participation and access to higher- paying jobs more intensively than private sector fi rms, 2014 2.11 117 Mobile phones improve sense of security and save time 3.5 156 Citizen use of e-government in Europe depends highly on income, 2014 2.12 119 : Labor share in national income is falling, driven by routine 3.6 156 Mobile phones are the main channel for labor citizens to interact with governments in Africa, 2014 2.13 119 Labor shares in national income are falling in many countries, including 3.7 159 e-government systems increase the some developing countries transparency of government budgets, 2014 2.14 120 Falling labor shares in national income are associated with rising inequality 3.8 162 The likelihood of being visited and the number of visits by offi cials after 2.15 121 The labor market is becoming polarized e-fi ling was introduced vary in select in both developed and developing European and Central Asian countries countries 3.9 163 e-fi ling and e-payment on average 2.16 124 Employment is becoming more reduced the time required to prepare intensive in the use of digital and pay technologies 3.10 163 e-procurement has no effect on fi rms’ 2.17 124 Nonroutine skills are becoming more likelihood of bidding for a government important over time contract or of being solicited for a bribe 2.18 125 In developing countries, one-third of in select European and Central Asian urban workers use digital technology countries at work B3.5.1 165 Success rate of large public sector ICT 2.19 125 Employment becomes more intensive projects in ICT use as economies grow B3.5.2 165 Performance of World Bank–funded 2.20 126 Lack of ICT skills is often a constraint to ICT projects employment 3.11 166 More complaints were resolved more 2.21 127 Nonroutine analytical and quickly in the Nairobi water utility after socioemotional skills are becoming the introduction of digital customer more important, especially in jobs feedback performed by younger cohorts 3.12 167 Citizens using ’s national 2.22 128 New economy skills, beyond levels of feedback portal (LAPOR) have few education, pay off actionable complaints, and these are 2.23 128 Digital technologies go hand in hand mostly for private , 2015 with nonroutine new economy skills 3.13 170 Government offi cials in Indonesia 2.24 129 From a technological standpoint, and the Philippines have generally two-thirds of all jobs are susceptible to low opinions of resource automation in the developing world, management practices but the effects are moderated by lower 3.14 172 Democracy has spread, but so have wages and slower technology adoption election irregularities—digital 2.25 131 The interaction between technology technologies can help make elections and jobs varies by occupation freer and fairer 2.26 132 The key policy challenge: Adapting the 3.15 175 Internet voting can increase voter skills agenda to expected labor market participation but can be biased toward disruptions more privileged groups CONTENTS xi

B3.8.1 178 Autocratic governments have promoted 5.8 260 Education that upgrades skills also e-government while censoring the facilitates higher adoption of digital internet technologies 3.16 179 Digital technology projects funded by 5.9 266 Even in advanced countries, youth are the World Bank are more successful often unable to think critically and in countries with higher-quality solve problems institutions 5.10 273 Countries with more accountable 3.17 180 Classifying public services and governments also adopt more digital activities as to their amenability technologies to improvement through digital technology 5.11 280 Digital safeguards in the WDR’s framework F3.1 191 Sequencing of e-health development in Montenegro B6.1.1 294 Stakeholders in S4.1 195 Different types of digital ID schemes 6.1 294 Concerns that have fueled the debate on across countries how the internet is governed B4.1.1 201 Global ICT access 6.2 297 The multistakeholder model of internet governance enjoys greater support than B4.1.2 202 Network buildout (subscriptions per other options 100 population) in OECD and low- and middle-income countries, 1990–2014 B6.2.1 298 Perceived barriers to buying over the internet in 2009 B4.3.1 207 Somalia’s rising mobile economy B6.2.2 298 Obstacles for enterprises not selling B4.4.1 209 Broadband in the Republic of Korea and online in 2013 other selected economies 6.3 299 Perception of U.S. fi rms on barriers to B4.5.1 213 The effect of history on internet prices, data fl ows as obstacles to trade, 2012 Pacifi c 6.4 300 A majority of respondents agree B4.6.1 216 How greater spectrum availability led to that their online data and personal lower prices in , 2003–09 information should be physically stored 4.1 217 Prices are falling for on a secure server in their own processing, storage, bandwidth, and 6.5 301 Changes in GDP, , and smartphones due to regulatory restrictions 4.2 218 If you want to make a mobile phone on data fl ows call, go to 6.6 305 The evolution of development by B4.10.1 231 How Israel stays ahead in high-tech information needs, 1973–2010 entrepreneurship 6.7 306 Proportion of international fi nancial F4.1 241 Smart cities: From data to intelligence institution projects with successful S5.1 245 World’s capacity to store information outcomes S5.2 245 Growth in telecommunications 6.8 306 High-quality M&E improves project capacity outcomes S5.3 246 Readiness, implementation, and impact S6.1 327 Gartner “hype” cycle applied to selected of open data digital technologies 5.1 249 Risks from digital technologies in the absence of complements Maps 5.2 251 Some services and sectors are more O.1 7 The internet is more evenly spread than amenable to digital technology than income others 1.1 53 Many more fi rms are using the internet 5.3 253 The quality of complements and in Vietnam technology rises with incomes 1.2 61 China’s destinations differ for 5.4 254 Regulations that encourage competition fi rms using online platforms, 2006 and also facilitate higher adoption of digital 2014 technologies 1.3 76 International online payment systems 5.5 255 Digital products are taxed as luxury for businesses are unavailable in many goods in some countries parts of Africa and Central Asia, 2012–14 5.6 256 Infrastructure complements 2.1 102 Mobile phones are the main source of 5.7 259 The types of skills needed in a modern connectivity in the developing world, economy but large gaps in internet access remain xii CONTENTS

B4.1.1 202 Price of mobile and fi xed broadband 2.5 132 Expected impacts of technological services change on employment and earnings B4.5.1 212 The effect of geography on internet S3.1 149 Relationships in different types of prices, Africa social media B4.6.1 215 Spectrum assignment in Latin America, 3.1 157 The impact of digital technology in MHz blocks on government capability to deliver 4.1 222 Evidence of internet content fi ltering services: A scorecard 4.2 226 National data protection and privacy 3.2 171 The impact of digital technology on laws and bills citizen empowerment: A scorecard 4.3 230 African tech hubs 3.3 176 Classifying the digital citizen engagement cases 6.1 296 A divided world: Country positions on the International Telecommunication 4.1 206 A policy framework for the supply of Regulations internet service 6.2 315 Availability of reports from weather B4.4.1 210 Broadband investment program, stations Republic of Korea 4.2 224 A basic framework for assessing the F6.1 323 Satellite estimates of average PM2.5 concentrations provide global coverage, costs of cybersecurity incidents 2010 5.1 252 Policy priorities for emerging, transitioning, or transforming Tables countries 5.2 262 Emerging countries: A skill O.1 14 Benefi ts of digital technologies for development agenda for a modern labor workers and consumers: A scorecard market O.2 24 Classifying the digital citizen 5.3 266 Transitioning countries: A skill engagement cases development agenda for a modern labor O.3 34 Priority policies for better service market delivery 5.4 270 Transforming countries: A skill 1.1 74 The internet impact is highest for development agenda for a modern labor data-intensive activities that involve market easy-to-enforce contracts 5.5 273 A framework for policies: How to 1.2 74 Many fi rms use the internet without improve services in different contexts changing their organizational 5.6 274 Emerging countries: An agenda for structures, limiting its impact, 2010–14 laying institutional foundations and 1.3 82 Economic activities with high potential improving services for fi rms to use digital technologies 5.7 277 Transitioning countries: An agenda more intensively are often protected for building capable and accountable from foreign or domestic competition institutions and improving services in developing countries, reducing productivity growth 5.8 279 Transforming countries: An agenda for deepening collaborative institutions 2.1 105 Digital technologies affect employment and improving services and earnings, the evidence shows 6.1 295 Multistakeholderism or multilateralism 2.2 106 Benefi ts of digital technologies for workers and consumers: A scorecard 6.2 301 Many countries have proposed comprehensive legislation on data 2.3 122 Interactions between technology and fl ows skills at work 6.3 304 Information as feedback and as input, 2.4 122 Recent evidence on skill-biased by expenditure type technological change Foreword

We fi nd ourselves in the midst of the greatest information and communications revolution in . More than 40 percent of the world’s population has access to the inter- , with new users coming online every day. Among the poorest 20 percent of households, nearly 7 out of 10 have a mobile phone. The poorest households are more likely to have access to mobile phones than to toilets or clean water. We must take advantage of this rapid technological change to make the world more prosperous and inclusive. This Report fi nds that traditional development challenges are preventing the digital revolution from fulfi lling its transformative potential. For many people, today’s increase in access to digital technologies brings more choice and greater convenience. Through inclusion, effi ciency, and innovation, access provides opportunities that were previously out of reach to the poor and disadvantaged. In Kenya, for example, the cost of sending remittances dropped by up to 90 percent after the introduction of M-Pesa, a digital payment system. New technologies allow women to participate more easily in the labor market—as e-commerce entrepreneurs, in online work, or in business-process outsourcing. The world’s 1 billion persons with disabilities—80 per- cent of whom live in developing countries—can lead more productive lives with the help of text, voice, and communication. And digital ID systems can provide better access to public and private services for the 2.4 billion people who lack formal identifi cation records, such as a birth certifi cate. While this is great , many are still left out because they do not have access to digital technologies. Those in have the most to gain from better commu- nication and access to information. Nearly 6 billion people do not have high-speed internet, making them unable to fully participate in the digital economy. To deliver universal digital access, we must invest in infrastructure and pursue reforms that bring greater competition to telecommunications markets, promote public-private partnerships, and yield effective regulation. The Report concludes that the full benefi ts of the information and communications transformation will not be realized unless countries continue to improve their business climate, invest in people’s education and health, and promote good governance. In countries where these fundamentals are weak, digital technologies have not boosted productivity or reduced inequality. Countries that complement technology investments with broader economic reforms reap digital dividends in the form of faster growth, more jobs, and better services. The stands ready to help countries pursue these priorities. We are already working with clients to promote competitive business environments, increase accountability, and upgrade education and skills-development systems to prepare people for the jobs of the future.

xiii xiv FOREWORD

While people around the world make more than 4 billion Google searches every day, 4 billion people still lack access to the internet. The fi ndings of this Report should be used by all who are working to end extreme poverty and boost shared prosperity. The greatest rise of information and communications in history will not be truly revolutionary until it benefi ts everyone in every part of the world.

Jim Yong Kim President The World Bank Group Acknowledgments

This Report was prepared by a team led by Deepak Mishra and Uwe Deichmann and com- prising Kenneth Chomitz, Zahid Hasnain, Emily Kayser, Tim Kelly, Märt Kivine, Bradley Larson, Sebastian Monroy-Taborda, Hania Sahnoun, Indhira Santos, David Satola, Marc Schiffbauer, Boo Kang Seol, Shawn Tan, and Desiree van Welsum. The work was carried out under the general direction of , Indermit Gill, and Pierre Guislain. World Bank President was an invaluable source of encouragement to the team. The team received guidance from an Advisory Panel cochaired by Kaushik Basu and and consisting of Salim Sultan Al-Ruzaiqi, Carl Bildt, Yessica Car- tajena, Dorothy Gordon, Richard Heeks, Monica Kerretts-Makau, Feng Lu, N.R. Narayana Murthy, , and Hal Varian. The team would like to acknowledge the generous support of Canada’s Department of Foreign Affairs, Trade, and Development and the International Development Research Cen- tre; Estonia’s Ministry of Foreign Affairs and Offi ce of the President; the French Develop- ment Agency; ’s Federal Ministry for Economic Cooperation and Development and the Deutsche Gesellschaft für Internationale Zusammenarbeit; Israel’s Ministry of Econ- omy; ’s Ministry of Foreign Affairs and the Norwegian Agency for Development Cooperation; ’s Ministry of Foreign Affairs; the multidonor Knowledge for Change Program; and the World Bank Research Support Budget. Consultation events were held in , , China, the , the Arab Republic of Egypt, Estonia, , , Germany, , Indonesia, Ireland, , Kenya, Morocco, the , , , the Philippines, Somalia, Swe- den, , , the , the , the United States, and Vietnam, with participants drawn from many more countries. Detailed information about these events can be found at http://www.worldbank.org/wdr2016/about. Interagency consultations were held with the European Commission, International Telecommunication Union, Organisation for Economic Co-operation and Development, United Nations Broad- band Commission, United Nations Conference on (UNCTAD), and United Nations Development Programme (UNDP). The initial fi ndings of the Report were also discussed at several conferences and workshops, including at the Brookings-Blum Roundtable; Columbia University; Nairobi’s iHub; International Conference of Agricultural in Milan; Oxford Internet Institute; the People-Centered Internet Conference at Stanford University; the , Technology and Innovation for Development conference in ; the Swedish Program for Information and Communication Technology in Devel- oping Regions; the UbuntuNet Alliance Connect Conference in ; the University of West Indies at Mona, Jamaica; the U.S. State Department; the ; and the World Summit on the Information Society. The team thanks the participants in all of these events for helpful comments and suggestions. Bruce Ross-Larson was the principal editor of the Report. The and logistics team for the Report comprised Brónagh Murphy, Mihaela Stangu, and Jason Victor, with xv xvi ACKNOWLEDGMENTS

contributions from Laverne Cook, Gracia Sorenson, Roza Vasileva, and Bintao Wang. Reboot was the principal graphic designer. Phillip Hay, Vamsee Kanchi, Mikael Ello Reven- tar, and Roula Yazigi provided guidance on communication strategy. The World Bank’s Publishing and Knowledge Division coordinated the copyediting, typesetting, designing, printing, and dissemination of the Report. Nancy Morrison and Dana Lane copyedited the Report. Diane Stamm and Laura Wallace edited the background papers and framing notes, respectively. Special thanks to Denise Bergeron, Jose de Buerba, Mary Fisk, Yulia Ivanova, Patricia Katayama, Stephen McGroarty, Andres Meneses, Chiamaka Osuagwu, Stephen Pazdan, and Paschal Ssemaganda, as well as the Translation and Interpretation Unit’s Bouchra Belfqih and her team, and the Map Design Unit. The team would like to thank Vivian Hon, Jimmy Olazo, and Claudia Sepúlveda for their coordinating roles. Elena Chi- Lin Lee, Surekha Mohan, and Joseph Welch coordinated resource mobilization. Jean-Pierre Djomalieu, Gytis Kanchas, Nacer Megherbi, Manas Ranjan Parida, and Pratheep Ponraj provided IT support. The team would like to acknowledge several people for insightful discussions, including Jenny Aker, George Akerlof, Robert Atkinson, David Autor, Arup Banerji, Eric Bartelsman, Vint Cerf, Carol Corrado, Claudia Maria Costin, Augusto de la Torre, Asli Demirgüç-Kunt, Shantayanan Devarajan, Laurent Elder, , Francisco Ferreira, Torbjorn Fredriksson, Carl Frey, Haishan Fu, Mark Graham, Caren Grown, , Jesse Kaplan, Loukas Karabarbounis, Phil Keefer, Michael Kende, , Taavi Kotka, Aart Kraay, Arianna Legovini, Norman Loayza, Epp Maaten, Michael Mandel, James Manyika, Magdy Martinez-Soliman, Njuguna Ndung’u, , Ory Okolloh, Tapan Parikh, Rich Pearson, Lant Pritchett, Martin Rama, Vijayendra Rao, Ana Revenga, John Rose, Sudhir Shetty, , Randeep , Larry Summers, Jan Svejnar, Syverson, Prasanna Tambe, Michael Thatcher, Hans Timmer, Kentaro Toyama, Nigel Twose, Bart van Ark, Tara Vishwanath, Stephanie von Friedeburg, Melanie Walker, and Darrell West. The contributors to the spotlight and sector focus pieces are Robert Ackland, Wajeeha Ahmad, Hallie Applebaum, Joseph Atick, Amparo Ballivian, Adis Balota, Biagio Bossone, Karan Capoor, Mariana Dahan, Alan Gelb, Aparajita Goyal, Dominic S. Haazen, Naomi Hale- wood, Mia Harbitz, Todd Johnson, Anna Lerner, Dennis Linders, Arturo Muente-Kunigami, Urvashi Narain, Thomas Roca, Zlatan Sabic, Marcela Sabino, Chris Sall, Randeep Sudan, Kyosuke Tanaka, Tatiana Tropina, Michael Trucano, and Darshan Yadunath. The Report draws on background papers and notes prepared by Karina Acevedo, Laura Alfaro, Maja Andjelkovic, Izak Atiyas, Ozan Bakis, Shweta Banerjee, Sheheryar Banuri, Johannes Bauer, Jessica Bayern, Zubair Bhatti, Miro Frances Capili, Xavier Cirera, Nicholas Crafts, Cem Dener, Joao Maria de Oliveira, Bill Dutton, Mark Dutz, Maya Eden, Ana Fer- nandes, Lucas Ferreira-Mation, Rachel Firestone, Jonathan Fox, Paul Gaggl, Jose Marino Garcia, Elena Gasol Ramos, Tina George, Daphne Getz, Itzhak Goldberg, Martin Hilbert, Sahar Sajiad Hussain, Leonardo Iacovone, Saori Imaizumi, Ali Inam, Melissa Johns, Todd Johnson, Patrick Kabanda, Chris Kemei, Doruk Yarin Kiroglu, Barbara Kits, Anna Kocha- nova, Gunjan Krishna, Arvo Kuddo, Filipe Lage de Sousa, Michael Lamla, Victoria Lemieux, Emmanuel Letouzé, Zahra Mansoor, Francisco Marmolejo, Aaditya Mattoo, Samia Melhem, Michael Minges, Martin Moreno, Huy Nygen, Stephen O’Connell, Brian O’Donnell, Alberto Osnago, Tiago Peixoto, Mariana Pereira-Lopez, Gabriel Pestre, Sonia Plaza, Rita Ramalho, , Seyed Reza Yousefi , Said Mohamed Saadi, Leo Sabetti, Simone Sala, Deepti Samant Raja, David Sangokoya, Bessie Schwarz, Sophiko Skhirtladze, Elisabeth Tellman, Kristjan Vassil, Patrick Vinck, Joanna Watkins, Robert Willig, Min Wu, Maggie Xu, Emilio Zagheni, and Irene Zhang. All background papers for the Report are available at http://www .worldbank.org/wdr2016 or through the WDR offi ce at the World Bank. The team received expert advice during several rounds of reviews from Christian Aedo, Ahmad Ahsan, Mohamed Ihsan Ajwad, Omar Arias, Cesar Baldeon, Morgan Bazilian, Kath- leen Beegle, Luis Beneviste, Christian Bodewig, Stefanie Brodmann, Shubham Chaudhuri, ACKNOWLEDGMENTS xvii

Karl Chua, Massimo Cirasino, Amit Dar, Ximena del Carpio, Deon Filmer, Adrian Fozzard, Samuel Freije, Roberta Gatti, Caren Grown, Mary Hallward–Driemeier, Robert Hawkins, Joel Hellman, Mohamed Ibrahim, , Luis Felipe Lopez Calva, Charlotte V. McClain-Nhlapo, Atul Mehta, Samia Melham, Claudio Montenegro, Reema Nayar, David Newhouse, Anna Olefi r, Pierella Paci, Cecilia Paradi-Guilford, Josefi na Posadas, Siddhartha Raja, Dena Ringold, David Robalino, Jan Rutkowski, Carolina Sanchez-Paramo, Joana Silva, Jin Song, Renos Vakis, Alexandria Valerio, Joao Pedro Wagner de Azevedo, Aleem Walji, Michael Weber, and William Wiseman, as well as from the World Bank Group regions, global practices, cross-cutting solutions areas, Legal Department, Independent Evaluation Group, and other units. Many people inside and outside the World Bank Group provided helpful comments, made other contributions, and participated in consultative meetings. The team would like to thank the following: Jamal Al-Kibbi, Mavis Ampah, Dayu Nirma Amurwanti, James Anderson, Elena Arias, Andrew Bartley, Cyrille Bellier, Rachid Benmessaoud, Natasha Beschorner, Zubair Bhatti, Phillippa Biggs, Brian Blankespoor, Blumenstock, David Caughlin, Jean-Pierre Chauffour, Michael Chodos, Diego Comin, Pedro Conceicao, Paulo Correa, Eric Crabtree, Prasanna Lal Das, Ron Davies, Valerie D’Costa, James Deane, Donato de Rosa, Niamh Devitt, Ndiame Diop, Dini Sari Djalal, Khalid El Massnaoui, Oliver Falck, Erik Feiring, Xin Feng, Nicolas Friederici, Doyle Galegos, Rikin Gandhi, John Garrity, Diari- etou Gaye, Daphne Getz, Ejaz Syed Ghani, Soren Gigler, Chorching Goh, Itzhak Goldberg, Simon Gray, Boutheina Guermazi, Suresh Gummalam, Stefanie Haller, Nagy Hanna, Jeremy Andrew Hillman, Stefan Hochhuth, Anke Hoeffl er, Bert Hofman, Mai Thi Hong Bo, Tim Hwang, William Jack, Sheila Jagannathan, Satu Kahkonen, Kai Kaiser, Jesse Kaplan, Rajat Kathuria, Anupam Khanna, Stuti Khemani, Zaki Khoury, Oliver Knight, Srivatsa Krishna, Kathie Krumm, Victoria Kwakwa, Somik Lall, Jason Lamb, Jessica Lang, Andrea Liverani, Steven Livingston, Augusto Lopez-Claros, Muboka Lubisia, Sean Lyons, Sandeep Mahajan, Shiva Makki, Will Martin, Selina McCoy, Stefano Mocci, , Partha Mukhopadhyay, Pauline Mwangi, Gb Surya Ningnagara, Tenzin Norbhu, Tobias Ochieng, Varad Pande, Douglas Pearce, Oleg Petrov, Jan Pierskalla, Maria Pinto, Martin Raiser, Achraf Rissafi , Nagla Rizk, Michel Rogy, Gabriel Roque, Karen Rose, Carlo Maria Rossotto, Frances Ruane, Onno Ruhl, Umar Saif, Daniel Salcedo, Apurva Sanghi, Arleen Seed, Shekhar Shah, Fred Shaia, Shehzad Sharjeel, Gurucharan Singh, Rajendra Singh, Alexander Slater, Karlis Smits, Vicenzo Spezia, Christoph Stork, Younas Suddique, Abdoulaye Sy, Maria Consuelo Sy, Noriko Toyoda, Rogier van den Brink, Adam Wagstaff, Ken Warman, Cynthia Wong, Bill Woodcock, Pat Wu, Elif Yonca Yukseker, and Breanna Zwart. The team also met with representatives from civil society and the private sector, includ- ing Airbnb; Alibaba (China); Babajob (India); Baidu (China); Diplo (Switzerland); Economic and Social Research Institute (ESRI; Ireland); Elance-oDesk (now Upwork); eLimu (Kenya); Enterprise Ireland; the Estonian e-Governance Academy; Facebook; Google; Groupe Speciale Mobile Association (GSMA); Human Rights Watch; Nairobi’s iHub; Internet Corporation for Assigned Names and Numbers (ICANN); Internet Society; Khan Academy; Let’s Do It! (Estonia); Lyft; MajiVoice (Kenya); McKinsey Global Institute; Microsoft; National Asso- ciation of Software and Services Companies (India); Nortal (Estonia); Olacabs (India); Postmates; Rovio Entertainment (Finland); Souktel ( and Gaza); the Start-Up Jamaica Accelerator; TransferWise (Estonia/United Kingdom); ; and Uber. The team apologizes to any individuals or organizations inadvertently omitted from this list.

Abbreviations

2G second-generation 3D three-dimensional third-generation fourth-generation 5G fi fth-generation ADB AfDB AI artifi cial intelligence APEC Asia-Pacifi c Economic Cooperation ATM automated teller machine AV autonomous vehicles B2B business-to-business BIA Bridge International Academies BISP Benazir Income Support Programme (Pakistan) BPO business process outsourcing C2C consumer-to-consumer CAL computer-assisted learning CDRs call data records CERT computer emergency response team CRM customer relationship management CSIRT Computer Security Incident Response Team CSO civil society organization DAI Digital Adoption DFID Department for International Development (United Kingdom) DRM disaster risk management DSL digital subscriber line EBRD European Bank for Reconstruction and Development EC European Commission ERP economic resource planning; Electronic EU European Union FCC Federal Communications Commission (United States) FDI foreign direct investment G-8 (Canada, France, Germany, , , the Russian Federation, the United Kingdom, and the United States) G2B government-to-business G2C government-to-citizen G2G government-to-government xix xx ABBREVIATIONS

GDP GIS geographic information system GNI GPS global positioning system GSMA Groupe Speciale Mobile Association (aka Global System for Mobile communications Association) GTAP Global Trade Analysis Project HEWs Health Extension Workers HMIS Health Management Information System HS harmonized classifi cation system I2D2 International Income Distribution Database (World Bank) IANA Internet Assigned Numbers Authority IATA International Air Transport Association ICANN Internet Corporation for Assigned Names and Numbers ICT information and communication technology ID identifi cation IDRC International Development Research Centre (Canada) IETF Internet Engineering Task Force IFAD International Fund for Agricultural Development IFC International Finance Corporation (of the World Bank Group) IoT internet of things IP intellectual property; internet protocol IPRs intellectual property rights ISP internet service provider IT information technology ITRs International Telecommunication Regulations ITU International Telecommunication Union IXP KILM Key Indicators of the Labour Market LDCs least developed countries LLU local loop unbundling LPI Logistics Performance Index LTE Long Term Evolution M&E monitoring and evaluation MDGs Millennium Development Goals (United Nations) MFN most favored nation MLM multilateral model MOOC massive open online course MSM multistakeholder model NGO nongovernmental organization NTM nontariff measure OECD Organisation for Economic Co-operation and Development OLPC One Laptop per Child OSI online service index OTT over-the-top PC personal computer PFR Program for Results (World Bank) PIAAC Programme for the International Assessment of Adult Competencies PISA Programme for International Student Assessment

PM2.5 particulates with a diameter of less than 2.5 micrometers ABBREVIATIONS xxi

PMR Product Market Regulation POP point of presence PPP public-private partnership; PTT public telephone and telegraph R&D research and development RFID radio frequency identifi cation RSS Rich Site Summary SCM supply chain management SDGs Sustainable Development Goals (United Nations) SIM subscriber identifi cation module SMEs small and medium enterprises SMS short message service STEM science, technology, engineering, and mathematics STEP Skills Towards Employability and Productivity (World Bank) SYNOP surface synoptic observations TFP total factor productivity TRCs Truth and Reconciliation Commissions UN United Nations UNCTAD United Nations Conference on Trade and Development USAID U.S. Agency for International Development USF Universal Service Fund VAT added tax W3C Consortium WDI World Development Indicators (World Bank database) WDR 2016 team team for the 2016 World Development Report WEF World Economic Forum WIPO World Intellectual Property Organization WITS World Integrated Trade Solution (World Bank database) WTO

Currencies $A Australian dollar ₹ Indian rupee € euro US$ U.S. dollar K Sh Kenyan shilling Y Chinese yuan

Units of measurement GB gigabyte kWh kilowatt-hour Gbit/s gigabits per second Mbit/s megabits per second GHz gigahertz MHz megahertz kbps kilobits per second Tbit/s terabits per second

Country and economy codes AFG AUS AGO Angola AUT Austria ALB AZE ARE United Arab Emirates BDI Burundi ARG Argentina BEL Belgium ARM Armenia BEN xxii ABBREVIATIONS

BFA HND BGD HRV BGR HTI BHR HUN BIH IDN Indonesia BLR IND India BLZ Belize IRL Ireland BOL IRN , Islamic Rep. BRA Brazil IRQ Iraq BRB Barbados ISL Iceland BRN Brunei Darussalam ISR Israel BTN ITA Italy BWA Botswana JAM Jamaica CAN Canada JOR CHE Switzerland JPN Japan CHL KAZ CHN China KEN Kenya CIV Côte d’Ivoire KGZ Kyrgyz Republic CMR KHM COD Congo, Dem. Rep. KOR Korea, Rep. COL Colombia KWT CPV Cabo Verde LAO Lao PDR CRI LBN Lebanon CYP LBR CZE Czech Republic LBY DEU Germany LKA Sri Lanka DJI Djibouti LSO DNK LTU DOM Dominican Republic LUX Luxembourg DZA LVA ECU MAR Morocco EGY Egypt, Arab Rep. MDA ESP MDG EST Estonia MDV ETH MEX Mexico FIN Finland MKD Macedonia, FYR FJI MLI FRA France MLT Malta MMR GBR United Kingdom MNE Montenegro GEO MNG GHA MOZ Mozambique GIN Guinea MRT GMB Gambia, The MUS GRC MWI GRD MYS GTM Guatemala NAM GUY NER ABBREVIATIONS xxiii

NGA STP São Tomé and Príncipe NIC SVK Slovak Republic NLD Netherlands SVN Slovenia NOR Norway SWE Sweden NPL SWZ Swaziland NZL SYC OMN Oman TCD Chad PAK Pakistan TGO PAN THA Thailand PER TJK PHL Philippines TKM Turkmenistan PNG TLS Timor-Leste POL TON PRT Portugal TTO PRY TUN Tunisia PSE West Bank and Gaza TUR Turkey QAT TZA ROU UGA RUS Russian Federation UKR RWA URY Uruguay SAU USA United States SEN Senegal UZB SGP VEN Venezuela, RB SLB Islands VNM Vietnam SLE YEM , Rep. SLV ZAF SOM Somalia ZMB SRB ZWE Zimbabwe Inclusion

Efficiency

Innovation

OVERVIEW Strengthening the analog foundation of the digital revolution

Digital technologies have spread rapidly in much of the world. Digital dividends—the broader development benefi ts from using these technologies—have lagged behind. In many instances digital technologies have boosted growth, expanded oppor- tunities, and improved service delivery. Yet their aggregate impact has fallen short and is unevenly distributed. For digital technologies to benefi t everyone everywhere requires closing the remaining digital divide, especially in internet access. But greater digital adoption will not be enough. To get the most out of the digital revolution, countries also need to work on the “analog complements”—by strengthening regulations that ensure competition among businesses, by adapting workers’ skills to the demands of the new economy, and by ensuring that institutions are accountable.

Digital technologies—the internet, mobile phones, e-commerce site, by signifi cantly reducing coordi- and all the other tools to collect, store, analyze, and nation costs, boosts effi ciency in China’s economy share information digitally—have spread quickly. and arguably the world’s. The M-Pesa digital pay- More households in developing countries own a ment platform, by exploiting scale economies from mobile phone than have access to electricity or clean automation, generates signifi cant fi nancial sector water, and nearly 70 percent of the bottom fi fth of innovation, with great benefi ts to Kenyans and others. the population in developing countries own a mobile Inclusion, effi ciency, innovation—these are the main phone. The number of internet users has more than mechanisms for digital technologies to promote tripled in a decade—from 1 billion in 2005 to an development. estimated 3.2 billion at the end of 2015.1 This means Although there are many individual success sto- that businesses, people, and governments are more ries, the effect of technology on global productivity, connected than ever before (fi gure O.1). The digital expansion of opportunity for the poor and the middle revolution has brought immediate private bene- class, and the spread of accountable governance has fi ts—easier communication and information, greater so far been less than expected (fi gure O.2).2 Firms are convenience, free digital products, and new forms of more connected than ever before, but global produc- leisure. It has also created a profound sense of social tivity growth has slowed. Digital technologies are connectedness and global community. But have mas- changing the world of work, but labor markets have sive investments in information and communication become more polarized and inequality is rising—par- technologies (ICTs) generated faster growth, more ticularly in the wealthier countries, but increasingly jobs, and better services? Indeed, are countries reap- in developing countries. And while the number of ing sizable digital dividends? democracies is growing, the share of free and fair Technology can be transformational. A digital elections is falling. These trends persist, not because identifi cation system such as India’s Aadhaar, by of digital technologies, but in spite of them. overcoming complex information problems, helps So, while digital technologies have been spreading, willing governments to promote the inclusion of dis- digital dividends have not. Why? For two reasons. First, advantaged groups. Alibaba’s business-to-business nearly 60 percent of the world’s people are still offl ine OVERVIEW 3

Figure O.1 Digital technologies have spread rapidly in much of the world

a. Digital adoption b. Digital adoption c. Digital adoption by businesses by people by governments 1.0 1.0 1.0

Global 0.8 0.8 0.8 average Global Global average 0.6 average 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

0 0 0 100 1,000 10,000 100,000 100 1,000 10,000 100,000 100 1,000 10,000 100,000 GDP per capita (constant 2005 US$) GDP per capita (constant 2005 US$) GDP per capita (constant 2005 US$)

Source: WDR 2016 team. Data at http://bit.do/WDR2016-FigO_1. Note: The fi gures show the diff usion of digital technologies across countries as measured by the Digital Adoption Index compiled for this Report and described in detail in chapter 5 of the full Report. GDP = gross domestic product.

and can’t participate in the digital economy in any institutions, amplify the voice of elites, which can meaningful way. Second, some of the perceived bene- result in policy capture and greater state control. And fi ts of digital technologies are offset by emerging risks because the economics of the internet favor natural (fi gure O.3). Many advanced economies face increas- , the absence of a competitive business ingly polarized labor markets and rising inequality—in environment can result in more concentrated markets, part because technology augments higher skills while benefi ting incumbent fi rms. Not surprisingly, the bet- replacing routine jobs, forcing many workers to com- ter educated, well connected, and more capable have pete for low-paying jobs. Public sector investments received most of the benefi ts—circumscribing the in digital technologies, in the absence of accountable gains from the digital revolution.

Figure O.2 The pessimism concerning the global outlook is not because of digital technologies, but in spite of them

a. Global productivity b. Global inequality c. Five-year moving average of median growth Percentage change in real income between Share of elections that are free and fair (%) of labor productivity per hour worked, 1998 and 2008 at different levels of world in percent, in 87 countries income distribution in 2003 prices

6 90 100

70 75 4 50 50 30 2 25 10

0 –10 0 5 1525354555 65 75 85 95 09 1985 0 1979198219851988199119941997200020032006200920122015 1973 1979 1991 1997 2003 2 2015 Percentile of world income distribution

Sources: Panel a: Conference Board (various years); WDR 2016 team. Panel b: Lakner and Milanovic 2013. Panel c: Bishop and Hoeffl er 2014. Data at http://bit.do/WDR2016-FigO_2. 4 WORLD DEVELOPMENT REPORT 2016

Figure O.3 Why digital dividends are not spreading rapidly—and what can be done

Making the internet Divide Connectivity Accessible Affordable Open and safe

Digital Digital Spreading benefits development technologies strategy INCLUSION EFFICIENCY INNOVATION

Dividends Reducing risks Complements

CONTROL INEQUALITY CONCENTRATION

Source: WDR 2016 team.

To maximize the digital dividends requires better foundation, consisting of regulations that create a understanding of how technology interacts with other vibrant business climate and let fi rms leverage dig- factors that are important for development—what the ital technologies to compete and innovate; skills that Report calls “analog complements.” Digital technol- allow workers, entrepreneurs, and public servants to ogies can make routine, transaction-intensive tasks seize opportunities in the digital world; and account- dramatically cheaper, faster, and more convenient. But able institutions that use the internet to empower most tasks also have an aspect that cannot be auto- citizens. The long-term development impact is by mated and that requires human judgment, intuition, no means defi nitive, being continuously shaped by and discretion. When technology is applied to auto- the evolution of technology (connectivity) and the mate tasks without matching improvements in the country’s choice of economic, social, and governance complements, it can fail to bring broad-based gains. arrangements (complements).4 Countries that are The digital revolution can give rise to new business able to swiftly adjust to this evolving digital economy models that would benefi t consumers, but not when will reap the greatest digital dividends, while the rest incumbents control market entry. Technology can are likely to fall behind (fi gure O.3 and box O.1). make workers more productive, but not when they The triple complements—a favorable business cli- lack the know-how to use it. Digital technologies can mate, strong , and good governance— help monitor teacher attendance and improve learn- will sound familiar—and they should because they ing outcomes, but not when the education system are the foundation of . But lacks accountability.3 digital technologies add two important dimensions. What should countries do? Making the internet First, they raise the of not undertak- universally accessible and affordable should be a ing the necessary reforms. They amplify the impact global priority. The internet, in a broad sense, has of good (and bad) policies, so any failure to reform grown quickly, but it is by no means universal. For means falling farther behind those who do reform. every person connected to high-speed broadband, With digital technologies, the stakes have risen for fi ve are not. Worldwide, some 4 billion people do not developing countries, which have more to gain than have any internet access, nearly 2 billion do not use a high-income countries, but also more to lose. Second, mobile phone, and almost half a billion live outside while digital technologies are no shortcut to develop- areas with a mobile signal. The unfi nished task of con- ment, they can be an enabler and perhaps an accel- necting everyone to the internet—one of the targets in erator by raising the quality of the complements. the recently approved Sustainable Development Goals Online business registries ease market entry for new (SDGs)—can be achieved through a judicious mix of and innovative fi rms. Well-designed internet-based market competition, public-private partnerships, and training helps workers upgrade their skills. New effective regulation of the internet and telecom sector. media platforms can increase citizen participa- Access to the internet is critical, but not suffi cient. tion. And digital enablers—digital fi nance, digital The digital economy also requires a strong analog identifi cation, social media, and open data—spread OVERVIEW 5

Box O.1 Frequently asked questions: The Report at a glance

What is the Report about? of mid-level offi ce jobs, could contribute to a hollowing It explores the impact of the internet, mobile phones, and out of labor markets and to rising inequality. And the poor related technologies on economic development. Part 1 record of many e-government initiatives points to high fail- shows that potential gains from digital technologies are ure of ICT projects and the risk that states and corporations high, but often remain unrealized. Part 2 proposes policies could use digital technologies to control citizens, not to to expand connectivity, accelerate complementary reforms empower them. in sectors beyond information and communication technol- ogy (ICT), and address global coordination problems. What should countries do to mitigate these risks? Connectivity is vital, but not enough to realize the full devel- What are the digital dividends? opment benefi ts. Digital investments need the support Growth, jobs, and services are the most important returns of “analog complements”: regulations, so that fi rms can to digital investments. The fi rst three chapters show how leverage the internet to compete and innovate; improved digital technologies help businesses become more pro- skills, so that people can take full advantage of digital ductive; people fi nd jobs and greater opportunities; and opportunities; and accountable institutions, so that gov- governments deliver better public services to all. ernments respond to citizens’ needs and demands. Digital technologies can, in turn, augment and strengthen these How do digital technologies promote development and complements—accelerating the pace of development. generate digital dividends? By reducing information costs, digital technologies greatly What needs to be done to connect the unconnected? lower the cost of economic and social transactions for Market competition, public-private partnerships, and eff ec- fi rms, individuals, and the public sector. They promote tive regulation of internet and mobile operators encourage innovation when transaction costs fall to essentially zero. private investment that can make access universal and They boost effi ciency as existing activities and services aff ordable. Public investment will sometimes be necessary become cheaper, quicker, or more convenient. And they and justifi ed by large social returns. A harder task will be increase inclusion as people get access to services that to ensure that the internet remains open and safe as users previously were out of reach. face cybercrime, privacy violations, and online .

Why does the Report argue that digital dividends are not What is the main conclusion? spreading rapidly enough? Digital development strategies need to be broader than ICT For two reasons. First, nearly 60 percent of the world’s peo- strategies. Connectivity for all remains an important goal ple are still offl ine and can’t fully participate in the digital and a tremendous challenge. But countries also need to economy. There also are persistent digital divides across create favorable conditions for technology to be eff ective. gender, geography, age, and income dimensions within When the analog complements are absent, the develop- each country. Second, some of the perceived benefi ts of the ment impact will be disappointing. But when countries internet are being neutralized by new risks. Vested business build a strong analog foundation, they will reap ample , regulatory uncertainty, and limited contestation digital dividends—in faster growth, more jobs, and better across digital platforms could lead to harmful concentra- services. tion in many sectors. Quickly expanding automation, even

benefi ts throughout the economy and society, fur- technological innovations. For Indonesia to reap the ther strengthening the interaction between technol- benefi ts of steamships took 160 years after their inven- ogy and its complements. tion and for Kenya to have electricity, 60 years; but for Vietnam to introduce , only 15 years. Mobile Digital transformations— phones and the internet took only a few years. More households in developing countries own a mobile digital divides phone than have access to electricity or improved san- The internet and related technologies have reached itation (fi gure O.4, panel a). Greater internet access has developing countries much faster than previous led to an explosion in the production and consumption 6 WORLD DEVELOPMENT REPORT 2016

Figure O.4 Digital transformation in action nearly 70 percent own a mobile phone. The lowest mobile penetration is in Sub-Saharan Africa (73 per- a. Digital technologies are spreading rapidly in developing countries cent), against 98 percent in high-income countries. 100 But internet adoption lags behind considerably: only 31 percent of the population in developing countries

80 had access in 2014, against 80 percent in high-income countries. China has the largest number of internet users, followed by the United States, with India, Japan, 60 and Brazil fi lling out the top fi ve. The world viewed from the perspective of the number of internet users 40 looks more equal than when scaled by income (map O.1)—refl ecting the internet’s rapid . % of the population 20 Connected businesses Internet adoption has increased across businesses in 0 1990 1995 2000 2005 2010 2015 all country income groups. Nearly 9 of 10 businesses in high-income OECD (Organisation for Economic Improved water Mobile phone Electricity Internet Co-operation and Development) countries had a Secondary school Mobile broadband broadband internet connection in 2010–14, compared Improved sanitation with 7 for middle-income countries and 4 for low-in- come countries. But adoption rates for more sophisti- b. A typical day in the life of the internet cated technologies such as secure servers, enterprise network, inventory management, and e-commerce are much lower in most developing countries. 186 million 152 million photos SKYPE 36 million Connected governments calls purchases Governments are increasingly going digital, and a greater share of government jobs in developing coun- 2.3 billion tries is ICT-intensive than in the private sector. By 2014, GB of WEB 803 million 8.8 billion TRAFFIC all 193 member states of the United Nations (UN) had TWEETS 4.2 billion YOUTUBE national websites: 101 enabled citizens to create per- videos watched GOOGLE searches sonal online accounts, 73 to fi le income taxes, and 60 207 billion to register a business. For the most common core gov- E-MAILS ernment administrative systems, 190 member states sent had automated fi nancial management, 179 used such systems for processing, and 159 for tax man- Sources: World Development Indicators (World Bank, various years); WDR 2016 team; http://www .internetlivestats.com/one-second/ (as compiled on April 4, 2015). Data at http://bit.do/WDR2016-FigO_4. agement. And 148 of them had some form of digital Note: In panel a, for some years data for electricity are interpolated from available data. GB = gigabytes. identifi cation, and 20 had multipurpose digital iden- tifi cation platforms. So far, developing countries have invested more in automating back-offi ce functions of information around the world (fi gure O.4, panel b). than in services directed at citizens and businesses. But while the internet has reached almost all coun- tries quickly, the intensity of its use has been lower The divide in digital access and use in poorer countries—in large part because it has not persists spread as widely within those countries. And despite The lives of the majority of the world’s people remain many great examples of the uses of new technologies largely untouched by the digital revolution. Only in developing countries, advanced economies have around 15 percent can afford access to broadband been using them even more effectively.5 internet. Mobile phones, reaching almost four-fi fths of the world’s people, provide the main form of inter- Connected people net access in developing countries. But even then, On average, 8 in 10 individuals in the developing world nearly 2 billion people do not own a mobile phone, own a mobile phone, and the number is steadily ris- and nearly 60 percent of the world’s population has no ing. Even among the bottom fi fth of the population, access to the internet. The world’s offl ine population is OVERVIEW 7

Map O.1 The internet is more evenly spread than income

a. Based on national income, 2014

b. Based on internet population, 2014

IBRD 42010 Source: World Bank. Data at http://bit.do/WDR2016-MapO_1. Note: Countries’ sizes are rescaled in proportion to national income and internet population. The darker the shade, the higher the national income (panel a; GDP at market exchange rates) and the higher the internet population (panel b).

mainly in India and China, but more than 120 million percent and the top 60 percent and between rural people are still offl ine in (fi gure O.5). and urban populations are falling for mobile phones The digital divide within countries can be as high but increasing for the internet. In Africa, the digital as that between countries. Worldwide, nearly 21 per- divide across demographic groups remains consider- cent of households in the bottom 40 percent of their able (fi gure O.6, panel a). Women are less likely than countries’ income distribution don’t have access to men to use or own digital technologies. Gaps are even a mobile phone, and 71 percent don’t have access to larger between youth (20 percent) and those more the internet. Adoption gaps between the bottom 40 than 45 years old (8 percent). 8 WORLD DEVELOPMENT REPORT 2016

Figure O.5 The internet remains unavailable, inaccessible, and unaff ordable to a majority of the world’s population

a. ICT access by population b. A closer look at the world’s offline population Total global population Congo, Dem. Rep. Philippines 68 million ~7.4 billion Mexico 63 million Ethiopia 70 million Russian Federation 95 million 55 million Brazil Iran, Islamic Rep. Within 98 million 54 million mobile coverage Nigeria Myanmar 111 m i l l i o n 53 million Bangladesh Vietnam 7 billion 148 million 52 million Pakistan United States 165 million 51 million Mobile phones Tanzania Indonesia 49 million 213 million Thailand China 48 million 5.2 billion 755 million Egypt, Arab Rep. 42 million Turkey 41 million

Total India Total Countries internet users 1.063 billion internet users outside of the top 20 3.2 billion 3.2 billion

High-speed High-speed internet internet 1.1 billion 1.1 billion

Sources: World Bank 2015; Meeker 2015; ITU 2015; GSMA, https://gsmaintelligence.com/; UN Population Division 2014. Data at http://bit.do/WDR2016-FigO_5. Note: High-speed internet (broadband) includes the total number of fi xed-line broadband subscriptions (such as DSL, cable modems, fi ber optics), and the total number of 4G/LTE mobile subscriptions, minus a correcting factor to allow for those who have both types of access. 4G = fourth generation; DSL = digital subscriber line; ICT = information and communication technology; LTE = Long Term Evolution.

The increased connectivity has had limited effect government. And their use of e-government is highly in reducing information inequality. For example, uneven—citizens in the top 20 percent of income in there are more contributions to Wikipedia from the most connected EU country are 45 times more Hong Kong SAR, China, than from all of Africa com- likely to use e-services than those in the bottom 20 bined, despite the fact that Africa has 50 times more percent of income in the least connected EU coun- internet users.6 The amount of information published try (fi gure O.6, panel b). Within countries, greater on the web, and its origin, often corresponds to what e-government use by individuals is associated with one sees in the offl ine world as well. For instance, education, employment, urban residence, being male, 85 percent of the user-generated content indexed and broadband access. by Google comes from the United States, Canada, and Europe, similar to the share of global scientifi c journals originating in these countries. In fact, the How the internet promotes information produced and consumed in the digital development economy has little bearing on the number of users of digital technologies. Given that nearly one-fi fth of the Digital technologies have dramatically expanded world’s population is illiterate, the spread of digital the information base, lowered information costs, technologies alone is unlikely to spell the end of the and created information goods. This has facilitated global knowledge divide. searching, matching, and sharing of information and Countries that have bridged the digital-access contributed to greater organization and collaboration divide often face a new divide in digital capabilities. among economic agents—infl uencing how fi rms In the European Union (EU), businesses are more operate, people seek opportunities, and citizens inter- likely than citizens to use the internet to interact with act with their governments. The changes are not lim- the government. Citizens use e-government mostly ited to economic transactions—they also infl uence for getting information and not for transacting with the participation of women in the labor force, the OVERVIEW 9

Figure O.6 The digital divide in access is high in Africa, and the divide in capability is high in the European Union

a. Africa b. European Union Within-country digital divide can be signifi cant Poor households use e-government less than the rich

25 100

20 80

15 60

10 40 45:1

5 20 % of individuals (ages 16–74) Individuals with internet access (%) 0 0 Bottom Upper Mature Young Rural Urban Women Men 0 20,000 40,000 60,000 80,000 40% 60% (45+) (15–24) GDP per capita (US$) Income Age Location Gender distribution Top income quartile Second quartile () Third quartile Bottom quartile

Sources: WDR 2016 team, based on data from Research ICT Africa (various years), ITU, and Eurostat (EC, various years). Data at http://bit.do/WDR2016-FigO_6. Note: For more details see fi gure 2.4 in the full Report. ease of communication for people with disabilities, Figure O.7 The internet promotes development and the way people spend their leisure. By overcom- through three main mechanisms ing information barriers, augmenting factors, and transforming products, digital technologies can make DIGITAL development more inclusive, effi cient, and innovative TECHNOLOGIES (fi gure O.7 and box O.2). Spotlight 1 in the full Report explores the links between these three mechanisms in the broader economic literature.

The internet promotes inclusion Automation and Before the internet arrived, some transactions were Search and Scale economies information coordination and platforms so expensive that a market for them did not exist. Two types of transactions fall into this category. First is when two parties to a potentially benefi cial transaction simply didn’t know about each other and INCLUSION EFFICIENCY INNOVATION faced exorbitantly high search and information costs. Second is when one party had a lot more information Source: WDR 2016 team. than the other. In the economics literature, such situa- tions are known as information asymmetries between creditworthiness. Or a small fi rm that cannot connect buyers and sellers, and in the absence of trust and with a potential buyer in another country and does not transparency, many transactions do not take place. know whether to trust a new business partner. Or a By reducing the cost of acquiring information freelancer willing to perform small tasks for a fee. Or and making more information available transpar- a homeowner looking to rent her spare room to local ently, digital technologies can make new transac- visitors. Or remote or marginalized population groups tions possible.7 Consider a poor farmer who cannot who fall outside the reach of the services that gov- access credit because the lender has no way to assess ernments provide. In all these cases, a fundamental 10 WORLD DEVELOPMENT REPORT 2016

Box O.2 e-commerce with Chinese characteristics: Inclusion, effi ciency, and innovation in Taobao villages

The dynamic growth and rapid spread of e-commerce Effi ciency. Besides the Taobao e-commerce site for in China is best illustrated by the Shaji phenomenon. consumers, Alibaba and other Chinese fi rms operate The economy of Dongfeng village in Shaji town (Jiangsu business-to-business platforms. They facilitate intra- and Province) shifted from pig farming in the to plastic inter-industry trade in China’s already effi cient production waste recycling in the 1990s. In 2006, a migrant from sector, as well as exports. They also make it easier for for- the village returned to open an online shop to sell simple eign fi rms to sell in China. Consumers benefi t from greater furniture. His success encouraged other villagers to do selection and convenience on online retail sites. Online likewise, and by the end of 2010, the village had 6 board trade has not only helped raise rural incomes but also processing factories, 2 metal parts factories, 15 logistics made shopping more effi cient. Purchasing power in rural and shipping companies, and 7 computer stores serving areas is only about one-third that in cities, but the aggre- 400 households engaged in online sales throughout China gate consumption of China’s 650 million rural residents and even in neighboring countries. Shaji was one of the is vast, contributing to the national goal of moving from fi rst “Taobao villages”—named after an online shopping an export- and investment-driven economy to one that is platform run by the Alibaba Group—where at least 10 per- more consumption based. And the boom in online trade cent of households are engaged in online commerce.a The has spawned numerous logistics companies that provide Taobao villages, and the rise of e-commerce in China more quick delivery—sometimes by bicycle in towns and villages. generally, illustrate how the internet promotes inclusion, effi ciency, and innovation. Innovation. Taobao and other e-commerce platforms are examples of innovation generated by the economies of Inclusion. While the economies of China’s coastal urban scale that emerge when transaction costs drop drastically. areas have grown rapidly over the last three , rural Since these platforms are highly automated, fees can be and western parts of the country have lagged behind. kept low, and operations are often fi nanced by But China’s large investments in rural connectivity are alone. Some problems cannot easily be solved solely by beginning to pay off . More than 90 percent of villages will automation, such as creating trust in the market and pre- have fi xed broadband access by the end of 2015. Online venting fraud. Online ratings, escrow services, and confl ict commerce has allowed producers in towns and villages to resolution mechanisms address them. One of the most participate in the national and even global economy. At the valuable assets Alibaba and other e-commerce operators end of 2014, there were more than 70,000 merchants in accumulate is data. Each transaction contributes to better 200 Taobao villages, and many more in other rural areas. knowledge about the economy and consumer behavior. Most of the stores are small, with an average of 2.5 employ- This information supports new business lines, such as ees. About one-third of owners are female, and one-fi fth extending credit to small fi rms based on automated eval- were previously unemployed. About 1 percent are persons uations of creditworthiness. This can also advance fi nancial with disabilities. One of Alibaba’s top “netpreneurs,” con- inclusion. In early 2015, for instance, Alibaba’s Ant Financial fi ned to a wheelchair after an accident, built a thriving teamed up with the International Finance Corporation to online livestock business. expand credit to female entrepreneurs in China.

Sources: WDR 2016 team, based on information from the China State Information Center, China Association for Employment Promotion, and Alibaba company reports. a. http://www.alizila.com/report-taobao-villages-rural-china-grow-tenfold-2014.

information problem makes it diffi cult to make a expanding trade, creating jobs, and increasing access deal or a match. Mobile phone records, business-to- to public services—and thus promoting inclusion.8 business e-commerce, the sharing economy, online reputation mechanisms, and digital identifi cation sys- The internet promotes effi ciency tems all help to overcome these information barriers. Perhaps the largest impact has been on transactions While they make the market more effi cient, the big- that existed before the arrival of the internet but are gest benefi t seems to be their market creation effects: now quicker, cheaper, or more convenient to carry out. OVERVIEW 11

This mechanism operates in two ways. First, the dra- Figure O.8 Many digital transactions involve all three matic decline in the price of digital technologies has mechanisms and a two-sided market led businesses and governments to replace existing factors—labor and non-ICT capital—with ICT capital INNOVATION and to automate some of their activities. Airlines use online booking systems to fi ll planes. Supermarkets substitute cashiers with automated checkout count- Platforms ers. Manufacturers use real-time inventory and sup- ply chain management systems. And governments invest in information management systems and offer INCLUSION EFFICIENCY online services for a wide range of tasks—from issu- ing drivers’ licenses to fi ling taxes. Second, digital technologies augment the factors Sellers Buyers not substituted and make them more productive. They help managers to better supervise their workers, poli- Drivers, hosts, On-demand/ Riders, guests, and ticians to monitor the service providers, and workers and freelancers sharing economy small businesses to use technology to become more productive, thus Employers, airlines Job seekers, travelers, Matching raising the returns to their human capital. By stream- and hotels, investors, entrepreneurs, and artists platforms lining tasks and raising the productivity of existing and consumers factors, the internet can greatly increase economic Traders and senders e-commerce and Customers and effi ciency across fi rms, workers, and governments. (money) digital payments recipients (money)

The internet promotes innovation Source: WDR 2016 team. The extreme case of effi ciency is when transactions are executed automatically, without human input, Many internet businesses or services use a platform and transaction costs fall to essentially zero. This is or “two-sided market” model. The platforms match the realm of the “new economy,” such as search or buyers with sellers or a service user with a provider. e-commerce platforms, digital payment systems, In a ride sharing service, the platform automatically e-books, streaming music, and social media. The fi xed matches drivers and passengers (innovation), the cost of building the platform may be large, but the driver takes advantage of a fl exible income-earning of carrying out another transaction or activity not otherwise accessible (inclusion), and the adding another user is tiny. This gives rise to increas- passenger benefi ts from greater convenience and ing , which stimulate new business often lower prices (effi ciency). Online crowdfunding, models and provide a major advantage to online job matching, room sharing, and music sites operate fi rms competing with their offl ine counterparts. The similarly (fi gure O.8). zero marginal cost attracts new sellers and buyers to the fi rm’s platform, creating virtuous network effects, where the benefi t to a buyer increases as more sellers The dividends: Growth, jobs, join in, and vice versa. An auction site attracts more and service delivery bidders the more the sellers use it, and a search engine learns and becomes more useful the more searches The benefi ts of digital technologies fi lter throughout are performed. Scale and zero marginal costs also the economy (fi gure O.9). For businesses, the internet explain why many of the sites have promotes inclusion of fi rms in the by become the preferred vehicles for social mobilization expanding trade, raises the productivity of capital, and political . By enabling almost frictionless and intensifi es competition in the marketplace, which communication and collaboration, the internet can in turn induces innovation. It brings opportunities to support new delivery models, encourage households by creating jobs, leverages human capital, action, and accelerate innovation. and produces consumer surplus. It enables citizens to The 2016 WDR presents many examples of how access public services, strengthens government capa- the internet promotes inclusion, effi ciency, and bility, and serves as a platform for citizens to tackle innovation. In the internet economy the three mecha- problems. These benefi ts are neither nisms often operate together. So the one-to-one map- automatic nor assured, but in numerous instances ping in fi gure O.7 simplifi es a more complex . digital technologies can bring signifi cant gains. 12 WORLD DEVELOPMENT REPORT 2016

Figure O.9 How the three mechanisms apply to has become an essential part of a country’s infra- businesses, people, and governments structure—and a factor of production in almost any activity in a modern economy. Isolating the impact of DIGITAL digital technologies is therefore diffi cult at an aggre- TECHNOLOGIES gate level. Firm-level analysis provides a more reli- able picture.9 The internet enables many small fi rms to participate in global trade, thus leading to more inclusion; it makes existing capital more productive, INCLUSION EFFICIENCY INNOVATION raising effi ciency; and by stimulating competition, it encourages innovation.

BUSINESSES Trade Capital utilization Competition Expanding trade The internet enables more products to be exported to PEOPLE Job opportunities Labor productivity Consumer welfare more markets, often by newer and younger fi rms. A 10-percent increase in internet use in the exporting Public Participation Voice country is found to increase the number of prod- GOVERNMENTS sector capability ucts traded between two countries by 0.4 percent. A similar increase in internet use of a country pair Source: WDR 2016 team. increases the average bilateral trade value per prod- The internet can lead to more trade, better uct by 0.6 percent.10 Firms selling on eBay in Chile, capital use, and greater competition Jordan, Peru, and South Africa are younger than fi rms The ICT sector is a fairly modest part of the overall in the offl ine markets.11 In Morocco, rural artisans, economy. Its share in GDP is around 6 percent in some of them illiterate, sell globally through the Anou OECD member countries and considerably less in crafts platform. At the other end of the spectrum, developing countries (fi gure O.10, panel a). In the businesses trade on global e-commerce sites such as United States, home to 8 of the world’s 14 largest tech- Alibaba’s in an online market that could reach more nology companies by revenue, the contribution of than US$6 trillion over the next fi ve years. Online the ICT sector to GDP is around 7 percent. The corre- platforms overcome trust and information problems sponding number for Ireland is 12 percent—a country through feedback and rating systems and by offering that does not boast its own Silicon Valley, but attracts escrow and dispute resolution mechanisms. Easier many foreign fi rms through its competitive business trade of intermediate products encourages further environment and favorable tax rates. In Kenya, which “unbundling” of production processes, not just in hosts one of the largest ICT sectors in Africa, the value the markets for goods but also for services.12 Firms added share of ICT services in GDP was 3.8 percent in India, Jamaica, and the Philippines have captured in 2013. a share of these global markets for services that range The contribution of ICT capital to GDP growth from traditional back-offi ce services to long-distance has been fairly constant over the past two decades. online tutoring. In high-income countries, it has fallen from 0.7 per- centage points in 1995–99 to 0.4 percentage points in Improving capital utilization 2010–14 (fi gure O.10, panel b). In developing countries, Perhaps the greatest contribution to growth comes the contribution of ICT capital to GDP growth has from the internet’s lowering of costs and thus from been fairly modest—around 15 percent of growth— raising effi ciency and labor productivity in practi- refl ecting lower digital adoption. With rapid diffusion cally all economic sectors. Better information helps of digital technologies into developing countries, this companies make better use of existing capacity, opti- number could rise in the future. In addition, the indi- mizes inventory and supply chain management, cuts rect contributions of ICT capital to economic growth, downtime of capital equipment, and reduces risk. In through improvements in total factor productivity the airline industry, sophisticated reservation and (TFP), could be large as well, although rigorous evi- pricing algorithms increased load factors by about dence linking the two is still missing. one-third for U.S. domestic fl ights between 1993 and The rapid adoption of digital technologies in 2007. The parcel delivery company UPS famously the economy has meant that its benefi ts are widely uses intelligent routing algorithms to avoid left turns, dispersed and its indirect growth impacts diffi cult time and about 4.5 million liters of petrol per to estimate. Like energy or transport, the internet year. Many retailers now integrate their suppliers in OVERVIEW 13

Figure O.10 The size of the ICT sector and its contribution to GDP growth is still relatively modest

a. Share in GDP, OECD countries, 2011 b. Contribution to GDP growth, 1995–2014 14 7

12 6

10 5

8 4

Percent 3

Percent 6

4 2

2 1

0 0 l d k e ia d ia nia c nd r o ada any nds Italy ium apan nlan la Spaing J i public France el celaPolan Aust Ireland Est SwedenF e enmar Gree loven B I Norway Hungary R Can ther S Portuga h D Germe Korea, Rep.Switzerland c UnitedLuxembourg States N 1995–992000–042005–092010–14 1995–992000–042005–092010–14 lovak Republic United Kingdom S Cze OECD average (6%) Developed countries Developing countries ICT capital All other factors Sources: OECD 2014; Conference Board Total Economy Database, January 2014; WDR 2016 team. Data at http://bit.do/WDR2016-FigO_10. Note: GDP = gross domestic product; ICT = information and communication technology; OECD = Organisation for Economic Co-operation and Development.

real-time supply chain management to keep inventory Figure O.11 Vietnamese fi rms using costs low. Vietnamese fi rms using e-commerce had on e-commerce have higher TFP growth, average 3.6 percentage point higher TFP growth than 2007–12 fi rms that did not use it (fi gure O.11). Chinese car com- 4 panies that are more sophisticated users of the inter- net turn over their inventory stocks fi ve times faster 3 than their less savvy competitors. And Botswana and Uruguay maintain unique ID and trace-back systems 2 for livestock that fulfi ll requirements for beef exports to the EU, while making the production process more 1 effi cient. points Percentage

0 Advancing competition Labor TFP growth When fully automated internet-based services drive productivity effect growth effect marginal transaction costs to zero, the consequences e-commerce for are somewhat ambiguous. Low Internet use marginal costs imply large , which Source: Nguyen and Schiff bauer 2015 for the 2016 WDR. Data at http://bit favor natural monopolies. In the offl ine world, such .do/WDR2016-FigO_11. sectors—for example, electricity production—often Note: For more details see fi gure 1.9 in the full Report. TFP = total factor require some form of regulation to protect consumer productivity. interests. But the characteristics of internet-based services could also encourage competition. Price- consumers based on search history, geographic loca- comparison websites, for example, should reduce tion, or other information collected about buyers. prices for consumers, even though the evidence The internet can also facilitate market entry. Inter- shows that price dispersion on the internet persists, net fi rms can start and scale up quickly with relatively in part because companies are getting better at price little staffi ng or capital investment. Cloud comput- discrimination—offering different prices to different ing—the leasing of computing and data storage 14 WORLD DEVELOPMENT REPORT 2016

services—reduces startup costs and allows fi rms to technology will have an advantage. But even the poor add capacity as the need arises, which also reduces benefi t to some extent through indirect job creation risk to investors. Although many internet fi rms seem and better access to work and markets. As governments to operate in separate markets, most if not all com- and the private sector get better at tailoring digital ser- pete with offl ine fi rms. Instant messaging apps com- vices to the poor, those gains will likely increase. pete with telecoms, search engines and social media sites compete with traditional media for advertising Creating jobs revenue, e-commerce fi rms compete with brick-and- The number of direct jobs created by digital technol- mortar fi rms, and mobile money competes with tra- ogies is fairly modest, but the number enabled by it ditional banks. Innovations triggered by this online- can be large. In developing countries, the ICT sector offl ine competition generally benefi t consumers, accounts for only about 1 percent of the workforce on especially when offl ine markets are distorted. Trans- average: less than 0.5 percent in Bolivia and Ghana, port service companies such as Uber, Lyft, Olacabs, and just under 2 percent in Colombia and Sri Lanka. and Didi-Kuaidi Dache have disrupted taxi markets In OECD countries, about 3–5 percent of the employ- that tend to be overregulated with restricted entry and ment is in this sector. Instagram, a photo sharing app, high prices. Similarly, TransferWise and Xoom have had just 13 employees in 2012 when it was bought by reduced regulatory rents in the fi nancial sector and Facebook for US$1 billion. Facebook had 5,000 employ- cut the prices of international transfers by up ees at the time—compared with 145,000 at Kodak at its to 90 percent. In Uganda, eKeebo allows independent peak in photographic fi lm in the 1990s. Yet Facebook’s or amateur chefs to provide and share home-cooked market value is several times what Kodak’s was back meals, circumventing restaurant licenses. then.13 ICT jobs, however, tend to pay well, and each high-tech job generates 4.9 additional jobs in other The internet supports job creation and sectors in the United States.14 In Kenya, the M-Pesa makes workers more productive digital payment system creates additional income People have an enormous desire to communicate for more than 80,000 agents. And China’s State Infor- and connect. The personal welfare gain from having mation Center estimates that the recent boom in the access to digital technology is clearly great. Does it country’s e-commerce sector has created 10 million also increase people’s economic opportunities? People jobs in online stores and related services, about 1.3 certainly use mobile phones and the internet more percent of the country’s employment. New opportu- for social purposes than for professional ones. But an nities for entrepreneurship and self-employment are emerging literature also indicates that people realize also growing rapidly in the digital economy. tangible economic benefi ts. Quantifying these benefi ts The internet’s ability to reduce transaction costs is diffi cult, but qualitative evaluation of the evidence increases opportunities for people who face barriers shows that benefi ts accrue most to those already better in fi nding jobs or productive inputs. This promotes off (table O.1). Those who have the skills to leverage inclusion for women, for persons with disabilities,

Table O.1 Benefi ts of digital technologies for workers and consumers: A scorecard Impact so far Potential impact Channel Poor Nonpoor Poor Nonpoor

Creating jobs In the ICT sector and occupations Negligible L Negligible L

In sectors that use ICT L M L M Increasing worker productivity Increasing returns to human capital L M L H

Connecting people to work and markets M H H H Benefi ting consumers Increasing consumer surplus M H H H Source: WDR 2016 team. Note: Poor refers to the bottom 20 percent of the welfare distribution. The diff erential impact summarizes the discussion in chapter 2 in the full Report and is a qualitative assessment of the evidence. ICT = information and communication technologies; L = low; M = medium; H = high. OVERVIEW 15

Box O.3 Bridging the disability divide through digital technologies

Over 1 billion people around the world have disabilities, and message service (SMS), instant messaging, telephone relay, 80 percent of them live in developing countries. Persons and video captions reduce communication barriers for with disabilities face barriers to communicate, interact, persons with hearing and speech disabilities. Hands-free access information, and participate in civic activities. navigation and gesture-controlled interfaces assist persons Digital technologies are helping overcome some of these with severe mobility impairments in using digital devices. barriers. Technology enables multiple means of commu- But the mere existence of technology is an insuffi cient nication—voice, text, and gestures—to access information condition to bridge the gaps in the socioeconomic inclu- and engage with others. Voice recognition, magnifi cation, sion of persons with disabilities. A supportive ecosystem and text-to-speech functionality benefi t persons with is needed to drive the implementation of accessible digital visual, cognitive, learning, and mobility disabilities. Short technologies.

Source: Raja 2015, for the WDR 2016.

and for people in remote areas (box O.3). Impact out- remain high—almost 15 percent for an additional year sourcing brings internet-based jobs to the poor and of tertiary education in developing countries. vulnerable. The government of the Indian state of The biggest gains from digital technologies for the Kerala up the Kudumbashree project to outsource poor are likely to come from lower information and information technology services to cooperatives of search costs. Technology can inform workers about women from poor ; 90 percent of the women prices, inputs, or new technologies more quickly and had not previously worked outside the home. Sama- cheaply, reducing friction and uncertainty.15 That can source and Rural Shores link clients in the United eliminate costly journeys, allowing more time for States and the United Kingdom with workers in work and reducing risks of crime or traffi c accidents Ghana, Haiti, India, Kenya, and Uganda. Of global (box O.4).16 online workers on the Elance freelancing platform, Using technology for information on prices, soil part of Upwork, 44 percent are women, and many quality, weather, new technologies, and coordination wish to balance work and life. Among respon- with traders has been extensively documented in dents to a survey of online workers for this Report, agriculture (see sector focus 1 in the full Report). In the ability to work fl exible hours from home is con- Honduras, farmers who got market price information sidered the greatest advantage of online work. via short message service (SMS) reported an increase of 12.5 percent in prices received.17 In Pakistan, mobile Increasing labor productivity phones allow farmers to shift to more perishable but For the economy as a whole, the most profound higher return cash crops, reducing postharvest losses impact of the internet on individuals is that it makes from the most perishable crops by 21–35 percent.18 workers more productive. By handing off routine and The impacts of reduced information asymmetries repetitive tasks to technology, workers can focus on tend to be larger when learning about information activities with higher value. Judicious use of massive in distant markets or among disadvantaged farmers open online courses (MOOCs) or online teaching who face more information constraints.19 tools like Khan Academy lets teachers spend more time fostering discussion and working with students Increasing the consumer surplus who fall behind. Researchers can dedicate more Where the internet has led to a full automation of time thinking and innovating rather than searching services, many jobs have been lost—few travel agents, for information or duplicating other people’s work. booksellers, or music store employees are left. But Managers can work more easily with teams across these same dynamics have been a boon to consum- borders. These benefi ts are largest for the higher ers. There are new digital —such skilled. In fact, there has never been a better time to as e-books, digital music, and search engines. And the be a high-, as the returns to education internet has transformed existing ones—such as taxi 16 WORLD DEVELOPMENT REPORT 2016

Box O.4 Digital dividends and the bottom billion

The poor benefi t from digital technologies, but only connected to the internet. In the , modestly in relation to the true potential. Nearly 7 of 10 one month of internet access costs more than 1.5 times the people in the bottom fi fth of the population in developing annual per capita income. Even mobile phones are expen- countries own a mobile phone, improving their access to sive: the median mobile phone owner in Africa spends over markets and services. In rural Niger, agricultural price infor- 13 percent of her monthly income on phone calls and text- mation obtained through mobile phones reduces search ing. And many poor lack the basic literacy and numeracy costs by 50 percent.a In rural Peru, access to mobile phones skills needed to use the internet. In Mali and Uganda, about boosted household real consumption by 11 percent between three-quarters of third-grade children cannot read. In 2004 and 2009, reducing poverty by 8 percentage points Afghanistan and Niger, 7 of 10 adults are illiterate. and extreme poverty by 5.4 percentage points.b In advanced economies the poor face the prospects The poor can benefi t from digital technologies even when of stagnant wages and fewer opportunities, as they are they don’t own a mobile phone or a computer. For exam- increasingly forced to compete with those displaced by ple, a digital ID, by giving millions of poor people an offi cial automation. Digital technologies can also exacerbate identity, increases their access to a host of public and private socioeconomic disparities. For example, the internet voting services. In Narma Dih—a village in Bihar, India, with no on municipal budget proposals in the state of Rio Grande electricity or all-weather roads—poor farmers benefi t from do Sul in Brazil and citizen engagement initiatives such as digitally enabled agricultural extension services from Digital Uganda’s U-report show that the new users are more likely Green, an NGO (nongovernmental organization) that trains to be male, young, university educated, and wealthy—those farmers using locally produced how-to videos.c already better off before the internet’s advent.d Yet the poor are capturing only a modest share of Rapid technological progress will increasingly enable the digital dividends. While a majority of the poor have the poor to aff ord and use many digital services. But their a mobile phone, they can’t access or aff ord the internet. ability to reap dividends from these investments will be In Latin America, fewer than 1 in 10 poor households is largely determined by providing the analog complements.

Source: WDR 2016 team. a. Aker and Mbiti 2010. b. Beuermann, McKelvey, and Vakis 2012. c. Chomitz 2015. d. Spada and others 2015; Berdou and Lopes 2015.

and hospitality services, health, education, and retail. an average of US$50 a month for services it now gets This has increased the variety of goods and services for free on the internet. Time-use data in the United available, including those for leisure. The internet States suggest that the median individual gains more thus enhances consumer welfare, but in ways that are than US$3,000 annually from the internet. In Estonia, hard to measure. digital signatures saved 20 minutes per transaction. People’s perceptions are that digital technologies And a study of the time costs of searching for infor- have certainly made them better off. In 12 countries mation shows that the average online search tends surveyed in Africa, 65 percent of people believe that to be 15 minutes faster, the results are more accurate their family is better off because they have mobile and relevant, and the experience more enjoyable than phones, whereas only 20 percent disagree (14.5 per- offl ine search in a library. On average, people might cent not sure).20 And 73 percent say mobile phones realize a consumer surplus as high as US$500 a year help save on travel time and costs, with only 10 per- from such services, adding up to vast benefi ts when cent saying otherwise. Two-thirds believe that having aggregated over all users. a mobile phone makes them feel more safe and secure. Some studies have attempted to quantify the The internet can make governments more economic value of these gains. A McKinsey survey capable and responsive of consumers in France, Germany, the Russian Fed- Governments provide services that are typically eration, Spain, the United Kingdom, and the United nontradable, often lack scale, and are not subject to States in 2010 found that a household is willing to pay market competition. Raising effi ciency in the public OVERVIEW 17 sector is thus challenging, and one might expect improved service effi ciency. E-procurement helped the internet to bring large benefi ts in public service India and Indonesia inject more competition into the provision. There are indeed many examples where process by increasing the probability that the win- the internet has raised the capabilities of the public bidder comes from outside the project’s region. sector. Better tools for communicating with citizens This also improved the quality of infrastructure. But and providing information also allow greater partici- a majority of public sector digital technology projects pation—through inclusion in government assistance fail to achieve the project objectives, resulting in con- programs, or feedback to and monitoring of public siderable fi scal waste.22 offi cials. And the internet helps citizens to connect Digital technologies can also improve management online and organize for collective action in order to by monitoring the performance of workers. A small but put pressure when government performance falls growing impact evaluation literature reports generally short of people’s expectations. positive effects of technology-based monitoring on worker absenteeism when combined with other insti- Expanding participation tutional reforms.23 In Uganda, where teacher absen- Lack of identity is an impediment for poor people to teeism is estimated at 27 percent, head teachers use exercise their basic democratic and human rights. mobile phones to record attendance and transmit data Where civil registration systems are weak or non- to a central database that generates weekly reports. existent, many of the poor are simply not counted. Combined with incentive pay for teachers tied to atten- Digital identifi cation can help overcome barriers dance, the program reduced absenteeism by 11 percent- to participation. Many countries have introduced age points. The internet also provides real-time data for general-purpose digital identity (ID) schemes or better planning and management of service facilities. specifi c systems for elections or to manage postcon- In Ghana, Kenya, Tanzania, and Zambia, health work- fl ict transfers—with numerous benefi ts, including ers use mobile phones to report counterfeit drugs and making the public sector more effi cient. Nearly 900 stock-outs. Aggregated in a central database and geo- million Indians have been issued digital IDs in the graphically mapped, this information helps adminis- past fi ve years, which they are using to open bank trators address drug and equipment . accounts, monitor attendance of civil servants, and Providing citizens the opportunity to give specifi c identify recipients of government subsidies. Nigeria’s feedback quickly has helped improve performance in e-ID revealed 62,000 public sector “ghost workers,” many instances. Mobile-phone apps like SeeClickFix saving US$1 billion annually. But the most important and FixMyStreet in the United States and the United benefi t may be in better integrating marginalized or Kingdom let users report potholes, graffi ti, and illegal disadvantaged groups into society. dumping. Governments can report back on fi xes, clos- Digital technologies also enable the poor to vote ing the feedback loop. Internet call centers enabling by providing them with robust identifi cation and citizens to report problems and track the status of by curtailing fraud and intimidation through better their requests are now standard in Barcelona, Buenos monitoring. Mobile phones enable citizens to report Aires, Muscat, Rio de Janeiro, Seoul, and Ulaanbaatar, instances of violence and voter intimidation, improv- to name a few cities. The Nairobi water company uses ing electoral participation. In Mozambique, SMS MajiVoice, and one of the electricity supply com- messages allowed citizens to report electoral irregu- panies in the Dominican Republic, EDE Este, uses a larities and increased voter turnout by 5 percentage similar system to receive complaints, track their reso- points.21 Ushahidi and Uchaguzi are crowdsourced lution through an automated workfl ow, and regularly applications that report and map election violence update citizens on progress. When implemented well, in Kenya. By multiplying the sources of information, citizens eagerly take up the opportunity to give feed- the internet can reduce the risk of media capture and back, and resolution time declines (see fi gure O.12). make censorship diffi cult. Advancing voice Improving public sector capability Governments, particularly those in digitally advanced The internet raises effi ciency and productivity countries like Estonia, the Republic of Korea, and through automation and data-driven management. Singapore, are beginning to take advantage of data Almost all countries have tried to automate tax and analytics and digital platforms for faster, more customs administration, as well as budget preparation, informed, and integrated policy making. The internet execution, and accounting. Results have been mixed. also opens new avenues for . E-fi ling reduces tax costs, and one-stop Iceland has experimented with crowdsourcing its computerized service centers and online portals have constitution, and Brazil and Estonia have explored 18 WORLD DEVELOPMENT REPORT 2016

Figure O.12 More complaints were resolved more quickly in the Nairobi water utility after the The risks: Concentration, introduction of digital customer feedback inequality, and control

120 60,000 So, the internet can be an effective force for develop- ment. But as the Report documents, the benefi ts too 100 50,000 often are not realized, and the internet sometimes makes persistent problems worse. Why? The key 80 40,000 insight is that for complex occupations, business activities, or public services, the internet usually can 60 30,000 make only a portion of tasks cheaper, more effi cient, or more convenient through automation. Another

Days to resolve 40 20,000 portion still requires capabilities that possess MajiVoice introduced in abundance but computers do not. Many traditional 20 10,000 tasks of an accountant or bank teller are now auto- Number of cumulative complaints mated, such as making calculations or processing 0 0 withdrawals. Others require complex reasoning or June December June December 2013 2013 2014 2014 socioemotional skills, such as designing tax strategies or advising clients. Likewise, many public services Resolution time Complaints submitted involving provision of information or routine permis- Complaints resolved sions can be automated. But others, such as teaching or policing, need a high degree of human discretion, Source: World Bank 2015. Data at http://bit.do/WDR2016-FigO_12. tacit knowledge, and judgment. Note: For more details see fi gure 3.11 in the full Report. Many problems and failures of the internet sur- face when digital technology is introduced but the important analog complements remain inadequate. participatory lawmaking. By dramatically lowering What are these complements? The main ones are the cost of communication and coordination, social regulations that ensure a high degree of competition, media can overcome the traditional barriers to citizen skills that leverage technology, and institutions that collective action. A growing empirical literature has are accountable (fi gure O.13). also shown that cellphones and the use of Twitter and Facebook aided protests during the Arab Spring in the • When the internet delivers scale economies for Arab Republic of Egypt,24 antiwar demonstrations in fi rms but the business environment inhibits com- the United States,25 and citizen mobilization across petition, the outcome could be excessive concen- Africa.26 tration of market power and rise of monopolies, inhibiting future innovation. Figure O.13 Without strong analog complements, • When the internet automates many tasks but opportunities may turn into risks workers do not possess the skills that technology augments, the outcome will be greater inequality, DIGITAL rather than greater effi ciency. TECHNOLOGIES • When the internet helps overcome information barriers that impede service delivery but govern- ments remain unaccountable, the outcome will be greater control, rather than greater empowerment and inclusion.

Information Automation Scale The interplay between internet investments and without accountability without skills without competition reforms in complementary areas is at the core of pol- icy debates about technology impacts. A 2008 study by Claudia Goldin and Lawrence Katz,27 drawing on CONTROL INEQUALITY CONCENTRATION earlier work by Jan Tinbergen, framed these dynam- ics in the labor market as a “race between education and technology.” As technology progresses, some Source: WDR 2016 team. skills become obsolete. Workers must acquire new OVERVIEW 19 skills that help them become more productive with Figure O.14 Factors explaining the lower adoption of the help of that technology. Adjustment takes time digital technologies by businesses and will be painful for many, but this is how econo- mies progress. The sections that follow discuss risks Vested and complements in the private sector, in labor mar- interests kets, and in the public sector. Regulatory uncertainty Growing concentration: The nexus Analog economy Digital between regulations and technology economy One of the main mechanisms for the internet to pro- (4–5% of GDP; 1–2% of jobs) mote economic growth is competition. Information fl ows increase and speed up so that customers have Digital more choice and can compare prices more easily. Firms that use technology more effectively will do Source: WDR 2016 team. well and force others to follow suit. There is consid- erable evidence that this is happening throughout the economy, but three potential problems could emerge. First, while the internet has spread quickly in the or . The risk of allowing underregulated private sector of some countries, adoption among entrants into a market must therefore be weighed non-ICT fi rms has been slow in other countries. against the benefi ts to consumers from lower prices Larger, fast-growing, skill-intensive, export-oriented, and greater convenience. and urban fi rms tend to use digital technologies The third potential risk comes from the dominant more. The causes of these differences are not well position of many online platforms and internet inter- understood. Differences in adoption rates may simply mediaries. shows that fi rms are refl ect differences in income, sector characteristics, tempted to exploit a dominant position. Large inter- and management capabilities, but they could also be net fi rms may be no exception. The economics of the due to barriers to adoption (fi gure O.14). One possibil- internet favor natural monopolies,28 and some plat- ity is high import duties for digital goods and services forms now dominate their markets. They enjoy such in some countries. Another is market distortions high profi ts that they can quickly capture new mar- and protections that allow fi rms to maintain profi ts kets by buying out competitors or developing a rival without threats from more innovative entrants. For service; local startups, including those in developing example, fi rms in Mexico that faced competition countries, are left with tiny niche markets. Some of from China increased the number of computers per the biggest internet fi rms now face scrutiny from employee and became twice as likely to use the inter- regulators. Google, which captures almost one-third net for purchasing as those that didn’t face signifi cant of global digital advertising revenue,29 has been inves- competition. tigated for preferential placement of its own prod- Second, when online businesses enter the turf of ucts, exploiting third-party content, and exclusionary their offl ine counterparts, disruption can be great, practices in its placement of advertising.30 Amazon, and regulators are often unsure whether or how to the largest sales platform for book publishers, has react. Recently, “on-demand economy” fi rms Uber used its market power to enforce its pricing policies. and Airbnb have challenged established taxi and Safaricom, operating the M-Pesa payment system, hotel industries. Their platform business model is resisted the entry of competing service providers. scalable and global and has spawned numerous local The vast amount of identifi able personal information imitators. In cities from Paris to to Beijing, that many of these companies collect raises further the reaction has been a scramble by offl ine incum- challenges (box O.5). bents to keep these new competitors out, usually by It is too early to tell whether these problems will appealing to regulators to enforce established sector diminish the overall economic benefi ts from the regulations such as the knowledge of the city (in the internet or be mitigated by the sector’s low entry costs case of London cabbies) or insurance requirements. and rapid technological change. Consumers have This can be a valid appeal when regulations protect generally benefi ted from the internet-based busi- public safety and ensure minimum service levels. But ness models of existing and new fi rms. Markets are these new models often succeed because they enter extremely dynamic, so many advantages from scale heavily distorted markets with virtual monopolies or moving fi rst may be temporary. And greater size 20 WORLD DEVELOPMENT REPORT 2016

Box O.5 What Facebook “Likes” reveal—the convenience-privacy trade-off

When economists such as George Stigler and Richard Data collectors often sell the data to others. One data Posner wrote about privacy and economics in the early broker assembled an average of 1,500 pieces of informa- 1980s, they raised many issues debated today, but at that tion about more than half a billion consumers worldwide time the “storage and retrieval of information, and its accu- from information people provided voluntarily on various rate dissemination, [were] often extremely expensive.”a websites. But even easily accessible data such as Facebook Today, an enormous collection of identifi able information is “Likes” can predict sensitive characteristics including making service delivery more effi cient and more relevant. “sexual orientation, ethnicity, religious and political views, Service providers can better target or price their products personality traits, intelligence, happiness, use of addictive based on known characteristics and preferences. Search substances, parental separation, age, and gender.”b And engines provide more relevant search results. Health and smartphone sensors can infer a user’s “mood, stress levels, auto insurers can better price premiums with verifi able personality type, bipolar disorder, demographics (e.g., gen- information about exercise or driving behavior. And gov- der, marital status, job status, age), smoking habits, overall ernments can use data systems to reduce the bureaucratic wellbeing, progression of Parkinson’s disease, sleep pat- burden for citizens. In Estonia’s e-government system, terns, happiness, levels of exercise, and types of physical citizens never have to provide the same information twice. activity or movement.”c The problem is that few people know how these large The risks? Cybercrime such as identity theft when data amounts of data are collected and used—and who controls stored insecurely fall into the wrong hands. Discrimination them. Users are not always aware of and providers often when customers are charged a higher premium or inter- don’t tell what information is collected. The secret snoop- est rate, or denied a job based on erroneous information ing by governments can be for legitimate law enforcement they can’t easily correct. Persistence of dated information reasons, but sometimes violates laws and rights, as the that denies protection from embarrassing but irrelevant Edward Snowden revelations about spying by the security information or a second chance, which prompted Europe’s agencies of the United States, the United Kingdom, and “right to be forgotten” ruling. And perhaps most impor- others have shown. A consequence has been a new “data tant, reduced trust and thus suboptimal use of the internet. ,” where countries are demanding that data These concerns vary across societies. Fifty-eight percent of about their residents be stored within their territory or Nigerians and 57 percent of Indians believe private infor- favoring domestic technology that may be inferior or more mation on the internet is very secure, but only 18 percent of expensive, but is trusted more. French and 16 percent of German respondents do.d

Sources: WDR 2016 team, based on Peppet 2014; Castro 2013; 2014; Kosinski, Stillwell, and Graepel 2013. a. Posner 1981. b. Kosinski, Stillwell, and Graepel 2013. c. See Peppet (2014) for individual references. d. CIGI and Ipsos 2014.

allows large fi rms to provide services and products at Rising inequality: The race between skills low cost or free of charge, and their high profi ts fuel and technology investments in research and development (R&D). At If the internet and related technologies promote the same time, it is clear that competition and market growth, how are the gains shared in the labor market? structure on the internet are in many ways not so dif- While digital technologies raise productivity and ferent from the offl ine world. Policies need to ensure enhance overall welfare, labor market disruptions that all innovative companies can enter markets and can be painful and can result in higher inequality. compete on equal terms. Otherwise, the economic Global trends provide some indication. One is that performance between fi rms of different size and in the share of national income that has gone to labor, different countries could diverge further and contrib- especially routine labor, has fallen quite sharply ute to similarly divergent performance of national in many developing countries—though Brazil and economies. Ukraine are exceptions (fi gure O.15).31 Inequality has OVERVIEW 21

Figure O.15 Labor shares of national Figure O.16 Falling labor shares in national income income are falling in many countries, are associated with rising inequality including some developing countries Change in Gini coeffi cient vs. growth in labor share in national income, 1995–2010 Trends in labor shares in output since 1975 30 percentage points every 10 years 25 CHN Poland 20 Mexico FIN 15 DNK Hungary LVA Estonia 10 Bahrain BGR EST Slovenia 5 CRI Lithuania GBR South Africa 0 ITA NOR HND BLR Norway –5 ESP Luxembourg GRC

consumption or income (%) ARG Micronesia, Fed. Sts. –10 PAN PRY Namibia Change in the TUN IRL Latvia –15 New Zealand –20 China –30 –25 –20 –15 –10 –5 0 5 10 15 20 25 Finland Tunisia Change in the share of national output going to labor (%) Argentina Slovak Republic Source: Eden and Gaggl 2015, for the WDR 2016. Data at http://bit.do/WDR2016-FigO_16. Germany Austria Sweden increased more where this shift in incomes toward France Italy capital and away from labor has been higher (fi gure Australia O.16). A number of recent studies have linked techno- Taiwan, China logical change to this rising inequality (see chapter 2 Canada in the full Report). Japan Denmark A related trend is the polarization—or “hollow- Switzerland ing out”—of the labor market, not only in advanced United States economies, but increasingly also in many developing Netherlands Belgium countries. The share of employment in high-skilled Czech Republic occupations is up, as is the share of low-skilled jobs. Spain The share of middle-skilled employment, in contrast, Singapore is down in most developing countries for which United Kingdom Portugal detailed data are available (fi gure O.17). And these Bolivia types of jobs are often near the top of the income Turkey distribution in low-income countries, as in Africa. Armenia A notable exception to these global trends is China, Colombia Kenya where growing mechanization in agriculture has led Thailand to a (perhaps temporary) increase in routine, mid- Costa Rica level labor. Exceptions also include some countries Iceland Belarus rich in natural resources and commodity exporters, Moldova which include several countries in Central Asia and Korea, Rep. Latin America. Ukraine What explains all this? Machines can increasingly Brazil perform routine tasks more quickly and cheaply than –15 –10 –5 0510 humans, and much of what is considered nonroutine Change in labor shares in national income today—such as translation, insurance underwriting, or even medical diagnostics—computers might do

Source: Karabarbounis and Neiman 2013. Data at http://bit.do just as well tomorrow. Unlike previous technolog- /WDR2016-FigO_15. ical transformations such as the mechanization of 22 WORLD DEVELOPMENT REPORT 2016

Figure O.17 The labor market is becoming more polarized in many developing countries

Annual average change in employment share, circa 1995–circa 2012

2.0

1.5

1.0

0.5

0

–0.5

ercentage points –1.0 P

–1.5

–2.0

–2.5 y s s a eru India P Turke nzania SerbiaBolivi China a Ukraine Bhutan Ghana Panama Malaysia MauritiusT Uganda Thailand JamaicaSri Lanka Namibia Pakistan Ethiopia Philippine Hondura Costa Rica Mongolia BarbadosNicaraguaBotswana Guatemala South Africa El Salvador Kazakhstan

Macedonia, FYR Egypt, Arab Rep. Dominican Republic

High-skilled occupations (intensive in nonroutine cognitive and interpersonal skills) Middle-skilled occupations (intensive in routine cognitive and manual skills) Low-skilled occupations (intensive in nonroutine manual skills)

Sources: WDR 2016 team, based on ILO KILM (ILO, various years); the International Income Distribution database (I2D2; World Bank, various years); National Bureau of Statistics of China (various years). Data at http://bit.do/WDR2016-FigO_17. Note: The fi gure displays changes in employment shares between circa 1995 and circa 2012 for countries with at least seven years of data. The classifi cation follows Autor 2014. High-skilled occupations include legislators, senior offi cials and managers, professionals, and technicians and associate professionals. Middle-skilled occupations comprise clerks, craft and related workers, plant and machine operators and assemblers. Low-skilled occupations refer to service and sales workers and elementary occupations. For more details see fi gure 2.15 in the full Report.

agriculture or the automation of manufacturing, the than in more advanced ones, where many of these internet affects well-paying white-collar jobs even jobs have already disappeared (fi gure O.18). But it will more than blue-collar jobs. likely take longer in lower-income countries. Most Some mid-level workers will have additional skills of them are still fairly low-tech, with only about one- that allow them to switch to better-paid nonroutine third of urban jobs in a sample of developing coun- occupations in which technology tends to augment tries using any ICTs at work. And rates are still human capital and make skilled workers more pro- low, with a larger share of manual nonroutine labor, ductive. These workers will gain from technological so investments in technology will be less profi table disruption. In developing countries, returns to educa- for fi rms. This does not mean, however, that lower - tion are highest among those with tertiary education, income countries need not pay attention to these and they are higher and rising faster in ICT-intensive trends. Most important, even without signifi cant occupations.32 Those who do not have such skills will employment shifts, the nature of jobs is changing need to seek work in lower-skilled, nonroutine occu- toward skills that remain hard for technology to emu- pations, such as janitorial services, hospitality, or per- late: that is, advanced cognitive and socioemotional sonal care. Demand for such services could increase, skills. The policy response, besides rethinking social but perhaps not enough to prevent downward wage protection systems, is better and more responsive pressure as the available workforce in these sectors education and training—areas where reforms take grows. These dynamics are consistent with the rising many years to pay off. returns to education and income inequality we see in It is important to keep in mind the historical per- many countries. spective that job displacement and job losses from The implications for developing countries depend technological change are an integral part of economic on the pace of technological disruption. The share of progress. It is precisely rising productivity—as tech- occupations that could experience signifi cant auto- nology replaces some human labor but augments the mation is actually higher in developing countries skills of remaining and new workers—that generates OVERVIEW 23

Figure O.18 From a technological standpoint, two-thirds of all jobs are susceptible to automation in the developing world, but the eff ects are moderated by lower wages and slower technology adoption

Estimated share of employment that is susceptible to automation, latest year

100

80

60

40

be computerized (%) be computerized 20 Share of employment that can 0

India Malta Latvia BoliviaNepal Serbia AngolaChina OECD Cyprus Georgia UkraineNigeria EthiopiaBulgaria Panama Ecuador Albania Croatia Tajikistan Mongolia Paraguay MauritiusMalaysia RomaniaThailand Lithuania Uruguay NicaraguaCambodia Costa Rica Seychelles Argentina Uzbekistan El SalvadorGuatemalaBangladeshSouth Africa Kyrgyz Republic Macedonia, FYR

West Bank and Gaza Dominican Republic

Adjusted (technological feasibility + adoption time lags) Unadjusted (technological feasibility)

Source: WDR 2016 team. See fi gure 2.24 in the full Report for more details. Data at http://bit.do/WDR2016-FigO_18. Note: For more details see fi gure 2.24 in the full Report. OECD = Organisation for Economic Co-operation and Development.

growth and frees human and fi nancial resources technology and education, and the winners will be for deployment in sectors with higher returns. It those who encourage skill upgrading so that all can also reduces the need for humans to do physically benefi t from digital opportunities. hard, repetitive, or dangerous work. Such trends will be welcome in countries that are rapidly aging or Engendering control: The gap between where the population is declining, or in professions institutions and technology where skills are in short supply. Telemedicine and The internet was expected to usher in a new era of automated diagnostics, for instance, allow medical accountability and political empowerment, with experts to serve many more people, even remotely in citizens participating in policy making and forming areas with a of doctors. self-organized virtual communities to hold gov- And fears of “technological unemployment” go ernment to account. These hopes have been largely back to the industrial revolution. Even such think- unmet. While the internet has made many govern- ers as and the ment functions more effi cient and convenient, it has writer Isaac Asimov submitted to this fallacy. Keynes, generally had limited impact on the most protracted in the 1930s, predicted 15-hour workweeks by the problems—how to improve service provider account- end of the 20th century, and Asimov, in a 1964 essay, ability (principal-agent problems) and how to broaden expected that one of the most pressing problems for public involvement and give greater voice to the poor humanity by 2014 would be boredom “in a society of and disadvantaged (collective action problems). enforced leisure.” Yet over the centuries, economies Whether citizens can successfully use the internet have adapted to massive changes in labor markets— to raise the accountability of service providers depends the largest by far, being the shift out of agriculture. In on the context. Most important is the strength of 1910, there were 12 million farmworkers in the United existing accountability relationships between policy States. One hundred years later, there were only makers and providers, as discussed in the 2004 World 700,000 in a population more than three times larger. Development Report, Making Services Work for Poor People. Still, nobody can predict the full impact of techno- An examination of seventeen digital engagement ini- logical change in coming decades, which may be tiatives for this Report fi nds that of nine cases in which faster and broader than previous ones. What is clear, citizen engagement involved a partnership between however, is that policy makers face a race between civil society organizations (CSOs) and government, 24 WORLD DEVELOPMENT REPORT 2016

Table O.2 Classifying the digital citizen engagement cases CSO partners Impact Additional offline with Collective Government Case Location mobilization government feedback Citizen uptake response

Por Mi Barrio Uruguay L H

I Change My City India M H

Lungisa South Africa L H

Pressure Pan Brazil H M

Rappler Philippines H M

Change.org World H M

U-report Uganda H L

Huduma Kenya L L

Daraja Maji Matone Tanzania L L

FixMyStreet Georgia L L

Check My School Philippines L L

Barrios Digital Bolivia L L

e-Chautari Nepal L L

I Paid a Bribe India M L

Mejora Tu Escuela Mexico L L

Karnataka BVS India L L

Sauti Za Wananchi Tanzania L L Source: WDR 2016 team, based on Peixoto and Fox 2015, for the WDR 2016. Note: Examples are arranged by degree of government response. CSO = civil society organization. L = low; M = medium; H = high.

three were successful (table O.2). Of eight cases that Grande do Sul, online voting increased voter turnout did not involve a partnership, most failed. This sug- by 8 percentage points, but online voters were dis- gests that, although collaboration with government proportionately wealthier and more educated (fi gure is not a suffi cient condition for success, it may well O.19). Even in developed countries, engaging citizens be a necessary one. Another ingredient for success is continues to be a challenge. Only a small, unrepresen- effective offl ine mobilization, particularly because tative subset of the population participates, and it is citizen uptake of the digital channels was low in most often diffi cult to sustain citizen engagement. There of the cases. For example, Maji Matone, which facil- is no agreement among social scientists on whether itates SMS-based feedback about rural water supply the internet disproportionately empowers citizens or problems in Tanzania, received only 53 SMS messages political elites, whether it increases polarization, or during its fi rst six months of operation, far less than whether it deepens or weakens , in some the initial target of 3,000, and was then abandoned. cases even facilitating organized violence. Political participation and engagement of the poor The use of technology in governments tends to be has remained rare, while in many countries the inter- successful when it addresses fairly straightforward net has disproportionately benefi ted political elites information and monitoring problems. For more and increased the governments’ capacity to infl uence demanding challenges, such as better management of social and political discourse. Digital technologies providers or giving citizens greater voice, technology have sometimes increased voting overall, but this has helps only when governments are already responsive. not necessarily resulted in more informed or more The internet will thus often reinforce rather than representative voting. In the Brazilian state of Rio replace existing accountability relationships between OVERVIEW 25 governments and citizens, including giving govern- Figure O.19 Internet voting can increase voter ments more capacity for surveillance and control (box participation but can be biased toward more O.6). Closing the gap between changing technology privileged groups and unchanging institutions will require initiatives Profi le of online and offl ine voters in a participatory budgeting vote in Rio Grande that strengthen the transparency and accountability do Sul, Brazil, 2011–12 of governments. 60

Making the internet 50 universal, aff ordable, 40 open, and safe First-generation ICT policies involving market 30 competition, private participation, and light-touch Percent regulation have led to near-universal access and 20 affordability of mobile telephony, but have so far been less successful in spreading internet services. Much 10 of the explanation lies in continued policy failures such as regulatory capture, troubled , 0 ineffi cient spectrum management, excessive taxation <750 of the sector, and monopoly control of international Female Higher 0–6,000 >6,000 Secondary 750–1,500 gateways. At the same time the absence of global con- Elementary 1,50 sensus in dealing with the next-generation issues— Gender Education level Income level (BRL/month) such as privacy, cybersecurity, censorship, and inter- net governance—is resulting in more circumspect Offline voters Online voters and diverse approaches to regulating the internet Source: WDR team, based on Spada and others 2015. Data at http://bit.do/WDR2016-FigO_19. (box O.7 and fi gure O.20). Note: BRL = Brazilian real.

Supply-side policies: Availability, line), and mandating the incumbent to make local accessibility, and aff ordability access lines available to competitors at wholesale A useful framework for analyzing supply-side ICT prices (local loop unbundling). policies is to consider the that stretches • The most critical portion of the invisible mile from the point where the internet enters a country involves spectrum management, which requires (the fi rst mile), passes through that country (the mid- increasing the amount of spectrum available, dle mile) to reach the end user (the last mile), and cer- ensuring competitive access, encouraging sharing tain hidden elements in between (the invisible mile). of essential facilities, such as radio masts, and liber- alizing the market for spectrum resale. • The fi rst mile can be improved by liberalizing the market for satellite dishes and eliminating monop- In addition to pure ICT policies, almost everything oly status over the international gateway and cable that the private sector, citizens, or governments do on landing stations. the internet requires some essential building blocks • Strengthening the middle mile involves liberalizing (box O.8). the market for building and operating backbone networks, encouraging to the incum- Demand-side policies: Open and safe bent’s network, requiring all major infrastructure internet use programs (such as roads, railways, pipelines, and The challenges facing internet stakeholders today are energy distribution) to include provision for an as much about how networks are used (demand) as optical fi ber link, setting up internet exchange how they are built (supply). Global interconnected- points, and creating local caches for frequently ness introduces new vulnerabilities in areas where used content. coordination mechanisms are weak, still evolving, or • Government policies can encourage the provision based on nongovernment models. Threats to cyber- of last mile connectivity by permitting competing security, and censorship are undermining confi dence facilities, especially for intermodal competition and trust in the internet and increasing costs to busi- (between cable, wireless, and digital subscriber nesses and governments, resulting in economic losses 26 WORLD DEVELOPMENT REPORT 2016

Box O.6 Nailing Jell-O to the wall—restrictions on the fl ow of information

Governments also interfere directly with digital networks result refl ects broad societal preferences. In autocratic to control access to information. An early internet pioneer, countries, where use of the internet in government is often John Gilmore, claimed, “the Net interprets censorship as as high as in democratic countries (fi gure BO.6.1), leaders damage and routes around it.”a And Bill Clinton in 2000 face a dilemma. If they permit open discourse on the inter- said, “trying to control the internet is like trying to nail net, they risk challenges to their authority. If they do not, Jell-O to a wall.”b Yet private software vendors and state they risk isolating themselves from the global information institutions have fi gured out ways to censor access to inter- economy. This is a balancing act, and countries are becom- net content, whether by shutting down the entire national ing more sophisticated in calibrating their control—for web domain, as the Arab Republic of Egypt did in 2011 for example, censoring content that might encourage collec- fi ve days; by preventing access to specifi c domestic or tive action, but not individual criticism. foreign websites; or by targeting individuals’ blog posts Internet fi ltering and censorship impose welfare and or other social media postings. Google received 6,951 economic costs. First, the cost of censoring or fi ltering inter- requests from governments in 2013 to remove content from net content diverts public funds from other uses. Monitoring search results, with the largest numbers from Turkey, the domestic internet traffi c and selectively blocking foreign United States, and Brazil. Other countries, including China websites requires large fi nancial resources, technical know- and the Islamic Republic of Iran, block Google and some how, and dedicated staff —all of which could be deployed other internet sites completely, although such restrictions for more productive tasks. Second, fi ltering and methods to may change in the future. circumvent it can slow the speed of internet access, which Governments of all types restrict access to content such hurts business users. Third, fi ltering can restrict access to as child pornography, , insults, or criticisms of economically or scientifi cally useful information, such as authority fi gures, challenges to cultural or religious morals, the Google Scholar search engine for academic papers— or reporting of uprisings or accidents. When accountable indispensable in universities and labs. Fourth, in the view governments determine what should be censored, the of the European Union, for instance, blocking foreign

Figure BO.6.1 Autocratic governments have promoted e-government while censoring the internet

a. e-government provision, b. Internet filtering level, by government type by government type 1.0 Substantial

0.8 Selective 0.6

0.4 Suspected

Online index service 0.2

0 None –10 –5 0 5 10 Autocracy Democracy Autocracy Democracy Type of content filtered Political Social Conflict and security

Sources: WDR team, based on Polity IV 2015; UN 2014; Open Net Initiative 2013. Data at http://bit.do/WDR2016-FigBO_6_1. Note: The Polity IV project defi nes government types based on characteristics such as competitiveness and openness of executive recruitment, constraints on the chief executive, and regulation and competitiveness of participation in the political process. The combined score varies from –10 for a pure autocracy to +10 for a pure democracy. See the Polity IV user’s manual for details.

(Box continues next page) OVERVIEW 27

Box O.6 Nailing Jell-O to the wall—restrictions on the fl ow of information (continued) websites may be considered a nontariff trade barrier. Local and domestic champions will not face as much innovation- companies will fi ll the gap. This could be considered an inducing competition. Fifth, widespread censorship means economic benefi t or transfer rather than a cost. But it pre- that people avoid discussing and exchanging ideas openly, vents domestic users’ access to possibly better products, a prerequisite for an innovative and productive society.

Sources: WDR 2016 team, based on Saleh 2012; King, Pan, and Roberts 2013; Bao 2013; HRW 2015. a. Elmer-Dewitt 1993. b. Clinton 2000.

Box O.7 Is the internet a public good?

The internet does not have all of the characteristics of a benefi t when everyone else is better informed and when pure public good. Access to the internet often requires a public services are provided electronically at lower cost. fee, so individuals can be eff ectively excluded from its use. The private sector should take the lead in providing But once on the internet, the consumption of information internet infrastructure and services because the business by one user does not reduce its availability to others, so in case is usually compelling. But public investment or inter- that sense it is nonrivalrous (although capacity constraints vention is sometimes justifi ed where the private sector is can slow down access). One way to describe the internet is unable to provide aff ordable access. Historical precedents as a club good that is excludable but nonrivalrous, similar include the United States Communications Act of 1934, to ; or if bandwidth is scarce, as a private which called for universal “wire and radio communication good with strong positive externalities—everyone benefi ts service,” even in remote rural areas. Some countries have as more people come online. As more essential services gone further. Finland, for example, has defi ned access to and information migrate to the web, anyone without access the internet at broadband speeds as a legal right and pur- almost becomes a second-class citizen. And all citizens sues a universal access policy.

Source: WDR 2016 team.

Box O.8 The four digital enablers

The WDR 2016 looks at how the internet increases produc- e-commerce. Electronic transfers reduce the cost of sending tivity of businesses, opportunities for people, and the eff ec- remittances. Peer-to-peer lending can vastly improve the tiveness of governments. Across these domains, four major fi nancial access of startups. Governments can make pay- enablers of digital development are critical. Four spotlights ments and social transfers at lower cost and with less fraud in the Report discuss their benefi ts and potential risks. and leakage. However, if fi nancial regulations don’t keep pace with the rapid technological progress, these innova- Digital fi nance. Banks have been early and eager adopters tions could risk aff ecting the stability of the overall system. of digital technologies, but many of the major innovations, such as online payments, mobile money, and digital cur- Social media. Social networks are fundamental to human rencies, have come from nonbank institutions, including society, and digital technologies have accelerated their telecom and internet companies. Some of these innovations formation. More than one-fi fth of the world’s population is took root fi rst in developing countries, where they over- now believed to be a member of one or more social media came shortcomings in traditional fi nancial systems. Their platforms. These platforms have been credited with facilitat- benefi ts are distributed widely. Secure online payments fuel ing economically benefi cial interactions, channeling users’

(Box continues next page) 28 WORLD DEVELOPMENT REPORT 2016

Box O.8 The four digital enablers (continued)

behavior in ways that are consistent with development, pro- system, including legally binding contracts and voting in viding a platform for information and dissemination during national elections. natural disasters and emergencies, and encouraging political mobilization and social change. Some analysts think that Data revolution. In harnessing data for development, atten- social media played a critical role in recent events such as tion focuses on two overlapping innovations: “big data” and the Arab Spring and Occupy Wall Street, and thereby were open data. Big data are voluminous or fast, and they come instrumental in spreading democratic ideas, although many from myriad sources—from satellites to sensors and from remain skeptical of their actual impact. There is still much to clouds to crowds. Big data analytics is being deployed to learn about the role social media can play in development. improve traffi c planning, estimate macro aggregates (also While a source for innovative ideas, social media also remain referred to as “nowcasting”), track the spread of epidemics, conduits for gossip, slander, misinformation, harassment, and improve credit scoring and job matching. Open data are , and crime. One important lesson is that the impact those that are freely and easily accessible, machine-read- of social media on development seems to be highly specifi c able, and explicitly unrestricted in use. Governments are, to context. Variation in access to technology, education, and or could be, the most important source of open data. broader sociopolitical context matters. For instance, there is Exuberant estimates of the current and potential economic evidence that people in more autocratic countries are less value of big data and open data range from hundreds of bil- likely to forward information (for example, by re-tweeting it). lions to trillions of dollars per year. Yet sustained, impactful, scaled-up examples of big data and open data in developing Digital identity. Being able to prove who you are may seem countries are still relatively rare. Most big data are in private trivial, but it can be transformational for those excluded hands—large telecom and internet companies—which are from jobs and services. Simple electronic identifi cation sys- reluctant to share it for fear of jeopardizing customer pri- tems, often using biometric characteristics, have become vacy or corporate competitiveness. Public agencies, too, are an eff ective platform for secure bank transactions, voting, reluctant to share data, even when they have large public accessing social services, paying utility bills, and much benefi ts. For example, of countries surveyed by the Open more. Many countries, from Moldova to Nigeria and Oman, Data Barometer, one-third of the high-income countries have introduced digital IDs. India is on track to register its and 85 percent of developing countries had made little or entire population using its Aadhaar digital ID. In Estonia and no progress in opening map data. Reasons include lack of other countries, thousands of diff erent types of public and technical skills, inadequate resources, and unwillingness to private transactions are verifi ed with a unique electronic ID expose data to scrutiny.

Figure O.20 A policy framework for improving connectivity

First mile New technologies Competition policies Supply side Middle mile

Last mile Public-private partnership Effective regulation Universalizing Invisible mile connectivity Censorship and content filtering Awareness and investment Cybersecurity Global cooperation Demand side Online privacy Multistakeholder approach Internet governance

Open-access ecosystem for innovation, tech hubs, and maker

Source: WDR 2016 team. OVERVIEW 29 as well as higher security spending. For privacy and • Workers, entrepreneurs, and public servants who data protection, different countries are taking quite have the right skills to take advantage of opportu- different approaches. That makes it harder to develop nities in the digital world global services. Ensuring safe and secure access will • An accountable government that effectively uses require enhanced international collaboration, based the internet to empower its citizens and deliver on a multistakeholder model. services.

Analog complements for a What these priorities is that core ele- digital economy ments of the development agenda—business regula- tions that ease market entry, education and training The internet has great potential to promote economic systems that deliver the skills that fi rms seek, and development, but only some of that potential has capable and accountable institutions—are becoming so far been realized. It disrupts established markets more important with the spread of the internet. Not for products, services, and labor, and it disrupts the making the necessary reforms means falling farther public sector—major reasons for the frequent reluc- behind those who do, while investing in both tech- tance to adopt and deploy the internet more broadly. nology and its complements is the key to the digital But the benefi ts will come to those who embrace the transformation. changes the internet brings, not to those who resist Internet use still varies greatly between countries, them. And the way to get internet-enabled inclusive as does the quality of complements, and both tend to growth without long-term disruption is to strengthen move up with income (fi gure O.21). Policy priorities the analog complements of digital investments (box change as countries move along the digital transfor- O.9). Three policy objectives emerge from the analysis mation (fi gure O.22). Countries where internet use is in the Report: still low should lay the foundation—such as removing barriers to internet access and adoption, promoting • A business environment where fi rms can leverage basic and digital literacy, and using the internet for the internet to compete and innovate for the benefi t elementary government functions like provision of of consumers information. As countries transition to higher levels

Box O.9 Technology and complements: Lessons from academic research

Recent research on growth, labor markets, and governance abilities will leverage technology to become more produc- has taken a fresh look at the interplay between technology tive. Consider a modern offi ce assistant who uses digital and other factors. These insights inform the discussion of technologies to perform routine tasks quickly, and now policy priorities in this Report. spends far more time on personal interaction, complex scheduling, and other tasks that computers cannot easily Rules. Technology interacts with rules (such as regulations perform. and standards) to create new ideas, such as new ways of producing goods and services. Technology is traded across Institutions. Technology interacts with discretion. Many markets and borders, while most rules are established tasks in government can also be automated, but others locally. When it encounters rules that do not match, tech- involve a high degree of judgment. That means that even as nology fails to deliver the expected benefi ts. New busi- the internet can make many public service functions more nesses can acquire internet technology to reduce prices effi cient, the benefi ts will be limited when government and increase convenience for consumers, but they will not offi cials and workers do not have the incentives to use the be able to enter the market and compete if local regulations technology for the public good. Teacher attendance can protect incumbents. be fairly easily monitored using digital technologies, but the quality of teaching depends on the teacher’s training, Skills. Technology interacts with workers’ skills. It allows resources, ability, and motivation. routine tasks to be automated. Workers with the right

Sources: Romer 2010; Autor 2014; Pritchett, Woolcock, and Samji, forthcoming. 30 WORLD DEVELOPMENT REPORT 2016

Figure O.21 The quality of complements and technology rises with incomes

1.0

TRATRANSFORMING FIN CHE SGP CAN 0.9 NORR NZL AUS DNK SWE NLD IRL BEL USA GBR EST SVN ISL LUX 0.8 DEU CYP FRA LTU LVA JPN PRT KOR MYS MLT AUT BRB ARE CZEZE CHLHL 0.7 HUN QAT ISR ESP MUS POL SSVK GEO HRV ROU ITA TTTO CRI BGR GGRCC MKD URYY SYC PANN JJOR 0.6 BWA SRB ARM GUY ALBB RUS LKA UKR MEX KAZ ZMB SLV TUN MMNG TUR KGZ THA COL GHA RWA ZAF SAU BHR 0.5 LSO BLZ PHL MAR BRN BEN UZB PER CHCHN SENEN VNM ARG IDN DOM LBR TJK KWT SWSWZ GAB NIC IRN

Complements NAM KEN GTM TZA ECU BRA CMR BGDGD 0.4 LAO BDI SLE NNPL IND EGY MOZMO GGMB HNHND PRY PAK BOL MWIWI DZA MRT NGAG 0.3 ETH YEMY ZWE MLI LBY TTRANSITIONINGN I G AGAGO VEN KHMKH 0.2

HTI

0.1 EMEREMERGINGM E RG N 0 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Technology

High-income Lower-middle-income Upper-middle-income Low-income

Source: WDR 2016 team. For more details see fi gure 5.3 in the full Report. Data at http://bit.do/WDR2016-FigO_21. Note: Technology is measured by the Digital Adoption Index (DAI). DAI is based on three sectoral subindexes covering businesses, people, and governments, with each subindex assigned an equal weight: DAI (Economy) = DAI (Businesses) + DAI (People) + DAI (Governments). Each subindex is the simple average of several normalized indicators measuring the adoption rate for the relevant groups. Similarly, complements is the average of three subindicators: starting a business; years of education adjusted for skills; and quality of institutions.

of internet use, they require effective competition the internet for most functions of government and regulation and enforcement—including easy fi rm for more participatory policy making. entry and exit; a greater focus on advanced cognitive and socioemotional skills that are augmented by tech- Regulations that promote competition and nology; and effective e-government delivery systems entry for provider management and citizen engagement. Digital adoption by fi rms varies among countries, Countries in advanced stages of the digital transfor- and there are reasons for it to be slow. Most funda- mation need to tackle some of the most challenging mentally, adoption requires knowledge about the tasks. They need to fi nd ways to facilitate “new econ- technology, access to it, and knowledge of how to omy” competition, to ensure lifelong learning and best apply it. But the most important driver is com- respond to the changing nature of work, and to use petitive pressure, as fi rms adopting new technology OVERVIEW 31

Figure O.22 Policy priorities for countries that are emerging, transitioning, or transforming

EMERGING TRANSITIONING TRANSFORMING

REGULATIONS Competition Remove barriers Platform regulation and that promote to adoption competition competition and entry enforcement

SKILLS Foundational skills Prepare for Facilitate to leverage and basic ICT careers lifelong digital opportunities literacy instead of jobs learning

e-government Participatory INSTITUTIONS Mobile phone– based services delivery and policy making that are capable and monitoring citizen and digital and accountable engagement collaboration

Note: ICT = information and communication technology. raise productivity and those who don’t fall behind. benchmarking exercises and information programs This highlights the critical role of a country’s business can be effective. And to allow more innovative com- climate. It includes laws and regulations that ensure panies to enter markets easily, countries need to easy entry and exit of fi rms, and an open trade regime improve fi rm registration and create greater market that exposes companies to foreign competition and transparency to reduce price collusion, market shar- investment. There is a dimension ing, and rigged public procurement. E-government to this as well—special interests infl uence regulators systems such as online business registration and to keep markets closed to competition. This lessens e-procurement systems can simplify these processes the need for fi rms to reach for the technological fron- and produce more openness. tier. Where banks are heavily regulated and protected from new market entrants, they have less incentive Increase competition through eff ective to invest in effi ciency-boosting technology that might regulation and enforcement also help them serve customers better or reach new State control in economic sectors, barriers to entre- ones. But competition policy and enforcement are preneurship, and restrictions on trade and investment complex, and many low-income countries lack capac- reduce the incentives for fi rms in protected sectors to ity to design and implement them effectively. use digital technologies. Most countries have a com- petition authority, although many were set up fairly Lower the barriers to digital adoption recently and enforcement varies, especially when In countries where the digital economy is still emerg- the state or politically connected fi rms benefi t from ing, the priority is to facilitate connectivity and market restrictions. Moreover, the internet makes develop the foundation for effective competition reg- it easy to deliver services online from anywhere ulation. Although 74 mostly middle- and high-income in the world, so how trade in services is regulated countries have unilaterally removed tariffs on ICT becomes increasingly important. Ethiopia, India, capital goods, computers and smartphones are still and Zimbabwe have the greatest restrictions on ser- treated as luxury goods in some countries, including vice trade, but many other countries restrict specifi c Turkey, where taxation adds almost half to the price of services such as legal or accounting tasks. Countries mobile handsets.33 Djibouti’s on computers is 26 can increase the competitiveness of their economies percent. Many countries treat their telecom fi rms as and encourage greater use of digital technologies by cash cows. Where fi rms may have limited knowledge gradually reducing market distortions while building about how the internet can improve their business, up effective competition enforcement. This applies as 32 WORLD DEVELOPMENT REPORT 2016

Box O.10 Opening the M-Pesa mobile money platform to competition

Safaricom’s mobile money system is a well-known success because Safaricom incurred high costs developing the sys- story. It was able to grow quickly because Kenya’s banking tem. But in 2014, Kenya’s Competition Authority changed regulators initially decided to take a hands-off approach. the rules and opened the system to alternative mobile For seven years, Safaricom maintained a dominant position operators. The of transfers of up to K Sh through exclusivity agreements locking agents into the 500 (US$4.91) fell from K Sh 66 to K Sh 44 (US$ 0.43). system. Initially such arrangements were perhaps justifi ed

Source: Plaza, Yousefi , and Ratha 2015, for the WDR 2016.

much to traditional businesses that use the internet have the benefi t of being able to learn from the expe- as to internet platforms (box O.10). rience in the transforming countries before devising their own solutions. Tailor “new economy” regulations to ensure competition Skills for the digital economy Internet fi rms create new business models and Technological change means that many routine tasks change market structure, posing new challenges for will soon be done by machines. In contrast to previ- regulatory authorities. On-demand economy fi rms ous episodes, the internet will also make many tasks like Uber and Airbnb scaled up traditional ride shar- carried out in white-collar jobs redundant. This puts ing and subletting to a global scale. But regulators a premium on different types of skills that automa- struggle to determine whether these companies are tion complements rather than replaces (fi gure O.23). taxi or hotel companies or simply software providers. Education systems have been slow to respond to this Offl ine competitors complain that they do not follow challenge. Furthermore, the pace of change is fast, the same regulations. Where these industries tend and the types of skills in demand change quickly. So to be overregulated and their markets distorted, as is workers will have to upgrade their skills frequently often the case in the taxi business, this new competi- throughout their careers. These dynamics already tion can encourage a general overhaul of the industry. play out in many transforming and some transition- In the United States, cities like New York and states ing countries, but even for emerging countries it is like Massachusetts have begun to develop appropri- not too early to prepare. ate regulations for these platforms, imposing safety and tax obligations but also reducing their competi- Start early with foundational skills tors’ regulatory burdens. Skills development starts at birth and lasts a lifetime. Similar regulatory puzzles are posed by fi rms Good parenting and early stimulation prepare chil- such as Amazon, Facebook, and Google. For example, dren for school, where cognitive and socioemotional Google is known as a search engine company but is foundations are laid. Technology can play a role. better described as an advertising fi rm. These fi rms Even though the record on simply providing laptops confound conventional competition law because they or tablets to students is mixed, videoconferencing do not act as traditional monopolies. Their services with English speakers from the Philippines has are often free to consumers. But given their domi- improved learning among fi rst graders in Uruguay. nance in the markets for online ads and books, they Khan Academy provides resources for independent have considerable leverage over marketers and book- learning, and using a gaming approach to math sellers. This is similar to credit card companies’ posi- teaching benefi ted grade four children in . tion with respect to retailers. Research by economists But in these and many other cases, one factor was such as Jean Tirole has shown that regulations in more important: the quality of teaching. It is no such industries must be carefully tailored to guaran- coincidence that Finland, one of the most connected tee competition and avoid harm to consumers. These and best-performing countries in educational test- are very challenging problems, and most pressing in ing, uses very little technology in the classroom. the transforming countries. Developing countries It takes time to improve the quality of teachers, OVERVIEW 33

Figure O.23 The types of skills needed in a modern economy

Cognitive Social and behavioral Technical

Literacy, numeracy, and Manual dexterity and the use Socioemotional skills and higher-order cognitive skills of methods, materials, tools, personality traits (for example, reasoning and instruments and creative thinking) Openness to experience, Technical skills developed through Raw problem-solving ability conscientiousness, postsecondary schooling or versus knowledge to extraversion, agreeability, training or acquired on the job solve problems and emotional stability

Verbal ability, numeracy, Self-regulation, , mind-set, Skills related to specific problem solving, memory, decision making, and occupations (for example, and mental speed interpersonal skills engineer, economist, IT specialist)

Source: WDR 2016 team, adapted from Pierre, Sanchez Puerta, and Valerio 2014. Note: IT = information technology. however. But technology can help here as well, as that traditional education systems do not teach Rio de Janeiro’s Educopedia platform shows (box and that are hardest to measure. Many countries O.11). Using technology to closely guide teaching are rethinking their approach. Singapore is moving is a second-best option that can improve learning from a fairly rigid “effi ciency driven” model that outcomes at modest cost where teacher training is tried to get the best results from inputs (teachers unlikely to improve quickly. This is the model used and fi nance) to an “ability driven” model that empha- by the for-profi t Bridge Academy in Kenya and else- sizes project work and fewer assessments in place of where, where scripted instruction and automated frequent testing. Colombia’s Escuela Nueva model, administrative tasks help provide education at low now serving 5 million students in 16 countries, also cost. Although yet to be evaluated rigorously, such focuses on group learning and problem solving. These approaches hold promise to improve education. approaches change the relationship between teacher and student. No longer simply sources of information, Rethink curricula and teaching methods teachers now must instruct students in how to fi nd Today’s education systems need to prepare students information and apply it in a new and unexpected for a career and not only a job. Modern labor markets context. This requires changes in teacher training. require creativity, teamwork, problem solving, and crit- There are many examples of how digital technologies ical thinking in ever-changing environments—skills can assist teachers and students—by allowing group

Box O.11 Mobilizing technology in teaching in Rio’s Educopedia

Rio de Janeiro’s education department developed the serves almost 700,000 students. It has yet to be formally Educopedia online platform of lessons and other resources evaluated, but together with other reforms it likely con- in 2010 to improve public school teaching. The system tributed to a more than 20-percent increase in the Basic focuses as much on providing materials for the teacher to Education Development Index in middle schools between improve lessons as on giving students access to learning 2009 and 2012. And 80 percent of Rio’s students agreed resources. The system uses multimedia resources including that Educopedia helps their learning eff orts. videos, interactive quizzes, and digital libraries. It now

Source: Bruns and Luque 2014. 34 WORLD DEVELOPMENT REPORT 2016

work among classrooms connected online, apps that Improve informational services and stimulate creativity and problem solving, and games monitoring designed for education (“gamifi cation”). Although internet access is still low in many emerg- ing economies, mobile phones are widespread and Develop advanced technological skills and have great potential for improving services. In the encourage lifelong learning health sector, simple mobile phone–based remind- As more and more parts of the economy rely heav- ers for taking medications have been effective for ily on the internet, demand for advanced ICT skills HIV patients in Malawi and for providing maternal will also grow. Only a small share of the workforce health information in the Democratic Republic of will be involved in developing software or systems Congo. Phones can support the monitoring of teach- design, but exposing children to coding and basic ers or other government workers where absenteeism ICT concepts can infl uence career choices for some is a problem, even if monitoring alone is insuffi cient and impart a basic understanding to many. NairoBits, to improve service quality or outcomes (box O.12). a youth organization in Kenya, exposes underpriv- Monitoring also becomes important in provision ileged young people from informal settlements to by nonstate entities in weak institutional contexts, web design and other ICT skills, while AkiraChix where for-profi t or non-profi t organizations deliver reaches out to “geek girls.” Women tend to be under- services often funded by the government. And digital represented in ICT fi elds, and encouraging girls to technologies can improve electoral accountability. In enter such professions and ICT companies to create Afghanistan, Kenya, and Mozambique, election mon- welcoming environments for women will increase itoring using cellphones helped uncover fraud and the available workforce in fi elds with rising demand reduced election violence. This can complement— for labor. With technology likely continuing to get or, even in low-capacity settings, replace—more more advanced and affecting ever more occupa- demanding approaches such as biometric identifi ca- tions, workers need to continuously reevaluate and tion (see spotlight 4 in the full Report). upgrade their skills. Much of that will happen outside the formal education system, but governments can Strengthen e-government delivery and provide the incentives for fi rms and workers to create citizen engagement the mechanisms for lifelong learning. Where investments to automate government service delivery have advanced, complementary improvements Institutions that are accountable to in regulations, interdepartmental cooperation, and citizens streamlining become more important. Rather than just Although the internet has enabled many governments replicating cumbersome processes such as business to provide some basic services more effi ciently, tech- regulations online, automation provides an opportu- nology so far has not strengthened accountability. For nity for simplifying steps, increasing the impact as well policies, this implies a dual strategy: tailoring the appli- as the transparency. E-procurement systems reduce the cation of digital technologies to environments with risk of , but countries have invested less in limited accountability in the short term, and strength- them than in more complex budget or treasury systems. ening institutions in the longer term (table O.3). With greater internet use in a country, the scope for

Table O.3 Priority policies for better service delivery Emerging countries: Laying the Transitioning countries: Building Transforming countries: foundation for more effective capable and accountable institutions Deepening collaborative institutions institutions • Improve information services to citizens • Strengthen government delivery systems • Improve collaboration across and beyond government • Strengthen monitoring of and payment • Strengthen provider management to providers • Enhance participatory policy • Get regular user feedback on service making • Establish population registers quality • Scale up nonstate provision of services • Increase transparency in priority areas • Increase electoral accountability OVERVIEW 35

Box O.12 Can continuous monitoring and small sanctions improve provider performance?

Traditional monitoring systems are expensive and complex. lower, violations become rarer. The idea could be extended New technologies lower these costs, allowing rewards or to public service monitoring. In Niger, a well-designed punishment to be more immediate and frequent. The idea monitoring system enabled by mobile phones motivated comes from criminal justice innovations. Usually lawbreak- teachers because they felt their far-away superiors cared ers face a low probability of being caught, but a large pun- about their work and looked out for them. ishment. When people face a high probability, but fi nes are

Sources: Romer 2013; Aker and Ksoll 2015.

digital engagement with citizens also increases. As long collection of identifi able information creates privacy as access is not universal, there is a risk of leaving those and security concerns. Automation changes work in unconnected behind. But citizen feedback systems ways that challenge existing social protections and have reduced problems such as petty corruption or poor reveal the inadequacy of existing labor laws. And scale services in the Dominican Republic, Nigeria, and Paki- economies create antitrust concerns. Digital safe- stan. As one Kenyan water utility manager said, “By guards that mitigate these risks become increasingly introducing an automated complaint management important as the digital transformation proceeds. system we took a noose and put it around our own necks. We are now accountable!” Developing privacy policies The fl ood of data collected on the internet brings many Deepen collaboration and participatory benefi ts to consumers and citizens but also raises the policy making risk of abuse through cybercrime, discrimination, Even in countries with advanced e-government or manipulation. As of 2014, some 107 countries had systems, their use remains surprisingly low. Many privacy laws, but only 51 of them were developing citizens prefer traditional ways of interacting with countries. The basic principles of privacy law are well the government such as phones or mail, so parallel established. They should give users more control systems remain in place and go unrealized. (and perhaps co-ownership) over their data and make Providing incentives such as faster tax refunds for e-fi ling or greater convenience through simplifi ed Figure O.24 Digital safeguards in the WDR’s and closely integrated services across agencies framework increases their use. Estonia’s X-Road framework inte- grates services from all parts of government as well as DIGITAL private or civil society groups according to protocols TECHNOLOGIES that govern data exchange and security standards.34 Practically any transaction—from paying the parking meter to voting in national elections—can be done from a smartphone. Tangible benefi ts for citizens will lead to universal use of e-government services, making such platforms also suitable for broad-based Information Informalization and Winner-take-all participatory policy making. deluge job displacement economy

Digital safeguards Strengthening analog complements will ensure a PRIVACY SOCIAL ANTITRUST high social and economic return from digital invest- POLICIES PROTECTION POLICIES ments. But a downside risk remains. Returning to the Report’s framework (fi gure O.24), large-scale Source: WDR 2016 team. 36 WORLD DEVELOPMENT REPORT 2016

opting out at the point of collection easier. Even when Governing the internet such laws exist, preventing abuse is diffi cult, espe- The internet emerged from U.S. government research cially where legal and enforcement capacity is weak. in the 1970s, but as it grew into a global network of networks, its governance structure has evolved. Adapting systems to Today, the internet is managed by an international changing labor markets coalition of governments, industry, technical experts, Better skills will help many workers cope with the and civil society—in what is referred to as the multi- effects of internet-enabled automation. But changes stakeholder model. U.S. users are now a small minor- in the labor market also require rethinking social ity of total internet users, given the dramatic growth protection and tax systems. The on-demand economy in the number of users in developing countries, espe- leads to more informal employment, transferring cially in Asia. Many countries are demanding more insurance and occupational obligations to freelance meaningful representation in the discussions of how workers. Strict labor regulations, common in devel- the internet should be governed. In addition, the lack oping countries, and overreliance on labor taxation of trust among nations following the Edward - encourage faster automation by making hiring more den episode, the surveillance by state agencies, and expensive. It would be better to strengthen work- the increasing confl icts between national policy and ers’ protection independently from work contracts regulations and global norms have raised questions by delinking from employment, about internet governance. offering independent social assistance, and helping Several countries have advocated for a mul- workers retrain and fi nd new employment quickly. tilateral model of governance, which would give In many countries this requires major reforms. And national governments greater power in overseeing countries just starting to develop social protection the internet, much the same way the United Nations, systems and deepening labor laws should design the International Telecommunication Union, or the them for the 21st-century workplace, rather than copy World Bank are currently governed. The promoters what industrialized countries created for a very dif- of “multistakeholderism” argue that state control of ferent world of work. the internet would not leave space for the range of players currently involved in internet governance Building antitrust enforcement capacity and could pave the for greater suppression of Regulatory reform that improves the business envi- privacy and restrictions on access to information ronment is the fi rst priority. But even in countries and on rights of free expression. The inability of the with comprehensive competition policies, including stakeholders of the internet to reach a consensus on easy market entry and exit, there will be cases in future governance mechanisms can be costly; some which mergers, collusion, or discriminatory pricing have even suggested that the internet is at risk of harm consumers by creating overly dominant enter- splitting up into several local or regional . prises or by keeping innovative competitors out. The broad-based, participatory approach involving Because the internet is still quite new and its impact all stakeholders is seen by many to be best suited to on markets constantly evolving, developing the ensure an effi cient and unrestricted global fl ow of capacity to investigate and prosecute complex vio- information essential for economic development. lations to competition law will take time. Cases pur- sued in high-income countries can provide guidance. Creating a global digital market The internet is encouraging more cross-border exchanges of goods and services, allowing consum- Global cooperation to solve ers and fi rms to bypass national borders. But cross- global problems border issues—such as barriers to data fl ows and unco- ordinated intellectual property rights regimes—are The internet is truly an international network. It can impairing the growth of internet fi rms and robbing be better managed with coordination across nations consumers of gains from increased digital trade. This and serve as a powerful platform to facilitate global has also meant that many startups from smaller coun- cooperation. Three priority areas are governing tries with relatively modest domestic markets, partic- the internet, creating a global digital market, and ularly in Europe (box O.13), are moving their business providing global public goods—including those to the United States as soon as they achieve a certain that promote and environmental scale. The small scale imposed by cross-border barriers sustainability. may also partly explain why e-commerce fi rms are OVERVIEW 37

Box O.13 European Union: A fragmented market for digital trade

Despite being a single market with free fl ows of goods, In May 2015, the European Commission (EC) announced services, and people for many decades, the European Union plans to create a Digital Single Market, in three main policy (EU) still functions like a fragmented market for digital trade. areas. First, the EC wants to increase the access for con- Consumers in the EU prefer to shop from online stores within sumers and fi rms to these digital goods and services by their national borders. While 44 percent of consumers made facilitating e-commerce, improving parcel delivery, and an online purchase from a domestic business in 2014, only 15 dealing with geoblocking, where access to online services percent did so from a business in another EU country. Firms or content is restricted to specifi c countries. Second, it will also face many diffi culties selling their goods and services examine the regulatory environment for telecoms, media, online into other EU markets. For example, Copenhagen, online platforms, and data protection. Third, it will encour- Denmark, and Malmo, Sweden, are separated only by an age more ICT investment and innovation through better 8-kilometer bridge, but a package sent from Copenhagen standards and interoperability, and more use of “big data” to Malmo costs €27, whereas the same package sent from and cloud computing. If EU reforms to create a common Malmo to Copenhagen costs €42. Firms facing large costs digital market are successful, they could become a model to adapt to various national laws believe that the costs out- for other world regions. weigh the benefi ts of selling online.

Source: European Commission (EC 2015).

often losing money in Africa while being profi table in problems—, , air pol- China and India. lution, epidemics—are features of globally intercon- Some countries are considering regulations that nected environmental, economic, and social systems make it legally binding for data of or about their and require global cooperation. What’s the role of citizens to reside within their national borders, also development agencies, nongovernmental organiza- referred to as data localization or data nationalism. tions (NGOs), and international organizations in a While such barriers may stem from legitimate world where their fi nancial heft is small? The data concerns about privacy and security for their citi- and technology revolutions arrive in time to bridge zens’ information, they can be costly. A study of six the gap between their resources and ambition by developing countries and the EU-28 found that such amplifying the impacts of action and including more regulations can reduce GDP by up to 1.7 percent, people in the formulation and execution of plans. But investments up to 4.2 percent, and exports by 1.7 for this to work, development actors must tackle pol- percent.35 Restrictions on data fl ows face the risk of icy constraints, internal and external. becoming a new tool for protectionism—disguised to Start with the how of development operations. impede trade and economic activity or to encourage With new technologies, development agencies can domestic data-driven sectors. At the same time, coun- be more inclusive by tapping on benefi ciary wisdom tries should make it easier for fi rms to protect their in designing interventions. They can raise effi ciency intellectual property (IP) rights—but within limits by using rapid feedback to refi ne and improve their that do not give excessive protections to large, well- actions through trial and error. But these approaches connected fi rms at the cost of stifl ing innovation and won’t come easily in organizations that emphasize creativity. The process to apply for IP licenses should spending and outputs over results, have burdensome be harmonized, streamlined, and globalized—so structures for accountability, and see any failures as fi rms need only to register their patent or trademark damning rather than informative. If traditional agen- in any signatory country to protect it across member cies can’t adapt, some of their business may be taken countries. up by disruptive newcomers. Next, the what. Development agencies can support Providing global public goods information services that help individuals and sys- Sustainable development and poverty reduction are tems managers make better decisions in ways that a focus of global partnerships. Many environmental promote poverty reduction. These services have fi xed 38 WORLD DEVELOPMENT REPORT 2016

set-up costs in software and data assembly, but can some of this. Diffi culties in measuring technolo- have near zero costs for distributing information. So gy’s role are another partial explanation for the gap the private sector will tend either to shy away from between individual cases of substantial benefi ts providing these services, or will price them at a level and modest macro effects. Technology impact is that shuts out poor people who could benefi t. One diffused throughout the economy, the world of work, and many aspects of personal life. And many area where the need for international cooperation benefi ts come in the form of higher quality or and support is particularly acute is the collection and convenience—nonmonetary benefi ts not refl ected distribution of data on weather, climate, and trans- in GDP numbers. boundary water fl ows, which are critically important 3. Even if rapid progress in artifi cial intelligence could to tackle climate change, improve natural resource solve some of these problems, it could take decades management, and support agriculture. (see spotlight 6 in the full Report). In the meantime, External agents and international organizations it would be unwise for policy makers to simply wait can help with targeted funding—for instance, fi lling and watch. the gap in African weather stations. They can support 4. Acemoglu and Robinson 2014. the complementary investments for information plat- 5. See Comin 2014. forms. And they can fi nd ways to encourage public and 6. See Graham and Foster 2014. private sectors, in both the developed and developing 7. While the internet reduces the cost of information, it does not necessarily reduce the effort it takes world, to open and share data for public goods. humans to process that information. In fact, infor- mation overload, in combination with behavioral biases, can promote behavior, amplify facts, or Reaping digital dividends even be abused for or manipulation. for everyone 8. Overcoming information problems also improves market effi ciency and could even lead to greater Digital technologies are transforming the worlds of innovation. For expositional simplicity, the Report’s business, work, and service delivery. These advances framework is simplifi ed and focuses on the most are making the leading parts of the economy and important development outcome associated with society more productive—even as many still wait for each mechanism that is enabled by the internet. the most basic benefi ts of the digital revolution. This 9. Moreover, cross-country regressions measuring the Report argues that to ensure that everyone will reap impact of digital technologies on growth could suf- fer from several other problems involving measure- the dividends of the internet, focusing on access to ment issues, endogeneity of variables, and small technology is essential but far from suffi cient. Why? sample size bias. Because technology needs to be complemented by 10. These results are based on Tan 2015; Osnago and Tan improvements in areas that determine whether fi rms, 2015. people, and governments can make effective use of 11. eBay 2013. new digital tools. The analog foundation cannot be 12. Baldwin 2011. strengthened overnight. It requires overcoming some 13. Brynjolfsson and McAfee 2014. of the most protracted development challenges: how 14. Moretti and Thulin 2013. to create an environment for fi rms to thrive, how 15. Goyal 2010; Aker and Mbiti 2010. to build effective education and training systems, 16. See Handel 2015; Best and others 2010; Jagun, Heeks, and how to make service providers more responsive and Whalley 2008; Aker 2011; Martin 2010. to citizens. The stakes are high, because the digital 17. Pineda, Aguero, and Espinoza 2011. 18. Asad 2014. revolution leaves behind countries that do not make 19. Aker and Mbiti 2010; Pineda, Aguero, and Espinoza the necessary reforms. For those that do, technology 2011. investments will produce ample digital dividends, 20. The survey was conducted by Research ICT for and these dividends will be widely shared among all Africa. stakeholders. 21. Aker, Collier, and Vicente 2013. 22. See box 3.5 in chapter 3 of the full Report. Notes 23. Dufl o, Hanna, and Ryan 2012. 24. Acemoglu, Hasan, and Tahoun 2014. 1. References to this and other data and citations in the 25. Bennet, Breunig, and Givens 2008. overview may be found in the full Report. 26. Hollenbach and Pierskalla 2014. 2. The lag between technology creation, adoption, 27. Goldin and Katz 2008. and learning to use it most effectively explains 28. Varian 2003. OVERVIEW 39

29. Data for 2014 of net digital worldwide ad revenue Bao, Beibei. 2013. “How Is Curb- shares from eMarketer, an online market research ing Innovation in China.” Atlantic, April 22. http:// company. www.theatlantic.com/china/archive/2013/04 30. Wood 2011. /how-internet-censorship-is-curbing-innovation 31. Eden and Gaggl 2014. -in-china/275188/. 32. WDR 2016 team estimates, based on household sur- Bauer, Matthias, Hosuk Lee-Makiyama, Erik Van der veys. See chapter 2 in the full Report for details. Marcel, and Bert Verschelde. 2014. “The Costs of Data 33. A landmark agreement that will further boost dig- Localization: Friendly Fire on Economic Recovery.” ital adoption around the world is the Information ECIPE Occasional Paper 3/2014, European Centre for Technology Agreement, concluded by World Trade International Political Economy, Brussels. Organization (WTO) members on July 24, 2015. It Bennet, W. L., C. Breunig, and T. Givens. 2008. “Commu- will eliminate tariffs on more than 200 ICT prod- nication and Political Mobilization: Digital Media ucts, valued at $1.3 trillion in global trade. and the Organization of Anti-Iraq War Demonstra- 34. Vassil 2015. tions.” Political Communication 25 (3): 269–89. 35. Bauer and others 2014. Berdou, E., and C. A. Lopes. 2015. “The Case of UNICEF’s U-Report (Uganda).” World Bank, Washington, DC. Best, Michael L., Thomas N. Smyth, John Etherton, and References Edem Wornyo. 2010. “Uses of Mobile Phones in Acemoglu, Daron, Tarek Hasan, and Ahmed Tahoun. Post-Confl ict Liberia.” Information Technologies and 2014. “The Power of the Street: Evidence from Egypt’s International Development 6 (2): 91–108. Arab Spring.” NBER Working Paper 20665, National Beuermann, Diether, Christopher McKelvey, and Renos Bureau of Economic Research, Cambridge, MA. Vakis. 2012. “Mobile Phones and Economic Develop- Acemoglu, Daron, and James A. Robinson. 2014. “The ment in Rural Peru.” Journal of 48 Rise and Decline of General Laws of Capitalism.” (11): 1617–28. NBER Working Paper 20766, National Bureau of Eco- Bishop, Sylvia, and Anke Hoeffl er. 2014. “Free and nomic Research, Cambridge, MA. Fair Elections: A New Database.” Working Paper Aker, Jenny C. 2011. “Dial A for Agriculture: A Review of WPS/2014-14, Centre for the Study of African Econo- Information and Communication Technologies for mies, Oxford, U.K. http://www.csae.ox.ac.uk/datasets Agricultural Extension in Developing Countries.” /free-fair-elections/. 42 (6): 631–47. Bruns, Barbara, and Javier Luque. 2014. Great Teachers: Aker, Jenny C., , and Pedro Vicente. 2013. “Is How to Raise Student Learning in Latin America and the Information Power? Using Mobile Phones and Free Caribbean. Washington, DC: World Bank. Newspapers during an Election in Mozambique.” Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second Working Paper 328, Center for Global Development, Machine Age: Work, Progress, and Prosperity in a Time of Washington, DC. http://papers.ssrn.com/sol3/papers Brilliant Technologies. New York: W. W. Norton. .cfm?abstract_id=2364162. Castro, Daniel. 2013. “The False Promise of Data Nation- Aker, Jenny C., and Christopher Ksoll. 2015. “Call Me alism.” Information Technology and Innovation Educated: Evidence from a Mobile Monitoring Exper- Foundation. http://www2.itif.org/2013-false-promise iment in Niger.” Working Paper 406, Center for Global -data-nationalism.pdf. Development, Washington, DC. http://www.cgdev China, National Bureau of Statistics of China. Various .org/publication/call-me-educated-evidence-mobile years. Population Census. Beijing. http://www.stats -monitoring-experiment-niger-working-paper-406. .gov.cn/english/Statisticaldata/CensusData/. Aker, Jenny C., and Isaac M. Mbiti. 2010. “Mobile Phones Chomitz, Kenneth. 2015. “Information as Interven- and Economic Development in Africa.” Journal of Eco- tion: A Visit to Digital Green.” Let’s Talk Development nomic Perspectives 24 (3): 207–32. (blog). http://blogs.worldbank.org/developmenttalk Asad, Saher. 2014. “The Crop Connection: Impact of Cell /information-intervention-visit-digital-green. Phone Access on Crop Choice in Rural Pakistan.” Job CIGI (Centre for International Governance Innova- market paper, George Washington University, Wash- tion), and Ipsos. 2014. “Global Survey on Internet ington, DC. Security and Trust.” https://www.cigionline.org Autor, David. 2014. “Polanyi’s Paradox and the Shape of /internet-survey. Employment Growth.” Draft prepared for the Federal Clinton, Bill. 2000. Speech presented at the Paul H. Nitze Reserve Bank of Kansas City. School for Advanced International Studies (SAIS) at Baldwin, Richard. 2011. “Trade and Industrialization after Johns Hopkins University, Washington, DC, March 8. Globalization’s Second Unbundling: How Building Comin, Deigo. 2014. “The Evolution of Technology and Joining a Supply Chain Are Different and Why Diffusion and the .” http://www It Matters.” NBER Working Paper 17716, National .dartmouth.edu/~dcomin/files/brookings%20 Bureau of Economic Research, Cambridge, MA. blum%20roundtable.pdf. 40 WORLD DEVELOPMENT REPORT 2016

Comin, Diego, and Bart Hobjin. 2010. “Technology Dif- Hollenbach, Florian, and Jan Pierskalla. 2014. “Voicing fusion and Postwar Growth.” NBER Working Paper Discontent: Communication Technology and Pro- 16378, National Bureau of Economic Research, Cam- test.” APSA Annual Meeting paper. http://papers.ssrn bridge, MA. .com/sol3/papers.cfm?abstract_id=2452306. Conference Board. Various years. “Total Economy Data- HRW (Human Rights Watch). 2015. “Human Rights base.” https://www.conference-board.org/data Watch Submission.” Background note submitted /economydatabase/. to the World Development Report 2016 team, Human Dufl o, , Rema Hanna, and Stephen P. Ryan. 2012. Rights Watch, Washington, DC. https://www.hrw “Incentives Work: Getting Teachers to Come to .org/sites/default/files/supporting_resources/hrw School.” 102 (4): 1241–78. _submission_re_wdr_2016_internet_for_develop eBay. 2013. “Commerce 3.0 for Development: The Prom- ment.pdf. ise of the Global Empowerment Network. An eBay ILO (International Labour Organization). Various years. Report Based on an Empirical Study Conducted Key Indicators of the Labor Market (KILM database). by Sidley Austin LLP.” eBay Inc., Washington, DC. International Labour Organization, Geneva, http:// https://www.ebaymainstreet.com/sites/default/fi les www.ilo.org/empelm/what/WCMS_114240/lang--en /eBay_Commerce-3-for-Development.pdf. /index.htm. EC (European Commission). Various years. Eurostat ILO (International Labour Organization) Laborsta (database). European Commission, Brussels, http:// (database). Various years. ILO, Geneva, http:// ec.europa.eu/eurostat/data/database. laborsta.ilo.org/. ————. 2015. “Strategy Paper on a Digital Single Market.” ITU (International Telecommunication Union). 2015. European Commission, Brussels. http://ec.europa.eu Facts and Figures. Geneva: ITU. http://www.itu.int /priorities/digital-single-market/. /en/ITU-D/Statistics/Documents/facts/ICTFacts Economist. 2014. “Defending the Digital Frontier: A Figures2015.pdf. Special Report on Cybersecurity.” July. http://www Jagun, Abi, Richard Heeks, and Jason Whalley. 2008. “The .economist.com/news/special-report/21606416 Impact of Mobile Telephony on -companies-markets-and-countries-are-increasingly Micro-Enterprise: A Nigerian Case Study.” Information -under-attack-cyber-criminals. Technologies and International Development 4 (4): 47–65. Eden, Maya, and Paul Gaggl. 2014. “On the Welfare Impli- Karabarbounis, Loukas, and Brent Neiman. 2013. “The cations of Automation.” August 20, 2015. https://belk- Global Decline of the Labor Share.” NBER Working collegeofbusiness.uncc.edu/pgaggl/wp-content Paper 19136, National Bureau of Economic Research, /uploads/sites/36/2014/09/Income_Shares_08-20-2015 Cambridge, MA. _WP.pdf. King, Gary, Jennifer Pan, and Margaret E. Roberts. 2013. ————. 2015. “The ICT Revolution: A Global Perspective.” “How Censorship in China Allows Government Crit- Background paper for the World Development Report icism but Silences Collective Expression.” American 2016, World Bank, Washington, DC. Political Science Review 107 (2): 1–18. Elmer-Dewitt, Philip. 1993. “First Nation in Cyberspace.” Kosinski, Michal, David Stillwell, and Thore Graepel. Time, December 6. 2013. “Private Traits and Attributes Are Predictable Frey, Carl, and Michael Osborne. 2013. “The Future of from Digital Records of Human Behavior.” Proceedings Employment: How Susceptible Are Jobs to Computer- of the National Academy of 110 (15): 5802–05. isation?” Working paper, Oxford University, Oxford. Lakner, Christoph, and Branko Milanovic. 2013. “Global U.K. Income Distribution: From the Fall of the Wall Goldin, Claudia, and Lawrence F. Katz. 2008. The Race to the Great .” Policy Research Working between Education and Technology. Cambridge, MA: Paper 6719, World Bank, Washington, DC. http:// Press. elibrary.worldbank.org/doi/pdf/10.1596/1813-9450-6719. Goyal, Aparajita. 2010. “Information, Direct Access to Martin, Brandie. 2010. Mobile Phones and Rural Livelihoods: Farmers, and Rural Market Performance in Central An Exploration of Mobile Phone Diff usion, Uses, and Per- India.” American Economic Journal: ceived Impacts of Uses among Small- to Medium-Size 2 (3): 22–45. Holders in Kamuli District, Uganda. Ames: Iowa State Graham, Mark, and Christopher Foster. 2014. “Geog- University. raphies of Information Inequality in Sub-Saharan Meeker, Mary. 2015. “Internet Trends 2015: Code Confer- Africa.” Oxford Internet Institute, University of ence.” http://www.kpcb.com/internet-trends. Oxford, U.K. http://cii.oii.ox.ac.uk/geographies-of Moretti, Enrico, and Per Thulin. 2013. “Local Multipliers -information-inequality-in-sub-saharan-africa/. and Human Capital in the United States and Sweden.” Handel, Michael. 2015. “The Effects of Information and Industrial and Corporate Change 22 (1): 339–62. Communication Technology on Employment, Skills, Nguyen, Huy, and Marc Schiffbauer. 2015. “Internet, and Earnings in Developing Countries.” Background Reorganization, and Firm Productivity in Vietnam.” paper for the World Development Report 2016, World Background paper for the World Development Report Bank, Washington, DC. 2016, World Bank, Washington, DC. OVERVIEW 41

OECD (Organisation for Economic Co-operation and Saleh, Nivien. 2012. “Egypt’s Digital and the Dic- Development). 2014. “ICT Value Added (Indicator).” tator’s Dilemma: An Evaluation.” Telecommunications doi: http://dx.doi.org/10.1787/4bc7753c-en. Policy 36 (6): 476–83. Open Net Initiative. 2013. https://opennet.net/. Spada, Paolo, Jonathan Mellon, Tiago Peixoto, and Fre- Osnago, Alberto, and Shawn W. Tan. 2015. “The Effects drik Sjoberg. 2015. “Effects of the Internet on Par- of the Internet on Trade Flows and Patterns.” Back- ticipation: Study of a Policy Referendum in Brazil.” ground paper for the World Development Report 2016, Policy Research Working Paper 7204, World Bank, World Bank, Washington, DC. Washington, DC. Peixoto, Tiago, and Jonathan Fox. 2015. “When Does ICT- Tan, Shawn W. 2015. “The Effects of the Internet on Firm Enabled Citizen Voice Lead to Government Respon- Export Behavior.” Background paper for the World siveness?” Background paper for the World Develop- Development Report 2016, World Bank, Washington, ment Report 2016, World Bank, Washington, DC. DC. Peppet, Scott R. 2014. “Regulating the Internet of Things: UN (United Nations). 2014. “E-Government Survey 2014: First Steps toward Managing Discrimination, Pri- E-Government for the Future We Want.” United vacy, Security, and .” Texas Law Review 93 (85): Nations, New York. 87–176. UN (United Nations) Population Division. 2014. Popula- Pierre, Gael, Maria Laura Sanchez Puerta, and Alexandria tion and Development Database, http://www Valerio. 2014. “STEP Skills Measurement Surveys: .un.org/en/development/desa/population Innovative Tools for Assessing Skills.” Working Paper /publications/development/population-development 89729, World Bank, Washington, DC. -database-2014.shtml. Pineda, Allan, Marco Aguero, and Sandra Espinoza. 2011. Varian, Hal. 2003. “Economics of Information Technol- “The Impact of ICT on Vegetable Farmers in Hon- ogy.” Revised version of the Raffaele Mattioli Lecture, duras.” Working Paper 243, Inter-American Develop- delivered at the Sorbonne on March 6, 2003. http:// ment Bank, Washington, DC. people.ischool.berkeley.edu/~hal/Papers/mattioli Plaza, Sonia, Seyed Reza Yousefi , and Dilip Ratha. 2015. /mattioli.pdf. “Technological Innovations and Costs.” Vassil, Kristjan. 2015. “Estonian E-Government Ecosys- Background paper for the World Development Report tem: Foundation, Applications, Outcomes.” Back- 2016, World Bank, Washington, DC. ground paper for the World Development Report 2016, Polity IV. 2015. http://www.systemicpeace.org/polity World Bank, Washington, DC. project.html. Wood, David. 2011. “EU Competition Law and the Inter- Posner, Richard A. 1981. “The Economics of Privacy.” net: Present and Past Cases.” Competition Law Interna- American Economic Review 71 (2): 405–09. tional (April): 44–49. Pritchett, Lant, Michael Woolcock, and S. Samji. Forth- World Bank. Various years. I2D2 (International Income coming. What Kind of Organization Capability Is Needed? Distribution database). World Bank, Washington, HKS faculty research working paper, Harvard Ken- DC, http://econ.worldbank.org/WBSITE/EXTERNAL nedy School, Cambridge, MA. /EXTDEC/EXTRESEARCH/0,,contentMDK:20713100 Raja, Deepti Samant. 2015. “Bridging the Disability ~pagePK:64214825~piK:64214943~theSitePK: Divide through Digital Technologies.” Background 469382,00.html. paper for the World Development Report 2016, World World Bank. Various years. World Development Bank, Washington, DC. Indicators (database). World Bank, Washington, Research ICT Africa. Various years. Household survey. DC, http://data.worldbank.org/data-catalog/world Capetown, South Africa. http://www.researchict -development-indicators. africa.net/home.php. ————. 2015. “MajiVoice: A New Accountability Tool Romer, Paul M. 2010. “What Parts of Globalization Mat- to Improve Public Services.” Water and Sanitation ter for Catch-Up Growth?” American Economic Review Policy Note, World Bank, Washington, DC. https://wsp Papers and Proceedings 100 (2): 94–98. .org/sites/wsp.org/files/publications/WSP-Maji ————. 2013. “Small Stakes, Good Measurement.” Urban- Voice-New-Accountability-Tool-to-Improve-Public ization Project (blog), July 2013. http://urbanization- -Services.pdf. project.org/blog/small-stakes-good-measurement/. 42

SPOTLIGHT 1 WORLD DEVELOPMENT REPORT 2016 somewhat interchangeably. “Internet” emphasizes the somewhat interchangeably. “Internet”emphasizes the “information and communicationtechnologies(ICTs)” terms “digitaltechnologies,”“internet,” andsometimes sharing ofdataandinformation. TheReportusesthe greatly facilitatethecreation, storage, analysis,and impacts ofdigitaltechnologies andservicesthat about specifi c technologies,itgenerallycoversthe Report. working defi nitionofthe technologiescoveredinthis tion costsaffectseconomicdevelopment.Butfi spotlight describeshowthedeclineofthesetransac- among businesses,people,andgovernments.This implications formarketandnonmarketexchanges have vastlyreducedmanyofthesecosts,withmajor years later,theinternetandotherdigitaltechnologies the costsofacquiringandsharinginformation.Many firm toproduceitin-house. than theprofit fromthatexchange,itisrationalfora intermediate goodorserviceinthemarketislarger them. Aslongasthecostofmakinganexchange or suppliers,andnegotiatingcontractsenforcing additional costs,suchastheeffortoffi using thepricemechanismincurredanumberof control environment.WhatCoaserealizedwasthat tend tooperateinaself-containedcommand-and- way toorganizeeconomicactivity,largecompanies economics considersthemarketmosteffi the Firm,”whichaskedwhyfi Prize inEconomics1991—published“TheNatureof In 1937,RonaldCoase—whowouldreceivetheNobel old economicsexplainstheneweconomyquitewell. understand whattheyactuallydo.Itturnsoutthat technologies affectdevelopment,itisimportantto To understandhowtheinternetandrelateddigital promotes development How theinternet ENABLING DIGITALDEVELOPMENT is not While theWorldDevelopmentReport2016 isnot Most oftheseCoasiantransactioncostsstemfrom rms exist. nding buyers 1

Even though cient rst, a distributed andaccessedeffortlesslyfromanywhere. these devicesarelinkedsothatinformationcanbe a massiveimpactonalmostallaspectsoflifeisthat the reasonthatallofthesetechnologieshavehadsuch and cheaperstorageareusefulintheirownright.But central importanceofconnectivity.Fastercomputers customers, and even forindividualswhocannot use made itmuch easier forsmallproducersto fi high. Theemergenceofe-commerce platformshas cases, thetransactioncostsare essentiallyinfi to confi way tofind eachotheroracquireenoughinformation not takeplacebecausethetwo parties simplyhadno instances, amutuallybenefi cial transactionmight can helpovercomeinformationproblems.Insome major, interrelatedways.Oneisthattheinternet internet affectseconomicdevelopmentinthree consequence, productioncosts. in turnhasloweredtransactioncosts—andoftenasa lower costofacquiringandusinginformation,which 1.7 trillionand73trillion. of acomputationhasdroppedbyfactorbetween manual computinginthemid-19thcentury,cost Nordhaus, in2007,estimatedthatsincetheeraof accelerating declineinthecostofcomputing.William 30 percentperyear.Thiscontinuesalong-termand that drivetheinternet—insomecasesbymorethan and increasedthespeedofalldigitaltechnologies Technology developmenthasvastlyreducedthecost innovation effi inclusion, promotes internet The By loweringthecostofthesetransactions, dentlyproceedwiththetransaction; insuch ciency, and 2 Theresulthasbeenafar nitely nd the world. products thathasattractedcustomersfromallover are illiterate,havesetupAnou,awebshopfortheir shows. RuralartisansinMorocco,someofwhom traditional marketingtoolslikeadvertisingortrade is amajordeterrenttolendingbybanks. cost ofgatheringinformationaboutpoorborrowers example ofextendingcredittopoorfarmers.Thehigh has muchmoreinformationthantheother.Take do nottakeplacewhenonepartytoatransaction ment thatnowservesmorecitizens. side ofthetransaction,suchasafi means expansionofamarketbythoseontheother delivery systems.Inclusionfortheindividualusually kets, insocialinteraction,orgovernmentservice is thatitfostersinclusion—innewandexistingmar- has manybenefits, butthemostimportant,arguably, and informationcosts,createsthesemarkets.This information aboutthecar’squalitythanbuyer. buying ausedcar,thesellerusuallyhasmuchmore for Lemons,”publishedin1970,bythefactthatwhen vated towritehismostfamouspaper,“TheMarket received theNobelPrizein2001.Akerlofwasmoti- lof, MichaelSpence,andJosephStiglitz,whojointly and asymmetriesbyeconomistssuchasGeorgeAker- products. own mobilephonesbuthavenoaccesstofi and creditworthinessoflow-incomehouseholdsthat the unbanked,byassessingsavingspotential The goalistopromotefi nancial inclusionamong nesses that,while individuallyoftennotspectacular, brought numerous effi of existingfactorsproduction. Theinternethas words, lowertransactioncosts raise theproductivity them faster,cheaper,ormore convenient. Inother before thedigitalrevolution,but theinternethasmade costs. Mosttransactionswere alreadytakingplace Coasian conceptsoftransaction andcoordination service records. many ownersithashad,andwhethercomplete including whetherithasbeeninanaccident,how States letbuyersresearchthehistoryofacaronline, Today, internetsitessuchasCarfaxintheUnited relate savingsbehaviorwithmobilephonerecords. World SavingsandRetailBankingInstitutetocor- phone records.InGhana,Cignifi ness ofapotentialborrowerbasedontheirmobile have developedmethodstojudgethecreditworthi- possess mobilephones.CompaniessuchasCignifi charge exorbitantinterestrates.Butmanyofthepoor therefore needtorelyoninformalmoneylenderswho Even whensearchcostsarelow,transactionsoften There isalargeliteratureoninformationproblems The secondmechanismisclosesttotheoriginal 3 Theinternet,byvastlyloweringsearch ciencyimprovements tobusi- workedwiththe rm oragovern- 4 Thepoor nancial

6 5

gives risetovarioustypesofscaleeconomies. scope forcustomizationofservices. problems) thishasspawned,andtheunprecedented possible, thenewbusinessmodels(andcompetition of thescaleeconomiesthatmakethisinnovation have importantimplicationsrelatedtothenature online newsand data. Sweden), e-books(Amazoninthe UnitedStates),or is alsopurelydigital,suchasdigital music(Spotifyin associated withthe“neweconomy.” innovation—the thirdmechanism—thatistypically now befullyautomated.Thishasledtoenormous what previouslyinvolvedroutinehumanlaborcan transaction costessentiallydropstozerobecause transaction incursverylittleextracost.Themarginal platform isinplace,eachadditionalcustomer,user,or fi up to7millionworkdaysoverall. citizen savesaboutfi veworkingdaysperyear,adding million in2003to340by2014.Asaresult,each queries madethroughX-Roadincreasedfromhalfa sector agenciestothecitizensonline.Thenumberof 3,000 servicesfrom900governmentandprivate X-Road isane-governmentsystemthatoffersnearly cles areoftenheldupbyoncomingtraffi technology thatminimizesleftturns,wherevehi- 1 milliongallonsofgaseachyearbyusingrouting The deliverycompanyUPSfamouslysavesabout ity utilizationforlogisticsandtransportcompanies. navigation, andschedulingsoftwareimprovecapac- inventory managementtotheirsuppliers.Tracking, vendors globallyinrealtime,essentiallyshifting planning. Retailersnowsharepoint-of-saledatawith supply chainmanagementandenterpriseresource munication andinformationprocessingimproves add uptoenormousaggregatebenefi 2003. needed justsixwarehousesto reach US$3billionin stores beforereachingUS$1billion insales;Amazon with just500engineers. few resources.Facebookreachedhalfabillionusers scale upextremelyquickly,evenwithrelatively initially, entrycostsarelowandsuchwebsitescan stores—are alsodominatedbyafewfi web searches,mobilepayments,oronlinebook- regulated. Manyinternet-basedmarkets—suchas entry costsarealsohigh,suchsectorstendtobe operateinsimilarenvironments.Because emergence ofnaturalmonopolies.Waterorelectric an increasingnumberoftransactions,favorthe Supply-side scaleeconomies,wherecostsdropwith brokerage ormatchmaking servicesfortravel, jobs, xed up-frontcostscanbehigh,butoncetheonline The coststructureofmanyinternetbusinesses For manyinternet-basedbusinessesorservices, HOW THEINTERNETPROMOTES DEVELOPMENT 9 Formanyofthesefirms, theproducttheysell 10 Otherssellhighly automated 8 Walmarthadtobuild276 These dynamics ts. Bettercom- rms. Atleast c. Estonia’s 7

SPOTLIGHT 1 43 44

SPOTLIGHT 1 WORLD DEVELOPMENT REPORT 2016 explosion ofvideocontent. traffi 30 companiesaccountforoverhalfofpeakinternet distributed acrossthousandsofcompanies,todayjust industries. Tirole showsthattwo-sidedmarketsexistinmany classic 2003paperbyJean-CharlesRochetandJean makes theothersideofmarketmorelucrative.A and increasetheuserbase(atverylittlecost),which both, itmakessensetoprovidetheserviceforfree who wantstoreachtheuser.Ratherthancharging typically theuserofserviceandanadvertiser The platformownerhastwodifferentcustomers, operating platformmarketsortwo-sidedmarkets. new businessmodels.Manyofthesearewebservices users tomaintainthesamelevelofutility. collective actionbyalargenumberofinterconnected very littleactualcostontheuser,butwouldrequire switch toadifferentsocialmediaplatformimposes benefi utility riseswithscale.Thesenetworkexternalities demand-side scaleeconomies,averagerevenueor scale economies,averagecostdropswithscale.With like M-PesainKenyaareexamples.Withsupply-side attracts. Socialmediasitesordigitalpaymentsystems able itbecomestoitsusersandthemorenew many services,themorepeopleuseit,valu- 10 years.Whileintheearly2000 high degreeofconcentrationhasoccurredinthelast countries. Buteventhere,inmanyofthesemarkets,a models havebeenreplicatedbyfi merchandise, orridesharing.Manyofthesebusiness platform owners,users,andadvertisers(fi led toaparticularlyeffectivegrandbargainbetween nano-economics—which studiesindividual,computer- have createda whole newbranchofeconomics— sive datavolumescollectedby internetplatforms sometimes anonymously, not.Themas- Most onlinebehaviorisautomatically monitored— is tremendousscopeforcustomization ofservices. their totalsales. publishers dependonAmazonforacrucialshareof is asingleorarejustfewbuyers).Forinstance,book forms couldexertmonopsonypower(becausethere of alldigitaladvertisingrevenue.And,dominantplat- Facebook, Baidu,andAlibaba—nowaccountforhalf advertising space.Justfourcompanies—Google, over users.Buttheycoulddosovendorsbuying a service,theydonotactuallyexertmonopolypower tion policy.Becauseplatformsoftendonotchargefor This modelraisesdifficult questionsaboutcompeti- Ultralow marginaltransactioncostshavepowered Scale economiesalsoexistonthedemandside.For Because most processes can be automated, there Because mostprocessescanbe automated,there c intheUnitedStates—muchofthisisdueto t users,butcanalsocreatelock-ineffects;to 13 Buttheeconomicsofinternethave 11 s internettraffi rms indeveloping 12 gure S1.1). c was access. ing history,andeventhetypeofdevicetheyusefor vary pricesbasedonusers’estimatedlocation,brows- pay andofferdifferentpricestousers. can analyzeuserbehaviortodeterminewillingness even pricediscrimination,whenautomatedsystems For theseller,itallowsmoretargetedadvertisingand preferences—although atthecostofgivingupprivacy. is thatservicescanbetailoredtoindividualneedsand mediated transactions. Users three mechanisms. in boxS1.1presents agraphicrepresentation of the now moreconvenient,faster,or cheaper.FigureS1.1.1 was typicallyavailablethrough anotherchannelis tion experiencesincreasedeffi action. Thecustomerattheother endofthetransac- inclusion inanotherwiseinaccessible markettrans- remote location.Forthem,itwill oftenbeacaseof a ridesharingplatformorfreelanceworkerin service, suchasaninformaldriverworkingthrough side ofthetransactionofteninvolvesaprovider fully automatedserviceisthemaininnovation,one ically involveallthree.Whiletheplatformrunninga For example,transactionsoninternetplatformstyp- than oneofthethreemechanismsmaybeatwork. better servicedelivery. but thepublicsectorhasbeenslowtoadaptthemfor or service.Businessesusesuchtoolsextensively, allow thecustomertoassessqualityofaproduct that helptheproviderimproveproduct,butalso commercial websitesprovidefeedbackmechanisms personal data for useful services useful for data personal S1.1 Figure Source: Free services (search, e-mail, cloud) There isampleevidencethate-commercesites Finally, inmany,ifnotmost,transactions,more Based on Kurbalija 2015. Kurbalija on Based 15

Informationcanflow bothways,sincemany Data (personal preferences) Google, Facebook,andsoon Internet users trade trade users Internet Goods andservices Platforms 14

Themainbenefi t totheuser Money

ciency. Aservicethat

ey Mon

Advertising space Advertising u c and User

Vendors

mer pr mer sto

of

i le . Shapiro and Varian1999;andFarrell 2004. 7. 6.Akerlof 1970. 5. Cignifi BanerjeeandDufl 4. 3. http://www.theanou.com. 2. Nordhaus 2007. 1. Coase 1937. Notes c. Varian and Farrell 2004. Farrell and Varian c. 2013. Clark and Gelb b. 2009. Tandon and Morrell, Melhem, a. Source: digital IDsto verify theireligibility. receive publicservices becausegovernments canuse work. in work outsidethehome, butcannow engageintele- with disabilitieshave sometimesbeenunableto engage possible promotes inclusionaswell asmarket expansion. transactions didnottake place. Makingthesetransactions pre-internet era, thatthere was essentiallynomarket—the some transactions for whichthecost was sohighinthe drop, andthree thingscanhappen.Ontheleft,there were duction oftheinternet. Withthe right. Theuppercurve shows thesecosts before theintro- impose, from most costly ontheleftto least costly onthe tions inaneconomy arranged by thetransaction costs they the eff Figure S1.1.1 provides asimplegraphic representation of Box S1.1 represent thelion’s share ofbenefits from theinternet. ally unspectaculareffi ciency gains may, intheaggregate, benefits extend to governments, aswell. Theseindividu- fulfi llment oremployee performance. Many ofthesame through job-matching services, andmonitoring contract negotiating withbuyers orsuppliers,fi nding workers has reduced costs for businesses whenconnecting and now bedonewithaclickortap. Similarly, theinternet erally usedto require atripto ashoporoffi a job, paying taxes, orrenewing adriver’s license gen- executing banktransactions, searching for ahomeor effi even before theadvent oftheinternet. Thisraises the cost ofexisting transactions—that is,thoseoccurring information problems, contributes to greater inclusion. These are allexamples where theinternet, by overcoming in globalmarkets, where they canearnhigherreturns. and smallfirms inpoorcountries cantrade theirservices ciency ofavast range ofactivities.Purchasing goods, For example, women withsmallchildren orpersons In themiddleoffi gure, theinternet lowers the a ect offalling transaction costs. Imaginealltransac- WDR 2016 team. 2016 WDR Many poorordisadvantaged populationswillnow andWSBI2014. Three waysThree promotes internet inwhichthe development o 2011. internet, many suchcosts b Andskilledworkers ce, but can but ce, 11. Congressional testimonyofCraigLabowitz, chief Shapiro andVarian1999. 10. Ellisonand2005. 9. Levin 8. 2011. promotes development internet ofhow the representation Figure SB1.1.1 Source: economy” firms are inthisspace. the more usefulitbecomes. by network eff to various types ofscaleeconomies, often reinforced cost isalsocloseto zero. Thiscost structure gives rise products, suchase-books,themarginal production labor cannow befullyautomated. For purely digital zero becausewhatpreviously involved routine human In many cases,themarginal transaction cost isessentially customer, user, ortransaction incursvery littleextra cost. but once theonlineplatform isinplace, eachadditional businesses orservices, fi xed up-front costs canbehigh, right-hand sideofthefi gure. For many internet-based

The most dramatic impactoftheinternet isonthe Transaction costs _missing/. in http://www.wired.com/2014/06/net_neutrality scientist ofthesoftwarecompany, Arbor,quoted HOW THEINTERNETPROMOTES DEVELOPMENT Transactions, arrangedfromhightolowcost WDR 2016 team. 2016 WDR Inclusion internet Post- ects,where themore usersasystem has, Pre-internet A graphic A graphic Efficiency c Most oftheso-called“new Innovation

SPOTLIGHT 1 45 46 WORLD DEVELOPMENT REPORT 2016

12. Switching costs are far higher when replacing a Kurbalija, Jovan. 2015. An Introduction to Internet Gover- widely used software package, since it will require a nance. 6th ed. DiploFoundation. lot of retraining and associated investments, such as Levin, Jonathan D. 2011. “The Economics of Internet Mar- enterprise resource planning or operating systems. kets.” NBER Working Paper 16852, National Bureau 13. Rochet and Tirole 2003. of Economic Research, Cambridge, MA. 14. Varian 2014; Varian and Farrell 2004. Melhem, Samia, Claudia Morrell, and Nidhi Tandon. 15. Pasquale 2015. 2009. “Information and Communication Technolo- gies for Women’s Socioeconomic Empowerment.” Working Paper 176, World Bank, Washington, DC. References Nordhaus, William D. 2007. “Two Centuries of Produc- Akerlof, George A. 1970. “The Market for ‘Lemons’: Qual- tivity Growth in Computing.” Journal of Economic His- ity Uncertainty and the Market Mechanism.” Quar- tory 67 (01): 128–59. terly Journal of Economics 84 (3): 488–500. Pasquale, Frank. 2015. The Black Box Society. Cambridge, Banerjee, Abhijit, and Esther Dufl o. 2011. Poor Economics: MA: Harvard University Press. A Radical Rethinking of the Way to Fight Global Poverty. Rochet, Jean-Charles, and Jean Tirole. 2003. “Platform New York: PublicAffairs. Competition in Two-Sided Markets.” Journal of the Cignifi and WSBI (World Savings and Retail Banking European Economic Association 1 (4): 990–1029. Institute). 2014. Mobile Phone Data as the Key to Promot- Shapiro, Carl, and Hal R. Varian. 1999. Information Rules: A ing Financial Inclusion. Cambridge, MA: Cignifi and Strategic Guide to the Network Economy. Cambridge, MA: WSBI. Press. Coase, Ronald H. 1937. “.” Econom- Varian, Hal. 2014. “Big Data: New Tricks for Economet- ica 4 (16): 386–405. rics.” Journal of Economic Perspectives 28 (2): 3–28. Ellison, Glenn, and Sara Fisher Ellison. 2005. “Lessons Varian, Hal, and Joseph Farrell. 2004. The Economics of about Markets from the Internet.” Journal of Economic Information Technology: An Introduction. Cambridge, Perspectives 19 (2): 139–58. MA: Cambridge University Press. Gelb, Alan, and Julia Clark. 2013. “Identifi cation for Devel- opment: The Biometrics Revolution.” Working Paper 315, Center for Global Development, Washington, DC. SPOTLIGHT 1 HOW THE INTERNET PROMOTES DEVELOPMENT 47 1. Accelerating growth

2. Expanding opportunities

3. Delivering services D CHAPTER 1 Accelerating growth

Digital technology creates opportunities to accelerate inclusion, effi ciency, and innovation. It promotes growth, but these are often missed because fi rms the inclusion of fi rms in the world economy by in sectors where technology’s impact is greatest are enabling more fi rms to trade new products to new frequently protected from innovative competitors. destinations. For instance, fi rms selling their goods Firms that face more competition use digital technol- online through Alibaba, China’s leading e-commerce ogy more intensively and effectively—it enables them company, are smaller and younger and export more to reduce their costs to outperform their competitors. products to different destinations than fi rms selling But fi rms in developing countries do not necessar- offl ine. It raises effi ciency by allowing fi rms to make ily have the incentive to adopt new technologies to better use of their capital and labor. For instance, real- increase their cost effectiveness because they are time data help equipment manufacturers in China often protected from domestic or foreign compe- turn over their inventory stocks fi ve times faster than tition. And it is precisely in protected sectors such suppliers not connected to the internet. It enhances as retail and wholesale trade, fi nance, transport, or innovation by enabling fi rms to exploit scale effects public utilities where digital technology can increase through online platforms and services that compete productivity the most. Harnessing those opportuni- with conventional business models in retail, trans- ties thus requires policies that lower the barriers to port, lodging, and banking, to name a few. These three competition and market entry, in addition to invest- mechanisms thus boost growth by expanding trade, ments in infrastructure and skills. Only then will increasing capital and labor utilization, and intensify- fi rms use new digital technologies more intensively ing competition (fi gure 1.1).1 and effectively—and only then will countries avoid But the benefi ts are neither automatic nor assured. falling behind. Despite great opportunities, fi rms’ use of digital Firms across the world are becoming more con- technologies differs substantially across countries nected. For instance, the share of fi rms with at least due to variations in skills and infrastructure and in fi ve employees using broadband internet in lower- barriers to competition and market entry. Competi- middle-income countries rose from 39 percent to 68 tion from China induced fi rms in member-countries percent from 2006–09 to 2010–14. And the growth of the Organisation for Economic Co-operation and rates and valuations of internet fi rms across the Development (OECD) to adopt new technologies to world are surging. Less visibly, but more importantly, escape the competition from low-cost producers, digital technologies have transformed traditional pro- accounting for 15 percent of their investment from duction structures, facilitating new, more cost-effec- 2000–07. Manufacturing fi rms in Mexico responded tive processes. Indeed, the vast majority of effi ciency to higher competition from low-cost Chinese pro- gains emerge outside the information and communi- ducers in the domestic and the U.S. (export) market cation technology (ICT) sector, where fi rms use the by using digital technologies more intensively and internet to sell and market their products online or productively. Manufacturing fi rms in Brazil facing share real-time information with suppliers to mini- an increase in competition are more likely to invest mize their inventory and with customers to optimize in e-commerce systems. Firms in Africa facing an their services. increase in competition are more likely to use the The impact of digital technology on economic internet to market their products or to manage their growth is mediated through three mechanisms— inventories. Firms’ use of digital technology also ACCELERATING GROWTH 51 varies with barriers to competition across sectors in Figure 1.1 A framework for the internet and economic the same country. Business process outsourcing in growth the Philippines has few entry barriers, and fi rms use digital technology intensively—the retail sector on DIGITAL the other hand faces substantial restrictions to entry TECHNOLOGIES and is dominated by incumbent fi rms, with few of them offering e-commerce. Harnessing the full growth potential of digital technology is thus predicated not just on investments in skills (chapter 2) and infrastructure (chapter 4) but Augment also on reforming regulatory barriers by overcoming Overcome Generate information barriers existing factors economies of scale vested interests to encourage all fi rms to compete by investing in these new technologies (chapter 5). This also involves overhauling regulatory regimes in the digital economy, especially in sectors where online INCLUSION EFFICIENCY INNOVATION and offl ine fi rms increasingly compete, such as retail, transportation, printing and publishing, lodging, and fi nance. The initial entry of internet fi rms into International trade Capital utilization Competition these sectors promotes competition and can disrupt traditional monopolies. But internet fi rms can be prone to anticompetitive behaviors by exploiting RISK: DIVERGENCE AND MONOPOLY POWER scale and network effects. So the regulators need to level the regulatory regime to guarantee entry and prevent market shares from becoming too Source: WDR 2016 team. concentrated. The greater digital adoption therefore needs to be accompanied by unifi ed standards, full interoperability, and competition across platforms More than 90 percent of fi rms in high-income and contracts. countries, and 46 percent in low-income countries, The goal is to have fi rms’ use of the internet pro- used electronic mail (e-mail) to communicate with mote competition, which encourages more fi rms to clients between 2010 and 2014 (fi gure 1.2, panel b). The use the internet. But that will not happen if vested differences are greater when the internet is used for groups are strong enough to capture regu- more demanding business activities. For instance, 42 lators and create new barriers to competition and percent of fi rms in upper-middle-income countries technology adoption. A level playing fi eld for busi- had a website, and 30 percent purchased goods or ser- ness was always important—digital technologies vices online. Only 14 percent of fi rms in low-income have made it an imperative. countries purchased goods or services online, and only 11 percent delivered them online. Connected businesses Developing countries such as Vietnam have invested heavily in the rollout of (broadband) internet The adoption of broadband internet has increased for infrastructure in recent years. As a result, the share fi rms in all country income groups. Almost all fi rms of manufacturing and service fi rms in Vietnam using in high-income OECD countries (with at least fi ve the internet for business activities rose to 71 percent employees) used a broadband internet connection in 2007 and 86 percent in 2011. Internet access was up between 2010 and 2014, with usage rising from 79 per- almost uniformly across all provinces (map 1.1).2 cent in 2006–09 to 92 percent in 2010–14 (fi gure 1.2, But many advanced digital technologies have not panel a). The increase between the two periods was yet diffused widely, even in high-income countries. even stronger for lower-income countries. The share Almost all European fi rms with at least 10 employees of fi rms in lower-middle-income countries using use a personal computer (PC) and broadband internet. broadband internet rose from 39 percent in 2006–09 About 80 percent have a website, and 60 percent use to 68 percent in 2010–14. The share in low-income supply chain management software that is integrated countries in 2010–14 is still fairly low (38 percent), but with the ICT systems of customers or suppliers out- with some notable exceptions. side of the fi rm (fi gure 1.3). But less than 20 percent 52 WORLD DEVELOPMENT REPORT 2016

Figure 1.2 More fi rms are using broadband internet

a. Share of firms using broadband internet, 2006–14 b. Type of internet use, 2010–14

100 100 90 90 80 80 70 70 60 60 50 50 Percent 40 Percent 40 30 30 20 20 10 10 0 0 Low- Lower- Upper- High- High- Low- Lower- Upper- High- High- income middle- middle- income income income middle- middle- income income countries income income non-OECD OECD countries income income non-OECD OECD countries countries countries countries countries countries countries countries 2006–09 2010–14 Internet Internet Website e-mail deliveries purchases

Source: Hussain (2015) based on World Bank Enterprise Surveys, various years. Data at http://bit.do/WDR2016-Fig1_2. Note: The data cover fi rms with at least fi ve employees. Panel a is based on 10,161 fi rms in 100 countries (2006–09) and 33,467 fi rms in 88 countries (2010–14). The survey question is, “Does this establishment have a high-speed, broadband internet connection on its premises?” Panel b is based on 100 countries in 2010–14 with 45,892 fi rms (e-mail), 45,872 fi rms (website), 20,480 fi rms (internet purchases), and 17,099 fi rms (internet deliveries). OECD = Organisation for Economic Co-operation and Development.

of European fi rms purchase or sell goods or services More productive fi rms are early adopters online and use any cloud computing services. Only 8 of the internet percent use cloud computing services to acquire man- More productive fi rms are more likely to adopt the agement software and computing power. And only 3 internet and use it more intensively. The percentage percent use radio frequency identifi cation devices, of fi rms having a website and selling or buying goods which are wireless microchips used to connect or services online tends to increase with fi rm pro- machines with one another (the “internet of things”). ductivity in all country income groups (fi gure 1.4).4 The use of digital technologies is still basic in most Firms in the most productive quintile are most likely developing countries. In Vietnam, only 2.2 percent of to have a website or use broadband internet.5 The cor- all fi rms sold their goods or services online in 2011 relation between fi rm productivity and e-commerce (up from 0.6 percent in 2007). In Turkey, 92 percent is stronger in upper-middle-income countries. In of all fi rms with at least 10 employees had internet lower-middle-income countries, less than 13 percent access in 2012, but only 71 percent used it for banking; of the most productive fi rms sold or bought goods 55 percent had a website; and 10 percent made orders or services online. The results suggest that only the or reservations online. In Mexico, 84 percent of fi rms more productive fi rms in developing countries over- with at least 10 employees used the internet in 2012, come (unobservable) barriers to using the internet but only 26 percent of employees had internet access; more effectively. and only 13 percent of fi rms used e-commerce for pur- African fi rms using the internet have on average chases, and 9 percent for sales. In Brazil, 63 percent 3.7 times higher labor productivity than nonusers of all manufacturing fi rms with at least 10 employees and 35 percent higher total factor productivity (TFP) bought or sold products online, but only 28 percent (fi gure 1.5). But the most productive fi rms that do provide an online catalogue or feature online orders not use the internet have TFPs comparable to high- on their website; and only 6 percent allow online pay- productivity internet users. The differences are also ments. And less than half the manufacturing fi rms found to be larger for labor productivity than for in Brazil have an information technology (IT) depart- TFP, implying that African fi rms using the internet ment or use software supporting management, either are not only more productive but also more capital for resource planning or customer relations.3 intensive.6 ACCELERATING GROWTH 53

Map 1.1 Many more fi rms are using the internet in Vietnam

a. Usage in 2007 b. Usage in 2011

Firms using the internet (%) 30–59 60–79 80–100

IBRD 42011 Source: Nguyen and Schiff bauer 2015. Data at http://bit.do/WDR2016-Map1_1. Note: The census of Vietnamese fi rms includes more than 300,000 observations each year.

Figure 1.3 Many advanced digital technologies have not yet diff used across fi rms in high-income countries, 2014

100 Maximum 75th percentile 80 Median

Mean 60 25th percentile Minimum 40 % of firms using these technologies these 20

0

PC ERP SCM CRM RFID e-sales ERP Internet Website Broadband e-purchases

Cloud computing Cloud computing Digital technologies Source: Eurostat, circa 2014 (EC, various years). Data at http://bit.do/WDR2016-Fig1_3. Note: For each technology, the chart shows the distribution across 32 high-income countries of the share of fi rms (with at least 10 employees) that use that technology. Data are for 2014 or the last available year. CRM = customer relationship management software; ERP = economic resource planning software; PC = personal computer; RFID = radio frequency identifi cation technologies; SCM = supply chain management software. 54 WORLD DEVELOPMENT REPORT 2016

Figure 1.4 Higher-productivity fi rms are more likely to use the internet, 2010–14

a. Website b. e-commerce 55 28

24 45

20

35 16

12 Percent of firms Percent

25 of firms Percent 8

15 4 1. Qtl 2. Qtl 3. Qtl 4. Qtl 5. Qtl 1. Qtl 2. Qtl 3. Qtl 4. Qtl 5. Qtl Distribution of firm productivity Distribution of firm productivity Low-income countries Lower-middle-income countries Upper-middle-income countries

Source: Hussain (2015) based on World Bank Enterprise Surveys, various years. Data at http://bit.do/WDR2016-Fig1_4. Note: The fi gures show the share of fi rms that have a website or use e-commerce among the diff erent labor productivity quintiles across country income groups. “Qtl” refers to log labor productivity quintiles. The data, pooled for all years between 2006 and 2014, cover fi rms with at least fi ve employees.

Figure 1.5 African fi rms using the Firms that use digital technologies intensively internet are more productive, 2014 also share other characteristics of high-productivity Distribution of sales per worker fi rms. They tend to be larger, fast-growing, skill- intensive, export-intensive, and located in the cap- 1 ital city.7 Larger fi rms use the internet more inten- sively across all country income groups (fi gure 1.6), .8 and 65 percent of large fi rms but only 21 percent of small fi rms use broadband internet in low- .6 income countries. The differences are comparable for fi rms that use the internet for more demanding business activities: 17 percent of small fi rms deliver .4 goods or services using the internet in lower- middle-income countries, compared with 43 percent .2 Firms not using the internet Firms of large fi rms. Census and detailed survey data from Mexico, 0 Turkey, and Vietnam confi rm the positive correla- 0.2.4.6.81 tion between internet use and the characteristics of Internet users high-productivity fi rms. Firms in Turkey using the Source: Cirera, Lage, and Sabetti 2015. Data at http://bit.do/WDR2016-Fig1_5. web for online orders or reservations are 11 percent Note: The fi gure shows the relative distribution of productivity for fi rms that more productive, 25 percent larger, and twice as use the internet and fi rms that do not. Productivity is measured by sales per 8 worker, estimated at the sector level. Results are similar using value added likely to export. Those having a website are twice instead of sales. Observations on the 45-degree line indicate that fi rms in as productive, twice as large, and more than twice as both groups have the same productivity. Firms not using the internet are less productive (above the 45-degree line) throughout the distribution of labor likely to export as fi rms that do not have a website. productivity. For instance, the median productivity fi rm among nonusers has In Mexico and Vietnam, fi rms are more productive about the same productivity as the 20th percentile fi rm of the productivity distribution among internet users. The surveys are representative of all fi rms if they have more computers per worker, conduct with at least fi ve employees in six African countries: the Democratic Republic e-commerce, or have a higher share of workers using of Congo, Ghana, Kenya, Tanzania, Uganda, and Zambia. the internet.9 ACCELERATING GROWTH 55

Figure 1.6 Larger fi rms use the internet more intensively across all income groups, 2006–14

100 90 80 70 60 50

Percent 40 30 20 10 0 Small Medium Large Small Medium Large Small Medium Large

Low-income countries Lower-middle-income countries Upper-middle-income countries Firm size and country income group Sell online Website Broadband

Source: Hussain (2015) based on World Bank Enterprise Surveys, various years. Data at http://bit.do/WDR2016-Fig1_6. Note: The fi gure shows the percentage of manufacturing and service fi rms that deliver orders over the internet, have a website, and use broadband internet by fi rm size across low-income, lower-middle-income, and upper-middle-income country groups. The data, pooled for all years between 2006 and 2014, cover only fi rms with at least fi ve employees. Small fi rms have between 5 and 19, medium fi rms between 20 and 99, and large fi rms more than 100 employees. The statistics are based on about 100 countries (some countries being surveyed twice) and 27,035 fi rms selling online, 94,083 with a website, and 43,628 using broadband.

More trade, higher (5.4 percent) in developing countries, consistent with their less intensive use of ICTs. Firms in developing productivity, and greater countries have considerable room to adopt digital competition solutions that have led to growth in high-income countries, such as using the internet for e-commerce Digital technologies raise growth, or inventory management. but quantifying their aggregate impact The true contribution to growth can be larger is diffi cult if ICT complements other production factors, but The accumulation of ICT capital accounted for almost also smaller if it substitutes for them. Productivity 20 percent of global growth between 1995 and 2014, externalities beyond measured production factors approaches suggest. However, the are ascribed to the residual TFP growth. A necessary results need to be regarded with some caution, as the (but not suffi cient) condition to accurately measure approach involves some severe measurement prob- productivity externalities is analysis at the fi rm level. lems (box 1.1). The impact was largest between 2005 Assessing the growth opportunities of the internet and 2009, when it raised aggregate annual growth by warrants more detailed analysis of the mechanisms 1.3 percentage points a year in developing countries, for it to affect growth. Against this background, it is out of a total of 6.6 percent among these countries, instructive to draw insights from the economics of or 19 percent (fi gure 1.7). And among high-income the internet as well as from the past industrial revo- countries, it contributed 0.4 percentage points out of lutions (see box 1.2). The internet reduces transaction 1.2 percent, or 38 percent.10 costs, allowing fi rms to enter new markets, enhance Digital technologies accounted for a higher share their effi ciency, and exploit economies of scale, lead- of aggregate growth in high-income countries, point- ing to innovation. It does this by reducing information ing to greater digital use. The accumulation of ICT frictions, search costs, and the costs to communicate. capital accounted on average for 27 percent of aggre- The decline can be dramatic if fi rms adapt their busi- gate growth (2.1 percent) from 1995 to 2014 in high- ness models to automate data-intensive transactions, income countries and 14 percent of aggregate growth generating economies of scale. 56 WORLD DEVELOPMENT REPORT 2016

Box 1.1 Tracing back growth to a single, new technology suff ers from severe measurement problems

My view is that pinpointing precisely which (explanatory) variables matter for growth is impossible. —Robert Barro, 2015

The limited number of observations relative to the seem- cross-country growth regressions are fairly small due to the ingly open-ended list of other potential growth correlates fi nite number of countries and the typically limited low- at the country level makes it almost impossible to reject frequency time-series variation in the data.b alternative interpretations of the same macroeconomic Growth accounting provides a less ambitious approach; it correlation. Faster-growing countries, for instance, have decomposes aggregate growth into the contribution of each more resources and economic opportunities available to production factor without getting to the issue of causality. invest in information and communication technology (ICT) But its precision also hinges on the ability to compute ICT infrastructure; thus, the direction of causality may run from capital stocks, which requires measuring depreciation rates growth to ICT rather than the other way around. Moreover, and price indexes for digital technologies that appropriately there are severe endogeneity issues, since diff erences in refl ect quality changes.c The approach also assumes that the provision of ICT infrastructure across countries are likely the contribution of each factor of production to output is to be positively correlated with (time-varying) unobserv- proportional to the corresponding share in total input costs. able, country-specifi c GDP correlates such as government So, the true contribution to growth can be larger or smaller. accountability and other institutional factors, leading to an Firm-level studies comprise large sample sizes and upward bias in ICT-growth estimates. allow comparisons of the performance among fi rms oper- Cross-country growth regressions are not an appropri- ating in similar institutional environments. ICT investment ate tool to draw inference on the impact of ICT on growth. is a fi rm’s decision—still it is not plausible to assume that Numerous studies fi nd a positive correlation between GDP it is independent of performance. A positive productivity growth and (lagged) values of diff erent forms of ICT capital correlation in the data might simply capture that more stocks based on cross-country growth regressions.a It is productive fi rms use digital technology more eff ectively, tempting to take the estimated ICT-growth elasticities at indicating that other potentially unobservable fi rm-specifi c face value to quantify the impact of digital technologies factors explain the positive correlation between digital on growth. But this approach has serious shortcomings. technologies and fi rm productivity. In fact, many fi rm-level Most important, compared to the seemingly open-ended studies claiming to fi nd productivity externalities still suff er list of potential growth correlates, the sample sizes in from measurement and reversed causality problems.d,e

a. See Czernich and others (2009), Koutroumpis (2009), or Niebel (2014) for more recent contributions using cross-country regressions to assess the impact of broadband infrastructure on GDP growth. Qiang, Rossotto, and Kimura (2009); Cardona, Kretschmer, and Strobel (2013); and Minges (2015) provide surveys of the literature using cross-country regressions to assess the correlation between ICT capital and broadband internet on GDP growth, respectively. b. Several researchers use , such as generalized method of moments (GMM) panel estimators, to address the endogeneity issue in cross- country growth regressions, but the results are typically not robust to small variations in the included countries, time periods, or control variables due to the large number of necessary country-level control variables, given the open-endedness of potential factors correlated with growth and ICT capital investments (Hauk and Wiczarg 2009). c. But the acceleration in computing power (Moore’s observation) involves a severe measurement problem, as conventional price indexes do not capture changes in quality of hardware or software. Jorgenson (2001) addresses this problem for hardware by constructing a constant quality index of computer prices using hedonic techniques. The Conference Board (2015) uses the World KLEMS data following this , harmonizing the most recent techniques across countries. d. See, for example, Gordon (2010, 2014) or Acemoglu and others (2014). e. For instance, Draca, Sadun, and Van Reenen (2007) argue that causality has not yet been convincingly demonstrated.

The reduction in transaction costs thus increases Inclusion—facilitating international trade inclusion (market access), effi ciency, and scale, which Online marketplaces can reduce differences in the translate into economic growth primarily through information available to buyers and sellers (for exam- the three channels of trade, capital utilization, and ple, information asymmetries), enabling more fi rms competition. in developing countries to engage in international ACCELERATING GROWTH 57

Figure 1.7 ICT capital accounted for nearly one-fi fth of global growth, 1995–2014

a. High-income countries b. Developing countries 7 3 6

5 2 4 Percent

Percent 3

1 2

1

0 0 1995–99 2000–04 2005–09 2010–14 1995–99 2000–04 2005–09 2010–14 Labor Labor quality Non-ICT capital ICT capital TFP

Source: Conference Board 2015. Data at http://bit.do/WDR2016-Fig1_7. Note: The fi gure shows the average contribution of factor inputs and TFP to aggregate growth for 39 high-income countries and 91 emerging economies. ICT = information and communication technology; TFP = total factor productivity.

Box 1.2 Is this time diff erent? Predicting labor productivity growth at the technological frontier based on lessons from past industrial revolutions

History shows that productivity growth driven by general purpose technologies can arrive in multiple waves; it need not simply arrive, give what it has, and fade away forever thereafter. —Chad Syverson, 2014

It is easy to fi nd echoes of today’s debates about the How do information and communication technologies internet in the history of the industrial revolution. Each (ICTs) stack up with past major general-purpose technol- technological breakthrough comes with its own idiosyn- ogies? For instance, electricity led to the invention of air cratic variations, but much can be learned by studying the conditioning, elevators, and home appliances half a century deliberations and experiences of the past. later. Arguably, the internet can be regarded as a major Technological revolutions often take a long time to have follow-up invention from the computer (or microchip), signifi cant impacts. The maximum impact of steam power just as the internal combustion engine led to the interstate on British productivity growth was not felt until the third highway system in the United States 60 years later. The quarter of the 19th century, nearly 100 years after James internet boosts the productivity of computers and mobile Watt’s patent.a The benefi ts of railroads were fairly small phones in the same way that highways boosted the produc- initially, but grew as railroad productivity improved and tivity of cars and trucks. rail output rose as a share of economic activity.b Similarly, U.S. data show that labor productivity increased on investments in electrical capital equipment did not have average by 2.4 percent annually after the inventions of the important spillovers until the 1920s.c Initially, factory owners internal combustion engine, electricity, and running water simply replaced large steam engines with large electric ones. between 1891 and 1972. Then it slowed, averaging only 1.4 It took nearly 40 years after electricity was widely available percent annually between 1973 and 1996. Between 1996 in the United States for organizational methods to catch up and 2004, it surged again, growing 2.5 percent annually, and develop more effi cient decentralized production lines. which scholars ascribe to the fast spread of the internet.d

(Box continues next page) 58 WORLD DEVELOPMENT REPORT 2016

Box 1.2 Is this time diff erent? Predicting labor productivity growth at the technological frontier based on lessons from past industrial revolutions (continued)

Since 2004 aggregate productivity growth has returned to Figure B1.2.1 U.S. labor productivity its pre-1996 level, averaging 1.3 percent annually. during the electrifi cation era Does the 2004–13 productivity slowdown in the United (1890–1940) shares remarkably States refl ect a petering out of the ICT revolution? Gordon common pat terns with the ICT era interprets this as a sign that the IT (or internet) revolution (1970–2012) does not have the same lasting impact as previous industrial revolutions triggered by the internal combustion engine or ICT era electricity.e But the diffi culty of tracing a country’s growth performance to a single new technology, given severe mea- 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 surement and endogeneity problems, allows alternative 180 interpretations of the same recent historical trend. 160 Syverson overlays U.S. labor productivity in the electrifi - 140 cation era and the ICT era to show that productivity growth 120 in both eras has exhibited remarkably common patterns so 100 f far (fi gure B1.2.1). During the electrifi cation era, labor pro- 80 ductivity growth over the fi rst 25 years was relatively slow, 60 just as it was in the ICT era from 1970 to 1995. Then both

Labor productivity (index) Labor 40 eras saw decade-long accelerations in productivity growth,

spanning 1915–24 for electrifi cation and 1995–2004 for IT. 1890 1895 1900 1905 1910 1915 1920 1925 1930 1935 1940 Analogous to the 2004–12 slowdown, labor productivity Electrification era growth slowed again in the electrifi cation era in 1924–32. Electrification ICT Then labor productivity growth in the electrifi cation era sped up again, averaging 2.7 percent annually in 1932–40. Sources: Syverson 2013; Crafts 2015. This does not necessarily suggest that labor productivity in Note: ICT series years are labeled on the upper horizontal axis; the series is indexed to a value of 100 in 1995. Electrifi cation series years the United States will pick up again in coming years. are labeled on the lower horizontal axis; the series is indexed to 100 in 1915. ICT = information and communication technology.

a. Von Tunzelmann 1978. b. Fogel 1965. c. David 1990. d. Gordon 2014; Syverson 2013. e. Gordon 2010, 2014. Byrne, Oliner, and Sichel (2013) use a growth decomposition of long-term time-series data and come to a similar conclusion. f. Syverson 2013.

trade (box 1.3). To purchase any product, the buyer places that operate internationally, like eBay, Ama- must fi nd a seller, make a payment before receiving zon, and Alibaba, operate websites in local languages. the goods, and trust that the seller will deliver the Online marketplaces also include rating systems, correct amount and quality on time. As a result of allowing the buyer and seller to assess each other’s these information asymmetries, only the most pro- performance. The ratings and individual comments, ductive fi rms export—typically the larger and older visible to anyone, build trust for future transactions fi rms.11 Online marketplaces help solve all three and encourage more responsible behavior. Many problems. They provide an organized marketplace online marketplaces also provide payment and for fi rms to advertise their products and fi nd buyers delivery services to reduce the cost of e-commerce. in overseas markets.12 They thus reduce the costs of Digital technologies also lower communication costs trade by enabling fi rms to avoid intermediaries to facilitating the unbundling of tasks, allowing fi rms establish trade connections or participate in costly to offshore production processes and services to trade fairs to market their products. Online market- developing countries at lower costs. ACCELERATING GROWTH 59

Box 1.3 Is the internet reshaping ? Not yet.

Fifteen years ago, a widely read study pronounced the delivered over the internet. But distance might be captur- death of distance and concluded that, with better and ing taste preferences that can infl uence online activities. cheaper communication technology, geography will no The eff ects of distance are not uniform across all website longer infl uence individuals and fi rms. It argued that the visits. Instead, websites selling digital goods that are dif- internet is making it easier and cheaper to communicate ferentiated and dependent on taste (music and games) over long distances. Goods and services, increasingly virtual are aff ected by distance, but more standard digital goods and weightless, can be delivered over the internet. That (software and fi nancial information) are not.f The eff ect of makes location irrelevant to what an individual consumes, to distance on online transactions also diff ers across product where an individual works, or to where the fi rm establishes categories, where goods that appeal to a local market (tick- its business.a ets and sports memorabilia) are most aff ected by distance.g Even though the internet has reduced the economic dis- The impact of the internet on the location decisions of tance between countries, geography still determines fl ows fi rms and the geography of jobs is less clear. The internet of goods and services. Bilateral trade between countries is facilitates the dispersion of fi rms by allowing better com- still negatively related to the distance between them after munication between workers performing complex tasks controlling for the eff ects of the internet.b And online trans- and greater disintermediation of production processes. actions are still negatively related to distance: transaction And it enables the dispersion of jobs, as individuals fi nd values on online market platforms like eBay in the United work regardless of their location through online labor States and MercadoLibre in Latin America decline when exchanges (chapter 2). But it also opens locations previ- the distance between buyers and sellers increases.c Internet ously considered economically unfeasible and allows fi rms websites hosted in countries closer to the United States to cluster in locations to take advantage of comparative receive more visits by U.S. consumers.d More generally, the advantage, further enabled by better communication tech- number of online transactions globally declines when the nology.h Jobs are clustered in these locations because the bilateral distance between buyers and sellers increases.e fi rms want to access the dense labor market and special- Even so, the relationship between trade fl ows and geography ized inputs suppliers. The equilibrium eff ect of the internet is weaker for online transactions than for offl ine transactions. on the location of fi rms and jobs is diffi cult to determine, The impact of distance on digital goods may at fi rst but the fi nal global pattern of fi rm location could be many seem counterintuitive, given that they are weightless and clusters connected by the internet.i

a. Cairncross 2001. b. Freund and Weinhold 2004. c. Hortaçsu, Martínez-Jerez, and Douglas 2009; Lendle and others 2012. d. Blum and Goldfarb 2006. e. Cowgill and Dorobantu 2014. f. Blum and Goldfarb 2006. g. Hortaçsu, Martínez-Jerez, and Douglas 2009. h. Leamer and Storper 2001. i. Venables 2001.

More internet use in a country is positively related internet use and the importer has low internet use, to the growth of bilateral exports of goods and ser- relative to two countries with low internet use.15 vices.13, 14 The relationship tends to depend on country income. Higher internet use in a developing country is Enabling more fi rms to reach new markets related to higher exports to high-income countries, but The internet makes it easier to reach new markets not necessarily to other developing countries. Higher and thereby increases the extensive margin of trade— internet use in developed countries, in turn, has no more fi rms start to export and more products get effect on their exports. The similarities in internet use exported. A 10-percent increase in internet use in an between trading partners can also affect how the inter- exporting country increases the number of products net affects trade. The trade between two countries is traded between two countries by 1.5 percent; this 25 percent higher if both countries have high internet increase in the extensive margin can account for as use and 31 percent higher if the exporter has high much as 78 percent of the total effect of the internet 60 WORLD DEVELOPMENT REPORT 2016

Figure 1.8 The internet enables more Looking forward, the rise of online marketplaces fi rms to reach new markets, 2001–12 can accelerate the integration of developing countries into world markets, opening substantial opportuni- 2.0 ties for trade and future growth. In China, Alibaba already enables smaller and younger fi rms to sell 1.5 more products and reach more new consumers or businesses in foreign countries. Online marketplaces also emerged in other developing regions, but so far 1.0 only on a (much) smaller scale or for market niches. The Moroccan online platform Anou enables artisans 0.5 in rural areas to directly export their products online, 22

Trade increase (%) cutting out traditional middlemen. The Kenyan 0 online platform iProcure prescreens its vendors to provide reliable local procurement services connect- 23 –0.5 ing agricultural businesses and institutional buyers. Internet use Internet use So, e-commerce platforms should have substantial in exporter in importer country country growth effects if they achieve suffi cient scale in devel- Intensive margin of trade oping regions (see box 1.4). Extensive margin of trade

Source: Osnago and Tan 2015. Data at http://bit.do/WDR2016-Fig1_8. Increasing the volumes of exported Note: Extensive margin of trade: more fi rms start to export and more products products get exported; intensive margin of trade: exporters increase the The intensive margin of trade—defi ned as the average amount of their exports for the same products. trade per fi rm or product—increases with the internet penetration of countries. A 10-percent increase in the on trade (fi gure 1.8). The impact of the internet is larg- internet use of a country pair increases the average est if both countries have high internet use.16 bilateral trade value per product by 0.6 percent. Online platforms allow smaller fi rms to become More specifi cally, the average export value per fi rm exporters. The fi rms selling on eBay in Chile, Jordan, increases by 1 percent if internet use increases by 10 Peru, and South Africa are younger and have smaller percent in the exporting country; it increases by 0.5 market shares than fi rms in the offl ine markets.17 The percent if internet use increases by 10 percent in the Alibaba platform has a larger fraction of small fi rms importing country.24 than the offl ine market.18 Firms exporting through The rollout of broadband infrastructure boosted Alibaba also sell more products per fi rm.19 exports and labor productivity in China even in the Online platforms enable fi rms to reach new era before Alibaba, from 1999 to 2007. The number of export destinations, changing trade patterns. Mar- internet users increased across all provinces between keting goods globally on the internet is cheaper, 1997 and 2007, though it was stronger in coastal areas and customizing the marketing information to suit in the earlier years and in several inland provinces in the local context and language is easier. So, fi rms later years.25 The value of real exports seems to have trading through online platforms can reach more followed a similar pattern.26 The increase in internet destinations than their offl ine peers. Chinese com- domains and users per capita had a positive impact panies selling on Alibaba reach an average of 3 and on fi rms’ manufacturing exports in ICT-intensive a maximum of 98 export destinations,20 up from an sectors.27 It raised the number of fi rms that export, average of one and a maximum of 50 export des- the fi rms’ share of export in total sales, and the real tinations for the offl ine fi rms.21 The set of export value of fi rms’ exports. The higher share of internet destinations also differs for fi rms exporting online domains and users also increased fi rms’ real output or offl ine in China. And a larger share of the online and labor productivity. exports reach emerging markets such as Brazil, Col- ombia, India, Nigeria, the Russian Federation, and Facilitating the unbundling of tasks Ukraine, compared with exporting offl ine, which Better communication technologies have facilitated is more intensive in low-income countries or tradi- the unbundling of tasks, the “second unbundling” tional high-income markets such as Germany, Japan, of international trade.28 Businesses can locate dif- and the United States (map 1.2). ferent stages of their production in different host ACCELERATING GROWTH 61

Map 1.2 China’s export destinations diff er for fi rms using online platforms, 2006 and 2014

Trade type Substantially more offline trade More offline trade More online trade Substantially more online trade Not applicable or no data

IBRD 41767 Source: Chen and Xu 2015. Data at http://bit.do/WDR2016-Map1_2. Note: Countries are grouped according to their share of total revenue in online trade relative to offl ine trade. Online trade is measured by Alibaba platform transaction data, and offl ine trade by customs data. Substantially more offl ine trade is a ratio below 0.5 for the share of a country’s online exports in world online exports divided by the share of its offl ine exports in world offl ine exports. More offl ine trade is a ratio between 0.5 and 1. More online trade is a ratio between 1 and 4.75 (75th percentile). Substantially more online trade is a ratio above 4.75.

Box 1.4 Successful online platforms account for local context and institutions

Successful online platforms cannot necessarily be trans- Alibaba’s success can be explained (at least in part) by planted to a foreign country without adapting to the its ability to adapt to local market conditions and culture to local market and local institutions. The entry of eBay into remove the information asymmetries unique to the Chinese China illustrates this. Both eBay and Taobao, Alibaba’s C2C market. The Taobao website, for instance, indicates consumer-to-consumer (C2C) platform, entered the Chinese whether sellers are online and allows buyers to communi- market in 2003. eBay immediately established its dominant cate instantly with the sellers through an online messaging market share of 85 percent by acquiring the Chinese market system.b Alibaba further guarantees foreign buyers in-time leader, EachNet.a Taobao, however, quickly assumed a dom- delivery of their money and implemented a system to verify inant position. By 2014, it accounted for over 80 percent of sellers on its website for business-to-business (B2B) trans- the Chinese C2C market (eBay exited the Chinese market actions. Firms can buy a gold supplier status to have a third- in 2006). party verifi cation company conduct on-site quality control.

a. Wulf 2010. b. Ou and Davidson (2009) note that Chinese consumers want to be able to communicate with their sellers and solve any problems relating to the transaction immediately. 62 WORLD DEVELOPMENT REPORT 2016

countries to achieve greater effi ciency in each stage. cal services can be supplied internationally over the This unbundling of tasks allows fi rms to outsource internet. Online platforms like Upwork (formerly production processes and services to developing Elance-oDesk) create marketplaces for freelancers to countries at lower cost, which in turn increases trade provide these services (see chapter 2). in intermediate goods and places more importance on global supply chains. This trend has, for example, Effi ciency—raising fi rms’ utilization of enabled the strong growth in business process out- capital and labor sourcing in , India, and the Philippines. Digital technologies help fi rms save costs by auto- The foreign value added content of exports increased mating data-intensive production processes and substantially along all major trade routes over the past reorganizing their business models, increasing their 20 years, particularly among Asian economies. productive use of capital and labor.34 Computers Digital technologies reduce the costs of offshor- and software allow fi rms to routinize processes, ing, especially when the host country has signifi cant increasing management effi ciency and replacing human capital. So, production structures become personnel with, for instance, human resource or more globalized, and more fi rms choose to offshore supply chain management software. The internet different stages of their production. Higher use further increases the opportunities for cost saving by of ICTs by fi rms allows headquarters to manage connecting machines, suppliers, and clients, so that increasingly complex functions, enabling them to fi rms can manage their supply chains and inventory confi dently send these functions overseas.29 As com- more effectively in real time. The extent to which dig- munication costs decline, exports of headquarter ital technologies raise labor productivity depends on services to foreign affi liates, such as professional and fi rms’ activities, but examples abound across coun- technical services, increase.30 These functions are, tries and economic sectors. however, complex, implying that fi rms select host The vast majority of effi ciency gains emerge countries based on their human capital.31 outside the ICT sector, where fi rms use the internet Better use of digital technology in a country to sell and market their products online or share encourages multinational fi rms to locate their sub- real-time information with suppliers to minimize sidiaries there. Many foreign subsidiaries are in their inventory and with customers to optimize their countries with high business use of the internet, and services. The digital economy accounts for about the entry of a multinational into a foreign market is 6 percent of GDP in OECD countries (box 1.5). Less positively correlated with business use.32 And the pro- visible but more important for growth, incumbent vision of digital technology in the host country has fi rms in traditional sectors invest in digital technolo- a stronger effect on the entry of multinational fi rms gies to save costs by optimizing their production and when the fi rm is in an industry that uses communi- management processes. Retail companies such as cation technology more intensively and has fewer Walmart, for instance, have integrated (global) sup- routine tasks. ply chains to minimize their inventory holdings by The unbundling of tasks has driven the tripling linking electronic cash registers at retail outlets and of service trade over the past 15 years, particularly for business-to-business ordering systems with order business, professional, and technical services such dispatch and transportation scheduling at remote as legal, advertising, consulting, and accounting. The factories. So far, the largest effi ciency gains from internet makes it easier for headquarters to transmit fi rms using digital technologies have been found information, supervise their factories, and coordinate in wholesale and retail trade, business services, the supply chain across borders, encouraging fi rms insurance, fi nance, and selected manufacturing to outsource not only manufacturing but also service sectors. The following section highlights selected tasks. Trade in education and professional services illustrations of how digital applications boost fi rms’ has also increased.33 Trade in education services prior effi ciency. to the internet was conducted through correspon- dence courses, with textbooks and course materials Increasing management effi ciency mailed between instructors and students. Now web- Crowdsourcing information through mobile phone sites provide massive open online courses, known applications is helping farmers in Tanzania to pre- as MOOCs, with video and other materials over the vent the outbreak of diseases. In Tanzania’s Lake internet. They also connect health service providers Zone, 60 groups of farmers from 10 districts belong- enabling the spatial unbundling of health services, ing to the Digital Early Warning Network prevent such as radiology. Similarly, professional and techni- the outbreak of the cassava disease by crowdsourcing ACCELERATING GROWTH 63

Box 1.5 The growth impact is largest when fi rms in traditional sectors use digital technologies to modernize their business

The digital economy accounts for about 6 percent of gross relatively small. Consider Amazon. In the United States, its domestic product (GDP) in OECD (Organisation for Economic home market, Amazon accounted for only 1.7 percent of the Co-operation and Development) countries (fi gure B1.5.1).a retail market in 2014; e-commerce accounted for about 7 In the United States, which hosts some of the biggest tech percent of the U.S. retail market in 2015. By contrast, offl ine companies, the value added of information and communica- sales accounted for more than 99 percent of all sales of tion technology sectors in GDP is 7 percent, compared with and beverages; 84 percent of clothing and furniture; and 59 13 percent in real estate, renting, and leasing; 12 percent in percent of books, magazines, and music.d wholesale and retail trade; and 8 percent in fi nance and insur- Less visible but more important for growth, incumbent ance or health and social services.b The value added of infor- fi rms in traditional sectors invest in digital technologies to mation and communication technology (ICT) sectors in GDP save costs by optimizing their production and management is the highest in Ireland (12 percent), thanks to large infl ows processes. The McKinsey Global Institute estimates that 75 of foreign direct investment. In Kenya, which hosts one of the percent of the economic impact of the internet in 12 large largest ICT sectors among African countries, the value added developing and developed countries originates in fi rms share of ICT services in GDP was 3.4 percent in 2013; that in traditional sectors—fi rms that would exist without the includes telecommunications (and thus mobile money).c internet but that use it to increase their cost-eff ectiveness.e Despite the strong growth and high market valuations The fi ndings are consistent with studies showing that the of internet fi rms off ering conventional services, their mar- bulk of U.S. productivity growth over the past 15 years orig- ket shares in these traditional sectors to date have been inated in ICT-using, not-producing sectors.f

Figure B1.5.1 The ICT sector accounts for 4–7 percent of GDP in most OECD countries, 2011

14 12 10 8 6 Percent 4 2 0

Italy Spain Austria Poland France Japan Norway IcelandBelgium Greece Canada FinlandSwedenEstonia Ireland Portugal Slovenia Germany Denmark Hungary Switzerland Korea, Rep. Netherlands LuxembourgUnited States Czech SlovakRepublic Republic United Kingdom Source: OECD 2015. Data at http://bit.do/WDR2016-FigB1_5_1. Note: ICT value added is the diff erence between the ICT sector’s gross output and intermediate consumption. The data are from 2011. The aggregate of information industries here includes ISIC rev. 4 division 26 (manufacture of computer, electronic and optical products) and section J (information and communication), which in turn comprises divisions 58–60 (publishing and broadcasting industries), 61 (telecommunications), and 62–63 (computer programming, and information service activities). ICT industries are broadly defi ned, as they include publishing and broadcasting, trade and repair activities, and media and content industries (Division 63.9). a. The McKinsey Global Institute (Manyika and others 2011) used a diff erent approach to compute the contribution of the internet to GDP. It computes the total value of three main activities based on the internet: activities using the internet as support (e-commerce, online content, or advertising); internet service providers; and software and internet service activities such as IT consulting, hardware manufacturers (computers, smartphones, or hardware equipment). They fi nd that these activities account on average for 3.4 percent of GDP among the G-8 countries, Brazil, China, India, the Republic of Korea, and Sweden in 2009. The impact is largest among the developed countries. The shares include publishing and broadcasting activities and media and content industries. b. The classifi cation includes internet fi rms such as search engines, but not online retailers. The market share of online retailers is, however, still relatively small: Amazon accounted for only 1.7 percent of the retail market in the United States in 2014; see http://www.bea.gov/. c. World Bank 2015. d. Economist 2014, 2015b. e. McKinsey Global Institute (Manyika and others 2011). The results are based on a survey of 4,800 small and medium enterprises in the following 12 countries: Canada, China, France, Germany, India, Italy, Japan, the Republic of Korea, the Russian Federation, Sweden, the United Kingdom, and the United States. f. See, for example, Jorgenson 2001, 2011. 64 WORLD DEVELOPMENT REPORT 2016

information.35 Each farmer has a topped-up phone effi cient real-time communication and coordination card and has been trained to recognize early symp- with clients and gives RW access to a large pool of toms of cassava mosaic disease and brown streak independent carrier trucks. Each truck has a global disease. The farmers use the mobile phones to send positioning system (GPS), so RW can track shipments monthly text messages to researchers about disease in real time and be notifi ed once the shipment is incidence. In return, they receive advice on disease delivered. The technology shortens the supply chain control, and disease experts visit the area when more gap between shipment and delivery in the domestic than 10 percent of the members of a group report the market, allowing RW to provide superior transport disease. The network, part of the Great Lakes Cassava services at more competitive rates. Initiative, supports more than 1 million farmers in Connected water metering systems help munici- six countries. palities identify leaks or changes in water use, reduc- Digital technologies help farmers in Botswana ing water losses by up to 10 percent. At its research meet quality and traceability requirements for beef center in France, U.S. manufacturer Itron developed exports to the European Union. The farmers use the water metering systems with wireless devices Livestock Identifi cation Trace-Back System, based on allowing remote over large distances, so transponders inserted into each animal, transmitting that municipalities can reorganize maintenance and information automatically to databases in 46 district respond immediately to water waste or losses in the offi ces. The centrally stored data allow exporters to distribution network. The data also improve water trace beef to the individual cattle and their owners, consumption analysis to optimize client service and necessary to meet traceability requirements for tariff structures. Itron’s production of communica- imports into the European Union. The technology tion devices has increased fi fteenfold since 2000, also helps improve veterinary services and save costs and its production of water meters threefold. Such by optimizing feeding schedules. connected metering systems have great potential to Digital interventions in agriculture seem to work save —more than 30 percent of the best when the information provided to farmers is drinking water produced worldwide never reaches simple, such as real-time prices or weather forecasts. the customer (often due to leaks).37 Based on simple mobile phone applications or inter- Looking forward, the internet of things should net kiosks, real-time data improve farmers’ access further increase the labor productivity of fi rms, to prices, local weather forecasts, and more detailed implying substantial opportunities for more effi - advice on agricultural practices and input use (see cient business models and thus future growth (see sector focus 1). spotlight 6). For instance, the German truck body and Real-time data help equipment manufacturers trailer manufacturer Schmitz Cargobull uses telemat- in China turn over their inventory stocks fi ve times ics (the integrated application of telecommunications faster than suppliers not connected to the internet. and data) in its trailers for real-time sharing of data Supply chains and logistics operations account for with drivers, freight agents, and customers on main- 10–20 percent of Chinese equipment manufacturers’ tenance, loaded weights, cargo temperatures, and costs, so companies using real-time data to optimize truck locations. The company adjusted its organiza- inventories and transportation routes can achieve tional structure and management practices to max- substantial effi ciency gains. Anji Logistics, a wholly imize effi ciency gains and escape competition from owned subsidiary of Shanghai Automotive Industry low-wage countries. As a result, the fi rm’s production Corporation, connects its IT systems with dozens pace necessary to meet customer demand fell by 90 of Chinese equipment manufacturers to manage percent, internal failures decreased by 92 percent, and logistics. General Motors China has cut costs and yearly accidents declined by 95 percent between 2000 improved customer satisfaction by building internet and 2013, as its market share for refrigerated trailers connectivity into cars, so that dealers can check faults in Europe rose to 50 percent.38 remotely and send maintenance alerts to owners. With only 20–25 percent of small manufacturing Enabling online marketing and distribution fi rms in China estimated to have used the internet in Firms are fi nding new and more effi cient ways to 2013, there clearly is more scope to improve manufac- market and distribute their products online. Carmak- turing effi ciency.36 ers represented on the internet in China can attend to The Egyptian logistics fi rm RW uses the online about 10 million searches each day on Baidu, reduc- platform En2ly to optimize supply chains, shortening ing their marketing and sales costs relative to non- the time to deliver goods. The platform ensures more connected manufacturers. Volkswagen, for instance, ACCELERATING GROWTH 65 is selling cars online to Chinese customers through Figure 1.9 Vietnamese fi rms using its own website and on the e-commerce site Tmall.39 e-commerce have higher TFP growth, Other manufacturers in China have been working 2007–12 with search engines to gain deeper insights into customer preferences to incorporate them in product 4 development. Vietnamese fi rms using the internet for 3 e-commerce had 3.6 percentage point higher pro- ductivity growth in subsequent years. The rollout of 2 broadband internet infrastructure in Vietnam (see map 1.1) is positively correlated with fi rm productiv- 1 Percentage points Percentage ity growth. Using the internet increased TFP growth by 1.9 percentage point; fi rms also doing e-commerce 0 Labor TFP growth increased TFP growth by an additional 1.7 percentage productivity effect point (fi gure 1.9). The effect of e-commerce is larger in growth effect sectors that use ICT more intensively, consistent with e-commerce Internet use a causal impact on productivity growth (see box 1.6).40 Source: Nguyen and Schiff bauer 2015 for the 2016 WDR. Data at http://bit Innovation—intensifying competition and .do/WDR2016-Fig1_9. Note: Results are based on a regression of TFP or labor productivity growth creating new business models on the one-year lag of a dummy if the fi rm used the internet and the one- Online services and platforms eliminate search and year lag of a dummy if the fi rm also conducts e-commerce. The regressions control for year fi xed eff ects, province fi xed eff ects, industry fi xed eff ects, communication costs, increasing price transparency fi rm age, fi rm size, foreign ownership, exports status, state ownership, and and lowering the fi xed costs to start a business. the share of workforce with a secondary schooling degree. The internet or e-commerce eff ects are signifi cantly larger in ICT-intensive industries, Lower fi xed and marginal costs enable new startups measured by World KLEMS two-digit sector-level data for Japan or the to exploit scale economies from the beginning, sup- United States; two-digit sector-level data on telecommunication expenses in China; or four-digit sector-level data from a Vietnamese fi rm census, porting their rapid growth. The transaction costs for 2007–12, with more than 300,000 observations each year. ICT = information each new customer is almost zero for some services, and communication technology; TFP = total factor productivity.

Box 1.6 Do digital technologies embed productivity externalities?

Firms’ total factor productivity (TFP) growth in high- infrastructure rollout must be independent from productiv- income countries is often positively associated with ity growth in the diff erent locations. Such a case is arguably information and communication technology (ICT) capi- provided by the limited funding of a public program rolling tal accumulation.a The relation has been stronger in the out broadband internet access points in Norway in the early United States relative to the European Union, particularly in ; the expansion of broadband increased fi rms’ pro- ICT-using sectors, such as retail and wholesale trade, ductivity.f Another study shows that establishments owned fi nance, and other business services.b The few studies for by U.S. multinational companies in the United Kingdom use developing countries show that the correlations between their ICT capital stock more productively (after takeovers) fi rms’ ICT capital stock and TFP growth in Brazil and India relative to domestic fi rms and establishments owned by are comparable to those estimated for high-income coun- multi nationals from other countries. This productivity dif- tries.c A few recent studies focus on the impact of rolling ferential is highest in exactly the same sectors that were out broadband infrastructure on fi rm productivity growth, responsible for the U.S. productivity surge from ICT invest- but the results are ambiguous.d ments in the 2000s, pointing to a causal impact of ICT on fi rm productivity among U.S. fi rms in these sectors.g Knowledge spillovers? Some recent studies suggest a causal impact of fi rms’ use Foreign direct investment spillovers? of digital technologies on their TFP growth.e Identifying a Foreign direct investment (FDI) in Jordan’s ICT sector causal impact requires that the spatial sequencing of ICT did not lead to growth spillovers among domestic fi rms

(Box continues next page) 66 WORLD DEVELOPMENT REPORT 2016

Box 1.6 Do digital technologies embed productivity externalities? (continued)

interacting with the foreign tech companies. Neither is, some knowledge can be more readily adapted for use domestic fi rms supplying goods or services to foreign in related production processes while other knowledge tech companies (backward linkages) nor domestic fi rms is limited in its application. So, products or industries consuming their services (forward linkages) grew as results embodying more applicable technologies also have more of such linkages. The lack of measurable growth spillovers scope for technology externalities. A recent study uses pat- could be due to the relatively small number of foreign ent citation data from the U.S. Patent and Trademark Offi ce multinationals in the ICT sector (8 in ICT-producing sectors, to trace the bilateral direction and intensity of knowledge and 160 in ICT-using sectors). And there are still relatively fl ows between technologies. It fi nds that digital technol- few domestic ICT companies in Jordan, such as software ogies typically are more widely applicable in driving the programmers, which are expected to benefi t most from discovery of new technologies.k linkages with foreign tech companies. Furthermore, foreign fi rms like Microsoft or Oracle might use Jordan primarily as Increasing international technology diff usion? a hub to support regional activities involving few linkages The internet reduces geographic distances, which are to the domestic economy. The lack of spillovers to ICT-using the main predictor of international technology diff usion. sectors in Jordan suggests that Jordanian fi rms import ICT Technology spillovers provide an indirect channel for digital services rather than depend on the presence of foreign tech technologies to boost effi ciency and thus fi rm productivity companies in the country.h growth beyond fi rms’ measured ICT capital stocks. For instance, the internet might facilitate the (international) Producing new products and processes? diff usion of technologies by providing access to relevant Broadband internet increases innovation. A study of information or inducing knowledge spillovers between German fi rms shows that broadband (DSL) infrastructure users. The literature suggests that geographic distances rollout between 2001 and 2003 increased fi rms’ process are barriers to international technology diff usion.l Likewise, and product innovation once broadband internet was avail- several studies fi nd that bilateral trade fl ows embed learn- able within their districts (measured by the postal code).i ing externalities, spurring the diff usion of technologies. A study of six African countries fi nds that more intensive This Report provides evidence that the internet enhances use of computers or the internet enabled more product and trade and reduces distance, suggesting that the internet process innovation in 2014.j facilitates the international diff usion of technologies, even Production technologies diff er in whether they are more though the growth contribution of this channel is diffi cult or less widely applicable across products or industries. That to verify empirically.

a. Cardona, Kretschmer, and Strobel 2013. b. See, for example, Inklaar, Timmer, and van Ark 2008; Schiff bauer, Serafi ni, and Strauch 2011; Bloom, Sadun, and Van Reenen 2012. c. Commander, Harrison, and Menezes-Filho 2011. d. Haller and Lyons (2015); Colombo, Croce, and Grilli (2013); and Bertschek, Cerquerab, and Kleinc (2013) do not fi nd a signifi cant impact of the rollout of broadband infrastructure on fi rm productivity in Ireland, Italy, and Germany, respectively. By contrast, Grimes, Ren, and Stevens (2012) and Akerman, Gaarder, and Mogstad (2015) fi nd a positive impact in New Zealand and Norway, respectively. The results might suggest that moving from basic inter- net access to broadband leads to only marginal productivity improvements. However, it might also be too early to measure the eff ect of high-speed internet in fi rm-level data, since fi rms’ use of the internet is often still relatively basic. e. Most studies attempt to address the endogeneity problem by using lagged ICT or human capital variables as instruments. However, unobserved tran- sitory shocks (or unobservable fi rm-specifi c factors) lead to serial correlation in ICT and productivity measures that make these types of instrumental variables invalid. f. Akerman, Gaarder, and Mogstad 2015. g. Bloom, Sadun, and Van Reenen 2012. h. See Lamla and Schiff bauer 2015. i. See Bertschek, Cerquerab, and Kleinc (2013). The authors analyze the impact of DSL infrastructure rollout in Germany in an early phase, when about 60 percent of German fi rms already used broadband internet. j. Cirera, Lage, and Sabetti 2015. k. Cai and Li 2015. l. See Keller 2002; Keller and Yeaple 2013.

enabling scale effects and increasing the profi ts from and fi nance. Innovations include mobile money, dig- innovation. Such scale effects inspire new business ital marketplaces, price comparator websites, online models based on the internet in services ranging media, and the sharing economy. The substantial from retail trade, transport, and logistics to tourism scale effects for some activities have also led to digital ACCELERATING GROWTH 67 goods trading exclusively online, as for e-books, Figure 1.10 Firm entry rates rose after online search, and streaming music and videos— countries introduced online registration making transport, storage, and distribution obsolete. systems, 2006–12

Intensifying competition 8 At 90th percentile, Price comparator websites enhance transparency in difference = ~2 prices and result in lower and less dispersed prices for 6 consumers. Consider term-life insurance, where aver- age prices have fallen up to 15 percent in the United 4 States after the introduction of comparator sites.41 At 75th percentile, difference = ~1 These websites emerged in 1996 and eliminated the previously high markups. The potential customer 2

Average firm density At median, fi lls out a medical questionnaire online, and the sites difference = ~0.1 report quotes from companies that offer a suitable policy. In almost all cases, the individual does not buy 0 Before After the product directly online but gets connected instead reform reform to the offl ine seller. Comparator sites essentially pro- Sources: World Bank Doing Business database, 2007–12; World Bank Entre- vide an information platform between the consumer preneurship Database, 2006–12. Data at http://bit.do/WDR2016-Fig1_10. and the life insurance company that formerly was Note: The entry densities are based on regression coeffi cients using the available only to brokers. In contrast, the prices of reform years in the Doing Business database. The reform year is either the year when the online business registration was introduced or when whole-life insurance, which were not covered by signifi cant digital measures were undertaken to make the online registry these sites, were not affected—whole-life insurance more eff ective. is a more complex product including built-in saving components, leaving more room for asymmetric boundaries between the online and offl ine econo- information across buyers and sellers; a decline in mies and can help break existing regulatory barriers search costs is thus less relevant. to entry in sectors that are often protected from Online registration systems can lower the cost of competition. entry for new players, increasing competitive pres- Mobile money spurs competition in fi nance. Safari- sure for incumbents. The number of newly registered com, the leading telecom fi rm in Kenya, launched the limited liability fi rms has increased, on average, 56 mobile money service M-Pesa in 2007, allowing users percent after the introduction of online registration to transfer money through a simple text-based menu systems (from 2.7 per 1,000 working-age population in years before the reform to 4.2 in the years after).42 But this positive average impact masks heterogeneity Figure 1.11 Two out of three fi rms report competitive across countries (fi gure 1.10). Thirty-three countries pressure from digital innovations, 2014 introduced online registration systems for fi rms between 2006 and 2012. The entry density declined Established companies somewhat in 8 of the 33 countries. that are using digital Two out of three fi rms experience moderate or technology to enter the severe competition from digital innovations. The market(s) for the first time share of fi rms reporting competitive pressure from Digital offerings from traditional competitors is somewhat higher than the traditional competitors share of fi rms reporting competitive pressure from digital technology startups (fi gure 1.11). Firms in tradi- tional sectors using digital technologies to modernize Startups driven by digital technology their business are thus an important source of com- petition (see also box 1.5).43 0 10 20 30 40 50 60 70 80 90 100 Bringing competition to incumbent fi rms Percent The internet has created new types of startups that Moderate-to-severe competitive pressure base their business model entirely on the web but Limited-to-no competitive pressure offer traditional services such as retail trade, fi nance, Not available transport, logistics, tourism, media, publishing, and Source: Economist 2015a. Data at http://bit.do/WDR2016-Fig1_11. advertising. These new business models dissolve the The data are based on a global survey with 561 respondents. 68 WORLD DEVELOPMENT REPORT 2016

available on the most basic mobile phone. By the end cial sector have been restrictive, with deposit interest of 2013, 17 million Kenyans, or more than two-thirds rates partially liberalized only recently. In contrast, of the adult population, were using the service to pay internet platforms for businesses are often loosely for taxi rides, electricity bills, or daily supermarket regulated, allowing them to offer fi nancial services purchases. M-Pesa also created new opportunities for that bypass regulations in the fi nancial sector.45 Yu’E innovation. Kopo Kopo partnered with Safaricom to Bao, Tencent’s online bank, and WeBank, for instance, offer mobile money services to businesses. One of the offer online banking services in China. They collect key factors underpinning M-Pesa’s success in Kenya deposits from households and lend money using big was the regulator’s decision to permit the scheme to data analytics to assess the creditworthiness of poten- proceed as an experiment without formal approval: tial borrowers. Yu’E Bao had 150 million subscribers similar schemes in other countries have typically been by the end of 2014. These internet fi rms threaten the held up by opposition from banks and regulators. profi tability of traditional banks, increasing com- Firm-level data show that M-Pesa reduces the prices petitive pressure. Similar startups have emerged in of competing services in the fi nancial sector.44 other countries. ZestFinance, LendUp, and Wonga Internet startups offering fi nancing based on “big currently target unbanked subprime customers in data” analysis to target unbanked customers increase Canada, Europe, South Africa, and the United States. the competitive pressure for traditional banking ser- German startup Kreditech and Hong Kong SAR, vices. For example, in China, regulations in the fi nan- China, startup Lenddo use information from social media profi les and networks to determine the cred- Figure 1.12 Prices of taxi medallions have started to itworthiness of potential customers. Lenddo targets decline following the entry of on-demand services the thriving middle class in developing countries, and reduced demand for traditional taxis which often lacks access to local fi nancial services. Ghanaian startup Cignifi uses borrowers’ mobile a. Sales price of New York City taxi medallion, 2002–15 phone records to judge their creditworthiness. 1,200,000 Online marketplaces exert competitive pressure 1,100,000 Uber started on brick-and-mortar retailers to innovate. The most 1,000,000 operation in the U.S. prominent are online retailers and wholesalers such 900,000 as Amazon and eBay in the United States, Alibaba in 800,000 700,000 China, Flipkart and Snapdeal in India, Ozon in Rus- 600,000 sia, Jumia and Konga in Nigeria, and Takealot and U.S. dollars U.S. 500,000 Kalahari in South Africa. In the United States, online 400,000 sales accounted for 25 percent of all retail sales for 300,000 computers, electronics, and appliances in 2013. The 200,000 competition from online retailers has led many large traditional retailers in the United States to combine 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 digital and physical consumer experiences by offer- ing online orders and in-store pickup services to b. Average monthly number of trips per traditional taxi improve their effi ciency. And some online retailers in , January 2012–August 2014 such as Amazon started to open brick-and-mortar 1,600 stores to enjoy potential (marketing) advantages of 1,400 traditional retailers. 1,200 Ride sharing services enhance competition and 1,000 the effi ciency of transport services. Many cities 800 around the world require drivers to hold a license 600 to operate a taxi; entering the market is often asso-

Number of trips 400 ciated with high fi xed costs since licenses are issued rarely and must be bought from current owners. 200 The licensing system, together with growing urban 0 populations, has led prices for taxi medallions to sky- rocket in large cities worldwide; in New York City, the

July 2012 July 2013 July 2014 Jan. 2012Mar. 2012May 2012 Sept. 2012Nov. 2012Jan. 2013Mar. 2013May 2013 Sept. 2013Nov. 2013Jan. 2014Mar. 2014May 2014 Sept. 2014 cost of a single-taxi medallion increased from about

Sources: Wei and Mozur 2014; Metropolitan Transportation Agency (New York), http://www.mta.info/; US$400,000 in 2004 to more than US$1,100,000 in the Golovin 2014. Data at http://bit.do/WDR2016-Fig1_12a. beginning of 2013 (fi gure 1.12, panel a). But the new ACCELERATING GROWTH 69 competition from ride sharing services reversed the business models. Airbnb operated in more than 40 trend. In San Francisco, the home city of these com- countries in 2014, enabling owners to let their homes panies, taxi use fell 65 percent (fi gure 1.12, panel b). for short-term rents. This puts competitive pressure on After decades of steady increases, single-taxi medal- the hotel and tourism industry, which has frequently lion prices in New York City started to slump, falling enjoyed high rents due to local market segmentation to about US$800,000 by the end of 2014 (the trend is or exclusive contracts in developing countries. The similar in other cities).46 The new competition also Estonian startup TransferWise and the U.S. startup forces taxi companies around the world to reduce Xoom match requests for international currency their prices or improve their services. Taxi companies transfers online, saving direct and indirect trans- in various cities have begun to develop joint smart- action fees by clearing reciprocal currency transfer phone applications enabling online payment, rating, requests. The startups reduce regulatory rents by and vehicle tracking in real time to compete with the reducing the prices of international currency transfers ride sharing services. The trend is not confi ned to by up to 90 percent. Postmates and Parcel provide high-income countries. While Uber operates in more local logistics services in U.S. urban centers and are than 57 countries and 300 cities, regional competitors starting to compete with traditional service provid- have emerged, such as Lyft and Sidecar in the United ers such as Federal Express, but also with existing States, Hailo and BlaBlaCar in Europe, Kuaidi-Didi e-commerce platforms by matching customers Taxi in China, Olacabs in India, and Easy Taxi in demanding any type of locally available goods with Nigeria. a pool of couriers. And Kenyan startup Sendy pro- Digital innovations increase welfare in many ways vides a platform accessible by simple mobile phones that are not necessarily captured by GDP statistics. connecting customers with motorcycle couriers to These welfare gains are even more diffi cult to quan- offer delivery services payable with mobile money.47 tify, but a growing literature is shedding more light to Upwork (formerly Elance-oDesk) matches fi rms in these “unmeasured benefi ts” of the internet (box 1.7). high-income countries with freelancers offering Internet platforms also reduce the entry costs professional service tasks in developing countries. in other service sectors contesting conventional The Ugandan internet platform eKeebo connects

Box 1.7 Much of the benefi t from the internet is unmeasured

The internet off ers many benefi ts to individuals that are not the prices in brick-and-mortar bookstores. It infers that the captured in the gross domestic product (GDP) statistics. buyers and sellers are better matched online.a This stylized Countries compute GDP based on activities measured in fact is an example of the “Long Tail” argument, which notes monetary terms and exclude activities that do not generate that online stores provide a larger variety of products and monetary transactions. But many online activities generate can sell niche products to more consumers.b The same study substantial benefi ts for the individual, such as time saved, also estimates that the consumer surplus is higher when con- consumer convenience, expanded choice, better quality sumers shift to purchasing used books online.c A study of the leisure time, and access to more knowledge. These benefi ts music industry fi nds that variety and diversity have improved can be understood as the consumer surplus: the diff erence since 2000, mostly because independent labels and musi- between the price individuals are willing to pay and the cians can operate online and release their music digitally.d actual price for the product or service, which is often free Another way to calculate the consumer surplus from the on the internet. internet is to directly measure individuals’ willingness to pay Economists are developing new techniques and col- for the internet (online products and services are often free lecting new data to present a more accurate estimate of and funded through advertisements). A McKinsey survey the consumer surplus from the internet. One method is to of consumers in France, Germany, the Russian Federation, analyze the price and quality outcomes of using the internet Spain, the United Kingdom, and the United States in 2010 in a sector. A recent study compares the online and offl ine found that an average household would be willing to pay prices of used books, fi nding that online prices are lower than approximately €38 a month for internet services that it

(Box continues next page) 70 WORLD DEVELOPMENT REPORT 2016

Box 1.7 Much of the benefi t from the internet is unmeasured (continued)

currently receives for free.e Similarly, an examination of procrastinate and distract oneself with social networking time-use data estimates that the median individual gains sites and cute animal videos. The aggregate eff ects of the more than US$3,000 annually from the internet.f internet use of individuals are diffi cult to determine—an A study comparing the outcomes of online searches to interesting avenue for further research. offl ine searches in a library fi nds that the average online But all this is not new. The initial gains from past techno- search tends to be faster by 15 minutes, the results are more logical change were also undercounted in GDP. For exam- accurate and relevant, and the experience is more enjoy- ple, the cost savings from railroads made up about three- able than the offl ine search.g Also attempting to measure fourths of the total gains in GDP in the mid-19th century; by consumer surplus, another study fi nds that users of Google the early 20th century, this had fallen to about one-fourth. obtain a surplus of US$500 a year, on average, or a total of Time saving became by far the more important benefi t as US$150 billion for a 300-million user base.h commute times shortened and the leisure time of working- The increase in time saved and in quality—and thus class customers increased. Similarly, U.S. the productivity of work from using the internet—can did not refl ect the output from automobiles for nearly 15 indirectly increase GDP statistics. Conversely, the internet years after the Ford Model T, the fi rst mass-produced car, has also reduced productivity by providing an easy way to was available.

Source: Crafts 2015. a. Ellison and Fisher Ellison 2014. b. See Anderson 2006. c. Ellison and Fisher Ellison 2014. d. Waldfogel 2013. e. McKinsey Global Institute (Manyika and others 2011). f. Goolsbee and Klenow 2006. g. Chen, Jeon, and Kim 2014. h. Varian 2011.

individuals who want an authentic home-cooked meal Digital technologies can with independent or amateur chefs, circumventing restaurant licensing, which can be a source of rent- lead fi rms and countries to seeking in developing countries. diverge Cloud computing can reduce the costs of entry in developing countries, implying substantial oppor- Divergence—with benefi ts short of tunities for innovation and competition and thus expectations future growth.48 It provides computing infrastructure Despite the opportunities, fi rms’ use of digital tech- (processing, memory, and storage of data), platform nologies differs substantially across sectors and coun- applications, and software services for fi rms with tries; the differences prevail when comparing the ICT access to the internet. Firms can use these services, intensities of the same sectors across countries with for a fee, without investing in the underlying hard- a similar GDP per capita. The share of fi rms that used ware or software infrastructure. Cloud computing the internet for banking in 2012, for example, was has signifi cantly reduced the fi xed costs of starting a below 20 percent in several middle-income countries, business in the last decade. Startups can use the latest but more than 80 percent in others (fi gure 1.13). computing infrastructure, video conferencing ser- This divergence in the intensity of internet use vices, or online payment systems at much lower cost. for fi rms in the same sector is confi rmed by more The reduction in entry costs has so far materialized detailed survey data from six African countries in mainly in high-income countries. Although local pro- 2014. The share of manufacturing and service fi rms viders have emerged in developing regions, such as with at least fi ve employees using the internet ranged Angani in Kenya and Data Park in Oman, these have from 22 percent in Tanzania to 73 percent in Kenya. not yet achieved suffi cient scale. Of manufacturing fi rms in Kenya, 41 percent used ACCELERATING GROWTH 71 it to manage their inventories, compared with 27 Figure 1.13 Firms’ use of online banking varies percent in Zambia and only 6 percent in Uganda substantially across countries at comparable incomes, (fi gure 1.14). Of service fi rms in Kenya, 41 percent used 2003–06 and 2008–13 the internet to manage their inventories, compared with 15 percent in Zambia, 12 percent in Uganda, and 100 ISL EST only 8 percent in the Democratic Republic of Congo BLR SVN FIN ISL DNK HRV CHE ESP BEL SWE and Tanzania. The divergence in the shares of man- 80 BRA AUT BRA CZE AUS ufacturing or service fi rms that sell goods online or UKR LTU ITA IRL URY DEU NOR ARG MLT GRC NLD use the internet for marketing was comparably large TUR LUX 60 SVK GBR POL PRT 49 LBN across these six countries. ARE SGP The share of retail fi rms that sell their products HUN CYP 40 online varies substantially across Latin American LVA SGP KAZ BGR countries with a similar GDP per capita. In Bolivia, URY AZE 52 percent of all fi rms with at least fi ve employees 20 MKD KOR PSE ROU in the retail sector sold their products online in 2010 EGY OMN % of firms doing internet banking RUS AZE THA (fi gure 1.15). Peru had a slightly higher GDP per 0 THA capita in 2010, but only 14 percent of retail fi rms with 0 20,000 40,000 60,000 80,000 at least fi ve employees sold online. Only 18 percent of retail fi rms in Brazil sold online in 2010, and only GDP per capita (constant 2005 US$) 27 percent in Panama. In Mexico and Uruguay, about Circa 2004 Circa 2012 half of all retail fi rms sold online; in Argentina 62 Source: UNCTAD, http://unctad.org/en/Pages/Statistics.aspx. Data at http://bit.do/WDR2016-Fig1_13. percent. Variations in the share of retail fi rms that do Note: The fi gure shows all countries with available data for the latest year with available data from e-commerce are comparably large in the other devel- 2003–06 and 2008–13. oping regions. The share of fi rms that use connected customer relationship management (CRM) platforms to facili- 1.16). In Austria, 60 percent of retail and wholesale tate sales, customer support, and related interactions fi rms use integrated CRM systems, but only 28 with customers or other businesses varies substan- percent do in the United Kingdom. Half of all man- tially across sectors and countries in Europe (fi gure ufacturing fi rms in Germany use these systems,

Figure 1.14 Firms’ use of the internet varies among six African countries, 2014

a. Manufacturing b. Services 45 50

40 45

35 40 35 30 30 25 25

Percent 20 Percent 20 15 15

10 10

5 5

0 0 Managing Selling online Marketing Managing Selling online Marketing inventory inventory

Tanzania Uganda Congo, Dem. Rep. Zambia Ghana Kenya

Source: Cirera, Lage, and Sabetti 2015. Data at http://bit.do/WDR2016-Fig1_14. Note: The fi gures show the shares of fi rms in the manufacturing and services sectors that use the internet to manage their inventory, sell their goods or services, and do marketing. The results are based on 2,843 fi rms (1,458 manufacturing and 1,385 service fi rms) in these six African countries in 2014. 72 WORLD DEVELOPMENT REPORT 2016

Figure 1.15 The share of fi rms in the retail sector that in developing countries contributes to cross-country sell their products online varies substantially among divergence in incomes. Latin American countries, 2010 Why are the differences so large? In most coun- tries, the physical infrastructure barriers to internet 80 CHL use are fairly small for most fi rms in urban areas.

COL Differences in the affordability of the internet across ECU 60 ARG countries have been an important factor (see chapter SLV BOL GTM 4). But equally important are other structural barri- URY MEX JAM ers that limit fi rms’ ability or incentives to use the 40 CRI internet more intensively. That three-fourths of all DOM NIC retail fi rms in Panama do not conduct e-commerce, PAN VEN for instance, might be related to a regulatory barrier 20 HND BRA PER PRY protecting domestic retailers: foreign-owned fi rms % of retailers selling online are not allowed to operate in Panama’s retail sector, 0 reducing competitive pressure among domestic

5,000 10,000 15,000 20,000 fi rms to increase their effi ciency (see fi gure 1.15). By contrast, there are no restrictions to foreign entry GDP per capita (PPP) into the retail sectors in Argentina, Mexico, and Sell online Fitted values Uruguay—where retailers use e-commerce more Source: World Bank Enterprise Surveys data 2010. Data at http://bit.do/WDR2016-Fig1_15. extensively. Note: GDP = gross domestic product; PPP = purchasing power parity. Market power—scale and network eff ects compared to only 32 percent in Norway, 20 percent can lead to anticompetitive behavior in Poland, 18 percent in Estonia, and 12 percent in The digital economy can be highly concentrated. Croatia. In Finland, 25 percent of professional service Facebook is the leading social network. Amazon sold fi rms sourced CRM software from the cloud in 2014, 40 percent of all e-books in 2014. Google is the dom- but only 5 percent did in France or Lithuania. inant search engine in most countries: it accounted These fi ndings are consistent with previous stud- for about 25 percent of the display-ad market, while ies showing that digital technologies diffuse much Yahoo and Facebook accounted for about 10 percent more slowly within developing countries than pre- each.51 And while the sharing economy is increasing vious major technologies (for example, electricity or the number of market players in various services, Air- the steam engine).50 This lack of internet penetration bnb and Uber are its major platforms so far.

Figure 1.16 The share of fi rms using integrated customer relationship management platforms varies substantially among sectors and countries in Europe, 2014

70 60 50 40 30 Percent 20 10 0

Italy Latvia France Malta Spain GreeceCroatia Iceland EstoniaPoland IrelandCyprus SwedenFinland Austria Hungary Romania Bulgaria Slovenia PortugalLithuania Denmark Norway Germany Luxembourg Netherlands Czech Republic Slovak Republic United Kingdom Manufacturing Wholesale and retail trade Professional services

Source: Eurostat (EC, various years). Data at http://bit.do/WDR2016-Fig1_16. Note: The fi gure shows the shares of European fi rms in manufacturing, wholesale and retail, and professional services that used integrated customer relationship management software in 2014. Customer relationship management systems allow fi rms to track, record, and store data to facilitate sales, customer support, and related interactions with customers or with other businesses. ACCELERATING GROWTH 73

Domestic mobile money markets are often con- Figure 1.17 Mobile money markets are often trolled by one or two operators. Safaricom, creator of controlled by one or two operators, 2014 the mobile money service M-Pesa, controls more than 100 two-thirds of the mobile money market in Kenya (fi g- 90 ure 1.17). The mobile money market in the Philippines 80 is a , with the largest operator controlling 77 70 percent of the market. 60 What matters for antitrust regulators, however, 50

is not the concentration of a sector per se, but the Percent 40 entry costs for new entrants or switching costs for 30 52 consumers. The concentration of various markets 20 in the digital economy—often two-sided markets in 10 which platforms match service providers or sellers 0 with users or buyers53—is not surprising. When the Kenya transaction costs to serve additional customers are Ghana Nigeria Sri Lanka Pakistan Uganda Zimbabwe IndonesiaPhilippines close to zero, the most innovative fi rm—say, the one Côte d’Ivoire with the best search algorithm or online platform— Congo, Dem. Rep. will be frequented by the most customers. As long as Market share of second largest provider the fi xed entry costs are low, the industry remains Market share of largest provider dynamic, in the sense that entrepreneurs program- Source: Evans 2015. Data at http://bit.do/WDR2016-Fig1_17. ming more effi cient search engines or applying superior business models will be able to enter and New digital solutions can also raise fi xed costs and disrupt the existing incumbent (through creative thus lower competition in sectors such as manufac- destruction). AltaVista, for instance, was once the turing due to high up-front costs of software, data most popular search engine, but it lost ground in 2001 storage, analytics, and security. to new market entrant Google, which provided more Safeguarding a level playing fi eld for business can effi cient services. thus also involve overhauling the regulatory regimes Network effects or anticompetitive behavior can, in the digital economy, including sectors in which however, create barriers for startups. For instance, the technology is dissolving the boundaries between more people who use Facebook, the higher the value online and offl ine fi rms. The initial entry of internet for users to create content on its website. Similar net- fi rms into these sectors can disrupt monopolies, but work effects (or switching costs) might exist for other regulators need to level the regulatory regime and online services. Google has been accused of using its prevent market shares from becoming excessively dominant position in the market for online display concentrated. advertising to curb competition and push companies toward its other services. Several Google competitors, including Microsoft and Yelp, had argued that Google The nexus of technology and unfairly demoted rivals in its search engine results to regulation direct users toward its services. Similarly, Amazon has been accused of abusing its market power by delaying The potential impact of the internet on fi rm pro- the shipping times of products from publishers that ductivity is highest for activities where the internet did not agree with its pricing mechanism. can achieve large-scale effects and where contracts The room for anticompetitive behavior varies are easier to enforce—and thus can be automated. among digital products and services. Network effects Economic activities can be classifi ed by the degree of and switching costs appear to be large for search their amenability to the internet, based on: engines, allowing them to create barriers to scale for new fi rms offering this service. By contrast, program- • Contract complexity. Some goods and services ming ride sharing platforms can be imitated fairly are contract-intensive, because they embody a easily, and drivers can use more than one platform at higher share of intermediate inputs that require the same time in the absence of exclusivity contracts. relationship-specifi c investments that are more Indeed, multiple ride sharing platforms competing diffi cult to enforce by written contracts.54 The less with each other have emerged. And traditional taxi complex that goods or services are, the easier it is companies have also started to use the technology. to enforce contracts and use internet platforms to 74 WORLD DEVELOPMENT REPORT 2016

Table 1.1 The internet impact is highest for data-intensive activities that involve easy-to-enforce contracts Data-intensive activities More scalable Less scalable Less complex products (contracts easier Retail and wholesale trade, transport, Legal services to enforce) insurance, banking

More complex products (contracts more Agriculture, education, health care, Construction diffi cult to enforce) hotels and restaurants, manufacturing, real estate, utilities Source: WDR 2016 team. Note: The approximate grouping of sectors is based on the literature; among others, on Bloom, Sadun, and Van Reenen (2012).

market these products and match buyers and sellers more effectively. Two such barriers for fi rms are the (table 1.1).55 skills of the workforce and the ability to reorganize • Scale eff ects. The impact of the internet is larger if the management processes to make better use of the internet can help achieve scale effects. For instance, effi ciency gains that digital technologies can provide. longer value chains require more intensive real- There is even evidence that investing in ICTs without time communication or sharing of large amounts business process reorganization can reduce fi rms’ of data between different stages of production or productivity growth.57 with suppliers and clients. The internet reduces The correlation between fi rm productivity and these transaction costs, enabling scale effects and ICT capital stocks among fi rms in India and Brazil increasing management effi ciency.56 increases signifi cantly after fi rms have reorganized their production structures or hired more skilled Firms using digital technologies need to labor. Firms in Vietnam with a more educated work- invest in skills and reorganization force or a higher share of managers relative to total As for any new technology, integrating digital solu- staff had a stronger correlation between fi rms’ use of tions into fi rms’ business models can be risky and lead the internet (such as selling products online) and TFP to costly failures. Digital technologies have often been growth in subsequent years.58 called on to help meet supply chain needs, but not all The majority of fi rms in Eastern Europe and Cen- have brought the expected widespread benefi ts. Five tral Asia that use broadband internet have not reor- or ten years ago, radio frequency identifi cation tags ganized their businesses, pointing to potential ineffi - for tracking individual goods were considered the ciencies in the way it is deployed. Only one-fourth of major source for effi ciency gains. U.S. retailer Walmart fi rms using ICTs have adjusted their organizational required its biggest suppliers to use the tags, but their structures or management practices (table 1.2). use has not taken off in the extended supply chain. The differences in these complementary invest- The successful use of digital technologies depends ments help explain the heterogeneity in the use and on fi rms’ complementary investments in skills and impact of digital technologies. For instance, fi rms’ organizational restructuring. That more productive investments in computerized information (digital fi rms use digital technologies (more intensively) technologies) are comparable among fi rms in the suggests that there are barriers to fi rms using them United States and Brazil (fi gure 1.18). But U.S. fi rms

Table 1.2 Many fi rms use the internet without changing their organizational structures, limiting its impact, 2010–14 percentage of fi rms

Firms that use Firms that use Firms that have Did your fi rm introduce new or broadband e-mail a website improved organizational structures or management practices? YesNoYesNoYesNo Yes 23 10 22 7 26 12 No 77 90 78 93 74 88 Source: Hussain (2015) based on World Bank Enterprise Surveys, various years. ACCELERATING GROWTH 75 invest substantially more in training and develop- Figure 1.18 Firms’ ICT investments as a share of GDP ment (skills) and in business process improvements in several countries are comparable to those of U.S. (reorganization), explaining why the impact of dig- fi rms, but they invest much less in complementary ital technologies on fi rm productivity is higher for skills and reorganization, 2006 U.S. fi rms. Similarly, fi rms in Japan and China had the highest investments in digital technologies, but 6 they invested less in skills than France, Germany, the 5 United Kingdom, and the United States. Complementary skills and reorganizations are 4 more important for more advanced digital technol- ogies, potentially explaining the slow diffusion of 3

some of these technologies across fi rms (see fi gure % of GDP 2 1.4). Implementing customer relationship or supply chain management software and integrating it with 1 ICT systems of customers and suppliers requires a well-trained workforce and the adoption of new orga- 0 Italy nizational processes incorporating real-time informa- Brazil Spain India Japan China France tion fl ows. Germany United States Countries need to invest in skills and United Kingdom Process improvements logistics to enable fi rms to use ICT more Training and development eff ectively Computerized information

Since a shortage of skills can lower aggregate pro- Sources: Corrado and others 2013; Hulten and Hao 2011; Hao, Hulten, and Jaeger 2013; Dutz and others ductivity growth despite investments in broadband 2012. Data at http://bit.do/WDR2016-Fig1_18. infrastructure, countries need to invest in education Note: The countries are ordered by their investments in computerized information (digital technology) as a share of GDP. to complement investments in new digital technolo- gies (chapter 2). They also need to invest in infrastruc- ture such as electricity, trade logistics, and payment systems. Asian countries (map 1.3). Consumers and fi rms need Online marketplaces for nondigital goods enable a way to pay for online transactions. e-commerce scale effects if trade infrastructure is adequate. After relies on the ability of fi rms and consumers to pay an online transaction has been completed, the timely for their products online. Credit cards are accepted delivery of (physical) goods requires offl ine trade on most online platforms, but many individuals do infrastructure.59 Countries thus need ports with not have credit cards: less than 10 percent of the pop- enough capacity to connect domestic fi rms to inter- ulation in developing countries had a credit card in national markets and effi cient domestic logistics 2012.60 Online payment systems providing business infrastructure to complete the “last mile” delivery. accounts, such as PayPal, MercadoPago, and PayU, Online retailers in some countries still struggle are alternatives. For instance, MercadoPago provides to deliver their parcels to domestic customers. In 38 online payment services to small businesses that do countries in 2012—27 of them in Africa—less than 50 not have bank accounts. percent of the population had access to home postal The low share of fi rms using e-commerce in delivery (fi gure 1.19). some developing countries can also originate from Faced with such diffi culties, many e-commerce network effects in internet use. Network effects are companies have come up with their own solutions. present because the value of the internet to a fi rm Alibaba has partnered with logistics fi rms in China to depends on the number of internet users (creating create a network for delivering parcels to Chinese cit- online content). Thus low rates of domestic internet ies within a day. Flipkart has partnered with the Dab- use by individuals, fi rms, and government reduce bawalas, the deliverymen famous for their reliable the benefi ts of e-commerce. But the argument is only delivery of hot lunches, to deliver parcels in India. relevant for fi rms that target the domestic market in Russian online retailer Ozon created its own logistics developing countries with few internet users; export- service company to increase the speed of deliveries. ing fi rms targeting world markets face over 3.5 billion Online payment services are available to fi rms in potential online customers. The rapid increase in the many countries but not in most African and Central number of internet users worldwide in recent years, 76 WORLD DEVELOPMENT REPORT 2016

Figure 1.19 Many countries still have poor postal delivery systems Percentage of population receiving home postal delivery in 2012 50 45 40 35 30 25

Percent 20 15 10 5 0

Togo Benin Haiti Kenya LiberiaNigerOman SudanBolivia Gabon Rwanda MalawiBurundiLesotho Guinea Senegal JordanAngola Ghana Nigeria Tanzania Uganda PanamaVietnam Lao PDR Jamaica Botswana Swaziland Mongolia NicaraguaZimbabwe Burkina Faso Mozambique

Congo, Dem. Rep.

Central African Republic Sources: Universal Postal Union (http://www.upu.int/en.html); UNCTAD 2015. Data at http://bit.do/WDR2016-Fig1_19. Note: The fi gure shows the 36 of 130 countries where less than 50 percent of the population receives mail delivery at home. Qatar and the United Arab Emirates have been excluded because they have a comprehensive public postal delivery system even though they usually do not deliver mail to homes but instead to local post offi ces.

Map 1.3 International online payment systems for businesses are unavailable in many parts of Africa and Central Asia, 2012–14

Major online payment systems None available PayPal PayU MercadoPago Two or more available No data available

IBRD 41768 Sources: World Bank staff calculations; UNCTAD 2015. Data at http://bit.do/WDR2016-Map1_3. ACCELERATING GROWTH 77 together with the lower transaction costs in online Consistent with the Schumpeterian mechanism, marketplaces should, in turn, encourage more fi rms fi rms in more contested industries use digital tech- (in developing countries) to use the internet to export nologies (more effectively). The allocation and selec- to foreign markets in coming years. tion mechanism in more contested markets drives fi rms to invest in new digital technologies.65 Compe- Countries need policies to encourage tition from China has induced the adoption of new competition technologies and ICTs in OECD countries, accounting Domestic or foreign barriers to entry and competition for 15 percent of their technology investment between reduce fi rms’ incentives to invest in digital technol- 2000 and 2007 and contributing to their productivity ogies or complementary skills and reorganization. growth.66 Arguably, the entry and innovations of Tao- Without competitive pressure, private fi rms lack bao, Alibaba’s consumer-to-consumer (C2C) platform, incentives to invest in costly or risky new technol- were at least in part encouraged by foreign compe- ogies. The complementary factors at the fi rm level tition from eBay. Similarly, fi rms in more contested correspond to competition policies at the country sectors in Vietnam, which have higher entry and exit level. Firms are more likely to invest in skills or orga- rates, are more likely to invest in broadband internet. nizational restructuring if they are subject to product And younger fi rms in Mexico use e-commerce more market competition.61 intensively: the share of online sales in total sales is European manufacturers are investing in machines almost twice as high in young fi rms (less than fi ve that communicate with one another (internet of years old) than in older fi rms.67 In France, automo- things) to escape the increasing competition from bile dealers enhanced their investments in ICT, such low-cost Asian producers. For instance, BuS Elektronik as human resource software and other innovations, invested in information technology to specialize in when facing more intense product market competi- custom-designed electronics components produced tion after the of the automobile distri- in lot sizes that are too small to spur the entry of Asian bution system in the European Union in 2002.68 electronics fi rms. The Daimler Group’s Smart auto- In high-income countries, the youngest fi rms use mobile plant uses modern information technology the internet more intensively, while in low-income to integrate assembly and production lines among countries, the oldest fi rms do (fi gure 1.20). In low- a network of seven large suppliers. The interlinked income countries, however, startups seem to face production system allows Daimler to optimize its barriers to more intensive internet use, which the old- value chain and to customize its products, with 10,000 est fi rms (typically also larger) tend to overcome. In variations of its vehicle cockpits. Low-cost factories in middle-income countries, the results are mixed. emerging countries cannot easily copy this strategy, Domestic fi rms in developing countries use the which requires proprietary data.62 internet more intensively when they face pressure The mechanism is illustrated in Schumpeterian from foreign competition. The probability that growth models predicting that each fi rm will invest domestic fi rms (with at least fi ve employees) facing in technologies new to the fi rm to reduce its costs foreign competition use broadband internet increases and escape competition—albeit only temporarily. by 41 percent; that they sell their products online, by But these incentives are attenuated when leading 29 percent; and that they purchase inputs online, by fi rms in the sector have cost advantages that trailing 36 percent (fi gure 1.21). The increase in internet use fi rms cannot overcome. The market leaders have little due to foreign competition is independent of fi rms’ incentive to invest in technologies new to the fi rm initial productivity. since they do not face competitive pressures to reduce Among six African countries in 2014, for instance, their costs; the laggard fi rms are too far away from the increased competition incentivized fi rms to use the frontier to bridge the cost gap and instead use vintage internet more intensively, leading to more product production technologies, focusing on local market and process innovation. Firms that had experienced niches to survive.63 Thus fast-growing sectors with a more competitive pressure, leading to a decline in high share of fi rms using new technologies exhibit their market shares over the past three years, invested specifi c fi rm dynamics, echoing high fi rm churning in more intensive uses of the internet. For example, and neck-and-neck competition market structures, a 10-percent reduction in the market share increased forcing fi rms to enhance their effi ciency by investing the probability that these fi rms used the internet to in more productive technologies. The positive links market their products, on average, by 11 percent, and among competition, technology adoption, and pro- the probability that they used the internet to manage ductivity growth are well documented.64 their inventory by 8 percent.69 Domestic fi rms in 78 WORLD DEVELOPMENT REPORT 2016

Figure 1.20 Young fi rms use the internet more intensively in high-income countries—old fi rms, in low-income countries, 2010–14

a. High-income OECD countries b. Low-income countries

50 18 Youngest Oldest Oldest 45 16 40 14 Youngest 35 12 30 10 25

Percent 8 20 Percent 6 15

10 4

5 2

0 0 Online Online Online Online Online Online buying selling research buying selling research Years in operation <5 5–9 10–19 20+ Source: Hussain (2015) based on World Bank Enterprise Surveys, various years. Data at http://bit.do/WDR2016-Fig1_20. Note: OECD = Organisation for Economic Co-operation and Development.

Figure 1.21 Domestic fi rms use the Kenya and Uganda used the internet more intensively internet more intensively when they in sectors with increasing competition from foreign face foreign competition, 2010–14 direct investment (FDI). The (more intensive) internet use due to higher competition had led to more prod- 45 uct and process innovation among fi rms in these six 40 African countries. 35 Manufacturing fi rms in Mexico are more likely to invest in digital technologies and use them more 30 productively when they sell products that are directly 25 competing with imports from China. Firms that 20 faced this external shock of higher foreign competi- Percent

15 tion between 2000 and 2008, either in the domestic or the U.S. (export) market, increased their number of 10 computers per employee, their share of labor using 5 the internet, and their share of online purchases 0 in total purchases in the subsequent four years, 2008–12 (see fi gure 1.22). As a result, the share of e-mail Client labor using the internet in 2012 was 11 percent higher Broadband Sell online Buy online for fi rms that faced more Chinese competition—and the share of online purchases was 114 percent higher. communications The more intensive use of digital technologies due Source: Hussain (2015) based on World Bank Enterprise Surveys data, 2010–14. Data at http://bit.do/WDR2016-Fig1_21. to Chinese competition translated into productivity Note: The bar chart shows the coeffi cients of six fi rm-level cross-section growth among fi rms in Mexico. By contrast, ICT use regressions of dummy variables for internet use on a dummy if the fi rm had no impact on labor productivity growth among reports pressure from foreign competition. All coeffi cients are signifi cant at the 5 percent level. Each regression controls for fi rm age, fi rm size, Mexican fi rms that did not face import competition foreign ownership, and sector- and country-specifi c eff ects. The data from China.70 include all fi rms with at least fi ve employees in more than 80 developing countries. The number of fi rms in the diff erent regressions varies from Manufacturing fi rms confronted with an increase 3,400 to 8,200. in competition in Brazil are more likely to implement ACCELERATING GROWTH 79

Figure 1.22 Firms in Mexico facing higher import competition from China use more ICTs more productively percentage-point change, 2008–12

a. Firms’ use of ICTs b. Percentage change in productivity associated with 1 standard deviation change in ICT measure with 1 standard deviation change in ICT measure 7 6

5 Computers 9.6 27.2 4 per worker 3 2 6.0 Percent Share of labor 18.0 1 using the internet 0 –1 e-commerce –9.9 10.2 –2 purchases/ total purchases –3 Share of labor Change in number Share of e-commerce using the internet of computers purchases to Low High per worker total purchases competition competition

Low Chinese competition High Chinese competition Not statistically significant

Source: Iacovone, Pereira-Lopez, and Schiff bauer 2015. Data at http://bit.do/WDR2016-Fig1_22. Note: Chinese competition is measured as the change in the share of China in Mexico’s imports in the period 2000–08. The two groups depicted in these fi gures are defi ned as below the median (low competition) and above the median (high competition). The Mexico ICT survey, ENTIC, represents over 52,000 manufacturing and services fi rms with at least 10 employees. ICT = information and communication technology. e-commerce systems. They are also more likely to for manufacturing fi rms: higher nontariff barriers to move to more complete e-commerce systems—pro- trade are associated with lower ICT use in manufac- viding an online order and integrating payment turing (fi gure 1.23, panel b).72 system on their website—when they face more com- Firms’ internet use varies with barriers to competi- petitive pressure.71 tion across sectors in the same country. The Philippine Consistent with these fi ndings, aggregate sector retail sector has substantial restrictions to domestic and country data suggest a negative correlation and foreign entry and is dominated by a few incum- between regulatory barriers to product market com- bent fi rms, while few fi rms use ICTs. Foreign retailers petition and fi rms’ investments in digital technolo- that aim to establish a commercial presence need to gies. More restrictive product market regulations on pass prequalifi cation procedures, meet minimum cap- fi rm entry in service sectors are associated with lower ital requirements, meet limitations to foreign equity ICT use (fi gure 1.23, panel a). The negative relation participation, and have the majority of the board of prevails when comparing restrictions to domestic or directors be Filipinos. Only about 20 percent of retail foreign entry in individual service sectors with fi rms’ fi rms (with at least fi ve employees) sell online in the internet use in these sectors across countries. Pro- Philippines.73 By contrast, the Philippine business pro- fessional service fi rms in Europe—for architecture, cess outsourcing sector is characterized by high entry design, consulting, legal, and accounting—are less rates and few regulatory barriers to competition. It likely to sell their services online in countries that is intensive in ICT-related services such as software have higher barriers to entry in these sectors. Trans- development, animation, contact centers, and tran- port service fi rms are less likely to purchase cloud scription. These ICT-specifi c services experienced computing services, such as CRM software, over high productivity growth in recent years and provided the internet in European countries that have higher about 1.2 million jobs in 2015. Similarly, Nigeria’s retail regulatory barriers for foreign fi rms to enter. Domes- sector, Indonesia’s banking sector, and India’s and tic retail fi rms are less likely to use the internet for Bulgaria’s ICT sectors are more internet intensive and online sales if they operate in countries with higher have fewer regulatory barriers to competition in oth- entry barriers for foreign retailers. The same is true erwise diffi cult business environments. 80 WORLD DEVELOPMENT REPORT 2016

Figure 1.23 Restrictive product market regulations in services and higher nontariff technical barriers to trade in manufacturing are associated with lower ICT use, 2010–14

a. Product market regulations in services b. Frequency of nontariff barriers in manufacturing IND NAM HND 3.5 11 GIN CHN NER MLI CPV 3.0 BRA LAO NIC TUR LKA GHA CRI 10 NPL SLVDOM BEN ISR 2.5 PAK TGO BFA MEX ZAF MLT ESP CYP ROU MWI PER LVACHL ISL COL RUS GRC COL IRL 9 2.0 HRV CZE KOR TUR SVN BEL CIV MDG RWA BGR AUS SWE NOR HUN POL FRADEU JPN NGA GTM LTU LUX TZA MUS VEN ARG RUS FINEST SEN GBR CHE of NTM frequency Log 1.5 PRY PER CRI BOLCHL Barriers to entry, 2013 CAN PRT DNK AUT GMB ECU CHN SVK NZL NLD 8 URY MEX ITA BRA 1.0

0.4 0.5 0.6 0.7 0.8 0.9 020406080 Digital Adoption Index % of manufacturing firms having a website

Sources: Panel a: Product Market Regulation Index (OECD, various years); Digital Access Index (DAI) (Internet Coaching Library, various years). Panel b: World Integrated Trade Solution (WITS) database (World Bank, various years); World Bank Enterprise Surveys (World Bank, various years). Data at http://bit.do/WDR2016-Fig1_23. Note: Panel a: The y-axis shows the barriers to entry subindex of the Product Market Regulation (PMR) Index (OECD, various years) for service sectors. The x-axis shows the Digital Adoption Index, as computed for this Report. The PMR index is available for 47 OECD and large developing countries, as well as for eight smaller Latin American countries. Panel b: The y-axis shows the (log) frequency of nontariff barriers (excluding class A, B, and D restrictions) across developing countries. The x-axis shows the share of manufacturing fi rms that have a website in the corresponding developing countries. ICT = information and communication technology; NTM = nontariff measure; OECD = Organisation for Economic Co-operation and Development.

Regulation to foster competition in the digital higher fees, despite having competitive rates for economy needs to consider network effects and customers. So greater digital use needs to be accom- switching costs, which can lead to new entry barri- panied by unifi ed standards, full interoperability, and ers in parts of the digital economy. To address these competition across platforms and contracts. changes in market structures, regulations must Overall, countries that pursue procompetitive reg- prevent anticompetitive behavior and ensure that ulations in most sectors grow faster, as fi rms in these potential entrepreneurs have fair market access. Reg- countries invest in digital technologies and use them ulatory authorities in many countries play a similar more effectively. Given the substantial variations in role for other network industries, such as electricity barriers to domestic or foreign competition and entry or telecommunications. Regulators in several coun- across countries, the divergence in fi rms’ use of the tries have made phone numbers portable across car- digital technologies across sectors and countries is riers—so that users can keep the same phone number not surprising. while switching to a new provider—minimizing switching costs and ensuring a level playing fi eld for The future of markets new entrants. Similar regulatory innovations might be required for social media, digital marketplaces, Uber, the world’s largest taxi company, owns digital payment systems, and the sharing economy, no vehicles. Facebook, the world’s most popular where users need to be able to change internet plat- media owner, creates no content. Alibaba, the most forms with ease and at zero cost. valuable retailer, has no inventory. And Airbnb, the But regulating the digital economy is far from world’s largest accommodation provider, owns no straightforward. It requires deep understanding of real estate. Something interesting is happening. technical characteristics of the corresponding tech- —Tom Goodwin, Havas Media nology. Most internet fi rms, operating in two-sided markets, can blur price signals in either of the two The internet promotes competition, which encour- markets. For instance, sharing economy platforms ages more fi rms to use the internet—a potential typically set the rates they charge customers as well virtuous cycle. So, competition is not only a comple- as suppliers (drivers or homeowners). Once the plat- mentary factor but also a mechanism for the internet form has a large set of users, it can charge suppliers to increase growth. Internet fi rms can circumvent ACCELERATING GROWTH 81 regulatory barriers to competition and entry in tra- Figure 1.24 The dominance of a few politically ditional sectors such as retail, transport, or fi nance. connected fi rms stifl es competition and innovation But this will not happen if vested interests are strong in Morocco, 2004 and 2007 enough to capture regulators and create new barriers 70 to competition and technology adoption. For instance, some taxi companies try to protect their lucrative businesses by pushing for new reg- 60 ulations that block ride sharing companies. Prices of the ride sharing companies, estimated before the 50 ride takes place, can easily be compared, introducing greater transparency—something that taxi regulators 40 attempted for years by requiring taxis to publish their price lists. Cities with tighter market controls like Bar- Share of firms (%) 30 celona, Berlin, and Paris recently saw widespread pro- tests by the taxi industry, as opposed to Dublin, which deregulated its taxi market and lifted restrictions on 20 Have a website Innovate processes Report more than the number of taxis practically overnight in 2000 five competitors (fares remain regulated).74 In India, taxi associations wrote to the Reserve Bank of India complaining that Connected firms Uber’s credit card transactions violate the country’s Unconnected firms in connected industries All firms in connected industries foreign exchange regulations. Uber collects the fare All firms in unconnected industries from a passenger’s credit card, transmits it abroad in Source: Saadi 2015. Data at http://bit.do/WDR2016-Fig1_24. a foreign currency, and then sends 80 percent of the Note: The analysis is based on 48 politically connected and 620 unconnected manufacturing fi rms with transaction value back to the driver. The taxi associa- at least fi ve employees in 2004 and 2007. See Saadi (2015) for more details. tion argues that this violates Indian laws that permit Uber to collect the commission but not the fare in a foreign currency. In Morocco, the market power of fi rms owned by politically connected fi rms.76 Only 28 percent of all infl uential politically connected businessmen stifl es fi rms innovate in politically connected industries, competition, ICT adoption, and innovation among compared with 37 percent in industries without polit- manufacturing fi rms. Firms that are directly or indi- ically connected fi rms (fi gure 1.24). rectly owned by the political rulers are larger, have Economic activities with a high potential growth higher market shares, and have higher profi ts.75 The impact from digital technologies are often protected larger scale seems to help these fi rms adopt digital from foreign or domestic competition in developing technologies and innovate (fi gure 1.24). While polit- countries. The potential impact of the internet on ically connected fi rms individually invest more in fi rm productivity is often highest for economic activ- ICT, their presence seems to discourage other fi rms ities that typically have higher barriers to competi- from doing so. Politically connected fi rms appear tion (table 1.3). These sectors include utilities that are to dominate their markets: only 30 percent of them often natural monopolies, which are more diffi cult to report more than fi ve competitors, compared with 65 regulate. But they also include banking, insurance, percent of all other fi rms; only 11 percent of politically transport, retail trade, and some professional ser- connected fi rms report price competition in domestic vices, which are often protected by regulations from markets, compared with 37 percent of all other fi rms. more foreign or domestic competition in developing There is no difference between politically connected countries (see fi gure 1.23). and unconnected fi rms for price competition in for- In sum, developing countries with an institutional eign (export) markets. The dominance of politically environment that safeguards competition and easy connected fi rms and the resulting lack of competition market entry will more likely harness the full growth seem to discourage all other fi rms in their industry opportunities of digital technologies and catch up from adopting digital technologies or innovating, faster with high-income countries. But countries leading to inferior aggregate technology adoption: will more likely remain poor if the institutional envi- only 32 percent of all fi rms in industries with at ronment allows fi rms to obtain profi ts by lobbying least one politically connected fi rm have a website, for protection rather than investing in new digital compared with 40 percent in industries without technologies.77 82 WORLD DEVELOPMENT REPORT 2016

Table 1.3 Economic activities with high potential for fi rms to use digital technologies more intensively are often protected from foreign or domestic competition in developing countries, reducing productivity growth

Sectors or activities with barriers to competition Potential impact Lowest Medium Highest High — Banking, insurance, retail and wholesale Transport trade

Medium Manufacturing Agriculture, hotels and restaurants, Education, health care, , technical services real estate, utilities

Low — Legal services Construction

Source: WDR 2016 team. Note: The approximate grouping of sectors for the potential impact of digital technology is determined by scalability and contract complexity of most products in the sector, as defi ned in table 1.1. Digital technologies have a high potential in sectors that are more scalable and less complex (upper left cell in table 1.1), a medium impact for sectors that are more scalable but more complex or less scalable but less complex, and a low impact for sectors that are less scalable and more complex (lower right cell in table 1.1). The approximate grouping of sectors for the degree of competition is based on sectors’ tradability, their average barriers to foreign or domestic entry across countries (for example, the Product Market Regulation Index [OECD, various years] and the Services Trade Restrictions Index), and their fi xed costs (natural monopolies or land intensiveness). — = not available.

Notes 10. The results here are somewhat higher than the ones in Jorgenson (2011), who fi nds that the accumulation 1. These channels can affect growth through more than of ICT capital raised aggregate growth by 13 percent one mechanism. For instance, the internet increases a year on average for more than 100 countries from trade not only by allowing more and smaller fi rms 1995 to 2008. Moreover, Manyika and others (2011) to export (inclusion) but also by allowing existing estimate that the average contribution of the inter- exporters to trade more intensively, exploiting econ- net (alone) to economic growth was 0.3 percentage omies of scale (effi ciency). points among seven high-income countries between 2. Nguyen and Schiffbauer 2015. 2004 and 2009, and 0.4 percentage points for fi ve 3. The results are based on ICT surveys linked to census large developing countries (Brazil, China, India, data of fi rms in manufacturing and service sectors. the Republic of Korea, and the Russian Federation). The data cover more than 300,000 fi rms in Vietnam These results should be regarded with caution, how- in 2011 (Nguyen and Schiffbauer 2015); 8,000 fi rms ever, since the growth contribution of the internet is with at least 10 employees in Turkey in 2011 (Atiyas effectively equated with the contribution of TFP. and Bakis 2015); more than 52,000 fi rms in Mexico 11. Melitz 2003. in 2009 (Iacovone, Pereira-Lopez, and Schiffbauer 12. The internet reduces information frictions by provid- 2015); and more than 3,000 manufacturing fi rms in ing a tool for fi rms to easily fi nd information about Brazil (Cirera, Lage, and de Oliveria 2015). new markets and advertise their products to multiple 4. See, for example, Cardona, Kretschmer, and Strobel buyers. In addition, the internet can reinforce busi- 2013; Bartelsman, Hagsten, and Polder 2013. ness and social networks and reduce communication 5. There are two exceptions: fi rms in the second- costs with potential customers (Rauch 1999; Rauch lowest productivity quintile have the largest share of and Trindade 2002; Fink, Mattoo, and Neagu 2005). broadband users in low-income countries, and fi rms 13. Freund and Weinhold (2004) examine the growth in the third-lowest quintile have the largest share of of bilateral merchandise exports of 56 countries broadband users in upper-middle-income countries. and (2002) consider the growth of U.S. imports and 6. Cirera, Lage, and Sabetti 2015. Note that fi rm labor exports of services. Both studies use data from 1995 productivity is higher due either to higher TFP or to to 1999. higher capital intensities. 14. There are two common measurements of internet 7. See, for example, Haller and Siedschlag 2011. usage in these studies. Freund and Weinhold (2002, 8. Atiyas and Bakis 2015. 2004) and Clarke and Wallsten (2006) measure inter- 9. The statistics are based on census data including net usage by the number of web hosts, while Osnago all manufacturing and service fi rms in Vietnam and Tan (2015) use the percentage of individuals between 2007 and 2012 (no size restrictions). In Mex- using the internet in a country. ico and Turkey, the statistics are based on a repre- 15. Osnago and Tan (2015) estimate a panel gravity equa- sentative sample for all manufacturing and service tion using aggregate bilateral trade fl ows from 2001 fi rms with at least 10 employees. to 2012 for all countries with country-pair and year ACCELERATING GROWTH 83

fi xed effects. They determine whether a country has fi xed effects controlling for time-invariant differ- high internet usage when it has an internet usage ences across fi rms and changes in any other export above the median for all countries, which changes determinants across provinces over time. yearly. 28. Baldwin (2011) describes the reduction of transpor- 16. The results are based on Osnago and Tan (2015). tation costs from the late 1800s as the fi rst unbun- 17. eBay 2013. dling, where consumption and production can be 18. Chen and Xu (2015) compare transaction-level data geographically separated. from AliExpress (a branch of Alibaba) and Chinese 29. This relationship has been confi rmed in Canada customs data across seven months. The online trade (Baldwin and Gu 2008), Germany (Rasel 2012), Ire- data are transaction-level data based on AliExpress land (Murphy and Siedschlag 2013), Italy (Benfra- from January to July 2014, and the offl ine trade data tello, Razzolini, and Sembenelli 2009), the United are Chinese customs data from January to July 2006. Kingdom (Abramovsky and Griffi th 2006), and glob- 19. Chen and Xu 2015. ally based on a large dataset of multinational fi rms 20. The average number of export destinations is rel- and their subsidiaries (Alfaro and Chen 2015). atively low since the study focuses on only nine 30. Cristea (2014) also fi nds that the negative relation- products that are narrowly defi ned at the 10-digit ship between the headquarter services exports and level (harmonized classifi cation system). Thus, the communication costs weakens with the education average fi rm is only exporting about two different of the foreign workers. specialized products, for instance, silk and pure cot- 31. Fort 2014. ton T-shirts. 32. Alfaro and Chen (2015) examine the patterns of entry 21. As a comparison, large fi rms selling on eBay reach an of multinational fi rms into 70 countries from 2005 average of 30 export destinations (eBay 2013); large to 2007. The data are from the Harvard Business fi rms are defi ned as having an annual export of more School’s Orbis Database (http://www.library.hbs.edu than US$10,000 in 2012. /go/orbis.html) and cover 1.2 million manufacturing 22. See http://www.theanou.com/. companies. 23. Andjelkovic 2015. 33. Computer and information services exports include 24. Tan 2015; Osnago and Tan 2015. Tan (2015) uses the computer services such as hardware- and software- World Bank Exporter Dynamics Database (World related services, agency services, and database Bank, various years) and estimates a panel regres- services. sion to examine the effect of the internet usage of 34. Cardona, Kretschmer, and Strobel (2013) provide a 4.5 million fi rms in 47 developing countries between survey of the recent microeconomic studies esti- 2002 and 2012. A product is defi ned as a harmonized mating the relationship between the use of digital system (HS) six-digit category. technologies and fi rm productivity growth. 25. This measure of physical infrastructure predicts 35. See UNCTAD 2013 and Ogodo 2009. fi rms’ internet use; it is strongly positively correlated 36. McKinsey Global Institute 2014; Economist 2014. with the number of internet users by province over 37. Chick, Huchzermeier, and Netessine 2014. time (correlation coeffi cient of 0.55). Therefore, the 38. Chick, Huchzermeier, and Netessine 2014. length of the fi ber-optic cables per population is also 39. McKinsey Global Institute 2014. a good proxy for the number of “last mile” connec- 40. The analysis is based on an annual panel of fi rm cen- tions to fi rms or households in a province. sus data from 2007 to 2012 with more than 300,000 26. The number of exporting fi rms and fi rms’ export observations each year. TFP is estimated following share follow a similar pattern across provinces over the methodology of Olley and Pakes (1996). Each time. regression controls for two-digit sector dummies, 27. Fernandes and others (2015) use changes in the year dummies, and fi rm characteristics (fi rm size, number of internet users per capita, Chinese inter- age, and ownership as well as export status). See net domains per capita, and the length of fi ber-optic Nguyen and Schiffbauer (2015) for details. cables per population to measure the rollout of 41. Brown and Goolsbee 2002. Their availability raised broadband infrastructure across Chinese provinces price dispersion initially and then reduced it as inter- between 1999 and 2007. The analysis is based on cen- net use became more widespread, which is consis- sus data for manufacturing fi rms. The ICT intensity with the theory of search costs by Stigler (1961) per industry is measured by the ICT capital services and Stahl (1989). share in total capital services based on World KLEMS 42. The results are based on a regression of the average two-digit sector-level data for the United States, entry density on a dummy that is one in the years two-digit sector-level data on telecommunication of or after the reform, controlling for year-specifi c expenses in China, and four-digit sector-level data effects. from a Vietnamese fi rm census. All empirical speci- 43. The positive link between the use of new technology fi cations include fi rm fi xed effects and year x province and competition is well documented. For instance, 84 WORLD DEVELOPMENT REPORT 2016

Collard-Wexler and De Loecker (2015) show that and the probability is higher for countries with bet- two-thirds of the higher productivity in the U.S. ter logistics. steel sector after the introduction of a major new 60. UNCTAD 2015. technology, the minimill, originated primarily from 61. Bloom, Sadun, and Van Reenen 2012; Inklaar, Tim- an increased competition effect and only one-third mer, and van Ark 2008. from a technology replacement effect. In particular, 62. Chick, Huchzermeier, and Netessine 2014. the minimill expansion drove a productivity resur- 63. See Aghion and others 2001. The theoretical frame- gence at the surviving vertically integrated produc- work is also closely related to Parente and Prescott ers and thus for industry as a whole. (1999). Aghion and others (2001) also discuss whether 44. Mbiti and Weil (2011) fi nd this effect based on fi rm- can reduce the incentives for level data comparing competing money transfer any innovation by reducing the discounted present services. value of associated rents, leading to an inverted 45. Wei and Mozur 2014. U-shaped relation between competition and growth. 46. Golovin 2014. They provide evidence, however, that the negative 47. Andjelkovic 2015. part in the competition-growth nexus is less empiri- 48. Van Welsum 2015. cally relevant (see also Aghion and others 2006, 2008, 49. Cirera, Lage, and Sabetti 2015. 2009). 50. Comin and Mestieri 2013. 51. According to data from the market-research fi rm 64. For instance, Aghion and others (2006, 2008, 2009) IDC. fi nd that more product market competition and 52. Rochet and Tirole 2003. entry lead to higher productivity 53. A two-sided market is one in which (1) two sets of growth. Buccirossi and others (2013) show that agents interact through an intermediary platform improvements in competition policy, measuring and (2) the decisions of each set of agents affect the restrictions to domestic and foreign entry and outcomes of the other set of agents, typically through restrictions to antitrust, led to higher TFP growth an (Rysman 2009). For instance, in an among 22 industries in 12 OECD countries. The e-commerce platform like Amazon, the two agents effect was strongest for antitrust activities. are the consumers and the online sellers whereby 65. Bartelsman, Hagsten, and Polder 2013. consumers benefi t from more online sellers using 66. Bloom, Draca, and Van Reenen 2011. the platform and vice versa. 67. The fi ndings are consistent for the other available 54. Nunn (2007) constructed a measure of contract measures of ICT, such as the share of fi rms that have intensity for tradable goods. Contract-intensive a website or use e-commerce in Vietnam (Nguyen goods include packaging machinery, musical instru- and Schiffbauer 2015), and the average share of a ments, and alcoholic beverages. Goods with low con- fi rm’s workforce using the internet or the share of tract intensity include wearing apparel, frozen food, fi rms’ e-commerce purchasing over total purchases plastics and rubber, base metals, or fertilizers. in Mexico (Iacovone, Pereira-Lopez, and Schiffbauer 55. The correlation coeffi cient between the contract- 2015). intensity four-digit sectors in Vietnam from 2007 to 68. Kretschmer, Miravete, and Pernías 2002. 2011 and the share of fi rms that use the internet (have 69. Cirera, Lage, and Sabetti 2015. The results are based a website) in these sectors is 0.44 (0.28); see Nguyen on more than 2,300 fi rms in these six African coun- and Schiffbauer 2015. tries in 2014. 56. Bloom, Sadun, and Van Reenen (2012) argue that 70. The effect is based on a regression of (log) labor digital technologies enhance managers’ information productivity from 2008 to 2012 on the ICT variables, set, improving their ability to monitor staff and the product-level degree of import competition from allowing them to delegate decisions to subordinate China from 2000 to 2008 (in the Mexican or U.S. management layers. They show that the impact of good management practices on fi rm productivity is markets), and the interaction term between these larger in ICT-intensive sectors. two variables. Each regression controls for fi rm fi xed 57. Bresnahan, Brynjolfsson, and Hitt 2002; Brynjolfs- effects and changes from 2008 to 2012 in other fi rm son and Hitt 2003; Crespi, Criscuolo, and Haskel characteristics (fi rm size, age, location, exports, for- 2007; Bartel and others 2009; Bloom, Sadun, and Van eign or state ownership, and share of skilled labor). Reenen 2012. See Iacovone, Pereira-Lopez, and Schiffbauer 2015. 58. See Commander, Harrison, and Menezes-Filho (2011) The results are robust when using the interaction for Brazil and India, and Nguyen and Schiffbauer term between household internet use at the province (2015) for Vietnam. level and sectors’ ICT intensity based on U.S. data as 59. Riker (2015) shows that internet access increases the an instrumental variable controlling for the endoge- probability of product entry into the United States, neity of the ICT variables. ACCELERATING GROWTH 85

71. Cirera, Lage, and de Oliveria 2015. An increase in Innovation and Productivity.” Review of Economics and competition is measured by a change in fi rms’ mar- Statistics 91 (1): 20–32. ket share in previous years. The results are based on Aghion, Phillipe, , Stephen J. Redding, and a multinomial logit regression from 2009 to 2012 Fabrizio Zilibotti. 2006. “Entry Liberalization and where the dependent variable indicates different Inequality in Industrial Performance.” Journal of the stages of the e-commerce system; the most advanced European Economic Association 3 (2/3): 291–302. e-commerce system also allows for online payments. ————. 2008. “The Unequal Effects of Liberalization: The regressions control for fi rm characteristics such Evidence from Dismantling the License Raj in India.” as fi rm size, age, export status, and ownership and American Economic Review 98 (4): 1397–1412. sector-specifi c effects. Aghion, Phillipe, Christopher Harris, Peter Howitt, and 72. Countries with available data that are not members John Vickers. 2001. “Competition, Imitation and of the World Trade Organization (Afghanistan, Growth with Step-by-Step Innovation.” Review of Eco- Kazakhstan, Lebanon, and Liberia) are excluded nomic Studies 68 (3): 467–92. because they do not face restrictions on using tar- Akerman, Anders, Ingvil Gaarder, and Magne Mogstad. iff rates as an alternative mode to restrict import 2015. “The Skill Complementarity of Broadband Inter- competition. net.” NBER Working Paper 20826, National Bureau of 73. See World Bank Enterprise Surveys (http://www Economic Research, Cambridge, MA. .enterprisesurveys.org/); the World Bank’s Services Alfaro, Laura, and Maggie Chen. 2015. “ICT and Multina- Trade Restrictions Index (Borchert, Gootiiz, and tional Activity.” Background paper for the World Devel- Mattoo 2013). opment Report 2016, World Bank, Washington, DC. 74. See Golovin 2014. To deal with the problem at the Anderson, Chris. 2006. The Long Tail: Why the Future of Busi- time, Ireland set up a “hardship fund” with payments ness Is Selling Less of More. New York: Hyperion. of up to €15,000 to alleviate the fi nancial hardships Andjelkovic, Maja. 2015. “What Is the Digital Economy suffered by license holders due to the devaluation and Why Does It Matter?” Background paper for the of their assets, although the general consensus was World Development Report 2016, World Bank, Washing- that the government was under no obligation to ton, DC. compensate the taxi industry. Atiyas, Izak, and Ozan Bakis. 2015. “ICT, Internet Use and 75. The analysis is based on 48 politically connected and Organizational Characteristics of Turkish Firms.” 620 nonconnected manufacturing fi rms with at least Background paper for the World Development Report fi ve employees in the World Bank Enterprise Surveys 2016, World Bank, Washington, DC. data in 2004 and 2007. The 48 politically connected Baldwin, John Russel, and Wulong Gu. 2008. “Outsourc- fi rms operate in almost all of the 22 two-digit man- ing and in Canada.” Statistics Canada. ufacturing industries. But within these, they operate http://www.statcan.gc.ca/start-debut-eng.html. in fewer than half of the more than 100 four-digit Baldwin, Richard. 2011. “Trade and Industrialization after subsectors in total. See Saadi (2015) for more details. 76. Almost all fi rms in connected and nonconnected Globalization’s Second Unbundling: How Building industries use e-mail. and Joining a Supply Chain Are Different and Why 77. See, among others, Parente and Prescott (1999) or It Matters.” NBER Working Paper 17716, National Bureau of Economic Research, Cambridge, MA. Acemoglu, Johnson, and Robinson (2005). Bartel, Ann P., Casey Ichniowski, Kathryn L. Shaw, and Ricardo Correa. 2009. “International Differences in References the Adoption and Impact of New Information Tech- Abramovsky, Laura, and Rachel Griffi th. 2006. “Out- nologies and New HR Practices: The Valve-Making sourcing and Offshoring of Business Services: How Industry in the United States and United Kingdom.” Important is ICT?” Journal of the European Economic In International Diff erences in the Business Practices and Association 4 (2): 594–601. Productivity of Firms, edited by Richard B. Freeman Acemoglu, Daron, David Autor, David Dorn, Gordon and Kathryn L. Shaw, 55–78. Chicago: University of Hanson, and Brenadan Price. 2014. “Return of the Chicago Press. Solow Paradox.” American Economic Review Papers and Bartelsman, E. J., E. Hagsten, and M. Polder. 2013. Proceedings 104 (5): 394–99. “Cross-Country Analysis of ICT Impact Using Firm- Acemoglu, Daron, Simon Johnson, and James Robin- Level Data: The Micro Moments Database and son. 2005. “Institutions as a Fundamental Cause of Research Infrastructure.” http://www.scb.se/Grupp Long-Run Growth.” In Handbook of Economic Growth /OmSCB/Internationellt/Dokument/esslait-mmd 1, edited by and Steven N. Durlauf, -fi nal.pdf. 385–472. Amsterdam: Elsevier. Benfratello, Luigi, Tiziano Razzolini, and Aghion, Phillipe, R. Blundell, R. Griffi th, P. Howitt, and Sembenelli. 2009. “Does ICT Investment Spur or S. Prantl. 2009. “The Effects of Entry on Incumbent Hamper Offshoring? Empirical Evidence from Micro- 86 WORLD DEVELOPMENT REPORT 2016

data.” Development Working Paper 276, University of Development Report 2016, World Bank, Washington, Milan, Centro Studi Luca d’Agliano, Milan. DC. Bertschek, Irene, Daniel Cerquerab, and Gordon Kleinc. Chen, Yan, Grace YoungJoo Jeon, and Yong-Mi Kim. 2014. 2013. “More Bits—More Bucks? Measuring the “A Day without a Search Engine: An Experimental Impact of Broadband Internet on Firm Performance.” Study of Online and Offl ine Searches.” Experimental Information Economics and Policy 25 (3): 190–203. Economics 17 (4): 512–36. Bloom, Nicholas, Mirko Draca, and John Van Reenen. Chick, Stephen, Arnd Huchzermeier, and Serguei 2011. “Trade Induced Technical Change? The Impact Netessine. 2014. “Europe’s Solution Factories.” Har- of Chinese Imports on Innovation, IT and Productiv- vard Business Review (April). https://hbr.org/2014/04 ity.” NBER Working Paper 16717, National Bureau of /europes-solution-factories. Economic Research, Cambridge, MA. Cirera, Xavier, Filipe Lage, and Joao Maria de Oliveria. Bloom, Nicholas, Raffaella Sadun, and John Van Reenen. 2015. “E-Commerce and Productivity in Brazilian 2012. “ Do IT Better: US Multinationals and Firms.” Background paper for the World Development the Productivity Miracle.” American Economic Review Report 2016, World Bank, Washington, DC. 102 (1): 167–201. Cirera, Xavier, Filipe Lage, and Leo Sabetti. 2015. “ICT Blum, Bernardo S., and Avi Goldfarb. 2006. “Does the Adoption, Innovation and Productivity in Africa.” Internet Defy the Law of Gravity?” Journal of Interna- Background paper for the World Development Report tional Economics 70 (2): 384–405. 2016, World Bank, Washington, DC. Borchert, I., B. Gootiiz, and A. Mattoo. 2013. “Policy Barri- Clarke, George R. G., and Scott J. Wallsten. 2006. “Has the ers to International Trade in Services: Evidence from Internet Increased Trade? Developed and Developing a New Database.” World Bank Economic Review. http:// Country Evidence.” Economic Inquiry 44 (3): 465–84. wber.oxfordjournals.org/content/early/2013/07/14 Collard-Wexler, Allan, and Jan De Loecker. 2015. “Reallo- /wber.lht017.abstract. cation and Technology: Evidence from the U.S. Steel Bresnahan, Timothy, Erik Brynjolfsson, and Lorin M. Industry.” American Economic Review 105 (1): 131–71. Hitt. 2002. “Information Technology, Workplace Colombo, Massimo G., Annalisa Croce, and Luca Grilli. Organization, and the Demand for Skilled Labor: 2013. “ICT Services and Small Businesses Productiv- Firm-Level Evidence.” Quarterly Journal of Economics ity Gains: An Analysis of the Adoption of Broadband 117 (1): 339–76. Internet Technology.” Information Economics and Policy Brown, Jeffrey R., and Austan Goolsbee. 2002. “Does the 25 (3): 171–89. Internet Make Markets More Competitive? Evidence Comin, Diego, and Martí Mestieri. 2013. “If Technol- from the Life Insurance Industry.” Journal of Political ogy Has Arrived Everywhere, Why Has Income Economy 110 (3): 481–507. Diverged?” INET Research Note 26, Institute for New Brynjolfsson, Erik, and Lorin M. Hitt. 2003. “Computing Economic Thinking, New York. Productivity: Firm-Level Evidence.” Review of Econom- Commander, Simon, Rupert Harrison, and Naercio ics and Statistics 85 (4): 793–808. Menezes-Filho. 2011. “ICT and Productivity in Devel- Buccirossi, Paolo, Lorenzo Ciari, Tomaso Duso, Giancarlo oping Countries: New Firm-Level Evidence from Spagnolo, and Cristiana Vitale. 2013. “Competition Brazil and India.” Review of Economics and Statistics 93 Policy and Productivity Growth: An Empirical Assess- (2): 528–41. ment.” Review of Economics and Statistics 95 (4): 1324–36. Conference Board. 2015. “Productivity Brief 2015: Global Byrne, David M., Stephen D. Oliner, and Daniel E. Sichel. Productivity Growth Stuck in the Slow Lane with No 2013. “Is the Information Technology Revolution Signs of Recovery in Sight.” Conference Board, New Over?” International Productivity Monitor 25 (Spring): York. 20–36. Corrado, Carol, Jonathan Haskel, Cecilia Jona-Lasinio, Cai, J., and N. Li. 2015. “The Composition of Knowl- and Massimiliano Iommi. 2013. “Innovation and edge and Long-Run Growth in a Path-Dependent Intangible Investment in Europe, Japan and the US.” World.” 2013 Meeting Papers 336. Society for Economic Oxford Review of 29 (2): 261–86. Dynamics. http://www.freit.org/WorkingPapers Cowgill, Bo, and Cosmina Dorobantu. 2014. “Worldwide /Papers/Development/FREIT848.pdf. Gravity in Online Commerce.” http://www.columbia Cairncross, Frances. 2001. The Death of Distance: How the .edu/~bc2656/papers/WWCurrent.pdf. Communications Revolution Is Changing Our Lives. Cam- Crafts, Nicholas. 2015. “The Economic Impact of ICT: A bridge, MA: Harvard Business Press. Historical Perspective.” Background paper for the Cardona, M., Tobias Kretschmer, and Thomas Strobel. World Development Report 2016, World Bank, Washing- 2013. “ICT and Productivity: Conclusions from the ton, DC. Empirical Literature.” Information Economics and Policy Crespi, Gustavo, Chiara Criscuolo, and Jonathan Has- 25 (3): 109–25. kel. 2007. “Information Technology, Organizational Chen, Maggie, and Min Xu. 2015. “Online International Change and Productivity Growth: Evidence from Trade in China.” Background paper for the World UK Firms.” CEP Discussion Paper 783, Centre for ACCELERATING GROWTH 87

Economic Performance, London School of Economics Fernandes, Ana, Aaditya Mattoo, Huy Nguyen, and Marc and Political Science. Schiffbauer. 2015. “The Impact of Broadband on Cristea, Anca D. 2014. “The Effect of Communication Chinese Exports in the Pre-Alibaba Era.” Background Costs on Trade in Headquarter Services.” January paper for the World Development Report 2016, World 30. http://pages.uoregon.edu/cristea/Research_fi les Bank, Washington, DC. /infofdi.pdf. Fink, Carsten, Aaditya Mattoo, and Ileana Cristina Czernich, Nina, Oliver Falck, Tobias Kretschmer, and Neagu. 2005. “Assessing the Impact of Communica- Ludger Woessmann. 2009. “Broadband Infrastruc- tion Costs on International Trade.” Journal of Interna- ture and Economic Growth.” CESifo Working Paper tional Economics 67 (2): 428–45. 2861, Center for Economic Studies and Ifo Institute, Fogel, R. W. 1965. “The Reunifi cation of Economic Leibniz Universität, Hannover, Germany. History with Economic Theory.” American Economic David, Paul A. 1990. “The Dynamo and the Computer: An Review 55 (1/2): 92–98. Historical Perspective on the Modern Productivity Fort, Teresa C. 2014. “Technology and Production Frag- Paradox.” American Economic Review 80 (2): 355–61. mentation: Domestic versus Foreign Sourcing.” Draca, Mirco, Raffaela Sadun, and John Van Reenen. Working paper, Tuck School of Business at Dart- 2007. “Productivity and ICTs: A Review of the Evi- mouth, Hanover, NH. http://faculty.tuck.dartmouth dence.” In The Oxford Handbook of Information and .edu/images/uploads/faculty/teresa-fort/Fort Communication Technologies, edited by Robin Mansell, _Fragmentation_March2014.pdf. D. Q. Chrisanthi Avgerou, and R. Silvestone, 100–47. Freund, Caroline, and Diana Weinhold. 2002. “The Inter- New York: . net and International Trade in Services.” American Dutz, Mark, Sérgio Kannebley Jr., Maira Scarpelli, and Economic Review 92 (2): 236–40. ————. 2004. “The Effect of the Internet on International Siddharth Sharma. 2012. “Measuring Intangible Trade.” Journal of 62: 171–89. Assets in an Economy: An Applica- Golovin, Sergiy. 2014. “The Economics of Uber.” Bruegel tion to Brazil.” Policy Research Working Paper 6142, (blog), September 30. http://www.bruegel.org/nc/blog World Bank, Washington, DC. /detail/article/1445-the-economics-of-uber/. eBay. 2013. “Commerce 3.0 for Development: The Prom- Goolsbee, Austan, and Peter J. Klenow. 2006. “Valuing ise of the Global Empowerment Network.” eBay Consumer Products by the Time Spent Using Them: report based on an empirical study conducted by An Application to the Internet.” American Economic Sidley Austin LLP. http://www.ebaymainstreet.com Review Papers and Proceedings 96: 108–13. /news-events/commerce-30-development-promise Gordon, Robert J. 2010. “Revisiting U.S. Productivity -global-empowerment-network. Growth over the Past Century with a View of the EC (European Commission). Various years. Eurostat Future.” NBER Working Paper 15834, National Bureau (database), Brussels, http://ec.europa.eu/eurostat/data of Economic Research, Cambridge, MA. /database. ————. 2014. “The Demise of U.S. Economic Growth: Economist. 2014. “Defending the Digital Frontier: A Spe- Restatement, Rebuttal, and Refl ections.” NBER cial Report on Cybersecurity.” http://www.economist Working Paper 19895, National Bureau of Economic .com/news/special-report/21606416-companies Research, Cambridge, MA. -markets-and-countries-are-increasingly-under Grimes, Arthur, Cleo Ren, and Philip Stevens. 2012. “The -attack-cyber-criminals. Need for Speed: Impacts of Internet Connectivity on ————. 2015a. “Hyperconnected Organization.” http:// Firm Productivity.” Journal of Productivity Analysis 37 www.economistinsights.com/technology-innovation (2): 187–201. /analysis/hyperconnected-organisations/fullreport. Haller, Stefanie, and Sean Lyons. 2015. “Broadband ————. 2015b. “The Truly Personal Computer: The Smart- Adoption and Firm Productivity: Evidence from Irish phone Is the Defi ning Technology of the Age.” http:// Manufacturing Firms.” Telecommunications Policy 39: www.economist.com/news/briefing/21645131 1–13. -smartphone-defining-technology-age-truly Haller, Stefanie, and Iulia Siedschlag. 2011. “Determi- -personal-computer. nants of ICT Adoption: Evidence from Firm-Level Ellison, Glenn, and Sara Fisher Ellison. 2014. “Match Data.” Applied Economics 43 (26): 3775–88. Quality, Search, and the Internet Market for Used Hao, Janet, Charles R. Hulten, and Kirsten Jaeger. 2013. Books.” Massachusetts Institute of Technology. “The Measurement of India’s Intangible Capital.” http://economics.mit.edu/fi les/10351. Conference Board, New York. Evans, David. 2015. “An Empirical Examination of Why Hauk, William R., and Romain Wiczarg. 2009. “A Monte Mobile Money Schemes Ignite in Some Developing Carlo Study of Growth Regressions.” Journal of Eco- Countries but Flounder in Most.” Research Paper 723, nomic Growth 14: 103–47. Coase-Sandor Institute for Law Hortaçsu, Ali F., F. Asís Martínez-Jerez, and Jason Douglas. and Economics, Chicago. 2009. “The Geography of Trade in Online Transactions: 88 WORLD DEVELOPMENT REPORT 2016

Evidence from eBay and Mercado Libre.” American Eco- Mbiti, I., and D. N. Weil. 2011. “Mobile Banking: The nomic Journal: 1: 53–74. Impact of M-Pesa in Kenya.” NBER Working Paper Hulten, Charles R., and Janet Hao. 2011. “The Role of 17129, National Bureau of Economic Research, Cam- Intangible Capital in the Transformation and Growth bridge, MA. of the Chinese Economy.” Unpublished manuscript, McKinsey Global Institute. 2014. “Offl ine and Falling University of Maryland. Behind: Barriers to Internet Adoption.” McKinsey Hussain, Sahar. 2015. “When Do Firms in Developing & Company. http://www.mckinsey.com/insights Countries Adopt New Digital Technologies?” Back- /high_tech_telecoms_internet/offline_and_falling ground paper for the World Development Report 2016, _behind_barriers_to_internet_adoption. World Bank, Washington, DC. Melitz, Marc J. 2003. “The Impact of Trade on Intra- Iacovone, Leonardo, Mariana Pereira-Lopez, and Marc Industry Reallocations and Aggregate Industry Pro- Schiffbauer. 2015. “The Complementarity between ductivity.” Econometrica 71 (6): 1695–1725. ICT Use and Competition in Mexico.” Background Minges, Michael. 2015. “Exploring the Relationship paper for the World Development Report 2016, World between Broadband and Economic Growth.” Back- Bank, Washington, DC. ground paper for the World Development Report 2016, Inklaar, R., M. P. Timmer, and B. van Ark. 2008. “Market World Bank, Washington, DC. Services Productivity across Europe and the U.S.” Eco- Murphy, Gavin, and Iulia Siedschlag. 2013. “Determi- nomic Policy 23 (53): 139–94. nants of Offshoring: Empirical Evidence from Ire- Internet Coaching Library. Various years. DAI (Digital land.” SERVICEGAP Discussion Paper 38, University Access Index). Miniwatts Marketing Group, Bogota. of Birmingham, Birmingham, U.K. http://www.internetworldstats.com/list3.htm. Nguyen, Huy, and Marc Schiffbauer. 2015. “Internet, Jorgenson, Dale W. 2001. “Information Technology and Reorganization and Firm Productivity in Vietnam.” the U.S. Economy.” American Economic Review 91 (1): Background paper for the World Development Report 1–33. 2016, World Bank, Washington, DC. ————. 2011. “Innovation and Productivity Growth.” Amer- Niebel, T. 2014. “ICT and Economic Growth: Comparing ican Journal of Agricultural Economics 93 (2): 276–96. Developing, Emerging and Developed Countries.” Keller, Wolfgang. 2002. “Geographic Localization of ZEW Discussion Paper 14–117, ZEW Centre for Euro- International Technology Diffusion.” American Eco- pean Economic Research, Mannheim, Germany. nomic Review 92: 120–42. Nunn, Nathan. 2007. “Relationship-Specifi city, Incom- Keller, Wolfgang, and Stephen R. Yeaple. 2013. “The Grav- plete Contracts, and the Pattern of Trade.” Quarterly ity of Knowledge.” American Economic Review 103 (4): Journal of Economics 122 (2): 569–600. 1414–44. OECD (Organisation for Economic Co-operation and Koutroumpis, Pantelis. 2009. “The Economic Impact of Development). Various years. Product Market Regu- Broadband on Growth: A Simultaneous Approach.” lation Index. OECD, Paris, http://www.oecd.org/eco Telecommunications Policy 33 (9): 471–85. /growth/indicatorsofproductmarketregulation Kretschmer, Tobias, Eugenio J. Miravete, and José C. Per- homepage.htm. nías. 2002. “Competitive Pressure and the Adoption ————. 2015. “ICT Value Added (Indicator).” doi: of Complementary Innovations.” American Economic 10.1787/4bc7753c-en. Review 102 (4): 1540–70. Ogodo, O. 2009. “Africans ‘Losing Out’ in North-South Lamla, Michael, and Marc Schiffbauer. 2015. “Spillovers Collaborations.” Science and Development Network. from FDI in ICT in Jordan.” Background paper for the http://www.scidev.net. World Development Report 2016, World Bank, Washing- Olley, G. S., and A. Pakes. 1996. “The Dynamics of Produc- ton, DC. tivity in the Telecommunications Equipment Indus- Leamer, Edward E., and Michael Storper. 2001. “The try.” Econometrica 64: 1263–97. Economic Geography of the Internet Age.” NBER Osnago, Alberto, and Shawn W. Tan. 2015. “The Effects Working Paper 8450, National Bureau of Economic of the Internet on Trade Flows and Patterns.” Back- Research, Cambridge, MA. ground paper for the World Development Report 2016, Lendle, Andrea, Marcelo Olarreaga, Simon Schropp, and World Bank, Washington, DC. Pierre-Louis Vezina. 2012. “There Goes Gravity: How Ou, Carol Xiaojuan, and Robert M. Davidson. 2009. “Why eBay Reduces Trade Costs.” Policy Research Working eBay Lost to TaoBao in China: The Global Advantage.” Paper 6253, World Bank, Washington, DC. Communications of the ACM 52: 145–48. Manyika, James, Matthieu Pélissié du Rausas, Eric Parente, Stephen L., and Edward C. Prescott. 1999. Hazan, Jacques Bughin, Michael Chui, and Rémi Said. “Monopoly Rights: A Barrier to Riches.” American Eco- 2011. “Internet Matters: The Net’s Sweeping Impact nomic Review 89 (5): 1216–33. on Growth, Jobs and Prosperity.” McKinsey Global Qiang, Christine Zhen-Wei, Carlo M. Rossotto, and Kaoru Institute report, McKinsey & Company. http://www Kimura. 2009. “Economic Impacts of Broadband.” .mckinsey.com/insights/high_tech_telecoms_internet In Information and Communications for Development /internet_matters. 2009: Extending Reach and Increasing Impact, chap. 3. ACCELERATING GROWTH 89

Washington, DC: World Bank. http://siteresources Van Welsum, Desiree. 2015. “Sharing Is Caring? Not .worldbank.org/EXTIC4D/Resources/IC4D Quite: Some Observations about the Sharing Econ- _Broadband_35_50.pdf. omy.” Background paper for the World Development Rasel, Fabienne. 2012. “Offshoring and ICT: Evidence for Report 2016, World Bank, Washington, DC. German Manufacturing and Service Firms.” ZEW Varian, Hal. 2011. “Economic Value of Google.” Paper Discussion Paper 12-087, ZEW Centre for European presented at The Data Frame: The Web 2.0 Summit Economic Research, Mannheim, Germany. 2011 Conference, San Francisco, October 17–19. http:// Rauch, James E. 1999. “Networks versus Markets in www.web2summit.com/web2011. International Trade.” Journal of International Economics Venables, A. J. 2001. “Geography and International 48 (1): 7–35. Inequalities: The Impact of New Technologies.” Jour- Rauch, James E., and Vitor Trindade. 2002. “Ethnic Chi- nal of Industry, Competition and Trade 1 (2): 135–59. nese Networks in International Trade.” Review of Eco- Von Tunzelmann, G. N. 1978. Steam Power and British nomics and Statistics 84 (1): 116–30. Industrialization to 1860. Oxford: Oxford University Riker, David. 2015. “The Internet and Product-Level Entry Press. into the U.S. Market.” Research Note 2015-05B, U.S. Waldfogel, Joel. 2013. “Digitization and the Quality of International Trade Commission, Washington, DC. New Media Products: The Case of Music.” Unpub- Rochet, Jean-Charles, and Jean Tirole. 2003. “Platform lished manuscript, University of Minnesota. Competition in Two-Sided Markets.” Journal of the Wei, Lingling, and Paul Mozur. 2014. “China’s Central European Economic Association 1 (4): 990–1029. Bank Leads Effort to Regulate Internet Finance.” Rysman, Marc. 2009. “The Economics of Two-Sided Mar- Wall Street Journal, February 11. http://www.wsj.com kets.” Journal of Economic Perspectives 23 (3): 125–43. /articles/SB100014240527023038745045793761110285 Saadi, Mohammed Said. 2015. “Innovation, Political Con- 23376. nectedness, and Competition.” Background paper for World Bank. Various years. Doing Business (database). the World Development Report 2016, World Bank, Wash- World Bank, Washington, DC, http://www.doing ington, DC. business.org/. Schiffbauer, Marc, Roberta Serafi ni, and Rolf Strauch. ————. Various years. Enterprise Surveys database. 2011. “The Impact of ICT and Human Capital on World Bank, Washington, DC, http://www.enterprise TFP: Implications for Recent EMU and US Growth surveys.org/. Differences.” Working paper, European , ————. Various years. Entrepreneurship Database. World , Germany. Bank, Washington, DC, http://econ.worldbank.org Stahl, Dale O. 1989. “Oligopolistic Pricing with Sequen- /WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH tial Consumer Search.” American Economic Review 79: /EXTPROGRAMS/EXTFINRES/0,,contentMDK: 700–12. 21454009~pagePK:64168182~piPK:64168060~the Stigler, George J. 1961. “The Economics of Information.” SitePK:478060,00.html. Journal of Political Economy 69: 213–25. ————. Various years. Exporter Dynamics Database. Syverson, Chad. 2013. “Will History Repeat Itself? Com- World Bank, Washington, DC, http://econ.worldbank ments on ‘Is the Information Technology Revolution .org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH Over?’ ” International Productivity Monitor 25 (Spring): /EXTPROGRAMS/EXTTRADERESEARCH/0,,content 37–40. MDK:23192718~pagePK:64168182~piPK:64168060~the Tan, Shawn W. 2015. “The Effects of the Internet on Firm SitePK:544849,00.html. Export Behavior.” Background paper for the World ————. Various years. WITS (World Integrated Trade Development Report 2016, World Bank, Washington, Solution) (database). World Bank, Washington, DC, DC. http://wits.worldbank.org/. UNCTAD (United Nations Conference on Trade and ————. 2015. Measuring the Urban Tech Sector. Unpublished Development). 2013. Information Economy Report manuscript. Washington, DC: World Bank. 2013: The Cloud Economy and Developing Countries. Wulf, Julie. 2010. “Alibaba Group.” Harvard Business Geneva: United Nations. http://unctad.org/en School Case Study 710–436, Harvard Business School, /PublicationsLibrary/ier2013_en.pdf. Cambridge, MA. http://www.hbs.edu/faculty/Pages ————. 2015. Information Economy Report 2015: Unlocking /item.aspx?num=38507. the Potential of E-Commerce for Developing Countries. Geneva: United Nations. http://unctad.org/en /PublicationsLibrary/ier2015_en.pdf. 90 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 1

Agriculture

Across the developing world, agriculture sustains the this disparity, suboptimal agricultural practices and majority of rural livelihoods. Ever since people have poor management are also to blame. New produc- grown crops, raised livestock, and caught fi sh, they tion technologies such as improved seed varieties, have sought information from one another. What is nutrient management, and pest control methods are the most effective planting strategy on steep slopes? not necessarily reaching farmers. Public extension Where can we buy improved seeds? Who is paying agents can overcome information barriers related to the highest price at the market? Over time, weather new agricultural practices and technologies, but such patterns and soil conditions change. Epidemics of extension programs have been burdened by limited pests and diseases come and go. Updated information scale, sustainability, and impact. allows farmers to cope with and benefi t from these Digital technologies help overcome these con- changes. straints. They are reviving agricultural extension and Providing such knowledge can be challenging. advisory services around the world. In cooperation Agriculture is location-specifi c, and farmers need with agricultural research and extension services, accurate local weather forecasts, advice on agricul- organizations such as Digital Green, the Grameen tural practices and input use, and real-time informa- Foundation, and TechnoServe deliver timely, rele- tion about prices and market logistics. Harnessing vant, and actionable information and advice to farm- the rapid growth of the internet and associated digital ers in , Latin America, and Sub-Saharan technologies such as mobile phones is critical to help- Africa, respectively, at a dramatically lower cost than ing farmers obtain the information they need and to traditional services can. Rather than always visiting a promoting transformative agricultural development. farmer, extension agents use a combination of phone Interest is especially keen in increasing access to calls, text, videos, and internet to reduce transac- extension services, improving marketing of outputs, tion costs and increase the frequency of interaction and arranging logistics. The body of rigorous, quan- with farmers. Governments, in partnership with titative evidence on the ways in which digital inno- mobile operators, use phones to coordinate distri- vations can help improve the lives of rural people is bution of seeds and subsidized fertilizers in remote growing steadily. areas through e-vouchers, as in Nigeria’s large-scale e-wallet initiative (spotlight 2, “Digital fi nance”). Enhancing on-farm Technology fi rms such as Climate Corp, based in Silicon Valley, are pioneering the provision of productivity agrometeorological services for early warning of Agricultural productivity varies dramatically around weather and climate risks. A number of innovations the world. While credit constraints, missing insurance aim for real-time and accurate weather monitoring markets, and poor infrastructure account for some of using remote sensing and technologies enabled with geographic information systems (GIS) for climate- This sector focus was contributed by Aparajita Goyal. resilient agriculture. AGRICULTURE 91

Figure F1.1 Introducing mobile phone service reduces price dispersion in local markets Sardine prices in three coastal markets in Kerala, India a. Region I

14 Phones added 12 10

₹ 8 6 4 Price ( /kg) Price 2 0

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 Survey week

b. Region II

14 Phones added 12 10 8 ₹ 6 4 Price ( /kg) Price 2 0

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 Survey week

c. Region III

14 Phones added 12 10

₹ 8 6 4 Price ( /kg) Price 2 0

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 Survey week

Source: Jensen 2007, © Oxford University Press. Reproduced with permission from Oxford University Press; further permission required for reuse. Note: kg = kilogram; ₹ = Indian rupee.

Facilitating market by farmers and traders, and hence overcome an important constraint in the context of limited infra- transparency structure. As Robert Jensen’s classic 2007 study of Agricultural product markets in many developing sardine fi shermen and wholesalers in Kerala, India, countries are poorly integrated. High search costs found, the introduction of mobile phone service dra- have tended to lower competition and create an matically reduced price dispersion and waste in the ineffi cient allocation of goods across markets. When sardine catch, increasing welfare for producers and the internet took off in the mid-1990s, it was often consumers (fi gure F1.1). Similar effects have been claimed that it would improve price transparency, shown for communication platforms such as Esoko cut out middlemen, and make markets more effi - in Ghana, e-Choupal in India, and telecenters in Peru, cient. Indeed, rapid adoption of digital technologies as well as studies on grain traders in Niger and farm- has dramatically reduced the search costs incurred ers in the Philippines.1 92 WORLD DEVELOPMENT REPORT 2016

Enabling effi cient logistics have cast doubt on the overall novelty of information provided to the farmer and the degree of competition and improving quality in many markets. One explanation of weak effects is control low take-up of fee-based price information services. But even when farmers are seemingly better informed, Digital technologies also improve the management they may not necessarily be able to act on that infor- of agricultural supply chains. With globalized food mation because of the inaccessibility of alternative systems, ensuring food safety has become more com- markets and the complex interlinked relationships plex. These trends have catalyzed innovations to trace between buyers and sellers in low-income developing the food supply from producer to consumer—which economies. Rather than assuming that an information is important for developing countries that want to and communication technology (ICT) approach will reach new export markets. Smallholder can always be cost-effective and yield a better outcome, turn to cooperatives and aggregators, who use digital a more nuanced understanding of the underlying institu- tools to improve collection, transportation, and qual- tional environment and constraints is warranted. ity control. By opening up new specialized market Third, technology-enabled interventions are no opportunities, the internet has improved consumer panacea in themselves, and need to be backed by protection and farmers’ livelihoods. complementary investments in physical infrastructure, including electricity and literacy. The versatility and Lessons for adapting digital near-constant innovation that characterize digital technologies for agriculture technologies can sometimes be a distraction that can cause interventions to focus more on the technology Why do many of these innovations fail to scale up than on the demands and priorities of the intended and achieve wider acceptance? One reason is market benefi ciaries and the trade-offs imposed by resource- fragmentation—even though market consolidation constrained environments. will, over time, enhance growth prospects. Another Finally, information technology (IT) policy and the reason is the lack of fi nancially sustainable business broader regulatory environment in a country should be models that will attract private sector investments discussed jointly. Whereas the expansion of mobile in innovative solutions for small-scale agriculture. phone access has been rapid and commercially self- There is great potential for the internet and related sustaining even among many of the poor, the same is technologies to improve rural economies, but several not true of the internet. In the long run, the internet lessons need to be kept in mind. can have an even greater impact on rural growth; First, agriculture is becoming increasingly much depends on fi nding sustainable business mod- knowledge-intensive and high-tech. Some of the els to encourage its spread in the poorest parts of the world’s newest industries have started to put money world. and tech talent into farming—the world’s oldest industry. Digital soil maps, remote sensing, and global positioning system (GPS) guidance are critical tools Note for modern farmers. “Big data” for precision agricul- 1. Nakasone, Torero, and Minten 2014. ture increases yields and effi ciency. These high-tech tools mostly benefi t big farms that can make large investments in technology. But there are also many References innovative ways in which illiterate and otherwise dis- Jensen, Robert. 2007. “The Digital Provide: Information advantaged people use digital technologies, such as (Technology), Market Performance, and Welfare in basic mobile phones. Greater efforts to close the digital the South Indian Fisheries Sector.” Quarterly Journal of divide in rural areas can have great payoffs. Economics 122 (3): 879–924. Second, basic price and market information sys- Nakasone, E., M. Torero, and B. Minten. 2014. “The Power tems can improve effi ciency and welfare. The evi- of Information: The ICT Revolution in Agricultural dence, although strong, is still limited to certain coun- Development.” Annual Review of Resource Economics. tries and certain contexts. A number of recent studies doi: 10.1146/annurev-resource-100913-012714.

94 WORLD DEVELOPMENT REPORT 2016

ENABLING DIGITAL DEVELOPMENT Digital finance

Until a few years ago, agricultural productivity in in the fi nancial sector, many of which have emerged Nigeria was declining, even though government in developing countries (box S2.1). The benefi ts per- spending had increased.1 Since then, agriculture has vade almost all areas discussed in this Report. Digital become a driver of economic growth. One reason has fi nance makes businesses more productive, allows been an innovative mobile wallet system initiated individuals to take advantage of opportunities in the jointly by the public and private sectors and run digital world, and helps streamline public sector ser- by Cellulant, a mobile services company. Nigeria’s vice delivery. smallholder farmers depend on subsidized fertilizer, Like all great opportunities, digital fi nance also but this crucial resource too often did not reach ben- comes with risks. What makes online fi nancial efi ciaries. The 2012 Growth Enhancement Support systems easy to use for customers also makes them Scheme introduced mobile technology to transfer susceptible to cybercrime. The entry of nontraditional fertilizer subsidies directly to farmers, taking the players poses new challenges for policy, regulation, government out of the business of procuring and dis- and supervision. And the ease of transferring funds tributing fertilizer. The support scheme now helps up across the globe—often anonymously, using means

SPOTLIGHT 2 to twice as many farmers, at one-sixth the cost. The such as cryptocurrencies—might increase illicit fi nan- transfer system relies on a database of more than 10.5 cial fl ows. million farmers, who, as registered recipients of the subsidies, now have a better chance of gaining access Benefi ts of digital fi nance to formal or regulated fi nancial services. Based on this initial success, the system is expanding, aided by Digital fi nance promotes fi nancial inclusion a digital identifi cation system and biometric signa- More than 2 billion people have no access to any tures, taking fi nancial services far into Nigeria’s rural fi nancial services. Overall, only about 59 percent of hinterland. men and 50 percent of women in developing coun- This example and many similar experiences cov- tries have an account at a regulated fi nancial insti- ered throughout this Report illustrate the large impact tution. Women, the poor, and small businesses often of the internet and related digital technologies on the rely on informal fi nancial services, even when they fi nancial sector.2 Digital fi nance has promoted fi nan- receive public transfers or remittances. cial inclusion, providing access to fi nancial services Digital payment systems help overcome barriers to to many of the 80 percent of poor adults estimated accessing fi nancial services. Mobile money schemes, to be excluded from the regulated fi nancial sector.3 in particular, allow people who own a phone but do It has boosted effi ciency, as the cost of fi nancial not have a bank account to make and receive pay- transactions has dropped and speed and convenience ments. In the right environment, these systems can have increased. And it has led to major innovations take off and reach massive size rapidly (fi gure S2.1). Digital payments can reduce costs to recipients.4 For WDR 2016 team based on Bossone (2015) and Tropina (2015). instance, farmers in Niger realized time savings for Sources: account (notinnationalcurrency) designedto promote Credits andlocaldigitalcurrencies without abankaccount. use thantraditional payment systems, even for those Safaricom. Thesesystems canoff phones. Awell-known example isM-Pesa,run by books, andenablepayments onlineorthrough mobile currency ascredit onsmartcards orasystem provider’s Mobile money Pay. internet intermediaries suchasGoogleWallet andApple Many are off payment systems suchascredit cards orbankaccounts. Wrappers There are four majorinnovations indigitalpayments. Box S2.1 Bank of England 2014; https://blockchain.info; company reports. company https://blockchain.info; 2014; England of Bank create adigitalinterface withtraditional ered by nontraditional providers, including

80 percent of households within four years four within ofhouseholds percent 80 S2.1 Figure Note: required for reuse. Source: Innovations payments indigital Percent of U.S. households adopting systems Shares of households in the United States, except for Kenyan M-Pesa and cellphone users. users. cellphone and M-Pesa Kenyan for except States, United the in households of Shares

Suri, Jack, and Stoker 2012. Adapted with permission from the National Academy of Sciences; further permission permission further Sciences; of Academy National the from permission with Adapted 2012. Stoker and Jack, Suri, technology, unless stated otherwise 100 40 50 60 70 80 90 10 20 30 store money inthenational 0 04 Color television Internet Radio Cellphone (Kenyan households, estimated) M-Pesa (Kenyan households,estimated) er lower fees andeasier Kenya’s M-Pesa payment system reached reached system payment Kenya’s M-Pesa are alternative units of Years sinceinitialpublicavailability

mobile payments onM-PesainKenya, itwas about transactions worldwide was aboutUS$23billion;for fl ows, consider this:In2014, thevolume ofBitcoin of transactions. use cryptographic techniques to ensure secure validation cies, includingBitcoin, are cryptocurrencies becausethey actions inapubliclyvisibleledger. Most digitalcurren- scheme andanew currency. Suchschemesrecord trans- Digital in computer games. spending inalocaleconomy orasameansofexchange it was US$4.7 trillion. it was US$228billion;andfor thecredit card issuerVisa, US$24 billion;for theonlinepayment platform PayPal, To getasenseofthemagnitudedigitalpayment are bothanew decentralized payment Automobile Telephone Computer Cellphone 201918171615141312111098765321 DIGITAL FINANCE

SPOTLIGHT 2 95 96 WORLD DEVELOPMENT REPORT 2016

each payment equivalent to an amount that would short-term loans with approval often in a matter of feed a family of fi ve for a week. Digital payments seconds based on M-Pesa usage history. Between increase control, since senders of remittances can November 2012 and early 2015, the system issued 21 have a greater infl uence on how recipients use the million loans to 2.8 million unique borrowers, with money, including for savings. Digital fi nance can an average loan amount of US$15. increase the incentive to save, through automatic deposits, text reminders, or default options. Texted Digital fi nance can increase effi ciency reminders increased savings in Bolivia, Peru, and the The internet reduces the cost of many fi nancial Philippines by up to 16 percent.5 Digital payments transactions by allowing their unbundling into sep- improve risk management by making it easier to arate components that can be automated or provided receive support from social networks that can act as by specialized entities. A retail payment consists of safety nets. M-Pesa users were better able to absorb pre-transaction, authorization, clearing, settlement, income shocks compared to nonusers.6 Digital pay- and post-transaction, each one again involving several ments speed up delivery, which is especially impor- steps. Specialized providers can execute individual tant in case of emergencies such as natural disasters. steps, yielding economies of scale that translate into And they increase security compared to traveling with savings. Such service providers are becoming more large amounts of cash, as is commonly necessary in widespread in developing and emerging markets. low- and middle-income countries. Governments can also lower the cost of fi nancial Digital fi nance can increase women’s economic transactions. Electronic payment cards reduced costs participation. In part, this is because digital payments of social transfers in Brazil’s conditional cash trans- can more easily be concealed by the recipient than fer program, Bolsa Familia, from almost 15 percent to cash, at least temporarily, which helps shift economic under 3 percent of total payments.8 McKinsey esti- decision making in favor of women. Access to savings mated in 2010 that digital payments could save the instruments also increases female empowerment government of India US$22 billion per year.9 A large and the consumption and productive investment of part of these savings comes from lower leakages and female entrepreneurs. There is a signifi cantly positive reduced fraud. Electronic payments create a clear relationship between female labor force participation digital record and can be traced, so the likelihood of and female bank account ownership.7 funds not reaching the benefi ciary or of duplicate Information asymmetries—when one party to a payments or payments to “ghost” recipients who transaction knows much more than the other—are do not exist will be lower. Evidence from India also pervasive in the fi nancial sector. In credit markets, shows that using smart cards rather than cash for especially those for informal enterprises and low- social security payments halved the incidence of

SPOTLIGHT 2 income borrowers, the lender usually has limited demands for bribes. information about a potential borrower’s ability to repay a loan, thus impeding lending. Digital tech- Digital fi nance spurs fi nancial innovation nologies help estimate credit scores from digital The fi nancial sector is transaction-intensive and has footprints. Alifi nance, a subsidiary of the Chinese always been at the forefront of adopting new tech- e-commerce fi rm Alibaba that is now part of nology. Automation has led to a signifi cant reduction Ant Financial, provides loans to vendors on its in fi nancial transaction costs. This has spawned e-commerce platform. Many vendors have small innovations, such as automated credit scoring using operations and face diffi culties obtaining loans in the advanced analytics and massive amounts of data. traditional fi nancial system. Alifi nance’s credit scor- Automating processes allows new, so-called fi n-tech ing model is based on at least three months of a ven- fi rms to offer services often at lower costs than tra- dor’s online activity and makes loan decisions auto- ditional providers, including money transfer across matically and almost instantly. Loans are Y 20,000 borders (Xoom, TransferWise) or short-term loans to Y 30,000 (US$3,500 to US$5,000), on average, with (Wonga, ZestFinance). Part of the Alibaba Group in fl exible repayment terms. Microlending elsewhere China, Yu’E Bao—or “leftover treasure” in Mandarin— that serves small farmers or small businesses func- is a fi xed-income investment fund into which cus- tions similarly, often using mobile phone payment tomers can easily transfer balances from their Alipay records to assess credit risk. M-Shwari, operated by digital payment accounts. Established in 2013, by the the Commercial Bank of Africa and Safaricom, is a end of 2014, the fund had 185 million users and about bank account linked to Kenya’s M-Pesa payment US$93 billion in assets. Peer-to-peer lending platforms system. It pays interest on deposits and provides operate without traditional fi nancial intermediaries, DIGITAL FINANCE 97 instead matching potential lenders and borrowers Second, digital fi nance is bringing large num- directly. bers of people into the fi nancial system for the fi rst The emergence of digital currencies has been time. This requires strong consumer education and another innovation made possible by the internet. , including promoting fi nancial , the most well-known, was created in 2009. literacy and fraud prevention, dispute resolution Its value in terms of national currencies has fl uctu- mechanisms, and data privacy. ated widely, its acceptance as a means of exchange A third concern is that fi nancial innovations could has been limited, and there have been widely publi- pose a systemic risk to a country’s banking sector, cized instances of fraud. But a recent analysis by the including credit, liquidity, operational, and consumer Bank of England suggests that the key innovation risk. Prudential regulation of digital fi nance reduces of such currencies is the distributed ledger that this risk, but may involve high compliance costs removes the need for accounting and settlement by that raise barriers to entry, and thus to competition. intermediaries such as banks.10 This model could also Concerns about risks to the banking system were work for other fi nancial assets such as loans, stocks, raised about Bitcoin, but analysis by the Bank of or bonds, although it is unlikely to spread widely England, for instance, suggests that digital currencies anytime soon. play too small a role to threaten fi nancial stability. A greater concern may be that fi nancial innovations Managing risks create distortions in fi nancial markets that could have larger implications. For example, if automation and The rapid development of whole new segments of “big data” approaches make it much easier to issue fi nance has raised policy questions. How should these consumer credit but not commercial credit, fi nancial new areas of fi nance be regulated and supervised, for institutions might overallocate to the former, poten- instance to ensure consumer protection? Do they tially creating a credit bubble and reducing credit pose signifi cant risks to fi nancial stability? And do availability for investments that increase productivity. they make it easier to commit fi nancial fraud or illicit Finally, there are concerns about increased fraud fl ows of funds? in the fi nancial system.11 With the rise of electronic One characteristic of digital fi nance has been the banking, cyberattacks on fi nancial institutions and rise of nontraditional providers of fi nancial services on other sectors processing electronic fi nancial trans- such as money transfers, savings, and lending. Some actions have also increased. Massive theft of credit of these are new companies such as peer-to-peer lend- card information from retailers has highlighted the ing fi rms like Kickstarter or LendingTree. Others are stakes involved. Larger fi nancial institutions have the nonfi nancial institutions setting up a fi nance arm resources and know-how to continuously upgrade

(or “nonbank”), such as e-commerce sites like eBay online and mobile security through tools such as SPOTLIGHT 2 (owner of PayPal) and Alibaba; internet intermediar- encryption or strong authentication. In fact, banks ies like Google; electronics and software developers have been at the forefront of developing secure trans- like Apple; and telecom operators like Safaricom. This action processes. But smaller and nonfi nancial institu- raises several concerns. tions may be more at risk. Apart from monetary losses, One concern is that traditional fi nancial regula- a large risk is also a loss of trust in digital fi nancial sys- tion does not always cover these companies or they tems that may hinder further innovation in the sector. are held to a different standard, such as reduced Besides fraud and theft, digital fi nance could facili- oversight, even though they can scale up quickly. tate fi nancial fl ows for illegal or illicit purposes.12 Such These problems are somewhat similar to the “shadow transfers could be (“cyberlaunder- banking problem” that preceded the global fi nancial ing”), when illegally obtained funds are turned into crisis, and regulators are exploring ways to shift seemingly legal assets through a process of deposit, from regulating entities to regulating activities. Tra- layering, and integration with legitimate funds. They ditional fi nancial institutions, in contrast, use regu- could represent payments for illegal goods or services latory arguments to keep out innovative providers such as drugs. Or they could be illicit fi nancial fl ows— of digital fi nancial services that could greatly benefi t often from developing countries to capital markets in consumers and fi rms. This has obstructed the growth industrialized countries—that represent the proceeds of online payment systems in , for from , corruption, or bypassing of capital example. Light regulation fosters innovation. M-Pesa controls. Some fl ows may be to circumvent arbitrary could not have grown as quickly had Kenya’s central regulations by unaccountable governments, but often bank erected strict regulatory hurdles. they are the result of illegal activity that deprives a 98 WORLD DEVELOPMENT REPORT 2016

Box S2.2 Technology can help unveil illicit money fl ows

Global communication networks have made it easier to The largest number of addresses tied to these off shore move illicit money around the world and to shelter assets accounts were from a number of large emerging econ- from domestic tax authorities. But recent high-profi le omies. Besides Off shore Leaks, yielded data cases also show how a combination of human interven- about secret bank accounts in Switzerland, and Lux Leaks tion and technology increases the chances of detection. documented strategies that international corporations In each case, large datasets were leaked by insiders to use to avoid taxes in countries where they make profi ts, tax offi cials or watchdog groups such as the International notably by channeling them through Luxembourg. While Consortium of Investigative Journalists (ICIJ). One case in many clients are from developing countries, they get help 2014, Off shore Leaks, yielded 260 gigabytes of accounts from western accountants, and much of the money ends data in off shore tax havens, including the British Virgin up in industrialized countries or their off shore territories. Islands and . Collaborating with media Increased transparency, thanks to technology’s ability outlets, experts from Costa Rica, Germany, Malta, and to sift through large leaked datasets, is encouraging the United Kingdom developed automated software reforms that will make it harder to hide assets in foreign tools for organizing and searching this massive dataset. jurisdictions.

Sources: http://www.icij.org; press reports.

country of fi nancial resources that, in aggregate, are As with international reporting obligations aimed at thought to exceed the value of .13 preventing fi nancing of illegal activities or terrorism, The anonymity, speed, ease of transaction, and a greater regulatory burden and reporting require- global reach of digital fi nance make illegal and illicit ments can tax the resources of smaller countries, transfers easier, including through the ability to split which may require assistance to comply. large transactions into small tranches. Internet sites Digital fi nance is developing rapidly, and constant such as online casinos and digital currencies that offer changes will challenge the ability of regulators to a high degree of anonymity aid such activity. One catch up. Both too much and too little intervention by

SPOTLIGHT 2 study found that online casinos offered customers a policy makers entails risks. Even in light of new types choice of 235 payment options.14 And Bitcoin has been of fi nancial crime online, however, the opportunities used to circumvent capital controls. Because these of digital fi nance for inclusion, effi ciency, and inno- fl ows as well as cybercrime generally cross borders, vation will likely outweigh the risks. it is not always clear which jurisdiction is responsible. Although the internet might make illegal money transfers easier, it can also help address the problem. Notes In contrast to cash, electronic transfers leave a trail that 1. Case study from Grossman and Tarazi 2014. can aid law enforcement. Digital technologies could 2. See Bossone 2015. also help establish registries of benefi cial ownership 3. World Bank 2014. of fi nancial and commercial holdings and transaction 4. World Bank 2014. monitoring systems, a current priority of the G-20. 5. World Bank 2014. This could reduce tax evasion and cyberlaundering— 6. World Bank 2014. although the barriers to developing such directories 7. Data from the World Bank Findex database, which covers 148 countries. tend to be due more to resistance by tax havens than 8. World Bank 2014. to technical reasons, and reforms have mostly been 9. McKinsey, as cited in World Bank 2014. prompted by high-profi le disclosures (box S2.2). 10. Bank of England 2014. More generally, technology as a tool for tackling the 11. Chatain and others 2011. problem of illicit fi nancial fl ows can complement, but 12. Council of Europe 2012; Tropina 2015. not substitute for, proper legal frameworks, interna- 13. Tropina 2015. tional cooperation, and public-private collaboration. 14. Tropina 2015. DIGITAL FINANCE 99

References Bank of England. 2014. “Innovations in Payment Tech- Suri, Tavneet, William Jack, and Thomas M. Stoker. 2012. nologies and the Emergence of Digital Currencies.” “Documenting the Birth of a Financial Economy.” Quarterly Bulletin 2014 Q3. Proceedings of the National Academy of Sciences 109 (26): Bossone, Biagio. 2015. “Internet, Finance and Develop- 10257–62. ment.” Background paper for the World Development Tropina, Tatiana. 2015. “Do Digital Technologies Facil- Report 2016. World Bank, Washington, DC. itate Illicit Financial Flows?” Background paper for Chatain, P. L., A. Zerzan, W. Noor, N. Dannaoui, and L. De the World Development Report 2016. World Bank, Wash- Koker. 2011. Protecting Mobile Money against Financial ington, DC. Crimes: Global Policy Challenges and Solutions. Washing- World Bank. Various years. Findex (database). World ton, DC: World Bank. Bank, Washington, DC, http://www.worldbank.org Council of Europe. 2012. “Criminal Money Flows /en/programs/globalfi ndex. on the Internet: Methods, Trends and Multi- ————. 2014. The Opportunities of Digitizing Payments. Stakeholder Counteraction.” http://www.coe.int/t Report prepared by the World Bank Development /dghl/monitoring/moneyval/. Research Group for the Australian presidency. Grossman, Jeremiah, and Michael Tarazi. 2014. Serving Washington, DC: World Bank. Smallholder Farmers: Recent Developments in Digital Finance. Washington, DC: CGAP (Consultative Group to Assist the Poor). SPOTLIGHT 2 CHAPTER 2 Expanding opportunities

Digital technologies can improve overall welfare costs when fi rms innovate and automate processes, and reduce poverty, but without complementary or from fully digital goods and services that allow investments, they can also worsen inequality. In fi rms to exploit economies of scale (fi gure 2.1 and Africa alone, 11 million youth are expected to enter the spotlight 1). labor market every year for the next decade.1 Born in But not everyone stands to benefi t automatically. the internet era, they live in a world full of new and Only by improving internet access and basic literacy exciting opportunities.2 Farmers use mobile phones and updating skill and training systems will the ben- to get price information and technical advice. Women efi ts be realized and broadly shared. For the world’s facing barriers to work outside their homes can work poor, the key is to leverage digital technologies to online and better balance work and family. And many improve the productivity of household enterprises, have found earning opportunities through online subsistence family farmers, and the informal sector. work and the on-demand economy. But these new Yet for every person connected to the internet in opportunities come hand in hand with fundamental developing countries, almost three are not; among and rapid changes in the world of work, as digital the poor, more than six are not; and many also lack technologies increase the demand for advanced access to such complementary infrastructure as skills, and many skills quickly become obsolete. electricity or roads. Even if connected, many cannot From a technological standpoint, fewer than half of read or use the information the internet provides. For today’s schoolchildren in China, Croatia, or Thailand workers in more organized labor markets, conditions can expect to fi nd a job in an occupation that exists are changing rapidly. New jobs require different skills today.3 But more than jobs disappearing, they will be from old jobs, and many new jobs are informal or transformed. The challenge for policy makers is to nonwage, without benefi ts or worker protections. ensure that all current and future workers can seize Greater computer power and internet connectivity the growing economic opportunities that accompany make some skills obsolete by substituting for work the spread of digital technologies. The risk is that that is codifi able and routine, and thus can be auto- rapid technological change will end up increasing mated. The remaining tasks require complex skills inequality and leaving many behind—blunting the that complement technology, such as creativity, criti- digital dividends. cal thinking, and problem solving. These skills remain The potential gains from technological progress hard for technology to emulate, but also for education for workers and consumers in developing countries and training systems to provide, leaving many work- are indeed large. Digital technologies can create jobs ers unprepared for the modern world of work. and increase earnings in the small information and Thus while digital technologies can raise produc- communication technology (ICT) sector—and much tivity and enhance overall welfare, associated labor more in the sectors that use ICT. They also increase market disruptions can be painful and can result in worker productivity by augmenting human capital higher inequality. High-skilled workers are the biggest and—especially critical for the poor—connecting peo- winners when paired with digital technologies. Glob- ple to work and markets. And they can benefi t con- ally, returns to education remain high, at 10 percent per sumers by lowering prices and expanding the variety year, and are even higher for those using technology of goods and services available, thus producing con- at work. The poor, with no access to technology and sumer surplus. Most of these consumer gains come lacking skills, see few of the direct downsides from from lower marginal production and distribution technological adoption but also only partial benefi ts. EXPANDING OPPORTUNITIES 101

But it is the middle class that can be hollowed out as Figure 2.1 A framework for the internet and economic jobs held by this segment—often medium skilled—are opportunities transformed, partly driven by technological change. Machine operators and clerical support workers, for DIGITAL example, perform many “routine” tasks that are easily TECHNOLOGIES automated. Since 1995, the share of routine employ- ment in total employment has fallen by 8 and 12 per- centage points in developing and developed countries, respectively.4 Such medium-skilled jobs, critical to the growth of the middle class and held disproportion- ately by the bottom 40 percent of the welfare distri- Overcome Augment Generate bution, give way either to high-skilled jobs that only information barriers existing factors economies of scale a small share of workers qualify for, or to low-skilled jobs that face increasing competition and most likely declining wages. INCLUSION EFFICIENCY INNOVATION So the race is on between skills and technology, and the outcome will determine whether the divi- dends from digital technologies are realized and the Job creation Labor productivity Consumer surplus benefi ts widely shared. It is important to bridge the digital divides both in access and in capabilities. This second divide separates the digitally savvy, who can RISK: INEQUALITY make productive use of digital technologies and have skills that complement them—and the digitally poor, who remain unconnected and unskilled. Providing Source: WDR 2016 team. current and future workers with the cognitive, tech- nical, and socioemotional skills that are augmented (fi gure 2.2). A technology diffuses rapidly when it is by technology—and not replaced by it—is a priority. low in cost, easy to use, has high potential benefi ts, Given the fast pace of technological change—with and fi ts well with the local context. Low-cost mobile more intense job creation, destruction, and realloca- phones—which can be shared, prepaid, billed in tion to be expected—labor regulations, taxation, and prices per second, and do not require much literacy social protection systems will have to support labor or numeracy for basic use—fi t this description, and mobility and adapt to the changing nature of work. are the technology of choice among the poor.8 In This is the analog foundation for workers to succeed Cameroon, Ethiopia, Rwanda, Tanzania, and Uganda, in a digital world. more than four in fi ve mobile phone owners have simple phones, not capable of browsing the internet.9 Personal computers and the internet, by contrast, Connected people require literacy and often foreign language (especially The world is more connected than ever. On aver- English) skills. Computers with internet capabilities age, 8 in 10 people in the developing world own a in the Warana subdistrict in Maharashtra, India, for mobile phone (map 2.1). Digital technologies, often instance, went largely unused except for transmitting low-end phones, connect the more than 60 percent market information to farmers—a function later sub- of the world’s people who did not have a landline stituted by mobile phones, which were cheaper and phone as late as 2000. More people have access to easier to use.10 a mobile phone than to secondary schooling, clean Communications, entertainment, and searching water, or sanitation.5 Internet adoption lags behind for information are the most common uses for mobile mobile phone access, but has tripled since 2005. In phones and the internet. In African countries, social developing countries today, 28 percent of the popu- networking, sending and receiving e-mails, instant lation reports access to the internet at home, and in messaging, and checking facts and defi nitions are the advanced economies, 80 percent.6, 7 most common uses of the internet (fi gure 2.3). This is Mobile phones are driving this interconnect- similar in Argentina, Brazil, Colombia, Mexico, Uru- edness, especially among the poor. All regions are guay, and the European Union countries, especially converging in mobile phone use, but South Asia and with social networking (between 50 and 80 percent Sub-Saharan Africa lag far behind in internet access of all internet users).11 The use of digital technologies 102 WORLD DEVELOPMENT REPORT 2016

Map 2.1 Mobile phones are the main source of connectivity in the developing world, but large gaps in internet access remain

a. Mobile phone adoption rates, circa 2014

High-income

91% 95% 98%

Bottom 40% National Upper 60%

Middle East and North Africa

87% 89% 93%

Bottom 40% National Upper 60%

Latin America and the Caribbean

79% 86% 91%

Bottom 40% National Upper 60%

Sub-Saharan Africa

56% 63% 73%

Bottom 40% National Upper 60%

South Asia Mobile phone adoption rates (%) 80% 87% 94% 0–19 Bottom 40% National Upper 60% 20–39 40–59 and Pacific 60–79 79% 85% 91% 80–100 Bottom 40% National Upper 60% No data

b. Internet adoption rates, circa 2014

High-income

66% 81% 90%

Bottom 40% National Upper 60%

Middle East and North Africa

21% 32% 43%

Bottom 40% National Upper 60%

Latin America and the Caribbean

17% 32% 41%

Bottom 40% National Upper 60%

Sub-Saharan Africa

5% 10% 15%

Bottom 40% National Upper 60%

South Asia

4% 11% 17% Internet adoption rates (%)

Bottom 40% National Upper 60% 0–19 20–39 East Asia and Pacific 40–59 14% 28% 37% 60–79 Bottom 40% National Upper 60% 80–100 No data IBRD 41770 Source: WDR 2016 team, based on Gallup World Poll, various years. Data at http://bit.do/WDR2016-Map2_1. Note: Adoption rates refer to the percentage of individuals who report owning a mobile phone (panel a) and having internet access at home (panel b). EXPANDING OPPORTUNITIES 103

Figure 2.2 All regions are converging in mobile phone access, but South Asia and Sub-Saharan Africa are falling behind in internet access

a. Individuals with mobile phone access b. Individuals with internet access 100 100

80 80

60 60 Percent Percent 40 40

20 20

0 0 2007 2008 2009 2010 2011 2012 2013 2014 2006 20072008 20092010 2011 2012 2013 2014

High-income Latin America and the Caribbean Europe and Central Asia South Asia Middle East and North Africa Sub-Saharan Africa East Asia and Pacific Source: WDR 2016 team, based on Gallup World Poll, various years. Data at http://bit.do/WDR2016-Fig2_2.

Figure 2.3 How people use mobile phones and the internet in Africa

Percentage of individuals who use mobile phones or internet reporting each type of use, 2011–12

a. Mobile phones b. Internet

e-mail Social networking Sending or receiving e-mail Social networks Posting information or instant messaging Web browsing Checking facts or word definitions Reading online newspapers and books Sending and receiving money Playing games Downloading and watching videos or radio International calls Formal education Photos and videos Getting information about schools Getting information about health services Transfer airtime Getting information about goods and services Music Looking for free educational content Getting information from government Games Calling over the internet Missed calls Distance learning or job training Purchasing goods and services Text messages Internet banking Phone calls Interacting with government

0 10 20 30 40 50 60 70 80 90 100 020406080100 Percent Percent Daily Weekly Occasionally Source: WDR 2016 team, based on Research ICT Africa surveys (various years). Data at http://bit.do/WDR2016-Fig2_3. Note: Data are simple averages across 12 African countries. 104 WORLD DEVELOPMENT REPORT 2016

for work, education, and health is more limited but gender (fi gure 2.4). In African countries, the bottom increasing. Across the European Union, around 60 per- 40 percent is only one-third as likely to have access cent of internet users search for health information, to the internet as the upper 60 percent;17 18 percent and 13 percent make appointments with health practi- of men report using the internet versus 12 percent of tioners online.12 In Brazil, 60 percent of internet users women, and 20 percent of youth versus 8 percent of use it for educational purposes, and in Mexico, 35 those more than 45 years old. Such demographic gaps percent.13 One in four individuals who use the internet persist also in Latin America, and even in high- and in African countries reports doing so to get health and middle-income European countries.18 education information.14 Uses vary across population The digital divide refl ects inequalities in access groups. In Brazil and Mexico, women, rural, and poorer and barriers to productive use. Many areas simply populations are less likely to use the internet for fi nan- remain unconnected (chapter 4). Even when a region cial transactions or dealing with public authorities, is connected to the internet, access is not easy. In but in both countries these same groups are equally Cameroon, Ghana, Kenya, and Uganda, more than or more likely to use it for educational purposes than three in four users still access the internet in com- men, urban, and richer populations.15 Across countries, mercial internet cafes, where high costs and slow children and youth are most likely to use the internet connections limit use.19 A survey of 25 developing for education. countries found that although commercial cafes were more expensive than telecenters, they had more The digital divide persists skilled staff and more reliable infrastructure and Despite the rapid spread of digital technologies, more service.20 But this is not the whole story. Illiteracy and than 800 million people lack mobile access worldwide lack of skills are important barriers. In a subsidized (63 percent of them in the bottom 40 percent of the internet and mobile telephony program in rural Peru, income distribution), and 4.3 billion lack internet mobile phone ownership increased, on average, by 12 access (49 percent in the bottom 40).16 For every percentage points, but internet use increased by only person connected to the internet in developing coun- 2 percentage points.21 Explaining this gap is the lower tries, almost three are not, and in some countries, 20 internet use among adults than among youth, and are not. Big gaps remain by income, age, location, and the lack of use among the uneducated. Even among the literate, internet use may be limited by a lack of content in local languages.22 Figure 2.4 The digital divide within countries remains wide, especially in internet use

Internet use in Africa, by demographic and socioeconomic characteristics, 2011–12 Creating jobs, boosting

25 labor productivity, and benefi ting consumers

20 The overall impacts of digital technologies on employment and earnings are positive, if highly heterogeneous (table 2.1). As discussed throughout 15 this chapter, rigorous evidence remains limited, but it suggests that most gains accrue disproportionally to the better educated, with technology complementing 10 the skills and assets of users. But there are exceptions. In Peru, between 2007 and 2009, new internet users had higher income growth than those who remained 5 nonusers, with larger gains in rural areas.23 And in the

Individuals who use the internet (%) United States the diffusion of broadband increased employment rates more in rural areas than in urban.24 0 What explains the aggregate differences? The Bottom Upper Mature Young Rural Urban Women Men 40% 60% (45+) (15–24) fi rst factor is the type of technology. Mobile phones can be particularly benefi cial to people in more dis- Income distribution Age Location Gender (household) advantaged groups—who often lack skills for using Source: WDR 2016 team, based on Research ICT Africa surveys (various years). Data at http://bit.do the internet, or work in agriculture, where mobile /WDR2016-Fig2_4. phones alone can pay off. Thus the benefi ts from Note: The chart shows a simple average for 12 countries. mobile phones tend to be widespread. In Peru, for EXPANDING OPPORTUNITIES 105

Table 2.1 Digital technologies aff ect employment and earnings, the evidence shows

Authors Country Technology Key fi ndings De los Rios (2010) Peru Internet Internet adoption was associated with labor income gains of between 13 and 19 percent. There was no effect on the probability of fi nding employment. Klonner and Nolen South Africa Mobile Mobile phone coverage increased (wage) employment by 15 percentage (2010) (rural) phones points, mostly due to increased employment among women, especially those without signifi cant child care responsibilities. Among men, it induced a shift from agricultural employment to other sectors. Kolko (2012) United States Internet Broadband expansion was associated with local population and employment growth, but average wages and overall employment rates were unaffected. Localities with broadband became more attractive, and the supply of workers responded to job opportunities. Marandino and Uruguay Laptops and In two years, the One Laptop per Child program had no impact on average Wunnava (2014) internet households’ labor earnings but led to a 33-percent increase in hourly labor income among households below the median income. Richer households al- ready had access to technology before the program. Ritter and Guerrero Peru Internet Access to internet and mobile phones increased wage employment, the pro- (2014) (rural) and mobile duction of processed goods, and the prices farmers received for their products. phones Mobile phones were the main driver for agricultural activity, while internet access was the main driver for employment outside agriculture. Source: WDR 2016 team.

both mobile and internet use, the better educated their productivity and earnings. Digital technolo- benefi ted most, but while very few of the less edu- gies can also connect people to work and markets cated adopted the internet, they benefi ted greatly and facilitate the accumulation of productive from their new mobile phones.25 Second is the labor assets. This increases effi ciency in the labor market market context. Technology makes a difference when and the overall economy by allowing workers and it helps overcome obstacles to employment or higher fi rms to better leverage existing assets. productivity. In Peru, mobile phones were more ben- • Benefi ting consumers. When digital technologies efi cial in agriculture, where lack of access to relevant automate processes and generate economies of and timely information kept people from accessing scale, they can lower prices and create new goods better opportunities, whereas internet access made and services, thus increasing consumer surplus. more of a difference outside agriculture, where employers seemed to demand ICT skills and internet While quantifying these benefi ts is diffi cult, use. If the constraints are elsewhere, technology will the evidence discussed in the rest of the chapter not make a difference. In rural South Africa, the roll- suggests that they accrue most to those already out of mobile phone networks increased employment better off (table 2.2). Employment in the ICT sector among women, but only for those who did not have and in ICT occupations is limited and mostly for signifi cant family responsibilities.26 high-skilled workers. In industries that use ICT, the These aggregate effects of digital technologies on potential is greater, especially for those who have opportunities are mediated through three mecha- skills that better complement the technology. But the nisms (see fi gure 2.1): largest payoff from digital technologies is increased labor productivity. Better and cheaper information • Creating jobs. Digital technologies promote inclu- that helps connect people to inputs, outputs, and sion by boosting employment and earnings in work is particularly promising for the poor, since the ICT sector or in ICT occupations across the it addresses a key barrier to raising incomes for economy. But most important, they support jobs people previously disconnected from markets. The and earnings in sectors that use ICT when fi rms lower communication costs associated with mobile and the self-employed adopt new technologies and phones, the most common technology for the poor grow, as well as through ICT-enabled outsourcing today, can increase the effi ciency of agriculture and and entrepreneurship. labor markets, raising household incomes and reduc- • Increasing worker productivity. By taking on some ing poverty. As governments and the private sector tasks previously performed by workers, digital get better at tailoring digital services to the poor, technologies augment workers’ skills, increasing those gains should increase. 106 WORLD DEVELOPMENT REPORT 2016

Table 2.2 Benefi ts of digital technologies for workers and consumers: A scorecard Impact so far Potential impact Channel Poor Nonpoor Poor Nonpoor Creating jobs In the ICT sector and Negligible L Negligible L occupations In sectors that use ICT L M L M

Increasing worker productivity Increasing returns to L M L H human capital Connecting people to M H H H work and markets Benefi ting consumers

Increasing consumer M H H H surplus Source: WDR 2016 team. Note: Poor refers to the bottom 20 percent of the welfare distribution. The diff erential impact summarizes the discussion in the chapter and is a qualitative assessment of the evidence. ICT = information and communication technology; L = low; M = medium; H = high.

Creating jobs production spillovers. The median hourly earnings in the ICT sector and in ICT occupations are 1.5 times In ICT sectors and occupations higher than in urban non-ICT sectors or non-ICT In terms of employment, the ICT sector is small, has occupations in developing countries.31 The high pay high entry barriers, and remains male dominated. refl ects a workforce that is better educated than aver- The ICT sector employs, on average, 1 percent of age. It likely also refl ects the relative of ICT the workers in developing countries (fi gure 2.5). workers in some countries, driving up the skill pre- ICT occupations—such as network administrator mium. These high-paying jobs create more demand and electrical and electronic engineer—are also and new jobs outside ICT. In the United States and 1 percent of employment in developing countries, Turkey, one job in the high-tech industry generates and 2–5 percent in member-countries of the Organi- an average of 3 to 5 additional jobs elsewhere in the sation for Economic Co-operation and Development local economy.32 Many of these additional jobs are 27 (OECD). Even in the United States, since 2000, new low-skilled or medium-skilled in such local services technology-related industries—such as e-commerce as retail, cleaning, and food preparation. Kenya’s and social networking—have accounted for only mobile money service M-Pesa uses more than 80,000 28 0.5 percent of employment. Nor is the sector labor agents or service locations that make an average intensive. Instagram, a photo sharing app, had just profi t of US$70 a month.33 Hormuud Telecom—the 13 employees in 2012, when it was bought by Face- largest operator in Somalia—employs 5,000 staff but book for US$1 billion. Facebook, in turn, had 5,000 supports 25,000 agents. Given the high skill require- employees at the time—compared with 145,000 at ments of the ICT industry, these employment spill- Kodak at its peak in photographic fi lm in the 1990s. overs are more likely to benefi t the poor. Yet Facebook’s market value is several times what Kodak’s was back then.29 And most of these jobs are Expanding businesses that use ICT high-skilled. In developing countries, on average, The greatest employment potential of digital tech- half of all workers in the ICT sector have a tertiary nologies lies outside the ICT sector. By improving education, compared with one-quarter elsewhere. productivity and boosting fi rms’ growth throughout The gender gap is also large, with men 2.7 times more the economy (chapter 1), digital technologies can likely than women to work in the sector and 7.6 times increase aggregate employment and earnings. In more likely to be in ICT occupations.30 China, between 1997 and 2007, the increase in internet While not numerous, ICT jobs pay well and can domains and users per capita had a positive impact generate additional jobs through consumption and on fi rms’ employment in ICT-intensive industries.34 EXPANDING OPPORTUNITIES 107

Figure 2.5 Employment in the ICT sector and in ICT occupations remains small

3.0

2.5

2.0

1.5

1.0 Percent of employment Percent of 0.5

0

Peru India Brazil Chile Nepal GhanaKenya Bolivia VietnamAngola Nigeria Lao PDR Tajikistan Sri Lanka Georgia Tanzania ArmeniaParaguay Uruguay Colombia Costa Rica Mongolia Cambodia Bangladesh Nicaragua Uzbekistan Kyrgyz Republic Macedonia, FYR Dominican Republic China, Yunnan Province ICT sector ICT occupations Source: WDR 2016 team, based on the Skills Towards Employability and Productivity (STEP) household surveys (World Bank, various years); Central Asia World Bank Skills surveys (World Bank, various years); Survey-based Harmonized Indicators Program (SHIP) (World Bank, various years); Socio-Economic Database for Latin America and the Caribbean (SEDLAC) (CEDLAS and World Bank); and South Asia Region MicroDatabase (SARMD) (World Bank, various years). Data at http://bit.do/WDR2016-Fig2_5. The STEP surveys used in this Report cover 11 countries: Armenia, Bolivia, Colombia, Georgia, Ghana, Kenya, Lao PDR, FYR Macedonia, Sri Lanka, Ukraine, and Vietnam, as well as China, Yunnan Province. Note: The ICT (information and communication technology) sector includes ICT manufacturing industries, ICT trade industries, and ICT services (OECD 2011). ICT occupations refer to ICT specialists (OECD 2004, 2014).

In Brazil, between 2009 and 2013, fi rms in industries future.38 These jobs are often in business processing, intensive in the use of ICTs had higher wage increases including call centers and bookkeeping. Such jobs across skill levels than the rest of the economy (fi gure include those that can be broken down into routine 2.6), although they did not experience faster employ- tasks but also those requiring high skills and judg- ment growth. was especially high for ment if they can be reliably performed and monitored workers moving across fi rms.35 remotely. Almost half of business process outsourc- Recent studies, despite measurement diffi culties, ing (BPO) is in banking and fi nancial services, and fi nd a positive causal effect of fi rms’ technology another 20 percent is in high-tech and telecommuni- adoption on employment and earnings, especially cations. Some tasks of radiologists and other medical in fi rms with skilled workers who can make the best services are increasingly also offshored. India, China, use of digital technologies. Thanks to a tax allowance the Philippines, and South Africa are the leading BPO program for ICT investments in small fi rms in the hosts.39 The BPO industry in India employs more than United Kingdom, digital technologies raised labor 3.1 million workers, 30 percent of them women.40 In productivity within fi rms and increased the demand the Philippines, it employs 2.3 percent of all workers for workers performing high-skilled tasks.36 In Nor- (box 2.1). In rural India, a three-year awareness pro- way, a 10-percentage-point increase in broadband gram on opportunities in the BPO industry increased availability in a municipality raised wages of skilled women’s enrollment in relevant training programs, as workers by about 0.2 percent, while lowering wages well as school enrollment among young girls, by 3–5 for low-skilled workers.37 percentage points.41 New technologies are now also challenging the Internet-enabled off shoring and status quo of the outsourcing industry. Many of outsourcing, including online work the characteristics that make jobs “offshorable” also Internet-enabled offshoring is an important source make them more vulnerable to automation. So, as of jobs in developing countries and for women. An technology improves and wages rise, some of the jobs estimated one in four jobs in the United States has typically offshored, such as call center jobs, could be already been offshored or could be offshored in the automated. Indeed, a local health care company in 108 WORLD DEVELOPMENT REPORT 2016

Figure 2.6 In Brazil, internet and software use by fi rms throughout the economy is associated with higher earnings

Average annual wage growth, by fi rms’ ICT intensity and workers’ wage levels, 2009–13

a. All workers b. New hires from last job 0.7 0.7

0.6 0.6

0.5 0.5

0.4 0.4 Percent

0.3 Percent 0.3

0.2 0.2

0.1 0.1

0 0 Low wages Medium wages High wages Low wages Medium wages High wages

Low-intensity ICT Medium-intensity ICT High-intensity ICT

Source: Dutz and others 2015, for the WDR 2016. Data at http://bit.do/WDR2016-Fig2_6. Note: The categories of wage levels are terciles. ICT = information and communication technology.

South Africa is already using Watson, IBM’s artifi cial cent of women earning at least US$1 a day. Nine in ten intelligence system, to assist in customer service.42 of the women had previously not worked outside the Outsourcing opportunities are increasing in other home.45 Samasource, RuralShores, and Digital Divide areas through online work, providing workers and Data are three private service providers. Samasource fi rms with access to larger, even global, employment splits jobs into microwork for almost 6,400 workers, marketplaces (box 2.2). People can work for any mostly in Ghana, Haiti, India, Kenya, and Uganda, on employer anywhere, with parties buying and selling average more than doubling their previous income.46 services that can be delivered online. Upwork (for- Online work can prove particularly benefi cial for merly Elance–oDesk), the largest online outsourcing women, youth, older workers, and the disabled, who service, had 2.8 million job postings worldwide in may prefer the fl exibility of working from home 2014.43 Online jobs range from the very simple that or working fl exible hours. On freelancer.com—an can be completed in a few minutes—such as sign- online outsourcing platform with more than 10,000 ups, forum participation, review writing, and website microworkers globally—57 percent of workers are testing—to the more complex, such as software devel- between 16 and 25 years of age.47 In Elance (part of opment, translation services, data entry, and admin- Upwork), 44 percent of workers are women, com- istrative support. The online work market, though pared with just 25 percent in the nonagricultural still a tiny fraction of overall employment, is worth economy (fi gure 2.7). In a survey on microworkers around US$1 billion annually, up from US$700 million .com carried out for this Report, 27 percent of workers in 2009.44 see being able to work at home and fl exible hours as Impact-outsourcing brings online work to vul- the main advantages of working online. These are nerable communities. Still in its infancy—employing the reasons most cited, even more often than earning around 150,000 workers, or 3 percent of the total BPO extra income, especially among women (fi gure 2.8). industry—it is taking hold in India, Kenya, and South Africa. The government of Kerala, India, outsources Internet-enabled entrepreneurship and information technology services to cooperatives self-employment of women from poor families through the Kudum- By lowering information barriers and costs, the bashree (“Prosperity of the Family”) project. Average internet increases experimentation and gives rise to earnings were US$45 a month, with close to 80 per- new opportunities for entrepreneurship and self- EXPANDING OPPORTUNITIES 109

Box 2.1 Business process outsourcing and jobs in the Philippines: Opportunities and challenges from technological change

The information technology (IT) and business process out- transcription), IT outsourcing (such as software and applica- sourcing (BPO) industry in the Philippines has been a driver tion maintenance), engineering services (such as engineer- of economic growth and job creation in the last decade. It ing design and digital mapping), and creative processes has grown at an average of 24 percent annually, and its share (such as art production and game testing and support)—are of the global off shore services market went from 5 percent intensive in nonroutine cognitive and interpersonal tasks. in 2006 to 11 percent in 2013. Direct employment reached Middle-skilled occupations are intensive in routine cognitive 1 million full-time employees in August 2014 from virtu- tasks, mostly in nonvoice BPO (such as back-offi ce fi nance ally zero in 1999, accounting for around 2.3 percent of the and accounting or human resources), but can also include country’s total employment. The industry has a robust voice many of the jobs in voice business processing (such as cus- sector (primarily call centers), accounting for 64 percent of tomer service and technical support). the industry’s revenue. Health care information management Progress in digital technologies and international com- employment grew by 47 percent from 2012 to 2013. IT out- petition increases the need to move to more high-value- sourcing revenues also grew by 52 percent from 2012 to 2013, added, nonroutine jobs. In the Philippine IT-BPO industry, while knowledge process outsourcing grew by 18 percent. 85 percent of the revenues are generated in jobs inten- Earnings and skill requirements vary across these sec- sive in routine cognitive tasks, with workers increasingly tors. Industry-specifi c jobs tend to be higher skilled than susceptible to automation. Low wages will delay this those that cut across industries (such as human resources process but are unlikely to halt it. As the global market business processing), as they require more technical knowl- for complex services grows, providing a larger volume of edge. In 2012, average annual compensation per employee high-value nonvoice services can promote the industry’s in the industry was around US$8,849, with the highest aver- sustainability. This transition requires a concerted eff ort age compensation in software development (US$17,383). among business, academia, and government to make sys- It was US$8,301 for contact centers and US$7,687 for other temic improvements across the IT-BPO ecosystem, such as BPOs. High-skilled, high-paid occupations—as are most enhancing telecommunications infrastructure and directing research and development–related jobs in knowledge industry-government partnerships toward more nonvoice process outsourcing (such as market research and medical training and skills development.

Source: WDR 2016 team, based on Capili 2015, for the WDR 2016.

employment (box 2.3). Many are in the ICT sector, sellers worldwide, with 31 percent of its sales outside but others are ICT-enabled (see also chapters 1 and the United States.50 4). Alibaba in China, the world’s largest e-commerce As online commerce, the on-demand or sharing platform by sales volume, supports an estimated economy—in which people rent assets or command 10 million jobs, or 1.3 percent of China’s workforce. services directly from one another, coordinated Online shop owners using Alibaba in China, on through the internet—is growing rapidly, if still small average, employ 2.6 additional workers. Four in ten overall and just emerging in developing countries. shop owners are women, 19 percent were previously More than two-thirds of internet users globally are unemployed, 7 percent were farmers, and about 1 willing to be part of the sharing economy.51 In Uganda, percent are persons with disabilities. Alibaba is esti- food delivery services online are used by the diaspora mated to support an additional 2 million jobs, most in to send in-kind remittances to family members. In logistics.48 In Morocco, home-based female weavers Kenya, Sendy connects customers with motorcycle sell rugs and other textiles over the internet to keep couriers for delivery services, which can then be a larger share of their profi ts.49 Etsy, a peer-to-peer paid with mobile money. As of March 2015, Airbnb— e-commerce platform focused on handmade or vin- the largest site for peer-to-peer accommodation— tage items, and similar services, can take this to scale. reported operating in more than 34,000 cities and 190 Etsy has 20 million active buyers and 1.5 million active countries, with more than 25 million guests.52 Peer-to- 110 WORLD DEVELOPMENT REPORT 2016

Box 2.2 The economics of online outsourcing

Online outsourcing or freelancing platforms match fi rms demand in the local economy. An individual in Mongolia and workers to perform work online. They can reduce with expertise in web development and a programming contracting costs and the time it takes to match employ- language such as Python can make close to US$40 an hour ers and employees. Through Upwork, the largest of these online. On Upwork, 90 percent of the work is off shored. On platforms, the time it takes to hire a worker can fall from 43 the largest platforms, most employers are located in devel- to 3 days.a On the workers’ side, around 45 percent of free- oped countries and most workers in developing countries. lancers on Upwork earn most or all of their income through Australia, Canada, and the United States are the largest online work. In oDesk—with Elance, one of the two original employers. On oDesk, Bangladesh, India, Pakistan, the platforms behind Upwork—hourly earnings are on average Philippines, and the United States have the largest number 14 times higher than minimum wages in developing coun- of contractors in relation to their populations (fi gure B2.2.1, tries (fi gure B2.2.1, panel a). This is partly explained by the panel b). fact that online workers are better educated than those in Yet the internet only partly overcomes labor market the general economy. High-paying work is concentrated segmentation, and some barriers to inclusion and to fur- in information and communication technology (ICT) areas ther expansion of online labor markets remain. On Upwork, such as software development, which can pay twice as a given worker is 1.3 times more likely to fi nd work in much as online work in writing and translation, and even her domestic market, and domestic contractors get paid three to four times more than that in customer support and more than international contractors for the same type of sales.b work.c Online outsourcing is most likely to function like Thanks to the internet and innovations in monitoring a truly global market when tasks are less complex and and feedback systems, online labor markets are becoming involve fewer local institutions and less communications. global. Online job platforms increase the pool of talent for Additional constraints that could be addressed by policy fi rms, especially for smaller enterprises, and provide an are related to language (mainly English), regulations, pay- opportunity to monetize skills that may not be in suffi cient ment platforms, and trust.d

Figure B2.2.1 Online labor markets provide work and fairly good pay for workers in developing countries

a. Mean hourly wage in oDesk b. Number of oDesk contractors 32.00 100,000 AUS CHN BLR POL USA GBR PHL 16.00 RUS IND VNM UKR CAN USA 8.00 IND IDN ROU BGD LKA PAK PAK KEN 4.00 BGD PHL 10,000 2.00 UKR

Number GBR 1.00 CAN ROU KEN CHN LKA IDN 0.50 AUS BLR TUN POL FRA 1,000 VNM BGR ESP BRA Mean oDesk hourly wage SRB ITA DEU MAR ARG for that country (U.S. dollars) ARM 0.10 NPL MEX 0 HRV ARE ZAF 0 10 100 900 0.10 0.50 1.00 2.00 4.00 8.00 16.00 32.00 Population of country (millions) Country minimum wage (U.S. dollars per hour) Source: Agrawal and others 2013. Note: oDesk is now part of Upwork, together with Elance. In panel a, the line inside the fi gure is the 45-degree line along which hourly wages in oDesk are the same as the country’s minimum wage; in panel b, it is a regression line. In panel b, the number of contractors refers to those who have ever worked on oDesk.

a. http://elance-odesk.com/online-work-report-global, accessed March 26, 2015. b. Agrawal and others 2013. c. Lehdonvirta and others 2014. d. Kuek and others, forthcoming. EXPANDING OPPORTUNITIES 111 peer car services such as Uber, which operates inter- Figure 2.7 Online work expands women’s access to nationally, or Didi Kuaidi in China, are growing fast. work Expanding the principles of the on-demand economy Global nonagricultural employment composition by gender, “offl ine” and online to the urban self-employed in developing countries percent of total employment could be particularly promising. These workers often lack references and documented work histories when a. Total nonagricultural b. Online work through Elance seeking jobs and depend on word of mouth to expand employment (part of Upwork) their customer base. In this setting, the sharing econ- omy’s decentralized, crowd-based rating systems can help control quality, build trust, and maintain a live Women, 25% “resume.” Plumbers and handymen can expand their client pool by building strong reputations online. Women, 44% These new jobs in the sharing economy have Men, advantages for workers, but they also come with trade- 56% offs. The main advantages for workers are supple- Men, mental income and fl exibility. In the United States, 61 75% percent of sharing economy participants say earning extra money is their main motive.53 Workers appear Sources: WDR 2016 team, based on World Development Indicators (World Bank, various years) and to sort into a working arrangement that suits their Elance Annual Impact Report “Work Diff erently,” June 2013. Data at http://bit.do/WDR2016-Fig2_7. preferences and family circumstances. Two-thirds Note: Results are population-weighted. China is not included. For panel a, latest available data between of Uber drivers in the United States vary their hours 2008 and 2013.

Figure 2.8 Flexibility in hours worked and the ability to work from home are the main advantages of online work, but relatively poor pay and lack of career prospects are concerns

Microworkers.com: Most important advantage (panel a) and disadvantage (panel b) of using an online work platform over a traditional job “offl ine”

a. Most important advantage b. Most important disadvantage

Payment is not Able to earn good enough extra money Friends and family do not understand what I am doing Higher chances of welfare welfare

getting higher earnings Income and Do not have Income and than in jobs offline any social benefits

Tasks posted expire and Able to work from home a vacancy is already and work flexible hours closed Require access to an Reduce time and cost online payment system for job search Require access to internet

Efficiency and Reduce time and cost for work requirements registering the profile to Require internet

find a job Efficiency and and computer skills work requirements Provide access to the Take a long time to job market which register and the process was limited before is cumbersome Do not necessarily Higher chances of getting get to work on the tasks a job that matches my that you applied for skills and interests The job is temporary, and it is not for a long-term career Many jobs in my area Career development of expertise Tasks are posted only for limited industries

Chance to receive Tasks are posted only on-the-job training Career development for limited types of occupations and tasks 0 10 20 30 40 50 020304010 50 Percent of respondents Percent of respondents Men Women Source: Survey of online workers at microworkers.com, where the majority of workers are from developing countries, especially South Asia. Based on Imaizumi and Santos, forthcoming, for the WDR 2016. Data at http://bit.do/WDR2016-Fig2_8. 112 WORLD DEVELOPMENT REPORT 2016

Box 2.3 Expanding opportunities through online music

Usman Riaz began piano lessons at age six in his native US$25 billion for live performances, whose revenues are Pakistan. He later wanted to explore new instruments and rising.a Pirating, YouTube, and music streaming have led to musical styles, but he could not fi nd the right teachers in an overall decline of music sales in developed and devel- his own country. Instead, watching videos online, he taught oping markets. Most of the revenue goes to the biggest himself percussive guitar, a style that uses the strings and stars: the top 1 percent of artists earn about 80 percent of the instrument’s body for percussive eff ects. Riaz also used recorded music revenue.b At the same time, fi le sharing has the internet to showcase his music. His song “Fire Fly” went raised the demand for concerts,c which has always been viral in 2011. A year later, he shared a stage in Edinburgh the biggest source of income for most musicians.d As the with Preston Reed, one of his online . cost of recording and distributing music has gone down, The internet has exposed people around the world to online music sharing could become an eff ective advertising new cultural infl uences—not just global pop culture hits, vehicle for musicians in developing countries—although but also in the tiniest niches. Whether it has also helped limited internet access and slow speeds make this harder. many more artists in developing countries gain access to Even if only a small minority break into national or even arts and entertainment markets is diffi cult to say, given the international entertainment markets, the welfare benefi ts lack of suitable data. Globally in 2014, revenues for recorded of being able to learn from others and share talents on the music were about US$15 billion and falling, compared with internet can still be signifi cant.

Source: Kabanda 2015, for the WDR 2016. a. PricewaterhouseCoopers 2015. b. Thompson 2015. c. Holland, Nosko, and Sorensen 2012. d. Connolly and Krueger 2006.

worked by more than 25 percent from week to week. Workers who are able to work with digital technol- This fl exibility is likely to be valuable for women and ogies and complement them are well positioned to youth and for people between jobs. But these benefi ts access more (and more rewarding) employment and come at a possible cost since these jobs do not provide higher wages. Returns to education have fallen only much work security or protection to workers.54 in Latin America. Everywhere, average private returns remain high—at 10 percent per year—despite large Increasing worker productivity increases in the supply of educated workers in the last few decades. Returns to tertiary education are the Increasing returns to human capital highest, at 14.6 percent; tertiary education is the only Digital technologies can complement human capital, educational level for which returns have not fallen allowing workers to focus their efforts on activities since the early to mid-1990s.55 That refl ects strong with higher value and making them more productive. demand for advanced skills, especially among women Farmers can use precision agriculture or track live- (fi gure 2.9). Returns to education are higher and have stock. Teachers can use massive open online courses been rising more rapidly in ICT-intensive occupations (known as MOOCs) or online teaching tools like Khan compared to the rest of the economy (fi gure 2.10). Academy, better using study time inside and outside the classroom to increase practice and discussion, Connecting people to work and markets and dedicating more time to children who fall behind. Digital technologies help overcome barriers to pro- Researchers can dedicate more time to thinking and ductive employment and connect workers and entre- innovating rather than spending time searching for preneurs to (global) markets, clients, and suppliers. information or duplicating other people’s work. Man- This is particularly important for the disadvantaged agers can work more easily with teams across borders. or often-excluded groups such as the poor, women, In fact, there has never been a better time to be a minorities, the disabled, and people in remote regions well-educated worker. Chapter 1 shows how digital (box 2.4). For all of them, high search costs, long dis- technologies increase labor productivity within fi rms. tances, and a lack of information are key obstacles. EXPANDING OPPORTUNITIES 113

The internet makes labor markets more effi cient Figure 2.9 Returns to education remain high despite by connecting a larger pool of individuals and fi rms at signifi cant expansion in the supply of educated lower cost.56 Indeed.com, Monster.com, and elempleo workers, especially for tertiary education .com are international platforms that aggregate job Average return to one additional year of education in tertiary education vacancies from different sources and allow fi rms to post job openings, and workers to apply for jobs and 25 post résumés. Similar services exist for local markets, such as aldaba.com in the Dominican Republic and kariyer.net in Turkey. Social and professional net- 20 working sites also provide information on potential workers, often on behavioral aspects and social ties not refl ected in traditional resumes. LinkedIn, the 15 largest online professional network, has more than

310 million registered members, 67 percent outside Percent 10 the United States. In Brazil, LinkedIn’s penetration rate is already at 8 percent.57 In most countries, online job search remains concentrated among youth and 5 the best educated and grows with income: online job search among the employed and unemployed is above 0 20 percent in urban Armenia and Georgia, but below 2004 2005 2006 2007 2008 2009 2010 2011 2012 5 percent in urban Bolivia, Ghana, the Lao People’s Women, low- and middle-income countries 58 Democratic Republic, Sri Lanka, and Vietnam. Men, low- and middle-income countries Online job boards, social media, and matching Women, high-income countries platforms can improve labor market effi ciency, espe- Men, high-income countries cially in developing countries and in the informal Source: WDR 2016 team, based on Montenegro and Patrinos 2014. Data at http://bit.do sector, where information failures are large. Online /WDR2016-Fig2_9. Note: Includes 97 countries and only wage employees. The regressions control for potential experience job matching is cheaper and faster than traditional and potential experience squared using individuals’ age. methods.59 In Peru, integrating mobile phones into traditional public intermediation services increased employment among job seekers by 8 percentage Figure 2.10 Returns to education points in the short term.60 In Germany, online job are particularly high in ICT-intensive seekers are better matched to jobs, are happier with occupations work, and have higher chances of promotion and job security.61 But other studies fi nd no effect of online Wage premium, beyond returns to education, for working in tools on the speed of matching or on the length of an ICT-intensive occupation 62 unemployment. The large number of applications 6 per vacancy and stale resumes and job posts makes it costly for employers to select workers and for work- 5 ers to fi nd available jobs. As online tools become more advanced, however, matching is becoming more 4 effective. In the United States, the average unemploy- 3

ment duration for internet searchers was 25 percent Percent shorter than for noninternet searchers, reversing 2 earlier results.63 Online tools can address many labor market fric- 1 tions, but much of this potential remains unrealized. 0 A fi rst challenge is reaching lower-skilled workers. Low- and middle-income High-income countries Some services, such as Souktel, are solving this via countries mobile phone (box 2.5). Babajob in India and Duma Women Men in Kenya have also implemented innovations to Source: WDR 2016 team, based on Monroy-Taborda, Moreno, and Santos, reach the bottom of the pyramid. They use text mes- forthcoming, for the WDR 2016. Data at http://bit.do/WDR2016-Fig2_10. Note: An ICT-intensive occupation scores 4 or higher in an index between 0 saging and “missed calls” to connect low-skilled and (no use of technology at work) and 6 (most use of technology at work). informal workers to vacancies. A second challenge is ICT = information and communication technology. 114 WORLD DEVELOPMENT REPORT 2016

Box 2.4 Bridging the disability divide through digital technologies

For most people, technology makes things easier. For people with disabilities, it makes things possible. —Mary Pat Radabaugh, formerly with the IBM National Support Center for Persons with Disabilities

Prakash lost his sight at birth. Today he is a successful But if not designed to be accessible, digital technolo- entrepreneur and programmer running his own information gies can widen the disparities between persons with and technology (IT) company in a mid-size city in India. Screen- without disabilities. Free and low-cost mobile apps off er reading and voice-recognition software enable him to use a increased functionality for persons across the disabil- computer and write computing programs, and the internet ity spectrum. Assistive software is available for feature helps him fi nd and connect with clients. Technology aug- phones. Accessibility enhancements for web browsers ments his business and his life. promote greater internet use by persons with disabilities. Around the world, more than 1 billion people have dis- Governments should focus on building the capacity of pub- abilities, 80 percent of them in developing countries. They lic bureaucracies, teachers, vocational trainers, employers, face infrastructure and environmental barriers to social, and information and communication technology (ICT) fi nancial, and civic participation, which digital technologies professionals to design accessible content and support ICTs can help overcome. Technology enables multiple means for persons with disabilities; developing the legal, policy, of communication—voice, text, and gestures—to access and regulatory foundation for accessible ICT; supporting, information and engage with others. Magnifi cation, voice through public-private partnerships, the development of recognition, and text-to-speech benefi t persons with accessible ICT, such as local language text-to-speech and visual, cognitive, learning, and mobility disabilities. Short voice-recognition software; and mainstreaming accessibil- message service (SMS), instant messaging, telephone ity in all public services off ered through ICT, such as disaster relay, and video captions reduce communication barriers warnings and communications, public services, and fi nan- for those with hearing and speech disabilities. Hands-free cial services. navigation and gesture-controlled interfaces help those with severe mobility impairments.

Source: Raja 2015, for the WDR 2016.

providing supporting services. There is high demand make it easier to work away from an offi ce. In the for supporting services in online work, especially European Union, telework doubled to reach 9 percent from women. In Souktel, 40 percent of women and in the fi rst half of the 2000s, and around 23 percent 30 percent of men report a need for career coaching.64 of enterprises in the EU-15 employed teleworkers in A third challenge is receiving up-to-date vacancy 2006, up from 16 percent in 2003 and 18 percent in information, since many postings are stale. Employ- 2004. In the United States, in 2009, one-quarter of ers also report a fairly high rate of no-shows for workers used telework regularly.66 interviews. A two-side rating system—quality control The rise in telework has been particularly rapid on vacancies’ expiration dates, and short message among female workers in Europe.67 Budget airlines, service (SMS) reminders to candidates selected for such as JetBlue, manage their customer support interviews—can address some of these shortcomings. centers largely with home workers, mostly women.68 Telework can also make it easier for youth to combine Making work more fl exible school and work and for older workers to work longer. Digital technologies can bring women and new These new work arrangements can address skill gaps entrants into the labor market, especially in white- and increase productivity. Facilitated by the internet, collar occupations, by allowing people to work on home-based work in a 16,000-employee travel agency different schedules or from different locations. In in China improved worker productivity by 13 per- Georgia, Romania, and Ukraine, more than 10 percent cent.69 And where there is a shortage of doctors, tele- of employment is part-time, up from less than 5 per- medicine and examinations of digital X-rays can be cent a decade earlier.65 Video conferences and e-mail very helpful. In Uruguay, through teleconferencing, EXPANDING OPPORTUNITIES 115

Box 2.5 Using digital technologies to match workers with jobs: Souktel in West Bank and Gaza

Souktel is an online job-matching service started in 2009 panel a). There are 200 registered employers, 80 percent of and operating primarily in West Bank and Gaza, but also which are medium in size. active in the Arab Republic of Egypt, Jordan, Morocco, The main benefi t that job seekers see in Souktel’s ser- Rwanda, and Somalia. There are more than 15,000 regis- vice is saving time and money. Most workers look for jobs tered job seekers. The service registers job seekers through that match their skills and experiences, but more than one- short message service (SMS) and online, provides informa- third look for jobs in new occupations or industries. Souktel tion on vacancies, advises on resume preparation, and does connects users to better-paying jobs. The share of workers active job matching by screening potential candidates for earning less than the overall average monthly earnings vacancies. Notifi cations on relevant job vacancies are sent (around US$500 per month) drops both for men and through voice calls and SMS once or twice a week. Job seek- women after fi nding a job through Souktel (fi gure B2.5.1, ers in the platform are young (80 percent are age 15–25) panel b). About 70 percent of users have been invited for and well educated (all have completed secondary and half a job interview at least once. Employers report that using have completed tertiary education). Among Souktel’s core online job platforms allows them to hire from a prescreened users, 4 of 10 do not have jobs. Women are 30 percent of pool of high-quality candidates and reduces the time and all online job seekers in Souktel, a share that is 15 percent- cost for hiring. Firms report a reduction in recruiting costs age points higher than at the national level (fi gure B2.5.1, of about 20 percent.

Figure B2.5.1 Online platforms improve female labor force participation and access to higher-paying jobs

a. Share of the labor force nationally b. Souktel job seekers earning less than and in Souktel US$500 a month in the job before 100 and after using Souktel 80

80 60 60

40 Percent

40 Percent

20 20

0 0 National Souktel Men Women Men Women Before Souktel After Souktel

Source: Imaizumi and Santos, forthcoming, for the WDR 2016. Data at http://bit.do/WDR2016-FigB2_5_1.

English is being taught to fi rst graders by teachers wider mobile phone coverage induced market par- from the Philippines, raising the children’s English ticipation of farmers in remote areas, especially scores and the English profi ciency of Uruguayan among those producing perishable crops.71 In India, teachers.70 e-Choupal has provided computers and internet access in rural areas. Farmers can place orders for Improving access to markets and inputs, and directly negotiate the sale of their produce productive inputs with buyers. With 6,500 kiosks, it now reaches 4 mil- Digital technologies, especially the internet, make lion farmers in more than 40,000 villages.72 it easier for people to sell products in new markets. Digital technologies can increase access to more E-commerce platforms are one example. In Uganda, productive assets and better quality services, raising 116 WORLD DEVELOPMENT REPORT 2016

Box 2.6 The impact of digital technologies on remittances

Online and mobile money transfer systems off er new but international digital remittance services have yet to cost-eff ective means of sending money. In Kenya, among the take off in a substantial way. As of early 2012, only 20 53 percent of adults who reported having sent remittances percent of 130 mobile banking operators worldwide off ered in the past year, 90 percent did so using a mobile phone.a such services. The value of international remittances Today, the of sending money is 8 percent of through mobile phones accounted for less than 2 percent the remitted value. Mobile technology can lower this cost of global remittances in 2013. by removing the need for the physical presence of staff and Policy action is called for on several fronts. First is customers, while ensuring timely and secure transactions. fostering innovative cross-border mobile money transfer Digital technologies make domestic and international technologies. That requires harmonizing banking and tele- remittances cheaper. In Kenya, in 2008 shortly after M-Pesa communications regulations to enable banks to participate entered the market, the cost of sending US$100 domesti- in mobile money transfers, mobile companies to off er cally was US$12 by MoneyGram, US$20 by bank wire, US$6 mobile money services without exclusivity agreements, by postal money order, and US$3 by bus, compared to and telecommunications fi rms to off er micro-deposit and US$2.50 by M-Pesa. In Cameroon, costs have declined by savings accounts. It also requires simplifying regulations 20 percent since mobile money entered the market. Prices aimed at stopping money laundering and the fi nancing of have also fallen for international remittances. In the United terrorism for small-value transfers. And it requires ensuring Kingdom-Bangladesh corridor, the cost of sending US$200 that mobile distribution networks are open to multiple through Western Union fell from 12 percent in 2008 to 7 per- international remittance service providers. cent in 2014 after the entry of digital competition. Between Second is increasing competition by eliminating telecom the United States and Mexico, Xoom charges 4.4 percent for monopolies and exclusivity contracts. The experience in the online money transfers, down from Western Union’s 6.2 per- United States-Mexico corridor shows how eliminating an cent. But the costs are still high for the poor, since they make exclusivity agreement between Western Union and Elektra many small transactions, which tend to be more expensive— can reduce prices. The interoperability of money transfer more than 5 percent for amounts less than US$5. operators in remittance markets, as in Indonesia, Pakistan, Across the board, traditional remittance service provid- Sri Lanka, and Tanzania, can reduce prices further. ers are building their own mobile and online capabilities,

Source: Plaza, Yousefi , and Ratha 2015, for the WDR 2016. a. Demirgüç-Kunt and others 2015.

an individual’s long-term productive capacity. They of identifying borrowers can improve access to fi nan- can help build human capital throughout the life cial and other services at the bottom of the pyramid.75 cycle, managing risks and increasing access to fi nan- Through social and professional networking sites cial capital and remittances (sector focus 2; spotlight and better connections with friends and family, 2; box 2.6). Since the poor are often most constrained the internet also enlarges, deepens, and leverages by these factors, they stand to benefi t the most. social capital to fi nd jobs and access resources (spot- Mobile money accounts can drive fi nancial inclu- light 3).76 In the United States, internet users have a sion. In Sub-Saharan Africa, 12 percent of adults have larger extended network activated when looking for a mobile money accounts, compared with just 2 percent job.77 Digital technologies can also increase agency worldwide, and 45 percent of them have only a mobile and modify aspirations, affecting social norms that money account.73 In Kenya, access to mobile money can be barriers to participation, employment, and has helped in managing risk. Statistically comparable productivity, especially for women,78 much the same households that were not connected to M-Pesa, the as with soap operas in Brazil and cable television in mobile money service, experienced on average a 6–10 India.79 In Africa, the internet appears to broaden percent reduction in consumption in response to sim- social interactions with groups with different politi- ilar shocks.74 And biometric tools that reduce the costs cal views or religious beliefs.80 The internet and EXPANDING OPPORTUNITIES 117 mobile phones also increase ties between migrants Benefi ting consumers and those remaining in the home country.81 My life became easy after I started using the internet. We can learn about any subject. It also helped me earn Improving access to information some income online by using freelancer sites. It helps me For the poor—who rely on mobile phones and often to fi nd health and beauty tips, to know current events and did not have access to a fi xed line—the biggest gains news. Through internet, I am able to stay connected with from digital technologies are likely to come from my friends and relatives. lower information and search costs. When making agriculture and labor market decisions, individuals —Young woman, microworker in Amazon often rely on informal sources, such as family and Mechanical Turk, September 201495 neighbors, or are left with no information: six in ten farmers in Boyaca, Colombia, do not know the prices Beyond earning opportunities, the internet offers in the capital city.82 Information technologies can many benefi ts to individuals as consumers, such as inform workers about prices, inputs, or new technol- consumer convenience, expanded choice, better qual- ogies more quickly and cheaply, reducing friction and ity leisure time, and access to more knowledge. These uncertainty, eliminating costly journeys, and reduc- benefi ts are consumer surplus, often not captured in ing the risks of accidents and crime.83 In rural Niger, GDP statistics (chapter 1). mobile phones reduced search costs by 50 percent.84 In Digital technologies have thus enhanced welfare. turn, these benefi ts can reduce poverty. In rural Peru, Across 12 countries in Africa, 62 percent of people mobile phones increased household real consumption believe that their family is better off because of mobile by 11 percent between 2004 and 2009 and reduced pov- phones, whereas only 21 percent disagree (17 percent erty by 8 percentage points.85 Mobile phones have also are not sure). And 76 percent of people say mobile been found to reduce poverty in East Africa.86 phones help save on travel time and costs. A majority Using technology for getting information on (62 percent) also believe that mobile phones make prices, weather, soil quality, and new technologies, them more secure (fi gure 2.11). The annual consumer and for coordinating with traders is becoming more surplus from Google search has been estimated at common in agriculture (sector focus 1). Among fi sh- ermen in the Indian state of Kerala, price information Figure 2.11 Mobile phones improve sense of security on mobile phones increased their profi ts 8 percent.87 and save time In Honduras, farmers who got market price informa- tion via SMS reported an increase of 12.5 percent in Africa: Respondents that agree with each statement on benefi ts and use of mobile prices received.88 In Argentina, the TRAZ.AR program phones, 2011–12 to track animals increased profi ts per kilo of meat by 8 percent.89 In Pakistan, thanks to mobile phones, I use it to check on safety of loved ones farmers shifted to more perishable but higher return Saves on travel time and cost cash crops, reducing postharvest losses from the most perishable crops by 21–35 percent.90 My family is better off because I have a mobile phone Digital technologies make the largest difference when learning about information in distant markets or It makes me feel more secure or safe among disadvantaged farmers who face more informa- It lets me get more done during the day tion constraints.91 An adult education program on using simple mobile phones in Niger increased internal labor It helps me to find work migration and communication with migrants on labor 92 I use it to mobilize the community market conditions in faraway regions. Gains and uses or for political events are more common when the information transmit- I use it to access free educational content ted is simple (as with prices or weather) rather than nuanced or diffi cult to convey (as in agricultural exten- 0 102030405060708090 93 sion). Similarly, when information provided through Percent technology is not relevant to local needs or when there Women are other constraints to economic activity—such as Men physical infrastructure or market structure—there are Source: WDR 2016 team, based on Research ICT Africa surveys (various years). Data at http://bit.do fewer or no gains, as in cases in Ethiopia and Nigeria.94 /WDR2016-Fig2_11. 118 WORLD DEVELOPMENT REPORT 2016

US$500 per user, or US$150 billion for the 300 million especially the case when the skill needs of new jobs users.96 In Estonia, digital signatures saved 20 min- are different from those of the old jobs. Beyond utes per signature.97 In Europe and the United States, skill-upgrading, the challenge is to ensure that labor consumers are willing to pay an average of US$50 a regulations facilitate and do not impede these tran- month for services that they now get for free on the sitions, and that social protection systems support internet.98 Large consumer surpluses have also been workers when they are between jobs or not working estimated for Brazil, China, and Mexico.99 regularly. The rapid adoption of digital technologies, despite A second risk relates to the changing nature of the costs, speaks for itself. In developing countries, 5 work and the quality of internet-enabled jobs, such percent of consumption goes into ICT, ranging from as microwork or jobs in the on-demand economy. 2.8 percent in the poorest households to 6.6 percent in These new forms of work provide workers and fi rms the richest. In Africa, in addition to covering the cost with fl exibility and improve effi ciency in the use of of the hardware, the median phone owner spends resources, but also come with a possible erosion of more than 6 percent of his or her monthly income workers’ bargaining power and a lack of benefi ts, on mobile phones for calls and SMSs. The share is such as unemployment and health insurance or over 13 percent when accounting for people who own severance pay. In most cases, workers are considered a mobile phone but do not work (mostly youth and independent contractors rather than employees. In a spouses).100 world where a job in a fi rm has been a pathway out Benefi ts to consumers, while signifi cant, also of poverty because fi rms help share risks and provide come with risks. There are concerns about loss of pri- capital, training, and technology,104 higher nonwage vacy (chapter 4); information overload, as more and employment and this “new informality” may not be more information—some relevant and some not—is desirable. at our fi ngertips; and “overconnectivity,” since people In developing countries, most work does not have are constantly online and reachable. The line between these benefi ts, but especially in advanced countries, a leisure and work is blurring. Digital technologies balance is needed between effi ciency and protection make leisure time more enjoyable and less costly, but to avoid a “race to the bottom” in terms of workers’ also make workers more productive and allow them protections. Already, some microwork platforms and to work away from the offi ce. More than one-third companies in the sharing economy provide insurance of internet users in the United States report working to workers and collect taxes, but these new develop- longer hours because of technology, despite also feel- ments in the labor market raise questions about ing more productive.101 traditional approaches for protecting workers. As dis- cussed in chapter 5, this will probably require reforms Labor market polarization not only in the new industries but also in the tradi- tional ones to ensure that all workers—irrespective of can lead to greater inequality their type of work contract—have basic protections. For workers, digital technologies generate new Perhaps the biggest risk from technological opportunities for employment and earnings, but change, however, is that of widening income inequal- also risks. One major risk is related to the speed of ity. Although technologies are becoming widespread, labor market changes and the destruction of jobs. the economic payoffs are not. The poor almost exclu- Nonstandard forms of work and shorter job ten- sively use only mobile phones not connected to the ures are likely to become more common, especially internet. And even if they had access to the internet, among youth. Internet access has been associated they lack the skills to use it productively, with many with more job-to-job fl ows, within the same fi rm and still unable to read in the fi rst place. Positive impacts across employers.102 Large-scale automation can also from using digital technologies—as with other tech- accelerate job destruction, especially in developed nologies in the past century—are most likely to be countries. In addition to factory automation, there is captured by those already better off. In the United automation of logistics and processing, digitization States, the adoption of advanced internet applications (data entry, publishing/printing), and self-service by fi rms led to substantial wage growth in the 6 per- (document creation and management versus clerical cent of counties that were the wealthiest, the most support, or retail self-checkout).103 educated, and had an IT-intensive industry, with no These changes are good for aggregate productiv- effect elsewhere. Technology explains more than half ity, as discussed in chapter 1, but can create challenges the difference in wage growth between already well- for individuals in the transition to new jobs. This is off counties and others.105 EXPANDING OPPORTUNITIES 119

The risk of rising inequality is evident in the Figure 2.13 Labor shares in national declining shares of (routine) labor in national income, income are falling in many countries, and the “polarization” of the labor market—that is, the including some developing countries declining employment in middle-skilled occupations Trends in labor shares in output since 1975 relative to those in low- and high-skilled ones, and percentage points every 10 years the heightened competition for low-skilled jobs. The concern is that the ladder to the middle class is pulled Poland away as middle-skilled jobs disappear or are funda- Mexico Hungary mentally transformed by digital technologies. Estonia Bahrain Declining shares of labor in national income Slovenia Various factors, including technology, are shifting Lithuania the distribution of income within countries away South Africa Norway from routine labor and toward nonroutine labor and Luxembourg 106 capital. In the past few decades, and especially after Micronesia, Fed. Sts. 2000, the share of national income going to workers Namibia has fallen steadily in developed and many developing Latvia countries, driven by a falling share of income going New Zealand China to workers performing mostly routine tasks that fol- Finland low exact, well-defi ned procedures that can be easily Tunisia automated (fi gures 2.12 and 2.13). In the United States, Argentina at the technological frontier, the share of income Slovak Republic going to routine labor has fallen from 38 to 23 percent Germany Austria since the late 1960s, with a simultaneous rise in the Sweden nonroutine labor share from 24 to 34 percent. In Hon- France duras and Romania, in the 2000s, the income share of Italy nonroutine labor increased from 28 to 32 and from 21 Australia to 25 percent, respectively, with declines in the share Taiwan, China of routine labor.107 Where the labor share has fallen Canada Japan most, inequality has risen most (fi gure 2.14). A grow- Denmark ing literature also links recent technological change Switzerland to widening inequality.108 United States Netherlands Figure 2.12 United States: Labor share Belgium in national income is falling, driven by Czech Republic Spain routine labor Singapore Share of routine and nonroutine labor in total income United Kingdom Portugal 45 Bolivia 40 Turkey 35 Armenia Colombia 30 Kenya 25 Thailand 20 Costa Rica Iceland 15 Belarus Income share (%) Income share 10 Moldova 5 Korea, Rep. Ukraine 0 Brazil

1968 1973 1978 1983 1988 1993 1998 2003 2008 2013 –15 –10 –5 0510 Routine labor Nonroutine labor Change in labor shares Source: Eden and Gaggl 2014, for the WDR 2016. Data at http://bit.do Source: Karabarbounis and Neiman 2013. Data at http://bit.do /WDR2016-Fig2_12. /WDR2016-Fig2_13. 120 WORLD DEVELOPMENT REPORT 2016

Figure 2.14 Falling labor shares in national income are Yet declining labor shares in income and job polar- associated with rising inequality ization are only symptoms. At their heart is the fact that digital technologies complement and augment Growth in Gini coeffi cient vs. growth in labor share in national income, 1995–2010 some skills (and thus some workers) while replacing 30 others. Since not everybody has the skills that go well 25 CHN with digital technologies, many can end up falling 20 behind. Education and skills thus determine whether FIN the promise of digital dividends is achieved, or 15 DNK LVA whether technological advances translate into more 10 BGR inequality in a race between skills and technology.113 EST 5 CRI Understanding this dynamic is critical: inequality GBR will increase if more workers do not acquire the mod- 0 ITA NOR HND BLR ern skills most in demand. But if education and train- –5 ESP GRC ing systems increase the supply of workers who meet ARG consumption or income (%) –10 PAN PRY the changes in skill demands, more workers would Change in the Gini coefficient in coefficient the Gini Change in IRL –15 TUN benefi t from technological change and inequality

–20 could decrease. –30 –25 –20 –15 –10 –5 0 5 10 15 20 25

Change in the share of national output going to labor (%) The race between skills and

Source: Eden and Gaggl 2015, for the WDR 2016. Data at http://bit.do/WDR2016-Fig2_14. technology With rising computing power, combined with the connectivity and informational value of the internet, Employment polarization digital technologies are taking on more tasks. They Declining labor shares coincide with the polariza- are particularly good at performing tasks that follow tion of labor markets, most notably in high-income explicit, codifi able procedures—that is, routine tasks. countries.109 Employment is growing in high-skilled, Some of these tasks are cognitive, such as processing high-paying occupations (managers, professionals, payrolls, keeping books, or doing arithmetic. Others technicians) and low-skilled, low-paying occupations are manual or physical, requiring simple motions and (elementary, service, and sales workers). Middle- muscle power, such as driving a train or assembling skilled, middle-paying occupations (clerks, plant goods. These tasks can be easily automated. Nonrou- and machine operators) are being squeezed (fi gure tine tasks, by contrast, are less amenable to automa- 2.15). In high-income countries, on average, the share tion. Doing research, maintaining personal relation- of routine labor in employment has fallen by about ships, and designing new products have proven hard 0.59 percentage points a year since 1995, or almost 12 to automate; so have manual tasks that involve signifi - percentage points for the period. In the United States, cant dexterity, such as cleaning and providing security local labor markets susceptible to automation due to services or personal care. Technology is even moving specialization in routine task-intensive occupations into areas once thought to be the exclusive domain of have no net decline in employment, but they expe- humans, such as driving or writing news articles. rience polarization in manufacturing and services.110 So, the impact of digital technologies on jobs There are signs that employment is also polarizing depends on the type of tasks and how technology in a number of low- and middle-income countries. The either complements or substitutes workers in those average decline in the share of routine employment tasks. A job comprises many tasks, each character- has been 0.39 percentage points a year, or 7.8 per- ized by the skills most used to perform it (cognitive, centage points for the period. China is an exception, socioemotional, or manual) and by how amenable it since the mechanization of agriculture increased the is to automation or codifi cation (table 2.3). In some share of routine employment.111 Labor markets in low- cases, technology augments labor by complementing income countries such as Ethiopia, with a large share workers. Both the researcher and the hairdresser do of employment in manual occupations, are also not tasks that are nonroutine and not easily programmed polarizing; neither is employment in Mongolia or Latin into a computer, but technology makes the researcher American countries where other factors—such as a (who uses more advanced skills at work) much more commodity-driven boom benefi ting low-skilled work- productive, while barely affecting the hairdresser. ers—could play a larger role in shaping labor markets.112 This means that technology is skill-biased. EXPANDING OPPORTUNITIES 121

Figure 2.15 The labor market is becoming polarized in both developed and developing countries

Annual average change in employment share, circa 1995–circa 2012

a. High-income countries 1.5

1.0

0.5

0

–0.5 ercentage points P

–1.0

–1.5 d d c Italy ranceSpain Israel nlan Chile Greece F Iceland AustriaPoland EstoniaIreland Fi Slovenia Sweden Norway Croatia Canada Portugal Uruguay rgentina Lithuania DenmarkGermany Australia Hungary BarbadosA Korea,Switzerlan Rep. Netherlands Luxembourg New Zealand United States United Kingdom Czech RepublicSlovak Republi Russian Federation

b. Low- and middle-income countries 2.0

1.5

1.0

0.5

0

–0.5

ercentage points –1.0 P

–1.5

–2.0

–2.5 y s s a eru India P Turke nzania SerbiaBolivi China a Ukraine Bhutan Ghana Panama Malaysia MauritiusT Uganda Thailand JamaicaSri Lanka Namibia Pakistan Ethiopia Philippine Hondura Costa Rica Mongolia BarbadosNicaraguaBotswana Guatemala South Africa El Salvador Kazakhstan

Macedonia, FYR Egypt, Arab Rep. Dominican Republic

High-skilled occupations (intensive in nonroutine cognitive and interpersonal skills) Middle-skilled occupations (intensive in routine cognitive and manual skills) Low-skilled occupations (intensive in nonroutine manual skills)

Sources: WDR 2016 team, based on ILO Laborsta (various years); I2D2 (World Bank, various years); National Bureau of Statistics of China (various years). Data at http://bit.do /WDR2016-Fig2_15. Note: The fi gures display changes in employment shares between circa 1995 and circa 2012 for countries with at least seven years of data. The classifi cation follows Autor 2014. High-skilled occupations include legislators, senior offi cials and managers, professionals, and technicians and associate professionals. Middle-skilled occupations comprise clerks, craft and related trades workers, plant and machine operators and assemblers. Low-skilled occupations refer to service and sales workers and elementary occupations. For the United States, comparable data could be accessed only for a short period (2003–08); consistent with Autor (2014), the observed polarization is limited in this period, with most of it having taken place in earlier years.

In other cases, workers are in jobs that are routine, cases, technology is labor-saving. The fundamental whether mostly manual or mostly cognitive, and questions then become: To what extent are different are susceptible to automation and to seeing their occupations and countries’ labor markets affected jobs profoundly transformed or vanishing. In these by skill-biased and labor-saving digital technologies? 122 WORLD DEVELOPMENT REPORT 2016

Table 2.3 Interactions between technology and skills at work

Ease of complementarity (technology is labor-augmenting)

High (tasks intensive in cognitive analytical Low and socioemotional skills) (tasks intensive in manual skills) 12Bookkeepers Machine operators High Proofreaders Cashiers (routine tasks) Clerks Typists Ease of automation Researchers Cleaners

(technology is labor-saving) Low Teachers Hairdressers (nonroutine tasks) 43Managers Street vendors

Source: WDR2016 team, adapted from Acemoglu and Autor 2011. Note: Workers in occupations in quadrant 4 can benefi t greatly because the majority of their tasks are diffi cult to automate, and the core of their work is in tasks in which digital technologies make them more productive. Occupations in quadrants 1 and 2 are composed of many tasks that can be easily automated. Productivity in occupations in quadrant 3 is by and large not directly aff ected by digital technologies.

And what are the typical characteristics of workers in to be a worker with only ‘ordinary’ skills and abilities occupations that, in the absence of effective policies, to offer, because computers, , and other digital stand to gain or lose from technological change? technologies are acquiring these skills and abilities at an extraordinary rate.”114 So, recent skill-biased tech- Skill-biased technological change and the nological change favors workers with advanced skills new digital divide (table 2.4). Not only is overall employment moving As Brynjolfsson and McAfee write in The Second toward occupations intensive in these more advanced Machine Age, “There’s never been a better time to be skills, but even within a given job, skill demands are a worker with special skills or the right education, similarly shifting. because these people can use technology to create Because of this, two sets of skills are increasingly and capture value. But there’s never been a worse time important in today’s labor markets: ICT skills and

Table 2.4 Recent evidence on skill-biased technological change Authors Country Findings Akerman, Gaarder, Norway Broadband adoption in fi rms complements skilled workers performing nonroutine tasks and and Mogstad (2015) substitutes for workers performing routine tasks. Autor, Katz, and United States Patterns of wage inequality are best explained by a modifi ed version of the skill-biased Kearney (2008) technical change hypothesis, which emphasizes information technology in complementing abstract (high-education) tasks and substituting for routine (middle-education) tasks. Autor, Katz, and United States Digital technologies widen wage differentials. Skill upgrading within industries accounts for Krueger (1998) most of the growth in the relative demand for college workers, especially in more computer- intensive industries. Berman, India Trade openness and reform promote technology adoption and diffusion and increase the Somanathan, and nonproduction worker shares of employment and total wages in manufacturing, even within Tan (2005) industries. Gaggl and Wright United Kingdom A tax allowance on ICT investments among small fi rms leads, in the short run, to an increase (2014) in demand for nonroutine cognitive-intensive work, some substitution of routine cognitive work, and no effect on manual work. Marouani and Malaysia Without skill-biased technological change, skilled wage earners should expect lower wages Nilsson (2014) and higher unemployment, and unskilled labor should expect higher wages and lower unemployment. Srour, Taymaz, and Turkey Domestic and imported technologies increase the demand for skilled labor fi ve to six times Vivarelli (2013) more than the corresponding demand for unskilled labor. Source: WDR 2016 team. Note: ICT = information and communication technology. EXPANDING OPPORTUNITIES 123

Box 2.7 Skills wanted: Key concepts

Beyond foundational cognitive skills, such as basic literacy work with a software for a person at a bank.b They also and math, a well-educated worker in a modern economy include ICT skills. ICT skills refer to the eff ective application needs to develop the following skills: of ICT systems and devices, and range from ICT specialists who have the ability to develop, operate, and maintain ICT Nonroutine, higher-order cognitive skills. These refer to systems, to basic ICT users, who are competent users of the ability to understand complex ideas, deal with complex the mainstream tools needed in their working life (e-mail, information processing, adapt eff ectively to the work envi- Excel, Outlook, PowerPoint, Word).c ronment, learn from experience, engage in various forms of reasoning, to overcome obstacles by critical thought.a More Nonroutine interpersonal, socioemotional skills. Socio- specifi cally, these include skills such as unstructured prob- emotional skills (also called soft or noncognitive skills) lem solving, and , learning, and reasoning. encompass a broad range of malleable skills, behaviors, attitudes, and personality traits that enable individuals to Technical skills, including information and communication navigate interpersonal and social situations eff ectively.d technology (ICT) skills. Technical skills are those abilities These include grit or the perseverance to fi nish a job or needed to carry out one’s job, such as the ability to repair achieve a long-term goal, working in teams, punctuality, a water leakage for a plumber, the knowledge to operate organization, commitment, creativity, and honesty. a machine for a worker at a factory, or the knowledge to

a. Neisser and others 1996. b. Cunningham and Villasenor 2014. c. European Commission (EC 2004); OECD 2004. d. Cunningham and Villasenor 2014.

higher-order cognitive and socioemotional skills (box what will enable them to participate productively 2.7). Since 2000, the ICT intensity of employment has in life, as in Albania, Indonesia, Jordan, Malaysia, or increased by almost 10 percent in low- and middle- Peru.117 About three-quarters of third-graders in Mali income countries, almost twice as fast, on average, as and Uganda cannot read.118 Beyond basic literacy, ICT in high-income economies (fi gure 2.16). The share of skills are defi cient. In Africa, 7 in 10 people who do not employment in occupations intensive in nonroutine use the internet say they just don’t know how to use cognitive and socioemotional skills has also increased it, and almost 4 in 10 say they do not know what the in low- and middle-income countries, from 19 to 23 internet is.119 In high-income Poland and the Slovak percent. However, the decline in occupations inten- Republic, one-fi fth of adults cannot use a computer.120 sive in routine skills was even larger, from 50 to 44 The use of ICT skills at work is unequal, but their percent (fi gure 2.17).115 This is the driving force behind importance is growing. On average, one-third of urban the polarization of labor markets. The new economy workers (and 20 percent of the bottom 40 in urban offers, therefore, a premium for ICT skills, strong areas) in developing countries use a computer at work, foundational cognitive and socioemotional skills, for example (fi gure 2.18). This number is likely to raise and for more advanced nonroutine 21st-century skills quickly, with ICT use increasing as countries become such as critical thinking, complex problem-solving, richer and as work becomes more complex (fi gures 2.16 creativity, and expert communication. In fact, work- and 2.19). ers using these “new economy” skills and technology Employers are looking for ICT skills, but cannot are better remunerated—by 25–40 percent—than fi nd them. In the former Yugoslav Republic of Mac- their peers with the same level of education but per- edonia, 43 percent of fi rms say ICT skills are very forming traditional tasks and jobs.116 important for workers, but more than 20 percent say that workers lack them.121 Although the same workers Poor digital literacy limits the productive lacking ICT skills often also lack other skills or face use of digital technologies other barriers to employment, digital literacy limits It is hard to use the internet when, even among youth, their employment opportunities (fi gure 2.20). The more than half have a level of functional literacy below use of digital technologies at work is associated with 124 WORLD DEVELOPMENT REPORT 2016

Figure 2.16 Employment is becoming more intensive in the use of digital technologies

Change in the ICT intensity of employment, 2000–12

35

30

25

20

15

10 Percent 5

0

–5

–10

Peru India Latvia Malta Turkey Mexico Croatia Jamaica Namibia PanamaBulgaria Ukraine Moldova RomaniaLithuania Malaysia BarbadosMauritius Thailand Costa Rica Philippines Cambodia El Salvador South Africa Bahamas, The Russian Federation Macedonia, FYR West Bank and Gaza Average, low- and middle-income countries Average, high-income countries

Source: WDR 2016 team, based on Monroy-Taborda, Moreno, and Santos, forthcoming, for the WDR 2016, using ILO Laborsta (various years). Data at http://bit .do/WDR2016-Fig2_16. Note: ICT (information and communication technology) intensity of employment is based on an index between 0 (no use of technology at work) and 19 (most use of technology at work), averaged by occupation (at the three-digit level) and weighted by employment.

Figure 2.17 Nonroutine skills are becoming more important over time

Employment composition by type of occupation according to skills requirements, 2000–12

a. High-income countries b. Low- and middle-income countries 60

45 40 50 35 30 40

25 30 20 15 20 10 10 5 Percent of total employment Percent of total employment 0 0

05 08 10 0 0 012 002 003 005 010 2000 2001 2002 2003 2004 2 2006 2007 20 2009 2 2011 2 2000 2001 2 2 2004 2 2006 2007 2008 2009 2 2011 Nonroutine cognitive or interpersonal Routine cognitive or manual Nonroutine manual

Source: WDR 2016 team, based on ILO Laborsta (various years). Data at http://bit.do/WDR2016-Fig2_17. Note: Data are simple cross-country averages. Classifi cation of occupations according to skills requirements follows Autor 2014 and refl ects the types of skills most intensely used in each occupation.

higher earnings, even after accounting for educa- High-order cognitive and socioemotional tional attainment. Across a sample of eight develop- skills are more important in the new ing countries, the return associated with using ICT at economy work is around 40 percent.122 In Brazil, and focusing Technological progress is redefi ning the nature and on workers most similar to one another, returns to content of jobs. Some of today’s jobs are new and internet use are about 10 percent.123 require new skills—software publishers, data scientists, EXPANDING OPPORTUNITIES 125 enterprise mobile developers. Others have long existed Figure 2.18 In developing countries, one-third of urban but have been transformed. Consider, for example, the workers use digital technology at work job description of an accountant in the 1970s and today. Urban workers who use a computer at work, conditional on working Specialized softwares have automated many of the accountant’s original tasks, shifting her job toward 50 more advisory services and critical thinking. Employers now demand more “new economy” skills: that is, high- 40 order cognitive and socioemotional skills, as evidenced in countries as diverse as Brazil, Malaysia, and FYR Macedonia (fi gure 2.21). 30 At the aggregate level and within fi rms, new technologies substitute for workers performing routine tasks, while making skilled workers who Percent 20 execute nonroutine abstract tasks more productive.124 Across 28 studies, more than half of the top fi ve skills 10 demanded by employers are socioemotional, another 30 percent are higher-order cognitive, and 16 percent 125 are technical. High-order cognitive skills include 0 reasoning, problem solving, and critical thinking, Men Women 15–29 30–49 50+ Upper Bottom 60% 40% while socioemotional skills refer to behavior, person- Gender Age group ality traits, and attitudes, such as grit, teamwork, self- Income distribution discipline, dependability, and leadership. These are Average things that are still hard for technology to replicate. Source: WDR 2016 team, based on STEP household surveys (World Bank, various years). Data at http:// These new economy skills pay off. Workers with bit.do/WDR2016-Fig2_18. strong cognitive and socioemotional skills in Arme- Note: “Average” refers to the average use of technology among all urban workers in 11 countries where the STEP survey was conducted. “Upper 60%” and “Bottom 40%” refer to asset distribution of individuals’ nia, Georgia, the Kyrgyz Republic, Tajikistan, and households. Vietnam are more likely to be employed and have better-quality jobs.126 In Vietnam, within a given industry, performing nonroutine analytical tasks car- Figure 2.19 Employment becomes more intensive in ries a 23-percent earnings premium, and interactive ICT use as economies grow (that is, interpersonal) tasks a 13-percent premium, Share of employment in high–ICT-intensity occupations, circa 2013 whether combined with routine or nonroutine tasks 60 (fi gure 2.22). In Armenia and Georgia, the premium NLD GBR BEL for problem solving and learning new things is close 50 DNK NOR CAN SWE 127 CZE FIN to 20 percent. More generally, in 12 of 16 mostly EST USA MKD AUT 40 SVK DEU developing countries, wages in nonroutine occu- ESP FRA IRL KOR ITA JPN POL pations increased signifi cantly more than wages in GEO 30 128 ARM routine occupations between 2005 and 2011. UKR

Percent VNM RUS In addition to technological change, other factors BOLCOL 20 KEN help explain these global labor market trends. Trade, LKA urbanization, structural transformation, and global- 10 GHA ization fundamentally shape labor markets and in a LAO number of cases are likely to be even more important 0 than technology. In some Latin American countries, 0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 for example, commodity-driven economic booms GDP per capita (constant U.S. dollars) have boosted low-wage earnings and show no polar- Source: Monroy-Taborda, Moreno, and Santos, forthcoming, for the WDR 2016, based on STEP (World Bank, various years), PIAAC household surveys, and World Development Indicators (World Bank, various ization at the aggregate level. Moreover, distinguish- years). Data at http://bit.do/WDR2016-Fig2_19. ing among factors is more diffi cult because they are Note: GDP = gross domestic product; ICT = information and communication technology; PIAAC = related to one another.129 Programme for the International Assessment of Adult Competencies. Yet technological change seems to be an important part of the explanation. First, occupations most inten- the weight of China in manufacturing supply chains, sive in the use of digital technologies are also most the decline in routine labor in the rest of the world intensive in nonroutine cognitive and interpersonal could simply be the result of a shift of routine labor skills (fi gure 2.23). Second, globalization and trade do to China’s manufacturing sector. Between 2000 and not fully explain the observed polarization.130 Given 2010, the share of employment in routine occupations 126 WORLD DEVELOPMENT REPORT 2016

Figure 2.20 Lack of ICT skills is often a constraint to sector too, digital technologies are substituting for employment workers performing routine tasks. In the Indonesian Treasury, electronic budget planning and execution Share of working-age individuals in urban areas who report that lack of ICT skills is a barrier to employment and higher earnings, circa 2013 is linked to job redefi nitions and reassignments for around 5,000 workers (of 8,000) who previously pro- 50 cessed payments, disbursements, and cash manage- ment.137 In Pakistan, the automation of systems in the 40 central bank made 3,000 of 12,000 employees redun- dant (mostly low-skilled staff). The savings boosted 30 the salaries of remaining employees.138 Two-thirds of all jobs could be susceptible to auto- Percent 20 mation in developing countries in coming decades, from a pure technological standpoint (fi gure 2.24). 10 Estimates for the United States and Europe range between 50 and 60 percent of jobs.139 Given expected 0 advances in artifi cial intelligence, falling ICT prices, and increased coverage of the internet, the potential Bolivia Kenya Ghana Vietnam Lao PDR Armenia Ukraine Georgia for automation is clear. Rapid automation of a large Colombia Sri Lanka number of jobs would be problematic: It may take Lao PDR (rural) Macedonia, FYRSri Lanka (rural) time for new jobs to be created, and even if they are,

China, Yunnan Province retraining takes time and may be diffi cult.

Source: WDR 2016 team, based on STEP household surveys (World Bank, various years). Data at http:// Yet, even if technologically feasible, large-scale net bit.do/WDR2016-Fig2_20. job destruction due to automation should not be a con- Note: ICT = information and communication technology. cern for most developing countries in the short term. Even in the United States, on average, there were no net employment declines in local labor markets most sus- in China rose from 19 to 27 percent (see fi gure 2.15, ceptible to automation (that is, those specialized in rou- panel b). Yet manufacturing employment in China tine, task-intensive occupations).140 New jobs and new is also polarizing, for the increase in routine labor tasks in existing occupations are created. Machines is explained by the mechanization of agriculture. and digital technology are not perfect or even good Moreover, across most countries with relevant data, substitutes for many tasks (at least not yet), especially employment is polarizing even within services, sug- those requiring adaptability, common sense, and cre- gesting an additional effect in skill demand over and ativity.141 The expansion of automated teller machines above what can be explained by trade or the structural (ATMs) in banks went hand-in-hand with an expan- transformation of developing economies.131 This evi- sion of bank transactions, branches, and employment. dence is also consistent with the evidence for OECD Cashiers continue to do some of the things that ATMs countries linking changes in skills requirements to do, but they also do other things, such as client support, technological changes, even within occupations.132 where human interaction remains important.142 Full automation of jobs takes time, even in the Labor-saving technologies: Automation developed world. In the United Kingdom, an exoge- and job displacement nous increase in ICT investment between 2000 and There is concern, especially in advanced countries, 2004 led to a short-term increase in the demand for that technology is killing jobs and depressing nonroutine skills, but only to the limited substitution wages.133 Manufacturers are using machines that of routine workers during that period.143 Why? Because substitute for workers in warehouses or auto plants. it takes time to make the necessary organizational More than 200,000 industrial robots come into use changes (chapter 1), and because labor reorganizations each year, and that number is rising.134 Increasingly, tend to happen in periods of recession rather than in automation is taking place in services.135 In call cen- booms.144 Not all disruptive technologies are adopted ters, technology can answer routine customer service quickly, implemented fully, or yield immediate bene- requests. In retail, technology and “big data” suggest fi ts.145 Barriers to technology adoption, lower wages, what to buy. Software is handling accounting, trans- and a higher prevalence of jobs based on manual dex- lations, and paralegal services. Travel agents have all terity in developing countries mean that automation but disappeared, with three-quarters of all travel in is likely to be slower and less widespread there (see the United States now booked online.136 In the public fi gure 2.24). But as wages rise, and in countries with EXPANDING OPPORTUNITIES 127

Figure 2.21 Nonroutine analytical and socioemotional skills are becoming more important, especially in jobs performed by younger cohorts

Evolution of the skills intensity of jobs, measured as mean skills percentile of base year, Brazil, Malaysia, and Macedonia, FYR, various years, 2001–11

a. Brazil (cohort born before 1955) b. Brazil (cohort born after 1974) 70 70

60 60

50 50

40 40

Change in skills intensity of jobs 30 Change in skills intensity of jobs 30 2002 2003 2004 2005 2006 2007 2008 2009 2002 2003 2004 2005 2006 2007 2008 2009

c. Malaysia (cohort born before 1955) d. Malaysia (cohort born after 1974) 70 70

60 60

50 50

40 40 Change in skills intensity of jobs Change in skills intensity of jobs 30 30 2001 2002 2003 2004 2005 2006 2007 2009 2010 2001 2002 2003 2004 2005 2006 2007 2009 2010

e. Macedonia, FYR (cohort born before 1955) f. Macedonia, FYR (cohort born after 1974) 70 70

60 60

50 50

40 40 Change in skills intensity of jobs Change in skills intensity of jobs 30 30 2007 2008 2009 2010 2011 2007 2008 2009 2010 2011

Nonroutine cognitive and interpersonal skills Manual skills Routine cognitive skills

Sources: WDR 2016 team, based on Socio-Economic Database for Latin America and the Caribbean (SEDLAC) (CEDLAS and the World Bank); East Asia and Pacifi c Poverty (EAPPOV) Database (World Bank, various years); and Europe and Central Asia Poverty (ECAPOV) Database (World Bank, various years); and following Autor, Levy, and Murnane 2003; Acemoglu and Autor 2011; Aedo and others 2013; Arias and others 2014. Data at http://bit.do/WDR2016-Fig2_21. Note: The y-axis represents the percentile of the skill distribution for jobs held by each cohort in any given year, with respect to the corresponding median skills intensity of jobs held by that cohort in the initial year. An increase means that jobs increased in intensity in that particular skill. For Malaysia (panels c and d), data for 2008 were unavailable. large manufacturing or offshored activities, there is the Industrial Revolution to Keynes during the Great more (and faster) scope for automation. Depression—have gone unrealized (box 2.8).146 The Concerns about automation are not new. Past assembly line, after all, replaced the artisans mak- fears of technology leading to mass unemployment ing carriages. Back in the 1960s, with the creation in (and boredom)—from the Luddites in the midst of the United States of the National Commission on 128 WORLD DEVELOPMENT REPORT 2016

Figure 2.22 New economy skills, beyond levels of fostering entrepreneurship and improving produc- education, pay off tivity and the allocation of resources, have led in due

Urban Vietnam: Average return to diff erent task combinations, controlling for time to more jobs elsewhere. education and demographics, 2012 Will this time be different? Large factories and electrifi cation did lead, as now, to a polarization of 25 employment by hollowing out the middle of the skill distribution. The share of employment of blue-collar 20 workers in manufacturing fell from 39 percent in 1850 to 23 percent in 1910, as new capital goods allowed 15 factory owners to unbundle and simplify tasks that could now be performed by unskilled workers. Elec- 10 trifi cation increased the relative demand for workers intensive in clerical and managerial skills compared

Percent 5 with manual and dexterity skills among white-collar workers. Among blue-collar workers, it increased the demand for manual workers relative to the demand 0 for workers performing tasks intensive in dexterity needed to operate machines before electrifi cation.147 –5 Despite these similarities, the biggest difference from past waves of technological progress is that –10 the polarization of the labor market today is affect- Analytical*** Interactive* Manual Analytical Interactive*** Manual ing both blue-collar and white-collar workers.148 It is Nonroutine Routine probably easier for white-collar workers to transition

Source: Bodewig and others 2014. Data at http://bit.do/WDR2016-Fig2_22. to other white-collar jobs, but in the aggregate, there Note: Returns are estimated using a wage regression that controls for education, sex, experience, and may be fewer well-paying jobs for a large and diverse economic sector. pool of potentially dislocated workers. Even if all Signifi cance level: * = 10 percent, *** = 1 percent. those jobs do not fully disappear—unlikely in a short period—they will be signifi cantly transformed. Technology, Automation, and Economic Progress, And here is where a second lesson from history there were also concerns about automation in this is relevant. Individuals and governments adapted to wave of technological change. Unemployment fears technological change, but this process took time and have gone unrealized because new technologies, by required deep institutional changes in education,

Figure 2.23 Digital technologies go hand in hand with nonroutine new economy skills

ICT intensity and skills intensity, by occupation

a. ICT intensity and nonroutine analytical skills b. ICT intensity and routine manual skills 5 5

4 4

3 3

2 2 Routine manual (scale 1 to 5) 1 1 Nonroutine analytical (scale 1 to 5) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 8 9 10 11 12 13 Intensity of ICT use at work Intensity of ICT use at work

Source: Monroy-Taborda, Moreno, and Santos, forthcoming, for the WDR 2016, based on STEP household surveys (World Bank, various years). Data at http://bit.do/WDR2016-Fig2_23. Note: Ninety-fi ve percent confi dence intervals. The y-axis is a standardized score (from 1 to 5) that refl ects the intensity of the use of the particular type of skills as estimated by Autor, Levy, and Murnane (2003) and expanded by Acemoglu and Autor 2011. The intensity of ICT use is an index between 0 (no use of technology at work) and 19 (most use of technology at work). ICT intensity is averaged by occupation. ICT = information and communication technology. EXPANDING OPPORTUNITIES 129

Figure 2.24 From a technological standpoint, two-thirds of all jobs are susceptible to automation in the developing world, but the eff ects are moderated by lower wages and slower technology adoption

Estimated share of employment that is susceptible to automation, latest year

100

80

60

40

be computerized (%) be computerized 20 Share of employment that can 0

India Malta Latvia BoliviaNepal Serbia AngolaChina OECD Cyprus Georgia UkraineNigeria EthiopiaBulgaria Panama Ecuador Albania Croatia Tajikistan Mongolia Paraguay MauritiusMalaysia RomaniaThailand Lithuania Uruguay NicaraguaCambodia Costa Rica Seychelles Argentina Uzbekistan El SalvadorGuatemalaBangladeshSouth Africa Kyrgyz Republic Macedonia, FYR

West Bank and Gaza Dominican Republic

Adjusted (technological feasibility + adoption time lags) Unadjusted (technological feasibility)

Sources: WDR 2016 team, based on STEP surveys (World Bank, various years); Central Asia World Bank Skills surveys (World Bank, various years); Survey-based Harmonized Indicators Program (SHIP) (World Bank, various years); Socio-Economic Database for Latin America and the Caribbean (SEDLAC) (CEDLAS and the World Bank); South Asia Region MicroDatabase (SARMD) (World Bank, various years); Europe and Central Asia Poverty (ECAPOV) Database (World Bank, various years); East Asia and Pacifi c Region Poverty (EAPPOV) Database (World Bank, various years); the I2D2 dataset (International Income Distribution Database; World Bank, various years); ILO Laborsta database (various years); the National Bureau of Statistics of China (various years); Frey and Osborne 2013; Comin and Hobjin 2010. Data at http://bit.do/WDR2016-Fig2_24. Note: The unadjusted probabilities of automation for occupation are from Frey and Osborne (2013), weighted by employment. The adjusted probabilities account for the slower pace of tech- nology adoption in poorer countries, using the adoption lag of earlier technologies (Comin and Mestieri 2013). See Monroy-Taborda, Moreno, and Santos, forthcoming, for the WDR 2016. OECD = Organisation for Economic Co-operation and Development.

Box 2.8 Concerns about technological unemployment are not new

We are being affl icted with a new disease of which some readers may not yet have heard the name, but of which they will hear a great deal in the years to come—namely, technological unemployment. This means unemployment due to our discovery of means of economising the use of labour outrunning the pace at which we can fi nd new uses for labour. —John Maynard Keynes “Economic Possibilities for Our Grandchildren,” 1931

The situation will have been made the more serious by the advances of automation. The world of A.D. 2014 will have few routine jobs that cannot be done better by some machine than by any human being. Mankind will therefore have become largely a race of machine tenders. Schools will have to be oriented in this direction . . . It is not only the techniques of teaching that will advance, however, but also the subject matter that will change. Even so, mankind will suff er badly from the disease of boredom . . . The lucky few who can be involved in creative work of any sort will be the true elite of mankind, for they alone will do more than serve a machine. —Isaac Asimov “Visit to the World’s Fair of 2014,” 1964 social protection, and labor regulations.149 The answer in the economy (chapter 1). It depends on how individ- to whether things may turn out differently this time, uals, fi rms, and policy makers respond to the change while unsatisfactory, is: “It depends.” It depends on in skills requirements. And it depends on how well the the ability and speed of creating new jobs elsewhere social protection system supports dislocated workers. 130 WORLD DEVELOPMENT REPORT 2016

The future of jobs 19th century in the United States led to an upskilling of employment as increases in middle-skilled sales Be an expensive complement (stats knowhow) to and clerical employment compensated for the decline something that’s getting cheaper (data). in blue-collar jobs. With electrifi cation, displaced —Hal Varian, Chief Economist, Google, 2014 workers typically moved to lower-skilled jobs at lower wages, such as truck drivers, but large increases Technological progress makes the jobs challenge in new middle-skill employment outweighed the decline in blue-collar employment.154 While the steam more complex. Digital and mechanical technologies, engine in the 18th century led to the disappearance and deindustrialization, mean that the manufactur- of coachmen for horse-drawn carriages155 and to the ing sector is likely to generate fewer jobs than in the substitution of workers in mining, in due time it also past, especially for unskilled workers.150 ICT manufac- was at the center of steam-powered boats, locomo- turing can be expected to create jobs, although mostly tives, and automobiles. Mechanics, road building and high-skilled and likely concentrated in specifi c coun- maintenance, and dealerships have created many jobs tries, as now. Since manufacturing jobs have been an of diverse skill profi les ever since. important source of better-paying jobs as workers Yet as a result of this process of job creation and move out of agriculture, this is an important devel- destruction, technological change disrupts labor mar- opment. The service sector is also increasingly being kets and can hurt individuals whose skills are substi- automated. Of particular concern are jobs that have tuted by technology, because they often do not have so far actually grown thanks to digital technologies, the skills required in many of the new jobs. Even for such as low- and middle-skilled call center jobs. those who stay within the same occupations, jobs will Employment in the ICT service industry, even more be transformed, requiring modern skills. The speed of than its manufacturing counterpart, is expected to these changes appears to be accelerating, intensify- grow, but also among higher-skilled workers in lead- ing and the pace of labor market ing countries. changes. Future employment growth is likely to come from Since digital technologies have different applica- jobs that cannot be fully or partially automated, largely bility to different kinds of work, the extent of disrup- outside the ICT industry. New jobs are also likely to tion across countries will refl ect differences in eco- emerge in the digital economy—in the analysis of data nomic and occupational structures. Numerical clerks such as data scientists, in the development and main- or secretaries, often users of digital technologies, also tenance of apps and other software, and in support ser- perform many tasks that can be easily automated. vices. New opportunities will also arise in ICT-enabled Managers or software developers, by contrast, while services, such as the on-demand economy. But the intensively using digital technologies, also comple- signifi cance for employment generation, especially in ment them well, so they are not easily substituted by most developing countries, is likely to be small. machines. And for occupations that use little technol- The potential for employment creation lies in the ogy, some are hard to automate, such as hairdressers, rest of the economy, as digital technologies allow while others could be automated, such as assemblers businesses to expand. Among the low-skilled, some (fi gure 2.25). services that must be delivered face-to-face or require More advanced economies can expect larger dis- awareness and situational adaptability (housekeep- ruptions in the near future since they use more tech- ers, hairdressers) are likely to grow. Among the nology at work and are experiencing faster changes in high-skilled, occupations will rely on modern skills skill requirements (fi gure 2.26).156 And while they have involving creativity and social interactions. Some smaller shares of employment in routine occupations observers call this the polarization in “the high-tech, susceptible to automation, their higher wages make it high-touch” economy.151 easier for automation to be economically viable. Low- Although impossible to predict in advance, jobs are and middle-income countries can also expect substan- likely to arise in new industries and occupations. New tial disruptions, albeit with a time lag, given their rapid industries arising from digital technologies since technological adoption and large number of workers 2000 account for only 0.5 percent of employment in in routine occupations. Their low skill base suggests the United States,152 and there is evidence of (young) important challenges ahead, however. In poorer coun- skilled workers having to take on less-skilled jobs.153 tries, where wages are lower and technological adop- But historically, economies have been able to create tion is slower (chapter 1), the disruptions are likely to enough jobs through technological change. The initial arrive more slowly, giving more time for policies and labor market polarization caused by factories in the institutions to adapt. All this has implications for EXPANDING OPPORTUNITIES 131

Figure 2.25 The interaction between technology and jobs varies by occupation

Probability of being computerized and intensity in use of ICT at work, by occupation

Average intensity in use of ICT at work 1.0 Agriculture workers Waiters and bartenders Clerks Numerical clerks 0.9 Subsistence farmers Secretaries (general) Cooks Metal-processing operators Tellers Food processors 0.8 Assemblers Finance professionals

Drivers Shop salespersons 0.7 Mining and Transport and storage laborers construction laborers Sales and purchasing Cleaners 0.6 agents and brokers Protective services Garment workers Business services agents workers 0.5 Average probability of being computerized Sales, marketing, and public 0.4 Handicraft relations professionals workers Administration managers 0.3

Retail and wholesale Hairdressers, Administration Managing directors 0.2 trade managers professionals Software and applications Probability of being computerized beauticians and chief executives

(technology substituting for workers) developers and analysts Mining, manufacturing, and Legislators 0.1 construction supervisors Street and related Hotel and restaurant Information and communication Secondary Medical doctors Legal professionals service workers managers technology service managers education teachers Sales managers 0 0246810121416 Intensity in use of ICT at work (technology complementing workers)

Sources: WDR 2016 team, based on STEP household surveys (World Bank, various years) and Frey and Osborne 2013. Data at http://bit.do/WDR2016-Fig2_25. Note: The probability of being computerized is obtained from Frey and Osborne (2013). ICT intensity is an index between 0 (no use of technology) and 19 (most use of technology). ICT = information and communication technology. The red lines represent the average values of ICT intensity (x-axis) and of computerization (y-axis) across the pooled sample of 10 developing countries with STEP household surveys.

whether countries need not just to develop modern changes in labor demand have in turn implications skills among children and youth, but also to come up for earnings. But employment polarization does not with a strategy for the retraining and lifelong learning necessarily mean wage polarization. of the current stock of (older) workers. Three interrelated factors mediate the impact of The challenge is to start reforms today to max- digital technologies on earnings: imize the digital dividends and to prepare for any disruptions. Even if expected labor market changes • Complementarity with technology. Workers in jobs are similar in Malaysia and South Africa, Poland and that use and complement technology are likely Turkey, or Finland and Italy, skill systems vary widely to see both an increase in employment and an and not all are prepared to equip workers with skills increase in earnings because of higher productivity. that complement technology. This process needs to This is the case for workers who use nonroutine start very early in life, and education and training cognitive skills and ICT skills. Workers in routine systems are notoriously diffi cult to change. So, any occupations, however, will see less demand for reform takes many years to have effects, which is why their skills, bringing down both their employment there is a race between skills and technology. Some and their wages. skill systems are well-positioned, but for many others, • Product demand. If workers produce goods or ser- skills—and hence, people—are losing the race. vices that consumers keep buying as they get richer or as the price declines, increases in productivity Making the internet work for everyone can translate into increases in wages. This is often To design policy responses to technological change, it the case for workers with nonroutine skills pro- is important to understand who the changes are likely ducing, say, knowledge, management expertise, or to affect the most, and how the process plays out both medical services. If not, increases in productivity in terms of employment and earnings. As discussed, can lead to lower employment and earnings in that employment is likely to polarize, with routine occupa- sector because fewer workers can satisfy demand, tions losing ground to nonroutine occupations. These as for many agricultural goods. 132 WORLD DEVELOPMENT REPORT 2016

Figure 2.26 The key policy challenge: Adapting the Table 2.5 Expected impacts of skills agenda to expected labor market disruptions technological change on employment and earnings Expected labor market disruption and quality-adjusted years of education Expected impact on Average expected Type of occupation labor market disruption 13 (by skills intensity) Employment Earnings CHE Nonroutine Positive Positive cognitive FIN NLD Routine cognitive Negative Negative IRL BELGBR NOR and manual EST DEU ISL CYP AUS SWE MYS SVN DNKMLT Nonroutine manual Positive Negative LUX FRA LTU LVACZE ISR Source: WDR 2016 team, based on Autor 2014. HUN AUT ESPHRV 5.6 ROUPOLSVK CHL PRT UKR BGR Average MUSCRI SYC ITA

Adaptability GEO ALB MKD quality-adjusted SRB GRCTJK PAN The young, the better educated, and those already CHN years of education UZBTHAURYARG SLVMNG ECU better off are most likely to benefi t from digital tech- IND TUR MEXKGZ ETH ZAF BOL nologies—with older workers, those with less educa- BGD NPL DOM (quality-adjusted years of education) KHM GTM NICNGA tion, and the poor falling behind. The former group is PRY more likely to have more advanced skills—especially AGO 0 cognitive and ICT skills—regardless of their occu- Low (0)Expected labor market disruption High (1) pation or work status.158 In addition, these groups GDP income group are disproportionally likely to be in, or to move into, High Upper-middle Lower-middle Low occupations that pay well and are likely to grow in the Sources: WDR 2016 team, based on STEP surveys (World Bank, various years); Central Asia World Bank future—those intensive in nonroutine skills (fi gure Skills surveys (World Bank, various years); SHIP (World Bank, various years); SEDLAC (Cedlas and 159 the World Bank); SARMD (World Bank, various years); ECAPOV (World Bank, various years); EAPPOV 2.27). Recent evidence from the United States shows (World Bank, various years); the National Bureau of Statistics of China (various years); ILO Laborsta that there has been a marked decline in the rate at database (various years); World Development Indicators (World Bank, various years); World Economic Forum’s Competitiveness Index (WEF, various years). Data at http://bit.do/WDR2016-Fig2_26. which workers transition into routine employment Note: Labor market disruption is an index that goes from 0 (no disruption) to 1 (highest disruption). It is (particularly among the young) but that women and the standardized summation of two components, equally weighted: the probability of an average job those with higher education levels have found it being computerized (Frey and Osborne 2013, and adjusting for adoption lags), and the intensity of ICT use at work. For each country, the ICT intensity of employment corresponds to the average for countries at easier to adjust to these changes by moving into the the next level of development, to be more forward-looking. The quality-adjusted years of education are high-paying, nonroutine cognitive jobs.160 constructed by adjusting average years of education for each country with the World Economic Forum’s quality-of-education indicator. For example, if a country has, on average, 10 years of education and A big challenge for policy makers, especially in rap- scores 3.5 on the indicator (which ranges from 0 to 7), its quality-adjusted years of education are 5. See idly aging societies, is managing skill obsolescence. Monroy-Taborda, Moreno, and Santos, forthcoming, for the WDR 2016. GDP = gross domestic product. Recall that the surge in the demand for new economy skills has been concentrated among young workers • Labor supply. The higher the skill requirements for a (see fi gure 2.21). Digital technologies accelerate the job, the more diffi cult it is for new workers to enter depreciation of skills and work experience, affecting that market. So, higher demand for workers would especially older workers (box 2.9). But obsolescence is translate into higher wages. If, however, it is easy to not destiny or the same for all types of skills. Most liter- retrain for a new job or skill requirements are low, ature argues that younger workers have a comparative there can be downward pressure on wages because advantage in tasks where problem solving, learning, of increased competition. Workers in nonroutine and speed are important (“fl uid” abilities), and older cognitive occupations are likely to see their higher workers have an advantage when experience and ver- productivity rewarded as higher earnings because bal abilities matter more (“crystallized” abilities).161 But entry barriers are high. But low-skilled workers evidence from Germany shows that workers in their in nonroutine manual occupations are likely to 50s experienced a more rapid growth in tasks intense see their earnings fall over time as middle-skilled in fl uid cognitive skills than those in their 30s.162 Box workers in routine occupations are displaced and 2.10 examines the gender impacts in more depth. start competing for the available jobs in low-paying occupations (table 2.5).157 A policy agenda Digital technologies improve overall welfare and can Therefore, the main winners from technological reduce poverty, but without complementary policies, change will have and use new economy skills and gain many benefi ts can go unrealized and inequality can employment in nonroutine cognitive occupations. increase. To capitalize on the benefi ts—and to do EXPANDING OPPORTUNITIES 133

Figure 2.27 The less educated and the bottom 40 percent of the welfare distribution are most vulnerable to technological changes in the labor market

Ratio of employment by occupation type to total employment

a. By educational attainment b. By socioeconomic status 3.0 1.4

2.5 1.2

1.0 2.0 0.8 1.5 Ratio Ratio 0.6 1.0 0.4

0.5 0.2

0 0 Nonroutine Routine cognitive Nonroutine Nonroutine Routine cognitive Nonroutine cognitive or manual manual cognitive or manual manual Secondary education or less Tertiary education Bottom 40% Upper 60%

Source: WDR 2016 team, based on the I2D2 dataset (International Income Distribution Database; World Bank, various years). Data cover 117 countries. “Bottom 40%” and “Upper 60%” refer to the welfare distribution (either of income or consumption) of individuals’ households. Classifi cation of occupations follows Autor 2014. Data at http://bit.do/WDR2016-Fig2_27. Note: A ratio higher than 1 means that workers with the given level of education are disproportionately likely to be in the given occupation type. A ratio lower than 1 means that workers are relatively unlikely to be in a given occupation type. so without leaving people behind—internet access in seizing the opportunities that the internet gener- (chapter 4) needs to be complemented with an “ana- ates (chapter 5). log” policy agenda. In addition to supporting entre- The most crucial element is skills development. preneurship and innovation to expand businesses A modern economy requires workers with mod- and job opportunities, this agenda should ensure that ern skills. For people to become online workers education and training systems, labor regulations, and traders, farmers to use technology to become and social protection institutions support all workers more productive, or for workers to access modern,

Box 2.9 The challenge of keeping up with new technologies in Mexico

Technology is making many skills obsolete and reducing explanations for the decline in relative wages of skilled the returns to experience, especially among older workers. workers: low-quality tertiary educated graduates, a skills In Mexico, as in other Latin American countries, wage mismatch among young workers, or skills obsolescence inequality has declined since 2000. A declining skill pre- among older workers. mium has been one of the drivers of this trend because the New research fi nds that the fall in the skills premium is wages of the low-skilled rose and those of the high-skilled driven at least partly by skills obsolescence among older fell. The average hourly wage for college-educated workers workers. First, the earnings of the oldest cohort, above age fell 2 percent a year between 2001 and 2014. 50, declined fastest (40 percent during the period). But for Why have high-skilled wages declined in Mexico, when the youngest cohort—aged between 23 and 30—earnings in most countries, especially outside of Latin America, increased. Second, earnings start to decline at a younger returns to tertiary education continue to rise (despite age over time. If age is a proxy for work experience, this increasing educational attainment everywhere)? It could means that, among older workers, the return to education simply be that the supply of educated workers has out- and the return to experience are falling. This could be evi- paced demand for them. But there are three alternative dence of an accelerated skills obsolescence.

Source: WDR 2016 team, based on Campos-Vazquez, Lopez-Calva, and Lustig, forthcoming. 134 WORLD DEVELOPMENT REPORT 2016

Box 2.10 Digital technologies and economic opportunities: A gender lens

Digital technologies can empower women economically agency and control over economic resources, and their and socially. Because social norms and time and mobility safety. For example, in Niger, greater privacy and control constraints are often more severe for women than for men, of mobile transfers, compared to manual cash transfers, women could benefi t greatly from technology.a shifted intrahousehold decision making in favor of women, Digital technologies can reduce gender gaps in labor the recipients of the transfer.k force participation by making work arrangements more But in many countries, gender gaps in technology use, fl exible, connecting women to work, and generating new and in working in the ICT sector, remain vast. Over 1.7 billion opportunities in online work, e-commerce, and the sharing women in low- and middle-income countries do not own economy. In Elance, a platform for online work now part of mobile phones. Women in those countries are 14 percent Upwork, 44 percent of workers are women, compared to less likely to own a mobile phone than men, on average. an average of 25 percent in the nonagricultural economy Women in South Asia are 38 percent less likely to own a globally. Business process outsourcing in India employs phone than men.l Barriers to access can be particularly more than 3.1 million workers, 30 percent of them women. salient in the case of the internet, especially in poor and Four in ten online shop owners on Alibaba are women. remote localities where access is predominantly outside Moreover, technology can help improve women’s access the home, and where social norms for socializing or safety and accumulation of productive assets. The digitization of concerns can become a barrier.m In Africa, for example, land registries, for example, can be particularly benefi cial women are 50 percent less likely to use the internet than for women.b These improved opportunities, in turn, can men.n In addition to ownership and access, lack of control increase investments in young girls’ human capital.c over the use of the technology can be an additional barrier Technology-driven shifts in skills demand can reduce for women. In the Arab Republic of Egypt and in India, for wage gaps, especially among the better educated. Women example, 12 percent of women stated that they did not are well positioned to gain from a shift in employment access the internet more often because they did not think it toward nonroutine occupations, and away from physical was appropriate, and more than 8 percent did not access it work.d In Germany and the United States, the use of digital more often because family or friends would disapprove.o In technologies at work explains a signifi cant share of the the labor market, women are also much less likely to work increase in female labor force participation and employ- in the information and communication technology (ICT) ment in the past few decades.e Similar changes have taken sector or in ICT occupations, which are well paid. These lat- place in Brazil, Mexico, and Thailand, and also explain ter gaps partly refl ect women’s low participation in science, a large part of the reduction in gender wage gaps.f New technology, engineering, and math (STEM) education, itself technologies level the playing fi eld, particularly among the a product of early gender-based biases in formal and infor- better educated, who take on jobs that use their compara- mal education (chapter 5). tive advantage in nonphysical work.g Recent evidence from Even with gender parity in ownership, access, and the United States shows that it is the better educated and control over digital technologies, gains are not automatic. women who are able to get the high-paying jobs intensive They need to be complemented with analog changes that in nonroutine skills as middle-skilled jobs shrink.h address the underlying barriers to women’s employment, Digital technologies also impact women’s voice and voice, and agency. In rural South Africa, for example, mobile agency. Increased access to information can aff ect gender phones increased employment mostly among women, as norms and aff ect aspirations, often faster than expected.i long as they did not have large child care responsibilities.p Social media is an additional outlet for women to participate In addition, technologies can have gender-diff erentiated in public discussions and voice their opinions (spotlight eff ects, as shown throughout this chapter. This suggests 3). “My Dress, My Choice” in Kenya, a social media move- that interventions around digital technologies would ment against female violence that mobilized thousands of become more eff ective if they were more gender-informed, Kenyans, including through street protests, eventually led including by having women participate early on in the to changes in relevant laws.j Some of the new innovations design of interventions and of the technologies themselves. of the digital economy, such as digital payments, mobile Moreover, by circumventing, rather than eliminating, money, and taxi sharing rides, can also increase women’s some of the barriers to employability faced by women,

(Box continues next page) EXPANDING OPPORTUNITIES 135

Box 2.10 Digital technologies and economic opportunities: A gender lens (continued)

the internet could actually delay necessary reforms. For seen as unacceptable for women or if there is no availability example, home-based work could help connect women of aff ordable child care, technology could end up delaying to work in environments where social norms or child care fundamental reforms. Addressing these underlying con- responsibilities are a barrier to women working outside straints remains key to the gender and overall economic the home. But if working outside the home continues to be agenda.

a. World Bank 2011. b. World Bank 2014b. In Rwanda, the Land Tenure Regularization Programme demarcated and digitized 10 million plots. Households that registered their land were more likely to invest in it, and this eff ect was twice as strong for female-headed households (Ali, Deininger, and Goldstein 2014). c. Oster and Millett 2013. d. Rendall 2010; Weiberg 2000. e. Black and Spitz-Oener 2007; Rendall 2010; Weiberg 2000. f. Rendall 2010; Autor and Price 2013; Black and Spitz-Oener 2007. g. WDR 2016 team, based on STEP household surveys (World Bank, various years). h. Cortes and others 2014. i. La Ferrara, Chong, and Duryea 2012; Jensen and Oster 2009. j. Seol and Santos 2015. k. Aker and others 2014. l. GSMA 2015. m. Gomez (2014) shows, for developing countries, that women—unlike men—prefer using the internet in public libraries rather than in private cybercafés because they are safer and despite poorer service. n. WDR 2016 team calculations, based on Research ICT Africa surveys (various years). o. and Dalberg Global Development Advisors 2012. p. Klonner and Nolen 2010.

better-paying jobs, skills need to be upgraded. Current taking a wide perspective, it is also used to include and future workers need to develop the lifelong cog- gains to consumers. nitive, technical, and socioemotional skills required 3. WDR 2016 team, based on STEP surveys (World of a well-educated worker in the 21st century. Work- Bank, various years); Central Asia World Bank Skills ers also need to be capable of processing the ever- surveys (World Bank, various years); Survey-based Harmonized Indicators Program (SHIP) (World increasing information available on the internet. Bank, various years); Socio-Economic Database Building these skills requires actions affecting all for Latin America and the Caribbean (SEDLAC) relevant environments for learning: families, schools, (CEDLAS and the World Bank); South Asia Region universities, training systems, and fi rms. Given the MicroDatabase (SARMD) (World Bank, various speed of technological changes, these skills will also years); Europe and Central Asia Poverty (ECAPOV) require constant updating throughout the life cycle as Database (various years); East Asia Pacifi c Poverty workers prepare for careers that last more than one (EAPPOV) Database (World Bank, various years); job. Digital technologies themselves can help (sector the I2D2 dataset (International Income Distribu- focus 2 and chapter 5). Complementary reforms are tion Database) (World Bank, various years); ILO also needed in , social protection, and labor Laborsta database (ILO, various years); and the National Bureau of Statistics of China (various market institutions to facilitate the transition of years). Automation probabilities adapted from Frey workers from old economy jobs to new economy jobs, and Osborne (2013). and address the distributional consequences of the 4. WDR 2016 team calculations, based on ILO Key digital revolution. Indicators of the Labour Market (KILM; various years), ILO Laborsta database (various years), World Notes Bank’s International Income Distribution Database (I2D2; various years), and the National Bureau of 1. World Bank 2014c. Statistics of China (various years). For more details, 2. Throughout this chapter, “opportunities” refer to see fi gure 2.15. people’s short- and long-term capacity to generate 5. WDR 2016 team calculations, based on World Devel- income (Bussolo and Calva 2014). In addition, and opment Indicators (World Bank, various years). 136 WORLD DEVELOPMENT REPORT 2016

6. ITU’s (International Telecommunication Union) 33. CGAP 2014. World Telecommunication/ICT Indicators (ITU, 34. Fernandes and others 2015. various years), Gallup World Poll, and Eurostat 35. Dutz and others 2015, for the WDR 2016. Information Society Statistics (EC 2015). 36. Gaggl and Wright 2014. 7. This, like the rest of the chapter, is based on house- 37. Akerman, Gaarder, and Mogstad 2015. hold survey data. Subscription data, often based on 38. Blinder and Krueger 2013. estimates, are widely available but have upward and 39. Kennedy and others 2013. downward biases. Although less readily available, 40. NASSCOM 2014. household surveys can better account for sharing of 41. Jensen 2012. mobile phones, or for one individual having more 42. IBM 2014. than one subscription or phone. 43. See http://elance-odesk.com/online-work-report 8. Handel 2015, for WDR 2016; Aker 2010b. -global, accessed October 2, 2014. 9. WDR 2016 team, based on Research ICT Africa sur- 44. Agrawal and others 2013. veys (various years). 45. Heeks and Arun 2010; Kennedy and others 2013; 10. Veeraraghavan, Yasodhar, and Toyama 2009. Monitor Inclusive Markets 2011. 11. WDR 2016 team, based on the Argentina National 46. Samasource 2015. Digital Divide Data has more Institute of Statistics and Census, Brazilian Inter- than 1,300 employees, with 10 percent of its data net Steering Committee, Colombia Directorate of management operators having physical disabilities Statistical Methodology and Production, European (Digital Divide Data 2014). Ruralshores is active in Commission Eurostat database (various years), remote rural areas in India and has 2,500 employees Mexico National Institute of Statistics and Geogra- (http://ruralshores.com/about.html). phy, and Uruguay National Institute of Statistics. 47. Imaizumi and Santos, forthcoming. 12. Eurostat (EC, various years), for the WDR 2016. 48. China Association for Employment Research 2014. 13. WDR 2016 team, based on Brazilian Internet Steer- 49. Schaefer-Davis 2005. ing Committee and Mexico National Institute of 50. See https://www.etsy.com/about/?ref=ftr, accessed Statistics and Geography. May 15, 2015. 14. WDR 2016 team, based on Research ICT Africa sur- 51. Based on an online survey of 60 countries world- veys (various years). wide done by Nielsen in 2013 (Van Welsum 2015). 15. WDR 2016 team, based on Brazilian Internet Steer- 52. See https://www.airbnb.com/about/about-us, ing Committee and Mexico National Institute of ac cessed March 11, 2015. Statistics and Geography. 53. See http://www.gravitytank.com/pdfs/info_graphics 16. WDR 2016 team, based on Gallup World Poll, vari- /SharingEconomy_web.pdf, as cited in Van Welsum ous years. 2015. 17. WDR 2016 team calculations, based on Research 54. In a study of Uber in the United States, Hall and ICT Africa surveys (various years). Krueger (2015) show that drivers aged 18–29 years 18. ICT surveys for Argentina, Brazil, Colombia, Mex- are 19 percent of all Uber drivers, compared to 8.5 ico, and Uruguay; and Eurostat (EC, various years). percent among regular taxi drivers and chauffeurs. 19. WDR 2016 team calculations, based on Research Female drivers are 13.8 percent of Uber drivers, ICT Africa surveys (various years). compared to 8 percent elsewhere. At the same time, 20. Gomez 2014. however, Uber drivers are less likely than tradi- 21. Ritter and Guerrero 2014. tional employees to have health insurance, and half 22. Galperin and Viecens 2014; Pimienta, Prado, and leave Uber within 50 weeks. Blanco 2009. 55. Montenegro and Patrinos 2014. 23. De los Rios 2010. 56. Bagues and Sylos 2009; Nakamura and others 2009; 24. Atasoy 2013. Stevenson 2009. 25. Ritter and Guerrero 2014. 57. See http://press.linkedin.com/. 26. Klonner and Nolen 2010. 58. WDR 2016 team, based on STEP household surveys 27. OECD Key Economic Indicators Database, latest (World Bank, various years). year available (circa 2011). 59. Kuhn 2014; Raja and others 2013. 28. Berger and Frey 2014. 60. Dammert, Galdo, and Galdo 2014. 29. Brynjolfsson and McAfee 2014. 61. Mang 2012. 30. WDR 2016 team, based on STEP household surveys 62. Kuhn 2014; Kroft and Pope 2014. (World Bank, various years). 63. Kuhn and Mansour 2014. 31. WDR 2016 team, based on STEP household surveys 64. Imaizumi and Santos, forthcoming, for the WDR (World Bank, various years). 2016. 32. Moretti and Thulin 2013 for the United States; 65. Arias and others 2014. Maloney and Valencia 2015 for Turkey. 66. OECD 2011. EXPANDING OPPORTUNITIES 137

67. World Bank 2011. show more turnover among young workers in 68. See http://blog.jetblue.com/index.php/2013/09/05 industries intensive in ICT in Brazil. See also /unpacked-working-from-home/, accessed June 19, Stevenson 2009. 2015. 103. Handel 2015, for the WDR 2016. 69. Bloom and others 2014. 104. World Bank 2012, 2013. 70. See http://www.philstar.com/education-and-home 105. Forman, Goldfarb, and Greenstein 2012. /2013/06/13/953332/fi lipino-teachers-uruguay. 106. Eden and Gaggl 2014; Karabarbounis and Neiman 71. Muto and Yamano 2009. 2013. Eden and Gaggl (2015, for the WDR 2016) sug- 72. e-Choupal, “The Status of Execution.” http://www gest that while the fall in the routine share of labor .itcportal.com/businesses/agri-business/e-choupal is linked to technological change, the shift in the .aspx. labor share compared with that of capital may be a 73. World Bank 2014a; Demirgüç-Kunt and others 2015. more complex story. In the United States, housing 74. Jack and Suri 2014. prices have been identifi ed as a key driver in the 75. Fingerprint scanning among paprika farmers in rise of capital shares (Bonnet and others 2014). This Malawi increased repayment rates (IFPRI and chapter focuses on the fall of routine labor within World Bank 2010). the labor share, as this is more clearly linked to 76. For a discussion on personal networks and technological change. labor markets, see Granovetter (1973) and Calvo- 107. From 16 to 14 percent in the case of Honduras, and Armengol (2004). from 18 to 16 percent in the case of Romania (Eden 77. Boase and others 2006. and Gaggl 2015, for the WDR 2016). 78. World Bank 2011. 108. Acemoglu 2002; Aghion and others 2015; Dabla- 79. La Ferrara, Chong, and Duryea 2012; Jensen and Norris and others 2015; Garicano and Rossi- Oster 2009. Hansberg 2006; Jaumotte, Lall, and Papageorgiou 80. WDR 2016 team, based on Research ICT Africa 2008; Autor, Katz, and Krueger 1998. surveys (various years). 109. Acemoglu and Autor 2011; Akcomak, Kok, and 81. De, Mohapatra, and Plaza, forthcoming, for the Rojas-Romagosa 2013; Autor and Dorn 2013; Goos, WDR 2016. Manning, and Salomons, forthcoming. 82. Camacho and Conover 2011; Jensen 2010. 110. Autor, Dorn, and Hanson, forthcoming; Autor, Dorn, 83. Goyal 2010; Aker 2010a; Best and others 2010; Aker and Hanson 2013. 2011; Martin 2010. 111. The number of workers in agriculture declined in 84. Aker and Mbiti 2010. China between 2000 and 2010, but the number of 85. Beuermann, McKelvey, and Vakis 2012. workers within the sector that were machine or 86. May, Dutton, and Munyakazi 2011. equipment operators almost doubled (WDR 2016 87. Jensen 2007. team, based on the National Bureau of Statistics of 88. Pineda, Aguero, and Espinoza 2011. China, various years). 89. Galiani and Jaitman 2010. 112. Dutz and others 2015; Messina, Oviedo, and Pica 90. Asad 2014. 2015 for Mexico and Peru. 91. Aker 2010a; Pineda, Aguero, and Espinoza 2011. 113. Goldin and Katz (2008) refer to a race between edu- 92. Aker 2010a. cation and technology when discussing the case of 93. Aker 2011. the United States, but the phrase was fi rst used by 94. Tadesse and Bahiigwa 2015; Jagun, Heeks, and Tinbergen (1975). Whalley 2008. 114. Brynjolfsson and McAfee (2014, 11). 95. Amazon Mechanical Turk is an online work plat- 115. Oviedo and others (forthcoming) use STEP house- form. The quote was obtained through an online hold surveys (World Bank, various years) from 10 questionnaire of online workers done in September developing countries to characterize the typical 2014 for this Report. tasks done across occupations, and show that, 96. Varian 2011. indeed, occupations that are considered intensive 97. 2015. in nonroutine cognitive and socioemotional skills 98. Interactive Advertising Bureau 2010. require workers to do more complex reading, write 99. Greenstein and McDevitt 2011. longer texts, use more advanced math, contact more 100. WDR 2016 team, based on Research ICT Africa sur- clients, collaborate more with others, as well as do veys (various years). more thinking, learning, supervising, and present- 101. Pew Research Center 2014. ing. By contrast, these occupations do fewer routine 102. In the United States, the median tenure for male and manual activities, such as operating machines wage and salary workers was lower in 2014 at 5.5 and doing physical activities. years, compared with 5.9 years in 1983 (Copeland 116. WDR calculations based on STEP household sur- 2015). Dutz and others (2015, for the WDR 2016) veys (World Bank, various years). 138 WORLD DEVELOPMENT REPORT 2016

117. This is referred to as functional literacy. Functional 139. Bowles 2014; Frey and Osborne 2013. illiteracy is defi ned as the proportion of exam tak- 140. Autor, Dorn, and Hanson, forthcoming. Their analy- ers (15-year-olds) who score below a level 2 on the sis examines the period between 1980 and 2007. Programme for International Student Assessment 141. Autor 2014. (PISA) reading test (WDR 2016 team, based on 142. Handel 2000. OECD PISA 2012 scores). 143. Gaggl and Wright 2014. 118. World Bank 2014c. 144. Jaimovich and Siu 2012. 119. WDR 2016 team, based on Research ICT Africa (var- 145. Handel 2015, for the WDR 2016. ious years). Among those who do not cite a lack of 146. Luddites were English textile workers who pro- connection as a reason for not using the internet, tested rapid automation in the early 19th century by 3 out of 10 say they do not use the internet because destroying the machinery used to replace them. they do not know how to. 147. Crafts 2015, for the WDR 2016; Katz and Margo 120. PIAAC survey. 2013; Gray 2013. 121. World Bank 2010. 148. MacCrory and others 2014. 122. Valerio and others, forthcoming. Most of this liter- 149. Crafts 2015, for the WDR 2016. ature remains focused on the returns to computer 150. For a discussion on “premature” deindustrializa- use (Spitz-Oener 2008; Sakellariou and Patrinos tion, see Rodrik 2015. 2003). Most studies are not based on random 151. Turner 2014. assignment of the technology, and estimates likely 152. Berger and Frey 2014. refl ect that these workers have other characteristics 153. Beaudry, Green, and Sand 2014. or skills that command an earnings premium. In a 154. Crafts 2015, for the WDR 2016. randomized experiment using fi ctitious resumes 155. Brynjolfsson and McAfee 2014. for white-collar occupations in Buenos Aires and 156. Countries in Europe and Central Asia are a case Bogota, ICT skills increased the probability of in point. Starting from a comparable institutional receiving a job callback by one percentage point and economic level in the early 1990s, countries (Lopez-Boo and Blanco 2010). that carried out the most economic reforms (for 123. Correa and de Sousa 2015, for the WDR 2016. example, the Czech Republic, Estonia, and Poland) 124. Akerman, Gaarder, and Mogstad 2015. have seen a more rapid shift toward new economy 125. Cunningham and Villasenor 2014. occupations (Arias and others 2014). Historically, 126. Valerio and others 2015a, 2015b; Ajwad and others GDP per capita has also been a strong determinant 2014a; Ajwad and others 2014b; Bodewig and others of technology adoption (Crafts 2015). 2014. 157. Autor 2014. 127. Valerio and others 2015a; Valerio and others 2015b. 158. WDR 2016 team, based on STEP household surveys 128. Eden and Gaggl 2015, for the WDR 2016. (World Bank, various years). 129. For example, it is often diffi cult to distinguish 159. Acemoglu and Autor 2011; Aedo and others 2013; between trade and technology (chapter 1). Arias and others 2014; Autor and Price 2013; Autor, 130. Akcomak, Kok, and Rojas-Romagosa 2013; Autor, Levy, and Murnane 2003; Bruns, Evans, and Luque Dorn, and Hanson, forthcoming; Eden and Gaggl 2012. 2014; Karabarbounis and Neiman 2013; Michaels, 160. Cortes and others 2014. Natraj, and Van Reenen 2014. 161. Skirbekk 2013. 131. WDR 2016 team, based on I2D2 data (World Bank, 162. Skirbekk 2013. various years). 132. Autor 2014; Autor and Dorn 2013; Autor, Levy, and Murnane 2003; Spitz-Oener 2008; MacCrory and References others 2014. Acemoglu, Daron. 2002. “Technical Change, Inequality, 133. Autor 2014; Summers 2014; Brynjolfsson and and the Labor Market.” Journal of Economic Literature McAfee 2014. 40: 7–72. 134. Ryder 2015. Acemoglu, Daron, and David Autor. 2011. “Skills, Tasks and 135. Autor, Dorn, and Hanson, forthcoming. Technologies: Implications for Employment and Earn- 136. Statista 2015. ings.” Handbook of Labor Economics 4 (12): 1043–1171. 137. World Bank team under the Government Financial Aedo, Cristian, Jesko Hentschel, Javier Luque, and Mar- Management and Revenue Administration Project. tin Moreno. 2013. “From Occupations to Embedded Other regulatory and business processes changes Skills: A Cross-Country Comparison.” Background also help explain the redundancy of workers. paper for the World Development Report 2013, World 138. WDR 2016 team interview with Dr. Ishrat Hussain, Bank, Washington, DC. former governor of the State Bank of Pakistan, Jan- Aghion, Philippe, Ufuk Akcigit, Antonin Bergeaud, Rich- uary 30, 2015. ard Blundell, and David Hemous. 2015. “Innovation EXPANDING OPPORTUNITIES 139

and Top Income Inequality.” Working paper, Harvard Argentina, National Institute of Statistics and Census. University, Cambridge, MA. http://scholar.harvard Various years. National Survey on the Access and Use .edu/aghion/publications/innovation-and-top-income of Information and Communication Technologies. -inequality. Argentina National Institute of Statistics and Census, Agrawal, Ajay, John Horton, Nico Lacetera, and Elizabeth Buenos Aires. Lyons. 2013. “Digitization and the Contract Labor Arias, Omar, Carolina Sanchez-Paramo, Maria Davalos, Market: A Research Agenda.” NBER Working Paper Indhira Santos, Erwin Tiongson, Carola Gruen, Nata- 19525, National Bureau of Economic Research, Cam- scha de Andrade, Gady Saiovici, and Cesar Cancho. bridge, MA. 2014. Back to Work: Growing with Jobs in Europe and Cen- Ajwad, Mohamed Ihsan, Stefan Hut, Ilhom Abdulloev, tral Asia. Washington, DC: World Bank. Robin Audy, Joost de Laat, Sachiko Kataoka, Jennica Asad, Saher. 2014. “The Crop Connection: Impact of Cell Larrison, Zlatko Nikoloski, and Federico Torracchi. Phone Access on Crop Choice in Rural Pakistan.” Job 2014a. The Skills Road: Skills for Employability in Tajiki- market paper, George Washington University, Wash- stan. Washington, DC: World Bank. ington, DC. Ajwad, Mohamed Ihsan, Joost de Laat, Stefan Hut, Jen- Asimov, Isaac. 1964. “Visit to the World’s Fair of 2014.” nica Larrison, Ilhom Abdulloev, Robin Audy, Zlatko New York Times, August 16. Nikoloski, and Federico Torracchi. 2014b. The Skills Atasoy, Hilal. 2013. “The Effects of Broadband Internet Road: Skills for Employability in the Kyrgyz Republic. Expansion on Labor Market Outcomes.” Industrial and Washington, DC: World Bank. Labor Relations Review 66 (2): 315–45. Akcomak, Semih, Suzanne Kok, and Hugo Rojas- Autor, David. 2014. “Polanyi’s Paradox and the Shape of Romagosa. 2013. “The Effects of Technology and Employment Growth.” Draft prepared for the Federal Offshoring on Changes in Employment and Task- Reserve Bank of Kansas City. Autor, David, and David Dorn. 2013. “The Growth of Content of Occupations.” Discussion Paper 233, CBP Low-Skill Service Jobs and the Polarization of the [Netherlands Bureau for Economic Analysis], The U.S. Labor Market.” American Economic Review 103 (5): Hague. 1553–97. Aker, Jenny C. 2010a. “Information from Markets Near Autor, David, David Dorn, and Gordon Hanson. 2013. and Far: Mobile Phones and Agricultural Markets in “The Geography of Trade and Technology Shocks Niger.” American Economic Journal: Applied Economics 2: in the United States.” American Economic Review 103: 46–59. 220–25. ————. 2010b. “Why Have Mobile Phones Succeeded ————. Forthcoming. “Untangling Trade and Technol- Where Other Technologies Have Not?” In Global ogy: Evidence from Local Labor Markets.” The Eco- Development: Views from the Center (blog). http://www nomic Journal. .cgdev.org/blog/why-have-mobile-phones-succeeded Autor, David, Lawrence Katz, and Melissa Kearney. -where-other-technologies-have-not. 2008. “Trends in U.S. Wage Inequality: Revising the ————. 2011. “Dial A for Agriculture: A Review of Informa- Revisionists.” Review of Economics and Statistics 90 (2): tion and Communication Technologies for Agricul- 300–23. tural Extension in Developing Countries.” Agricultural Autor, David, Lawrence Katz, and Alan Krueger. 1998. Economics 42 (6): 631–47. “Computing Inequality: Have Computers Changed Aker, Jenny C., Rachid Boumnijel, Amanda McClelland, the Labor Market?” Quarterly Journal of Economics 1134: and Niall Tierney. 2014. “Payment Mechanisms and 1169–1213. Anti-Poverty Programs: Evidence from a Mobile Autor, David, Frank Levy, and Richard Murnane. 2003. Money Cash Transfer Experiment in Niger.” Working “The Skill Content of Recent Technological Change: Paper 268, Center for Global Development, Washing- An Empirical Exploration.” Quarterly Journal of Econom- ton, DC. ics 118 (4): 1279–1333. Aker, Jenny C., and Isaac M. Mbiti. 2010. “Mobile Phones Autor, David, and Brendan Price. 2013. “The Changing and Economic Development in Africa.” Journal of Eco- Task Composition of the U.S. Labor Market: An nomic Perspectives 24 (3): 207–32. Update of Autor, Levy, and Murnane (2003).” Working Akerman, Anders, Ingvil Gaarder, and Magne Mogstad. paper, Massachusetts Institute of Technology and 2015. “The Skill Complementarity of Broadband Inter- National Bureau of Economic Reseach, Cambridge, net.” NBER Working Paper 20826, National Bureau of MA. Economic Research, Cambridge, MA. Bagues, Manuel, and Mauro Sylos. 2009. “Do Online Ali, Daniel Ayalew, Klaus Deininger, and Markus Gold- Labor Market Intermediaries Matter? The Impact of stein. 2014. “Environmental and Gender Impacts of Alma Laurea on the University-to-Work Transition.” Land Tenure Regularization in Africa: Pilot Evidence In Studies of Labor Market Intermediation, edited by from Rwanda.” Journal of 110: David Autor, 127–54. Chicago: University of Chicago 262–75. Press. 140 WORLD DEVELOPMENT REPORT 2016

Beaudry, Paul, David A. Green, and Benjamin M. Sand. Brynjolfsson, Erik, and Andrew McAfee. 2014. The Second 2014. “The Declining Fortunes of the Young since Machine Age: Work, Progress, and Prosperity in a Time of 2000.” American Economic Review 104 (5): 381–86. Brilliant Technologies. New York: W. W. Norton. Berger, Thor, and . 2014. “Industrial Bussolo, Maurizio, and Luis Lopez Calva. 2014. Shared Renewal in the 21st Century: Evidence from U.S. Cit- Prosperity: Paving the Way in Europe and Central Asia. ies.” University of Oxford, Oxford, U.K. http://www Washington, DC: World Bank. .oxfordmartin.ox.ac.uk/publications/view/1849. Calvo-Armengol, Antoni. 2004. “Job Contact Networks.” Berman, Eli, Rohini Somanathan, and Hong Tan. 2005. Journal of Economic Theory 115 (1): 191–206. “Is Skill-Biased Technological Change Here Yet? Camacho, Adriana, and Emily Conover. 2011. “The Impact Evidence from Indian Manufacturing in the 1990s.” of Receiving SMS Price and Weather Information in Policy Research Paper 3761, World Bank, Washington, Colombia’s Agricultural Sector.” Working Paper 220, DC. Inter-American Development Bank, Washington, DC. Best, Michael L., Thomas N. Smyth, John Etherton, and Campos-Vazquez, Raymundo M., Luis Lopez-Calva, and Edem Wornyo. 2010. “Uses of Mobile Phones in Nora Lustig. Forthcoming. “Declining Wages for Post-Confl ict Liberia.” Information Technologies and College-Educated Workers in Mexico: Degraded Ter- International Development 6 (2): 91–108. tiary Education or Skills Obsolescence?” World Bank, Beuermann, Diether, Christopher McKelvey, and Renos Washington, DC. Vakis. 2012. “Mobile Phones and Economic Develop- Capili, Miro. 2015. “The BPO Industry in the Philippines: ment in Rural Peru.” Journal of Development Studies 48 An Overview.” Background paper for the World Devel- (11): 1617–28. opment Report 2016, World Bank, Washington, DC. Black, Sandra, and Alexandra Spitz-Oener. 2007. “Explain- CEDLAS (Center for Distributive, Labor and Social Stud- ing Women’s Success: Technological Change and the ies) and World Bank. Various years. Socio-Economic Skill Content of Women’s Work.” NBER Working Database for Latin America and the Caribbean Paper 13116, National Bureau of Economic Research, (SEDLAC), La Plata, Argentina. http://cedlas.econo Cambridge, MA. .unlp.edu.ar/eng/index.php. Blinder, Alan, and Alan Krueger. 2013. “Alternative Mea- CGAP (Consultative Group to Assist the Poor). 2014. “10 sures of Offshorability: A Survey Approach.” Journal of Myths about M-Pesa: 2014 Update.” http://www.cgap Labor Economics 31 (2): S97–S128. .org/blog/10-myths-about-m-pesa-2014-update. Bloom, N., J. Liang, J. Roberts, and Z. Ying. 2014. “Does China Association for Employment Research. 2014. Working from Home Work? Evidence from a Chinese “Internet Business Employment and Social Security Experiment.” Working paper, Stanford University, Research.” For an English summary of the report, Stanford, CA. see http://www.alizila.com/e-commerce-force-china Boase, Jeffrey, John Horrigan, Barry Wellman, and Lee -jobs-not-job-security-report. Rainie. 2006. “The Strength of Internet Ties.” Pew China, National Bureau of Statistics of China. Various Internet and American Life Project. http://www.pew years. Population Census. Beijing. http://www.stats internet.org/fi les/old-media/Files/Reports/2006/PIP .gov.cn/english/Statisticaldata/CensusData/. _Internet_ties.pdf. Colombia, Directorate of Statistical Methodology and Bodewig, Christian, Reena Badiani-Magnusson, Kevin Production. Various years. Household Integrated Macdonald, David Newhouse, and Jan Rutkowski. Survey. National Administrative Department of 2014. Skilling Up Vietnam: Preparing the Workforce for Statistics, Bogota. a Modern Market Economy. Washington, DC: World Comin, Diego, and Bart Hobjin. 2010. “An Exploration of Bank. Technology Diffusion.” American Economic Review 100 Bonnet, Odran, Pierre-Henri Bono, Guillaume Chapelle, (5): 2031–59. and Étienne Wasmer. 2014. “Does Housing Capital Comin, Diego, and Martí Mestieri. 2013. “If Technol- Contribute to Inequality? A Comment on Thomas ogy Has Arrived Everywhere, Why Has Income Piketty’s Capital in the 21st Century.” Sciences Po Eco- Diverged?” INET Research Note 26, Institute for New nomics Discussion Paper 2014-07, Paris Institute of Economic Thinking, New York. Political Studies, Paris. Connolly, Marie, and Alan Krueger. 2006. “Rockonomics: Bowles, Jeremy. 2014. “The Computerisation of European The Economics of Popular Music.” In Handbook of the Jobs.” Bruegel (blog), July 17. https://www.bruegel.org. Economics of Art and Culture, edited by V. A. Ginsburg Brazilian Internet Steering Committee. Various years. and D. Throsby, 667–719. Amsterdam: North-Holland. Survey on the Use of Information and Communica- Copeland, Craig. 2015. “Employee Tenure Trends 1983– tion Technologies in Brazil. Brazilian Network Infor- 2014.” Employee Benefi t Research Institute Notes 36 (2). mation Center, São Paulo. Correa, Paulo, and Filipe Lage de Sousa. 2015. “Internet Bruns, Barbara, David Evans, and Javier Luque. and Wages: The Case of Brazil.” Background paper 2012. Achieving World-Class Education in Brazil: The Next for the World Development Report 2016, World Bank, Agenda. Washington, DC: World Bank. Washington, DC. EXPANDING OPPORTUNITIES 141

Cortes, Matias, Nir Jaimovich, Christopher Nekarda, and ————. 2015. “The Global Distribution of Information and Henry Siu. 2014. “The Micro and Macro of Disappear- Communication Technologies: The Role of Industrial ing Routine Jobs: A Flows Approach.” NBER Working Composition.” Background paper for the World Devel- Paper 20307, National Bureau of Economic Research, opment Report 2016, World Bank, Washington, DC. Cambridge, MA. Estonia, Government of. 2015. E-Estonia presentation. Crafts, Nicholas. 2015. “The Economic Impact of ICT: A Tallin, November. Historical Perspective.” Background paper for the Fernandes, Ana, Aaditya Mattoo, Huy Nguyen, and Marc World Development Report 2016, World Bank, Washing- Schiffbauer. 2015. “The Impact of Broadband on ton, DC. Chinese Exports in the Pre-Alibaba Era.” Background Cunningham, Wendy, and Paula Villasenor. 2014. paper for the World Development Report 2016, World “Employer Voices, Employer Demands, and Impli- Bank, Washington, DC. cations for Public Skills Development Policy.” Policy Forman, Chris, Avi Goldfarb, and Shane Greenstein. 2012. Research Working Paper 6853, World Bank, Wash- “The Internet and Local Wages: A Puzzle.” American ington, DC. Economic Review 102 (1): 556–75. Dabla-Norris, Era, Kalpana Kochhar, Nujin Suphaphip- Frey, Carl, and Michael Osborne. 2013. “The Future of hat, Frantisek Ricka, and Evridiki Tsounta. 2015. Employment: How Susceptible Are Jobs to Computer- “Causes and Consequences of Income Inequality: isation?” Working paper, Oxford University, Oxford, A Global Perspective.” Staff Discussion Note 15/13, U.K. International Monetary Fund, Washington, DC. Gaggl, Paul, and Greg Wright. 2014. “A Short-Run View Dammert, Ana C., Jose Galdo, and Virgilio Galdo. 2014. of What Computers Do: Evidence from a UK Tax “Mobile Phones for Labor Market Intermediation: A Incentive.” Discussion Paper 752, University of Essex, Multi-Treatment Experimental Design.” http://www Colchester, U.K. .iza.org/conference_fi les/worldb2014/galdo_j2620.pdf. Galiani, Sebastián, and Laura Jaitman. 2010. “Trace- De, Supriyo, Sanket Mohapatra, and Sonia Plaza. Forth- ability Applied to SMEs in Argentina: A Case Study.” coming. “Impact of Migration on ICT Adoption in Inter-American Development Bank, Washington, DC. Sub-Saharan Africa.” Gallup. Various years. World Poll. Washington, DC, De los Rios, Carlos. 2010. Welfare Impact of Internet Use http://www.gallup.com/services/170945/world-poll on Peruvian Households. : Instituto de Estudios .aspx. Peruanos. Galperin, Hernan, and Fernanda Viecens. 2014. “Con- Demirgüç-Kunt, Asli, Leora Klapper, Dorothe Singer, nected for Development? Theory and Evidence and Peter Van Oudheusden. 2015. “The Global Fin- about the Impact of the Internet Technologies on dex Database 2014: Measuring Financial Inclusion Poverty Alleviation.” Universidad de San Andres, around the World.” Policy Research Working Paper Buenos Aires. http://papers.ssrn.com/sol3/papers 7255, World Bank, Washington, DC. .cfm?abstract_id=2397394. Digital Divide Data. 2014. Building a Global Platform: 2014 Annual Report. New York: Digital Divide Data. Garicano, Luis, and Esteban Rossi-Hansberg. 2006. “Orga- Dutz, Mark, Lucas Ferreira-Mation, Stephen O’Connell, nization and Inequality in a .” and Robert Willig. 2015. “Technological Change Quarterly Journal of Economics (November): 1383–1435. and Labor Market Segmentation in the Developing Goldin, Claudia, and Lawrence F. Katz. 2008. The Race World: Evidence from Brazil.” Background paper for between Education and Technology. Cambridge, MA: the World Development Report 2016, World Bank, Wash- Harvard University Press. ington, DC. Gomez, Ricardo. 2014. “When You Do Not Have a Com- EC (European Commission). 2004. E-skills for Europe: 2010 puter: Public-Access Computing in Developing and Beyond. Brussels: European Commision, Enter- Countries.” Information Technology for Development 20 prise and Industry Directorate General. (3): 274–91. ————. 2015. Eurostat Information Society Statistics. Goos, M., A. Manning, and A. Salomons. Forthcoming. European Commission, Luxembourg, http://ec “Explaining Job Polarization: Routine-Biased Tech- .europa.eu/eurostat/statistics-explained/index.php nological Change and Offshoring.” American Economic /Information_society_statistics_-_households_and Review 104 (8): 2509–26. _individuals. Goyal, Aparajita. 2010. “Information, Direct Access to ————. Various years. Eurostat (database). European Farmers, and Rural Market Performance in Central Commission, Luxembourg, http://ec.europa.eu/ India.” American Economic Journal: Applied Economics eurostat/data/database. 2 (3): 22–45. Eden, Maya, and Paul Gaggl. 2014. “The Substitution of Granovetter, Mark. 1973. “The Strength of Weak Ties.” ICT Capital for Routine Labor: Transitional Dynam- American Journal of Sociology 78 (6): 1360–80. ics and Long-Run Implications.” Working paper, Gray, Rowena. 2013. “Taking Technology to Task: The available at the Social Science Research Network. Skill Content of Technological Change in the Early 142 WORLD DEVELOPMENT REPORT 2016

Twentieth Century United States.” Explorations in Eco- Interactive Advertising Bureau. 2010. “Consumers Driv- nomic History 50 (3): 351–67. ing the Digital Uptake: The Economic Value of Online Greenstein, Shane, and Ryan McDevitt. 2011. “The Global Advertising-Based Services for Consumers.” http:// Broadband Bonus: Broadband Internet’s Impact on www.iab.net/media/fi le/White-Paper-Consumers- Seven Countries.” In The Linked World: How ICT Is driving-the-digital-uptake_FINAL.PDF. Transforming Societies, Cultures and Economies, edited ITU (International Telecommunication Union). Various by B. Van Ark, 35–42. New York: Conference Board. years. World Telecommunication/ICT Indicators GSMA (Groupe Speciale Mobile Association). 2015. (database). ITU, Geneva, http://www.itu.int/en “Bridging the Gender Gap: Mobile Access and Usage /ITU-D/Statistics/Pages/publications/wtid.aspx. in Low- and Middle-Income Countries.” http://www Jack, William, and Tavneet Suri. 2014. “Risk Sharing and .gsma.com/newsroom/press-release/results-of-new Transactions Costs: Evidence from Kenya’s Mobile -mobile-phone-gender-gap-survey/. Money Revolution.” American Economic Review 104 (1): Hall, Jonathan, and Alan Krueger. 2015. “An Analysis of 183–223. the Labor Market for Uber’s Driver-Partners in the Jagun, Abi, Richard Heeks, and Jason Whalley. 2008. “The United States.” https://s3.amazonaws.com/uber-static Impact of Mobile Telephony on Developing Country /comms/PDF/Uber_Driver-Partners_Hall_Kreuger Micro-Enterprise: A Nigerian Case Study.” Information _2015.pdf. Technologies and International Development 4 (4): 47–65. Handel, Michael. 2000. “Models of Economic Organiza- Jaimovich, Nir, and Henry Siu. 2012. “The Trend Is the tion and the New Inequality in the United States.” Cycle: Job Polarization and Jobless Recoveries.” NBER PhD diss., Harvard University. Working Paper 18334, National Bureau of Economic ————. 2015. “The Effects of Information and Communi- Research, Cambridge, MA. cation Technology on Employment, Skills, and Earn- Jaumotte, Florence, Subir Lall, and Chris Papageorgiou. ings in Developing Countries.” Background paper 2008. “Rising Income Inequality: Technology, or Trade for the World Development Report 2016, World Bank, and Financial Globalization?” Working Paper 08/185, Washington, DC. International Monetary Fund, Washington, DC. Heeks, Richard, and Shoba Arun. 2010. “Social Outsourc- Jensen, Robert. 2007. “The Digital Provide: Information ing as a Development Tool: The Impact of Outsourc- (Technology), Market Performance, and Welfare in ing IT Services to Women’s Social Enterprises in Ker- the South Indian Fisheries Sector.” Quarterly Journal of ala.” Journal of International Development 22 (4): 441–54. Economics 122 (3): 879–924. Holland, Julie, Chris Nosko, and Alan Sorensen. 2012. ————. 2010. “The (Perceived) Returns to Education and “Supply Responses to Digital Distribution: Recorded the Demand for Schooling.” Quarterly Journal of Eco- Music and Live Performances.” Information Economics nomics 125 (2): 515–48. and Policy 24 (1): 3–14. ————. 2012. “Do Labor Market Opportunities Affect IBM (International Business Machines Corporation). Young Women’s Work and Family Decisions? Exper- 2014. “IBM Announces First Commercial Applica- imental Evidence from India.” Quarterly Journal of tion of IBM Watson in Africa.” http://www.ibm.com Economics 127 (2): 753–92. /news/za/en/2014/10/08/Z027713X96105D22.html. Jensen, Robert, and Emily Oster. 2009. “The Power of TV: IFPRI (International Food Policy Research Institute) and Cable Television and Women’s Status in India.” Quar- World Bank. 2010. “Innovations in Rural and Agri- terly Journal of Economics 124 (3): 1057–94. culture Finance.” In Focus 18. Washington, DC: World Kabanda, Patrick. 2015. “The Internet, Music and Devel- Bank. opment.” Background paper for the World Development ILO (International Labour Organization). Various years. Report 2016, World Bank, Washington, DC. Key Indicators of the Labour Market (KILM). ILO, Karabarbounis, Loukas, and Brent Neiman. 2013. “The Geneva, http://www.ilo.org/empelm/what/WCMS Global Decline of the Labor Share.” NBER Working _114240/lang--en/index.htm. Paper 19136, National Bureau of Economic Research, ————. Various years. Laborsta (database). International Cambridge, MA. Labour Organization, Geneva, http://laborsta.ilo.org/. Katz, Lawrence, and Robert Margo. 2013. “Technical Imaizumi, Saori, and Indhira Santos. Forthcoming. Change and the Relative Demand for Skilled Labor: “Linking Workers to Jobs: Using Mobile Phones and The United States in Historical Perspective.” NBER Online Job Search Platforms and Online Outsourcing Working Paper 18752, National Bureau of Economic Platforms in Developing Countries.” Background Research, Cambridge, MA. paper for the World Development Report 2016, World Kennedy, Robert, Sateen Sheth, Ted London, Ekta Jhaveri, Bank, Washington, DC. and Lea Kilibarda. 2013. Assessing the Opportunity for Intel and Dalberg Global Development Advisors. 2012. Building a Thriving Industry. Ann Arbor: William David- “Women and the Web.” http://www.intel.com son Institute at the University of Michigan. /content/dam/www/public/us/en/documents/pdf Keynes, John Maynard. 1931. Essays in . Lon- /women-and-the-web.pdf. don: Macmillan. EXPANDING OPPORTUNITIES 143

Klonner, Stefan, and Patrick Nolen. 2010. “Cell Phones Mondialisation]. http://EconPapers.repec.org and Rural Labor Markets: Evidence from South /RePEc:dia:wpaper:dt201424. Africa.” In Proceedings of the German Development Martin, Brandie. 2010. Mobile Phones and Rural Livelihoods: Economics Conference, Hannover, 56. http://econpapers An Exploration of Mobile Phone Diff usion, Uses, and Per- .repec.org/paper/zbwgdec10/. ceived Impacts of Uses among Small- to Medium-Size Farm Kolko, Jed. 2012. “Broadband and Local Growth.” Journal of Holders in Kamuli District, Uganda. Ames: Iowa State 71 (1): 100–13. University. Kroft, Kory, and Devin Pope. 2014. “Does Online Search May, J., V. Dutton, and L. Munyakazi. 2011. “Information Crowd Out Traditional Search and Improve Match- and Communication Technologies as an Escape from ing Effi ciency? Evidence from Craigslist.” Journal of Poverty Traps.” PICTURE Africa Research Project, Labor Economics 32 (2): 259–303. Nairobi. Centre d’Économie de la Sorbonne. http:// Kuek, Siou Chew, Cecilia Paradi-Guilford, Toks Fayomi, www..net/publication/253341123. Saori Imaizumi, and Panos Ipeirotis. Forthcoming. Messina, Julian, Ana Maria Oviedo, and Giovanni Pica. The Global Opportunity in Online Outsourcing. Washing- 2015. “Inequality and Polarization in Latin America: ton, DC: World Bank. Patterns and Determinants.” Unpublished paper, Kuhn, Peter. 2014. “The Internet as a Labor Market Match- World Bank, Washington, DC. maker.” IZA World of Labor 2014 (article 18). http://wol.iza Mexico, National Institute of Statistics and Geography. .org/articles/internet-as-a-labor-market-matchmaker. Various years. Survey on the Availability and Use Kuhn, Peter, and Hani Mansour. 2014. “Is Internet Job of Information and Communication Technology Search Still Ineffective?” Economic Journal 124 (581): in Households. National Institute of Statistics and 1213–33. Geography, Aguascalientes. La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. Michaels, Guy, Ashwini Natraj, and John Van Reenen. 2012. “Soap Operas and Fertility: Evidence from Bra- 2014. “Has ICT Polarized Skill Demand? Evidence zil.” American Economic Journal: Applied Economics 4 (4): from Eleven Countries over Twenty-Five Years.” 1–31. Review of Economics and Statistics 96 (1): 60–77. Lehdonvirta, Vili, Helena Barnard, Mark Graham, and Monitor Inclusive Markets. 2011. “Job Creation through Isis Hjorth. 2014. “Online Labour Markets: Leveling Building the Field of Impact Sourcing.” Working the Playing Field for International Service Markets?” paper, , New York. https:// http://ipp.oii.ox.ac.uk/sites/ipp/fi les/documents/IPP www.rockefellerfoundation.org/app/uploads/Job 2014_Lehdonvirta_0.pdf, accessed May 6, 2015. -Creation-Through-Building-the-Field-of-Impact Lopez-Boo, Florencia, and Mariana Blanco. 2010. “ICT -Sourcing.pdf. Skills and Employment: A Randomized Experiment.” Monroy-Taborda, Sebastian, Martin Moreno, and Indhira IZA Discussion Paper 5336, Institute for the Study of Santos. Forthcoming. “Technology Use and Changing Labor, Bonn. Skills Demands: New Evidence from Developing MacCrory, Frank, George Westerman, Yousef Alham- Countries.” Background paper for the World Develop- madi, and Erik Brynjolfsson. 2014. “Racing with ment Report 2016, World Bank, Washington, DC. and against the Machine: Changes in Occupational Montenegro, Claudio, and Harry Patrinos. 2014. “Compa- Skill Composition in an Era of Rapid Technologi- rable Estimates of Returns to Schooling around the cal Advance.” Paper presented at the Thirty-Fifth World.” Policy Research Working Paper 7020, World International Conference on Information Systems, Bank, Washington, DC. Auckland, December 14–17. http://aisel.aisnet.org/cgi Moretti, E., and P. Thulin. 2013. “Local Multipliers and /viewcontent.cgi?article=1052&context=icis2014. Human Capital in the US and Sweden.” Industrial and Maloney, William, and Felipe Valencia. 2015. “Techno- Corporate Change 22 (1): 339–62. logical Adoption, Human Capital and Job Creation.” Muto, M., and T. Yamano. 2009. “The Impact of Mobile Working paper, World Bank, Washington, DC. Phone Coverage Expansion on Market Participation: Mang, Constantin. 2012. “Online Job Search and Match- Panel Data Evidence from Uganda.” World Develop- ing Quality.” IFO Working Paper 147, Institute for ment 37 (12): 1887–96. Economic Research, Munich. Nakamura, Alice, Kathryn Shaw, Richard Freeman, Emi Marandino, Joaquin, and Phanindra Wunnava. 2014. Nakamura, and Amanda Pyman. 2009. “Jobs Online.” “The Effect of Access to Information and Commu- In Studies of Labor Market Intermediation, edited by nication Technology on Household Labor Income: David Autor, 27–65. Chicago: University of Chicago Evidence from One Laptop per Child in Uruguay.” Press. IZA Discussion Paper 8415, Institute for the Study of NASSCOM. 2014. The IT-BPM Sector in India: Strategic Labor, Bonn. Review 2014. : NASSCOM. Marouani, Mohamed, and Bjorn Nilsson. 2014. Neisser, Ulric, and others. 1996. “Intelligence: Knowns “The Labor Market Effects of Skill-Biased Tech- and Unknowns.” American Psychologist 51 (2): 77–101. nological Change in Malaysia.” Working Paper OECD (Organisation for Economic Co-operation and DT/2014/24, DIAL [Développement, Institutions et Development). Various years. OECD Key Economic 144 WORLD DEVELOPMENT REPORT 2016

Indicators (KEI) Database. OECD, Paris, http://stats Agricultural Production in Rural Villages in Peru.” ..org/mei/default.asp?rev=4. Working paper, University of Piura, Piura, Peru. ————. 2004. Information Technology Outlook. Paris: OECD. http://udep.edu.pe/cceeee/files/2014/07/3B_2 ————. 2011. The Future of Families to 2030. Paris: OECD. _RitterGUERRERO.pdf. ————. 2014. Measuring the Digital Economy: A New Perspec- Rodrik, Dani. 2015. “Premature Deindustrialization.” tive. Paris: OECD. NBER Working Paper 20935, National Bureau of Eco- Oster, Emily, and Bryce Millett. 2013. “Do IT Service nomic Research, Cambridge, MA. Centers Promote School Enrollment? Evidence from Ryder, Guy. 2015. “How Can We Create Decent Jobs in the India.” Journal of Development Economics 104: 123–35. Digital Age?” In World Economic Forum (blog). https:// Oviedo, Ana Maria, Maria Laura Sanchez Puerta, Indhira agenda.weforum.org/2015/01/how-can-we-create Santos, Emmanuele Dicarlo, Salvatore Lo Bello, and -decent-jobs-in-the-digital-age/. Sebastian Monroy-Taborda. Forthcoming. The Skill Sakellariou, Chris, and Harry Patrinos. 2003. “Technol- Content of Occupations across Low- and Middle-Income ogy, Computers, and Wages: Evidence from a Devel- Countries: Evidence from Harmonized Data. Washington, oping Economy.” Policy Research Paper 3008, World DC: World Bank. Bank, Washington, DC. Pew Research Center. 2014. “Digital Life in 2025: Technol- Samasource. 2015. http://www.samasource.org. ogy’s Impact on Workers.” http://www.pewinternet Schaefer-Davis, Susan. 2005. “Women Weavers Online: .org/fi les/2014/12/PI_Web25WorkTech_12.30.141.pdf. Rural Moroccan Women on the Internet.” In Gender PIAAC (Programme for the International Assessment and the Digital Economy: Perspectives from the Developing of Adult Competencies). Various years. OECD, Paris, World, edited by Cecelia Ng and Swasti Mitter, 159–85. http://www.oecd.org/site/piaac/. New Delhi, India: SAGE Publications. Pimienta, Daniel, Daniel Prado, and Alvaro Blanco. 2009. Seol, Boo Kang, and Indhira Santos. 2015. “#MyDressMy Twelve Years of Measuring Linguistic Diversity in the Inter- Choice: Tackling Gender Discrimination and Vio- lence in Kenya One Tweet at a Time.” Let’s Talk Develop- net. Paris: UNESCO. ment (blog), May 4. http://blogs.worldbank.org Pineda, Allan, Marco Aguero, and Sandra Espinoza. 2011. /developmenttalk/mydressmychoice-tackling “The Impact of ICT on Vegetable Farmers in Hon- -gender-discrimination-and-violence-kenya-one duras.” Working Paper 243, Inter-American Develop- -tweet-time, accessed May 4, 2015. ment Bank, Washington, DC. Skirbekk, Gordo. 2013. “Skill Demand and the Compara- Plaza, Sonia, Seyed Reza Yousefi , and Dilip Ratha. 2015. tive Advantage of Age: Jobs Tasks and Earnings from “Technological Innovations and Remittance Costs.” the 1980s to the 2000s in Germany.” Background paper for the World Development Report 22: 61–69. 2016, World Bank, Washington, DC. Spitz-Oener, Alexandra. 2008. “The Returns to Pencil Use PricewaterhouseCoopers. 2015. “Global Entertainment Revisited.” Industrial and Labor Relations Review 61 (4): and Media Outlook 2015–2019.” http://www.pwc 502–17. .com/gx/en/global-entertainment-media-outlook/. Srour, Ilina, Erol Taymaz, and Marco Vivarelli. 2013. Raja, Deepti Samant. 2015. “Bridging the Disability “Skill-Biased Technological Change and Skill- Divide through Digital Technologies.” Background Enhancing Trade in Turkey: Evidence from Longitu- paper for the World Development Report 2016, World dinal Microdata.” IZA Discussion Paper 7320, Insti- Bank, Washington, DC. tute for the Study of Labor, Bonn. Raja, Siddhartha, Saori Imaizumi, Tim Kelly, Junko Statista. 2015. New York, http://www.statista.com Narimatsu, and Cecilia Paradi-Guilford. 2013. “Con- /statistics/285344/sources-used-to-plan-leisure necting to Work: How ICTs Could Help Expand -personal-business-trips-2013/. Employment Opportunities.” World Bank, Washing- Stevenson, Betsy. 2009. The Internet and Job Search. In ton, DC. http://documents.worldbank.org/curated Labor Market Intermediation, edited by David Autor, /en/2013/09/18221189/connecting-work-information 67–86. Chicago: University of Chicago Press. -communication-technologies-help-expand Summers, Lawrence. 2014. “Lawrence H. Summers on -employment-opportunities. the Economic Challenge of the Future: Jobs.” Wall Rendall, Michelle. 2010. “Brain versus Brawn: The Real- Street Journal, July 7. http://online.wsj.com/articles ization of Women’s Comparative Advantage.” Work- /lawrence-h-summers-on-the-economic-challenge-of ing Paper 306, University of Zurich, Department -the-future-jobs-1404762501. of Economics, Center for Institutions, Policy and Tadesse, Getaw, and Godfrey Bahiigwa. 2015. “Mobile Culture in the Development Process. Phones and Farmers’ Marketing Decisions in Ethio- Research ICT Africa. Various years. Household survey. pia.” World Development 68: 296–307. Capetown, South Africa. http://www.researchict Thompson, Derek. 2015. “The Death of Music Sales.” africa.net/home.php. Atlantic, January 25. Ritter, Patricia, and Maria Guerrero. 2014. “The Effect Tinbergen, Jan. 1975. Income Distribution. Amsterdam: of the Internet and Cell Phones on Employment and North Holland Publishers. EXPANDING OPPORTUNITIES 145

Turner, Adair. 2014. “The High-Tech, High-Touch Econ- ————. Various years. Europe and Central Asia Poverty omy.” Project Syndicate, April 16. (ECAPOV) database, World Bank, Washington, DC. Uruguay, National Institute of Statistics. Various years. ————. Various years. I2D2 (International Income Dis- Survey on the Use of Information and Communi- tribution Database). World Bank, Washington, DC, cation Technology. National Institute of Statistics, http://econ.worldbank.org/WBSITE/EXTERNAL Montevideo. /EXTDEC/EXTRESEARCH/0,,contentMDK: Valerio, Alexandria, Katia Herrera-Sosa, Sebastian 20713100~pagePK:64214825~piPK:64214943~the Monroy-Taborda, and Dandan Chen. 2015a. “Armenia SitePK:469382,00.html. Skills toward Employment and Productivity (STEP) ————. Various years. Skills Towards Employability and Survey Findings (Urban Areas).” World Bank, Wash- Productivity (STEP) household surveys (database), ington, DC. World Bank, Washington, DC, http://microdata ————. 2015b. “Georgia Skills toward Employment and .worldbank.org/index.php/catalog/step/about. Productivity (STEP) Survey Findings (Urban Areas).” ————. Various years. South Asia Region MicroDatabase World Bank, Washington, DC. (SARMD) (database), World Bank, Washington, DC. Valerio, Alexandria, Maria Laura Sanchez, Namrata ————. Various years. Survey-based Harmonized Indi- Tognatta, and Sebastian Monroy-Taborda. Forthcom- cators Program (SHIP) (database), World Bank, ing. “The Skills Payoff in Low and Middle Income Washington, DC, http://web.worldbank.org/WBSITE Countries: Empirical Evidence Using STEP Data.” /EXTERNAL/COUNTRIES/AFRICAEXT/EXTPUBREP World Bank, Washington, DC. /EXTSTATINAFR/0,,contentMDK:21102610~menu Van Welsum, Desiree. 2015. “Sharing Is Caring? Not PK:3084052~pagePK:64168445~piPK:64168309~the Quite. Some Observations about the Sharing Econ- SitePK:824043,00.html. omy.” Background paper for the World Development ————. Various years. World Development Indicators Report 2016, World Bank, Washington, DC. (database). World Bank, Washington, DC, http://data Varian, Hal. 2011. “Economic Value of Google.” In “The .worldbank.org/data-catalog/world-development Data Frame: The Web 2.0 Summit 2011 Conference,” -indicators. San Francisco, October 17–19. http://www.web2 ————. 2010. Skills Demand Survey for FYR Macedonia. summit.com/web2011. Washington, DC: World Bank. Veeraraghavan, Rajesh, Naga Yasodhar, and Kentaro ————. 2011. World Development Report 2012: Gender Equality Toyama. 2009. “Warana Unwired: Replacing PCs with and Development. Washington, DC: World Bank. Mobile Phones in a Rural Sugarcane Cooperative.” ————. 2012. World Development Report 2013: Jobs. Washing- Information Technologies and International Development ton, DC: World Bank. 5 (1): 81–95. ————. 2013. World Development Report 2014: Risk and WEF (World Economic Forum). Various years. Competi- Opportunity: Managing Risk for Development. Washing- tiveness Index (database). WEF, Davos, http://www ton, DC: World Bank. .weforum.org/reports/global-competitiveness-report ————. 2014a. The Opportunities of Digitizing Payments. -2014-2015. Prepared by the World Bank Development Research Weiberg, Bruce. 2000. “Computer Use and the Demand Group for the G20 Australian Presidency. Washing- for Female Workers.” Industrial Relations Review 53 (2): ton, DC: World Bank. 290–308. ————. 2014b. Levelling the Field: Improving Opportunities World Bank. Various years. Central Asia World Bank for Women Farmers in Africa. Washington, DC: World Skills surveys (database), World Bank, Washington, Bank. DC. ————. 2014c. Youth Employment in Sub-Saharan Africa. ————. Various years. East Asia Pacifi c Region Micro- Washington, DC: World Bank. Database, World Bank, Washington, DC. 146 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 2

Education

Educational policy makers and planners face a per- island nation of Papua New Guinea. These areas are sistent challenge related to the potential use of infor- characterized by poverty, low levels of literacy, geo- mation and communication technologies (ICTs) in graphic remoteness, linguistic diversity (over 800 remote, low-income communities around the world: languages are spoken across the island nation), low most products, services, usage models, expertise, and teacher quality, and a lack of teaching and learning research related to ICT use in education come from materials. Many rural classrooms have few books— high-income contexts and environments. One conse- and some none at all. Few students read at grade level, quence is that “solutions” enabled by technology are and teachers often do not know what materials they imported and “made to fi t” in environments that are are meant to cover in a given week—or how to teach often much more challenging. Sometimes this works; them effectively. Most of them do, however, have sometimes it doesn’t. The One Laptop per Child proj- mobile phones. ect in Peru provided hundreds of thousands of pieces The SMS Story project demonstrates that it is pos- of low-cost computing equipment to students in sible to orient and motivate teachers in useful ways rural schools. But early research found no evidence using technologies already at hand. It sent daily text of increased learning in math or language.1 This is message and teaching tips to teachers as an aid one high-profi le example of the diffi culties faced in to help improve student reading. The result? Teachers introducing hardware-centric educational technology were reminded and motivated to teach reading every projects conceived in highly developed environments single day. While reading comprehension did not into less developed places without suffi cient atten- noticeably improve, the intervention halved the num- tion to local contexts. ber of children who could not read anything. While An alternate approach would consider how to inno- this was hardly a “silver bullet” solution, SMS Story vate using existing technology that is already avail- offers one model for using connected technologies able in a local environment. In addition to exploring in simple ways to help address some long-standing, the uses of new technologies, it might also be useful seemingly intractable challenges facing educators to ask, How can we innovate using what is already avail- in rural schools in Papua New Guinea. While no one able? In many low-resource communities, the best would suggest that these results imply that these technology is the one that people already have, know schools do not need books, such simple interventions how to use, and can afford. In most circumstances, can be modestly transformative in ways that are this is the mobile phone. The SMS Story project in inclusive, effi cient, and scalable.2 rural Papua New Guinea is one example of an inno- The evidence base for investments in educational vative approach to using a “new” technology in ways technologies in middle- and low-income countries that meet local needs and that the original designers is weak—but growing. Until recently, there were of the technologies may not have dreamed of. few rigorous randomized studies on the impact of Few educational environments are more challeng- investments in educational technologies that could ing than those found in remote locales in the Pacifi c inform related policy decisions; now there are almost two dozen. In addition, a signifi cant body of practical This sector focus was contributed by Michael Trucano. evidence has also been collected based on experiences EDUCATION 147 in middle- and low-income countries. Characteristics educational planners to keep up. When it comes to of many successful technology-enabled educational the use of technology in education, educational sys- projects in such places include: tems may therefore be fated to exist effectively in a state of permanent experimentation. That said, exper- • A focus on the “guided use” of technology, and not iments should, by defi nition, teach something. A sad just providing tools for general use; fact of too many experiments in educational tech- • Providing relevant curricular materials; nology use around the world remains that they have • The shared use of devices in school settings; implemented technological “solutions” to problems • Attention to pedagogy, teacher support, and that have not been well understood. If you are pointed development; in the wrong direction, technology may help you get • Using technologies in ways that are supplemental there more quickly. At a fundamental level, many of and practical; and these efforts are not really failures of technology, but • Using evaluation mechanisms that go beyond rather a result of poor planning and an inability to outputs.3 learn from failure and adapt. As such, they are not a result of technology failures, but rather human ones. A number of key challenges remain that inhibit the potential positive impact of new technologies Notes on teaching and learning. Foremost among these are a pernicious focus on providing technology alone, 1. Cristia and others 2012. and a related that educational challenges can 2. Kaleebu and others 2013. be overcome simply by providing more and better 3. Drawn from Arias Ortiz and Cristia (2014). devices and connectivity. The so-called Matthew 4. McEwan 2014; Trucano 2015. Effect of educational technology holds that those most likely to benefi t from the use of new technol- References ogies in educational settings are those who already enjoy many privileges related to wealth, existing lev- Arias Ortiz, Elena, and Julián Cristia. 2014. The IDB and els of education, and prior exposure to technology in Technology in Education: How to Promote Eff ective Pro- other contexts. Policies that neglect to consider this grams? Washington, DC: Inter-American Develop- phenomenon may result in projects that exacerbate ment Bank. existing divides within an education system, and Cristia, J., P. Ibarraran, S. Cueto, A. Santiago, and E. indeed within larger society. Severin. 2012. “Technology and Child Development: Education challenges cannot be overcome simply Evidence from the One Laptop per Child Program.” IDB by providing more and better ICT devices and con- Working Paper IDB-WP-304, Inter-American Devel- nectivity. There is sometimes talk that technology will opment Bank, Washington, DC. replace teachers. In reality, experience from around the Kaleebu, N., A. Gee, R. Jones, and A. H. A. Watson. 2013. world demonstrates that, over time, the role of teachers SMS Story Impact Assessment Report, VSO, Papua New becomes more central—and not peripheral—as a result Guinea. Papua New Guinea Department of Education, of the introduction of new technologies.4 That said, VSO (Voluntary Services Overseas), and Australian while technology will not replace teachers, teachers who use Aid. http://www.vsointernational.org/sites/vso technology will replace those who do not. These teachers, in _international/files/sms-story-impact-assessment addition to having a suite of basic technology-related -report_tcm76-41038_0.pdf. skills, will be asked to take on new, often more sophis- McEwan, P. 2014. “Improving Learning in Primary ticated duties and responsibilities in ways that will Schools of Developing Countries: A -Analysis challenge the existing capacity of many educational of Randomized Experiments.” Review of Educational systems to prepare and support teachers over time. Research Month (October 7), Wellesley College, Welles- Understanding the local education challenge ley, MA. and context before proposing a technology-driven Trucano, M. 2015. “Mobile Phones & National Educa- solution is a critical fi rst step. A number of recent tional Technology Agencies, Sachet Publishing & the articles in the popular press have proclaimed many Khan Academy: What’s Happening with Educational high-profi le efforts to use new technologies within Technology Use in Developing Countries?” Excerpts educational systems, including variations of the One from the World Bank’s EduTech blog (Volume VI). Laptop per Child initiative in countries around the World Bank, Washington, DC. http://siteresources world, as “failed experiments.” The speed of techno- .worldbank.org/EDUCATION/Resources/EduTech logical change almost always outpaces the ability of Blog2014_all_the_posts.pdf. 148 WORLD DEVELOPMENT REPORT 2016

ENABLING DIGITAL DEVELOPMENT Social media

When considering another person in the world, harassment, bullying, and infectious disease, as well a friend of your friend knows a friend of as job opportunities and innovative ideas. their friend, on average. One of the clearest channels by which networks affect economic development is the diffusion of —Lars Backstrom and colleagues, innovations. Networks provide the medium through “Four Degrees of Separation” (2012) which people learn of an innovation. Adoption by a trusted person then initiates a chain reaction of That remarkable statement cannot yet be applied to adoptions through the network, accelerating as peer everyone in the world, but a 2011 study showed that pressure bears on laggards. The impacts can be large. it pertained to the active users of Facebook—some 721 An experimental study on the adoption of weather million people at the time, and their 69 billion friend- insurance (an unfamiliar innovation) by Chinese ships.1 The average of 3.7 intermediary links between farmers found that, while attending an intensive randomly selected people in the global Facebook pop- information session on the product boosted take-up ulation contrasts with the 5.7 links between Nebraska rates by 43 percent, merely having a friend who residents and Bostonians documented in Milgram’s attended such a session had almost half that effect, SPOTLIGHT 3 famous “six degrees of separation” study from 1967.2 and was equivalent in impact to a 15-percent subsidy.4 From 2011 to early 2015, the network’s user base had Impacts were larger when those initially informed doubled, with each user having an average of 338 were more central to the network. Social networks are “friends.” This is more than twice Dunbar’s number, also important channels for transmission of health which asserts that 150 is the maximum number of stable relationships that humans can retain.3 These behavior, and there is a tendency for people with poor statistics show some of the ways in which social net- health behavior to clump in networks isolated from 5 works are expanding and changing with the advent of those with healthier behaviors. social media on the internet. Enter the internet, in the form of Facebook, Social networks are fundamental to human soci- LinkedIn, and Twitter based in the United States; ety. They constitute the fabric of relationships that and WeChat in China; VK in the Russian support trust, reputation, and social cohesion. For Federation; and many other social media platforms instance, a debtor may be less likely to default on a designed to encourage the formation of social links. loan if both she and her lender share a dense network There are many types of social media, but social net- of mutual friends. Social networks exert working sites and micro blogs are most relevant in on behavior and shape aspirations for work, marriage, the context of social and economic development. A and consumption. By defi ning social boundaries, simple typology is based on whether communication they can be instruments both of social inclusion and is directed at specifi c recipients, and whether ties are exclusion. They also are a conduit of gossip, slander, explicit (prompted by invitation, acceptance, or refer- ence within a social network) or not (table S3.1). WDR 2016 team, incorporating contributions from Robert Social scientists distinguish between weak ties and Ackland and Kyosuke Tanaka. strong ties in social networks. Weak ties exist between people’s voice. in emergencysituations,andhowtheycanincrease how theypromptchangesinbehavior,help evidence onhowtheyaffecteconomicdevelopment, social mediaisstillatanearlystage,buttheresome network. Researchonthedevelopmentimpactsof visibility ofopinionsandsomebehaviorsacrossthe information betweenlinks.Third,theyincreasethe physical distance.Second,theyspeedthediffusionof works, sincelinksareveryeasytoform,regardlessof encourage expansionofthescopeanddensitynet- social networksinatleastthreeways.First,they megaphone shouting“Checkthisout!” ter islikesomeonestandingatastreetcornerwith surrounded bypeopleyouknowandlike,whileTwit- compared Facebooktoacocktailpartywhereyouare ties, whilemicroblogsfosterweakties.Oneobserver are consideredmoreimportantfordevelopingstrong groups linkedbystrongties.Socialnetworkingsites ior withinthegroup.Socialcapitalislikelyhigherin high reputationalcostofbadorunproductivebehav- trust andfacilitatecoordination,inpartbecauseofthe to outsiderscanbeimportant,asthismayincrease ties within“closed”groupswithrelativelyweak group arelikelytohavethesameinformation.Strong source ofnovelinformation,sincemembersaclose emotional supportbutmaybelessimportantasa (family, friends,orclosecolleagues),areasourceof ties, betweenpeoplewhohavemoreincommon ple withwhominteractionismoresporadic.Strong new knowledge,sinceinformationcomesfrompeo- information abouttechnologiesorotherusefuland than friends.Suchtiescanbeusefultotransmitnovel people whocouldbeconsideredacquaintancesrather Source: harmful, interactions. ResearchinNigeriashowed costs foreconomically benefi and informationexchange,thus reducingtransaction Social mediaactasachannel forcommunication development economic Social and media Table S3.1 Table Type ofties Implicit Explicit Social mediaplatformschangethedynamicsof Ackland and Tanaka 2015. Tanaka and Ackland 6 Relationships in diff redhpntok Fcbo,Gol+ Microblognetworks(SinaWeibo, Twitter)social taggingsystems) Semantic networks(recommendationsystems, Friendship networks(Facebook,Google+) cial,aswellpotentially Directed erent types of social media media ofsocial types erent Direction oftiesandexamples including about consumer preferences and complaints. including aboutconsumerpreferencesandcomplaints. sites canalsobeasourceforeconomicallyusefuldata, not betransferabletoeveryotherplace.Socialmedia study—they weremiddle-classurbanresidents—may for theirproducts,althoughthespecifi c contextofthis preneurial activitiesandhelpedthemfi that theiruseofsocialmediaencouragedentre- owners. ing informationfl that socialmediasupportedmicrofinance byfacilitat- cials, citizens,and theprivatesector. radio ortelevision, butinalldirectionsbetween offi fl key contributionisthatinformation doesnotjust tool, suchastochannelrequests forassistance.The disseminating informationand asamanagement urgent crises,socialmediaplatforms areusefulfor Before, during,andafternatural disastersorother emergencies Social and media tured network. practice diffusedmorequicklythaninalooselystruc- were networkedintightclusters,apromotedhealth online experimentshowedthatwhenparticipants to encouragedesiredoutcomes.Forinstance,an show that,inprinciple,networkscanbemanipulated competition tomaintainmotivation.Experiments their socialnetwork,thususingpeerpressureand progress inmeetingexerciseorlearninggoalswith instance, individualscanbeencouragedtosharetheir in waysthatareconsistentwithdevelopment.For ships. Butonlinebehaviorcanalsobechanneled personal preferences,buyinghabits,andrelation- they exploitthehugeamountsofinformationabout of socialmedia—oftenformarketingpurposes—as Commercial venturestakeadvantageofthefeatures behavioral change Social and media ow onewayfromoffi cials to citizensasthrough A study of a group of women in Jakarta showed A studyofagroupwomeninJakartashowed News groups,blogs 7

owtothepoorandsmallbusiness Undirected SOCIAL MEDIA nd customers ndcustomers -

SPOTLIGHT 3 149 150 WORLD DEVELOPMENT REPORT 2016

Social media and Notes community voice 1. By March 2015, Facebook had 1.4 billion active monthly users worldwide. Recent events such as the Occupy Wall Street move- 2. Milgram 1967. ment or the Arab Spring have raised questions about 3. Dunbar 1992. the role of social media in facilitating social change. 4. Cai, de Janvry, and Sadoulet 2015. Some analysts think social media played a central 5. Centola 2011. role, including by spreading democratic ideals across 6. See Ackland and Tanaka 2015 for individual borders. They also make movements less dependent references. on charismatic leaders or ideologues. Others are more 7. Centola 2010. skeptical, arguing that revolutions occurred long before social media, and the apparent inability to References form strong leadership through such tools can reduce the chances for lasting change. An additional issue is Ackland, Robert, and Kyosuke Tanaka. 2015. “Develop- the role of social media in spreading false information ment Impact of Social Media.” Background paper or to correct misinformation. The intensity of a crisis for the World Development Report 2016. World Bank, changes the quality of information shared on Twitter. Washington, DC. And, interestingly, fact-checking conversations on Backstrom, Lars, Paolo Boldi, Marc Rosa, Johan Ugan- Twitter are more likely to occur between strangers der, and Sebastiano Vigna. 2012. “Four Degrees of than between friends. Whether social media can Separation.” Presented at ACM Web Science Con- encourage collective fact-checking will infl uence its ference 2012, Evanston, IL, June 22–24. role in social learning in the context of social and eco- Cai, Jing, Alain de Janvry, and Elisabeth Sadoulet. nomic development. 2015. “Social Networks and the Decision to Insure.” There is still much to learn about the role that American Economic Journal: Applied Economics 7 (2): social media can play in development. One important 81–108. lesson is that their impact is country-specifi c and Centola, Damon. 2010. “The Spread of Behavior in an context-specifi c. Variations in access to technology Online Social Network Experiment.” Science 329 and education obviously matter. But there are also (5996): 1194–97. large variations in how people use social media and ———— . 2011. “An Experimental Study of how they share information. People will more likely in the Adoption of Health Behavior.” Science 334 share information broadly if they think they can (6060): 1269–72. infl uence events or policy. There is evidence that peo- Dunbar, Robin. 1992. “Neocortex Size as a Constraint

SPOTLIGHT 3 SPOTLIGHT 3 ple in more authoritarian countries are less likely to on Group Size in Primates.” Journal of Human Evolu- forward (re-tweet) information. tion 22 (6): 469–93. Finally, general-purpose social media platforms Milgram, Stanley. 1967. “The Small World Problem.” may have less of an impact in poor communities Today 2 (1): 60–67. compared with more targeted social media interven- tions. Examples in other parts of the Report show that purpose-built platforms set up by public institutions or development agencies as part of program deliv- ery may be more effective. They could, for instance, deliver extension services to farmers or connect par- ticipants with leaders to foster aspirational change. SOCIAL MEDIA 151 ENABLING DIGITAL DEVELOPMENT SPOTLIGHT 3 SPOTLIGHT 3 CHAPTER 3 Delivering services

Have digital technologies made governments more ing communication costs through digital platforms, willing and able to deliver services to their citizens? they enable citizens to connect with one another at Not yet. Governments have invested heavily in dig- unprecedented scale, fostering citizen voice and col- ital technology over the past two decades, and these lective action. efforts have made it easier in many countries for busi- But the impact of these mechanisms on capability nesses to fi le taxes, and for the poor to get an offi cial and empowerment depends on the strength of gov- identity allowing them to receive welfare payments ernment institutions. Institutions are the formal and and vote in elections. Digital technologies have also informal rules that shape the incentives and behav- enabled governments to receive regular feedback iors of politicians, government offi cials, and citizens. from service users, improving service quality. But Strong institutions create incentives for politicians all too often the successes are isolated—limited to a to work toward the public good and for bureaucrats few activities, sectors, or locales. Many investments to be accountable to those politicians. Digital tech- in e-government fail to have any impact other than nologies, aligned with the incentives of politicians, wasting scarce fi scal resources. Digital channels government offi cials, and service providers, can thus for mobilizing citizens to pressure policy makers to be highly effective in improving outcomes. By con- become more responsive have had little effect. And trast, politicians in clientelist political institutions the new technologies have enhanced the capacity are largely accountable to a small set of elites and of unaccountable governments for surveillance and reject digitally enabled reforms that hurt vested inter- control. In sum, digital technologies have helped will- ests; instead, they use these technologies for greater ing and able governments better serve their citizens, control. And patronage-based bureaucracies resist but they have not yet empowered citizens to make e-government advances that would reduce discretion unwilling governments more accountable. and rent-seeking. This misalignment between digital Effi cient service delivery requires a capable gov- technologies and weak or unaccountable institutions ernment that can implement policies and spend pub- creates the twin risks of increasing elite control lic resources effectively. It also requires an empow- and wasting scarce public resources on ineffective ered citizenry able to hold politicians and policy e-government projects. makers to account so that governments serve citizens Why do these institutional constraints persist? and not themselves or a narrow set of elites.1 Digital Digital technologies, particularly social media, have technologies can strengthen government capability uncovered egregious abuses of public authority and and empower citizens through three mechanisms triggered both virtual and physical citizen protests (fi gure 3.1). They overcome information barriers and around scandals. But they haven’t sustained collective promote participation by citizens in services and in action around failures of service delivery. Service elections. They enable governments to replace some delivery problems are more diffi cult to monitor and factors used for producing services through the auto- attribute to specifi c government offi cials, and only mation of routine activities, particularly discretion- rarely become salient in fl ashpoint events that trig- ary tasks vulnerable to rent-seeking, and to augment ger protests—like the excessive use of force by police. other factors through better monitoring, both by And they cannot be improved by one-off actions that citizens through regular feedback on service quality change a bad policy or pass a new law. Addressing ser- and within government through better management vice delivery failures requires sustained citizen voice of government workers. And by dramatically lower- through effective offl ine mobilization by civil society DELIVERING SERVICES 153 organizations. But in the absence of willing and able Figure 3.1 A framework for digital technologies and governments to partner with, civil society initiatives government service delivery often remain small, with limited citizen interest or impact. DIGITAL The dependence of digital technology on initial TECHNOLOGIES institutional conditions, though, varies by service and activity. For services and activities based on more routine tasks that are easy to monitor, digital technologies can improve outcomes rapidly and Overcome Augment Generate signifi cantly—even when institutions are relatively information barriers existing factors economies of scale weak. In effect, the technology substitutes for the institutions, as with cash transfers, licensing and registration services, and monitoring elections, all INCLUSION EFFICIENCY INNOVATION much improved by digital technologies in many low-income countries. But for services and activities that require more discretion from workers and are Participation State capability Voice hard to monitor, the quality of institutions is much more important, and digital technologies have an RISK: ELITE CONTROL AND FISCAL WASTE incremental effect. The technology only comple- ments or augments the initial institutions, as with Source: WDR 2016 team. teaching, health care, and organizational manage- ment, where digital technologies have an impact than in transactional government-to-citizen and only if institutions are strong. government-to-business services like e-fi ling of taxes Given the importance of institutions for service and portals that enable citizens to access a variety of delivery, the policy agenda is to use digital technol- services from one website (fi gure 3.2). Policy priorities ogies to strengthen institutions. The variation in are also evident in cross-national patterns of the interaction of institutions and digital technologies detailed indexes of different e-government systems by service and activity, and the considerable hetero- constructed for this report.3 The level of e-government geneity of institutions even within countries, can unsurprisingly increases with per capita income, guide policy and open many possibilities for service but the gap between high-income and low-income improvements in even the most challenging con- texts—in short, for digital dividends. Figure 3.2 Low-income countries have invested Connected governments heavily in e-government Main e-government systems, 2014 The advent of the internet in the mid-1990s triggered the rapid diffusion of e-government systems to auto- e-customs mate core administrative tasks, improve the delivery Financial management of public services, and promote transparency and accountability. By 2014, all 193 member-states of the e-procurement United Nations had national websites: 101 enabled Tax management citizens to create personal online accounts, 73 to fi le Human resources income taxes online, and 60 to register a business.2 In all, 190 countries had automated government fi nan- e-ID cial management, 179 had automated customs, and e-filing 159 had automated tax systems. And 148 countries had digital identifi cation schemes, although only 20 e-services portal had multipurpose digital identifi cation for such ser- 0 20 40 60 80 100 vices as voting, fi nance, health care, transportation, Percent and social security. High-income Middle-income Low-income Developing countries have invested more in Sources: WDR 2016 team, based on World Bank Global e-Government Systems database 2015b; World core government administration systems like fi nan- Bank Global Digital Identifi cation for Development (ID4D) database 2015c; UN 2014. Data at http://bit.do cial management, customs, and tax management /WDR2016-Fig3_2. 154 WORLD DEVELOPMENT REPORT 2016

Figure 3.3 The priority in low-income countries has been core e-government systems, 2014

Indexes of e-government systems, by income and region

a. Core systems b. Online services c. Digital ID

All countries All countries All countries

High-income: OECD High-income: OECD High-income: OECD

Upper-middle-income High-income: Non-OECD High-income: Non-OECD

Lower-middle-income Upper-middle-income Upper-middle-income

High-income: Non-OECD Lower-middle-income Lower-middle-income

Low-income Low-income Low-income

North America North America South Asia

Europe and Central Asia Europe and Central Asia Middle East and North Africa

Latin America and the Caribbean Middle East and North Africa Europe and Central Asia

South Asia Latin America and the Caribbean Latin America and the Caribbean

Middle East and North Africa East Asia and Pacific Sub-Saharan Africa

East Asia and Pacific South Asia East Asia and Pacific

Sub-Saharan Africa Sub-Saharan Africa North America

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 Index Index Index

Sources: WDR 2016 team, based on World Bank (2015b) Global e-Government Systems database; World Bank Global Digital Identifi cation for Development (ID4D) database; UN 2014. Data at http://bit.do/WDR2016-Fig3_3. Note: and right of each box plot represent the fi rst and third quartiles, and the line in the middle of each box plot is the median. The two ends of the whiskers represent the 10th and the 90th percentile observations. The core e-government index in panel a measures the automation and integration of countries’ fi nancial management information systems (including budget formulation and execution functions, source of public fi nance data, and treasury account structure); human resources management information systems (including personnel management and payroll); e-tax systems (including e-fi ling and tax administration systems); e-customs; and e-procurement. The online services, panel b, index is prepared by the United Nations Department of Economic and Social Aff airs as part of its overall e-government index. The Digital ID index, panel c, measures the extent to which electronic identifi cation can be used to access services, whether it features magnetic-strip or smart-chip technology, and whether it has embedded digital signature capabilities. OECD = Organisation for Economic Co-operation and Development.

countries is lowest for administrative systems and For example, more than 60 percent of public sector highest for online services for citizens and businesses, workers are in jobs that require intensive use of tech- with digital identifi cation systems in the middle (fi g- nologies in Bolivia, Colombia, and Kenya, compared ure 3.3). Fifteen middle-income countries score in the with only approximately 20 percent of private sector top quartile of administrative e-government systems, workers. But the gap persists in 22 of the countries with Brazil, Colombia, Mexico, and Peru in the top 10 when the public sector is compared only with large worldwide, and Nepal, Rwanda, and Uganda scoring private sector fi rms (those with more than 50 employ- highly among low-income countries. India and Paki- ees). Governments are also more intensive users of stan have more advanced digital identifi cation sys- technology in high-income countries like Germany, tems than high-income North American countries, Japan, and the United States. pointing to the potential for “leapfrogging” that these Whether this substantial investment in technologies offer. e-government has strengthened state capability and Governments are more intensive users of infor- improved outcomes will be explored at length in the mation technology than fi rms. Survey data on the use next section. One indication of the limited return of digital technologies (computers, mobile phones, on these investments is citizen use of e-government and the internet) in nonagricultural jobs from 30 services, which has lagged behind supply (fi gure 3.5, countries show that in all but one country, a higher panel a). Even in countries where internet access is not share of public sector workers are in jobs that require a constraint, as in Europe, citizens use e-government frequent use of the technology in their daily work as mostly to get information (on average 44 percent of compared to private sector workers (fi gure 3.4). This individuals in the European countries for which data gap is not surprising for developing countries, where are available visited a government website at least once most fi rms are small and in the informal sector. in the past year to get information) and not for trans- DELIVERING SERVICES 155

Figure 3.4 Governments use digital many citizens use online services in the richest coun- technologies more intensively than tries than in the poorest, with a similar gap between private sector fi rms, 2014 the highest and lowest income quartiles within countries (fi gure 3.5, panel b). Within countries, use Share of workers in jobs that require intensive use of digital technologies (private or public sector) is unsurprisingly associated with age, education, and urban residence. Businesses are more likely Kenya than individuals to use the internet to transact with Bolivia the government (fi gure 3.5, panel c), and the use of Colombia e-services by business is also less sensitive to income Vietnam United States across countries. Japan In low-income countries, limited internet access Estonia means that mobile phones, not the internet, drive Spain interconnectedness, as chapter 2 discussed. Surveys Canada of the use of digital technologies in 12 African coun- Netherlands Ghana tries reveal that only 5 percent of individuals had used Germany the internet to obtain information from, or interact Poland with, the government, compared to 63 percent of Macedonia, FYR respondents who had used their mobile phones to Lao PDR Italy contact health workers (fi gure 3.6). Internet-based Sri Lanka e-services will therefore be biased against the poor in United Kingdom developing countries. Mobile e-government portals Czech Republic have become more common, nearly doubling from Slovak Republic 25 countries in 2012 to 48 in 2014, but still are not on France 5 Belgium par with internet services. More than three-quarters Armenia of countries archive information on their websites Ireland for education, health, and fi nance services, but fewer Austria than half send e-mail or Rich Site Summary (RSS) Finland Georgia updates, and only a handful send short message ser- Denmark vice (SMS) updates. Norway Sweden Greater state capability and 0 20 40 60 80 Percent citizen participation Private sector firms What impact have these signifi cant investments in Large firms (50+ employees) Public institutions e-government had on strengthening government capability to deliver services and expanding citizen Sources: WDR 2016 team, based on STEP 2014 and PIAAC 2014 household surveys. Data at http://bit.do/WDR2016-Fig3_4. participation? If politicians and policy makers want Note: Excludes the agricultural sector. Public institutions include government to improve services, they need a capable government administration, public utilities, and state-owned corporations. The countries to convert policies and spending into better outcomes. are sorted in descending order of the gap between the public and the private sectors. Resources must be collected and spent without leak- age. Teachers and health workers must be present in classrooms and clinics, and then must be motivated acting with government (only 29 percent returned a to work. Benefi ciaries of government programs must form online in the past year). These patterns are also be properly identifi ed and reached. And governments refl ected in surveys of citizens in Australia, Canada, often must help citizens make better decisions, and New Zealand—countries that rank 8, 10, and 15, overcoming bad habits or a lack of information. But respectively, in the provision of online services. The as underlined in the 2004 World Development Report, majority of respondents use government websites for Making Services Work for Poor People, and subsequent information and prefer to use traditional channels like research, weaknesses in each of these links in the voice calls to undertake transactions such as register- service delivery chain make services fail in develop- ing their children for day care.4 ing countries. Improving government capability or Both between and within European countries, the implementation capacity is key for effective service use of e-services depends on income. Three times as delivery. 156 WORLD DEVELOPMENT REPORT 2016

Figure 3.5 Citizen use of e-government in Europe depends highly on income, 2014

a. e-government use, by type of activity c. e-government use across Europe, by type of user

Any business use 100 LTU e-procurement SVKSVN FIN SWE POL PRT FRAAUTNLD Obtain information EST GRC DNK LVA MLT BEL IRL NOR LUX HRVCZE Return forms DNK HUN NOR Obtain forms 80 CYPESP FIN SWE BGR DEU Any individual use ITA NLD Return forms MKD LUX Obtain forms FRA Obtain information 60 AUT SVK ROU BEL LVA SVN EST DEU IRL 0 20 40 60 80 100 HUN ESP GRC Percent of businesses and individuals PRT 40 LTU MLT CYP CZE b. e-government use, by characteristic of individual user HRV POL Total population ITA Percent of businesses and individuals and businesses Percent of 20 BGR Education: none and low MKD

55–74 years of age ROU Income: 1st quartile Rural 0 Female 0 20,000 40,000 60,000 80,000 Income: 4th quartile Income, per capita (PPP), US$ Education: high Businesses Individuals 0 20406080100 Percent of businesses and individuals

Source: WDR 2016 team, based on Eurostat (EC 2014). Data at http://bit.do/WDR2016-Fig3_5. Note: Panels a and b show how businesses (in blue) and individuals (in green) use e-services for 29 European countries by type of activity and individual characteristics. Panel c shows the variation in use across the European countries. Individuals refer to those aged 16 to 74; businesses refer to those with 10 or more employees. PPP = purchasing power parity.

Figure 3.6 Mobile phones are the main channel for Digital technologies can help improve govern- citizens to interact with governments in Africa, 2014 ment capability and citizen participation by:

• Informing citizens and giving them an offi cial identity Used internet to get so that individuals can make better decisions for information from government their health, their safety, and education for their Used internet to read or download children, and can access a variety of publicly and news, magazines, or books privately provided services; • Streamlining processes to reduce discretion and Used internet to opportunities for rent-seeking, ensuring that public interact with government resources are collected and spent effi ciently, with- Used mobile phone to out leakage; contact health care workers • Receiving feedback from service users to regularly track satisfaction, identify problems, and improve Used mobile phone to mobilize community service quality; • Improving service provider management through better 0 20 40 60 80 monitoring so that government workers both show Percent of individuals up at work and are productive. Sources: WDR 2016 team, based on Research ICT Africa surveys 2014 of 11 countries. Data at http://bit .do/WDR2016-Fig3_6. Digital technologies have had varying impacts Note: The left and right of each box plot represent the fi rst and third quartiles, and the line in the middle of each box plot is the median. The two ends of the whiskers represent the 10th and the 90th percentile across these channels (table 3.1). The clearest impact observations. involves addressing a lack of information and weak DELIVERING SERVICES 157

Table 3.1 The impact of digital technology on government capability to deliver services: A scorecard

Impact of Main problem Channel technology to address Do digital technologies solve the problem?

Informing citizens Poor information • Yes, when poor information is the main barrier to and giving them H and communication improving service outcomes an identity

Streamlining High transaction • Sometimes, given the considerable heterogeneity processes costs; rent-seeking of impacts across countries and the high risks and M rewards of reforms

Receiving user High transaction • Yes, when citizens have an incentive to complain feedback costs; rent-seeking and services can be easily monitored M • No, otherwise

Improving Information • Yes, for reducing ghost workers and absenteeism service provider L asymmetries management • No, for improving provider accountability for services that cannot be easily monitored

Source: WDR 2016 team. Note: Channels are arranged by degree of technology impact. L = low; M = medium; H = high. communication—citizens’ lack of knowledge about Informing citizens and giving them an issues that affect their welfare, their inability to com- identity municate with one another and with government, Digital technologies are enabling governments to and government’s inability to be informed about overcome barriers of geography, low physical con- citizens. For example, the provision of information nectivity, and limited administrative capacity to pro- to citizens, particularly to poor citizens in remote vide information and deliver services to previously locations through mobile phones, has helped people unreached citizens, improving citizen participation make better decisions on a variety of issues. And dig- and choice. The most popular and promising inno- ital identifi cation, by giving poor citizens a verifi able vations in poor countries entail the use of mobile identity, has enabled them to participate in public and phones to implement a variety of health interven- private services previously denied to them. tions—so-called “m-health” (mobile health) initiatives. Digital technologies have been less effective in These can focus on either health providers, such as solving government failures associated with rent- point-of-service data collection, disease surveillance, seeking and in strengthening the incentives of gov- health-promotion campaigns, and telemedicine; or ernment bureaucrats and service providers to per- on citizens, such as SMS reminders to parents to get form due to principal-agent problems associated with children vaccinated or to patients to undertake ther- information asymmetries. Automating government apy (sector focus 3). While hundreds of pilot m-health functions and services has been at best partially suc- initiatives are underway, robust evidence in devel- cessful; these are often complex and costly systems to oping country contexts is limited largely to the pos- implement. The failure rate is high, and so is the risk itive effects of reminding patients to adhere to their of wasting public funds. Citizen feedback has been antiretroviral therapy schedules.6 Initial results from effective when citizens have an incentive to provide pilots in Bangladesh, India, South Africa, and Tan- feedback and the services are fairly easy to monitor, zania suggest that regular SMS communication can and when governments have the capacity to respond induce similar behavior changes in expectant moth- to this feedback. It is not effective when these condi- ers by providing information on neonatal health.7 tions are absent. Aside from a few isolated examples The initial results are also promising from m-health of reducing absenteeism, digital technologies have initiatives that assist health providers in recording not signifi cantly improved service provider manage- patient information, monitoring pregnancies, and ment in government bureaucracies. reporting drug stock-outs.8 158 WORLD DEVELOPMENT REPORT 2016

Box 3.1 Digital technology and crisis management

Digital technologies have helped monitor and manage and many thousands homeless. Although donations came human and natural disasters, such as wars, fl oods, earth- in from all over the world, ensuring that aid reached the quakes, and epidemics. The internet can help in coordi- areas most aff ected proved a challenge. One response, nating data fl ows and improving the timeliness of decision coordinated by Living Labs,b was to develop making, while mobile phones can disseminate vital infor- a dynamic map (quakemap.org) as a centralized platform mation. Social media can also help survivors communicate for displaying incident reports and off ers of assistance. their status and mobilize volunteer resources and donations The map is built around the OpenStreetMap and uses outside the disaster zone. A few examples of responses to the Ushahidi platform for incident reports—both based recent crises show the range of possibilities: on open-source software and crowdsourced information.c • Recovering from confl ict and requires long- • The Ebola epidemic in Guinea, Liberia, and Sierra Leone term nation building, and more than 40 countries have created logistical challenges for the organizations that adopted Truth and Reconciliation Commissions (TRCs) responded, not least how to pay fi eldworkers in remote to assist in this process.d The Liberia TRC received some and rural locations, many of them not local. The United 20,000 statements from Liberians, at home and in Nations Development Programme set up a scheme to the diaspora. Many depositions were audio and video use mobile phones to make direct payments to the Ebola recordings, now stored as a permanent memorial at the response workers in Sierra Leone and Liberia, who num- TRC website (http://trcofl iberia.org/). Interestingly, the bered some 60,000 at the height of the crisis.a videos have been viewed four times more often than • The earthquakes that struck Nepal in April and May 2015 the TRC’s fi nal report, indicating the reach of multimedia left at least 8,500 people dead, more than 17,000 injured, for storytelling at the human scale.e

a. See http://www.sl.undp.org/content/sierraleone/en/home/presscenter/articles/2015/05/12/mobile-pay-for-thousands-of-ebola-workers.html. b. See http://www.kathmandulivinglabs.org/. c. See http://www.bbc.com/news/world-asia-32603870. d. Kelly and Souter 2014. e. Best 2013.

This use of digital technologies to overcome geo- Weaknesses in civil registration in developing graphic, infrastructure, and administrative bottle- countries mean that approximately 2.4 billion people necks to information and communication is particu- in the world do not have birth certifi cates and other larly powerful in emergencies like disease outbreaks, offi cial documents to prove who they are, preventing natural disasters, and confl ict (box 3.1). many from opening bank accounts, registering prop- Digital technologies can also strengthen social erty, or receiving public transfers.13 Digital identifi - norms and promote civic behavior like voluntary tax cation (ID) programs that use biometrics to register compliance or energy conservation through “peer and authenticate individuals, although complex and comparisons”—making an individual’s actions more thus risky to implement, provide a robust solution to observable to others and making others’ actions this problem. They enable poor countries to leapfrog more observable to the individual.9 For instance, tax the decades-long processes that developed countries evasion is pervasive in developing countries. Some went through to build their traditional, paper-based governments collect less than half the taxes due to civil registration systems (see spotlight 4). These them.10 Simple behavioral nudges, such as naming schemes are expanding; 148 of 197 countries have and shaming tax delinquents by publishing their some form of electronic ID. But only 20 countries names on websites, can be a cost-effective inter- have a multipurpose ID with digital signature capa- vention even in low government capacity contexts. bility that can be used for multiple online and offl ine Such peer comparisons have improved tax com- services; 100 have digital identifi cation systems lim- pliance in Norway and the United States, although ited to a few functions (services like voting or public the developing country evidence is more mixed.11 sector payroll that cover only subsets of the general Comparisons of a household’s power consumption population); and 49 have no digital ID of any kind.14 with that of neighbors, as done by companies like Evidence on the impact of these schemes, while Opower, can also help conserve energy through peer largely anecdotal, points to signifi cant potential ben- pressure.12 efi ts of reducing leakages in government welfare DELIVERING SERVICES 159

Box 3.2 Empowering women through digitally enabled social programs

Safety net programs are becoming more eff ective through have to wait in line for cash or worry about keeping their digital systems for registering, authenticating, and pay- payments safe. The Niger pilot led to improved nutrition for ing benefi ciaries, promoting inclusion and empowering children, in part because of the time savings for mothers. women. Female empowerment is one major benefi t of the BISP Consider the Absa Sekulula (“It’s easier”) welfare pay- and the Niger programs. Women in BISP mostly receive the ment card in South Africa, the Benazir Income Support payments on behalf of their households, increasing their Programme (BISP) in Pakistan, and a pilot mobile money authority to make decisions on how the cash is spent, their cash transfer experiment in rural Niger. The programs in mobility, and their engagement with formal institutions. Pakistan and South Africa rely on a debit card payment The Niger mobile transfer, by reducing the observability of mechanism linked to an effi cient management information the amount and timing of the cash transfer, increased wom- system and biometric identity database. When benefi - en’s bargaining power in the household, changing house- ciaries register, their personal data and biometrics are hold agricultural activities and expenditures in ways that captured and authenticated in minutes, a biometric debit benefi t women. To quote Amina Bibi, a BISP benefi ciary, card is issued in a few days, and the system automatically “Women are being encouraged and empowered. Because credits cardholder accounts monthly with the appropriate of this program people who did not let their women leave grant amount. In Pakistan, 4.5 million BISP benefi ciary the house earlier are allowing them to do so. Women now households receive their payments through debit cards have confi dence that they, too, can do something for the within 72 hours of disbursement from the government. So, betterment of their family.” the elderly, parents, and people with disabilities no longer

Sources: WDR 2016 team, based on http://worldbank.org/safetynets/howto; http://www.bisp.gov.pk; Oxford Policy Management 2015; Aker and others 2014.

programs and empowering women through fi nancial Figure 3.7 e-government systems increase the inclusion and access to government cash transfers (box transparency of government budgets, 2014 3.2). Biometric registration, authentication, and pay- 0.6 ments in India’s National Rural Employment Guarantee NZL Scheme, the world’s largest workfare program, reduced ZAF the time for paying benefi ciaries by 29 percent and leak- SWE NOR 15 ages by 35 percent (see spotlight 4). But these schemes 0.4 SVNUSA CZE are complex and have all the risks of wastage associ- SVK GBR UGABGR ated with large government information technology HRV CHL AF G PRT FRA JOR KOR projects. They also raise concerns about government 0.2 NAM PNG GEO BGD HND PER MW I MNG COL BIHBWAUKR GTMCRI surveillance, violation of privacy, and data integrity. KENGHA PAK POL MO ZTZAESP ALB A more recent initiative around digital informa- MLILBRLKA BRAKAZNPLROU SLV RUS PHL NIC SLE AZ E MEX ITAIDNDEU TTO SRB tional services is the open data movement, publishing 0 TLS MARTURMKD LBN government data in machine-readable format as a plat- VEN MYS STP DOM ECU form for the private sector to develop applications to AG O THA BFA Budget transparency (index) IND ZWE KGZ deliver new e-services to citizens. Countries continue to –0.2 COD TJK KHM VNM join the Open Government Partnership, but initiatives BOL YEM SEN TUN CMR RWA are in their infancy. According to the Global Open Data DZA FJI NER NGA ZMB TCD EGY Index, only 13 percent of the datasets in 110 countries –0.4 BEN CHN were “open”—public, machine-readable, and nonpropri- QAT etary.16 Many governments have no desire to open their –0.4 –0.2 0 0.2 0.4 data or lack the capability to do so. Making meaningful, Core e-government systems good-quality data publicly available depends on sound Sources: WDR 2016 team, based on Open Budget Index 2014 data (International Budget core administrative systems—countries with more Partnership 2014) and the World Bank (2015b) Global e-Government Systems database. Data at http:// advanced core systems have greater budget transpar- bit.do/WDR2016-Fig3_7. ency (fi gure 3.7)—and on effective interministerial Note: Controlled for income. Budget transparency is measured by the Open Budget Index. 160 WORLD DEVELOPMENT REPORT 2016

cooperation and coordination. The necessary support- routine manual processes prone to error and manip- ive ecosystem of a vibrant private sector and civil soci- ulation, reduces the number of government interme- ety to use open data is also often missing. diaries in a transaction, and establishes audit trails to The impact after governments open up is also curtail rent-seeking. Digital registration and payment unclear. In the United States and the United Kingdom, systems have, for example, signifi cantly reduced leak- the take-up of open data has been limited to a few sec- ages in transport subsidies for the poor in Colombia tors where there is consumer demand, like real estate, (box 3.3). Given the importance of mobilizing and man- transport, energy, and health care, so the private aging resources, almost all countries have tried to auto- sector has an incentive to develop applications on mate budget preparation, execution, and accounting the open data platform.17 In Kenya, Moldova, and the and some aspects of tax and customs administration. Philippines, which have opened hundreds of govern- Management information systems are also exten- ment datasets, the motive has been more to increase sively used in education, health, land management, transparency than to spur business opportunities for and social protection services. In the past decade, the private sector. In locales in Kenya and the Philip- one-stop shops or service centers that provide citizen pines, fewer than 15 percent of survey respondents and business services—such as registration, licensing, with internet access had heard of the Open Data Ini- records, bill payments—in a single physical space or tiative.18 Some media and civil society organizations web portal have become popular, with initiatives in use open data in their advocacy campaigns, but such many countries, including Azerbaijan, Bangladesh, initiatives generally are few and fragmented. Brazil, Cambodia, India, Kenya, Moldova, Mozam- bique, Oman, Peru, and Uganda (box 3.4). Streamlining processes Given the methodological diffi culties in measuring government productivity, the impact of e-government Automation applied to an effi cient operation will mag- must be assessed indirectly through, for instance, nify the effi ciency . . . automation applied to an ineffi cient changes in fi rms’ perceptions of tax compliance costs, operation will magnify the ineffi ciency. the competitiveness of government procurement, and —Bill Gates corruption in taxation and government contracting after the introduction of e-government systems in a The effi ciency of some government tasks and services country. These indirect measures, from regular sur- can be improved through automation that eliminates veys of fi rms in select European and Central Asian

Box 3.3 Targeted public transit benefi ts in Bogotá

Public transit is essential for 64 percent of the 8 million peo- makes it easier and more fl exible to provide targeted ple living in Bogotá, Colombia. As in many major cities, hous- demand-side subsidies with lower scope for leakage ing is often distant from work, especially for low-income resi- and abuse. Most public transit systems, in contrast, rely dents. But using public transit can be costly for the poor, who on undiff erentiated supply-side subsidies. Linking the spend about 25 percent of their income on transportation, transit user and SISBEN databases makes targeting effi - compared to 3.5 percent for high-income residents. To help cient. Introduction of the smart card system has not been overcome this spatial mismatch, the city built an integrated without startup problems. The smart cards do not work public transport system (SITP) anchored by the Transmilenio on the entire system yet. Someone traveling on diff erent bus rapid transit and collective bus transport. To increase portions (“Phases”) of Transmilenio and on the collective aff ordability for low-income users, the city introduced a system will need two smart cards and possibly cash as well. subsidy—in 2014; it gave a 50-percent discount for 40 trips These parallel systems need to be integrated into the SITP. per month. Eligibility is tied to a database of benefi ciaries of Technically, this is straightforward (for example, by using social programs (SISBEN), and by February 2015 it reached turnstiles that receive the smart cards in all the buses and about 26 percent of the target population. stations). The challenge is to achieve consensus between The subsidy is transferred on smart fare cards activated the bus operators, the collection fi rm in charge of the smart and replenished at charging stations. This technology cards, and the city’s government.

Source: Rodriguez and Monroy-Taborda, forthcoming, for the WDR 2016. DELIVERING SERVICES 161

Box 3.4 Streamlining services through one-stop service centers

With committed political leadership, one-stop computer- and ID services are most popular, with the majority of users ized service centers can provide citizens with access to a from lower-income groups. broad range of public services from multiple government The UAI were designed to overcome institutional con- departments at one location quickly and easily—expand- straints that plagued the state’s prior attempt at service ing options, saving time, speeding delivery, and reducing integration. The earlier generation of service centers was opportunities for corruption. staff ed by low-performing civil servants seconded by par- In rural Karnataka, India, for instance, the more than 800 ticipating agencies and lacked standard operating proce- Nemmadi Centers operating since 2006 provide birth and dures. Long wait times could be abused by citizens willing death certifi cates, caste certifi cates, income certifi cates, to pay to advance in the queue. ration cards, land records, and pensions, among other ser- The new system strengthened oversight from the vices. Compared with typical government offi ces, citizens Secretariat for Planning and Management, which reports get these services in the centers with 3.4 fewer visits, 58 directly to the governor. It introduced a fi rst-come, fi rst- fewer minutes spent in the visits, 23 days faster processing, served ticketing system. And citizen complaints submitted and 50 percent less chance of being asked for a bribe.a online must be resolved within two days. Coordinators at But this experience of Karnataka is not replicated in many the UAI “situation room” monitor the average wait time, other states of India. One study found that the scope and the number of citizens served, and the number of atten- impact of service centers vary considerably across states. dants on at each service center to reallocate resources The greater the prevalence of petty corruption, the fewer to manage demand in real time. A private contractor staff s services off ered, particularly services for which bribes can the service centers, thus bypassing restrictions for civil be extracted. servants and allowing the centers to off er extended hours The integrated citizen assistance units (unidades de to citizens. Each participating agency provides an on-site atendimento integrado, or UAI) in Minas Gerais, Brazil, supervisor to oversee its programs. Confl icts occasionally provide another model. With 30 shops and 1,800 employ- arise between UAI coordinators and the agency-employed ees, the UAI off er access to services from 15 government supervisors. As one government offi cial noted, “When you agencies. In 2012, they handled more than 6.2 million citi- tell a police offi cer he is no longer the authority in that zen transactions, up sevenfold in three years. Employment place, it doesn’t go down very well.”b

Sources: Bussell 2012; Majeed 2014. a. Bussell 2012. b. Majeed 2014.

countries, show that the impact of e-government is with Belarus, Kazakhstan, and having the mixed and varies considerably across countries.19 biggest declines in inspections, and Bosnia and Her- Filing of taxes electronically (e-fi ling) is meant to zegovina, Montenegro, and Tajikistan reporting an reduce the administrative burden to businesses and increase in inspections after the introduction of the citizens by minimizing interactions with tax offi cials e-government system. E-fi ling also lowered fi rms’ and reducing opportunities for rent-seeking—and perceptions of paying bribes to tax offi cials only mar- by lowering the time it takes businesses to fi le tax ginally, with medium-size fi rms and fi rms that export returns, make payments, and receive refunds. Indeed, less likely to pay bribes to tax offi cials after the intro- advanced e-fi ling systems in a country (e-fi ling of tax duction of e-fi ling. forms and e-payment of taxes) lowered the likelihood E-fi ling and e-payment help on average to reduce of being inspected by tax offi cials by an average 13 per- the time it takes businesses to prepare and pay taxes, cent in select countries in Europe and Central Asia.20 as measured by the assessments of experts in the But shifting only to the simpler e-fi ling system that Doing Business data for 75 countries that introduced did not enable online payment of taxes had no effect these systems between 2005 and 2014 (fi gure 3.9). (fi gure 3.8, panel a). Nor did e-fi ling on average reduce Overall the time taken declined by roughly 25 percent the number of visits by tax offi cials (fi gure 3.8, panel in the fi ve years after the system was introduced b). These effects vary considerably across countries, across the sample, with the time dropping by more 162 WORLD DEVELOPMENT REPORT 2016

Figure 3.8 The likelihood of being visited and the of taxpayers and tax offi cials to use these systems. number of visits by tax offi cials after e-fi ling was The importance of complementary reforms is illus- introduced vary in select European and Central trated by the contrasts of South Africa and Belarus Asian countries with Tajikistan and Uganda. South Africa’s tax mod- ernization reforms, which began in 2007, coupled the a. Likelihood of being visited by tax officials technology initiatives with simplifying the tax code e-filing Armenia and reforming business processes in the tax authority. Uzbekistan For example, revenue targets became a compulsory Kyrgyz Republic feature of the performance scorecards of managers. Kosovo Average (8 countries) Online tax fi ling increased from 40,000 individual tax Serbia returns in 2006 to more than 2 million in 2009, with Montenegro signifi cant reductions in tax compliance costs.22 In Bosnia and Herzegovina Belarus, e-fi ling was similarly part of a broader reform Tajikistan e-filing and e-payment to lower compliance costs for citizens, including sim- Kazakhstan plifying the tax code, setting up taxpayer facilitation Georgia services, and reaching out to the business community.

Average (3 countries) But in Tajikistan, e-fi ling was not mandatory, use was Belarus low, and most businesses continued to submit paper –20 –10 0 10 20 30 returns, in part because they did not trust that online Percent change submissions would be secure but also because tax Statistically significant inspectors continued to pester fi rms.23 And in Uganda, Statistically not significant the electronic forms were more complicated than the manual ones, and taxpayers had to fi le paper returns in addition to the e-fi ling.24 As a result, the time to b. Change in the number of visits by tax officials e-filing prepare and pay taxes increased in Uganda, and tax Kosovo inspections increased in Tajikistan. Serbia Effective service delivery requires effi cient and Uzbekistan transparent government procurement of goods and Armenia Average (8 countries) services. Procurement is particularly vulnerable to Kyrgyz Republic collusion and corruption; e-procurement, through Montenegro greater transparency and less discretion, can result Bosnia and Herzegovina in signifi cant budgetary savings and better-quality Tajikistan e-filing and e-payment infrastructure. In India and Indonesia, e-procurement Belarus increased competition by making information on gov- Kazakhstan ernment contracts widely available and by encourag- Average (3 countries) Georgia ing more fi rms, particularly those outside the region where the project was located, to bid. It also improved –80 –60 –40 –20 0 20 40 the quality of the infrastructure projects through Percent change shorter delays and better construction.25 Statistically significant Firms’ perceptions of the impact of e-procurement Statistically not significant in Europe and Central Asia, however, are quite mixed. Source: Kochanova, Hasnain, and Larson, forthcoming, for the WDR 2016. Data at http://bit.do Systems that provide only information on govern- /WDR2016-Fig3_8. ment contracts online or that also enable fi rms to bid Note: The panels show the impact of introducing electronic tax fi ling after controlling for fi rm-specifi c for these contracts online had no effect on average characteristics and fi xed eff ects. The bars stand for the average eff ects in each country. Based on a sample of 25,969 fi rms (panel a) and 14,232 fi rms (panel b) from the World Bank Enterprise Survey. on fi rm perceptions of the competitiveness of public procurement (fi gure 3.10, panel a). In Belarus, Georgia, and the Slovak Republic, fi rms were more likely to than half in Belarus (from 987 hours to 183 hours), bid for government contracts after an e-procurement Costa Rica (from 402 hours to 163 hours), and Kenya system was introduced, but in Armenia, Bosnia and (from 432 hours to 202 hours).21 Herzegovina, and Tajikistan they were less likely. To have impact, e-government systems need to Transactional e-procurement systems increased the be accompanied by regulatory and administrative competitiveness of procurement only in the higher- reforms—changing laws and management practices, income countries in the sample, suggesting that simplifying tax procedures, and increasing the ability nontechnological factors correlated with a country’s DELIVERING SERVICES 163 level of development determine their impact. Sur- Figure 3.9 e-fi ling and e-payment on average prisingly, e-procurement was not associated with reduced the time required to prepare and pay taxes any reduction on average in fi rm perceptions of cor- ruption in government contracting, but again with 125 considerable variation (fi gure 3.10, panel b). Nor were there signifi cant average differences across types of fi rms or in the functionality of the systems. This negative fi nding is striking because one of the main 100 reasons for investing in these systems is reducing corruption in contracting and increasing the value- for-money of government capital spending. E-procurement is also conditional on regulatory, 75 institutional, and human resource factors for impact. In one of India’s leading reformers in e-procurement,

Karnataka state, reform was underpinned by an (year of e-filing introduction = 100) Time required to prepare and pay taxes 50 amended procurement law that made it mandatory to use the government e-procurement portal for all −10 −5 0 5 10 state departments and agencies. Run from the chief Years before and after introduction of e-filing minister’s offi ce to give it authority, the system was Sources: WDR 2016 team, based on World Bank (2015b) e-Government Core Systems database and the implemented under a public-private partnership to Doing Business database for 2014 data. Data at http://bit.do/WDR2016-Fig3_9. bring in skills from the market.26 Note: Shaded area shows the 95 percent confi dence interval for 75 countries.

Figure 3.10 e-procurement has no eff ect on fi rms’ likelihood of bidding for a government contract or of being solicited for a bribe in select European and Central Asian countries

a. Change in the number of firms b. Change in the number of firms bidding for government contracts being solicited for bribes

Informational Informational Bosnia and Herzegovina Uzbekistan Croatia Bosnia and Herzegovina Average (7 countries) Moldova Moldova Azerbaijan Turkey Average (6 countries) Azerbaijan Croatia Kyrgyz Republic Kyrgyz Republic Uzbekistan

Transactional Transactional Tajikistan Russian Federation Armenia Slovenia Slovenia Tajikistan Russian Federation Belarus Estonia Macedonia, FYR Ukraine Average (11 countries) Average (12 countries) Ukraine Macedonia, FYR Mongolia Estonia Kazakhstan Georgia Georgia Armenia Belarus Mongolia Slovak Republic Kazakhstan

–20 –10 0 10 20 –100 –50 0 50 Percent change Percent change Statistically significant Statistically significant Statistically not significant Statistically not significant

Source: Kochanova, Hasnain, and Larson, forthcoming, for the WDR 2016. Data at http://bit.do/WDR2016-Fig3_10. Note: The fi gures show the impact of introducing informational and transactional e-procurement systems, after controlling for fi rm-specifi c characteristics and fi xed eff ects. Informational systems provide only information on government contracts online; transactional systems also enable fi rms to submit bids online. The bars stand for the average eff ects in each country. Based on a sample of 25,961 fi rms (panel a) and 4,343 fi rms (panel b), from the World Bank Enterprise Survey. 164 WORLD DEVELOPMENT REPORT 2016

The varied impact of e-fi ling, e-procurement, and provide feedback, and the service provider must have one-stop centers reveals a broader, sobering trend: the incentive and capacity to respond and resolve the Many government digital technology projects are complaint. Citizens are more likely to complain if they unsuccessful. They are either abandoned before use the service frequently, if it is a private good (so completion or underused, wasting substantial public there are no problems of free-riding), and if it is easy funds (box 3.5). A widely accepted explanation for to identify the service failure. Governments are more this poor record is the big gap between the techno- likely to respond if the information provided by citi- logical capability and the institutional capacity of the zens on the service failure is specifi c and actionable, government, as measured by processes, objectives, if it is clear who in government needs to resolve the values, staffi ng numbers, skills, and management complaint, and if the complaint also integrates with systems and structures.27 document management systems within the agency so that both the complaint and the response can be Receiving user feedback prioritized and tracked. Government responsiveness One potential catalyst for service improvements is can further incentivize citizens to complain, creating user feedback on service quality. More than a decade a virtuous of feedback and responsiveness. For ago, the private sector discovered customer relation- example, users of FixMyStreet—the platform in the ship management—the use of digital technologies United Kingdom that citizens use to report local road to integrate all aspects of a fi rm’s interaction with a problems, which are then routed to the relevant local customer to improve personalized communication authorities—are 54 percent more likely to submit a and provide real-time information so that customers second report if their fi rst complaint was resolved.28 can track the status of their service requests. Gov- Household water supply and electricity are exam- ernments have only recently discovered this man- ples of private goods that citizens use daily and have agement approach, with most innovation in cities an incentive to monitor. The responsibility for deliver- in the developed countries. Mobile phone apps like ing these services lies clearly with a particular utility, SeeClickFix and FixMyStreet, active in many munici- and it is easy for citizens to know when the service is palities in the United States and the United Kingdom, failing. MajiVoice is a complaint mechanism in the and internet call centers allow users to report service water and sewerage company in Nairobi, and there is issues and enable governments to report back to a similar one in use in EDE Este, the electricity supply citizens on fi x requests through integrated back-end company for the eastern Dominican Republic. Given systems. They are now standard in Barcelona, Boston, the limited internet use in the country, consumers Chicago, London, New York, Seoul, and Singapore. initiate complaints mostly in person at the service Developing countries are following suit. Given the desk in the utility or through phone calls, generating a considerable ineffi ciencies in public sector delivery ticket number as a tracking tool for both the customer systems, the gains from embedding digitally enabled and the utility. The customer receives an SMS message citizen feedback in management are potentially from the utility when the complaint is entered and greater than in developed countries. These feedback another when the complaint is resolved. This tracking channels come in many varieties. They can be for a is possible because each complaint is registered on a limited set of services tied to particular agencies or dashboard that enables management to categorize for multiple services and issues, usually housed in a the complaint, delegate the issue to a specifi c staff center-of-government entity, either local or national. member on duty, and track resolution. The dashboard They can be complaint portals that invite citizens enables the utility regulator to track performance, and to report service issues, or government-initiated customers of EDE Este are also randomly contacted (“proactive”) feedback mechanisms through which after the complaint is resolved to gauge their satisfac- government offi cials contact citizens to inquire about tion with the service. their service experience. Complaint portals and call Initial results are impressive. Before MajiVoice centers for citizen feedback are spreading rapidly was initiated, the Nairobi water company received on across developing countries, particularly in cities— average 400 complaints a month. Since 2013, when Buenos Aires, Muscat, Rio de Janeiro, and Ulaanbaa- MajiVoice was launched, the utility has been getting tar, to name a few. Proactive feedback mechanisms about 3,000 complaints a month, resolution rates are more recent and thus limited in number—those in climbed from 46 percent to 94 percent, and time to Albania and Pakistan are examples. resolution dropped by 90 percent (fi gure 3.11).29 Since Two conditions are required for user feedback its implementation in 2011, the feedback channel in to have impact: Citizens must have an incentive to EDE Este has also recorded rising resolution rates for DELIVERING SERVICES 165

Box 3.5 The high failure rate of e-government projects

Despite spending more than $600 billion on IT over the past decade, the federal government has achieved little of the productivity improvements that private industry has realized from IT. Too often, federal IT projects run over budget, run behind schedule, or fail to deliver promised functionality.

—U.S. Offi ce of Management and Budget 2010

Many public sector digital technology projects fail. Figure B3.5.2 Performance of World Although the evidence is limited, various estimates from Bank–funded ICT projects surveys of government offi cials, audit reports, and coun- try cases suggest that about 30 percent of these projects 100 are total failures, with the project abandoned before completion. Another 50 to 60 percent are partial failures, 80 with signifi cant budget and time overruns and only a limited number of the project objectives achieved. Fewer than 60 20 percent are successes.a In some cases, even if e-govern-

ment projects are successfully implemented they may actually Percent worsen outcomes as, without proper regulatory safeguards in 40 place, automation makes it easier to perpetrate fraud and cor- rupt practices, and to erase records or avoid capturing them 20 altogether, thereby eroding transparency mechanisms.b Risky for the private sector, these projects do worse in 0 governments. One survey of U.S. information technology ICT projects All projects projects found success rates of 59 percent in retail, 27 Highly satisfactory percent in manufacturing, and 18 percent in government.c Satisfactory Moderately satisfactory Larger projects are more likely to fail than smaller ones, Moderately unsatisfactory with success rates of only 13 percent on average (fi gure Unsatisfactory B3.5.1) and a skewed distribution, with a small proportion Highly unsatisfactory of projects accounting for a major chunk of budget and Sources: WDR 2016 team, based on Standish Group 2014 and World Bank (2015a) Digital Governance Projects Database. Data at http://bit.do Figure B3.5.1 Success rate of large /WDR2016-FigB3_5_2. public sector ICT projects Note: Data are for 1995 to 2015. ICT = information and communication technology. time overruns. One comprehensive review of more than 13% 1,400 public sector information and communication tech- 29% nology (ICT) projects found that a sixth went over budget by 200 percent, posing “systemic” fi scal risks.d World Bank–funded projects have not fared much bet- ter: 27 percent of the roughly 530 information technology 58% projects from 1995 to 2015 were self-evaluated by the World Bank as moderately unsatisfactory or worse, in line with all projects (fi gure B3.5.2). But far fewer were rated “satisfactory” or above, with 35 percent for ICT projects and Failed Partially failed Succeeded 56 percent for all projects. Sources: WDR 2016 team, based on Standish Group 2014 and World Bank e-government scholars provide numerous explanations (2015a) Digital Governance Projects Database. Data at http://bit.do for these stark numbers. A widely quoted study blames a /WDR2016-FigB3_5_1. large gap between the regulatory, political, management, Note: “Large” is above US$6 million. Data are for 2010 to 2014. ICT = information and communication technology. process, and skill in government and the ambitions

(Box continues next page) 166 WORLD DEVELOPMENT REPORT 2016

Box 3.5 The high failure rate of e-government projects (continued)

of e-government projects.e Another points to the “danger- vague to allow fl exibility during implementation. Ministers ous enthusiasms” of technological infatuation and faddism and senior offi cials, by contrast, are explicitly forbidden by in large IT projects.f Many studies also point to the rigidi- government rules to exercise this level of discretion and ties in government procurement rules, IT vendors’ lack of instead must purchase on the basis of detailed specifi ca- understanding of government processes, and a failure to tions. Surveys of technology professionals and executives understand the country context.g In the private sector, exec- in both the public and private sectors underline user utives can choose vendors based on personal experience involvement, leadership support, and a clear statement of and jointly determined guidelines that are deliberately kept project requirements as critical factors for success.h

a. These estimates are based on the various reports issued by the Standish Group and Heeks 2008. b. Lemieux 2015. c. Standish Group 2014. d. Budzier and Flyvbjerg 2012. e. Heeks 2008. f. Gauld and Goldfi nch 2006. g. Dunleavy and Carrera 2013; Fountain 2001; Heeks 2006; Bhatnagar 2009. h. Standish Group 2014.

Figure 3.11 More complaints were resolved more salary increases) based on worker performance. In quickly in the Nairobi water utility after the the Nairobi utility, data from monthly management introduction of digital customer feedback reports are also used as a basis for performance incen- tives for staff. 120 60,000 Property registrations, welfare payments, and licensing services are also private goods that are easy 100 50,000 to monitor, and citizens have an incentive to give feedback on them. With multiple agencies usually 80 40,000 delivering these services, citizen-initiated feedback is more likely to have impact if the services are consol- 60 30,000 idated in one-stop service centers that make it easier for citizens to provide feedback and for governments

Days to resolve 40 20,000 MajiVoice to integrate the feedback in administrative systems introduced and to monitor the responsiveness of agencies. One- 20 10,000

Number of cumulative complaints stop centers in Azerbaijan, in Brazil’s Minas Gerais state, and in Moldova show how automating the 0 0 June December June December service delivery chain with citizen feedback can drive 2013 2013 2014 2014 service improvements.31 Resolution time By contrast, it is much more diffi cult for citizens Complaints submitted to determine the quality of education or curative Complaints resolved health care and to attribute poor outcomes to the pre- Source: World Bank 2015d. Data at http://bit.do/WDR2016-Fig3_11. cise cause—whether, for example, their child’s poor Note: Days to resolve represents the three-month trailing average; numbers of complaints are cumulative. learning (to the extent that it is assessed through standardized tests) or their poor health is due to poor reported problems, improvements in customer satis- service providers, their own negligence, or environ- faction, and a reduction in reported corruption.30 This mental factors. Beyond issues like provider absentee- improved tracking is also triggering improvements in ism, complaints are less likely to be actionable. And human resource management. In EDE Este, the feed- citizens have fewer incentives to complain about the back is systematically used to inform sanctions (such weak provision of public goods like roads and munic- as administrative procedures) and rewards (such as ipal services because of free-riding. DELIVERING SERVICES 167

Figure 3.12 Citizens using Indonesia’s national feedback portal (LAPOR) have few actionable complaints, and these are mostly for private goods, 2015 a. Percentage of actionable complaints b. Distribution of actionable complaints by service

Other Actionable

Social welfare 10%

Fuel subsidies

Education

Transportation Not actionable 90% Public works

Roads

0 5 10 15 20 25 30 35 Percent

Source: WDR 2016 team, based on 2015 LAPOR data. Data at http://bit.do/WDR2016-Fig3_12.

These problems are evident in general citizen on acquiring actionable information. In Pakistan, feedback portals recently introduced in Botswana, under the Punjab Citizen Feedback Model, service Indonesia, and the Philippines. The government of providers record citizens’ mobile numbers. A gov- Indonesia’s complaint-handling system (LAPOR) ernment call center sends SMS messages and voice allows citizens to submit reports on many issues, calls to public service users to make targeted inqui- ranging from delays in welfare payments to damaged ries about satisfaction with 16 services, ranging from roads. It has received about 800 complaints a day property registration to primary health care and on average from all over Indonesia since the portal’s potential incidents of corruption. The data are logged launch in 2012. Of these complaints, only 10 percent and tracked on dashboards.35 This system has been are specifi c and actionable (fi gure 3.12, panel a) and deployed at a very large scale, with more than 7 mil- are forwarded to the relevant agency for resolution. lion citizens contacted since 2012—250,000 a month Of these actionable complaints, 28 percent and 18 per- on average—and about 1 million having provided cent, respectively, are for private benefi ts like welfare feedback from across the province.36 But the impact of payments or fuel subsidies; less than 5 percent are for the feedback on service delivery is unclear. Many offi - public goods like roads and public works (fi gure 3.12, cials have tried to subvert the initiative, as refl ected panel b).32 The complaint portals of the governments in the high and steady proportion of invalid phone of Botswana and the Philippines have similarly pro- numbers recorded by service providers—40 percent vided limited actionable information to the govern- for citizens registering property, a service particularly ment, particularly on public goods.33 prone to petty corruption. The government has taken Citizen feedback portals are also likely to be dispro- more than 6,000 administrative actions against offi - portionately used by better-off, more educated, and cials based on the feedback. But given the protections more digitally savvy individuals, potentially biasing afforded staff under civil service rules, the actions government responses. For example, most users of have been mostly warnings and formal apologies FixMyStreet in the United Kingdom are male, older, from the concerned offi cial to the citizen, and only a more educated, and less likely to be from an ethnic handful have been suspensions or dismissals. minority than the general population.34 Similarly, the Governments in high-income countries are users of LAPOR are disproportionately from Jakarta, moving beyond user feedback to engage citizens in the largest city and the capital, than from the poorer collaborative policy making. The northern European and more remote regions in eastern Indonesia. countries are on the vanguard in these moves to more Government-initiated feedback can overcome direct democracy, as exemplifi ed by Iceland’s crowd- these biases in citizen reporting and be more focused sourced constitution, Estonia’s e-legislation portal, 168 WORLD DEVELOPMENT REPORT 2016

and Reykjavik’s digital participatory budgeting.37 absenteeism among government providers, who are But even in these digitally and socioeconomically diffi cult to dismiss and discipline.40 Technology must advanced countries, e-participatory policy making be accompanied by good management. In India, the has engaged only a small, nonrepresentative subset monitoring of health workers did improve attendance, of the population, has elicited limited government but many local governments did not use the data to responsiveness, and is diffi cult to sustain. In Estonia, sanction absent workers because of cumbersome a global leader in e-services and e-voting, with high civil service rules or for political reasons. In Uganda, citizen use across all socioeconomic groups, only absenteeism was reduced only by combining teacher 8 percent of people polled had heard of http://www incentive pay with monitoring technology. In Paki- .osale.ee, the government’s e-legislation and policy stan, smartphone-based monitoring of district health making portal. supervisors in rural clinics doubled inspections of health facilities and reduced medical worker absen- Improving service provider management teeism, but the results were highly contingent on the severity of patronage politics in a locality. In Haiti, the experiment failed because government teachers had Technology can improve systems that not been paid for months and resented the additional are already working . . . but it doesn’t fi x unfairness imposed on them by digital monitoring. systems that are broken. There is no knowledge In Niger, by contrast, the experiment was conducted management without management. with contract teachers hired by a nongovernmental —Kentaro Toyama, University of Michigan organization, and mobile phone–based monitoring School of Information by itself was suffi cient to improve teacher effort and motivation as well as student learning outcomes. User feedback is effective when providers and man- Reducing absenteeism, while necessary, is not suf- agers have the incentive and the ability to respond to fi cient to ensure that doctors, teachers, and managers, citizens, which depends on strong relationships of once they show up to work, actually are motivated to accountability between policy makers and providers. perform well and to serve the public. Recent innova- By better monitoring of workers and facilities and tions in the private sector have extended these teacher through better organization and performance man- monitoring technologies to the actual practice of agement in bureaucracies, digital technologies can teaching, through standard lesson plans and system- strengthen accountability within the government by atic monitoring of teacher activities by management. reducing information asymmetries between policy The initial results are promising, again conditional makers and service providers. on the strong relationships of accountability that can The absenteeism of teachers, doctors, nurses, and exist in private schools (box 3.6). But applying these agricultural extension workers is pervasive in Africa lessons to a government bureaucracy will be chal- and South Asia.38 Estimates suggest that in India lenging. The typical developing country government roughly one-quarter of government teachers and over bureaucracy is unable to recruit the best and brightest one-third of government doctors in primary health workers because of low pay and excessive infl uence centers are absent without a legitimate reason on any by politicians. It promotes staff on the basis of senior- given day, with similarly alarming numbers in several ity and not performance—in part because measuring African countries.39 Physical monitoring of providers performance is diffi cult and seniority is the only cred- is costly, particularly in rural areas, and the monitors ible standard—and is unable to motivate workers to are just as likely to shirk as the providers or to collude work hard and serve the public. For many providers, with providers. Using mobile phones to record atten- the reason for joining the public sector is job security, dance—either with the photographs or thumbprints not promoting student learning or . of the providers as evidence—and transmit data to a These weaknesses are evident from perception central database to generate monitoring reports can surveys of government workers in Indonesia and the be a cost-effective solution to this seemingly intrac- Philippines. Many staff believe that the best univer- table problem. sity graduates prefer working in the private sector, Evidence from impact evaluations in Haiti, India, that their coworkers are often not productive, that Niger, Pakistan, and Uganda shows that digital mon- promotions are based on politics rather than merit, itoring can reduce absenteeism in general, but that and that their coworkers in other government agen- technology on its own is not suffi cient for curbing cies cannot be trusted (fi gure 3.13). Anecdotal evidence DELIVERING SERVICES 169

Box 3.6 Digitally enabled teacher management in private schools

Private schools targeted to the world’s poorest have system, expense-processing system, standardized assess- become a new model for education in developing countries ments, and evaluation tools. Even paying school fees is with low-quality public education. Bridge International standardized and technology-enabled; tuition payments Academies (BIA) is an innovative for-profi t private model are made electronically with M-Pesa or through Equity operating in underserved communities in Kenya and Bank Kenya. The standardized instruction, school man- Uganda that leverages standardization and technology to agement, and assessment ensure that each student in a educate more than 118,000 children. Bridge Academy receives the same education, regardless The BIA business and educational model is driven by of location or instructor. a standardized set of curricular and managerial materials Initial evaluations show that this educational style to keep classes consistent, students on task, and schools may be eff ective. In 2013, a self-evaluation found that running smoothly. The core of these materials is an BIA students scored 0.32 standard deviations higher than internet-enabled tablet computer, which downloads more peer students in public schools in reading fl uency and than 3,400 lesson plans and teacher scripts at the beginning comprehension, translating into an additional 16 words per of each term. These plans include step-by-step instructions minute and 252 additional learning days. In math skills, BIA on how to prepare and present each lesson, homework students scored 0.51 standard deviations higher, equivalent assignments, and tests. BIA management tracks class- to an additional 288 learning days. These fi ndings need to room progress—such as the speed of lessons and teacher be viewed cautiously, however, as the diff erences in out- attendance—with the tablet through automatic syncing. If comes may be entirely explained by diff erences in students a teacher falls behind in lessons, BIA managers in Nairobi between BIA and public schools. and BIA’s U.S. headquarters know immediately and can This type of private education is not unequivocally follow up with the teacher hundreds or thousands of miles embraced. The scripted lesson plans are particularly con- away. In addition to teacher monitoring, student progress troversial, with questions about the eff ectiveness of a script is tracked by teacher-inputted student test scores and to create a dynamic learning environment for children to progress measures. This automation gives BIA headquarters develop higher-order thinking skills. In addition, Bridge instant access to school data and helps BIA manage their Academies rely on large class sizes and teachers with only academies. a secondary education to keep costs low. Low-cost private In addition to classroom management, standardization education is also criticized because of the potential to and technology are leveraged to improve school adminis- undermine public education systems. Countries that rely tration. Each “academy manager”—akin to a head teacher on private education may not invest as heavily in their or principal—receives a smartphone to use for communi- public systems, and private schools often attract the best cation and monitoring, and as a modem for the tablets. students, further diminishing the public education system. Academy managers follow standardized instructions on Even so, in contexts where public schools are failing, a stan- managing fi nances, personnel, and student and parent dardized private sector model may be a viable option for relationships, as well as on how to use a central payroll improving learning outcomes.

Sources: WDR 2016 team, based on Bridge Academies 2013; Heyneman and Stern 2014; Rangan and Lee 2010; Ross 2014. suggests that these problems are even more severe in ers reported that they had never had a meeting with low-income countries. Studies of teacher and health the head teacher to discuss their performance, and worker management in South Asia and Africa reveal two-thirds of the head teachers had themselves been a fundamentally broken system, with pervasive prob- absent from school at least once in the last fi ve days.42 lems of politicization, weak school management, low The low productivity of some government bureau- provider motivation, and shirking.41 A recent survey of cracies refl ects fundamental structural constraints public schools in Tanzania revealed that 67 percent of that cannot easily be substituted for by technology. the teachers who were present in the school did not Where digital technologies can help is in monitoring show up in classrooms to teach. Nor did school princi- goals and performance targets for organizational units pals hold them accountable: Three-quarters of teach- and workers. For the private sector, fi rms that adopt 170 WORLD DEVELOPMENT REPORT 2016

Figure 3.13 Government offi cials in Indonesia and the Philippines have generally low opinions of human resource management practices

a. Indonesia b. Philippines

Your ministry easily recruits The best and brightest join high-quality staff the private sector

You recommend jobs in your Your coworkers are not department to friends and family productive at work

Your coworkers are not productive at work Promotions are based on merit Promotions are based on politics rather than merit Underperformers are routinely disciplined Underperformers are routinely disciplined

Most people in other Most people in other departments can be trusted departments can be trusted

0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents Disagree or strongly disagree Neither Agree or strongly agree

Sources: World Bank surveys of government offi cials, 2011 and 2013. Data at http://bit.do/WDR2016-Fig3_13. Note: The survey in Indonesia was conducted in 2011 and covered about 3,000 government offi cials; the survey in the Philippines was done in 2013 and covered 2,500 offi cials.

more structured performance management practices ment and level of trust in an organization—which can are more profi table, and digital technologies com- take years to build. There is little evidence, even in the plement performance monitoring and performance OECD (Organisation for Economic Co-operation and incentives.43 Goal-setting and performance incentives Development) countries, that digital technologies are diffi cult—but not impossible—in government have made any fundamental differences in the way bureaucracies. The multiple demands on a public orga- government bureaucracies are managed, particularly nization, and the multiple interests it needs to serve, for the better integration of policy making and service make it diffi cult to defi ne goals, and performance delivery across government. Even in New Zealand, incentives can trigger a host of perverse incentives. known for its highly competent and innovative public The key variable is the extent to which the tasks and administration, two-thirds of public offi cials surveyed outputs of agencies can be routinized and monitored.44 were skeptical that digital technologies would induce As MajiVoice and the EDE Este feedback systems greater collaboration and integration across govern- show, eliciting citizen feedback on service quality, ment departments because the notion of a “joined-up tracking complaint resolution, producing audit trails government” confl icted with the annual agency-based of worker effort, and offering performance incentives budget appropriation process.45 for staff can be a powerful combination for trans- In sum, these digitally enabled management forming government bureaucracies. This combina- improvements are isolated examples. Either they are tion is also being used in one-stop service centers in pilots and experiments limited to a subset of activities, many countries. such as addressing absenteeism, that form the basic Digitally enabled performance monitoring is much minimum required to improve teaching or health care, more diffi cult for services that are highly discretion- or they are limited to a few locales, sectors, or agencies ary and hard to monitor, such as teaching and curative and not taken to scale in the government as a whole. health care, and for policy and regulatory functions. Developing country government bureaucracies now Assessing performance is necessarily more subjective have a digital veneer over a largely unchanged struc- and therefore conditional on the quality of manage- ture, culture, and performance orientation. DELIVERING SERVICES 171

Digital technologies too Emerging evidence suggests that digital technolo- gies have made elections freer and fairer by improv- often fail to empower ing voter registration and reducing errors in voting, citizens and by better monitoring them to curb electoral fraud and violence. These technologies have also helped cit- As discussed, digital technologies can be effective even izens vote out corrupt politicians, if the information in weak government bureaucracies in some areas— comes from a credible source. But signifi cant barriers notably through digital identifi cation, streamlining to more informed voting remain, and digital tech- routine tasks, and citizen feedback for certain services nologies, by giving elites new ways of manipulating that citizens have the incentive and capacity to moni- information to their advantage in election campaigns, tor. But in general, to have an impact, e-government can also disempower the poor. These disparities can requires effective leadership by politicians and policy be countered if traditional media bridge the digital makers—to make the necessary changes to govern- divide, but such a role may be limited to providing ment rules and management practices, to overcome information that is more salient and newsworthy, resistance from vested interests, and to respond to cit- such as corruption scandals, rather than service deliv- izen feedback on service quality. Strengthening gov- ery failures. Digital technologies, particularly social ernment capability requires a willing government. The media, have galvanized citizen protests, but except question, then, is whether digital technologies can where governments are willing and able, they have encourage good leadership by empowering citizens to not sustained collective action and citizen voice to hold policy makers and providers accountable. improve service delivery (table 3.2). Mechanisms for empowering citizens include Freer and fairer elections • Free and fair elections in government systems in Is the growth of digital technologies spreading demo- which leaders are chosen in elections—ensuring cratic ideals around the world, as many believe?46 that all legitimate votes can be cast and counted so Democracy has indeed spread across the world, but that the poor in particular are not excluded from so have election irregularities.47 As the number of participating developing country democracies more than doubled • More informed voting—informing voters to increase from 1990 to 2012, the proportion of elections that electoral participation and reducing information were “free and fair” halved from almost 80 percent asymmetries so that votes can be more accurately to under 40 percent over the same period (fi gure 3.14, based on the performance of politicians panel a). In a free election, the electoral rules and their • Citizen voice and collective action—empowering citizens implementation leading up to an election enable all to individually and collectively mobilize to pressure adult citizens to be registered, to exercise their right policy makers and providers to improve services. to vote, and to join political parties and to campaign

Table 3.2 The impact of digital technology on citizen empowerment: A scorecard

Impact of Main problem Channel technology to address Do digital technologies solve the problem?

Free and fair Lack of information; • Yes, monitoring reduces errors and fraud elections H high transaction in voting costs

More informed Information • Yes, for blatant abuses of office; no, for less voting asymmetries newsworthy public service failures M • Increase ability of elites to manipulate information

Greater citizen Collective action • Effective only when governments are already voice failures willing to listen to citizens L • Must be complemented by offline mobilization by civil society groups

Source: WDR 2016 team. Note: Channels are arranged by degree of technology impact. L = low; M = medium; H = high. 172 WORLD DEVELOPMENT REPORT 2016

Figure 3.14 Democracy has spread, but so have election irregularities—digital technologies can help make elections freer and fairer

a. Democracies and elections, 1985–2012 b. Internet use and perceptions of electoral integrity, 2015

80 10 CRI

TON 70 GRD BRA MUS 8 TUN STP SLV GEO 60 BWA ZAF MNG IND COL ROU BTN NIPDNBOLUKPRAN SRB SLB NAMRWAPRY ECU PAK PHL MALDBA TUR 50 MDV BBIGHR 6 CMR IRN JOR GIMNLI FJI HUN MWI IRQ MKD 40 THEGAY SWZ HND MRT KEN MNE integrity score MDG 4 ARM MOZ DKZHAM Number and percent 30 ZWE VEN TGO MYS Perceptions of electoral BGD TJK AZE 20 DJI TKMAFG 2 10 1985 1990 1995 2000 2005 2010 2015 0 0.2 0.4 0.6 0.8

Percentage of elections that are free and fair Digital Adoption Index Number of electoral democracies

Sources: WDR 2016 team, based on Polity IV 2015; Bishop and Hoeffl er 2014; Perceptions of Electoral Integrity 2015. Data at http://bit.do/WDR2016-Fig3_14. Note: The fi gures are limited to low-income, lower-middle-income, and upper-middle-income countries, for which election data are available. Election percentages are the fi ve-year trailing average.

freely. In a fair election, all voters and all votes are More informed voting treated equally on the day of the election.48 Electoral Improving the mechanical aspects of voting is a min- irregularities can affect poor voters disproportion- imum requirement for electoral accountability. The ately. They may not be able to cast their votes because more fundamental problem is that citizens are often they lack proper identifying documents, or their votes uninterested and unengaged in politics, face signifi - may not be counted because of error or manipulation. cant informational barriers in assessing politician Digital technologies are helping make elections performance, or base their votes on tribal, ethnic, or freer and fairer. Across nations, perceptions of the clientelist factors and not the ability and performance integrity of elections in developing countries are posi- of politicians.53 This lack of engagement is under- tively correlated with internet use (fi gure 3.14, panel b). standable. Acquiring political information is costly, Rigorous studies reinforce this cross-national correla- the sources of the information, like political parties, tion and demonstrate the powerful impact of digital are often not credible, and the impact of an individual technology in increasing electoral participation, par- citizen’s more informed vote is negligible to the out- ticularly for the poor. In Brazil, pre-internet electronic come of an election. The question, then, is whether voting machines signifi cantly reduced voting errors digital technologies change this calculus to result in in the 1990s, leading to the enfranchisement a more engaged and informed citizenry—and more of poor and less educated citizens, which in turn led to effective elections. more pro-poor spending.49 In Afghanistan, the moni- There is no shortage of popular opinion that the toring of votes cast through cellphone photographs of internet—by exponentially increasing transparency vote totals in polling stations reduced electoral fraud and forging connections among voters, particularly in the vote aggregation process.50 In Mozambique, through social media—will strengthen accountability SMS messages allowing citizens to report electoral through greater voter engagement and participation.54 irregularities increased voter turnout by 5 percentage These sentiments echo those of earlier generations points.51 And although not evaluated, the use of crowd- about television, which was also viewed as nothing sourcing by Ushahidi and its successor Uchaguzi, short of revolutionary for enhancing accountability. combined with civil society monitoring, advocacy, But television’s impact has turned out to be much and partnerships with government, has helped curtail more mixed. In Sweden, the rollout of commercial election violence in Kenya (box 3.7).52 television indeed improved citizen knowledge of DELIVERING SERVICES 173

Box 3.7 Improving the integrity of elections through crowdsourcing and collaboration

Improving electoral integrity requires eff ective media, citizen crowdsourcing with analysis by specialized civil nongovernmental organizations, and partnerships between society organizations with experience monitoring elections technologists, advocacy groups, and government watch- to provide actionable information to local and national dogs to gather, analyze, disseminate, and act upon rele- government offi cials, in particular the Interim Independent vant information. Consider the contrasting experiences of Electoral Commission. This partnership was a success, with Ushahidi and Uchaguzi, two digital election-monitoring the government responding to a majority of the reports. initiatives in Kenya. Ushahidi, one of the best-known digital Ushahidi’s implicit belief was that the platform would platforms in the world, was launched in the aftermath of by itself be the agent of change. Uchaguzi—or “Ushahidi Kenya’s tumultuous 2007 election and used “citizen report- 2.0”—recognizes that the platform can work only if it ers” to crowdsource and map incidents of violence after amplifi es existing institutions. The main lesson is that the election. While the platform galvanized considerable governments may be dysfunctional, but can only rarely interest initially—with more 45,000 users in its fi rst few be bypassed. As one of the founders of the initiative months—many of the reports were not actionable, and there commented after the local police—infamous in Kenya for was limited response from the authorities. their venality and brutality—successfully responded to an Learning from this experience, Ushahidi and a group online citizen report of a machete-armed mob gathering of journalists and civil society organizations launched outside a local polling station, “We can’t compel organi- its successor, Uchaguzi, to monitor Kenya’s 2010 consti- zations to act. We can support institutions, but we can’t tutional referendum. But this time, the eff ort combined replace them.”

Sources: WDR 2016 team, based on interviews with Ushahidi management; Fung, Russon Gilman, and Shkabatur 2013; Wrong 2013.

politics and government policy and increased voter news media landscape, as internet political opinion is turnout.55 In the United States, however, it reduced dominated by a small group of bloggers and outlets voter turnout and knowledge of policy issues by that receive the bulk of search engine hits.60 Finally, crowding out local radio and newspapers, previously the internet can give voice to extreme opinions by the main source of political information.56 enabling highly partisan media outlets to get suf- Digital technologies can certainly provide a vari- fi cient readership and encouraging clustering of ety of tools to improve the informational content of extreme views, particularly among the more polit- voting. For example, Voting Advice Applications are ically organized and aware, potentially increasing websites that enable citizens in several European polarization in politics.61 countries to compare the platforms of competing Given that the internet is the main source of news political parties with their own preferences on salient for less than 2 percent of the population in South Asia issues; these increased voter turnout among youth and Sub-Saharan Africa,62 the informational effects of in the Netherlands.57 But the internet also offers a digital technologies are likely to be mediated through variety of other entertainment options that compete traditional media in those regions, as well as else- for people’s time and can disengage them from poli- where in the developing world. A growing number tics. In Germany and Italy, the internet reduced voter of advocacy organizations in developing countries turnout by crowding out television, a greater source partner with media to provide voters with easy-to- of political information.58 understand information about politicians. For exam- The internet also provides political parties, blog- ple, Mumbai Votes tracks legislators from Mumbai at gers, and opinion leaders with resources to shape all levels of government in India, reporting on their public opinion. For example, a simple “I Voted” legislative activity and any court cases fi led against message on individuals’ Facebook pages increased them. The Fair Play Alliance of the Slovak Republic voter turnout by more than 300,000 in the 2010 U.S. tracks campaign fi nance and procurement contracts midterm elections.59 The “winner take all” feature of to uncover corrupt relationships between businesses the internet economy is also replicated in the online and politicians.63 174 WORLD DEVELOPMENT REPORT 2016

For this digitally enabled political information to countries by lowering the cost of voting, since votes have impact, it must be easy to understand, be clearly can be cast from the convenience of one’s home. In attributable to a particular politician, and be credi- Estonia, where about one-third of voting is online, ble. A good example is the uncovering of corruption the share of internet voters has increased by more by public offi cials if the information comes from a than 4 percentage points in almost every election trusted source, such as an independent anticorrup- since 2005, contributing to an overall increase in tion commission or reliable media outlet. In 2003, voter turnout (fi gure 3.15, panel a). Studies show no the federal government of Brazil launched a national differences in the demographic characteristics of anticorruption program based on random audits of internet versus offl ine voters.71 In Brazil, by contrast, municipal governments by the national audit author- internet voting on municipal budget proposals in the ity (the municipalities to be audited in a given month state of Rio Grande do Sul increased voter turnout were chosen by lottery), with results publicly dissem- by 8 percent (and these were new voters), but the inated over radio, television, and print media. Voters online voters were more likely to be male, university punished incumbents who were found to be corrupt educated, and higher-income (fi gure 3.15, panel b).72 by not reelecting them. The effects were stronger in These changes in the demographic profi le of voters municipalities with local radio stations, pointing to have not yet changed policy preferences, perhaps the role of local media in informing voters.64 These because the offl ine participatory mechanisms to fi ndings have been replicated in studies of municipal determine the budget proposals that are then voted audits in Mexico, again pointing to the importance on limit the range of options subject to the vote. But of a credible source for information on corruption they show how the internet can be biased against (a national audit by an independent and competent certain groups. institution) and local television and radio to dissem- Overall, then, the evidence suggests that digital inate that information for electoral accountability.65 technologies have helped curtail blatant abuses of By contrast, provision of information about fail- political offi ce, conditional on supportive institutions ures in service delivery such as poor education or of accountability (such as independent audit or anti- health care, or weak performance of legislatures— corruption bodies) and free media (to generate and which is more complex and diffi cult to attribute help disseminate the relevant information to voters). to particular politicians—has been less effective.66 Service delivery failures have not been salient enough Information campaigns over local radio on public to command the attention of even digitally enabled health and primary education in rural Benin had no media to infl uence voting decisions by the poor. effect on the incentives of politicians to provide better health and education services or on provider respon- Citizen voice and collective action siveness.67 Informing Indian voters about the perfor- mance of their legislators in legislative activity had “Home run” cases in which a technological no effect on voting behavior; but informing them on intervention almost by itself produces dramatic spending by parliamentarians on physical infrastruc- increases in accountability . . . are exceedingly ture from funds allotted to them to spend on their rare . . . The greatest opportunities [for technology constituencies did affect voting behavior, since this are in] amplifying NGO and governmental 68 information was more salient to voters. Similarly, strategies of accountability. providing information on legislator performance to voters in Uganda had no effect on parliamentarians’ —Archon Fung, Kennedy School of Government, reelection rates.69 The experience of Open Data Ini- and colleagues tiatives is also quite sobering. The mainstream media in Kenya, the Philippines, and Uganda have not been One of the more widely held beliefs about the trans- deeply engaged with the Open Data Initiatives, either formative power of digital technologies, particularly in using data as a source of evidence in their report- social media, is its ability to catalyze citizen voice and ing or in reporting on efforts surrounding open data. collective action to hold governments accountable. Journalists interviewed said there was no “scoop” in The internet promotes transparency by increasing the open data story.70 the richness and customizability of the information Digital technologies can also reinforce socioeco- for citizens. Less obvious—and unique to these tech- nomic disparities in voter participation and knowl- nologies—is that they also enable citizens to generate edge, particularly in developing countries. Internet their own information (user-generated content), voting has increased participation in some European which can be aggregated at low cost to provide col- DELIVERING SERVICES 175

Figure 3.15 Internet voting can increase voter participation but can be biased toward more privileged groups

a. Increase in internet voting in various b. Profile of online and offline voters in a participatory , 2005–15 budgeting vote in Rio Grande do Sul, Brazil, 2011–12 60 European Parliament National 30 50

National 40

Municipal 20 30

European Percent Parliament 20 Municipal 10 participating voters 10

e-voters as a percentage of 0

<750 National –6,000 Female Higher 0–1,500 >6,000 75 ElementarySecondary Municipal 1,500 0

2005 2010 2015 Gender Education level Income level (BRL/month) Source: WDR team, based on Vassil 2015, for the WDR 2016. Data at http://bit.do Offline voters Online voters /WDR2016-Fig3_15a. Source: WDR team, based on Spada and others 2015. Data at http://bit.do /WDR2016-Fig3_15b. Note: BRL = Brazilian real. lective voice through crowdsourcing or community technology-enabled effects hold for collective action mapping. Social media can facilitate collective action around the more mundane service delivery failures. by generating peer pressure as individuals observe Identifying the precise government failure is much the behavior and actions of others, improving within- more diffi cult for services—these after all are hidden group monitoring, reducing free-riding, and enabling from the public eye in classrooms, rural health cen- groups to better coordinate their actions around ters, and in the substandard materials used to con- specifi c events such as protests.73 These technologies struct roads—just as is clearly attributing the failure also provide new platforms for citizens to engage itself to the actions of a particular individual or group with the government, lowering the costs to citizens of individuals. These failures only rarely become of providing information, and enabling policy makers salient in fl ashpoint events like scandals or egregious and service providers to seek information and track abuses of public authority, such as the excessive use the feedback loop. of force by the police or a patient dying because of As a growing empirical literature shows, cell- poor care. And they cannot be addressed by the sim- phones and the use of Twitter and Facebook aided ple “stroke of the pen” actions that protests tend to protests during the Arab Spring in the Arab Republic mobilize around, such as jailing a corrupt politician. of Egypt,74 widespread demonstrations on social Instead, they require sustained reforms to improve issues in Chile in 2011,75 antiwar demonstrations in government capability, which in turn requires sus- the United States,76 and citizen mobilization across tained citizen voice and collective action. Africa, with the effect more pronounced in nondemo- Given these diffi culties in monitoring service cratic political systems where traditional channels for failures, it is not surprising that the large academic articulating interest were limited.77 The same mecha- empirical literature on nondigital instruments of col- nisms also imply that digital technologies can trigger lective action—citizen and community report cards, harmful collective action, like ethnic confl ict.78 complaint mechanisms, community monitoring of Social media have the unique ability to make service providers—presents frustratingly mixed evi- scandals and highly emotive government abuses dence of impact.79 The main lesson is that the success of authority “go viral” and trigger parallel physical of citizen collective action around service delivery citizen mobilization. But it is less clear that similar failures depends on contextual factors, and each of 176 WORLD DEVELOPMENT REPORT 2016

the links in the causal chain between transparency, what is likely to work and under what conditions.82 citizen action, and state action can break down if these The focus here is on digital channels that are initiated factors are not aligned. How credible, salient, and by civil society organizations (CSOs) and donors to comprehensible is the information to citizens?80 How pressure governments, and that publicize the inputs willing and able are citizens to act on the information provided by citizens so as to expose the behavior of individually or collectively? How effective are inter- service providers to public scrutiny. This analysis mediary organizations—civil society, media, interest excludes government-initiated portals, which are groups—to organize citizens or to lobby governments more accurately characterized as user feedback on their behalf? And most important, do state actors mechanisms to improve service quality, and not as have the incentive and capacity for action, thereby citizen empowerment or accountability initiatives, “closing the loop”? Action and impact are more likely and which were discussed earlier. if the social accountability initiative is embedded in The analysis distinguishes 17 cases in 12 countries a citizen-state interface and synchronized with top- by whether the mechanism for expressing citizen down government accountability initiatives.81 preferences and views is individual or collective; Digital citizen voice initiatives for improving ser- whether the CSO or donor that led the initiative also vice delivery have multiplied rapidly, but no rigorous had explicit partnerships with the concerned gov- studies have evaluated their impact. An organizing ernment; and whether there was also parallel offl ine framework is needed to classify these examples and mobilization accompanying the digital voice channel draw the necessarily tentative conclusions about (table 3.3). Impact is measured in two ways: citizen

Table 3.3 Classifying the digital citizen engagement cases

CSO partners Impact Additional offline with Collective Government Case Location mobilization government feedback Citizen uptake response

Por Mi Barrio Uruguay L H

I Change My City India M H

Lungisa South Africa L H

Pressure Pan Brazil H M

Rappler Philippines H M

Change.org World H M

U-report Uganda H L

Huduma Kenya L L

Daraja Maji Matone Tanzania L L

FixMyStreet Georgia L L

Check My School Philippines L L

Barrios Digital Bolivia L L

e-Chautari Nepal L L

I Paid a Bribe India M L

Mejora Tu Escuela Mexico L L

Karnataka BVS India L L

Sauti Za Wananchi Tanzania L L Source: WDR 2016 team, based on Peixoto and Fox 2015, for the WDR 2016. Note: Examples are arranged by degree of government response. CSO = civil society organization; L = low; M = medium; H = high. DELIVERING SERVICES 177 uptake, given that uptake can be considered a neces- result: The Supreme Court declared these funds sary condition for government responsiveness; and unconstitutional.85 government action to resolve the service issue, which What stands out, though, is the high proportion of is the ultimate objective of the citizen voice initiative. cases that have both low citizen uptake and low gov- The criterion for selecting the cases is simple avail- ernment response. Most of the cases that lacked part- ability of information. The examples have received nerships between CSOs and governments were unsuc- media and donor attention, and some data are avail- cessful. Maji Matone (Tanzania), for example, received able on citizen use of the channel and on government only 53 SMS messages during its fi rst six months of response. operation, against an initial target of 3,000, and the Of the 17 cases, 3 had high impact in terms of mobile platform was abandoned. Other disappoint- government responsiveness, 3 had medium impact, ments are Huduma (Kenya),86 I Paid a Bribe (India), and and 11 had low impact. All the high-impact examples Check My School (the Philippines). Despite high expec- involve partnerships between CSOs and government, tations, they have not generated much citizen uptake suggesting that platforms that both channel citizen or government resolution of the reported problem—in voice and link them to the government’s internal work some cases, despite government collaboration. order systems to effi ciently address the complaints Uganda’s U-report, perhaps the highest profi le are more likely to succeed. For example, Por Mi Barrio initiative, is an SMS-based platform that runs weekly in Uruguay and I Change My City in India, two suc- polls with registered users (U-reporters) on issues cessful CSO platforms, are connected to existing gov- ranging from child marriage to access to education, ernment complaint systems. That enables urban resi- which are then both broadly disseminated and tar- dents to report problems using both the CSO and the geted to members of parliament for action. The plat- government channel, with the complaints and gov- form has had considerable citizen uptake, but mostly ernment responses displayed on a map, thus inform- by the more privileged groups. Of close to 300,000 ing and validating the citizen action and naming and U-reporters, nearly half have some university edu- shaming the nonperforming government units. But cation, and one-quarter are government employees, not all initiatives involving government-CSO partner- raising questions about whose voices are being pro- ships led to high responsiveness. For instance, I Paid a jected.87 This high uptake has not led to any noticeable Bribe (India) and Check My School (Philippines) both government action. Surveys of online political partic- had low impact. Seen together, these fi ndings seem to ipation in the United States also fi nd that better-off suggest that while collaboration with government is and more-educated citizens are more likely to partic- not a suffi cient condition for success, it may well be ipate in both online and offl ine political activities.88 a necessary one. In sum, the success of digital citizen voice is con- Another ingredient for success is effective offl ine ditional on willing and able policy makers, a collab- mobilization, particularly because citizen uptake of orative approach with government, and signifi cant the digital channels was low in most of the cases.83 offl ine activism by strong civil society organizations Online petitions on Change.org are more likely to be so that the collective voice of citizens can pressure successful when sponsored by an organization, and governments. However, digital platforms can also citizen campaigns through Pressure Pan (Brazil) are enhance unaccountable governments’ capacity for three times more likely to succeed when supported surveillance and control (box 3.8). by Pressure Pan staff.84 Lungisa, a platform for Cape Town residents to report service delivery problems, involves signifi cant follow-up by the CSO staff to The gap between ensure that the responsible government agency acts technology and institutions on the complaints. Rappler, a media and advocacy organization in the Philippines, combines media, The internet largely, but not entirely, reinforces technology, and the power of crowdsourcing to iden- rather than replaces preexisting relationships of tify and amplify governance issues, with traditional accountability between governments and citizens, citizen mobilization strategies using investigative and it complements rather than substitutes for exist- journalists and social mobilizers. Rappler orga- ing government capabilities. The main explanation nizes community protests, the most dramatic one for the varying impact of digital technologies lies being the Philippines’s fi rst protests organized with in the mismatch—or gap—between rapidly chang- social media against a corruption scandal involving ing technology and slowly changing political and congressional discretionary funds. The eventual administrative institutions. Political institutions can 178 WORLD DEVELOPMENT REPORT 2016

Box 3.8 Digital technologies can strengthen control

The internet presents a “dictator’s dilemma” for autocra- United Nations’ online service index compared to the ones cies. Restricting the internet can hurt economic develop- in the middle (fi gure B3.8.1). This pattern suggests that ment, but leaving it unfettered can threaten the govern- improvements in basic service delivery can be independent ment by increasing citizens’ access to political information of a country’s political system.a Autocracies are, however, and facilitating collective action. Autocratic governments becoming more adept at monitoring the internet to censor have responded to this dilemma in sophisticated ways, criticism entirely or circumvent protests, to spread pro- taking advantage of the tools that the internet itself off ers paganda, or to learn about public grievances to improve to improve services while strengthening control. government responsiveness while also intervening to pre- Autocracies have invested in e-government, particularly vent specifi c instances of collective action that challenge in e-services. The relationship between online services and the government’s power.b Digital technologies therefore government type is U-shaped, with both more democratic provide opportunities for autocracies to improve service and more autocratic governments scoring high on the delivery while maintaining control.

Figure B3.8.1 Autocratic governments have promoted e-government while censoring the internet

a. e-government provision, by government type b. Internet filtering, by government type

1.0 Substantial

0.8

Selective 0.6

0.4

Filtering level Suspected

Online service index 0.2

0 None –10 –5 0 5 10 Autocracy Democracy Autocracy Democracy Type of content filtered Political Social Conflict and security

Sources: WDR 2016 team, based on Polity IV 2015; UN 2014; Open Net Initiative 2013 data. Data at http://bit.do/WDR2016-FigB3_8_1. Note: The Polity IV project defi nes government type based on characteristics such as competitiveness and openness of executive recruitment, constraints on the chief executive, and regulation and competitiveness of participation in the political process. The combined score varies from –10 for a pure autocracy to +10 for a pure democracy.

a. Corrales and Westhoff 2006; Rød and Weidmann 2015. b. King, Pan, and Roberts 2013; Morozov 2012; Pierskalla and Hollenbach 2013; Shapiro and Weidmann 2013; Rød and Weidmann 2015. See also HRW 2015.

be distinguished by whether they are clientelist or pro- and policy makers need to have the incentives to poor, and bureaucracies can be classifi ed by whether make these investments. they are patronage-based or performance-oriented.89 Bureaucracies can have a disincentive to reor- These institutions, and their characteristics, matter ganize their work to take full advantage of digital because effective adoption of technology in organi- technologies. In the private sector, market competi- zations requires signifi cant investments in skills and tion forces fi rms to change and rewards them with changes in working arrangements, and politicians higher profi ts for their investments. In contrast, in DELIVERING SERVICES 179 government, effi ciency improvements can reduce an Figure 3.16 Digital technology projects agency’s budget and staff.90 Adopting new technolo- funded by the World Bank are more gies also requires learning new skills and changing successful in countries with processes, which can add to the workload, at least ini- higher-quality institutions tially, with unclear career rewards in patronage-based civil service systems, particularly given the diffi culties 2 in measuring the productivity of government agen- cies. Ministries and departments compete against one 1 another for scarce budgetary resources, leading to a silo mentality and resistance to horizontal collabora- tion. Most important, the automation of business pro- 0 cesses and improved monitoring reduce bureaucratic discretion and opportunities for petty corruption in patronage-based systems. –1 Project outcome In the absence of market competition, overcom- ing bureaucratic resistance to digital technologies –2 requires strong political leadership. But political incentives may also not be aligned with digital tech- nologies. Many e-government projects take a long –3 time to implement and thus elicit little interest from −2 −1 0 1 2 political leaders whose preferences are governed by Quality of institutions short election cycles. In clientelist political institu- tions, politicians are accountable largely to a narrow Sources: WDR 2016 team, based on World Bank (2015a) Digital Governance Projects Database; Worldwide Governance Index 2014; Denizer, Kaufmann, and group of elites. Digital technologies, by promoting Kraay 2013. Data at http://bit.do/WDR2016-Fig3_16. transparency and enhanced tracking and monitoring Note: Each observation represents an e-government project, for a total of of resources, can curtail avenues for rent-seeking— 530 projects. Quality of institutions is measured by the combined Worldwide Governance Index. Project outcomes are measured on a six-point scale based on “grand corruption,” in contrast to the petty corruption assessments by the World Bank’s internal evaluation unit. The graph controls for of service providers—that underpin these informal economic growth, project duration, and project size. political institutions. And citizens continue to face signifi cant barriers to voice and collective action importance of these complementarities of skills, orga- that digital technologies have not signifi cantly ame- nizational changes, and issues specifi c to the public liorated. Lifting these barriers depends more on the sector, such as procurement rules, as conditioning the strength of an independent media, supreme audit and impact of e-government. anticorruption agencies, and civil CSOs as sources of The extent of the institutional dependence of dig- credible information and as intermediaries between ital technologies in achieving impacts varies by the the citizen and the state. type of service and activity, based on three factors: the E-government projects funded by the World Bank degree to which citizens have an incentive to monitor are more successful in countries with stronger insti- the service and provide feedback; the extent to which tutions (fi gure 3.16). While project performance var- the processes for the production and delivery of the ies considerably within countries, project outcomes service or activity are based on tasks that can be made across countries, as self-evaluated by the World Bank, routine and standardized; and the measurability and are strongly positively linked to the quality of gov- extent to which the outputs and outcomes from the ernment institutions, as measured by the Worldwide task can be attributed to particular public actors Governance Indicators. or actions (fi gure 3.17). Services and activities that There is considerable evidence of the comple- citizens have an incentive to monitor, and that are mentarity between technology and the organization measurable and attributable to the efforts of specifi c of work in attaining productivity improvements in government offi cials, are compatible with political the private sector.91 The few empirical studies that incentives. Improving them yields immediate politi- investigate these complementarities for government cal benefi ts, as citizens care about these services, can have similar fi ndings. For example, investments in assess the service improvement, and can attribute information technology by police departments in this improvement to actions by policy makers and the United States lowered crime rates only when providers. If their delivery is based on repeatable, rou- accompanied by signifi cant organizational changes.92 tine tasks, these are more amenable to automation or The mixed record of automation also underlines the digitally enabled monitoring. These services are less 180 WORLD DEVELOPMENT REPORT 2016

Figure 3.17 Classifying public services and activities as to their amenability to improvement through digital technology

1 2 3

Do citizens have Is the delivery Can outcomes be based on the incentive to easily measured Examples monitor? routine tasks? and attributed? MORE amenable to improvement YES Cash transfers, registration, filing taxes through digital technology YES NO Family-oriented health services and self-care YES YES Utility services (electricity, water) NO NO Curative health care

YES Constructing roads YES NO Preventive health care, financial management NO YES Procurement of complex goods LESS NO amenable to improvement NO Teaching, policing, management through digital technology

Sources: WDR 2016 team, based on Batley and Mcloughlin 2015; Pritchett and Woolcock 2002; Wilson 1989; World Bank 2003.

conditional on the strength of the initial institutions utility services like household water and electricity. and relatively amenable to improvements through The tasks to deliver some of these services, like cash digital technology. transfers or fi ling taxes, are largely rule-based and By contrast, services that citizens do not have an clerical, or follow a standardized set of procedures, incentive to monitor and that are less measurable and ensuring the timely processing of work orders and attributable do not yield political benefi ts to is enough to deliver the service. These features help politicians. If the delivery of these services is based explain the successful uses of digital technologies for on tasks that are highly discretionary, policy makers welfare payments, water provision, property and busi- have less infl uence over the providers responsible ness registration discussed earlier, including through for these services. These services and activities are citizen feedback on service quality. It also explains the much more dependent on the quality of existing success of various m-health initiatives. Automating institutions; improvements through the application these tasks does require breaking down departmental of digital technologies are only incremental. This silos and changing administrative processes, but the variation helps explain the differential impact of dig- quick, easily visible, and easily attributable service ital technologies across the elements of government improvements to citizens can yield political benefi ts capability and citizen empowerment summarized that even clientelist politicians might have an interest in the table 3.1 and 3.2 scorecards, and why digital in supporting, though the political economy consid- technologies can substitute for poor institutions for erations vary by activity. The mixed impact of e-fi ling certain activities and can only complement existing systems for example, refl ects how these reforms can institutions for others. often confl ict with elite interests and are likely more Citizens and businesses have an incentive to dependent on institutional complements. monitor private goods or services that they use very By contrast, teaching, policing, and management frequently. These include the variety of registration are tasks for which no “user manual” can be written, and licensing services offered in one-stop centers; since providers are confronted daily with unique fi ling taxes; welfare payments; family-oriented circumstances and must exercise signifi cant judg- health services and self-care such as neonatal health ment on how to respond.93 As a result, these tasks are and patients’ adherence to treatment schedules; and particularly susceptible to problems of asymmetric DELIVERING SERVICES 181 information. As discussed, it is diffi cult for digital institutions, technology can substitute for weak ini- technologies to make a difference beyond addressing tial institutions for certain services, and in the process basic problems of provider absenteeism. Citizens also also help improve those institutions. That has been often lack the incentive or capacity to monitor these evident in the example of MajiVoice in Kenya, where services, either because they are public goods (for automation quickly transformed a patronage-based example, security), or are less visible (classroom teach- water utility to a performance-oriented one, and in ing or government administration). These diffi culties the examples of institutional transformation trig- in measurability, and therefore “attributability,” also gered by digital identifi cation. Moreover, the consid- make it hard for citizens to hold providers and pol- erable heterogeneity of institutions within countries, iticians accountable for performance through elec- varying by sector and locale, creates many openings tions—beyond deterring electoral fraud and voting for innovations that are specifi c to a given context.94 out corrupt politicians. They also make it diffi cult for How should this institutional heterogeneity be citizen collective action to be effective in the absence classifi ed to help guide policies? Under clientelism, of a strong civil society or willing governments to policies are more likely to be adopted if they generate partner with. These are services and activities where immediate and highly visible service improvements institutions matter much more for outcomes. Digital to citizens without reducing the rents of vested inter- technologies cannot replace weak institutions. ests. In patronage-based bureaucracies, reforms are Between these two extremes are the mixed cases more likely to succeed if they do not require signifi - where digital technologies can improve some aspects cant changes in administrative processes or much of the service or activity. Roads and other public infra- collaboration across agencies, or threaten the bureau- structure, fi nancial management, and government cracy with staffi ng and budgetary reductions. Under procurement are activities with low incentives for cit- pro-poor political institutions, there can be greater izens to monitor because the service is a public good, political ownership in reforming bureaucracies to or not visible to citizens; however, the measurability strengthen government capability, and in transpar- and attributability of outcomes may be high, and the ency and accountability initiatives. Performance- delivery process may have some elements that can oriented bureaucracies have the incentive to engage be automated (such as construction). E-procurement with ambitious initiatives, including the long- of complex goods, such as e-government systems, duration and less visible administrative automation cannot entirely remove discretion in the evaluation that often elicits little enthusiasm from politicians. of different bids, but it does establish audit trails and More generally, the variation in institutional performance indicators. Curative and preventive dependency by the nature of the service and activ- health care are private and public goods, respectively, ity opens reform possibilities and can be a guide for and so differ in the citizens’ incentive to monitor policies in different country contexts, as discussed in (although vaccinating children is a preventive mea- chapter 5. sure that parents are likely to have an incentive to monitor). But the highly discretionary nature of Notes clinical services and the diffi culties in measuring the outcomes of treatment render curative health care 1. The 2004 World Development Report, Making Ser- vices Work for Poor People, underlined that service more diffi cult to improve through the application of delivery failures are largely about failures of account- technology than preventive health campaigns that ability and weak government implementation follow more relatively standardized procedures (such capacity, not about poor policies or lack of resources. as control of communicable diseases). It showed how relationships of accountability— between citizens and policy makers, between policy makers and service providers, and between citizens The future of public services and providers—determine outcomes. Unaccount- This varying interaction between institutions and able politicians target policies and public spending digital technology provides entry points for reform to a narrow set of elites rather than providing public goods to benefi t the general population. Managers in different country contexts to improve public ser- and providers are unresponsive to policy makers vice delivery. This institutional perspective does not and to citizens, often using their positions to extract imply that digitally enabled reforms must wait for rents. And poor citizens are unable to organize col- a country’s institutions to improve to have impact. lectively to pressure policy makers and providers to Digital technologies can also strengthen institutions. address service delivery failures. See also Pritchett, In countries with clientelist and patronage-based Woolcock, and Andrews 2010. 182 WORLD DEVELOPMENT REPORT 2016

2. Based on the Online Service Index (OSI), published 12. Ayres, Raseman, and Shih 2009. by the United Nations Department of Economic 13. World Bank 2015c. and Social Affairs. The OSI assesses the range and 14. World Bank 2015c. functionality of government-to-citizen (G2C) and 15. Muralidharan, Niehaus, and Sukhtankar 2014. government-to-business (G2B) services offered 16. See http://global.census.okfn.org/. on the national website of each of the UN’s 193 17. Chopra 2014; Sunstein 2013; Goldstein and Dyson member states, from simple one-way provision of 2013. information, to two-way interaction, to integrated 18. Bayern 2015. “one-stop” portals. This functionality is based on the 19. These perceptions are based on the World Bank standard three-stage classifi cation of e-services in Enterprise Survey panel dataset (2008/09 and the literature. Informational services are basic web- 2012/13) from the European and Central Asian sites where citizens can obtain information on gov- countries. ernment ministries and agencies, access archives, 20. Kochanova, Hasnain, and Larson, forthcoming. and download forms to apply for public services. 21. World Bank 2015b; WDI. Transactional services are chiefl y two-way online 22. Deloitte 2012; Yilmaz and Coolidge 2013. communications for governments to share informa- 23. World Bank 2014a. tion and solicit feedback, and often on government 24. Doing Business (World Bank). websites that process fi nancial transactions, such as 25. Lewis-Faupel and others 2014. electronic tax fi ling. Connected services are citizen- 26. Krishna 2015. centric, whole-of-government services. Institutional 27. Heeks 2003. reform and advanced technologies enable citizens to 28. Sjoberg, Mellon, and Peixoto 2014. automatically receive benefi ts based on life-cycle or 29. World Bank 2015d. economic characteristics. They also allow agencies 30. Peixoto and Fox 2015. to easily share data to facilitate service delivery. And 31. Presentation given at the World Bank on ASAN they encourage accountability through monitor- Xidmet centers by the government of Azerbaijan; ing and feedback mechanisms. Depending on the Majeed 2014. intended benefi ciary, these services can be classifi ed 32. Based on data received from LAPOR. as government-to-government (G2G), G2C, or G2B 33. Based on discussions with government offi cials. (UN 2014). 34. Cantijoch, Galandini, and Gobson 2014. 3. For this Report, the World Bank has developed 35. Bhatti, Kusek, and Verheijen 2015. two indexes to measure e-government. The Core 36. WDR 2016 team based on data received from the Systems index measures the automation and inte- government of Punjab. gration of core G2G fi nancial and human resource 37. Astrom and others 2013. management systems, as well as revenue-related 38. Chaudhury and others 2006. G2B and G2C systems, in 198 countries, drawing 39. Dhaliwal and Hanna 2014; Muralidharan and others on a global dataset of e-government systems devel- 2014. oped by the World Bank. The Digital Identifi cation 40. Dufl o, Hanna, and Ryan 2012; Cilliers and others Systems index measures the presence and scope of 2013; Callen and others 2014; Aker and Ksoll 2015; digital identifi cation systems in government that Dhaliwal and Hanna 2014; Adelman and others 2015. can serve as a foundational platform technology for 41. WDR 2016 team based on data from the government myriad G2G, G2C, and G2B services, also drawing of Punjab. on a global dataset of these systems developed by 42. Oxford Policy Management 2015. the World Bank. These datasets were prepared by a 43. Bloom and others 2013; Aral, Brynjolfsson, and Wu World Bank team comprising Cem Dener, Sophiko 2012. Skhirtladze, Irene H. Zhang, and Doruk Yarin 44. World Bank 2014b. Kiroglu. 45. Kraemer and King 2006; Baldwin, Gauld, and Gold- 4. Reddick and Turner 2012; Gauld, Goldfi nch, and fi nch 2012. Horsburgh 2010. 46. For example, Bill Clinton, former president of the 5. UN 2014. United States, reportedly said that “in the new 6. Tomlinson and others 2013. century, liberty will spread by cell phone and cable 7. Agarwal and Labrique 2014. modem” (see http://www.techlawjournal.com 8. World Bank 2012. /trade/20000309.htm). 9. Kradt-Todd and others 2015. 47. Effective service delivery requires sound mecha- 10. Corbacho, Cibils, and Lora 2013. nisms for holding policy makers and bureaucracies 11. Perez-Truglia and Troiano 2015; Bø, Slemrod, and accountable. Democracy is one, but not the only, Thoresen 2014. For a review of the literature, see political system for establishing strong relation- Luttmer and Singhal 2014. ships of accountability. DELIVERING SERVICES 183

48. Based on data from Bishop and Hoeffl er 2014. Six 86. This is the Huduma citizen complaints portal based indicators relate to the freeness of the electoral on the Ushahidi platform, not to be confused with process before election day: the legal framework, the Huduma service centers recently established by electoral management bodies, electoral rights, voter the government of Kenya. register, ballot access, and campaign process. Four 87. Berdou and Lopes 2015. indicators relate to the fairness of voting and events 88. Schlozman, Verba, and Brady 2010. on or immediately after election day: media access, 89. World Bank 2003. voting process, role of offi cials, and counting of 90. Fountain 2001. votes. Each indicator has several criteria that must 91. Bresnahan, Brynjolfsson, and Hitt 2002; Brynjolfs- be met for it to be coded as “1” for the given election. son and Hitt 2000. If there is a suffi cient body of data to conclude that 92. Garicano and Heaton 2010. the conditions are not met, the indicator is coded 93. Lipsky 1980. as “0.” An election is considered free when at least 94. Banerjee and Dufl o 2014. four of the six relevant indicators are coded as “1.” An election is considered “fair” when at least two of the four relevant indicators are coded as “1.” References 49. Fujiwara 2015. Acemoglu, Daron, Tarek Hasan, and Ahmed Tahoun. 50. Callen and Long 2015. 2014. “The Power of the Street: Evidence from Egypt’s 51. Aker, Collier, and Vicente 2013. Arab Spring.” NBER Working Paper 20665, National 52. Fung, Russon Gilman, and Shkabatur 2010. Bureau of Economic Research, Cambridge, MA. 53. For a review of the literature, see Anderson 2007. Adelman, Melissa, Moussa Blimpo, David K. Evans, 54. This “liberation technology” literature is consider- Atabanam Simbou, and Noah Yarrow. 2015. “Can able. See, for example, Diamond and Plattner 2012. Information Technology Improve School Effective- 55. Prat and Stromberg 2005. ness in Haiti? Evidence from a Field Experiment.” 56. Gentzkow 2006. Unpublished paper, World Bank, Washington, DC. 57. Gemenis and Rosema 2014. Agarwal, Smisha, and Alain Labrique. 2014. “Newborn 58. Falck, Gold, and Heblich 2014. Health on the Line: The Potential mHealth Applica- 59. and others 2012. tions.” Journal of the American Medical Association 312 (3): 60. Hindman 2008. 229–30. 61. Farrell 2012; Sunstein 2009. Aker, Jenny C., Rachid Boumnijel, Amanda McClelland, 62. Gallup World Poll, http://www.gallup.com/services and Niall Tierney. 2014. “Payment Mechanisms and /170945/world-poll.aspx. Anti-Poverty Programs: Evidence from a Mobile 63. Fung, Russon Gilman, and Shkabatur 2013. Money Cash Transfer Experiment in Niger.” Working 64. Ferraz and Finan 2008. Paper 268, Center for Global Development, Washing- 65. Larreguy, Marshall, and Snyder Jr. 2014. ton, DC. 66. Keefer and Khemani 2003. Aker, Jenny C., Paul Collier, and Pedro Vicente. 2013. “Is 67. Keefer and Khemani 2011. Information Power? Using Mobile Phones and Free 68. Banerjee and others 2011. Newspapers during an Election in Mozambique.” 69. Humphreys and Weinstein 2012. Working Paper 328, Center for Global Development, 70. Bayern 2015. Washington, DC. http://papers.ssrn.com/sol3/papers 71. Alvarez, Hall, and Trechsel 2009. .cfm?abstract_id=2364162. 72. Spada and others 2015. Aker, Jenny C., and Christopher Ksoll. 2015. “Call Me Edu- 73. Diamond and Plattner 2012; Little 2015; Shirky 2008. cated: Evidence from a Mobile Monitoring Experi- 74. Acemoglu, Hasan, and Tahoun 2014. ment in Niger.” Working Paper 406, Center for Global 75. Valenzuela 2013. Development, Washington, DC. http://www.cgdev 76. Bennett and Segerberg 2011. .org/publication/call-me-educated-evidence 77. Hollenbach and Pierskalla 2014. -mobile-monitoring-experiment-niger-working 78. Pierskalla and Hollenbach 2013. -paper-406. 79. Banerjee and others 2010; Björkman and Svensson Alvarez, Michael R., Thad E. Hall, and Alexander H. 2009; Lieberman, Posner, and Tsai 2014; Olken 2007. Trechsel. 2009. “Internet Voting in Comparative 80. Fung, Graham, and Weil 2007. Perspective: The Case of Estonia.” Political Science and 81. Fox 2014; Grandvoinnet, Aslam, and Raha 2015. Politics 42 (3): 497–505. 82. Peixoto and Fox 2015. Anderson, Christopher J. 2007. “The End of Economic 83. Fung, Russon Gilman, and Shkabatur 2010. Voting? Contingency Dilemmas and the Limits of 84. Peixoto and Fox 2015. Democratic Accountability.” Annual Review of Political 85. Bayern 2015. Science 10 (1): 271–96. 184 WORLD DEVELOPMENT REPORT 2016

Aral, Sinan, Erik Brynjolfsson, and Lynn Wu. 2012. WPS/2014-14, Centre for the Study of African Econo- “Three-Way Complementarities: Performance Pay, mies, Oxford, U.K. http://www.csae.ox.ac.uk/datasets HR Analytics and Information Technology.” Manage- /free-fair-elections/. ment Science 58 (5): 913–31. Björkman, Martina, and Jakob Svensson. 2009. “Power to Astrom, Joachim, Hille Hinsberg, Magnus Jonsson, and the People: Evidence from a Randomized Experiment Martin Karlsson. 2013. “Citizen Centric E-Partici- on Community-Based Monitoring in Uganda.” Quar- pation: A Trilateral Collaboration for Democratic terly Journal of Economics 124 (2): 735–69. Innovation.” PRAXIS Center for Policy Studies, Tallin, Bloom, Nicholas, Erik Brynjolfsson, Lucia Foster, Ron Jar- Estonia. min, Ita Saporta-Eksten, and John Van Reenan. 2013. Ayres, Ian, Sophie Raseman, and Alice Shih. 2009. “Management in America.” Discussion Paper 13-01, “Evidence from Two Large Field Experiments That Center for Economic Studies, Washington, DC. Peer Comparison Feedback Can Reduce Residential Bø, Erland E., , and Thor O. Thoresen. 2014. Energy Usage.” NBER Working Paper 15386, National “Taxes on the Internet: Deterrence Effects of Public Bureau of Economic Research, Cambridge, MA. Disclosure.” Discussion Paper 770, Statistics Norway Baldwin, Norman J., Robin Gauld, and Shaun Goldfi nch. Research Department. 2012. “What Public Servants Really Think of E-Gov- Bond, Robert M., Christopher J. Fariss, Jason J. Jones, ernment.” Public Management Review 14 (1): 105–27. Adam Kramer, Cameron Marlow, Jaime E. Settle, and Banerjee, Abhijit, Rukmini Banerji, Esther Dufl o, Rachel James H. Fowler. 2012. “A 61-Million-Person Experi- Glennerster, and Stuti Khemani. 2010. “Pitfalls of Par- ment in Social Infl uence and Political Mobilization.” ticipatory Programs: Evidence from a Randomized Nature 489: 295–98. Evaluation in Education in India.” American Economic Bresnahan, Timothy, Erik Brynjolfsson, and Lorin M. Journal: Economic Policy 2 (1): 1–30. Hitt. 2002. “Information Technology, Workplace Banerjee, Abhijit, and Esther Dufl o. 2014. “Under the Organization, and the Demand for Skilled Labor: Thumb of History? Political Institutions and the Firm-Level Evidence.” Quarterly Journal of Economics Scope for Action.” NBER Working Paper 19848, 117 (1): 339–76. National Bureau of Economic Research, Cambridge, Bridge Academies. 2013. “The Bridge Effect: Comparison MA. of Bridge Pupils to Peers at Nearby Schools.” EGRA- Banerjee, Abhijit, Selvan Kumar, Rohini Pande, and Felix EGMA Evaluation Programme, Fall 2013 Results, Su. 2011. “Do Informed Voters Make Better Choices? Bridge International Academies, Mombassa, Kenya. Experimental Evidence from Urban India.” Unpub- http://www.bridgeinternationalacademies.com lished manuscript, Harvard University, Cambridge, /wp-content/uploads/2013/01/Bridge-International MA. http://www.povertyactionlab.org/node/2764. -Academies_White-Paper_The-Bridge-Effect Batley, Richard, and Claire Mcloughlin. 2015. “The Pol- _Nov-2014_Website.pdf. itics of Public Services: A Service Characteristics Approach.” World Development 74: 275–85. Brynjolfsson, Erik, and Lorin M. Hitt. 2000. “Beyond Bayern, Jessica. 2015. “Investigating the Impact of Open Computation: Information Technology, Organiza- Data: Case Studies of Kenya, Moldova, Philippines, tional Transformation and Business Performance.” and Uganda.” Background paper for the World Devel- Journal of Economic Perspectives 14 (4): 23–48. opment Report 2016, World Bank, Washington, DC. Budzier, Alexander, and Bent Flyvbjerg. 2012. “Over- Bennett, W. Lance, and Alexandra Segerberg. 2011. “Dig- spend? Late? Failure? What the Data Say about IT ital Media and the Personalization of Collective Project Risk in the Public Sector.” In Commonwealth Action: Social Technology and the Organization of Governance Handbook 2012/13: Democracy, Development Protests against the Global Economic Crisis.” Informa- and Public Administration, edited by Commonwealth tion, Communication & Society 14 (6): 770–99. Secretariat, 145–57. London: Commonwealth Berdou, E., and C. A. Lopes. 2015. “The Case of UNICEF’s Secretariat. U-Report (Uganda).” World Bank, Washington, DC. Bussell, Jennifer. 2012. Corruption and Reform in India: Best, Michael. 2013. Cables, Commissions and Cybercafés: Public Services in the Digital Age. Cambridge, MA: Cam- ICTs in Postconfl ict Liberia. Washington, DC: World bridge University Press. Bank. Callen, Michael, Saad , Ali Hasanain, and Yasir Bhatnagar, Subhash. 2009. Unlocking E-Government Poten- Khan. 2014. “The Political Economy of Public tial: Concepts, Cases and Practical Insights. New Delhi, Employee Absence: Experimental Evidence from India: SAGE Publications. Pakistan.” Working paper, Harvard University, Cam- Bhatti, Zubair, Jody Kusek, and Tony Verheijen. 2015. bridge, MA. “Logged On: Smart Government Solutions from Callen, Michael, and James D. Long. 2015. “Institutional South Asia.” World Bank, Washington, DC. Corruption and Election Fraud: Evidence from a Bishop, Sylvia, and Anke Hoeffl er. 2014. “Free and Field Experiment in Afghanistan.” American Economic Fair Elections: A New Database.” Working Paper Review 105 (1): 354–81. DELIVERING SERVICES 185

Cantijoch, Marta, Silvia Galandini, and Rachel Gobson. Falck, Oliver, Robert Gold, and Stephan Heblich. 2014. 2014. “Civic Websites and Community Engagement: “E-lections: Voting Behavior and the Internet.” Amer- A Mixed Methods Study of Citizen Empowerment ican Economic Review 104 (7): 2238–65. Initiatives.” APSA Annual Meeting paper. http://ssrn Farrell, Henry. 2012. “The Consequences of the Internet .com/abstract=2454315. for Politics.” Annual Review of Political Science 15 (June): Chaudhury, Nazmul, Jeffery Hammer, Michael 35–52. doi: 10.1146/annurev-polisci-030810-110815. Kremer, Karthik Muralidharan, and F. Halsey Ferraz, Claudio, and Frederico Finan. 2008. “Exposing Rogers. 2006. “Missing in Action: Teacher and Corrupt Politicians: The Effects of Brazil’s Publicly Health Worker Absence in Developing Countries.” Released Audits on Electoral Outcomes.” Quarterly Journal of Economic Perspectives 20 (1): 91–116. doi: Journal of Economics (May 2008): 703–45. http://eml 10.1257/089533006776526058. .berkeley.edu/~ffi nan/Finan_Audit.pdf. Chopra, Aneesh. 2014. Innovative State: How New Technol- Fountain, Jane E. 2001. Building the Virtual State: Informa- ogies Can Transform Government. New York: Atlantic tion Technology and Institutional Change. Washington, Monthly Press. DC: Brookings Institution Press. Cilliers, Jacobus, Ibrahim Kasirye, Clare Leaver, Pieter Fox, Jonathan. 2014. “Social Accountability: What Does Serneels, and Andrew Zeitlin. 2013. “Improving the Evidence Really Say?” Working paper, Global Teacher Attendance Using a Locally Managed Mon- Partnership for Social Accountability, Washington, itoring Scheme: Evidence from Ugandan Primary DC. Schools.” Policy Note 14/0186, International Growth Fujiwara, Thomas. 2015. “Voting Technology, Political Centre, London. Responsiveness, and Infant Health: Evidence from Corbacho, A., V. Fretes Cibils, and E. Lora. 2013. More Than Brazil.” Econometrica 83 (2): 423–64. Revenue: Taxation as a Development Tool. Washington, Fung, Archon, Hollie Russon Gilman, and Jennifer Shka- batur. 2010. Impact Case Studies from Middle Income DC: Inter-American Development Bank. and Developing Countries. London: Transparency and Corrales, Javier, and Frank Westhoff. 2006. “Information Accountability Initiative, Open Society Foundation. Technology Adoption and Political Regimes.” Interna- http://www.transparency-initiative.org/wp-content tional Studies Quarterly 50: 911–33. http://onlinelibrary /uploads/2011/05/impact_case_studies_fi nal1.pdf. .wiley.com/doi/10.1111/j.1468-2478.2006.00431.x/pdf. ————. 2013. “Six Models for the Internet + Politics.” Deloitte. 2012. “eTransform Africa: Modernising Gov- International Studies Review 15 (1): 30–47. doi: 10.1111 ernment through ICTs.” Deloitte Consulting. http:// /misr.12028. siteresources.worldbank.org/EXTINFORMATION Fung, Archon, Mary Graham, and David Weil. 2007. Full ANDCOMMUNICATIONANDTECHNOLOGIES Disclosure: The Perils and Promise of Transparency. Cam- /Resources/282822-1346223280837/Modernizing bridge, MA: Cambridge University Press. Government_Fullreport.pdf. Garicano, Luis, and Paul Heaton. 2010. “Information Denizer, Cevdet, Daniel Kaufmann, and Aart Kraay. 2013. Technology, Organization, and Productivity in the “Good Countries or Good Projects? Macro and Micro Public Sector: Evidence from Police Departments.” Correlates of World Bank Project Performance.” Jour- Journal of Labor Economics 28 (1): 167–201. nal of Development Economics 105 (November): 288–302. Gauld, Robin, and Shaun Goldfi nch. 2006. Dangerous http://www.sciencedirect.com/science/article/pii Enthusiasms: E-Government, Computer Failure, and Infor- /S0304387813000874. mation System Development. Dunedin, New Zealand: Dhaliwal, Iqbal, and Rema Hanna. 2014. “Deal with the Otago University Press. Devil: The Success and Limitations of Bureaucratic Gauld, Robin, Shaun Goldfi nch, and Simon Horsburgh. Reform in India.” NBER Working Paper 20482, 2010. “Do They Want It? Do They Use It? The ‘Demand National Bureau of Economic Research, Cambridge, Side’ of E-Government in Australia and New Zea- MA. land.” Government Information Quarterly 27 (2): 177–86. Diamond, Larry, and Marc F. Plattner. 2012. Liberation doi: 10.1016/j.giq.2009.12.002. Technology: Social Media and the Struggle for Democracy. Gemenis, Kostas, and Martin Rosema. 2014. “Voting Baltimore, MD: John Hopkins University Press. Advice Applications and Electoral Turnout.” Electoral Dufl o, Esther, Rema Hanna, and Stephen P. Ryan. 2012. Studies 36 (December 2014): 281–89. “Incentives Work: Getting Teachers to Come to Gentzkow, Matthew. 2006. “Television and Voter Turn- School.” American Economic Review 102 (4): 1241–78. out.” Quarterly Journal of Economics 121 (3): 931–72. doi: Dunleavy, Patrick, and Leandro Carrera. 2013. Growing the 10.1162/qjec.121.3.931. Productivity of Government Services. Cheltenham, U.K.: Goldstein, Brett, and Lauren Dyson. 2013. Beyond Trans- Edward Elgar. parency: Open Data and the Future of Civic Innovation. EC (European Commission). 2014. Eurostat (database). San Francisco: Code for America Press. European Commission, Luxembourg, http://ec.europa Grandvoinnet, Helene, Ghazia Aslam, and Shomikho .eu/eurostat/data/database. Raha. 2015. Opening the Black Box: Contextual Drivers 186 WORLD DEVELOPMENT REPORT 2016

of Social Accountability Eff ectiveness. Washington, DC: Sector Capability?” Background paper for the World World Bank. Development Report 2016, World Bank, Washington, Heeks, Richard. 2003. “Most eGovernment-for-Devel- DC. opment Projects Fail: How Can Risks Be Reduced?” Kradt-Todd, G., Erez Yoeli, Syon Bhanot, and David Rand. iGovernment Working Paper 14, Institute for Devel- 2015. “Promoting Cooperation in the Field.” Current opment Policy and Management, Manchester, U.K. Opinion in Behavioral Sciences 3 (June): 96–101. doi: http://unpan1.un.org/intradoc/groups/public/docu 10.1016/j.cobeha.2015.02.006. ments/NISPAcee/UNPAN015488.pdf. Kraemer, Kenneth L., and John Leslie King. 2006. “Infor- ————. 2006. “Understanding and Measuring E-Govern- mation Technology and Administrative Reform: Will ment: International Benchmarking Studies.” Unpub- E-Government Be Different?” International Journal of lished paper, University of Manchester, Manchester, Electronic Government Research 2 (1): 1–20. U.K. Krishna, Gunjan. 2015. “Improving Service Delivery ————. 2008. “Success and Failure Rates of eGovernment through the Use of Digital Technologies: A Case in Developing/Transitional Countries.” http://www Study of Karnataka e-Procurement.” Background .egov4dev.org/success/sfrates.shtml. paper for the World Development Report 2016, World Heyneman, Stephen P., and Jonathan M. B. Stern. 2014. Bank, Washington, DC. “Low Cost Private Schools for the Poor: What Public Larreguy, Horacio, John Marshall, and James M. Snyder Policy Is Appropriate?” International Journal of Educa- Jr. 2014. “Revealing Malfeasence: How Local Media tional Development 35 (March): 3–15. Facilitates Sanctioning of Mayors in Mexico.” NBER Hindman, Matthew. 2008. The Myth of Digital Democracy. Working Paper 20697, National Bureau of Economic Princeton, NJ: Princeton University Press. Research, Cambridge, MA. Hollenbach, Florian, and Jan Pierskalla. 2014. “Voicing Lemieux, V. 2015. “One Step Forward, Two Steps Back? Discontent: Communication Technology and Pro- Does E-Government Make Governments in Develop- test.” APSA Annual Meeting paper. ing Countries More Transparent and Accountable?” HRW (Human Rights Watch). 2015. “Human Rights Background paper for the World Development Report Watch Submission.” Background note submitted to 2016, World Bank, Washington, DC. the World Development Report 2016 team, Human Rights Lewis-Faupel, Sean, Yusef Neggers, Benjamin A. Olken, Watch, Washington, DC. https://www.hrw.org/sites and Rohini Pande. 2014. “Can Electronic Procure- /default/fi les/supporting_resources/hrw_submission ment Improve Infrastructure Provision? Evidence _re_wdr_2016_internet_for_development.pdf. from Public Works in India and Indonesia.” NBER Humphreys, Macarten, and Jeremy Weinstein. 2012. Working Paper 20344, National Bureau of Economic “Policing Politicians: Citizen Empowerment and Research, Cambridge, MA. Political Accountability in Uganda: Preliminary Lieberman, Evan, Daniel Posner, and Lily Tsai. 2014. Analysis.” Working paper, Columbia University, New “Does Information Lead to More Active Citizenship? York. International Budget Partnership. 2014. “Open Budget Evidence from an Education Intervention in Rural Index.” http://internationalbudget.org/opening Kenya.” World Development 60: 69–83. -budgets/open-budget-initiative/open-budget-survey/. Lipsky, Michael. 1980. Street-Level Bureaucracy: Dilemmas of Keefer, Philip, and Stuti Khemani. 2003. “Democracy, the Individual in Public Service. New York: Russell Sage Public Expenditures and the Poor.” World Bank, Foundation. Washington, DC. Little, Andrew. 2015. “Communication Technology and ————. 2011. “Do Informed Citizens Receive More . . . or .” Journal of Politics 78 (1). http://www.jstor.org Pay More? The Impact of Radio on the Government /stable/10.1086/683187. Distribution of Public Health Benefi ts.” Policy Research Luttmer, Erzo, and Monica Singhal. 2014. “Tax Morale.” Working Paper 5952, World Bank, Washington, DC. Journal of Economic Perspectives 28 (4): 149–68. doi: https://openknowledge.worldbank.org/bitstream 10.1257/jep.28.4.149. /handle/10986/3236/WPS5952.pdf?sequence=1. Majeed, Rushda. 2014. “A Second Life for One-Stop Kelly, Tim, and David Souter. 2014. The Role of Information Shops: Citizen Service Centers in Minas Gerais, Bra- and Communication Technologies in Post-Confl ict Recon- zil, 2003–2013.” Innovations for Successful Societies, struction. Washington, DC: World Bank. http://www Princeton University, Princeton, NJ. .infodev.org/infodev-fi les/fi nal_report_10_feb.pdf. Morozov, Evgeny. 2012. The Net Delusion: The Dark Side of King, Gary, Jennifer Pan, and Margaret E. Roberts. 2013. Internet Freedom. PublicAffairs. “How Censorship in China Allows Government Crit- Muralidharan, Karthik, Jishnu Das, Alaka Holla, and icism but Silences Collective Expression.” American Aakash Mohpal. 2014. “The Fiscal Cost of Weak Political Science Review 107 (2): 1–18. Governance: Evidence from Teacher Absenteeism in Kochanova, Anna, Zahid Hasnain, and Bradley R. Larson. India.” NBER Working Paper 20299, National Bureau Forthcoming. “Does E-Government Improve Public of Economic Research, Cambridge, MA. DELIVERING SERVICES 187

Muralidharan, Karthik, Paul Niehaus, and Sandip Sukh- Rød, Espen Geelmuyden, and Nils B. Weidmann. 2015. tankar. 2014. “Building State Capacity: Evidence from “Empowering Activists or Autocrats? The Internet in Biometric Smartcards in India.” NBER Working Authoritarian Regimes.” Journal of Research 53 (3): Paper 19999, National Bureau of Economic Research, 1–14. Cambridge, MA. Rodriguez, Camila, and Sebastian Monroy-Taborda. Olken, Benjamin. 2007. “Monitoring Corruption: Evi- Forthcoming. Targeted Public Transit Benefi ts in Bogotá. dence from a Field Experiment in Indonesia.” Journal Background note for the World Development Report of Political Economy 115 (2): 200–49. http://economics 2016, World Bank, Washington, DC. .mit.edu/fi les/2913. Ross, Terrance. 2014. “Is It Ever Okay to Make Teachers OMB (U.S. Offi ce of Management and Budget). 2010. Read Scripted Lessons?” The Atlantic. http://www.the 25-Point Implementation Plan to Reform Federal Informa- atlantic.com/education/archive/2014/10/is-it-okay-to tion Technology Management. OMB, Washington, DC. -make-teachers-read-scripted-lessons/381265/. Open Net Initiative. 2013. https://opennet.net/. Schlozman, Kay Lehman, Sidney Verba, and Henry E. Brady. 2010. “Weapon of the Strong? Participatory Oxford Policy Management. 2015. “EQUIP-Tanzania Inequality and the Internet.” Perspectives on Politics 8 Impact Evaluation.” Briefi ng Note 2, Oxford Policy (2): 487–509. Management. Shapiro, Jacob N., and Nils B. Weidmann. 2013. “Is the Peixoto, Tiago, and Jonathan Fox. 2015. “When Does ICT- Phone Mightier Than the Sword? Cellphones and Enabled Citizen Voice Lead to Government Respon- Insurgent Violence in Iraq.” International Organization siveness?” Background paper for the World Develop- 2: 247–74. ment Report 2016, World Bank, Washington, DC. Shirky, Clay. 2008. Here Comes Everybody: The Power of Orga- Perceptions of Electoral Integrity. 2015. https://dataverse nizing without Organizations. London: Penguin Group. .harvard.edu/dataverse/PEI. Sjoberg, Fredrik, Jonathan Mellon, and Tiago Peixoto. Perez-Truglia, Ricardo, and Ugo Troiano. 2015. “Tax Debt 2014. “The Effect of Government Responsiveness on Enforcement: Theory and Evidence from a Field Future Political Participation.” Unpublished paper, Experiment in the United States.” Meeting Paper 134, World Bank, Washington, DC. Society for Economic Dynamics. Spada, Paolo, Jonathan Mellon, Tiago Peixoto, and Fre- PIAAC (Programme for the International Assessment of drik Sjoberg. 2015. “Effects of the Internet on Par- Adult Competencies). 2014. OECD, Paris, http://www. ticipation: Study of a Policy Referendum in Brazil.” oecd.org/site/piaac/. Policy Research Working Paper 7204, World Bank, Pierskalla, Jan H., and Florian M. Hollenbach. 2013. Washington, DC. “Technology and Collective Action: The Effect of Cell Standish Group. 2014. “Chaos Report.” http://www Phone Coverage on Political Violence in Africa.” Amer- .standishgroup.com/. ican Political Science Review 107 (2): 207–24. Sunstein, Cass R. 2009. Republic 2.0. Princeton, NJ: Prince- Polity IV. 2015. http://www.systemicpeace.org/polity ton University Press. project.html. ————. 2013. Simpler: The Future of Government. New York: Prat, Andrea, and David Stromberg. 2005. “Commercial Simon & Schuster. Television and Voter Information.” Discussion Paper Tomlinson, Mark, Mary Jane Rotheram-Borus, Leslie 4989, Centre for Economic Policy Research, London. Swartz, and Alexander C. Tsai. 2013. “Scaling Up Pritchett, Lant, and Michael Woolcock. 2002. “Solutions mHealth: Where Is the Evidence?” PLOS Medicine 10 When the Solution is the Problem: Arraying the Dis- (2): e1001382. doi: 10.1371/journal.pmed.1001382. Toyama, Kentaro. 2015. Geek Heresy: Rescuing Social Change array in Development.” Center for Global Develop- from the Cult of Technology. New York: PublicAffairs. ment, Washington, DC. UN (United Nations). 2014. “E-Government Survey 2014: Pritchett, Lant, Michael Woolcock, and Matt Andrews. E-Government for the Future We Want.” UN, New 2010. “Capability Traps? The Mechanisms of Per- York. sistent Implementation Failure.” Center for Global Valenzuela, Sebastián. 2013. “Unpacking the Use of Social Development, Washington, DC. Media for Protest Behavior: The Roles of Informa- Rangan, V. Kasturi, and Katherine Lee. 2010. “Bridge tion, Opinion Expression, and Activism.” American International Academies: A School in a Box.” Harvard Behavioral Scientist 20 (10): 1–23. Business School, Cambridge, MA. Vassil, Kristjan. 2015. “Estonian E-Government Ecosys- Reddick, Christopher, and Michael Turner. 2012. “Chan- tem: Foundation, Applications, Outcomes.” Back- nel Choice and Public Service Delivery in Canada: ground paper for the World Development Report 2016, Comparing E-Government to Traditional Service World Bank, Washington, DC. Delivery.” Government Information Quarterly 29 (1): 1–11. Wilson, James Q. 1989. Bureaucracy: What Government doi: 10.1016/j.giq.2011.03.005. Agencies Do and Why They Do It. New York: Basic Books. Research ICT Africa. 2014. Household surveys. Cape- World Bank. Various years. Doing Business (database). town, South Africa. http://www.researchictafrica.net World Bank, Washington, DC, http://www.doing /home.php. business.org/. 188 WORLD DEVELOPMENT REPORT 2016

————. Various years. Skills Towards Employability and World Bank. 2015a. Digital Governance Projects (data- Productivity (STEP) household surveys (database). base). World Bank, Washington, DC, http://data World Bank, Washington, DC, http://microdata .worldbank.org/data-catalog/dg-projects-database. .worldbank.org/index.php/catalog/step/about. ————. 2015b. Global E-Government Systems (database). ————. 2003. World Development Report 2004: Making World Bank, Washington, DC, http://data.worldbank Services Work for Poor People. Washington, DC: World .org/data-catalog/pfm-systems-eservices-dataset. Bank. ————. 2015c. ID4D (Identifi cation for Development ————. 2012. “eTransform Africa: The Transformational database). World Bank, Washington, DC, http://data Use of Information and Communication Technolo- .worldbank.org/data-catalog/id4d-dataset. gies in Africa.” World Bank, Washington, DC. ————. 2015d. “MajiVoice: A New Accountability Tool to ————. 2014a. “Impact Evaluation of E-Filing and Improve Public Services.” World Bank, Washington, In-Depth Study of Risk-Based Audits in Tajikistan.” DC. World Bank, Washington, DC. Wrong, Michela. 2013. “Geeks for Peace.” New York ————. 2014b. “Pay Flexibility and Government Per- Times, February 28. http://latitude.blogs.nytimes formance: A Multicountry Study.” Working Paper .com/2013/02/28/geeks-for-peace/?_r=0. 88486, World Bank, Washington, DC. http://docu Yilmaz, Fatih, and Jacqueline Coolidge. 2013. “Can ments.worldbank.org/curated/en/2014/06/19630016 E-Filing Reduce Tax Compliance Costs in Developing /pay-flexibility-government-performance-multi Countries?” Policy Research Working Paper 6647, country-study. World Bank, Washington, DC.

190 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 3

e-health

Providing rural health care is a major challenge in full range of uses of information and communication countries with large rural populations. For example, technologies (ICTs)—from traditional administrative in Ethiopia, more than 80 percent of the population reporting systems to broader Health Management lives outside of urban areas and more than 30 per- Information Systems (HMIS) to telemedicine, elec- cent of the rural population is poor. Since 2003, the tronic medical records, clinical decision support, and Ethiopian government has trained and deployed over patient portals—and a full range of technologies, 40,000 Health Extension Workers (HEWs) to serve including internet and mobile applications.3 rural and other hard-to-reach populations. However, Public health and clinical care cannot be delivered HEWs are often isolated and lack the capacity to pri- safely, with high quality and in a cost-effective man- oritize urgent but unpredictable antenatal and post- ner, without seamless, sustainable, and secure data natal care. and information exchange at all levels of the health To improve information fl ows, the World Bank, the system. By 2011, 93 of 112 health systems in countries African Development Bank, and Addis Ababa Univer- surveyed by the World Health Organization (WHO) sity developed the FrontlineSMS platform.1 HEWs can had already adopted some form of an e-health or register pregnant women and newborns and receive m-health (mobile phone–based) approach, mostly for automated short message service (SMS) reminders information programs, emergencies, and telemedi- to notify them of key appointments and to track the cine. Yet overall, deployment has been slow. Too often stock of essential medicines. An evaluation showed data are captured in a way that cannot be shared as that by using existing mobile networks and low-cost needed because of interoperability issues or a lack of feature phones, the system improved the ability standards regarding the exchange of health informa- of health workers to deliver services and improve tion. Sometimes data are captured multiple times in health outcomes. More women had skilled assistance multiple ways, leading to duplication, inaccuracies, with their delivery, more women delivered in health and delays. Often they are not captured at all. centers, and more women received antenatal care. Low- and middle-income countries can do better. The system improved HEWs’ capacity to respond in First, they can build on the emerging experience of a timely manner and shows that in a context where developed countries and adapt systems to local con- internet coverage is low, mobile phones can be an ditions to benefi t from e-health without repeating effective way to improve health system performance.2 others’ mistakes. For example, Montenegro chose The internet and associated technologies have “strategic interoperability” as a leading principle the potential to expand health services in developing in developing its e-health system (fi gure F3.1). This countries, increase health system effi ciency, and lead includes integrative and long-term investment to better patient outcomes. E-health encompasses the planning and phased implementation; central devel- opment for common resources such as shared code- This sector focus was contributed by Dominic S. Haazen, Zlatan books and key registries of insured persons, drugs, Sabic, and Adis Balota. health professionals, and health institutions; shared e-HEALTH 191

Figure F3.1 Sequencing of e-health development in Montenegro

4

After that, Screening programs integration becomes almost trivial Transfusiology Institute IS Public Health Institute IS 3 Drug agency IS Use savings as an incentive and source of Hospitals IS—phase II further investments in complex systems Hospitals IS—phase I 2 Push for Emergency care IS efficiency Private dentists IS

Primary Health Care IS—full implementation 1 “Start where Primary Health Care IS support to PHC reform—pilot in 3 locations the money is” Health Insurance Fund Information System—migration to online mode

Control of drug distribution and use—Montefarm IS

Health Insurance Fund—infrastructure investments, introduction of new ID cards and forms

Health Insurance Fund Information System

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Note: IS = information system; PHC = primary health care. information system infrastructure; and central devel- expands the reach and impact of the often very small opment of key applications, including electronic number of highly trained physicians, assists clinical health records, e-prescriptions, and e-referrals. Imple- staff in rural and remote areas in making better diag- mentation sequencing “starts where the money is” noses and treatment decisions, and helps make the (the Health Insurance Fund Information System) and best use of limited health care funding. progresses along the elements of e-health develop- There is early evidence to suggest that e-health solu- ment, taking effi ciency and monetary incentives into tions, while costly to implement, can bring signifi cant account. Most business processes are now computer- cost savings. This is because the implementation of ized, serving 340 locations and 4,600 users, including human resources information systems, logistics man- general practitioners and nurses in primary and agement information systems, clinical decision sup- emergency care, hospitals, pharmacies, and private port tools, digital payments, fi nancial management dentists. Systems are integrated and optimized by use information systems, and SMS reminder systems can of shared resources. Routine reporting systems pro- address a variety of health system problems, including vide complete and reliable data. Information is used system ineffi ciencies, overuse of procedures, inappro- to make policy-making and management decisions, priate hospital admissions, corruption and fraud, and and effi ciency has improved throughout the system.4 missed appointments. Second, the absence of “legacy systems” can be Effective country ownership, good governance, an advantage. Countries can now make use of cloud and strong institutional and human capacity are computing to lower system costs and mobile tech- core to e-health planning and implementation. This nology to expand services to even the poorest and includes a strong legal basis for managing health- most remote locations. The potential value of e-health related data with appropriate safeguards.5 In addition can arguably be greater in poorer countries, since it to ensuring that health workers are able to effectively 192 WORLD DEVELOPMENT REPORT 2016

use such systems, this also implies the need for strong Notes health informatics training programs in order to develop a qualifi ed e-health workforce. National plan- 1. Otto and others 2014. ning, enterprise architecture, standardization, and 2. Bilal and others 2011. 3. WHO and ITU 2012. interoperability are essential for successful e-health 4. Case study of the Montenegro Health System implementation. Improvement Programme (MHIP) by Adis Balota; User-centered health care systems should leverage Montenegro Republican Health Insurance Fund; the unique capacity of citizens to contribute infor- University of Montenegro, Podgorica; and Zlatan mation and feedback. This enables health systems Sabic. to connect with clients when and where needed, but 5. EU 2012. clients can also access information and care, at their 6. See, for example, http://openmrs.org, http://www convenience. .village reach.org/impact/openlmis, https://open Replacing paper-based patient registers with source.com/health. electronic registers should help improve local health care quality and inform management decision mak- References ing. Similarly, increasing the use of e-health and m-health approaches and tools can support improved Bilal, N. K., C. H. Herbst, F. Zhao, A. Soucat, and C. decision making by frontline providers, including Lemiere. 2011. “Health Extension Workers in Ethiopia: Improved Access and Coverage for the Rural Poor.” In GPS-enabled tools and harnessing the revolution that Yes Africa Can: Success Stories from a Dynamic Continent, smartphone access to broadband content will bring edited by P. Chuhan-Pole and M. Angwafo, 433–43. about in developing countries. In this regard, more Washington, DC: World Bank. emphasis is needed to expand and improve the use EU (European Union). 2012. eHealth Task Force Report: Rede- and functionality of open-source software platforms signing Health in Europe for 2020. Brussels: European (for example, OpenMRS, OpenLMIS, and iHRIS); Union. develop new open-source platforms (for example, for Otto K., M. Shekar, H. C. Herbst, and R. Mohammed. health insurance and training); and support open- 2014. Information and Communication Technologies for source frameworks (for example, OpenHIE).6 Health Systems Strengthening Opportunities—Criteria for Information and communication technology Success, and Innovation for Africa and Beyond. Washing- platforms (web, social media, SMS campaigns, direct ton, DC: International Bank for Reconstruction and Development/World Bank. access to personal data in electronic health records) WHO (World Health Organization) and ITU (Interna- can be leveraged to enhance accountability, trans- tional Telecommunication Union). 2012. National parency, and empowerment of citizens to be active eHealth Strategy Toolkit Overview. Geneva: WHO and contributors to governance in health and central to ITU. health care delivery.

194 WORLD DEVELOPMENT REPORT 2016

ENABLING DIGITAL DEVELOPMENT Digital identity

Individuals need mechanisms to identify one another an electronic legal representation of an individual. and to identify themselves to their communities and Through the use of personal identifi cation numbers governments. While this point may be obvious, it (PINs) to authenticate the holder against a digital card is profoundly important for people’s welfare. Sim- credential, people can access public services remotely ple mechanisms—familiarity, appearance, perhaps and even sign legal documents and contracts with the vouching by an elder—are suffi cient in small, intimate same legal validity as if they were signed in person. communities. Wider societies and economies require more formal systems—traditionally physical tokens, Country-specifi c use of such as a paper-based identifi cation (ID) card that includes the signatures or representations of their digital identity holders, and is verifi ed against documents stored in Most developing countries have some form of digital a central registry. But these formal systems are fail- ID scheme tied to specifi c functions and serving a ing in the developing world. Nearly 2.4 billion people subset of the population, but only a few have a multi- are not registered. They are usually the poorest and purpose scheme that covers the entire population. most marginalized members of society; about one-

SPOTLIGHT 4 Eighteen percent of developing countries have a 1 quarter are children. They are excluded from a range scheme that is used for identifi cation purposes only; of rights and services, such as health care, enrollment 55 percent have digital IDs that are used for specifi c in school, social welfare, and fi nancial services. functions and services like voting, cash transfers, Identity should be a public good. Its importance is or health; and only 3 percent have foundational ID now recognized in the post-2015 development agenda, schemes that can be used to access an array of online specifi cally as a Sustainable Development Goal (SDG) and offl ine services (fi gure S4.1). Twenty-four percent target to “promote peaceful and inclusive societies for of developing countries have no digital ID system. sustainable development, provide access to justice Although the concept of digital ID is universal, for all, and build effective, accountable, and inclusive it plays somewhat different roles depending on the 2 institutions at all levels.” One of the indicators is to country context. In high-income countries, digital ID “provide legal identity for all, including birth regis- represents an upgrade from well-established, robust tration, by 2030.” The best way to achieve this goal legacy physical ID systems that have worked reason- is through digital identity (digital ID) systems, central ably well in the past. Belgium, Estonia, Finland, France, registries storing personal data in digital form and the Republic of Korea, and Singapore are some of the credentials that rely on digital, rather than physical, countries leveraging existing physical identity infra- mechanisms to authenticate the identity of their structure to create digital ID ecosystems, enabling holder. India’s massive Aadhaar program, which has them to deliver public services more effi ciently. enrolled over 950 million people, has dispensed with Low-income countries, by contrast, often lack the physical ID card altogether. Estonia has created robust civil registration systems and physical IDs Contributed by Joseph Atick, Mariana Dahan, Alan Gelb, and and are building their ID systems on a digital basis, Mia Harbitz. leapfrogging the more traditional physically based digital ID, withaccess India, Moldova,andPakistan. e-services. SuccessfulexamplesarefoundinAlbania, services whilesupportingtheemergenceofsome ening andprogressivelyreplacingphysicalidentity and continuestobeunregistered. is thatinmanycasesthe0–18populationexcluded, tion systemswithoutasolidcivilregistrationsystem risk associatedwithleapfroggingtocivilidentifi oped inBangladesh,Guinea,andKenya.Onepotential main immediategoal.Suchsystemsarebeingdevel- Identifi system. lion peryearwhen appliedthroughoutthecountry. uefi transfers toAadhaar-linkedbank accountstobuyliq- India’s fuelsubsidyprogram, implementingcash benefi that thosewhoareentitledto receivesubsidiesor accounts linkedtoauniqueidentifi er. Thisensures Digital IDsenabletargetedcash transferstobank and welfare distribution Effi cientmanagement ofsocial programs sanctity ofelections. from thegovernmentpayroll,andimproving social welfareprograms,removingghostworkers in atleastthreekeyareas—effi anecdotal, butthereisagrowingbodyofresearch Evidence oftheimpactdigitalIDisstilllargely Evidence ofimpact Figure S4.1 Figure schemes across countries across schemes /WDR2016-FigS4_1. Source: to alargenumber In middle-incomecountries,digitalIDisstrength- ed petroleumgascylinderssaved aboutUS$1bil- Identification and of onlineservices access tolimited World Bank ID4D database (various years). Data at http://bit.do at Data years). (various database ID4D Bank World ts areactuallygettingthem. Forexample,in purposes only Multipurpose Identification services None cation, ratherthan e-services,isthe Diff 0 erent types of digital ID ID ofdigital types erent Percentage ofcountries Develo High-income countries 20 cient management of p in g countries 40 ca- 60 3

and effi government expendituresthroughreducedleakage tal ID,potentiallysavingoverUS$11billionperyearin that arebeingconvertedtodirecttransfersusingdigi- This isjustoneofmanysubsidyprogramsinIndia an areaoffurtherresearch. ough enrollment processofthetargetedpopulation. These inturnhave beenestablishedthrougha thor- systems, databases,andcivilor populationregistries. Digital IDschemesrelyonabackbone ofconnected ID schemes Developing eff 2015 elections. Nigeria useddigitalIDstopreventvoterigginginits Improving electoral integrity from thebiometricvoterIDs. Kenya andSomalia,havenotreapedthesamebenefi by digitalidentity.However,othercountries,suchas results inthefaceoftransparencybroughtabout were cast,anditwasdiffi cult torigorcontestthe the electionwasconductedsuccessfully:allvotes there weresomeoperationalchallengesatthepolls, thus preventing4millionduplicatevotes.Although chip) andusedcardreaderstoauthenticatevoters, encoded thefi ngerprints oftherightfulholderona voters usingbiometrics(issuingvotercardsthat nearly 20,000percentinoneyear. annually, andprovidingareturnoninvestmentof about 62,000suchghostworkers,savingUS$1billion ID systemforcivilservantsthatenabledittoremove fiscal savingsandeffi ciency gainsfromdigitalID. 40 percentinZimbabwe,pointingtothesubstantial ranging from10percenttoashighanestimated ghost workersisevenworseinmanyothercountries, even bealive. not actuallyworkforthegovernment,andmay imply thatmanyindividualspaidfromthepayrolldo Ghana, Uganda,andZimbabwe,butweaksystems payroll occupiesthebulkofnationalbudgets for capitalinvestments.Forexample,thepublic from bloatedcivilservicewagesthatleavelittleroom The budgetsofmanydevelopingcountriessuffer government payroll Removing ghost employees from the and Turkey. of Egypt,Ghana,Indonesia,Pakistan,SouthAfrica, corruption areoccurringinChile,theArabRepublic schemes duetoduplicates,“ghost”benefi or securityprograms,healthinsurance,andpension of digitalIDinreducingleakagesforsocialprotection ciency gains. 5 Nigeriarecentlyimplementedadigital 8 Thesystemenrolledabout68million 4 Otherexamplesofthebenefi ective digital 9 Therefore,thisremains DIGITAL IDENTITY 6 Theimpactof ciaries, and 7 ts ts

SPOTLIGHT 4 195 196

SPOTLIGHT 4 WORLD DEVELOPMENT REPORT 2016 on theholder’sfingerprints oririsscan. altogether, providingremoteauthenticationbased India’s Aadhaarprogramdispenseswiththecard able toactasabankcard,drivinglicense,andsoon. only ofidentity)ormultifunctional,withthecard can alsobesingle-function(andprovideevidence codes ormoreadvancedchip-basedsmartcards;they credentials, whichmaybecardsequippedwithbar identity andoneuniqueidentifyingnumber. ensure thateachindividualhasonlyoneregistered programs toeliminateduplicateenrollmentshelp ric dataandtraditionalbiographicaldata,aswell Many programsnowincludetheuseofbothbiomet- and thelike),referredtoasfunctionalschemes. protection orsecurity,pensions,healthinsurance, response toaspecifi capplication(elections,tax,social secure platform. nication technology(ICT)todevelopareliableand throughout thecountryininformationandcommu- functional) registries;andcoordinatedinvestments and interoperabilitywithother(administrativeor system; cleardefi nitions fortheinterconnectivity ers theaccessibilityandprotectivemeasuresof rooted inanupgradedlegalframeworkthatconsid- tion, storage,andmanagementofdigitalidentity. technology withconsiderablepotentialforregistra- world’s 7billionmobilesubscriptions,makingthisa oping worldishometomorethan6billionofthe region haveamobilephonesubscription.Thedevel- ID, butmorethantwo-thirdsoftheresidentsin population insomecountriesiswithoutoffi Saharan Africa,forexample,morethanhalfofthe and widespreadaccesstodigitalservices.InSub- governments seekingtoprovideidentitycredentials confidence inelectronictransactions. include e-signatures,e-seals,andtimestamps—toadd tion features—suchaselectronictrustservices,which mobile environmentsrequireenhancedauthentica- tional identity legal identityacrossallapplications,knownasfounda- capable ofsupportingtheentirerangeneedsfor may bedevelopedasuniversalmultipurposesystems tional schemeor throughharmonizationofthe mul- identity forall,eitherthrough auniversalfounda- the countrycontextis,priority shouldbetoconfer ID hasbecomedefactonational ID).Nomatterwhat foundational (inBangladesh,Haiti, andMexico,voter over time;oftenfunctionalones evolvetobecome tional andfoundationalsystems isnotimmutable The digitalrecordisthebasisforissuanceof Digital identification systemsmaybedeveloped in For adigitalIDsystemtobeeffective,itmust Mobile devicesofferacompellingpropositionfor schemes. Thisdistinctionbetweenfunc- 10 Onlineand 11

Or they cial • • Institutional andadministrativeconcernsaboutthe Legal andregulatoryconcernsabouthowtobestdeter- • as critical: world. experience ofdigitalIDschemes’rolloutsaroundthe guidelines thathaveemergedfromthecollective ful implementationcanbemitigatedbyadopting high expectations.Risksassociatedwithunsuccess- politicized, andaresubjecttofailuredeliveron Digital IDschemestendtobecomplex,areoften mitigation and Risks Business modelsandprocurementconcerns engendered • • Technological concerns aboutworkingwiththepri- totality theyachievefullcoverage. titude ofexistingfunctionalsystems,sothatintheir schemes. eligibility andnotabirthright,asinfoundational ipation infunctionalIDsisamatterofprogram risk ofexclusionwouldalsobehigher,aspartic- would lackinteroperabilityandconsistency.The is ariskofpatchworkcompetingschemesthat government. Withouteffectivecoordination,there affi cies independentofanylineministry(orloosely affairs; andmorerecentlyinspecial-purposeagen- located intheministryofinterior,justice,orhome The legalorfoundationalregistriesaretraditionally authenticate identitiesofbenefi registries orlineministriesthatneedtoverify registries, andtheirinteractionwithfunctional institutional locationofthecivilandidentifi involving registrationorcertifi from accessingservices,orincreasedrent-seeking onerous mandatesthatcoulddeterindividuals consequences suchasinadvertentexclusions, primary legislationorrulestoavoidunintended the privacyofpersonaldata;andhowtocraftnew lected underdigitalIDschemes;howtosafeguard mine thetypes,extent,anduseofinformationcol- larity andopen standards; lackofcostingguidance dors; lackofopenarchitecture anchored onmodu- by technologysolutionsthatare tiedtospecifi security ofidentityrecords. metrics, aswellthelong-term accessibilityand ensure datasecurity,particularly intheuseofbio- different servicesandsolutionproviders; authentication protocolsfordataexchangeamong exclusion; ensureinteroperabilityandtrusted structure thatcanreachremoteareasandprevent vate sectortodevelopasustainabledigitalinfra- liated withone),andreportingtothe centerof 12 Inthisrespect,severalareasoffocusemerge 13 cates. ciaries or clients. or ciaries cation c ven- . 7. Proceedings ofParliamentZimbabwe,February GelbandClark2013. 6. 5. Public disclosuresinmediabyfi nance ministersof Banerjee 4. 2015. 3. Barnwal 2015. DahanandGelb2015. 2. 1. World BankID4Dglobaldataset(April2015).This Notes inclusive oversightandconcertedglobalaction. the global,regional,andnationallevelstoensure results. Equallyimportantisdonorcoordinationat and reformswillallbecriticalinachievingtangible users. Incentives,technology,foreignassistance, and humanresources,and,critically,thetrustof interagency cooperation,mobilizationoffi strong leadership,asupportivelegalframework, • Country-specifi collected andanalyzedonanunprecedentedscale. in “bigdata”analyticsthatallowinformationtobe purposes maybeenabledbytherecentadvances for trackingofethnicgroupsandothernefarious citizens’ credentials.UsesofdigitalIDschemes standards forthemutualrecognitionofforeign tions, evenastheworldhastakenstepstodefi tials. Thiscandifferacrosscountriesandapplica- constitutes acceptableuniqueidentifyingcreden- uptake. business modelsanddigitalID–enabledservices’ of variousITcomponents;andabsenceviable Overcoming thesechallengesandbarriersrequires 28, 2012. International andID4Africa. several countries,monitoredbyIdentityCounsel istered children. number includesapproximately600millionunreg- c andcross-borderconcernsaboutwhat nancial ne World Bank.Variousyears.ID4D(Identifi Gelb, Alan,andJuliaClark.2013.“Identifi cationforDevel- Dunning, Casey,AlanGelb,andSnehaRaghavan.2014. Dahan, Mariana,andAlanGelb.2015. Barnwal, Prabhat.2015.“CurbingLeakageinPublic Banerjee, ShwetaS.2015.“FromCashtoDigitalTransfers References 1. 13. For example,childrenarenoteligibletoregisterin GelbandClark2013. 12. 11. This terminologywasfi Dunning,Gelb,andRaghavan2014. rst adoptedbyGelband 10. GelbandClark2013. 9. BasedonvariousreportsintheNigerianmedia. 8. -dataset. DC. http://data.worldbank.org/data-catalog/id4d Development database).WorldBank,Washington, 315, CenterforGlobalDevelopment,Washington,DC. opment: TheBiometricsRevolution.”WorkingPaper ment, Washington,DC. Agenda.” PolicyPaper46,CenterforGlobalDevelop- “Birth Registration,LegalIdentity,andthePost-2015 /role-identifi Washington, DC.http://www.cgdev.org/publication World BankandCenterforGlobalDevelopment, tion inthePost-2015DevelopmentAgenda.”Essays, /subsidyLeakageUID.pdf. New York.http://www.columbia.edu/~pb2442 University SchoolofInternationalandPublicAffairs, Evidence fromIndia’sFuelSubsidies.”Columbia Programs withBiometricIdentifi -Cash-to-Digital-Transfers-in-India-Feb-2015_0.pdf. http://www.cgap.org/sites/default/files/Brief-From Group toAssistthePoor(CGAP),Washington,DC. in India:TheStorySoFar.”CGAPBrief,Consultative included inpovertyprograms. voter rosters,whilemiddle-incomefamiliesarenot Clark (2013). cation-post-2015-development-agenda. DIGITAL IDENTITY “Role ofIdentifi cto Systems: cation cto for cation ca-

SPOTLIGHT 4 197 4. Sectoral policies

5. National priorities

6. Global cooperation

CHAPTER 4 Sectoral policies

Making the internet followed by private participation, and added mobile and internet networks later. But developing countries universal, aff ordable, open, are jumping straight to mobile networks, built by the and safe private sector. This may leave gaps in the backbone infrastructure, especially in rural areas, possibly Access to mobile phones is close to universal, and requiring investment through public-private part- prices are falling in most countries, thanks to policies nerships (PPPs) for the full benefi ts of high-speed based on market competition, private participation, networks to be enjoyed by all. and light-touch regulation. But today’s digital econ- While availability, accessibility, and affordability omy also requires universal access to the internet—at remain concerns, the challenges facing internet stake- broadband speeds. First-generation policies for the holders today are as much about how networks are information and communication technology (ICT) used (demand) as how they are built (supply). Global sector, aimed at universal access and affordability, interconnectedness introduces new vulnerabilities have proved successful for phone service, and sup- in areas where coordination mechanisms are weak, ply-side policies should also work well for the internet. still evolving, or based on nongovernment models. But with more than half the world still offl ine, the ben- Threats to cybersecurity are undermining confi dence efi ts of the internet are unevenly distributed. Next- in the internet and increasing the costs to businesses generation policies must also focus on demand-side and governments, resulting in economic losses as issues of digital literacy, as well as privacy, cybersecu- well as higher security spending. For privacy and rity, and internet governance, where a global consen- data protection, different countries are taking quite sus has yet to emerge. different approaches, making it harder to develop In the last decade, all countries have benefi ted global services. Ensuring safe and secure access will from the rapid spread of mobile communication require greater international collaboration based on a networks. But only 15 percent of the world’s citizens multistakeholder model. 1 have access to affordable high-speed internet, and Converting connectivity into digital dividends the prices for service vary enormously. This refl ects will work best where an open access internet ecosys- policy failures in some countries, such as regulatory tem allows content creation and applications devel- capture, troubled privatizations, ineffi cient spec- opment to thrive. ICT clusters tend to form naturally, trum management, excessive taxation of the sector, and governments do not need to intervene to create or monopoly control of international gateways. To them. But they can help clusters along and avoid sti- achieve better development outcomes, governments fl ing growth unintentionally through high tariffs or need to address these failures through open consulta- restrictions on openness. Most countries have found tive policy-making processes involving the industry it useful to develop national ICT sector strategies for and users. broadband, for e-government, and for local content. Developing countries are following a different The process of developing these strategies, through route from developed ones. Most member-countries multistakeholder consultations, can be just as useful of the Organisation for Economic Co-operation and as the strategies themselves—and ensures that tar- Development (OECD) benefi ted from initial state-led gets are realistic and actionable. Policy challenges are investment in their fi xed telephone infrastructure, summarized in box 4.1. SECTORAL POLICIES 201

Box 4.1 Policy challenges for digital development

Offl ine, and missing out Only around 15 percent of the world’s population currently 2 billion people remain largely untouched by information and has aff ordable high-speed access to the internet (fi gure communication technologies (ICTs), and half a billion live B4.1.1). Use of mobile phones, reaching almost three- outside areas with a mobile signal. The world’s offl ine popu- quarters of the world’s population, provides the main form lation is mainly in India and China, but more than 100 million of internet access in developing countries. But the lives of people are also offl ine in North America, mainly in Mexico.

Figure B4.1.1 Global ICT access

a. ICT access by population b. A closer look at the world’s offline population

Total global population Congo, Dem. Rep. Philippines 68 million ~7.4 billion Mexico 63 million Ethiopia 70 million Russian Federation 95 million 55 million Brazil Iran, Islamic Rep. Within 98 million 54 million mobile coverage Nigeria Myanmar 111 million 53 million Bangladesh Vietnam 7 billion 148 million 52 million Pakistan United States 165 million 51 million Mobile phones Tanzania Indonesia 49 million 213 million Thailand China 48 million 5.2 billion 755 million Egypt, Arab Rep. 42 million Turkey 41 million

Total India Total Countries internet users 1.063 billion internet users outside of the top 20 3.2 billion 3.2 billion

High-speed High-speed internet internet 1.1 billion 1.1 billion

Sources: World Bank 2015; Meeker 2015; ITU 2015; http://GSMAintelligence.com; UN Population Division 2014. Data at http://bit.do /WDR2016-FigB4_1_1. Note: High-speed internet (broadband) includes the total number of fi xed-line broadband subscriptions (such as DSL, cable modems, fi ber optics), and the total number of 4G/LTE mobile subscriptions, minus a correcting factor to allow for those who have both types of access. 4G = fourth-generation; DSL = digital subscriber line; ICT = information and communication technology; LTE = Long Term Evolution.

Connected, but in the slow lane Megabucks for megabytes Developed and developing countries are following dif- ICT prices are falling globally, but large diff erences remain ferent routes to the information society (fi gure B4.1.2). In (map B4.1.1). Although Europe has some of the highest the OECD (Organisation for Economic Co-operation and prices in the world for mobile calls and texts, prices are Development) countries, fi xed-line networks came fi rst, generally cheaper for data. North, South, and Central and now form the backbone for internet access. But most America have high prices for mobile data, in part due to developing countries jumped straight to mobile networks bundling, but cheaper prices for fi xed data. Northern Asia without investing fi rst in connectivity. The consequence for is generally cheaper than southern Asia. In Africa, prices are many users in developing countries is a second-class inter- generally cheaper for mobile than for fi xed-line data. Price net: slow, expensive, and rarely “always on.” diff erences refl ect policy failures as much as market failures. Governments need to go further in liberalizing market entry, making available more spectrum, and encouraging invest- ment to achieve more aff ordable prices.

(Box continues next page) 202 WORLD DEVELOPMENT REPORT 2016

Box 4.1 Policy challenges for digital development (continued)

Figure B4.1.2 Network buildout (subscriptions per 100 population) in OECD and low- and middle-income countries, 1990–2014

140 Fixed Fixed 24.8% 9.5% 120

Mobile Mobile Mobile 100 75.2% 90.5%

OECD countries, Low- and middle- 80 2014 income countries 2014 60

40 Fixed 20 Subscriptions population per 100 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014

OECD, fixed Low- and middle-income, fixed OECD, mobile Low- and middle-income, mobile

Source: Adapted from ITU World Telecommunication/ICT Indicators database. Data at http://bit.do/WDR2016-FigB4_1_2. Note: OECD = Organisation for Economic Co-operation and Development.

Map B4.1.1 Price of mobile and fi xed broadband services

US$, purchasing power parity, 2015

a. Mobile broadband services, price per gigabyte a month

Price of 1 GB of data per month, in US$ PPP (based on nearest offer to 1 GB data usage per month, Europe inset commercially available from the largest operator in the local market, in US$ PPP, implied 2014 rates) Caribbean inset 2.00–10.99 11.00–16.49 16.50–22.99 23.00–31.99 Above 32.00 No data/no service available

IBRD 41652 (Box continues next page) SECTORAL POLICIES 203

Box 4.1 Policy challenges for digital development (continued)

Map B4.1.1 Price of mobile and fi xed broadband services (continued)

US$, purchasing power parity, 2015

b. Fixed broadband services, residential, price per 1 Mbit/s

Prices for fixed, residential monthly broadband service, in US$ PPP (based on lowest price of nearest offer to 1 Mbit/s Europe inset usage per month, 10 GB data usage cap, residential DSL, in US$ PPP, implied 2014 rates) Caribbean inset 0–5.99 6.00–22.99 23.00–49.99 50.00–109.99 110.00 and above No data/no service available

IBRD 41653 Sources: WDR 2016 team, with additional data from Oxford Internet Institute, Google, http://blogs.oii.ox.ac.uk/roughconsensus/2014/02/could -you-aff ord-facebook-messenger-in-cameroon-a-global-map-of-mobile-broadband-prices/ and Plot and Scatter. Data and an interactive map of mobile broadband prices, as well as aff ordability measures, at http://bit.do/WDR2016-MapB4_1_1a and fi xed broadband prices at http://bit.do /WDR2016-MapB4_1_1b. Note: DSL = digital subscriber line; GB = gigabyte; Mbit/s = megabits per second; PPP = purchasing power parity.

Shaping the digital economy owner and operator. That role is now pared back to policy maker and regulator, establishing an enabling Government policies and regulation of the internet environment for the private sector to do most of the help shape the digital economy. Particularly through work. For mobile networks and the internet, govern- their policies for the ICT sector, governments and ments have been less directly involved, but many of regulatory agencies create an enabling environment them are seeking a more active role in shaping the for the private sector to build networks, develop ser- digital economy. Broadband internet, in particular, is vices, and provide content and applications for users. seen as a general-purpose technology,2 essential for Increasingly, governments seek to cooperate across the competitiveness of nations,3 and governments borders on issues such as cybersecurity, privacy, and have invested more than US$50 billion in broadband cross-border data fl ows. networks since 2009 as part of stimulus packages.4 Internet-enabling policies have evolved over Most also have national broadband plans.5 time, especially those for the ICT sector (chapter 5 The internet has transformed telecommunication looks at complementary policies such as those for networks. In the same way that containerization rev- skills, regulations, and institutions). In the past, the olutionized physical trade,6 so the “packetization” of government’s main role was in building the fi xed- data has commoditized digital trade.7 Networks have line telecommunication network and acting as both shifted from primarily carrying voice tele phony to 204 WORLD DEVELOPMENT REPORT 2016

Box 4.2 Is the internet a public good?

In considering what policies to apply to the internet, a key externalities (a club good). While the private sector can question is whether it is a public good. If it is, government and should take the lead in providing internet networks intervention would be easier to justify. The internet does not and services, achieving universal and accessible internet is have all the characteristics that economists conventionally a legitimate public policy goal. The bulk of the investment associate with a public good—such as being nonexcludable required to achieve such a goal can come from the private (people can be prevented from using it) or nonrivalrous sector, though many countries have invested public money (a user who hogs bandwidth may impair the quality of ser- to improve internet access by poor households and those vice for other users). Even so, the internet provides public living in remote areas. In a recent global survey, some access to a vast library of useful information, it uses shared 83 percent of users said they believe that aff ordable inter- resources such as the spectrum, and it relies on network net access should be a basic human right.a Several govern- security; all of which may be considered public goods. ments have followed Finland’s lead in defi ning access to the Moreover, government intervention in promoting the inter- internet at broadband speeds as a legal right and a goal of net may be justifi ed because it has the characteristics of a universal access policy.b High-speed, aff ordable broadband general-purpose technology, like electricity. has been described as a foundation stone of modern soci- A more accurate way of expressing the characteristics ety.c But it is something that more than fi ve-sixths of the of the internet is that it is a private good with positive global population still lacks.d

a. The Global Survey on Internet Security and Trust (2014), conducted by the Centre for International Governance Innovation and Ipsos, a market research company, polled over 23,000 internet users in 24 countries; see https://www.cigionline.org/internet-survey. b. World Bank/ITU, ICT Regulation Toolkit at http://www.ictregulationtoolkit.org/en/toolkit/notes/PracticeNote/3270. c. UN Broadband Commission 2014, 8. d. ITU 2014.

now providing a wide array of multimedia applica- and the latter on content. Internet regulation has tions, with internet protocol (IP) as the shared lan- some characteristics of both, but with generally a guage.8 Fixed-line networks continue to be important much lighter touch than either telecom or TV regula- in the developed countries, but in the developing tion. Indeed, some would argue that internet regula- world, especially in Africa, mobile networks are now tion by government is neither helpful nor necessary, the main means of delivering services. As the uptake particularly if older models of regulation are applied.11 of ICTs grows around the world, the policy focus is But the trend is toward greater government control, shifting from solving supply-side challenges (basic not less—even in the United States, where the regu- access and affordability) to addressing demand-side latory body recently proposed to regulate the internet dilemmas (how to ensure that networks are open as a common carrier telecommunication service—to and safe).9 Policy decisions intended to facilitate the preserve the concept of net neutrality, or treating all take-up and safe use of ICT products and services will “bits” of information alike, irrespective of their con- shape the digital economy and wider developmental tent or value.12 As this chapter shows, all governments outcomes. have been obliged to consider which policies are most In broad terms, services in sectors regulated by appropriate for the internet, and their answers vary governments account for just under half of the US$4.2 widely (box 4.2). trillion in revenues generated by the global ICT sector (telecom services, TV services, and internet), with the rest (hardware, software, and computer services) Supply-side policies: largely unregulated. Services delivered directly over Availability, accessibility, the internet account only for just over 7.5 percent, and aff ordability but they are the fastest growing segment, having more than doubled since 2010 to reach an estimated The supply side of the internet is conditioned by US$309 billion in 2014.10 Separate regulations have rules on market competition, shaped by the respec- been traditionally applied to telecom and TV services, tive roles of the public and private sector, and medi- with the former focused on carriage (transmission) ated by the degree to which regulation of the sector SECTORAL POLICIES 205 is independent of government and the operators. A From these different market opening moves, a useful framework for analyzing supply-side policies policy consensus emerged around three basic ingre- is to consider the value chain that stretches from the dients—market competition, private participation, point where the internet enters a country (the fi rst and independent regulation of the ICT sector.21 In mile), and passes through that country (middle mile), developing countries, the main push toward market to reach the end user (last mile). As a general rule, growth came in the 1990s with the arrival of digital the market works best closest to the end user (last (second-generation, or 2G) mobile communication mile). Public-private partnership is more likely to be services, which allowed for competition, often for the needed in the fi rst and middle miles, or where cus- fi rst time. Worldwide, only a handful of countries, tomers live in areas that are diffi cult to serve. In addi- including Djibouti, Eritrea, and Ethiopia in the Horn tion to the visible elements of the network, certain of Africa region, still maintain state-run monopolies in hidden elements are vital to ensuring the integrity of the provision of mobile services and the internet, and the value chain—call them the “invisible mile” (table they have generally not fared as well as their neighbors. 4.1). (This framework is used a bit later in this chap- Mobile penetration is only half the level in the coun- ter.) The goals of supply-side policies differ between tries that have retained monopolies than in Kenya, countries, but the general objective is to ensure that which has had mobile competition since 2000, or in the internet is universally available, accessible, and Sudan, since 2005. The cover photo of this report shows affordable. migrants in Djibouti straining to receive cellphone sig- nals from Somalia, where competition in the telecom Market competition sector is fi erce and prices are much lower (box 4.3).22 A Today’s internet runs on yesterday’s legacy networks. study of Sub-Saharan Africa showed that telecom rev- In the OECD countries, and some others, the same enue averaged 5.6 percent of GDP in liberalized econo- copper networks built in the 1960s and 1970s to carry mies, but only 3.5 percent in nonliberalized ones, and voice telephone calls and cable TV (the last mile) was growing twice as fast in liberalized economies.23 have now been upgraded and repurposed to stream The recipe of competition, private participation, movies and social media, carried over IP-based net- and independent regulation worked so well for works. Farther away from the user (the middle mile), mobile telephony that, as of June 2015, there were the networks are more likely to be newer, and based more than 7.5 billion subscriptions worldwide—more on fi ber, but the wired access networks that connect than the human population.24 The same recipe should users were largely built in an era of government-run also largely work well for internet services, to extend monopolies. All but a handful of countries had state- access to the 4 billion or so people now without owned public telephone and telegraph companies affordable service. But there is a wrinkle. For network (known as PTTs) when they were building out their operators, extending internet access may cannibalize networks.13 revenues from existing voice and text services, at Outside the United States, which followed a least in the short term. This is one outcome of mar- different path of private investment, market com- ket entry by third-party content and service provid- petition in telecommunications began in the 1980s, ers that ride “over-the-top” (OTT) of the operator’s with new market entrants arriving in the United IP-based networks: Kingdom (1981),14 the Republic of Korea (1982),15 and Japan (1985).16 In the European Union (EU), a coordi- • For telephony, voice over internet protocol services, nated process was set in motion by a Green Paper on such as Skype and Viber, substitute paid voice calls telecommunications liberalization in 1987 and a Full with calls made “free of charge” over the internet.25 Competition Directive in 1996, setting a timetable for • For text (short message service, or SMS), which full liberalization of the telecommunications sector had been highly profi table for mobile operators,26 by January 1, 1998.17 The EU is now committed to instant messaging OTT services such as WhatsApp achieving a single digital market by 2020.18 Globally, and WeChat provide a more attractive substitute some 69 countries19 made commitments to liberalize for a fraction of the price. telecommunication markets and allow foreign invest- • For video (such as cable or satellite TV), streaming ment in their telecommunication sectors as part of services like Netfl ix and YouTube—which together the World Trade Organization (WTO) Basic Telecom- account for almost half the traffi c delivered to users munications Agreement in February 1997, when these in the United States27—offer low-priced content and countries accounted for some 93 percent of global give more freedom to consumers to watch content telecommunications revenue.20 when and where they like, on multiple devices. 206 WORLD DEVELOPMENT REPORT 2016

Table 4.1 A policy framework for the supply of internet service The last mile The invisible mile The fi rst mile The middle mile (the connection between (other, less visible (the point at which the (the national, intercity users and their network components internet enters internet backbone nearest internet point and potential a country) of a country) of presence) bottlenecks) Network International internet National backbone Local access network, Nonvisible network components access, including and intercity network, including local loop, components, including submarine cable landing including fi ber backbone, central offi ce exchanges, spectrum, border stations, satellite dishes, microwave, IXPs, local wireless masts crossings, databases, domain name registration hosting of content SIM cards, cybersecurity

Market • Authorization of satellite • Licensing/authorization • Licensing/authorization • Market mechanisms competition dishes of nationwide facilities- of local facilities-based (such as auctions and • Designation of domain based operators and operators and service resale) for spectrum name registry and service providers providers assignments, especially registrars • Interconnection • Authorization of for 3G and 4G bands • Licensing of competing arrangements mobile virtual network • Arrangements for access international service • Infrastructure sharing operators to essential network providers and orbital arrangements • Authorization of value- facilities, including slots • Cross-sectoral added network service national numbers, • Authorizations for participation (such as providers, including for address database landing stations, and cable TV and alternative mobile money access (co-location) to infrastructures) • Unbundling the local international gateway • Licensing mobile virtual loop facilities network operators

Public-private • / • Privatization of the • Dominantly private • Negotiation of transit partnership liberalization of incumbent operator operation and and access to virtual international gateway • Industry consultation on ownership, with PPP landing stations (for • Development of a network master plan approach where market landlocked countries) government data centers • Establishment of fails (as in rural areas) • CSIRTs at national and • Participation in national and local IXPs • Stakeholder consultation institutional levels international cable and • Local hosting of content, on a national • Open access to short satellite consortia including government broadband plan code numbers, as for • Regulation of legal data center • Universal service SMS intercept obligations (as for emergency services and accessibility for disabled)

Effective • Open access to • Coordinating • Open access rules for • Spectrum management, regulation international facilities rights-of-way for local loop and central including arrangements • Open to foreign linear infrastructures offi ce exchanges for allocation of bands ownership and • Safeguards on • Coordination of and refarming investment signifi cant market power planning permission for • SIM card registration • Avoiding excessive • Open access rules for public works among arrangements import and taxes national backbone operators and utilities, • Data protection and • National representation • Promotion of local and authorizations for privacy guidelines at relevant national and content and hosting construction of wireless regional bodies, such as masts ITU, ICANN, and WTO • e-waste recycling guidelines

Source: WDR 2016 team. Note: Policy examples are indicative, not exhaustive. Policy actions shaded in red are particularly suitable for emerging economies, in green for transitioning economies, and in blue for transforming economies, but all are good options to pursue at any stage (see chapter 5 for classifi cation of economies). 3G = third-generation; 4G = fourth-generation; CSIRTs = Computer Security Incident Response Teams; ICANN = Internet Corporation for Assigned Names and Numbers; IP = internet protocol; ITU = International Telecommunication Union; IXP = internet exchange point; PPP = public-private partnership; SIM = subscriber identifi cation module; SMS = short message service; WTO = World Trade Organization. SECTORAL POLICIES 207

Box 4.3 Fragile states, resilient digital economies

Somalia is emerging from 20 years of civil war, during which Somalis were paid via a mobile phone, compared with just time its fi xed telecom infrastructure was destroyed. Yet infor- 25 percent of Kenyans (fi gure B4.3.1, panel b), despite the mation and communication technologies (ICTs) are one of of M-Pesa, Safaricom’s mobile money service, the bright spots in its economy. With seven mobile operators there.b Some 55 percent of Somalis used mobile phones to and multiple internet service providers (ISPs), Somalia has a receive remittances; this has become indispensable recently higher rate of penetration than its monopolistic neighbors, as other fi nancial channels, such as Hawalas, have been despite the fact that it has the lowest gross domestic product blacklisted as part of a crackdown on their suspected links per capita in the East Africa region (fi gure B4.3.1, panel a). with terror. Furthermore, telecom operators contribute to Telecommunications can play an important role in post- the national treasury, unlike the banking sector, which con- confl ict reconstruction.a Not only does the sector generate tributed nothing in 2014. Somalia has benefi ted since early jobs, entrepreneurship opportunities, and foreign currency 2014 from a fast connection to the global internet, following from incoming phone calls, but using mobile phones to the completion of the EASSy undersea cable to Mogadishu, make payments or trade airtime also provides an attractive in which the International Finance Corporation is a stake- alternative to the local currency. In Somalia, few citizens holder. The Shabaab terrorist organization has succeeded have faith in the Somali shilling, and prefer U.S. dollars, but in blocking the use of mobile broadband in much of the dollars are frequently in short supply. The World Bank’s 2014 country, because of traceability, but it still uses social media Findex survey showed that some 38 percent of wage-earning for recruitment and spreading its message.

Figure B4.3.1 Somalia’s rising mobile economy

a. Mobile cellular subscriptions in the b. Mobile payments use, 2014 Horn of Africa, per 100 population, 2003–14 80 60 70 50 60 50 40 40 30 30 Percent Subscriptions 20 20 10 0 10 n.a. 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 0 Paid school Paid utility Received wages Kenya (4) Somalia (7) Ethiopia (1) fees using a bills using a through a Sudan (3) Djibouti (1) Eritrea (1) mobile phone mobile phone mobile phone Somalia Kenya Ethiopia Sudan

Sources: Adapted from ITU World Telecommunication/ICT Indicators and World Bank World Development Indicators, and World Bank Findex survey. Data at http://bit.do/WDR2016-FigB4_3_1. Note: Figures in parentheses in panel a indicate the number of facilities-based mobile operators. n.a. = not available.

a. Kelly and Souter 2014. b. http://go.worldbank.org/1F2V9ZK8C0.

Although network operators may be losing revenue demands of users, especially on mobile networks, from traditional voice, text, and video services, they where spectrum may be scarce. Cisco estimates that benefi t from the demand for data traffi c generated mobile data traffi c grew 69 percent in 2014 and will by OTT services. But they face a heightened invest- continue to grow at 61 percent a year between 2014 ment challenge to keep up with the service-quality and 2019.28 Mobile operators in Africa, for instance, 208 WORLD DEVELOPMENT REPORT 2016

are having to spend on average one-quarter of their countries, such as the Democratic Republic of Congo annual revenue on capital expenditure as they build or , may never get a fi xed-line access net- out their networks, while in North America, where work, even in urban areas. An optimistic assessment network buildout took place much earlier, the capital is that developing countries are leapfrogging a whole expenditure burden is just 15 percent of revenue.29 But stage of network development, but more realistically network operators in developing countries may fi nd they may be missing an essential stage. Available evi- it a challenge to persuade users with low incomes to dence suggests that access to the internet from big- pay for higher data use, which has a lower perceived screen devices (PCs), with always-on fl at-rate access, value than voice calls. Their users may employ OTT provides a bigger boost to economic activities than services primarily for reducing, rather than increas- access from small-screen devices (mobile phones), ing, their monthly phone credit use. According to which generally have use-based pricing.31 As long as reports from the Global System for Mobile commu- mobile operators dominate the provision of voice and nications Association (GSMA), while average revenue internet services in developing countries, compa- per user in North America has remained steady at nies whose business models are based on long-term above US$65 per month since 2010, it has fallen in investment in fi xed assets, without revenues from Africa from US$14 to below US$10.30 So, African net- mobile services, will fi nd it diffi cult to survive. That work operators are caught in a triple bind of rising is why some degree of government intervention, user expectations (due to higher data usage) requir- through public-private investment, may be necessary ing higher capital expenditures, but declining user in smaller developing countries to build up resilient spending (due to OTT services). international connectivity, and an open access back- For developing countries, the threat of diluting bone network, in order to retrofi t the missing stages traditional revenue streams is particularly worrying of network development. because they are following a trajectory in their net- work buildout different from the one most developed Public-private partnership countries follow. Figure B4.1.2 shows network build- After market competition, the next essential ingre- out in developed (OECD) and developing countries dient in the ICT policy recipe is private investment. (low- and middle-income). Both sets of countries Private companies have driven network investment show a similar pattern of mobile networks growing to throughout the world, especially in mobile networks. overtake fi xed-line networks in subscriber numbers, The biggest network operator of them all, China since 2001 and 2003, respectively, and fi xed-line sub- Mobile, which had over 800 million subscriptions at scriptions declining thereafter. But the big difference the end of 2014, is still majority state-owned.32 But it is that OECD countries had already achieved univer- is the partial exception, and its public shareholding sal fi xed-line access (roughly, more than 90 percent is listed on the stock exchanges of Hong Kong SAR, household penetration) before 2001. China and New York. Network operators dominate the In developing countries, the decline in fi xed-line ICT sector in revenues and customer connections. But access occurred before it had reached even one- stock markets appear to prize internet companies that quarter of the universal service level, and is now well sell content and OTT services more highly than tele- below that. This is signifi cant because wireless net- com operators that build networks. Alibaba, a Chinese works (using spectrum) are not fully substitutable e-commerce company that went public in September for fi xed networks (using copper or fi ber), either in 2014, is valued at more than US$200 billion—a level usage (which rarely offers fl at-rate pricing, without similar to China Mobile, although it has only one- data limits) or in performance (where speeds are tenth the network operator’s revenue.33 Low market generally lower). So, despite procompetitive policies valuations for network operators, and the cannibaliza- that have encouraged infrastructure competition and tion of their voice and text revenues, make it harder allowed for the development of OTT services, many for them to invest in network capacity, particularly in developing countries are stuck with a second-class fi ber backbone networks or fourth-generation (4G) internet that may fail to deliver the expected benefi ts, mobile networks. Private investment in telecommuni- especially for business users. cation networks in 2013 had fallen by almost one-third Indeed, most developing countries are unlikely since its peak in 2008, suggesting that operators are ever to attract, or generate, suffi cient investment to fi nding it harder to justify capital expenditure at a extend a nationwide backbone (the middle mile) or time when future revenues are uncertain.34 create fi xed-line networks in rural areas without some One response is to revive public investment in the kind of public-private partnership. Some developing internet backbone. In the United States, the internet SECTORAL POLICIES 209 benefi ted from public funding from the 1960s until role for the Internet Corporation for Assigned Names 1993, when the government began “privatizing” the and Numbers (ICANN) to the global multistakeholder internet by transitioning the role of the National community.36 In other countries, too, the government Science Foundation to the private sector, including took an early role in building out the national back- responsibility for registering domain names and bone infrastructure for the internet, as in the Republic managing network access points.35 This process is due of Korea (box 4.4). This role continues, for instance, in to be completed in September 2016, when the Depart- National Research and Education Networks (NRENs), ment of Commerce is expected to transfer its oversight where many governments subsidize the higher edu-

Box 4.4 How public-private partnership helped build the internet backbone in the Republic of Korea

Until other developed countries caught up around mid- Convergence Network that followed it from 2005 to 2014 2006, the Republic of Korea’s fi xed broadband penetration saw government investment of just under US$1 billion in each was well ahead of its competitors (fi gure B4.4.1 and table phase. Private investment dominated in the initial phase, as B4.4.1). In fi ber-based ultrafast broadband, it is still ahead the backbone network was established and larger cities were of the game. It has the highest percentage of fi ber among served. During this phase, government money was used fi xed broadband connections (66 percent) of any country mainly to purchase bandwidth for government’s own needs. except Japan. Korea also leads in the “internet of things” Since 2005, government spending has been proportionately and sensor technology (see spotlight 6). greater, as network investment reached out to rural areas, One secret to Korea’s success was a public-private where there was less incentive for the private sector to take partnership (PPP) that combined government funding the lead. The government of Korea has followed up with a and policy direction with private infrastructure investment program to upgrade network performance, through the Ultra and management. The Korea Information Infrastructure Broadband Convergence Network, with around one-third of program, which ran from 1995 to 2005, and the Broadband total investment coming from the government.

Figure B4.4.1 Broadband in the Republic of Korea and other selected economies

Broadband per 100 inhabitants, 2002 Q4 through 2014 Q2

45 40 35 30 25 Republic of 20 Korea’s early Percent lead in 15 broadband 10 5 0

2002 2003Q4 2003Q2 2004Q4 2004Q2 2005Q4 2005Q2 2006Q4 2006Q2 2007Q4 2007Q2 2008Q4 2008Q2 2009Q4 2009Q2 2010Q4 2010Q2 2011Q4 2011Q2 2012Q4 2012Q2 2013Q4 2013Q2 2014Q4 Q2 Year and quarter Korea, Rep. Japan Netherlands OECD average Norway Source: OECD Broadband Portal, http://www.oecd.org/sti/broadband/oecdbroadbandportal.htm. Data at http://bit.do/WDR2016-FigB4_4_1. Note: The gap between the Republic of Korea and the two nearest countries of the Netherlands and Japan in 2002–04 indicates Korea’s early lead in broadband. OECD = Organisation for Economic Co-operation and Development. (Box continues next page) 210 WORLD DEVELOPMENT REPORT 2016

Box 4.4 How public-private partnership helped build the internet backbone in the Republic of Korea (continued)

Table B4.4.1 Broadband investment program, Republic of Korea US$ million and percent of total

Information infrastructure, Broadband convergence Total, Investment 1995−2005 network, 2005−14 1995−2014 Government 806 (2.4%) 981 (38.0%) 1,787 (5.1%) Private 31,721 (97.5%) 1,599 (62.0%) 33,320 (94.9%) Total (US$ million) 32,527 2,580 35,107

Sources: Kim, Kelly, and Raja 2010; World Bank and Korean Development Institute 2015.

cation community, such as Internet2 in the United governments. But even in the best of circumstances, States, SURFNet in the Netherlands, and KENET in regulatory agencies may lack suffi cient high-caliber Kenya.37 Governments can also to staff, or they may be captured by commercial or gov- negotiate lower bandwidth rates for universities and ernmental interests. In the many countries where government departments, for example. governments still hold a stake in the incumbent oper- ator, true regulatory independence is rare. The worst Eff ective regulation case is where commercial and governmental interests After market competition and private participation, coincide, in what is sometimes called “crony capi- the third key policy ingredient is independent regu- talism.” In Tunisia during the Ben Ali regime, three lation: establishing ICT regulatory agencies that are telecom companies with ownership related directly to independent of leading operators and of government the Ben Ali family generated revenues of US$86.8 mil- departments. The International Telecommunica- lion, or 8.8 percent of total turnover from the sector in tion Union (ITU) records that some 159 of its 194 2010, but they accounted for some 42.7 percent of total member-states (82 percent) had separate ICT regu- gross profi t generated by the sector.41 High demand latory agencies by the end of 2013, a doubling since for scarce spectrum also creates incentives for cor- 1998.38 Effective regulation creates a level playing ruption. In India, the sale of 2G cellular licenses in fi eld for operators and helps promote market entry. 2007–08 was rigged to create an artifi cial scarcity Governance structures and mandates for regulators that favored some companies over others, ending up differ, but most have responsibility for spectrum in a scandal and a jailed minister.42 Regulators bear management and for issuing and overseeing licenses a responsibility to protect consumers against such for network operators. Most regulators are run on a vested interests. They must also work closely with cost-recovery basis, and some generate a signifi cant competition authorities to ensure that consolidation surplus from license fees and spectrum auctions.39 in the number of operators does not reduce true com- Telecom sector regulation should work primarily petition in the sector. in favor of the consumer—addressing , fostering effective competition, protecting consumer Where markets fail interests, and increasing access to technology and By almost every measure, the policy recipe for pro- services.40 In particular, the regulator should seek to moting the supply side of the ICT industry, based ensure that the benefi ts from technological change, on market competition, private investment, and greater effi ciency and reduced costs, are passed on independent regulation, has been tremendously to consumers rather than appropriated in higher effective in extending coverage, at least for mobile profi ts for private fi rms or extortionate taxes to communications. The estimated 7.5 billion mobile SECTORAL POLICIES 211 subscriptions in use worldwide generate revenues phones. Public telecom operators use geographical of US$1.13 trillion a year, as of May 2015.43 And some averaging of tariffs to offer the same price to users 95 percent of the world’s inhabitants live within range in both urban and rural areas. Historically, they have of a mobile signal,44 with two-thirds of them served also used profi ts from lucrative international and by a 3G signal with theoretical access to the internet if long-distance services to cross-subsidize loss-making they have a suitably equipped device.45 Even in Africa, local services. But with privatization and the shift to the region where mobile coverage is lowest, at just more competitive markets and IP-based networks, 88 percent in 2012, more-effi cient markets could this became unsustainable.50 close all but 4.4 percent of the remaining gap, without An alternative, followed by more than 70 countries, needing a cross-subsidy.46 is a Universal Service Fund (USF) to channel payments But that is not the whole story. Markets fail where by operators to fund infrastructure in rural areas or the private sector underinvests—for instance, because to provide access to libraries, schools, and hospitals. the private return may be less than the social return. Some USFs have performed well, notably those that This appears to be happening in the ICT sector in at use competitive mechanisms to distribute funds, least three areas. such as least-cost subsidies in Pakistan or reverse auctions in rural Chile.51 But USFs generate funds that • Remote areas. Although only less than 5 percent of often remain unspent or go for unintended purposes. the global population remains unserved with cellu- Unspent USFs amounted to more than US$11 billion in lar mobile coverage, that still represents almost half 2012, and in Côte d’Ivoire and Paraguay more than 0.6 a billion people worldwide. Even where it is techni- percent of GDP. In several countries, USFs continue to cally viable to serve these people, there may be little fund basic fi xed-line telephony long after user demand commercial incentive to do so. The costs of reaching had shifted to mobile and broadband internet.52 them are high in relation to the commercial return, By far the most successful measure to extend as they live mainly in rural locations with low pop- access is to license competing mobile service provid- ulation density, or in geographically remote areas ers and internet service providers. Governments often (box 4.5). But the social costs of remaining unserved include specifi c network rollout obligations in the are high and growing, so universal service policies license conditions of mobile operators, and these are may be required. frequently exceeded thanks to strong demand, at least • Unattractive markets. Competitive market entry may for basic mobile telephony. But for advanced mobile not take place even in markets that are nominally networks, suitable for carrying data services such as open to competition, especially in fragile or con- high-speed internet, rollout into areas of sparse popu- fl ict-affl icted states, or in small island developing lation density has been slower. For voice services, pro- states.47 A lack of scale and the failure to use com- vision of coverage to rural areas by private investors petitive tendering for infrastructure investment, could sometimes be justifi ed by the volume of incom- for instance, may also result in prices that are unaf- ing calls. But for data services, although the direction fordable for users. of traffi c might be similarly asymmetric, only the • Uneconomic services. More worrying, even if basic ability of the local population to pay for service justi- services can be delivered, delivering more advanced fi es investment in network upgrading, as there is no networks suitable for carrying data services, such payment for incoming traffi c. as high-speed internet, may not be economic. On Given the increasing importance of broadband mobile broadband networks, this requires third- for modern life, better provision in rural areas could (3G) or preferably fourth-generation (Long Term help revitalize the local economy in secondary towns Evolution, LTE) networks, which need higher levels and rural areas.53 Here are some possible solutions for of investment and generally a denser network of rural broadband: 54 base stations and masts.48 Some developing coun- tries have yet to launch mobile broadband. Even • USFs can be repurposed to focus on broadband. In the where services are launched, coverage can be low, United States, since a 2011 decision by the regula- reaching just 1 percent of the rural population in tory body, the Federal Communications Commis- Zambia or 11 percent in Namibia, for example.49 sion (FCC), universal service subsidies have been channeled to the Connect America Fund, to the To extend affordable access, governments have tune of around US$4.5 billion a year.55 A similar tried price controls, regulated prices, public facili- initiative in Europe, using regional development ties (such as telecenters), and mandated public pay funds, is the Connecting Europe Facility.56 Already, 212 WORLD DEVELOPMENT REPORT 2016

Box 4.5 The last (1,000) mile(s)

The “last mile” problem is a term commonly used in com- miles,” plagues remote rural communities. Specifi cally, munications and to describe the rela- what is the best way to bring the internet to areas “land- tively high cost of building the infrastructure to link to end locked,” far from the cable landing stations on the coast, users: the access network. Technological change, particu- or “sealocked,” small islands with small populations that larly the development of cellular mobile communications cannot justify an investment in undersea cable? and data compression techniques, has greatly reduced Where undersea fi ber-optic cable is available, it will infrastructure investment costs and is helping to solve the generally trump all other solutions in speed, performance, last mile problem, at least in urban areas. But a diff erent and cost. But for low-density, dispersed populations or problem, perhaps better described as the “last 1,000 remote islands, satellite off ers a quick and easy alternative.

Map B4.5.1 The eff ect of geography on internet prices, Africa

Ghana Malawia Mauritius Gabon South Africa São Tomé and Príncipe a Zimbabwe MAURITANIA MALIMAL Sudan NIGERG SSENEGALAL CHAD SUDANSUDA Mozambique 3,000 BURKINAURKINA FASOFAS Seychelles GUINEA-BISSAUGUINEAGUINEA-BISS DJIBOUTIJIBOUT BENIN NIGERIAGERIA 2,000 Kenya CÔTE ETHIOPIA Mauritania D’IVOIRED IVOIRE GHANAG A SOMALIAS MAALIAL A Somalia TOGOTOG CAMEROCAMEROON 1,000 EQUATORIALQUATORIAL GUINEA Côte d’Ivoire SÃO TOMÉ AND PRÍNCIPE KENYAYA 600 Senegal GABGABON Zambiaa 500 a TANZANIA Botswana SEYCHELLES Nigeria a COMOROS 400 Ethiopia ANGOLA MALAWIMALA Djibouti ZAMBIA 300 Angola MOZAMBIQUEMOZAMZ Q E Namibia ZIMBABWE MADAGASCARMADAGASCAAADAGAAGAGASCA MAURITIUS NAMIBIAAMIBIA Madagascar BOTSWANAOTSWANA 200 Coastal average 206.61 Tanzania 100 SOUTHSOUT LESOTHOLESOT Togo AFRICA

Comoros 0 Benin Not available Burkina Fasoa Landlocked average 438.82 Guinea-Bissau Lesothoa Nigera Malia Cameroon Chada

0 100 200 300 400 500 600 1,000 2,000 3,000 Price (per Mbit/s in US$ PPP)

IBRD 42012 Source: WDR 2016 team. Data at http://bit.do/WDR2016-MapB4_5_1. Note: Price per Mbit/s in US$ PPP a month in 2014 Q4/2015 Q1 for fi xed, residential broadband service. Mbit/s = megabits per second; PPP = purchasing power parity. a. Indicates landlocked country; the gradient indicates greater cost.

(Box continues next page) SECTORAL POLICIES 213

Box 4.5 The last (1,000) mile(s) (continued)

Usage costs have come down, with the entry of new (VSATs) serving its population of just over 10 million, who players, such as Avanti Communicationsa and O3B.b The pay expensive retail rather than wholesale rates for the Cook Islands, with fewer than 14,000 people spread over limited capacity provided. Geography matters. Landlocked 15 inhabited islands across 2.2 million square kilometers countries generally pay higher prices for bandwidth than of ocean, recently opted for a satellite network from O3B.c coastal countries. In Africa, for instance, being landlocked But satellite has the disadvantage of being more expensive adds an average of US$232 to the monthly price for fi xed than fi ber-optic cable per unit of data, and, with older broadband access (map B4.5.1). But history matters, too. generations of satellite, suff ers from higher latency (delay), In the Pacifi c, countries fi rst reached by cable, such as which makes them unsuitable for real-time uses like video Fiji, tend to have lower internet prices than those reached gaming. Even the Cook Islands are now considering an more recently (fi gure B4.5.1). Once served with fi ber, small undersea cable. island states may fi nd that they use only a tiny amount Although satellite off ers distributional advantages, it of the capacity available; Tonga, for instance, barely uses makes it hard to aggregate demand and therefore to nego- 10 percent. Moving to fl at-rate pricing (“all you can eat”) tiate lower bandwidth costs. South Sudan, for instance, for bandwidth may be the best way to recoup the initial has an estimated 3,000 very small aperture terminals investment.d

Figure B4.5.1 The eff ect of history on internet prices, Pacifi c

Price per 1 GB of mobile data (US$) Digicel 79 0.9 a Our Telekom 69 0 Vanuatu Telekom 64 0.9 a Solomon Islands beMobile 58 0 Samoa Blue Sky 22 5 Samoa Digicel 17 5 Tonga Digicel 13 1.5 Tonga TTC 13 1.5 Fiji Digicel 6 14.8 Fiji Vodafone 4 14.8 Years since arrival of undersea cable

Source: See http://www.theprif.org/index.php/news/53-media-releases/169-prif-ict-study. Data at http://bit.do /WDR2016-FigB4_5_1. Note: Price, per gigabyte (GB) of data in US$ a month in 2014, for prepaid mobile broadband. a. Cable to Solomon Islands under consideration.

a. http://www.avantiplc.com/. b. http://www.o3bnetworks.com/. c. http://www.islandsbusiness.com/2014/4/business-intelligence/cooks-opt-for-satellite-telecom-network/. d. PRIF 2015.

from 2007 to 2014, some €14.7 billion (US$16.5 bil- also share infrastructure from other sectors, such lion) had been committed to broadband networks as electricity or transport. Mutualization is slightly under PPP funding, using EU structural funds, in different, with a wholesale operator created to sell more than 100 different projects in the European only to other operators and not directly to users. Union.57 But few developing countries have this Increasingly, specialist wireless mast companies, level of resources to commit. like Indus or Reliance Infratel, two Indian com- • Infrastructure sharing and mutualization58 can also panies, are emerging for wireless infrastructure, reduce costs for operators. Infrastructure sharing while cable backbone companies include Botswana’s refers to the operators sharing one another’s net- BoFiNet and the Burundi Backbone System. Sharing work infrastructure, or at least some elements of assets can improve management effi ciency, though it, such as wireless masts or cable ducts. They may it may also lead to disputes. 214 WORLD DEVELOPMENT REPORT 2016

• Technological solutions are emerging that promise a fresh is a complex matter, involving contract rights with approach to rural broadband. From drones to balloons private parties and not administrative fi at. But the to nanosats, there is no shortage of inventive solu- natural shortage of attractive, or easy to remember, tions to providing wide area coverage. But these domain names creates opportunities for profi teering. new technologies will need to leverage the physical and market infrastructure that the cellular mobile ICT prices are falling . . . industry has built in order to become commercially The trend toward declining prices in the ICT sec- viable and achieve scale. And efforts to bring new tor has been a long-term and predictable driver of technology will have to be complemented by more growth. For hardware, it is encapsulated in the pre- effi cient use of spectrum, such as spread spectrum diction made by Gordon Moore, the cofounder of and digital dividend spectrum, which releases for Intel, in a 1965 paper that the number of transistors cellular commercial use the highly valuable spec- in an integrated circuit would double about every two trum (for instance, in the 700 MHz band previously years, with consequent improvement in price and used for terrestrial TV broadcasts), and in the “white performance.61 For memory storage, this means that, spaces” between digital channels. These spectrum in 2014, a typical price to store a gigabyte of data was bands have wider coverage and are therefore ideal just 3 U.S. cents, whereas 20 years earlier it was more for rural areas. than US$500.62 Similar rates of progress are observ- able in the unit price of computer processing power Managing spectrum and other scarce and in the availability and price of bandwidth63 (fi gure resources 4.1). Manufacturers have, to some extent, compen- Managing scarce resources—such as numbers, rights- sated for this by building greater functionality into of-way, and especially spectrum—presents regulatory devices for the same price. But as they have started to challenges. Policy makers are turning to market chase mass markets, device prices have also started to mechanisms, such as auctions, to deliver the best out- fall—since 2011 for smartphones, and earlier for older comes, and this can result in lower prices and higher technologies such as laptops and televisions. Smart- growth, as in Guatemala (see box 4.6). More fl exible phones, with more computing power than NASA had approaches to spectrum sharing between services, use at the time of the moonshots, can now be purchased of spectrum-hopping technologies, and refarming of for less than US$40 (although the typical cost is much spectrum will also help. But demand for bandwidth, higher), and it is forecast that, by 2020, 80 percent of and thus also for spectrum, continues to grow rapidly, adults around the world will own one.64 especially as video entertainment shifts from tele- Predictable, rapid price declines create an interest- vision sets to mobile devices. By 2020, around 2 GHz ing dynamic: it is possible to foresee, with reasonable of total spectrum will be needed in major markets accuracy, at what point services and devices will fl ip for cellular services.59 Today, most developing coun- from narrow to mass markets, as the price of owner- tries have only around 500 MHz allocated, and some ship and use falls. But there is a tendency to overesti- have less than 300 MHz.60 Spectrum availability and mate the effect of a technology in the short term and allocation is one of the factors determining the future to underestimate its effects in the long.65 This may wealth of nations, and governments will have a vital in part explain why the initial impact of the internet role in maximizing the benefi ts from this resource. caused a catastrophic swing in the markets, starting The internet brings new challenges for allocating around 1997 and peaking in March 2000, since known scarce resources, including domain names and the as the dot-com bubble. The aspirations of many of the transition to longer (IPv6) addresses. While spectrum startups of the time—like Broadcast.com, an internet and numbers are regulated primarily by the public radio company, or Pets.com, an e-commerce supplier— sector—internationally by the ITU and nationally by simply could not be met by the slow-speed dial-up government ministries or regulators—domain names internet access available, and their business models and IP addresses are controlled almost entirely by the were often unrealistic. The value of stock markets Internet Corporation for Assigned Names and Num- worldwide fell by some US$5 trillion in the 18 months bers (ICANN), a not-for-profi t corporation, that followed.66 But that period of creative destruction and the entities to which it subcontracts. The man- also saw the birth of many of the giants that dominate agement of top-level domain names, such as .org and the internet today, including Google and Tencent .com, IP addresses, and even country code domains, (both founded in 1998) and Alibaba (in 1999). such as .za (South Africa), generally is not the prov- Perhaps the most sustained example of how enance of government regulation. Their regulation falling prices drive market expansion comes with SECTORAL POLICIES 215

Box 4.6 Guatemala: An early pioneer of spectrum auctions

In most countries, the management of the civil radio fre- newly created regulatory agencies. But growing demand quency spectrum was carried out on a “fi rst come, fi rst for using the airwaves for providing mobile communication served” basis, with incumbent operators taking the lead. As services meant that supply soon exceeded demand. countries separated the functions of operator and regulator, Guatemala was one of the fi rst countries to respond to starting in the 1980s, spectrum management moved toward the changing dynamics of the marketplace, and in its 1996

Map B4.6.1 Spectrum assignment in Latin America, in MHz blocks

BAHAMAS, THE

CUBA DOMINICAN REPUBLIC MEXICO HAITI BELIZE ANTIGUA AND BARBUDA JAMAICA ST. KITTS AND NEVIS HONDURAS DOMINICA ST. LUCIA BARBADOS GUATEMALA NICARAGUA GRENADA EL SALVADOR ST. VINCENT AND THE GRENADINES PANAMA TRINIDAD AND TOBAGO COSTA RICA VENEZUELA, RB GUYANA COLOMBIA

ECUADOR

PERU BRAZIL

BOLIVIA

PARAGUAY Spectrum assignment, in block MHz 200–214 226–255 URUGUAY CHILE 260–288 ARGENTINA 295–320 Above 330 No data

IBRD 41705 Source: Adapted from GSMA (unpublished spectrum database). Data at http://bit.do/WDR2016-MapB4_6_1.

(Box continues next page) 216 WORLD DEVELOPMENT REPORT 2016

entry costs, and extensive use of discounts. Six of the Box 4.6 Guatemala: An early ten cheapest countries to use a mobile phone are in pioneer of spectrum auctions South Asia, where the cost of ownership is typically (continued) below US$5 a month (fi gure 4.2). Seven of the top ten most expensive countries are in Western Europe. On affordability, measured by the monthly price of a bas- Figure B4.6.1 How greater spectrum ket of services as a percentage of income, developed availability led to lower prices in countries still do better, as might be expected, though Latin America, 2003–09 the gap is narrowing. Costa Rica 2003 0.80 . . . but ICT prices still vary widely Why do prices for ICT services vary so widely, by a factor of more than 40 for mobile cellular prices? 0.60 Economies of scale seem to matter, with both India and China having prices below US$5 a month, while 0.40 small island states, such as Vanuatu or the Marshall Panama Nicaragua Islands, are generally more expensive than the global average, although there are many exceptions. But 0.20 El Salvador geography and population density appear to matter 2009 Guatemala less than might be expected, with mountainous Average price per minute (US$) 0 Bhutan and Nepal among the cheapest countries for 50 100 150 200 mobile service, and the fl at and densely populated Available spectrum (MHz) Netherlands one of the more expensive. The practice Source: Adapted from Marino García 2015b. Data at http://bit.do of geographical averaging of pricing, applying the /WDR2016-FigB4_6_1. same price throughout a country, is still the norm— Note: The start of each arrow is a 2003 data point, and the end of each arrow is a 2009 data point. suggesting that, once basic coverage is established, rural areas are not necessarily more expensive to General Telecommunications Law, recognized the serve than urban ones. Mobile prices appear to be economic value of the spectrum.a The law gave opera- demand-driven rather than cost-based, with some of tors the right to request that any unallocated portions the cheapest prices in countries where the ability to of spectrum be off ered for sale, through auctions, and pay is lowest. once awarded, it gave operators the right to use, lease, Instead, to understand differences in ICT prices, it and resell the spectrum, creating a tradable market. is necessary to look at policy and regulatory explana- Guatemala benefi ted from a faster reduction in prices tions. For mobile cellular services in Western Europe, for mobile services than the rest of the region (see fi g- larger operators found it profi table to set high rates ure B4.6.1). Other countries adopted similar policies, for terminating each other’s calls, and especially those following Guatemala’s early lead,b and indeed have from fi xed-line operators. Higher prices for off-net now overtaken Guatemala in the spectrum allocated calls locked in users, while high roaming charges to mobile communications (see map B4.6.1). drove profi ts. Regulatory interventions, to oblige operators to reduce termination and roaming rates, a. Ibarguen 2003. b. IDRC 2010. were only partially effective because operators simply absorbed lower interconnection payments by raising prices for outgoing calls in a “waterbed” effect.68 In mobile phone services. When the dot-com bubble Canada and the United States, the unusual system burst in March 2000, there were fewer than 700 mil- of “both parties pay” pricing, where users pay both lion mobile phone subscriptions (SIM cards)—but to receive and make calls with their mobile phones, now, there are more than 7 billion, with over three- should in theory obviate the problem of high mobile quarters in developing countries. That success story termination rates, in that “sender keeps all” is used has come about, in part, because of the falling price of (there are no payments between operators), which ownership and usage of mobile phones, particularly means there are no interconnection payments.69 in South Asia, where it has been labelled the “budget But in practice, both appear among the most telecom model,”67 based on prepaid billing, very low expensive countries for mobile calls, based on the SECTORAL POLICIES 217

Figure 4.1 Prices are falling for computer processing, storage, bandwidth, and smartphones

a. Global storage cost trend, 1992–2012 b. Global bandwidth cost trend, 1999–2013 1,000 10,000 US$569

100 US$1,245 1,000 Storage cost, –38% per year Bandwidth cost, –27% per year 10 100 1 US$16.32 US$ per gigabyte US$ per gigabyte 10 0.10

US$0.02 0.01 1 1999 2001 2003 2005 2007 2009 2011 2013 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

c. Global computing cost trend, 1992–2012 d. Global smartphone cost trend, 2008–2013

1,000 450 US$222 US$430 Smartphone cost, –5% per year 100 Computing cost, –33% per year 400 10

1 350 0.10 US$335 US$ per 1 million transistors US$0.05 US$ per typical smartphone 0.01 300 2008 2009 2010 2011 2012 2013 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

Source: Deloitte Shift Index 2013; see http://www2.deloitte.com/us/en/pages/center-for-the-edge/topics/deloitte-shift-index-series.html. Data at http://bit.do/WDR2016-Fig4_1.

OECD low-user basket. Excessive bundling of dif- a minimum speed of 256 kbit/s and monthly use of ferent services, notably in North, South, and Central 1 gigabyte, GB) and the Central African Republic America, keeps prices for individual services high, (US$584.97). For mobile broadband too, the range is and “locked” SIM (subscriber identifi cation module) more than 100-fold between Pakistan (US$1.48 for a cards limit consumers’ ability to choose among oper- GB of data a month, downloaded to a mobile handset) ators. Again, it is hard to escape the conclusion that and São Tomé and Príncipe (US$169.38).71 In contrast operators are using demand-based, rather than cost- to mobile voice services, European countries, particu- based, pricing, as operators in these countries have larly in Scandinavia and Eastern Europe, are generally largely saturated markets. One more positive devel- among the cheapest in the world, thanks to vibrant opment is the dropping of mobile roaming charges in competition (see map B4.1.1). Mobile broadband in East Africa in October 2014, which led, for instance, Europe does not suffer from the termination rates to a 950-percent increase in traffi c from Rwanda to that keep prices high for voice calls and SMS. Kenya compared with the previous month.70 Price differences between countries are even more What policy and regulatory options are evident for the internet. The range between countries available? for fi xed-line broadband is almost 200-fold between What can regulators do to address the shortfalls in Vietnam ($2.93 a month, for an entry-level price with availability and the wide price differences for internet 218 WORLD DEVELOPMENT REPORT 2016

Figure 4.2 If you want to make a mobile phone call, go to Sri Lanka

a. Monthly price for a mobile b. Monthly price for a mobile cellular basket, cellular basket, US$, 2013 as a % of GNI per capita, 2008–13

Brazil 14 Ireland United Kingdom 12 Netherlands Greece France 10 Spain United States 8 Israel Italy 6 Average

Pakistan 4

Vietnam % of GNI per capita Ethiopia 2 India Uzbekistan Bhutan 0 Nepal 2008 2009 2010 2011 2012 2013 Iran, Islamic Rep. Bangladesh Developing countries Sri Lanka World Developed countries 0102030 40 50 U.S. dollars

Source: ITU 2014. Data at http://bit.do/WDR2016-Fig4_2. Note: The basket of services used is based on the OECD low-user basket, which includes 30 outgoing calls a month (on and off -net, peak and off -peak), plus 100 SMS messages. Prices were sampled in the fourth quarter of 2013. Panel a is based on 167 economies, and panel b on a simple unweighted average for 140 countries with a complete data set. GNI = gross national income; OECD = Organisation for Economic Co-operation and Development; SMS = short message service.

prices worldwide? Table 4.1 offers some suggestions, is very wide—between 6.4 Mbit/s in Luxembourg based on applying the principles of market compe- to just 146 bit/s in the Central African Republic.74 tition, public-private partnership, and independent Governments can encourage a higher availability regulation to the different phases of the internet of international bandwidth, for instance, by liberal- supply value chain. The fi rst step is benchmarking, izing the market for satellite dishes, including very to gather up-to-date price data to allow for compari- small aperture terminals, and allowing open access sons, both within the country (between operators and to submarine cables, satellite consortia, and landing over time) and between countries, using appropriate stations. Government procurement can also lead comparators. Price comparisons are tricky and gen- in bringing down prices by aggregating demand erally require using a predefi ned basket of services among different services—all the universities in a and methodology, such as those used by the OECD72 region, for example—to negotiate better prices for or ITU.73 Operators often resist price comparisons, bandwidth with international providers. so it is essential to give the ICT regulatory agency a • The international gateway is the point at which a legal mandate for collecting, and publishing, relevant country connects to the global internet, and this pricing and quality of service data (box 4.7). can often become a bottleneck. Although the Arab Armed with data, the next step is to work out in Republic of Egypt liberalized its market for internet which part of the value chain for the supply of inter- service providers in 1996, the incumbent, Telecom net the market may be failing. Table 4.1 suggests sev- Egypt through its subsidiary TE Data, still has a legal eral distinct submarkets: monopoly over the international gateway and cable The fi rst mile. This refers to the point at which the landing station, in Alexandria, at least until June internet enters a country. 2016. Competitors pay as much for interconnection in Alexandria and backhaul to Cairo as they pay for • International connectivity. The connection between a international connectivity, resulting in some of the country and the global internet can be measured in highest prices for internet in the region.75 In such bits per second per internet user. The global range cases, regulating wholesale prices for connectivity, SECTORAL POLICIES 219

Box 4.7 How better ICT data can lead to cheaper services

Better data on how information and communication tech- on ICT use, but in developing countries they often lack the nologies (ICTs) are used in developing countries can spur resources to carry out household surveys or do not have policy changes that benefi t the poor and increase their dig- the know-how to collect policy-relevant data. A few donor ital access. In some parts of the world, one of the best ways agencies provide support, such as the funding of Research to spur policy reform is to shame governments into action ICT Africa’s household surveys by Canada’s International by comparing them to countries with better records. When Development Research Centre (IDRC)b or the funding of a report from Research ICT Africa revealed that prepaid LIRNEasia’s research on mobile phone usage at the base mobile prices were higher in South Africa than in 34 other of the pyramid by the United Kingdom’s Department for African countries, pressure from the Parliament obliged the International Development (DFID).c The latter helped to regulator to cut mobile termination rates, resulting in a wave persuade the government of Sri Lanka to forgo new mobile of price cuts by mobile operators.a But such examples are taxation policy.d But more should be done. Key areas for rare, and there is an alarming lack of accurate and timely future action include: data on ICT use in the developing world. This makes it hard to understand the mechanisms by which greater informa- • At the national level, ensuring that service licenses require tion access among the poor drives economic opportunity. It operators to report data to regulators, and encouraging also leaves important policy questions unanswered, such as National Statistical Offi ces to include questions in their whether new “free” internet services (or “zero-rated” ser- household surveys regarding ICT access and use vices like “Facebook zero”) drive the take-up of broadband • At the global level, reigniting global coordination or whether they deny access to the “free and open” internet. bodies, such as the Partnership for Measuring ICT for In general, supply-side indicators (such as subscriptions, Development,e to compile, coordinate, and improve ICT or domain name registrations) are better reported than data gathering demand-side use and applications data. The International • Exploring the use of new data sources such as “big data,” Telecommunication Union has collected such data from its crowdsourced data, and social media to complement ICT membership since the 19th century. But gaps in its database access and usage statistics are increasing, as operators in more competitive environ- • Creating mechanisms for telecommunications compa- ments have become more wary of releasing commercially nies to share data, such as anonymized call records, for sensitive data. National Statistical Offi ces could collect data public research, policy, and planning purposes.

a. http://www.researchictafrica.net/docs/Fair_Mobile_Prices%20Q2-v04.pdf and http://www.researchictafrica.net/home_archive_reader .php?aid=118. b. http://www.researchictafrica.net/home_archive_reader.php?aid=128. c. http://lirneasia.net/projects/2010-12-research-program/teleusebop4/. d. http://lirneasia.net/2007/10/lirneasias-policy-infl uence-on-mobile-tax -issue-in-sri-lanka/. e. http://lirneasia.net/2007/10/lirneasias-policy-infl uence-on-mobile-tax -issue-in-sri-lanka/.

Contributed by Laurent Elder.

or structurally separating the supply of wholesale to liberalize the market for building and operating from retail services, could help. backbone networks, and to encourage open access to the incumbent’s network at the wholesale level. The The middle mile. This refers to the national, intercity risk is that the most popular routes—say, between internet backbone of a country. the two main cities—are “superserved,” while the rest of the country is underserved. So, many • The national backbone network. The internet back- governments favor public-private partnerships to bone network in a country provides backhaul from “direct” private investment, as in the Republic of cable stations or satellite stations to major cities and Korea (see box 4.4). Governments can also help by towns. Ideally, it should include fi ber-optic cables, making rights-of-way available to investors—and but microwave and even copper links can also be by requiring that all major infrastructure programs used. The simplest step a government can take is (such as roads, railways, pipelines, and energy 220 WORLD DEVELOPMENT REPORT 2016

distribution) include provision for an optical fi ber while cable modems offered the same facility for link,76 and follow the principle of “dig once, dig cable TV networks. In developing countries, where smart,” by obliging contractors to anticipate future copper-based local access networks serve only a demand for fi ber when digging trenches.77 few areas, wireless-based access networks offer the • The internet exchange point (IXP). The IXP is where most popular alternatives. Just over half the world’s IP-based traffi c is exchanged within a country. As inhabitants live within coverage of advanced, third- of May 2015, there were some 446 IXPs around the (3G) or fourth-generation (4G) networks, but even world, but 87 countries still lacked a single one.78 second-generation (2G) networks can be adapted Average traffi c handled by each exchange ranges for slow-speed internet use. from some 4.47 terabits a second on the Amsterdam • Government policies to encourage the spread of fi xed-line Internet Exchange, founded in 1997, to just 21 kbit/s broadband networks include permitting competing on the BurundiX IXP, founded in 2014. Setting up facilities, especially for intermodal competition an IXP is one of the most cost-effective steps a (between cable, DSL, and wireless), and local loop country can take to enhance local connectivity. unbundling (LLU), or mandating the incumbent Research by the Internet Society shows that the to make local access lines available to competitors Kenya IXP (KIXP), operating since November 2001 at wholesale prices. The widespread enforcement despite being initially declared illegal by the regu- of LLU in Europe since 2000, following EU direc- lator, saves local internet service providers (ISPs) tives,81 is one reason its average broadband prices some US$1.5 million a year in reduced costs for are among the lowest in the world. LLU is most international connectivity. And it reduces latency effective once a minimum level of penetration has from 200–600 milliseconds to 2–10 milliseconds, been achieved; below that, it can act as a deterrent on average, by exchanging traffi c locally.79 for investment.82 LLU also offers a foothold for new • Local hosting of content. Promoting local hosting of entrants that can quickly offer nationwide service content and creating a local cache for frequently and then later develop their own infrastructure, used content from elsewhere can also enhance as has been the case for free.fr, a new entrant in the effi ciency of the network and reduce latency, France. increasing usage as users experienced shorter wait- ing times for websites to load. In Rwanda, 14 of the The invisible mile. This refers to the other, less vis- top 20 sites with Rwandese content were commer- ible network components and potential bottlenecks. cial sites, and all were hosted outside the country so that a typical website could save around US$100 • Effi cient spectrum management. Critical at all stages a year and enjoyed a more reliable service. But this of the internet supply chain but especially in the imposed additional costs of US$13,500 a year for local access network, effi cient management of the local ISPs that had to bring the content back in civil radio frequency spectrum includes increasing over expensive international links—costs which are the amount of spectrum available, ensuring com- then passed on to users. Promoting local hosting, petitive access, encouraging sharing of essential already required for government sites in Rwanda, facilities such as radio masts, and liberalizing the could avoid this, and help improve service quality.80 market for spectrum resale and leasing to allow the creation of mobile virtual network operators. In The last mile. This refers to the connection between this sense, spectrum resale in the wireless world is the user and their nearest internet point of presence the equivalent of LLU in the fi xed-line world. Policy (POP). makers can also help by making more unlicensed spectrum available, especially for innovative uses • The local access network. The most costly part of the such as cognitive radio (which hops between network, and the hardest to duplicate, is the local frequency bands to avoid interference), and by access network, which connects the user to the opening up underused government spectrum for nearest internet POP. In the early days, this was commercial applications. typically achieved through dial-up, using a modem, • Over-the-top services. Policies to encourage a wide over ordinary copper telephone lines. Starting in range and diversity of OTT services can provide the late 1990s, a technology called digital subscriber a wider choice to consumers at lower cost. Poli- line (DSL) allowed that same telephone network cies to encourage OTT growth include removing to be used for always-on broadband connections, regulatory barriers to providing voice over IP and SECTORAL POLICIES 221

mandating that operators provide access to essen- Innovation also depends on the ability to protect tial facilities such as billing and app stores. Of and monetize intellectual property. Avoiding piracy course, OTT services introduce many regulatory will rely on striking the right balance between provid- issues of their own, not least the concept of “net ing access to information and protecting intellectual neutrality.” A recent consultation on net neutrality83 property rights. Generally, the illegal distribution of by the Federal Communications Commission, the copyright material tends to lessen when legitimate U.S. regulatory agency, attracted a record number ways exist to access that material, for a fair price. of comments, around 4 million.84 Ultimately, with Thus, contrary to expectations, peer-to-peer fi le shar- support for net neutrality from the U.S. president, ing systems that blossomed in the early 2000s for free the FCC ruled in February 2015 in favor of regulat- sharing of music, such as Napster and Kazaa, started ing broadband internet as a public utility, although to decline once legal alternatives, such as iTunes or legal challenges could continue for years.85 Google Play, became available, without any noticeable • Mobile money. For developing countries, one of the impact on the quantity or quality of artistic output.90 most signifi cant OTT services is mobile money, The action has now shifted to fi lm and TV fi le shar- which is also fraught with regulatory challenges ing, but again the availability of legitimate commer- (see spotlight 2, “Digital fi nance”). cial services, such as Netfl ix or Hulu, is progressively displacing illegal services as they are extended to more territories and have greater content coverage. Demand-side policies: But there remain large parts of the world where Open and safe internet use content is not available for download legally, because content licensing and copyright are so complex and The internet is at once unique, complex, and one of fragmented that small markets are too often ignored the most used global communications media. It has (chapter 6).91 different layers of infrastructure and applications, An issue that is more diffi cult to resolve surrounds and different stakeholders are involved in its opera- the restriction, or fi ltering, of certain types of online tion, use, and governance. What features affect the content. Every country has different redlines on con- stability and security of the global internet to engen- tent it regards as dangerous or offensive. Germany has der trust and therefore encourage use of the internet? restrictions on hate speech, while Thailand monitors What is the best way to balance stakeholder interests comments about its king. There are also certain abso- in these areas? This section addresses the creation of lute values, such as restrictions on child pornography. an enabling environment of “trust” for the internet to Such restrictions are understandable and justifi ed achieve its full potential. where they represent a societal consensus adopted by an accountable government. The Supreme Court Censorship and content fi ltering threaten of India, for example, recently reconsidered its posi- the internet’s utility as an engine of growth tion on free speech issues, potentially lifting existing The internet thrives on open information exchange restrictions on online speech.92 But few governments and freedom of expression.86 One global view shows have the capability to enforce such content moni- the range of internet fi ltering of political speech, from toring on their own. Instead, they must cooperate none to pervasive fi ltering (map 4.1).87 Another shows, with the major websites and search engines, which perhaps paradoxically, that concerns over freedom of then have to apply their own judgment. Google, for expression are actually greater among users in coun- instance, publishes a transparency report that pro- tries that have more recently come online than among vides statistics on government requests to remove users from countries with a longer history of internet content, which have grown signifi cantly since 2011.93 access.88 One account of fi ltering blames fi ltering In the six months ending December 31, 2013, Turkey policies and software for poor internet performance submitted the most removal requests (895), followed (slow access speeds).89 This is an example of a direct by the United States (481). cost associated with fi ltering, but such fi ltering also Some governments try to block access to content leads to indirect costs from the drag on innovation directly, for instance by imposing national fi rewalls that comes from self-censorship, and the resulting or restricting the use of certain internet applications, loss of freedom. In this discussion, “fi ltering” (or cen- such as virtual private networks or voice over IP ser- sorship of public content) does not include “surveil- vices. Google reports that, in the fi rst nine months lance” (or monitoring of private content). of 2015, six countries had experienced disruptions of 222 WORLD DEVELOPMENT REPORT 2016

Map 4.1 Evidence of internet content fi ltering

Europe Inset Caribbean Inset Level of political content filtering No evidence Selective Substantial Pervasive No data available

IBRD 41702 Source: Open Network Initiative, http://map.opennet.net/fi ltering-pol.html. Data at http://bit.do/WDR2016-Map4_1.

traffi c to its products—China, the Democratic Repub- includes “cybercrime.” The more common threats lic of Congo, the Islamic Republic of Iran, Pakistan, to cybersecurity are , denial of service, and Tajikistan, and Turkey.94 Excessive restrictions on use phishing attacks (attempts to acquire sensitive infor- can increase the costs of doing business, particularly mation online by someone who is masquerading as for international businesses, which may be dissuaded a trusted entity), but cyberincidents are increasingly from investing.95 In addition, inconsistent and com- perpetrated by disaffected insiders. So, cybersecurity plex content fi ltering requirements can make it dif- usually refers to securing data and infrastructure in a fi cult to offer internet services and applications that civilian context; but acts that might previously have are global in nature. Policy makers in governments been considered civilian attacks are now being uncov- that impose such restrictions have to choose between ered as acts of states against states via nonstate actor the relative perceived advantages of controlling cer- proxies, blurring the lines between acts of cybercrime tain kinds of information and the economic costs, and cyberwar or cyberterrorism. Threats to personal including lost opportunities, of maintaining such security online, such as theft, are systems. also growing. This may erode public confi dence in e-commerce and e-government applications, and Cybersecurity: Trust in the internet make internet use less attractive, thus suppressing its will come from balancing the security use for certain applications. Likewise, cybersecurity of networks and information with the can include securing critical information infrastruc- protection of individual rights ture from acts of nature, such as developing backup The term “cybersecurity” is a convenient shorthand facilities in alternative locations. for a very complex set of issues. It commonly refers However defi ned, weak cybersecurity is a sig- to systems and actions aimed at securing data and nifi cant problem worldwide, with both the scale of communications over the internet and even the infra- fi nancial losses, and the costs of preventing them, structure of the internet itself.96 It also sometimes growing (box 4.8). New breaches of security over SECTORAL POLICIES 223 the internet are reported almost weekly. Reliable estimates of the size of the problem are hard to come Box 4.8 The costs of cybercrime by because of the defi nitional problem that com- pounds a lack of common indicators—and because Estimates of the costs of cybercrime abound, but fi rms and governments suffering cyberincidents are many reports are based on weak evidence or overly unwilling to disclose losses and thus to reveal vul- simplifi ed assumptions. Often the methodology is nerabilities. Some of the published sources are based not disclosed, complicating an assessment of its on fl awed assumptions, and all reports raise as many validity. Damage, typically assessed at a highly aggre- questions as they answer. The growing centrality of gated level, is diffi cult to link to specifi c incidents. the internet in daily lives around the world, and the Furthermore, most estimates are developed by com- increasing use of mobile devices, combined with the panies directly involved in the sector, which may have “internet of things” (see spotlight 6), appear to have an interest to overestimate the risks. Nevertheless, the increased exposure to cyberrisks. cost estimates are both high and growing: Equally important, the incentives to deal with cyberrisks need realignment. For example, individ- • A 2014 study put the global costs of cybercrime uals sometimes do not take proper precautions to at between US$375 billion and $US575 billion, or secure their devices or data, passing on the cost of about 0.6 percent of global GDP.a security to society at large. Vendors of hardware and • A 2014 study showed the average per person cost of software are in a highly competitive environment data breaches ranged from US$51 in India to US$201 and may push products to market at the expense of in the United States, and had risen 15 percent in a ensuring proper security features. Cybersecurity is year.b expensive and complex (whether through incurring • A 2013 study estimated that the global costs to up-front costs of “prevention” or dealing with costs consumers (excluding businesses) were around of restoring security in “recovery and resilience” US$113 billion, and had risen by half in one year.c regimes). It may be economically rational to accept • A 2008 study estimated that the global eff ects of some degree of insecurity for the convenience to malware were around 0.5 percent of global GDP.d transact online.97 That said, there are also various ways of identify- Source: Adapted from Bauer and Dutton 2015, for the WDR 2016. ing the costs involved, including direct and indirect a. CSIS and McAfee 2014. b. Ponemon Institute 2014. costs, and two types of costs to be considered. First are c. Symantec 2013. the costs of the breach (actual loss) and of the remedi- d. Bauer and others 2008. ation efforts to fi x it. Second are costs associated with prevention and providing an environment of “trust,” as well as understanding the incentives of actors to provide security. The methodology encapsulated in both developed and developing economies are taking table 4.2 provides a systematic framework for more action at the national level to address cybersecurity precisely evaluating “cybercosts” and identifying concerns. But because of the global nature of the which stakeholders are most likely affected by partic- internet and the cross-border nature of cyberinci- ular costs. Aggregating the cumulative costs across all dents, governments should be encouraged to do more players and cost categories yields an estimate of the to protect themselves and their citizens through total direct, the total indirect, and the total implicit cooperation at the international level, for instance by costs. This kind of assessment will be key for policy exchanging information on threats. Because of the planning. role of nonstate actors in the provision of infrastruc- Public safety and security in the analog world is a ture and services, governmental efforts will have to public good, ensured by governments. In the cyber- involve public-private partnerships (working with a world, governments also have an obligation, through variety of nonstate actors beyond the private sector) their policies, laws, and institutions, to ensure the and fi nd ways of addressing and even overcoming protection of data, communications, and critical jurisdictional boundaries and barriers. Technical infrastructure. One particularly vexing problem in solutions will help, as in combating spam, but these government attempts to address cybersecurity is that need to be backed by legal measures and enforce- much of the infrastructure, and most of the commu- ment. Collaboration and openness are key. nications, are controlled by the private sector or other In the areas of cybersecurity, there are few obvi- nonstate actors. Around the world, governments of ous policy recommendations, and in these areas— 224 WORLD DEVELOPMENT REPORT 2016

Table 4.2 A basic framework for assessing the costs of cybersecurity incidents

Direct costs Indirect costs

Market players Repair costsData loss Security measuresFraud Patch developmentCustomer and supportAbuse deployment managementInvestigationLaw at theenforcement organizationalCost of infrastructure levelLost productivityRevenue lossConfidentialityReputation-related breach Security revenue countermeasuresSlower loss ICT adoptionSlower collateral ICT impactinnovation End users

Home

Business

e-commerce companies

Infrastructure

Software vendors

Internet service providers

Hosting providers

Registrars Computer emergency response teams Law enforcement

Society at large

Source: Bauer and Dutton 2015, for the WDR 2016, citing Bauer and others 2008. Note: ICT = information and communication technology.

perhaps more than others—governments can play a Protecting personal privacy and data role in developing effective policies. The “perimeter online is essential in building trust in the security” that pervades today, born in an internet era of a few centralized mainframe computers, needs The other side of the “balance” is protecting privacy revisiting, putting users—not devices—at the center and data online.99 One of the key drivers in the digital of the discussion, and thus implying a great role for economy is the fl ow of personal data, of which an .98 Public policy could be used to estimated 90 percent has been added in the last two change misaligned incentives to achieve security. years. Collecting and analyzing data about individu- Given the pace of technological change, these policies als is integral to how some of the largest companies should be guidelines or principles, not prescriptive, in the world do business. For example, Facebook is a and be technology neutral. They should encourage company with a US$230 billion capitalization largely interoperability among regimes and legal systems through its sales of ads that reach Facebook users.100 to investigate and prosecute cybercriminals across But data also allow small and medium companies to jurisdictions and to avoid the creation of safe havens monetize their services. for cybercriminals. Finally, work could be done to The World Economic Forum (WEF) has identi- develop a common set of indicators to measure the fi ed data-driven enterprises as part of a strategy for economic impact of cybercrime and other cyberrisks. economic development in developing countries.101 SECTORAL POLICIES 225

Protecting personal data online is key for the data- Like big data, the internet of things also poses driven economy, since it will increase trust in the new privacy challenges, notably in the ability it internet, and greater trust will foster more use. And offers to develop detailed user profi les.111 The number privacy is not just a developed-country issue. Some of connected devices in use worldwide is set to rise studies show how concerns over maintaining privacy signifi cantly as consumer goods companies, auto online have more to do with “awareness” of threats to manufacturers, health care providers, and other busi- privacy posed by the internet, rather than with relative nesses continue to invest in connected devices.112 Data economic development, debunking the notion that collected are generally anonymized, but users may be privacy is mainly a western preoccupation.102 But the unwittingly relinquishing control of their personal issue is not as simple as having and enforcing national data. Open data initiatives by governments raise sim- laws that protect personal data. Data fl ows nowadays ilar concerns about “anonymization,” data breaches, are global, and privacy regimes need to be interopera- and unauthorized or unintended re-use. ble with one another to really enable the internet to be Privacy concerns need to be balanced against an engine of innovation and economic growth. other important issues such as transparency, freedom Privacy concerns range from the personally sensi- of expression, proportionality, and security.113 There is tive (personal health information or precise locational tension between protecting the privacy of individuals information) to the seemingly trivial (search or brows- and keeping them safe from terrorism or criminals. ing history). But even the most innocuous data can Law enforcement and national security agencies become harmful when used by third parties in unau- need access to CDRs, and legal call interception, to thorized or unexpected ways.103 Most internet users track criminals, but they should do this by having are willing to surrender personal data—or control appropriate safeguards in place to seek authorization, over it—for more convenience (gift suggestions based and not by using blanket surveillance measures. The on their past purchases), for possible gain (the chance exposure of indiscriminate snooping by governments to win a holiday), and to avoid payment (notably, has, ironically, encouraged large internet content when downloading and using free mobile apps). But providers to make far wider use of encryption, and few users have the means to challenge data use poli- manufacturers are following suit.114 cies or even make the effort to read the small print.104 Outside the domain of law enforcement, legislation Privacy online is also a balance. It may be economi- should give users more control of their data at the cally rational to relinquish some degree of privacy for point of use (if not at the point of collection). Con- the convenience of transacting online. But beyond sider the Electronic Health Record System in Estonia, respecting users’ choice in how they transact online, which is based on the principle that citizens own their public policy choices involve determining what other own health records and can easily access them and measures are necessary to afford appropriate privacy transfer them between doctors.115 protections. The key is to ensure that users are made Privacy policy concerns also need to be balanced aware of the risks of diminished privacy. against countervailing public policy issues such as One of these privacy challenges105 is the growing freedom of expression and government transpar- use of “big data.”106 Data mining has been routinely ency in the internet context. A recent decision from used, especially by social media and e-commerce the European Court of Justice,116 popularly known companies, to create value. The business model of as the right to be forgotten, highlights this debate. Twitter or Facebook, for instance, is to grant users Users in Europe now have greater control of the free storage, communication, and functionality in data about them on the internet.117 This right can be return for privileged access to the content they cre- exercised against search engines, such as Google or ate. Such companies then aggregate and analyze Bing, and would include seeking to remove search user-generated content to sell value to advertisers results pointing to documents held elsewhere in the wishing to reach targeted audiences. This classic two- cloud. But the decision does not offer clear guidelines sided market approach can make for a more intelli- for applying the right. Beyond that, once something gent and convenient user interface, but at the risk of is online, it is virtually impossible to ensure that all alienating users if they fi nd the use of their data too copies are deleted. Complying with user demands to intrusive.107 Mobile call data records (CDRs), which delete personal data will impose additional costs on record basic data like the time, duration, location, and businesses and governments. direction of calls, can also be mined to extract trend One sign of the growing awareness of public policy information.108And many applications are potentially concerns is the surge in new privacy laws, even while valuable, as in traffi c analysis109 or epidemiology.110 there has been divergence among countries (such as 226 WORLD DEVELOPMENT REPORT 2016

in the European Union and the United States) in their a long-standing concern in the United States about approach.118 According to the United Nations Confer- which Europe had been more hesitant. The U.S. ence on Trade and Development (UNCTAD), 107 coun- administration unveiled a blueprint for a consumer tries had privacy laws or bills in place as of 2014, but privacy bill of rights that would address consumer only 51 of them were developing countries (map 4.2).119 privacy more comprehensively. At the same time, At the international level the United Nations Gen- in reaction to the Edward Snowden revelations (dis- eral Assembly has adopted a Resolution, introduced cussed below), the European Court of Justice issued by Brazil and Germany, on the Right to Privacy in the its opinion in the Schrems case effectively invalidating Digital Age.120 The Africa Union on Cyber- the “Safe Harbor” agreement reached between the security includes data protection.121 The OECD, in 2013, European Union and the United States permitting the published its updated Privacy Guidelines, revising work handling of European data in the United States.126 originally carried out in the 1980s.122 The Asia-Pacifi c Different approaches, however, are making it Economic Cooperation followed up on its Privacy diffi cult to develop truly global internet services and Framework with a cross-border privacy arrangement applications because local adaptations are required to in 2010.123 The European Union is conducting an over- comply with differing national privacy laws, impos- haul of its data protection framework.124 while the ing additional compliance burdens and costs on busi- United States is also considering options.125 nesses (see chapter 6).127 Diverging rules also stifl e The data protection story is becoming more trade and innovation. In addition, the threat of “data complex. There is some evidence of convergence: nationalism”128—the idea that a country’s data should the newly proposed EU data protection framework be stored within its borders, recently embodied in a includes a directive looking closely at the protection new Russian law, for example, that requires the local of data as they relate to security and defense issues, storage of the personal data of Russian citizens—has

Map 4.2 National data protection and privacy laws and bills

Europe inset Caribbean inset

Data protection laws Comprehensive data protection law enacted Pending effort or obligation to enact law No law or information No data available

IBRD 41703 Source: UNCTAD 2015. Data at http://bit.do/WDR2016-Map4_2. SECTORAL POLICIES 227 grown since the revelations by Edward Snowden So, depending on where one sits, the issue can be about the U.S. government’s surveillance of foreign taken up as an issue of scarce bandwidth or of free internet traffi c, which was subsequently declared to speech. As with other resources where capacity is have been illegal.129 That puts an additional burden scarce, as on mobile networks, carriers attempt to use on companies seeking to build services that rely on price discrimination to manage traffi c. Network pro- cloud-based networks.130 viders argue that regulators should grant them more What is clear, however, is that getting the data fl exibility to manage their traffi c fl ows. But others protection and privacy piece of the puzzle right is, argue that traffi c management should not become together with cybersecurity, a key element in engen- an excuse to block certain traffi c streams, content, or dering trust in and confi dence in use of the internet. expression, to give to others, or to impede Even before “Snowden,” users in countries of all competition. The recent trend to develop services in regions were concerned that they should be careful which some basic content can be accessed free of data about what they say online, as more and more per- charges (such as Facebook’s Free Basics or Internet. ceived that their actions would be monitored.131 While org), while other content is subject to data charges, data protection and privacy are essentially matters would appear to be the antithesis of net neutrality of local law, a preponderance of privacy laws around and a distortion of markets.133 Nevertheless, some the world are based on a common set of international, defend the practice as a means of extending internet durable, and recognized principles. Moreover, inter- use in low-income countries.134 In India, following operability of data protection regimes will encourage protests, some participating organizations withdrew cross-border data fl ows and decrease the propensity support to Airtel Zero and Internet.org, as the regu- for data nationalism. As discussed, data protection is lator indicated that these platforms do not provide not solely a preoccupation of advanced economies. equal treatment to all online services.135 Unlike investments in infrastructure, ensuring that An open and free internet is also a key contrib- good practice in data protection is part of the overall uting factor to innovation in the digital economy, internet enabling environment makes sense for a making it critical to protect this openness. Care country interested in the internet as a means of eco- should be taken to ensure that users have the nomic development. greatest possible access to internet-based content, applications, and services of their choice. But traffi c Does net neutrality matter? management measures, while legitimate, should not One of the more confounding issues around the reduce the enjoyment of fundamental rights and internet today is “net neutrality.” The term, euphemis- freedoms, particularly freedom of expression. The tically derived out of debates in the United States, has balance here should be carefully calibrated so that become a shorthand for a debate about the manage- network operators continue to have incentives to ment and prioritization of scarce resources—notably build out and continuously improve networks and bandwidth—over the internet. Thus the debate has pit- network capacity. Recognizing the balance to be ted users and content providers against network pro- sought in this debate, the Council of Europe in 2010 viders.132 In oversimplifi ed terms, content providers announced its policy on “net neutrality” (without want an “open” and “free” internet where every data calling it such), emphasizing the rights-based aspect bit, or IP packet, is treated equally, while network pro- of the issue.136 viders seek to charge higher rates for services chewing Much of the early debate on net neutrality in up more bandwidth. In the United States, the debate the United States was focused on the classifi cation over net neutrality was mainly about the bandwidth of the internet under the 1996 Telecommunications consumed by video downloads. After announcing Act; in the rest of the world, the issues argued under new rules in February 2015, opponents of the rules the same moniker look at the impact of traffi c man- in the United States (mainly network providers) have agement techniques, such as deliberately slowing claimed that compelling the openness of the internet (or “throttling”) data streams, on human rights, and imposes on their rights as a corporation to determine the continuing impact of the internet as a driver of which content to make available. In other parts of the innovation. Access to information is not solely a world, the debate over net neutrality is recognized as a “developed” or “developing” country preoccupation. matter of freedom of expression or access to informa- Not only is it a fundamental human right, inherent tion for individuals—a human rights issue. in Article 19 of the International Covenant on Civil 228 WORLD DEVELOPMENT REPORT 2016

and Political Rights,137 it is also an essential element and golf courses.142 Specifi cally, the serendipitous in the online innovation ecosystem, and therefore “collisions” between like-minded individuals that an economic development issue. In whatever form a create innovative ideas are more likely to occur in country would wish to use the internet for develop- high-density urban environments than in lower- ment purposes, its public policies should ensure that density locales where high-technology industry has technical management of internet traffi c is not used traditionally fl ourished.143 to suppress a tool of innovation. Governments are generally not very good at pick- ing technology winners. Nor can they easily create Promoting the digital technology clusters, which tend to grow organically, where the right conditions are in place, as fi rms draw economy upon the same talent pool and startups spin off from established enterprises. From the recent development The internet as an open access ecosystem of tech hubs and FabLabs—for instance, there are now Once the internet is widely available and affordable in more than 555 FabLabs across 77 countries144—it is a country, and public trust that the internet is a safe unclear if all of them would have come up organically, and open platform for doing business or storing per- though government involvement in most of them has sonal information has been established, it can become been minimal (box 4.9 and map 4.3). But government an unrivaled platform for promoting entrepreneurial policy can help to sustain an emerging technology opportunity. Indeed, 9 of the top 20 people in the cluster, for instance by instituting favorable tax Forbes 2015 billionaires list made their money from regimes or liberal policies on awarding employment the sector.138 Initially, internet entrepreneurship was permits to skilled workers (“tech visas”). Israel shows associated with the dot-com bubble of the late 1990s, how government can stimulate the growth of an ICT but the networks of the day lagged behind the aspi- cluster once it becomes established (box 4.10). Govern- rations of entrepreneurs. As one subsequently said, ment procurement can also help create opportunities “A lot of the business plans were deeply fl awed, but for local companies to fl ourish without infringing a lot of the ideas would have worked had there been commitments to , for instance by unbun- broadband.”139 Since 2008, when broadband crossed dling large ICT projects into smaller components. But the threshold of 10-percent penetration worldwide, a governments also can unwittingly undermine a local new and more sustained period of internet entrepre- ICT sector, with lax rules on intellectual property neurship has developed. protection or burdensome taxes on the import of ICT The U.S. economy has so far been the main bene- goods and components. fi ciary of the internet as a source of innovation and entrepreneurship.140 This success is exemplifi ed by National ICT strategies the Silicon Valley technology cluster. Many coun- In recent years, it seems to have become ever more tries have tried to imitate it.141 Few have succeeded. complicated to develop a coherent set of policies to The ingredients of success for ICT clusters have regulate and promote a nation’s ICT sector. Rapid traditionally included close collaboration between technological change, coupled with disruptive academia and industry, easy access to venture capital, changes in the supply side of the industry and grow- high levels of government research spending, and ing challenges on the demand side, requires fl exi- a physical environment and climate that is attrac- bility. But industry players demand predictability in tive to footloose, highly paid workers. However, in policy setting and a level playing fi eld. ICT users are recent years, certain cities with doubtful climates becoming ever more vocal, and expertise that may have also developed a vibrant tech entrepreneur- once have been centralized is now dispersed through- ship scene, notably , Berlin, Hangzhou, out government. So, coordination becomes all the London, Nairobi, and New York. This suggests that more important. the internet is creating a new set of geographical When countries take a conscious decision to preconditions for innovation districts that are more develop a national strategy for broadband, or their about density and bandwidth than about sunshine ICT sector more generally, they are rewarded with SECTORAL POLICIES 229

Box 4.9 Tech hubs in Africa

The recent development of tech hubs across Africa (see and the longer-established Ushahidi), and then to nearby map 4.3) exemplifi es how technology clusters create a Strathmore University (where iBiz and iLab are located) snowball eff ect whereby initial preconditions for success and along Ngong Road to the GreenHouse and 88 mph/ generate additional, mutually reinforcing innovation Startup Garage.a drivers. Though nowhere similar in scale to Silicon Valley, As with the creative destruction of the original dot-com technology clusters in Africa nevertheless demonstrate bubble, many new startups, including some of the hubs that close collaboration between academia, government, themselves, have overestimated market demand for their and the private sector can help develop a vibrant eco- products and services and ended up with short lifespans. system that facilitates ongoing innovation and market Thus the turnover of hubs and incubators in Africa has been entry. Tech hub clusters, such as in Cape Town, Lagos, relatively rapid over the last fi ve years,b but the growth and Nairobi, show that once a certain threshold is passed, path is still upward, with a net increase of around 15 percent new hubs and ongoing entrepreneurial activity can be since the start of 2014. Tech hub performance also depends stimulated through enhanced access to fi nance, services to on context-specifi c dynamics, and the situation of the local entrepreneurs, and the demonstration eff ects of successful business community. However, the comparative success of “fi rst mover” startups. The development of Nairobi’s tech certain clusters, while others remain stagnant, suggests cluster dates to the founding of iHub in March 2010. Growth that organic, multistakeholder ecosystems work better has spread, fi rst to the rest of the Bishop Magua Centre than initiatives led by government, the private sector, or where it is located (including Nailab, m:lab East Africa, academia alone.

a. Firestone and Kelly 2015. b. http://www.gsmaentrepreneurshipkenya.com/GSMA_KENYA-AR2014-060214-WEB-SINGLE-PGS.pdf.

higher rates of service take-up.145 As of mid-2013, oping them and whether they can be easily imple- some 134 countries had already developed national mented. In the same way that international internet broadband plans, and a further 12 planned to do so. governance has evolved toward a multistakeholder Once developed, plans need to be regularly refreshed model, so too national strategies can benefi t from and updated, ideally on a cycle of three to fi ve years, ensuring that there is an opportunity for all stake- according to the United Nations Broadband Commis- holders to express their views and that those views sion, which has set a target that all countries should are refl ected in policy making and regulation. In have a by 2015, with at least an era in which mobile subscriptions exceed the 40 percent of households in developing countries world’s population, and with each internet user hav- served with broadband.146 National broadband plans ing within easy reach a vast library of global knowl- are also more effective when they include specifi c edge, ICT policy is too important to be left to the benchmarks, or targets, that are relevant, measurable, bureaucrats. Broad multistakeholder collaboration realistic, and actionable. For example, Estonia has set is the key. Equally important is the fact that digital a target that at least 100 Mbit/s broadband service development strategies need to be broader than they should be available to each citizen by 2015. are today, strengthening the country’s ICT infra- What seems to be important is not so much the structure, but also the foundation underpinning plans themselves, which vary enormously in quality its digital economy—an issue explored in the next and scope, but the consultation process for devel- chapter. 230 WORLD DEVELOPMENT REPORT 2016

Map 4.3 African tech hubs

Egypt, Arab Rep. 1. Cairo Hackspace Civ 2. The District Civ Tunisia 3. Flat6Labs Hy 1. Wiki Start Up Civ 4. IceCairo Civ 5. Fab Lab Egypt Civ Morocco 6. The Greek Campus Hy 1. Jokkolabs Casablanca Civ 7. Technology Innovation Gvt 2. New Work Lab Civ Entrepreneurship Center Civ 3. Girls in Tech (GIT) – Morocco Civ 8. Alexandria Hackerspace Chapter 4. OCP Entrepreneurship Network Hy

Senegal 1. Jokkolabs Dakar Civ 2. CTIC Dakar Civ MaliMMMli 3. Africa Living Lab Aca CameCCaCameroonCCC roon 1.1 JokkolabsJokkolabs BamakBamakokoo Civ 4. Mobile Senegal Hy 1.1. ActivSpacesActivSSpaces CivCiv 5. Jjiguene Tech Hub Civ EthiopiaEthihihiopia Burkina Faso 1. IceAddisIceAdAddAddis Aca 2.2.22 xHubbbbbb Civ 1. Yam Pukri Civ 2. Jokkolabs Ouagadougou Civ Kenya The Gambia Uganda 1. iHub Civ Civ 1.1.1. Jokkolabs Banjul Civ 1. HIVE Colab 2. m:lab East Africa Hy Civ 2. Woe Lab Civ 2. @TheHub 3. GrowthHub Civ 3. Outbox Hub Civ 4. Nailab Civ 4. Grameen Foundation AppLab Civ 5. C4D Lab Aca Liberia 5. Women in Technology Uganda Civ 6. AkiraChix Civ (WITU) 7. Lake Hub Civ 1. iLab Liberia Civ 8. iBiz Africa Aca 9. iLab Africa Aca Côte d d’Ivoire Ivoire 10. FabLab Nairobi Aca Hy 1. Jokkolabs Civ 11. 88 mph /Startup Garage 2. Theee W Hub Civ 3. Akendewa Civ Rwanda 4. AMN Co-working Space Civ 1. kLab Hy 5.5 MwasiMwasiTech Tech Hub Civ 2. The Office Civ 3. THINK Technology Incubator Civ Ghana Tanzania 1. mFriday Civ 2. Meltwater Entrepreneurial Aca 1. Kinu Innovation and Co-Creation Civ School of Technology (MEST) Space [Incubator] 2. TANZICT Hy 3. gSpace Civ 3. Dar Tekinohama Business Civ 4. iSpace Civ Incubator (DTBI) 5. Ghana Multimedia Incubator Gvt Congo,Congo Rep. 4. Buni Hub Civ Centre 1. BantuHubBantuH Civ 6. Mobile Web Ghana Civ 7. Kumasi Business Incubator Hy Madagascar 8. Oguaa Business Incubator Civ Congo,, Dem. Rep. 1. I-Hub Malagasy Civ 9. Hub Hy 1. Imani HubH Civ 2. Habaka Civ

Togo Angolaa 1. Mara Launchpad Civ 1.1 AngolanAngolan Institute of Support for Gvt Benin Micro, S Small and Medium MauritiusMauritius EnterprEnterprisesrises [INAPEM] 1. Ebène AcceleratorAcccelerator Gvt 1. e-TRILABS Civ (ICT IncIncubator)cubator) 2.2 JokkolabsJkklbCt Cotonou CivCiCivCiv ZambiaZambia Nigeria Zimbabwe 1.1 BongohiveBongohhive Zambia Civ France (La(LaRéu Réunion) 1. L5 Lab Civ 1. Sky Hub Initiative Civ 1. Cyclotronn RéunionRéunio Indian Aca Civ 2. Hypercube Hub Civ 2. Co-creation Hub NamibiaNamibbia Ocean (CYROI(CYYROI GIP) GIP 3. Wennovation Hub Civ 3. Muzinda Umuzi Hub Civ 2. TechnopôleTechnopôôle de La Réunion Hy 4. Technology Incubation Centretre Gvt 1. NamibiNamibiaia Business Innovation Aca 5. Minna Tech Incubation Centretre Gvt CentreCt (NBIC)(NBIC)(NBIC) 6. Information Developers Hy 2. Fablab Namibia Hy Entrepreneurship Acceleratortor Mozambique (IDEA) Botswana 1. MICTI Technology and Business Gvt 7. Focus Hub Civ Gvt Centre 8. Enspire Incubator Hy 1. Botswana Innovation Hub Hy 9. Calabar Technology Incubationtion Gvt 2. First Steps Venture Center Center South Africa 10. 88 MPH (400 NG) Hy 1. Jozihub Civ 12. Smart Xchange Hy 2. Capetown Garage Civ 13. CodedinBraam Civ 3. Black Girls Code Civ 14. SoftStart Technology Civ 4. ThoughtWorks Civ 15. The House 4 Hack Civ 5. TechinBraam Hy 16. mLabbS Southern h Africa f Civ Tech hubs in Africa 6. Silicon Cape Initiative Civ 17. Start-up Garage/98MPH Hy Gvt 7. Impact Amplifier Civ 18. IMPACT HUB Civ Government led 10 8. Codebridge Civ 19. Rlabs Civ 9. Angel Hub Civ 20. Invo Tech Incubator Hy Civ 10. Eastern Cape Information Civ 21. The Innovation Hub (TIH) Gvt Civil society led 78 Technology Initiative 22. Startup 90 Civ 11. Cape Innovation and Technology Civ 23. Grindstone Civ Aca Academic institution led 9 Initiative (CITI) Hy Hybrid 20 Total tech hubs by Sept. 2015 = 117

IBRD 41639 Source: WDR 2016 team. Interactive map and source data available at http://bit.do/WDR2016-Map4_3. SECTORAL POLICIES 231

Box 4.10 Israel as a startup nation

The term “startup nation” was used to describe Israel in 2000), supporting entrepreneurs in the very fi rst stages of a 2011 best-selling book.a It takes as its starting point the development; Magneton (since 2001), supporting industrial/ 6,000 startups created between 2000 and 2010, more than academic cooperation; the Public Technology Incubator twice the rate of the previous decade. In 2013, the informa- Program set up in the early 1990s to help assimilate a wave tion and communication technology (ICT) sector in Israel of immigrants from the former ; and the Yozma made up 16 percent of GDP and 26 percent of exports and Fund, a venture capital fund set up in 1993 for technology employed 8.5 percent of its workforce.b It was also the fi fth companies. As a result, Israel consistently scores the highest largest exporter of computer software and services in 2012, level of gross domestic expenditures on research and devel- according to the World Trade Organization,c as ICT exports opment among OECD economies, although the Republic of have increased almost tenfold since 1990 to over US$20 Korea is catching up (fi gure B4.10.1, panel b). billion (fi gure B4.10.1, panel a).d A key ingredient in the But government support for fi rms in the ICT sector is not success of the startup nation has been the role of the Israeli the whole story. Israel’s government has promoted a high government, which designed its interventions so as not to level of investment in education, and the sector has bene- hamper the emergence of the private sector. fi ted from the innovativeness of the fi rm’s themselves, which Many Organisation for Economic Co-operation and have been particularly successful in fi elds such as cyberse- Development (OECD) governments play an active role in curity, mobile phone apps, and voice over internet protocol. funding innovation but Israel’s program, from the very Recent tech startups include Taboola (a content distribution start in the late 1960s, has been particularly geared toward and discovery platform), Silverbyte Systems (providing hotel “sector-neutrality”—meaning that the government did not management software), and Ceragon Networks (a wireless try to “pick winners.” The program’s design avoided crowd- backhaul provider), while longer-established ones include ing out private investment and stimulated additionality.e Waze, which provides a location-based navigation app, and Recent programs have included the Tnufa program (since Wix, off ering a cloud-based web development platform.

Figure B4.10.1 How Israel stays ahead in high-tech entrepreneurship

a. Israel’s ICT exports, 1990–2014 b. Gross domestic expenditures on R&D as a percentage of GDP, selected economies, 2004–13

25 5.0 4.5 20 4.21 4.0 15 4.15 3.49 3.5 10 3.32 3.0 5

Percent 2.81 2.5 Billions of U.S. dollars Billions of U.S. 0 2.40 2.0 2.02 1990 1995 2000 2005 2010 2014 1.5

1.0 2004 2005 2006 2007 2008 2009 20102011 2012 2013 Israel Japan United States China Korea, Rep. Finland OECD average

Source: OECD STI Indicators, available from http://www.oecd.org/sti/msti.htm. Data at http://bit.do/WDR2016-FigB4_10_1. Note: In panel b, the percentages shown on the fi gure are for 2013, except for the U.S. percentage, which is for 2012. GDP = gross domestic product; ICT = information and communication technology; OECD = Organisation for Economic Co-operation and Development; R&D = research and development. a. Senior and Singer 2011. b. Getz and Goldberg 2015. c. See https://www.wto.org/english/res_e/statis_e/its2014_e/its14_trade_category_e.htm. d. Israel, Central Bureau of Statistics 2014. e. Lach, Parizat, and Wasserteil 2008. 232 WORLD DEVELOPMENT REPORT 2016

Notes 1. “High-speed internet” in this chapter is equated Thus Somali migrants cannot roam in neighboring with fi xed-line broadband service (such as digital countries without changing their SIM cards. subscriber line, cable modem, and fi ber optics) and 23. Ampah and others 2009. fourth-generation (4G) mobile services. Internet 24. See https://gsmaintelligence.com/, although other usage is much higher, at around 40 percent, but this sources such as Gallup World Poll show lower is generally at slower speeds and is not necessarily coverage. affordable or “always on.” 25. ITU 2001. 2. See chapter 1 and Kim, Kelly, and Raja (2010). 26. See “WhatsApp Shows How Phone Carriers Lost 3. Markovitch 2013. Out on US$33 Billion,” Business Standard, February 28, 4. Qiang 2010. 2014, at http://www.business-standard.com/article 5. UN Broadband Commission 2013. /international/whatsapp-shows-how-phone 6. Levinson 2007. -carriers-lost-out-on-33-bn-114022200022_1.html. 7. Internet protocol (IP)–based networks used packet- 27. In the fi rst half of 2014, Netfl ix accounted for 35 switching, a technology developed in the 1960s, that percent and YouTube 13 percent of internet data packages information into uniform chunks of dig- traffi c to users in the United States, during peak ital data, or “packets,” that can be used to transmit periods on fi xed networks, according to Sandvine voice, text, video, or data. IP-based networks thus (2014). combine into a single network those functions that 28. Cisco 2015. were previously handled by separate and distinct 29. Cisco 2015. networks (Mathison, Roberts, and Walker 2012). 30. Regional average data from GSMA at www.gsma 8. Wu 2010. intelligence.com. 9. See Heeks (2014), notably fi gure 8 on the changing 31. Minges 2015. focus of ICT4D priorities over time. 32. Evans, Groves, and Croft 2014. 10. IDATE 2014. 33. Stock market valuations change constantly, but on 11. Cleland 2014. May 18, 2015, China Mobile had a market valuation 12. Wheeler 2015. of US$276.45 billion on revenue of US$103.21 bil- 13. The United States followed a different model of lion, while Alibaba had a market valuation of private ownership, but even there the monopoly US$217.26 billion on revenue of just US$12.26 of AT&T initially prevailed, as this was considered billion, according to Yahoo Finance. essential to achieving “universal service,” a tele- 34. Investment in constant U.S. dollars fell to US$248 phone in each home. An antitrust suit launched in billion in 2013, down from US$357 billion in 2008, 1974 eventually broke up AT&T into seven separate before the fi nancial crisis, according to data on regional operators and a national long-distance and private sector investment in telecoms, sourced by international carrier a decade later, opening the the Public-Private Infrastructure Advisory Facility door for competition (Crandall 1991). (PPIAF) and available at http://data.worldbank.org 14. Harper 1997. /indicator/IE.PPI.TELE.CDcountries/1W?display= 15. Cho 2002. default. 16. OECD 1999. 35. Offi ce of Legislative and Public Affairs 2003. 17. Ungerer 2000. 36. The announcement of the transition was on March 18. See http://ec.europa.eu/priorities/digital-single 14, 2014, by the National Telecommunications and -market/. Information Administration of the U.S. Department 19. The 69 World Trade Organization member coun- of Commerce, but the completion date was later tries that made commitments in telecommuni- delayed by one year to September 2016. For more cations in 1997 have now risen to 108; see https:// information, see https://www.icann.org/resources www.wto.org/english/tratop_e/serv_e/telecom_e /pages/process-next-steps-2014-10-10-en and http:// /telecom_e.htm. www.ntia.doc.gov/blog/2015/update-iana-transition. 20. ITU 1997. 37. For more information on different national 21. Wellenius and Stern 1995. research and education network models, see the 22. The cover photo illustrates a failing in all three South-South Exchange of Research and Education ingredients for a successful digital policy: Djibouti Experience, at https://openknowledge.worldbank Telecom is fully state-owned and enjoys monopoly .org/handle/10986/12367. status, contributing in part to its high prices for 38. See the regulatory information section of ITU’s ICT mobile service. But neighboring Somalia currently Eye database, at http://www.itu.int/icteye. lacks a national communications act and sector 39. As an example, the Communications Authority of regulator, which means that local operators func- Kenya, for the fi nancial year ending June 30, 2012, tion without internationally recognized licenses. records a surplus of around K Sh 7.0 billion (around SECTORAL POLICIES 233

US$74 million) on income of K Sh 8.8 billion (US$98 /sectors/telecommunications/broadband_decisions million), a surplus of around 80 percent. This is .pdf). over and above other taxes and other fi nancial con- 58. Marino García 2015a. tributions from the sector. See the Communications 59. Coleago Consulting 2013. Authority of Kenya Annual Report, 2012–13, http://www 60. Plum Consulting 2013. .ca.go.ke/images//downloads/PUBLICATIONS 61. Moore 1965. /ANNUALREPORTS/Annual%20Report%20for%20 62. See http://www.mkomo.com/cost-per-gigabyte the%20Financial%20Year%202012-2013.pdf. -update. 40. Blackman and Srivastava 2011. 63. Biggs and Kelly 2006. 41. Rijkers, Freund, and Nucifora 2014. 64. Economist 2015. 42. Economist 2012. 65. This is sometimes referred to as Amara’s Law. 43. See GSMA at https://gsmaintelligence.com/. 66. See http://articles.latimes.com/2006/jul/16/business 44. ITU estimates that some 95 percent of the world’s /fi -overheat16. population lived within range of a mobile signal in 67. Samarajiva 2010. 2015 (ITU 2015). 68. Genakos and Valletti 2014. 45. This aggregate statistic (68 percent) is derived from 69. Littlechild 2006. GSMA (http://www.gsmaintelligence.com), which 70. Kemei and Kelly 2015. estimates that, as of Q4 2014, some 98 percent of the 71. ITU 2014. citizens of developed economies and 62 percent of 72. http://www.oecd.org/sti/broadband/price-baskets developing ones lived within range of a 3G signal. .htm and http://www.oecd-ilibrary.org/docserver The equivalent fi gures for 4G/LTE were 88 percent /download/5k92wd5kw0nw.pdf?expires= and 15 percent, giving an aggregate of 26 percent. 1419284221&id=id&accname=guest&checksum= 46. Mayer and others 2008. 70B5A3123B6297371A59AC6F4C97AC55. 47. As an example, the Comoros opened its mobile 73. ITU 2014. market to competition in 2007, but the incumbent 74. ITU 2014. operator, Comoros Telecoms, was successfully able 75. Boston Consulting Group 2012. to resist interconnection due to weak regulation, 76. World Bank, forthcoming. and the market entrant, Twama Telecom, never 77. See https://www.washingtonpost.com/news started service (World Bank 2014). The Comoros /the-switch/wp/2015/09/22/dig-once-the-no announced a license award for a new second opera- -brainer-internet-policy-the-white-house-just tor, won by Telma, in October 2015. -endorsed/. 48. Third-generation (3G) and fourth-generation (LTE) 78. 2015. operate in multiple spectrum bands, but mainly at 79. Kende and Hurpy 2012. 1,800, 2,300, and 2,600 MHz. In countries where the 80. Kende and Rose 2015. transition to digital broadcasting is complete, the 81. Regulation (EC) No. 2887/2000 of the European Par- more valuable spectrum of 700 MHz can be released liament and of the Council of December 18, 2000, on for mobile communications. This is more attractive unbundled access to the local loop, at http://europa to investors as fewer cells are needed to cover the .eu/legislation_summaries/information_society same territory (Dahlman and others 2007). /internet/l24108j_en.htm. 49. See Partnership for Measuring ICT for Develop- 82. Klein and Wendel 2014. ment 2014. 83. See https://apps.fcc.gov/edocs_public/attachmatch 50. Ros and Banerjee 2000. /DOC-327104A1.pdf. 51. Wellenius 2002. 84. See http://www.washingtonpost.com/blogs/the 52. GSMA (Global System for Mobile communications -switch/wp/2014/12/23/fcc-confirms-that-680000 Association) 2013. -net-neutrality-comments-weremissing-from-the 53. Minges and others 2014. -public-record/. 54. UN Broadband Commission 2015, 64–69. 85. See http://www.nytimes.com/2015/02/27 55. See http://www.fcc.gov/encyclopedia/connecting /technology/net-neutrality-fcc-vote-internet-utility -america. .html?_r=0. 56. See http://ec.europa.eu/digital-agenda/ and https:// 86. UNESCO. See, specifi cally, Article 19 of the Inter- ec.europa.eu/digital-agenda/en/connecting-europe national Covenant on Civil and Political Rights -facility#the-connecting-europe-facility-to-support (ICCPR); Article 13 of the African Charter of Human -eu-infra. Rights (ACHR); Article 9 of the African Charter; 57. This calculation is based on projects that have and Article 10 of the European Court of Human received a waiver from the European Commission Rights. for the use of state aid in the deployment of broad- 87. See Open Network Initiative at http://map.opennet band networks (see http://ec.europa.eu/competition .net/fi ltering-pol.html. 234 WORLD DEVELOPMENT REPORT 2016

88. Dutton and others 2013. DNA samples interfered with an individual’s right 89. Clover 2015. to privacy. 90. Waldfogel 2011. 114. See http://www.bbc.com/news/technology 91. Internet Society 2014. -29276955. 92. See http://www.ndtv.com/india-news/supreme 115. Doupi and others 2010. -court-reserves-orders-on-validity-of-section-66a 116. Google Spain and Google vs. Agencia Española de Protec- -of-it-act-742758. ción de Datos (AEPD) and Mario Costeja González [Case 93. See http://www.google.com/transparencyreport/. No. 131/12, 2014] European Court of Justice. 94. See http://www.google.com/transparencyreport 117. Ausloos 2012. /traffi c/. 118. See Schwartz and Solove (2014). The EU generally 95. See http://blogs.wsj.com/chinarealtime/2015/02/12 prefers legal instruments that position privacy as a /china-internet-restrictions-hurting-business human right, whereas the United States relies more -western-companies-say/. on self-regulation, seeing privacy as a consumer 96. This chapter similarly adopts an expansive view of protection issue. “cybersecurity” and delves more deeply into some 119. UNCTAD 2015 aspects of cybersecurity—such as privacy and 120. UN General Assembly 2013. cybercrime—separately. 121. See the Convention on Cyber- 97. See Bauer and Dutton (2015), citing other sources. security and Personal Data Protection (July 27, 98. Bauer and Dutton 2015. 2014). 99. Article 17 ICCPR; Article 11 ACHR; Article 8 ECHR; 122. OECD 2013. The OECD Privacy Guidelines incorporate and UNGA resolution 68/167 (2013) on the right to a “Recommendation Concerning Guidelines Gov- privacy in the digital age. erning the Protection of Privacy and Transborder 100. YCharts, Facebook Market Cap (June 10, 2015); see Flows of Personal Data.” http://ycharts.com/companies/FB/market_cap. 123. See http://www.apec.org/Groups/Committee 101. WEF 2014. -on-Trade-and-Investment/Electronic-Commerce 102. Dutton and others 2013. -Steering-Group/Cross-border-Privacy-Enforcement 103. FTC 2013. -Arrangement.aspx. 104. See http://www.washingtonpost.com/blogs/the 124. European Commission (EC 2010). -switch/wp/2014/11/13/facebook-rewrites-its 125. White House 2012. -privacy-policy-so-that-humans-can-understand-it/. 126. Maximilian Schrems vs. Data Protection Commissioner 105. Polonetsky, Tene, and Jerome 2014. [Case No. C-362/14, 6 October 2015] European Court 106. World Bank and Second Muse 2014. of Justice. 107. See http://georgemdallas.wordpress.com/2014/06/05 127. The recent decision by the European Court of /making-sense-of-internet-platforms-network Justice about the “right to be forgotten” (Google Spain -effects-and-two-sided-markets/. and Google vs. Agencia Española de Protección de Datos 108. Haddad and others 2014. [AEPD] and Mario Costeja González [Case No. 131/12, 109. Berlingerio and others 2013. 2014]), which requires search engines to remove 110. See http://www.economist.com/news/leaders links to data upon request if they are inadequate /21627623-mobile-phone-records-are-invaluable or irrelevant, throws into question not only the bal- -tool-combat-ebola-they-should-be-made-available. ance of data protection and freedom of expression 111. See Article 29 Data Protection Working Party 2014. rights, but also extraterritorial application of law 112. See https://www.ftc.gov/system/files/documents and perhaps irreconcilable compliance obligations /reports/federal-trade-commission-staff-report of multinational companies processing data across -november-2013-workshop-entitled-internet-things borders. -privacy/150127iotrpt.pdf. 128. See https://www.huntonprivacyblog.com/2015/01 113. Recent jurisprudence from both the European /articles/deadline-for-compliance-with-russian Court of Justice (ECJ) and the European Court of -localization-law-set-for-september-1-2015/. Human Rights (ECHR) support striking this bal- 129. See http://www.theguardian.com/us-news/2015 ance. In Ireland Ltd v. Ireland (Joined /may/07/nsa-phone-records-program-illegal-court. Cases C-293/12 and C-594/12 Seitlinger and Others, 130. Castro 2013. April 8, 2014), the ECJ ruled the Data Retention 131. Dutton 2015. Directive to be in violation of the EU Charter of 132. The origins of the debate around net neutrality are Fundamental Rights. In S and Marper v. UK 30562/04 the U.S. Communications Act of 1996, as amended. [2008] ECHR 1581 (December 4, 2008), the ECHR, Under the act, “telecommunications services” are using a proportionality analysis, found the United regulated under common carrier rules, and other Kingdom to be in breach of Article 8, holding that communications services, such as data, are regu- the long-term retention of both fi ngerprints and lated differently. In the United States, the internet SECTORAL POLICIES 235

has been classifi ed as a data service and therefore lenges to an Open and Global Internet.” Background not subject to the same regulatory rules as telecom- paper for the World Development Report 2016, World munications services. In most other jurisdictions, Bank, Washington, DC. the distinctions found in the act do not apply, Bauer, Johannes, Michel van Eeten, Tithi Chattopadhyay, although the term net neutrality (though derived and Yuehua Wu. 2008. “ITU Study on the Financial out of specifi c statutory language) has come to be Aspects of Network Security: Malware and Spam.” applied globally. http://www.itu.int/ITU-D/cyb/cybersecurity/docs 133. See http://trak.in/tags/business/2015/04/17/trai /itu-study-financial-aspects-of-malware-and-spam -airtel-zero-internet-org-against-net-neutrality .pdf. -violate-free-internet/. Berlingerio, M., F. Calabrese, G. Di Lorenzo, R. Nair, 134. West 2015. F. Pinelli, and M. L. Sbodio, eds. 2013. “All Aboard: 135. See http://www.bbc.com/news/technology Smarter Urban Dynamics.” Dublin, Ireland: IBM -32334181. Research. 136. Council of Europe 2010. Biggs, Phillippa, and Tim Kelly. 2006. “Broadband Pricing 137. See http://www.ohchr.org/en/professionalinterest Strategies.” Info 8 (6): 3–14. /pages/ccpr.aspx. Blackman, Colin, and Lara Srivastava. 2011. Telecommuni- 138. See http://www.forbes.com/billionaires/list cation Regulation Handbook: Tenth Anniversary Edition. /#version:static. Washington, DC: World Bank. 139. Citation from entrepreneur Mark Simon, quoted in Boston Consulting Group. 2012. “Egypt at a Crossroads: Madslien 2010. How the Internet Is Transforming Egypt’s Economy.” 140. Kushida 2015; Minges 2015. http://www.bcg.com/documents/fi le124361.pdf. 141. Bresnahan, Gambardella, and Saxenian 2001. Bresnahan, Timothy, Alfonso Gambardella, and Anna- 142. Katz and Wagner 2014. lee Saxenian. 2001. “‘Old Economy’ Inputs for ‘New 143. See http://fortune.com/2012/05/09/the-random Economy’ Outcomes: Cluster Formation in the New -collision-theory-of-innovation/. Silicon Valleys.” Industrial and Corporate Change 10 (4): 144. FabLabs are similar to Tech Hubs, but focus more on 835–60. making physical prototypes than software, content Castro, Daniel. 2013. “The False Promise of Data Nation- and applications development; see, for instance, alism.” http://www2.itif.org/2013-false-promise-data Tokushima and Tanaka (2015). -nationalism.pdf. 145. The United Nations Broadband Commission (UN Cho, Sung-Hye. 2002. “Telecommunications and Infor- Broadband Commission 2013) reports that the 134 matization in .” Networks and Communica- countries that had a national broadband plan as of tion Studies 16 (1/2): 29–42. mid-2013 had, on average, some 2.5 percent higher Cisco. 2015. “Cisco Visual Network Index: Global Mobile fi xed broadband penetration and 7.4 percent higher Data Traffi c Forecast Update, 2014–2019.” http://www mobile broadband penetration. .cisco.com/c/en/us/solutions/collateral/service 146. See http://www.broadbandcommission.org/about -provider/visual-networking-index-vni/white /Pages/default.aspx. _paper_c11-520862.pdf. Cleland, Scott. 2014. “The Top Ten Reasons to Oppose References Broadband Utility Regulation.” http://dailycaller .com/2014/05/28/the-top-ten-reasons-against-broad Ampah, Mavis, Daniel Camos, Cecilia Briceño- band-utility-regulation/. Garmendia, Michael Minges, Maria Shkaratan, and Clover, Charles. 2015. “Chinese Internet: Com- Mark Williams. 2009. “Information and Communica- merce and Control.” Financial Times, March 13. tions Technology in Sub-Saharan Africa: A Sector Coleago Consulting. 2013. “Revised Spectrum Forecasts Review.” Africa Infrastructure Country Diagnostic Using the New Spectrum Model: Spectrum Required Study, Background Paper 10, African Development for Various Mobile Communications Markets in Bank Group, Abidjan, Côte d’Ivoire. 2013.” Report for Groupe Speciale Mobile Associa- Article 29 Data Protection Working Party. 2014. “Opinion tion. http://www.gsma.com/spectrum/wp-content 8/2014 on Recent Developments on the Internet of /uploads/2014/01/Coleago-Report-on-Spectrum Things.” http://ec.europa.eu/justice/data-protection -Demand-Model-Results.pdf. /article-29/documentation/opinion-recommendation Council of Europe. 2010. “Declaration of the Commit- /fi les/2014/wp223_en.pdf. tee of Ministers on Network Neutrality.” Adopted Ausloos, Jef. 2012. “The Right to Be Forgotten—Worth by the Committee of Ministers on 29 September Remembering?” Computer Law and Security Review 28 2010. https://wcd.coe.int/ViewDoc.jsp?id=1678287. (1): 143–52. Crandall, Robert. 1991. After the Breakup: U.S. Telecommu- Bauer, Johannes, and Bill Dutton. 2015. “Addressing the nications in a More Competitive Era. Washington, DC: Cybersecurity Paradox: Economic and Cultural Chal- Brookings Institution. 236 WORLD DEVELOPMENT REPORT 2016

CSIS (Center for Strategic and International Studies) Haddad, Ryan, Tim Kelly, Teemu Leinonen, and Vesa and McAfee. 2014. “Net Losses: Estimating the Global Saarinen. 2014. Using Locational Data from Mobile Cost of Cybercrime.” http://www.mcafee.com/us Phones to Enhance the Science of Delivery. Washington, /resources/reports/rp-economic-impact-cybercrime2 DC: World Bank. .pdf. Harper, John. 1997. Monopoly and Competition in British Dahlman, Erik, Stefan Parkvall, Johan Sköld, and Per Telecommunications: The Past, the Present and the Future. Beming. 2007. 3G Evolution: HSPA and LTE for Mobile London: Pinter. Broadband. Oxford, U.K.: Elsevier Academic Press. Heeks, Richard. 2014. “ICT4D 2016: New Priorities for Doupi, P., E. Renko, S. Giest, J. Heywood, and J. Dumort- ICT4D Policy, Practice and WSIS in a Post-2015 ier. 2010. “eHealth Strategies: Country Brief, Estonia.” World.” Centre for Development Informatics Work- European Commission. http://ehealth-strategies.eu ing Paper 59, University of Manchester, Manchester, /database/documents/Estonia_CountryBrief U.K. _eHStrategies.pdf. Ibarguen, Giancarlo. 2003. “Liberating the Radio Spectrum Dutton, William. 2015. “Multi-Stakeholder Internet in Guatemala.” Telecommunications Policy 27 (7): 543–54. Governance.” Background paper for the World Devel- IDATE. 2014. DigiWorld Yearbook 2014: The Challenges of the opment Report 2016, World Bank, Washington, DC. Digital World. Montpellier: IDATE. Dutton, William, Ginette Law, Gillian Bolsover, and IDRC (International Development Research Centre). 2010. Soumitra Dutta. 2013. “The Internet Trust Bubble: Radio-Spectrum Allocation and Management in Central Global Values, Beliefs and Practices.” World Economic American Countries and Their Impact in the Development Forum, Geneva. of the Mobile Telecommunications Services Sector. https:// EC (European Commission). 2010. “A Comprehensive idl-bnc.idrc.ca/dspace/bitstream/10625/49429/1 Approach on Personal Data Protection in the European /IDL-49429.pdf. Union.” Communication from the Commission to Internet Society. 2014. Global Internet Report 2014: Open and the European Parliament. http://ec.europa.eu/justice Sustainable Access for All. https://www.internetsociety /news/consulting_public/0006/com_2010_609_en.pdf. .org/sites/default/fi les/Global_Internet_Report_2014 Economist. 2012. “Indian Telecoms Scandal: Megahurts.” _0.pdf. February 11. Israel, Central Bureau of Statistics. 2014. “Data on the ————. 2015. “The Truly Personal Computer: The Smart- Israeli Economy.” http://www.cbs.gov.il/reader. phone Is the Defi ning Technology of the Age.” ITU (International Telecommunication Union). Various February 28. years. World Telecommunication/ICT Indicators Evans, David, Jon Groves, and Will Croft. 2014. “Oper- database. http://www.itu.int/ti. ator Group Ranking, 2Q 2014.” GSMA Intelligence. ————. 1997. World Telecommunication Development Report https://gsmaintelligence.com/analysis/2014/09 1997: Trade in Telecommunications. Geneva: ITU. /operator-group-ranking-q2-2014/444/. ————. 2001. ITU Internet Reports 2001: IP Telephony. Firestone, Rachel, and Tim Kelly. 2015. “Africa’s Thriving Tech Hubs.” Background paper for the World Develop- Geneva: ITU. ment Report 2016, World Bank, Washington, DC. ————. 2014. Measuring the Information Society. Geneva: FTC (U.S. Federal Trade Commission). 2013. “Mobile Pri- ITU. vacy Disclosures: Building Trust through Transpar- ————. 2015. Facts and Figures. Geneva: ITU. ency.” http://www.ftc.gov/sites/default/files Katz, Bruce, and Julie Wagner. 2014. The Rise of Innovation /documents/reports/mobile-privacy-disclosures Districts: A New Geography of Innovation in America.” -building-trust-through-transparency-federal-trade Washington, DC: Brookings Institution. -commission-staff-report/130201mobileprivacy Kelly, Tim, and David Souter. 2014. The Role of Information report.pdf. and Communication Technologies in Postconfl ict Recon- Genakos, Christos, and Tommasso Valletti. 2014. “Evalu- struction. Washington, DC: World Bank. ating a Decade of Mobile Termination Rate Regula- Kemei, Christopher, and Tim Kelly. 2015. “One Network tion.” Discussion Paper 1282, Centre for Economic Area in East Africa.” Background paper for the World Performance, London. Development Report 2016, World Bank, Washington, Getz, Daphne, and Itzhak Goldberg. 2015. “Best Practices DC. and Lessons Learned in ICT Sector Innovation: A Case Kende, Michael, and Charles Hurpy. 2012. “Assessment Study of Israel.” Background paper for the World Devel- of the Impact of Internet Exchange Points: Empirical opment Report 2016, World Bank, Washington, DC. Study of Kenya and Nigeria.” http://www.internet GSMA (Groupe Speciale Mobile Association). 2013. society.org/sites/default/fi les/Assessment%20of%20 “Survey of Universal Service Funds: Key Findings.” the%20impact%20of%20Internet%20Exchange%20 http://www.gsma.com/publicpolicy/wp-content Points%20%E2%80%93%20empirical%20study%20 /uploads/2013/04/GSMA-USF-Key-fi ndings-fi nal.pdf. of%20Kenya%20and%20Nigeria.pdf. SECTORAL POLICIES 237

Kende, Michael, and Karen Rose. 2015. “Promoting Local ground paper for the World Development Report 2016, Content Hosting to Develop the Internet Ecosystem.” World Bank, Washington, DC. http://www.internetsociety.org/doc/promoting-local Minges, Michael, Kaoru Kimura, Natasha Beschorner, -content-hosting-develop-internet-ecosystem. Robert Davies, and Guanqin Zhang. 2014. Information Kim, Yongsoo, Tim Kelly, and Siddhartha Raja. 2010. and Communications in the Chinese Countryside: A Study Building Broadband: Strategies and Policies for the Devel- of Three Provinces. A World Bank Study. Washington, oping World. Washington, DC: World Bank. DC: World Bank. Klein, Gordon, and Julia Wendel. 2014. “The Impact of Moore, Gordon. 1965. “Cramming More Components Local Loop and Retail Unbundling Revisited.” Discus- onto Integrated Circuits.” Electronics 38 (8): 114–17. sion Paper 163, Dusseldorf Institute for Competition OECD (Organisation for Economic Co-operation and Economics, Dusseldorf, Germany. Development). 1999. Regulatory Reform in Japan: Regu- Kushida, Kenji. 2015. “The Politics of Commoditization latory Reform in the Telecommunications Industry. Paris: in Global ICT Industries: A Political Economic Expla- OECD. nation of the Rise of Apple, Google and Industry ————. 2013. The OECD Privacy Framework. Paris: OECD. Disruptors.” Journal of Industry, Competition and Trade Offi ce of Legislative and Public Affairs. 2003. A Brief His- 15 (1): 49–67. tory of NSF and the Internet. Washington, DC: National Lach, , Shlomi Parizat, and Daniel Wasserteil. 2008. Science Foundation. http://www.nsf.gov/od/lpa/news “The Impact of Government Support to Industrial /03/fsnsf_internet.htm. R&D on the Israeli Economy.” Report prepared for the Packet Clearing House. 2015. “Full Exchange Point Israeli Ministry of Finance and the Offi ce of the Chief Dataset.” https://prefi x.pch.net/applications/ixpdir Scientist, Ministry of Industry, Trade and Labor. /menu_download.php. Levinson, Marc. 2007. The Box: How the Shipping Container Partnership for Measuring ICT for Development. 2014. Made the World Smaller and the World Economy Bigger. “Final WSIS Targets Review: Achievements, Chal- Princeton, NJ: Princeton University Press. lenges and the Way Forward.” http://www.itu.int Littlechild, Stephen. 2006. “Mobile Termination Charges: /en/ITU-D/Statistics/Pages/publications/wsistar Calling Party Pays versus Receiving Party Pays.” Tele- gets2014.aspx. communications Policy 30: 242–77. Plum Consulting. 2013. “Harmonised Spectrum for Madslien, Jorn. 2010. “Dotcom Bubble Burst: Ten Years Mobile Services in ASEAN and South Asia.” Report for On.” BBC website, March 9. http://news.bbc.co.uk/2 Axiata Group Berhad. http://www.plumconsulting /hi/business/8558257.stm. .co.uk/pdfs/Plum_Jan2014_harmonised_spectrum Marino García, José. 2015a. “The Economics and Policy _for_mobile_asean_south_asia.pdf. Implications of Infrastructure Sharing and Mutu- Polonetsky, Jules, Amer Tene, and Joseph Jerome. 2014. alisation and Its Role for the Development of ICT “Benefi t-Risk Analysis for Big Data Projects.” Future Networks in Africa.” Companion paper for the World of Privacy Forum, Washington, DC. Development Report 2016, World Bank, Washington, Ponemon Institute. 2014. “Cost of Data Breach Study: DC. Global Analysis.” Study for IBM. http://public.dhe ————. 2015b. “Spectrum Management in Guatemala: ..com/common/ssi/ecm/se/en/sel03027usen The Big Bang Reform.” Companion paper for the /SEL27USEN.PDF. World Development Report 2016, World Bank, Washing- PRIF (Pacifi c Region Infrastructure Facility). 2015. ton, DC. “Economic and Social Impact of ICT in the Pacifi c.” Markovitch, Steven. 2013. “U.S. Broadband Policy and http://www.theprif.org/index.php/resources Competitiveness.” Background paper, U.S. Council on /document-library/121-prif-ict-study-report-2015. Foreign Relations, Washington, DC. Qiang, Christine Zhen-Wei. 2010. “Broadband Infrastruc- Mathison, Stuart, Larry Roberts, and Phil Walker. 2012. ture Investment in Stimulus Packages: Relevance for “The History of Telenet and the Commercialization of Developing Countries.” Info 12 (2): 41–56. Packet-Switching in the United States.” IEEE Commu- Rijkers, Bod, Caroline Freund, and Antonio Nucifora. nications Magazine 50 (5): 28–45. 2014. “All in the Family: State Capture in Tunisia.” Mayer, Rebecca, Ken Figueredo, Mike Jensen, Tim Kelly, Policy Research Working Paper 6810, World Bank, Richard Green, and Alvaro Federico Barra. 2008. Washington, DC. “Costing the Needs for Investment in ICT Infra- Ros, Agustin J., and Aniruddha Banerjee. 2000. “Telecom- structure in Africa.” Africa Infrastructure Country munications Privatization and Tariff Rebalancing: Diagnostic Study, Background Paper 3, African Evidence from Latin America.” Telecommunications Development Bank Group, Abidjan, Côte d’Ivoire. Policy 24 (3): 233–52. Meeker, Mary. 2015. “Internet Trends 2015: Code Confer- Samarajiva, Rohan. 2010. “Leveraging the Broadband ence.” http://www.kpcb.com/internet-trends. Telecom Model to Bring Broadband to the People.” Minges, Michael. 2015. “Exploring the Relationship Information Technologies and International Development between Broadband and Economic Growth.” Back- 6: 93–97. 238 WORLD DEVELOPMENT REPORT 2016

Sandvine. 2014. “Global Internet Phenomena Report.” Waldfogel, Joel. 2011. “Bye, Bye Miss American Pie? The https://www.sandvine.com/downloads/general Supply of New Recorded Music since Napster.” NBER /global-internet-phenomena/2014/1h-2014-global Working Paper 116882, National Bureau of Economic -internet-phenomena-report.pdf. Research, Cambridge, MA. Schwartz, Paul, and Daniel Solove. 2014. “Reconciling WEF (World Economic Forum). 2014. “Delivering Digital Personal Information in the United States and the Infrastructure: Advancing the Internet Economy” 7 European Union.” California Law Review 102: 877–916. (April). http://www3.weforum.org/docs/WEF_TC Senior, Dan, and Saul Singer. 2011. Start-Up Nation: The _DeliveringDigitalInfrastructure_InternetEconomy Story of Israel’s Economic Miracle. New York: Hachette _Report_2014.pdf. Book Group. Wellenius, Bjorn. 2002. “Closing the Gap in Access to Symantec. 2013. “The Norton Report 2013.” https://msisac Telecommunications: Chile, 1995–2002.” Discussion .cisecurity.org/resources/reports/documents/b Paper 430, World Bank, Washington, DC. -norton-report-2013.pdf. Wellenius, Bjorn, and Peter Stern. 1995. Implementing Tokushima, Yutaka, and Hiroya Tanaka. 2015. Economic Reforms in the Telecommunications Sector: Lessons from Development Using an Enabling Environment for Con- Experience. Washington, DC: World Bank. textualized Innovation: The Case of the Poverty Reduction West, Darrell M., ed. 2015. Digital Divide: Improving Inter- Project by Building-Up the Innovation Environment Using net Access in the Developing World through Aff ordable Ser- FabLab. Bohol, Philippines: Fablab Asia Foundation. vices and Diverse Content. Washington, DC: Brookings UN Broadband Commission. 2013. “Planning for Prog- Institution. ress: Why National Broadband Plans Matter.” http:// Wheeler, Tom. 2015. “This Is How We Will Ensure www.broadbandcommission.org/documents/report Net Neutrality.” Wired online edition, February NBP2013.pdf. 4. http://www.wired.com/2015/02/fcc-chairman ————. 2014. “The State of Broadband 2014: Broadband -wheeler-net-neutrality/. for All.” http://www.broadbandcommission.org White House. 2012. “Consumer Data Privacy in a Net- /Documents/reports/bb-annualreport2015.pdf. worked World: A Framework for Protecting Privacy ————. 2015. “The State of Broadband 2015: Broadband as and Promoting Innovation in the Global Digital a Foundation for Sustainable Development.” http:// Economy.” http://www.whitehouse.gov/sites/default www.broadbandcommission.org/documents /fi les/privacy-fi nal.pdf. /reports/bb-annualreport2015.pdf. World Bank. 2014. “Comoros Policy Notes: Accelerating UN General Assembly. 2013. “The Right to Privacy in the Economic Development in the Union of Comoros.” Digital Age. Resolution 68/167, Adopted on 18 Decem- http://wbdocs.worldbank.org/wbdocs/viewer ber 2013.” http://www.un.org/ga/search/view_doc /docViewer/index1.jsp?objectId=090224b08249c087 .asp?symbol=A/RES/68/167. &standalone=true&respositoryId=WBDocs. UN Population Division. 2014. Population and Develop- ————. 2015. World Development Indicators 2015. http://data ment Database, http://www.un.org/en/development .worldbank.org/products/wdi. /desa/population/publications/development/popula ————. Forthcoming. “Infrastructure Sharing Toolkit.” In tion-development-database-2014.shtml. Broadband Strategies Toolkit. http://broadbandtool kit UNCTAD (United Nations Conference on Trade and .org/. Development). 2015. Information Economy Report: World Bank and Korean Development Institute. 2015. The Unlocking the Potential of E-Commerce for Developing Korean Digital Governance and Performance Experience: Countries. Geneva: UNCTAD. Lessons and Experience. Washington, DC: World Bank. Ungerer, Herbert. 2000. “Access Issues under EU Regula- World Bank and Second Muse. 2014. “Big Data in tion and Antitrust Law: The Case of Telecommunica- Action for Development.” http://live.worldbank.org tions and Internet Markets.” Research paper, Weath- /sites/default/files/Big%20Data%20for%20Develop erhead Center for International Affairs, Harvard ment%20Report_fi nal%20version.pdf. University, Cambridge, MA. Wu, Tim. 2010. The Master Switch. New York: Alfred A. Knopf.

240 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 4

Smart cities

Cities are where problems and solutions meet. They “fancy” technology over less exciting but more impor- are our centers of commerce and innovation—and tant foundational investments. the gateways to the global economy. With the world While evidence of sustained impact remains elu- rapidly urbanizing, cities will only grow in impor- sive, governments allocate signifi cant sums on smart tance; the urban population of the developing world city projects, including in the developing world. is expected to double between 2000 and 2030, adding China has launched a reported US$70 billion “smart 2 billion city dwellers. This creates urgency to get our city” credit line and a US$8 billion investment fund.2 cities “right” because the global response to our most India is seeking to leverage its homegrown informa- pressing challenges—from climate change to rising tion technology (IT) industry to construct 100 smart inequality—will likely succeed or fail in cities. cities, with a fi rst-year budget of US$1.2 billion. 3 Rio Innovative urban leaders have begun to tap into de Janeiro implemented a fi rst-of-a-kind Intelligent new streams of data on the state and performance Operations Center, bringing together data from of their cities, often in real time, to realize a forward- over 30 agencies and services in a command center looking vision of a “smart city”—a city that leverages like that of the U.S. National Aeronautics and Space the latest in technology and connectivity to make Administration (NASA). Even resource-constrained better decisions and achieve the urban aspirations cities in Africa are embarking on the smart city jour- of its residents.1 Specifi cally, smart cities collect lots ney, such as in Konza Techno City near Nairobi. of data through instrumentation, bring these data Pragmatically, old and new cities alike have begun together through integration, and then analyze the to incorporate smart technologies into the everyday integrated data for intelligence on how to improve the fabric and complexities of their existing urban cen- city’s services and (fi gure F4.1). Smart ters to drive greater effi ciencies in city operations; cities use this three-step approach to tackle everything provide a platform for innovation at a citywide scale; from reducing traffi c congestion to fi ghting crime to and promote social inclusion through heightened improving air quality. accountability, citizen empowerment, and smarter Despite widespread enthusiasm, however, dis- decision making. cussions held under the World Bank’s Urbanization Knowledge Partnership indicate that most city Effi ciency—to do “more leaders struggle to understand how to best invest in intelligent infrastructure and connectivity to deliver with less” long-term value. In addition, the concept of a smart By collecting large amounts of data and then trans- city has grown somewhat controversial. Proponents lating these data into insights, cities are able to boost argue that smart city innovations offer a genuine rev- the effi ciency and responsiveness of their operations. olution in city management. Skeptics see empty hype Data help cities better match the supply of public that risks wastefully distorting the investments of services with real-time needs and uncover emerging resource-constrained governments as they prioritize problems before they turn into crises. Smart city tech- This sector focus was contributed by Dennis Linders. nologies make this possible in several ways. Automated SMART CITIES 241

Figure F4.1 Smart cities: From data to intelligence

INCLUSION Instrumentation Integration Intelligence Collect (a lot of) data. Connect and bring these data Analyze integrated data together from across the city. for insights and trends to EFFICIENCY make smarter decisions. INNOVATION

Sensors (examples: traffic, water, energy) Systems (example: building automation) “Internet of things” Big data analytics Ubiquitous connectivity Predictive analytics Satellites (example: weather patterns) Smart grid Data-driven optimization Society (example: social media)

Source: Adapted from Palmisano 2008.

optimization translates data from cameras, sensors, Governments can facilitate collaboration through and anonymized cellphone records into intelligence three platforms. Through local open data, cities share to, for example, help optimize traffi c fl ows in real local data with the public, promoting transparency, time. Predictive analytics uses such data to track and accountability, and collaborative problem solving. predict everything from rainfall to crime hot spots to Through “living labs,” governments designate parts possible landslide areas. Evidence-based decision making of the city as test beds to collaboratively pilot new and planning can continuously monitor milestones ideas. Through urban analytics centers, cities partner and targets to ensure cities can quickly take corrective with local universities and industry to seed cross- actions as needed to achieve their goals. disciplinary research centers with systematic access to local urban data. Innovation—through collaboration at city scale Inclusion—to ensure everyone benefi ts Most smart city innovations have their origins in the private sector. Engineers argue that a city is essen- City leaders should focus smart city efforts on the tially a of systems, and each of these needs of all residents. Three valuable urban practices systems generates data that can be analyzed to make in the developing world are worth noting. First, use data cities smarter. But for individual “smart systems” to to target the most vulnerable, as São Paulo did by devel- add up to a “smart city,” innovations must be on a oping a comprehensive geographic database of socio- citywide scale. That requires contributions and ideas economic and physical indicators to prioritize housing not just from commercial fi rms but also from gov- and investments. Second, open up data ernments and citizens through public-private-people to promote accountability, including grassroots initia- partnerships. tives such as the mapping of facilities, pollution, and Open data, social media, and cellphones enable community needs in Kibera, Nairobi’s largest informal governments, fi rms, and citizens to exchange vast settlement. Third, tap mobile connectivity to expand civic amounts of information at virtually no cost—making participation, as cities in the Philippines have done for it far easier to share knowledge and ideas that are dis- participatory budgeting and for crowdsourcing the tributed throughout society. These tools also enable identifi cation of smoke-belching vehicles. real-time collaboration, enabling governments to view their citizens and fi rms not just as passive customers The road ahead—the city as of public services, but as key partners in innovative problem solving. Amsterdam and Singapore have a laboratory begun to use this collaborative approach to bring The explosion of urban data will slowly give rise to together citizens, businesses, and government to a “science of cities.” City leaders can speed up this experiment with innovations to lower their city’s envi- process by turning their cities into laboratories for ronmental footprint, one neighborhood at a time. smart innovations that translate local experiments 242 WORLD DEVELOPMENT REPORT 2016

into global knowledge, and global knowledge into halls everywhere should bring together governments, local solutions. fi rms, and the public to jointly design a vision for the This exchange of ideas will no longer fl ow only future of their city with long-term goals and short- from the West to the rest, since rapid urbanization term priorities to guide the development of smart city enables cities in the developing world to leapfrog projects that deliver value for all the city’s residents. their global peers by incorporating smart from the start in ways that are cheaper and easier than retrofi t- ting legacy infrastructure. Cities in the “global south,” Notes from Rio to Shanghai, often push the boundaries of 1. Townsend 2013. innovation as aggressively as their peers in the “global 2. People’s Daily, “China Prepares for Smart City Con- north.” In fact, smart cities may be the fi rst instance struction Boom,” October 16, 2013, http://en.people of a large-scale urban transformation that is truly .cn/90882/8426660.html. global in origin, with innovations and ideas fl owing 3. Times of India, “Government to Set Up 100 Smart in all directions simultaneously—and with pockets of Cities,” July 23, 2014, http://timesofi ndia.indiatimes .com/india/Government-to-set-up-100-smart-cities innovation appearing in unexpected places. /articleshow/38919516.cms. Accelerating this progress will require actions at all levels. Global institutions like the World Bank can facilitate the sharing of experiences. Cities can work References together to establish open standards to avoid vendor Palmisano, Samuel J. 2008. “A Smarter Planet: The Next “lock-in” and to make it easier to share solutions—a Leadership Agenda.” Speech delivered at the Coun- community-developed app in, say, Chicago, can then cil on Foreign Relations, New York, November 6. be rapidly deployed in Chengdu and Caracas. Local http://www.cfr.org/technology-and-foreign-policy governments can address the often fragmented struc- /smarter-planet-next-leadership-agenda/p17696. ture of their bureaucracies and outdated procurement Townsend, Anthony M. 2013. Smart Cities: Big Data, Civic systems that are incompatible with the design and Hackers, and the Quest for a New Utopia. New York: implementation of integrated IT systems. And city W. W. Norton.

244 WORLD DEVELOPMENT REPORT 2016

ENABLING DIGITAL DEVELOPMENT Th e data revolution

A report commissioned by the United Nations, A Open data are those that are freely and easily acces- World That Counts, proclaims, “Data are the lifeblood sible, machine-readable, and explicitly unrestricted of decision-making and the raw material for account- in use. Open data aren’t necessarily big, and big data ability. Without high-quality data providing the right aren’t necessarily open. Governments are, or could information on the right things at the right time, be, important sources of data on population, public designing, monitoring and evaluating effective poli- budgets, education and health facility usage and sta- cies becomes almost impossible.”1 tus, weather, and trade. When opened, these data can Recognizing the potential for harnessing the ongo- be combined and recombined in ways that directly ing explosion of data, but mindful of gaps between benefi t the public (for instance, by increasing the the developing and developed world, the report calls transparency and accountability of government) and for a data revolution that would aid in the achieve- provide the basis for commercial, value-added services ment of the Sustainable Development Goals. (such as apps for navigating public transit). Box S5.1 The world is witnessing an unprecedented explo- presents some examples of both big and open data. sion of data. Digital data overtook analog around 1998, Exuberant estimates of the current and potential

SPOTLIGHT 5 and in 2013 amounted to 46 billion trillion bytes. That’s economic value of big data and open data range from equivalent to about 400 trillion printed copies of this the hundreds of billions to the trillions of dollars per Report, which when stacked would reach from Earth year. The clearest benefi ts so far have been for map to well beyond Pluto. Figure S5.1 shows the ongoing and weather data. Openly available global positioning upsurge of data, as measured by total storage capacity. system (GPS) data supported markets for geospatial Figure S5.2 shows the evolution of telecommunica- data and services worth US$56 billion in 2013 in the tions capacity. Although the absolute gap between United States alone.3 At least four companies valued higher- and lower-income countries is increasing, at over US$1 billion—Zillow, Zoopla, Waze, and the growth in the latter has been faster since 2008. Climate Corporation—process and resell open data In harnessing this data explosion for development, about real estate, traffi c conditions, and weather. attention focuses on two overlapping innovations: Yet sustained, impactful, scaled-up examples of “big data” and open data. Big data are voluminous big and open data in the developing world are still rel- or fast. They come, for instance, from satellite and atively rare. Open data has far to go. Figure S5.3 shows ground sensors and as by-products (“data exhaust”) that readiness, implementation, and impact of open from electronic transactions and from mobile phone data are all highly correlated with GDP per capita, but calls. The promise of big data is to provide informa- that there are shortcomings in high-income coun- tion of unprecedented scope, detail, or rapidity. For tries, as well. instance, Global Watch crunches massive Data impact requires willing suppliers and eager amounts of open satellite data in order to generate demanders. On the supply side, private holders of near real-time, global maps of tropical deforestation.2 data may be reluctant to share it for fear of jeopardiz- WDR 2016 team, incorporating contributions by Amparo Balli- ing customers’ privacy or corporate competitiveness. vian and Thomas Roca (2015). Yet some of these data, if pooled and shared, could benefits forall. public, private,andacademicactorsthatincorporates that seektocreateasymbioticrelationshipamong generate publicbenefi Source: in optimallycompressed bytes information S5.1 Figure Technological capacity zettabyte Source: independent cross-check. http://www.pricestats.com. are more timelythanoffi ing price datafrom theweb. computes dailyinflation datafor 22economies by scrap- Real-time independentmeasures ofinflation. .worldpop.org.uk/nepal/. the 2015 earthquake to aidinrelief eff used itto estimate populationmovements following worked outprocedures for accessing thisdataandhas a Swedish nongovernmental organization (NGO), has cal, confi dentiality, orcompetitive reasons. Flowminder, operators are often reluctant to share thisdata for techni- real-time information onpopulation,butcellphone Cellphone locationdata canprovide comprehensive, displaced populationsfor effi earthquake. Informing relief eff Box S5.1 exabyte Hilbert 2015. Data at http://bit.do/WDR2016-FigS5_1. at Data 2015. Hilbert 10 10 10 10 10 10 10 16 17 18 19 20 21 22 WDR 2016 team. 2016 WDR 1986 2.6 exabytes Acriticalneedindisaster relief isto track

1990 tostore capacity World’s Total 16 exabytes “Big data” and open data inaction data open and data” “Big

1994 ts. Thereareemergingmodels orts inthewake oftheNepal

cial numbers, andprovide an 1998 54 exabytes Analog in lgsis planning. logistics cient 309 exabytes

These infl ation statistics 2002

2006 orts. http://www orts. 4.6 zettabytes Digital 2010 PriceStats

2014 ment isvaluableforthecompanies—ithelpsthem ments. Dataonaveragevehiclespeedbyroadseg- unwilling tosharedetaileddataonvehiclemove- in optimallycompressed kbps capacity telecommunications Fig Note: Source: est available price. http://med-db.medicines.sadc.int/. member country, ifmedicineswere purchased atthelow- on theweb, andestimates thepotential savings, for each African Development Community. Itpublishesprice data ceutical procurement volume andprices intheSouthern (SARPAM) InfoHub assemblesinformation aboutpharma- Programme onAccess to MedicinesandDiagnostics in SouthernAfrica. Promoting effi cient procurement ofpharmaceuticals ized. http://subsidiosalcampo.org.mx/. and otherfi nancial fl ows, andallows datato bevisual- iest 10percent ofrecipients. Awebsite now tracks these that 57 percent ofthebenefits were goingto thewealth- large PROCAMPOsubsidyprogram. Thedatashowed Mexican Ministry ofAgriculture to openitsdataonthe Analysis andResearch, aMexican NGO, persuadedthe Accountability for subsidies inMexico.

For instance,taxicompaniesmayjustifi Telecommunications capacity kbps = kilobits per second. per =kilobits kbps ure S5.2 10,000 12,000 14,000 Hilbert 2015. Data at http://bit.do/WDR2016-FigS5_2. at Data 2015. Hilbert 2,000 4,000 6,000 8,000 0 2003 Δ ≈91 ≈

2004 100 Growth in 9 2003

2005

KBPS KBPS TheSouthernAfrican Regional KBPS ≈11 2006 High-income 890 Δ ≈890

2007 ≈ 940 51

2008 THE DATA REVOLUTION 2007

KBPS KBPS

2009 KBPS 10,900 Δ ≈10,900 ≈18 Fundar Center for ≈

2010 12,400 1,500 2013

Rest 2011

KBPS KBPS

ably be ably 2012 KBPS ≈ 8 2013

SPOTLIGHT 5 245 246 WORLD DEVELOPMENT REPORT 2016

Figure S5.3 Readiness, percent of the remainder had made little or no prog- implementation, and impact of ress in opening map data. One reason is that cash- open data strapped agencies support themselves by selling data. This shuts out many potential users, even though the 100 GBR additional cost of providing the data to them would 90 NZL be negligible. A dramatic example of reversing this USA trade-off is the decision of the U.S. government to 80 NLD SWE reduce the cost of a digital Landsat satellite image 70 CHL AUS from US$600 to zero in 2008. The annual number of BRA DEU 60 CZE ESP downloads increased from fewer than 25,000 to 3 mil- ITA 50 PER DNK lion. The direct benefi ts are estimated at US$1.8 billion POL PRT EST RUS ECU HUN IRL AUT a year for U.S. users and US$400 million for interna- 40 MYS CHE 4 RWA SGP tional users; there are substantial indirect benefi ts as 30 KAZ COL well. The loss of from selling the

Implementation-scaled VNM BEL CHN 20 UGA THA KEN ARE TUN 25,000 images was miniscule by comparison. So if the BEN MAR TZA EGY fi xed costs of data maintenance can be fi nanced, the 10 PAK YEM QAT benefi ts from free distribution can be large. CMR NGA 0 MMR MLI There are other reasons for public agencies to be 01020 30 40 50 60 70 80 90 100 reluctant to share. Agencies may lack technical skills Readiness-scaled to share data. They may be unwilling to expose the data to scrutiny because of quality concerns. They Impact-scaled Income 0 Low-income may value the political leverage afforded by monop- 20 Lower-middle-income olizing data. Nonetheless, both the G-8 countries and 40 Upper-middle-income the African Union’s High-Level Conference on the 60 High-income Data Revolution have endorsed the concept of mak- 80 ing offi cial data open by default. 100 for data may also be weak, even

Source: WDR team based on 2014 data from World Wide Web Foundation if there is underlying interest. Of course, accessibility 2015. http://barometer.opendataresearch.org/report/about/data.html. to the internet is a prerequisite. Outdated or poor- Data at http://bit.do/WDR2016-FigS5_3. quality data will discourage demand. Because data lit- eracy is critical, intermediaries can play a key role in with dispatching—and for city managers it provides interpreting, analyzing, and disseminating open data.

SPOTLIGHT 5 a useful gauge of congestion. The World Bank is But capacity or incentive to do this is often lacking. exploring ways to provide companies with software There has not been as much uptake by journalists as that distills their data down to average vehicle speed. some have hoped. And while a benefi t of open data This summarized data could then be shared between is enhanced sharing among government agencies, companies and with the public, yielding more accu- relatively few governments are making optimal use rate and comprehensive estimates of travel speeds. of existing statistics for planning purposes, according In Australia, a tech startup called The Yield is work- to PARIS21. However, Bangladesh, Cambodia, and ing with public sector regulators to test ways to help Niger scored high on statistical capacity (according to farmers share data for the common good. The $A 100 PARIS21’s IDR Metadatabase), showing that it is pos- million Australian oyster industry loses nearly $A 34 sible for low-income countries to excel at generating million a year due to harvesting closures by regulators and using statistics.5 to protect human health when runoff is contaminated. The pioneering Kenyan Open Data Initiative The idea is to equip the oyster farms with real-time shows both the promise and challenges of open data. sensors on water quality. The data will help individual A champion, Permanent Secretary Bitange Ndemo, farmers with farm management, while allowing regu- persuaded ministries to open their data, and the lators to optimize decisions, avoiding costly, unneces- program was inaugurated in 2011. The World Bank, sary closures and still maintaining safety. Google, and Ushahidi provided technical and fi nan- Public agencies are also reluctant to share data, cial support. Drawing on Nairobi’s vibrant tech scene, even when it has large public benefi ts. For example, the ancillary Code4Kenya initiative spurred the of the 86 countries surveyed by the Open Data Barom- development of apps that presented health, safety, eter, one-third of the high-income countries and 85 and other information in engaging and useful ways. THE DATA REVOLUTION 247

A journalist, Irene Choge, attended a Data Journalism Notes Boot Camp. She then used the Open Data Platform to trace a chain of links: funding for school toilets had 1. Independent Expert Advisory Group Secretariat gone missing, so the student-to-toilet ratio for many 2014. 2. http://www.globalforestwatch.org/. schools was intolerably high, leading to open defe- 3. http://gpsworld.com/the-economic-benefi ts-of-gps/. cation; consequently students were suffering from 4. Miller and others 2013. illnesses associated with poor sanitation, and female 5. http://datarevolution.paris21.org/metabase. students, especially, were discouraged from attend- 6. https://www.youtube.com/watch?v=A58R2yNQtio. ing school; the end result was poor performance on tests. The fi ndings prompted government action.6 Despite this promising beginning, the Open Data References Initiative lost steam. Data were not updated, the apps Ballivian, A., and Thomas Roca. 2015. “New Data for struggled to fi nd users, and a legal framework for Development.” Background paper for the World Devel- open data was not in place by mid-2015. Kenya’s score opment Report 2016. World Bank, Washington, DC. on the Open Data Barometer dropped from 43 to 26 Hilbert, Martin. 2015. “Quantifying the Data Deluge and between 2013 and 2014. the Data .” Background note for the World This example underscores the global need for sus- Development Report 2016. World Bank, Washington, tained commitment to openness and for investment DC. Independent Expert Advisory Group Secretariat. 2014. in high-quality data if the dream of a data revolution A World That Counts: Mobilizing the Data Revolution for is to be realized. Countries can seek ways to discour- Sustainable Development. New York: United Nations. age data hoarding, by adequately funding data hold- Miller, Holly M., Leslie Richardson, Stephen R. Koontz, ers and showing them that they can attain a higher John Loomis, and Lynne Koontz. 2013. “Users, Uses, profi le by opening their data. Donors and the interna- and Value of Landsat Satellite Imagery—Results tional community can support these actions through from the 2012 Survey of Users.” Open-File Report funding, capacity building, and ensuring that their 2013–1269. U.S. Geological Survey. http://pubs.usgs own data are open. .gov/of/2013/1269/. SPOTLIGHT 5 CHAPTER 5 National priorities

Analog foundations for a parts: one that can be automated, and one that cannot. The automatable part of the task consists of repeat- digital economy able, routine activities that produce measurable Estonia is one of a handful of countries closest to outputs and outcomes, and therefore are amenable becoming a digital society. After regaining its indepen- to automation. Tasks performed by, say, bank tellers, dence in 1991, it decided to promote the use of digital bookkeepers, or clerks—and services such as regis- technologies in all areas of society and the economy. tration and licensing—can to a large extent be done The nation was not rich by European standards, so one with digital technologies. In contrast, many tasks per- objective was to take advantage of effi ciency gains. formed by teachers, researchers, or managers—and While investing in information and communication services such as policing or those performed in health technologies (ICTs), Estonia also greatly improved its care—involve activities where the providers must business climate, human capital, and governance. The exercise considerable judgment in deciding what to greater ease of doing businesses spawned a host of do or how to respond. ICT projects often fail when technology-intensive startups, including Skype and they focus solely on technology without also address- TranferWise, a company disrupting the money trans- ing shortcomings in the complements that cannot be fer industry. The country today ranks high in the PISA automated. The line that divides automatable activi- (Programme for International Student Assessment) ties from those that are not is, of course, continuously educational rankings and invests in digital literacy for shifting. But solving the most diffi cult development its older citizens. The greater accountability in gov- problems will almost always require more than just ernment boosted it from 78th to 40th between 1996 technology. and 2010 on a ranking of 144 countries on their control Chapters 1−3 of this Report presented evidence of of corruption. Today, have access to 3,000 the problems that can arise from greater use of tech- e-government, e-banking, and other services, saving nology and identifi ed the main risks that countries each of them an average of 5.4 work days a year.1 face as a consequence. To mitigate these risks, invest- Estonia demonstrates that even small and devel- ments in digital technologies must be accompanied oping or transitioning countries can seize the oppor- by improvements of their analog complements tunities the internet offers by implementing a smart (fi gure 5.1): and comprehensive digital development strategy. Many other countries have also eagerly invested • Lower-than-expected internet adoption means that in digital technologies but have failed to create the many fi rms in low- and middle-income countries environment for it to support development. So, while forgo considerable productivity benefi ts. The rea- the internet has spread rapidly, development has sons include a poor business climate and vested advanced much more slowly. Chapter 4 discussed interests that hinder market entry and reduce the policies that ensure universal, affordable, safe, and pressure to innovate. Without improved regula- open access to the internet. These supply-side policies tions, especially those governing competition, econo- are critically important, but are not enough. mies of scale brought by the internet could well lead Why is technology by itself unlikely to solve to harmful concentration and monopolies—and thus persistent development problems? The key insight to greater divergence between and within countries, is that a typical task in development has two broad rather than convergence and catching-up. NATIONAL PRIORITIES 249

• Rising shifts of income from labor to capital and Figure 5.1 Risks from digital technologies in the the drop in mid-level jobs in many countries sug- absence of complements gest that the gains from greater use of technology may not be equitably shared. Without complemen- DIGITAL tary investments in the skills that workers need to TECHNOLOGIES leverage the internet, automation could exacerbate inequality rather than promote greater opportunity and shared prosperity. • E-government projects have a poor record in many countries; governments have too often been unable, and sometimes unwilling, to use the inter- Scale Automation Information net to improve service delivery and increase public without competition without skills without accountability participation. Without more capable and more accountable public sector institutions, technology investments could lead to more control and not to CONCENTRATION INEQUALITY CONTROL greater empowerment of citizens.

Business regulations that ensure a high degree Source: WDR 2016 team. of competition, skills that prepare workers for the 21st-century workplace, and accountable public insti- tutions that use the internet to improve services— more inclusive. The internet thus tends to amplify together these constitute the analog foundation that existing strengths and weaknesses, and progress supports an inclusive, effi cient, and innovative digi- in these areas therefore becomes even more urgent. tal transformation. Otherwise, there is a risk of falling further behind The idea that technology alone is not enough to for those who do not make the necessary reforms. As have a widespread impact on development is sup- the following sections will show, technology does not ported by a growing academic literature. Research- just increase the value of complements. It can also ers focus on different angles, but come to similar often raise their quality—for instance, through online conclusions (box 5.1). In a growth context, the level business registration programs, online education, or of analysis is a country or sector, and what matters better monitoring. is how technology interacts with rules—that is, with So how can businesses, people, and governments regulations and similar endowments. In a labor mar- reap the greatest digital dividends? The Report ket context, the level of analysis is occupations or suggests many detailed steps to strengthen analog specifi c tasks, and what matters is the interaction of complements, but policy makers should also keep in automation and worker skills. And in a government mind a set of overarching principles concerning the context, the level of analysis is a public service, and interaction of technology and complements. what matters for delivery is the accountability of institutions. Guiding premises

The inability to scale up is one of the most The interdependence commonly cited reasons for failed digital between technology and interventions complements The popular media is fi lled with inspiring stories involving technology: digital marketplaces are Technology and its complements feed on one another. helping artisans in remote hillsides to access global Improvements in the business climate, human capi- markets; the on-demand economy is creating new tal, and public sector governance are already high on startups and digital entrepreneurs; and social media the reform agenda in most countries. With greater tools are helping citizens organize and rally around use of technology, the complements also become issues of common concern. But after a promising more valuable because they interact. Thanks to tech- start, many initiatives seem to fl ounder. For every nology, fi rms operating in a well-regulated business successful online commerce platform, nearly four environment are more innovative. Skilled workers fail to achieve scale. Despite the growing popularity are more effi cient. And accountable institutions are of the on-demand economy, the fi rm startup rate in 250 WORLD DEVELOPMENT REPORT 2016

Box 5.1 Three ideas about the interaction between technology and its complements

The relationship between technology and complements can be done by computers faster, better, and more cheaply. that is at the core of this Report is also the topic of recent With increasing computing power and better software, work by a number of prominent researchers. the rule of thumb is that if you can fairly easily explain a job to someone, it can probably be automated. Tasks that Technology and rules (competition). Paul Romer develops require problem solving, adaptability, and creativity, in a simple growth model in which technology interacts with contrast, are very hard to automate. Again, the diffi culty in “rules”—including, importantly, regulations shaping the automating a job tends to match the diffi culty in describing business climate. Technology is tradable and can spread it. Autor stresses the interaction of technology with non- to new places quickly. Rules are established locally. Even if routine tasks. Nonroutine jobs cannot easily be automated, they are negotiated internationally, they will not be simply but they can benefi t from automation because technology imported like, say, capital goods. So there will always be tends to complement skills. Workers in nonroutine jobs get a complex, context-specifi c political process that leads to more productive as the automation of other tasks increases. their adoption and determines how eff ectively technology is deployed. Together, technology and rules enable the cre- Technology and discretion (accountability). Lant Pritchett, ation or adoption of new “ideas”—new or improved ways of Michael Woolcock, and Salimah Samji focus on public ser- producing, working, or serving citizens. The quality of rules vices and develop a typology of organizational capabilities varies across countries. Romer cites the municipal water for specifi c public sector tasks. They do not specifi cally look sector privatization to a French company in Argentina to at the role of technology, but their framework provides illustrate his argument. The program was initially success- insights about when technology could improve services. ful, but—like in many other countries—eventually ran into Informed by the accountability framework in the 2004 popular opposition, which prompted renationalization. World Development Report, Making Services Work for Poor The technology required and the rules and preferences People, their main criterion is whether a task involves local about regulating utility monopolies were present in France discretion and therefore some judgment by a civil servant. but not in Argentina. Technologies and rules were thus Others are whether a task is transaction-intensive, and thus poorly matched. In contrast, ideas about setting up and involves many people; whether it is a service (education) or operating mobile phone systems spread quickly to many an obligation (policing); and whether it can draw on existing countries—even to fragile states like Somalia. Mobile phone knowledge or requires innovation. Categorizing public ser- technology, in contrast to landlines or municipal water, vice tasks in this framework helps explain why e-government easily allows competition among providers even in places eff orts often fail—when organizational capabilities are poorly with low regulatory capacity. For mobile, local conditions matched to technology and the requirements of the task. (or rules) have accommodated the new technology. These studies make very similar arguments. First, tech- Technology and skills. David Autor is most concerned nology alone will not be enough. Second, increasing the with eff ects of technology on labor markets. Looking at use of technology needs to be matched by complementary occupations and even individual tasks, he considers how reforms. And third, knowing the best combination of tech- much of a person’s job is routine and could be done by a nology and complements for diff erent tasks, occupations, machine—and how much is nonroutine and not easily auto- and services is the key to translating technological progress mated. Those tasks that follow easily codifi able procedures into development.

Sources: Romer 2010; Autor 2014; Pritchett, Woolcock, and Samji, forthcoming; World Bank 2004.

the United States continues to decline, with the larg- The internet is no shortcut to development, est decline in the retail and service sectors.2 And as but it can be an accelerator the Arab Spring showed, citizen mobilizations based For policy makers, using technology to solve a prob- on social media are as easily crushed by governments lem is often more tempting than fi an ailing as they were stitched together by citizen activists. All institution. A manifestation of this is the popularity this because it is easy to scale up the technology, but of distributing laptops to schoolchildren as an elec- diffi cult to improve its complements. tion pledge. There is also growing demand to wire NATIONAL PRIORITIES 251 schools, hospitals, and all government buildings. A digital strategy is more about the Connectivity is important, but it is not enough. adoption of ICTs than their production Technology can rarely bypass or substitute for other Many policy makers wish to have a Silicon Valley in shortcomings. When technology is deployed in an their country, dotted with large ICT fi rms like Apple, environment of weak complements, the gains will be Facebook, Google, Huawei, and Samsung. But the limited. But when technology and its complements widespread adoption of ICTs, not their production, work together, the impacts can be profound and is responsible for most of the growth and job cre- lasting. While the internet is no shortcut, it can be an ation.3 The ICT sector in developing countries is fairly accelerator because many digital tools improve the small, accounting for only 3–4 percent of GDP and complements. 1 percent of the total labor force. But when ICT fi rms enter traditional sectors such as retail, banking, man- Understanding the interaction between ufacturing, and transport, they spur competition, technology and human complements raise productivity, and expand opportunities. For should guide how much to invest in each example, Alibaba has 30,000 employees, but provides Some sectors, occupations, and services are more a platform for more than 10 million ancillary jobs. amenable to technology than others (fi gure 5.2). Gains Uber has a few hundred coders, but it supports the from technology will be larger in public services such livelihoods of around 1 million on-demand drivers. as cash transfer programs and utility services—where The online outsourcing industry creates millions of outcomes can be easily monitored by citizens—than freelancing opportunities in professional services. in services that require a lot of discretion, such as And in Mexico, manufacturing fi rms subject to com- policing and management. Similarly, occupations petition from Chinese imports invest more in digital involving routine tasks will be more affected by tech- technologies, thus boosting productivity and creating nology than those requiring a high degree of judg- more employment opportunities. ment and intuition. With technological advances, more tasks could become easier to automate, so the A digital strategy needs to be broader than relative roles of technology and complements could an ICT strategy change over time. Since outcomes are easier to mon- itor for private businesses, as a rule of thumb, fi rms The digital economy transcends the ICT sector, are likely to be more successful than governments in encompassing most sectors of the economy and 4 using technology to solve problems. Thus in a weak society. Yet many governments continue to treat the institutional environment, the priority should be digital economy as a sector, with exclusive emphasis to encourage greater adoption of technology by the on developing ICT infrastructure and creating an private sector, while continuing to strengthen the information technology (IT) workforce. Our analy sis complements in the public sector. shows that a successful digital strategy needs to stand

Figure 5.2 Some services and sectors are more amenable to digital technology than others

Ease of automation Extent of Business sectors Public services complements Cash transfers, Retail and wholesale licensing, filing trade, transport, taxes, monitoring Moderate insurance and banking elections Agriculture, Utility services, Significant High hotel and restaurant, public works Sectors manufacturing, (roads), and real estate procurement services Inspection, Low Legal services financial Moderate management through digital technology

Significant Curative health, More amenable to improvements Construction policing, teaching, policy formulation

Source: WDR 2016 team. 252 WORLD DEVELOPMENT REPORT 2016

on two pillars—one digital and the other analog. The skills. Use the internet to provide information and digital pillar deals with the supply-side issues, focus- monitor service providers. ing on policies to make the internet universally acces- • Transitioning countries. Countries transitioning sible, affordable, open, and safe (chapter 4). The ana- toward universal internet use need to address log pillar aims to strengthen the demand-side policies harder problems: Build effective competition regula- by creating regulations that encourage competition, tion and enforcement capacity. Teach advanced cog- basic modern skills for all, and public institutions nitive and socioemotional skills that complement that are accountable. Countries such as Rwanda, technology. Move toward effective e-government whose vision is to become an information-rich, systems. knowledge-based society, have gradually tried to • Transforming countries. Countries transforming broaden their digital strategy to include both the digi- into digital societies need to tackle complex prob- tal and the analog foundations.5,6 lems: Regulate platform competition and privacy. Advance ICT and science, technology, engineer- A country typology for digital ing, and mathematics (STEM) skills and lifelong development learning. Move toward ubiquitous e-government As digital development proceeds from emerging services and participatory policy making. to transitioning and then to transforming, policy reforms become more complex. There are no hard- Classifying countries into various groups is neces- and-fast rules, but it is useful to consider a sequence sarily somewhat ad hoc (fi gure 5.3). The labels applied of policy priorities in line with technology’s increas- in the following sections should therefore be treated ing penetration (table 5.1): with considerable discretion. Technology adoption and the quality of complements vary not only across • Emerging countries. In countries where the digital countries but also across sectors and across fi rms in economy is still emerging and internet use is low, the same sector. So a country classifi ed as emerging the priority is to lay the foundations: Remove bar- for its government institutions could be categorized riers to internet adoption such as high duties on as transitioning for its skills and as transforming for ICT capital imports. Improve the business climate, its regulations. Some sectors in a country could be including physical infrastructure critical for online transforming while the rest of the economy is emerg- businesses with an offl ine footprint. Promote digi- ing. So the boundary lines across country groups and tal literacy and basic cognitive and socioemotional complementary factors are more indicative than fi xed.

Table 5.1 Policy priorities for emerging, transitioning, or transforming countries Stage in the digital transformation Policy goals Emerging Transitioning Transforming Regulations (competition): • Low barriers to internet • Effective competition • Critical “new economy” A business environment in adoption (including access, regulation and enforcement regulation (including platform which fi rms can leverage affordability, and basic (including ease of market competition and the legal the internet to compete and open and safe issues; trade entry) basis for private sector data innovate for the benefi t of and basic competition collection) consumers issues) Skills: Workers, entrepreneurs, • Digital literacy and • Higher-order cognitive and • Advanced ICT skills and and public servants who foundational basic socioemotional skills STEM education can take advantage of cognitive skills, and • Lifelong learning opportunities in the digital socioemotional skills world Institutions (accountability): • Adoption of informational • Effective e-government • Widespread citizen use of An accountable government services and monitoring by delivery systems, provider e-government services and that effectively uses the public sector and nonstate management, and citizen participatory policy making internet to empower its providers engagement • Social protection systems for citizens and deliver services • Increased electoral • Trust and safeguards against a changing labor market accountability privacy and security Source: WDR 2016 team. Note: ICT = information and communication technology; STEM = science, technology, engineering, and mathematics. NATIONAL PRIORITIES 253

Figure 5.3 The quality of complements and technology rises with incomes

1.0

TRATRANSFORMING FIN CHE SGP CAN 0.9 NORR NZL AUS DNK SWE NLD IRL BEL USA GBR EST SVN ISL LUX 0.8 DEU CYP FRA LTU LVA JPN PRT KOR MYS MLT AUT BRB ARE CZEZE CHLHL 0.7 HUN QAT ISR ESP MUS POL SSVK GEO HRV ROU ITA TTTO CRI BGR GGRCC MKD URYY SYC PANN JJOR 0.6 BWA SRB ARM GUY ALBB RUS LKA UKR MEX KAZ ZMB SLV TUN MMNG TUR KGZ THA COL GHA RWA ZAF SAU BHR 0.5 LSO BLZ PHL MAR BRN BEN UZB PER CHCHN SENEN VNM ARG IDN DOM LBR TJK KWT SWSWZ GAB NIC IRN

Complements NAM KEN GTM TZA ECU BRA CMR BGDGD 0.4 LAO BDI SLE NPLN IND EGY MOZMO GGMB HNDHN PRY PAK BOL MWIWI DZA MRT NGAG 0.3 ETH YEMY ZWE MLI LBY TTRANSITIONINGN I G AGAGO VEN KHMKH 0.2

HTI

0.1 EMERGINGEMERM E RG N 0 00.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Technology

High-income Lower-middle-income Upper-middle-income Low-income

Sources: World Bank World Development Indicators (various years); World Bank Enterprise Surveys (various years); Gallup World Poll (various years); World Bank Global e-Government Systems database; World Bank ID4D database; United Nations 2014; WDR 2016 team. Data at http://bit.do/WDR2016-Fig5_3. Note: Technology is measured by the Digital Adoption Index (DAI). DAI is based on three sectoral subindexes covering businesses, people, and governments, with each subindex assigned an equal weight: DAI (Economy) = DAI (Businesses) + DAI (People) + DAI (Governments). Each subindex is the simple average of several normalized indicators measuring the adoption rate for the relevant groups. Similarly, complements is the average of three subindicators: starting a business; years of education adjusted for skills; and quality of institutions. See fi gures 5.4, 5.8, and 5.10 for more details on the construction of sectoral subindexes.

Regulations: Helping to adopt digital technologies (chapter 1). So those countries that are creating a level playing fi eld for businesses connect and their businesses by embracing competition—by mak- compete ing it easier for new fi rms to enter and for existing fi rms to reorganize and exit, while preventing incum- A competitive market encourages fi rms to use the bent fi rms from acquiring monopoly power—are internet to access new markets, increase productiv- expected to perform well in a digital world. ity, and achieve scale. Firms in more contested indus- The lack of competitive pressure and absence of tries—facing more domestic or foreign competitors basic infrastructure are holding back many fi rms and higher fi rm entry and exit rates—are more likely in developing countries from taking full advantage 254 WORLD DEVELOPMENT REPORT 2016

of the internet. It is not uncommon for incumbent Figure 5.4 Regulations that encourage fi rms to secure profi ts by seeking protection through competition also facilitate higher regulatory means rather than by competing in the adoption of digital technologies open market. The problem arises when policy mak- ers oblige. There are also instances when regulations 100 are genuinely outdated or refl ect to technological change and the consequent disruption. For example, regulators around the world seem to be 80 confl icted on whether ride sharing services should be regulated as transport service companies or as soft-

ware companies. Is mobile money a banking or a tele- 60 com product? And in highly connected markets, could the absence of competition across various digital plat- forms inhibit future innovation? In some low-income countries, the lack of supporting infrastructure—such 40 as electricity, roads, ports, warehouses, distribution, and a well-functioning postal system—could hinder Number of days to start a business fi rms from investing in digital technologies. 20 0 0.2 0.4 0.6 0.8 1.0 Appropriate regulation at various levels of Technology: digital adoption Digital Adoption Index (Business) Policy reform priorities are likely to vary across coun- Sources: Doing Business database (World Bank, various years); WDR 2016 tries depending on their level of digital adoption. team. Data at http://bit.do/WDR2016-Fig5_4. Matching the country typology with appropriate Note: The Digital Adoption Index (Business) is the simple average of four regulatory reforms leads to the following taxonomy normalized indicators: the percentage of businesses with websites, the number of secure servers, the speed of download, and 3G (third-generation) (fi gure 5.4): coverage in the country.

• Emerging countries are characterized by low dig- ital adoption and protected markets. They are able and capable governments that provide good also likely to suffer from poor infrastructure and physical infrastructure and enforce business reg- weakly accountable institutions. Investing in basic ulations that promote competition. But they face physical and digital infrastructure (chapter 4); two distinct sets of problems. First, because they lowering tariffs on digital products; encouraging are early adopters of digital technologies, many adoption of low-tech, disruptive applications such of their fi rms have well-functioning but perhaps as mobile money and social media; and reducing less productive and nonscalable “legacy systems.” product market regulations are some of the policy These countries would benefi t from relaxing their priorities these countries should consider. regulatory constraints to level the playing fi eld • Transitioning countries have mixed regulations, with between incumbent fi rms and (internet) startups in some sectors open to competition. Many of their all sectors, but especially in the legacy sectors. Sec- business regulations are codifi ed, easily available, ond, given the universality of digital technologies, and increasingly carried out online. These countries these countries have witnessed rapid growth of also have fairly good physical infrastructure. Tran- digital platforms in selected sectors, some of which sitioning countries thus need to remove regulatory have achieved a dominant position in their mar- barriers across major sectors to incentivize fi rms to kets. They need to fi nd ways to encourage greater invest in more effi cient digital solutions and encour- competition across platforms: say, by eliminating age the entry of startups that can put competitive exclusivity conditions and introducing portability pressure on incumbent fi rms. When required, they and seamless transfers of data and information should establish regulatory clarity between online across platforms. and offl ine businesses within the same sector. A majority of developing countries and some devel- Emerging countries: Invest in infrastructure oped countries fall into this category. and enforce product market competition • Transforming countries are open to competition in For many lower-income countries, creating basic most economic sectors and generally have account- business-sustaining infrastructure is a priority, NATIONAL PRIORITIES 255

requiring fi nance, local expertise, and innovative Another infrastructure constraint to a digital solutions. Digital technologies can help overcome economy in emerging countries is the absence of some of these problems. Revenues from telecoms— domestic delivery systems. There were 38 countries customs duties on hardware, auctioning and man- in 2012 where less than 50 percent of the population aging spectrums, and value added tax (VAT) or had access to home postal delivery (see chapter 1). And sales taxes on services—have been one of the fastest several countries had no “last mile” domestic delivery growing sources of revenue in many developing system in place at all: Botswana, Burkina Faso, the countries. But regulatory capture by powerful tele- Central African Republic, the Democratic Republic of com fi rms could undermine this effort, as in Somalia. Congo, Gabon, Kenya, Rwanda, Swaziland, Togo, and Many other countries in Africa, however, have the opposite problem—they have used the telecom sector as a “cash cow,” taxing it so heavily as to impede its Figure 5.5 Digital products are taxed as luxury goods growth prospects. Even so, there are private sector in some countries solutions to fi nancing and managing infrastructure services using digital technologies that hold the most a. Computers and laptops promise. For example, Mobisol, a German company, 20 is supplying off-grid energy to villagers in Rwanda by combining solar energy technology with mobile 15 phone–based loan payments.7 Countries need to look at the digital economy as a 10 source of growth and jobs and not just as a source of 5 revenue. Some countries impose signifi cant import Import tariffs (%) tariffs on computers, laptops, and mobile phones to meet short-term revenue objectives (fi gure 5.5). For 0 example, the most favored nation (MFN) tariff rate Brazil Chad Chile on computers in Djibouti is as high as 26 percent. Fiji GabonLiberiaBelarus Samoa CambodiaArgentina Cameroon charges MFN tariff rates of more than 30 percent for Madagascar Mozambique mobile phones. The tariffs on using digital technolo- Congo, Dem. Rep. gies raise the costs of using digital technologies for Central African Republic Computers Laptops fi rms and households, thus delaying the emergence of the digital economy and its attendant positive b. Mobile phones effect on growth and jobs. In general, the higher 35 revenue buoyancy in the medium term can more than make up for short-term revenue shortfall from 30 reduced tariffs. 25 In addition to electricity, countries need to invest in trade infrastructure and customs administration 20 to improve physical connectivity and facilitate the 15 growth of the digital economy. Several countries 10 still have ineffi cient ports and other trade infra- Import tariffs (%) structure. Among the weak performers on trade 5 logistics in 2007, only a few signifi cantly improved 0 their trade logistics infrastructure between 2007 and h a Fiji oa s lize dos e a B 2014, including Afghanistan, Nepal, Nigeria, Rwanda, Sam Ghana babwe rbud rb uyana Pakistan m GrenadaG Jamaica Zi d Ba Ba Dominica Serbia, and Tanzania (fi gure 5.6). Countries can also Cabo VerdeBanglade improve their overall logistics performance by low- Dominican Republic ering their nontariff barriers and tackling behind- Antigua an the-border issues. Firms with improved logistics are better positioned to export their products and ser- Source: WITS 2014. Data at http://bit.do/WDR2016-Fig5_5. Note: Panel a shows most favored nation (MFN) tariff rates for computers and laptops. The Harmonized vices using online marketplaces once these platforms Commodity Description and Coding System (HS) code for computers is HS 8471.49, and the HS code for gain in importance in their region. But several coun- laptops is HS 8471.30. The MFN tariff for computers in Djibouti (not shown) is 26 percent. Panel b shows MFN tariff rates for mobile phones. For some countries, the MFN tariff rates may exaggerate the level tries continue to lag behind, including Cameroon, the of taxation, especially if they source a large part of their digital imports from Free Comoros, Eritrea, Somalia, and Sri Lanka. partners, as is the case with Chile. 256 WORLD DEVELOPMENT REPORT 2016

Figure 5.6 Infrastructure complements common anticompetitive practices, even though most countries have competition laws in place.9 Trade logistics in many countries haven’t improved in the past seven years Emerging countries should thus focus fi rst on reducing the implementation gap in enforcing exist- ing laws. Required changes for implementation could 3 SRB be handled through secondary legislation or guide- RWA NGA lines. Such changes include transparently defi ning NPL appropriate fi nes for anticompetitive behavior, grant- TZA ing suffi cient authority and investigating power to implementing agencies to fi ght anticompetitive 2 AFG

LPI in 2014 regulations in all economic sectors, reducing anti- SOM competitive practices by government agencies that facilitate cartel behavior, and measuring and commu- nicating the harmful effects of cartel agreements to 1 encourage support for competition policy.

1 2 3 Transitioning countries: Enhance LPI in 2007 competition by removing regulatory

Source: World Bank Logistics Performance Index (various years). Data at http://bit.do/ barriers and encouraging entry of internet WDR2016-Fig5_6. startups in traditional sectors Note: The LPI (Logistics Performance Index) covers six areas of logistics: customs, infrastructure, international shipments, logistics quality and competence, tracking and tracing, and timeliness. The index varies from 1 to 7. The fi gure shows only those countries with relatively weak trade logistics in Competition law and competition 2007 (index below 2.5). Observations above the yellow line indicate that countries improved their authorities logistics performance between 2007 and 2014. Anticompetitive regulations, while designed with good intentions, can be misused. Vested interests Tanzania. Online retailers in these countries thus and privileged fi rms can use them to limit entry into struggle to deliver parcels to their customers and are domestic markets; to access government subsidies, forced to invest in their own logistics solution. contracts, or land; and to insulate them from foreign Reliable online payment systems are also not competition. Such anticompetitive regulations seem available in many African and Central Asian coun- to discourage fi rms in some developing countries— tries, limiting the potential for e-commerce. While including Brazil, the Democratic Republic of Congo, most individuals in developing countries do not Ghana, Kenya, Mexico, Morocco, Tanzania, Uganda, have access to a credit card, online payment systems Vietnam, and Zambia—from using digital technolo- for businesses, such as the PayPal business account, gies more intensively (chapter 1). provide reliable alternative solutions. In several Latin Most countries have a competition authority, American countries, small businesses that do not have though many were set up fairly recently, and enforce- a bank account can use PagosOnline for payments. ment varies, especially when the state or privileged Without reliable and legal online payment systems, fi rms benefi t from market restrictions. Limited capac- fi rms have been resorting to riskier solutions, such ity and the diffi culty of proving harm to consumers as Bitcoin, as in Ghana. Not only are such activities are further barriers. A priority for transitioning coun- illegal in some countries, but when undertaken on tries is to institute strong competition laws and boost a large scale and in a weak regulatory environment, the capacity of implementing institutions to enforce they could have a destabilizing effect on the fi nancial the law transparently and effectively. Policy measures sector. include leniency programs allowing participants Emerging countries should focus on enforcing of cartels to cooperate with government bodies in existing business regulations and competition and exchange for lower sanctions—and ex post investiga- antitrust laws. Most countries have laws that defi ne tion of potential collusive practices in procurement. some degree of competition or antitrust regulation but that are seldom enforced.8 Moreover, business Disruption as a strategy to create regulations are sometimes implemented in a way competition that discriminates between fi rms, increasing regu- Transitioning countries should encourage the entry latory uncertainty. Agreements among competitors of internet fi rms offering traditional services in order on setting prices, sharing markets, and bidding for to disrupt protected sectors. The internet has created public procurement, for instance, are one of the most a new wave of entrepreneurship and innovation NATIONAL PRIORITIES 257 around the world. Internet fi rms now compete with highlights the tight-rope walk for regulators in the offl ine incumbents in banking, insurance, retail trade, digital economy (box 5.2). transport, logistics, international currency transfers, The internet fi rms operating in the fi nancial accommodation, tourism, media, publishing, adver- sector are often not subject to the same level of reg- tising, programming, design, and other professional ulations, enabling Yu’E Bao, an online trust fund in service tasks. China, to enroll 150 million depositors and collect Governments may consider abstaining from Y 700 billion in deposits in less than two years. It imposing regulatory barriers on new internet fi rms offered higher yields on the deposits by making use before they achieve a certain scale. Eliminating of underregulation of the fi nancial sector outside the protection, often against the interests of infl uential formal banking system. The ongoing reforms in the elites benefi ting from the status quo, is diffi cult and fi nancial sector in China have narrowed this regula- takes time. But internet fi rms can do the work of tory arbitrage. the regulators and introduce competition in a sec- tor overnight. The mobile money service M-Pesa in Transforming countries: Encourage Kenya was able to achieve scale because of the regu- digital innovation but also regulate the lator’s initial decision to regulate it as a telecom and digital economy to level the playing fi eld not as a banking sector product. The mobile-money between incumbent and internet fi rms schemes in other countries have often been held up by opposition from traditional banks. At the same Critical “new economy” regulations time, lack of adequate regulation on the telecom side While most of the internet’s impact will be in tradi- meant that M-Pesa enjoyed dominant market power tional sectors that use technology to boost effi ciency, in Kenya for a fairly long time, which ultimately “pure” internet fi rms raise new challenges that had to be curbed by the competition authorities. This regulatory authorities even in advanced economies

Box 5.2 Mobile money: A success story and yet a regulatory minefi eld

Mobile money started informally, with users making pay- exclusivity contracts with the agents posed an entry barrier ments of airtime credit to each other (see also spotlight for other telecom operators. One reason Safaricom justifi ed 2). This system is still operating in many countries, such as keeping its network closed was that it had incurred high Somalia, where users exhibit more faith in storable and trans- costs to develop the existing infrastructure. ferable credit than in the local currency, the Somali shilling. Recently, the Competition Authority of Kenya ruled: But elsewhere, notably in Kenya, a more formal system of “(a) that all restrictive clauses in the agreements between mobile money developed, using a separate store of credit Safaricom Limited and its Mobile Money Transfer Agents be on the user’s SIM (subscriber identifi cation module) card. expunged immediately, but in any event not later than July Safaricom introduced its M-Pesa mobile payment service in 18, 2014; (b) that the Mobile Money Agents be at liberty to 2007, and other operators developed rival services. By 2013, transact the Mobile Money Transfer Businesses of any other transactions of US$21.9 billion—equivalent to just under half mobile money transfer service providers; (c) that oversight of Kenya’s GDP—were transacted on mobile devices. M-Pesa by Safaricom Limited be thereafter limited to its business was introduced two years later in Tanzania. Although initially with the Agents; and (d) that each Mobile Money Service slow to take off , it now equates to a higher percentage of Provider be responsible for ensuring compliance with GDP (53.2 percent) and will soon overtake Kenya in volume. Central Bank of Kenya Regulations.”a Safaricom had a monopoly in the Kenyan market This ruling came weeks after M-Pesa permitted its for almost seven years. It had established exclusivity agents to work with other mobile operators. Eliminating its arrangements with M-Pesa agents, meaning the agents exclusivity agreements reduced the cost of sending money could only off er products and services within the M-Pesa from M-Pesa users to other unregistered users. For exam- network. Agents were thus locked in to a single operator, ple, the cost of transfers of between K Sh 101 and K Sh 500 and so were most users. Since Safaricom controlled more fell from K Sh 66 to K Sh 44 within six months after the end than two-fi fths of the mobile money transfer business, the of the exclusivity agreement.

Sources: Plaza, Yousefi , and Ratha 2015, for the WDR 2016; Di Castri and Gidvani 2014. a. http://africanantitrust.com/2014/10/09/antitrust-enforcer-subjects-mobile-payment-operator-to-central-bank-oversight/. 258 WORLD DEVELOPMENT REPORT 2016

struggle with. Sharing economy fi rms like Uber could consider orienting their regulations to those and Airbnb scaled up traditional ride sharing and under development in the European Union (EU) or subletting to a global scale. But regulators struggle North America. to determine whether these companies are taxi and hotel companies or simply software providers. Off- Skills: Making the internet line competitors complain that they do not follow the same regulations. Where these industries tend work for everyone to be overregulated—often the case in the traditional If you compared our world today with the world one taxi business, for example—this new competition hundred years ago, you would encounter amazing can encourage a general regulatory overhaul of the advances in science, commerce, health care, transportation, industry. and other areas. But if you were to compare the classroom Once internet fi rms offering traditional services of a hundred years ago with an average classroom today, achieve a certain scale, regulators need to modernize you would recognize it immediately: students lined up in and impose sector-specifi c regulations to level the rows, paper and pencil in hand; a teacher at the blackboard playing fi eld between online fi rms and their offl ine jotting down facts; students furiously copying all that competitors. New York City regulators started to is written and said, expecting to memorize the facts require Uber drivers to be licensed, to drive vehicles and spit them out on an exam. with livery plates, and to have commercial insurance to create fairer competition between the sharing —Robert Hawkins (2002) economy and taxi drivers. In other cases, require- ments that taxi drivers have full knowledge of local Technological changes are disrupting the employ- streets have become obsolete with global positioning ment landscape. Occupations are becoming more systems (GPS) and should be scrapped. Airbnb is technology intensive, and “old economy” jobs are required to pay local in several cities where giving way to “new economy” careers. In many coun- its services compete with those of the hotel industry. tries, this transformation is just starting and will take time. But even in these countries, the time for Creating competition across platforms action is now. Changes in education and training take Concentration in the digital economy paired with a generation to have an effect, and reforms need to network effects or switching costs can lead to anti- start early so that skills do not become a bottleneck competitive behavior. But internet fi rms confound as countries advance in their digital transformation. traditional competition law because they do not act as traditional monopolies. Their services are often A changed world with unchanged free or more convenient for consumers. It is also classrooms less than straightforward to establish their real line The skills mix needed to succeed in the labor market of business. Google, known as a search engine com- is changing, and today’s education and training sys- pany, is better described as an advertising fi rm. It tems are often failing to keep up. The use of digital accounts for nearly 90 percent of the online search technologies requires basic cognitive skills, such as market in some countries, and around 25 percent of literacy and numeracy. But a well-educated worker the display-ad market, while Yahoo and Facebook in the 21st century also needs skills that are easily account for about 10 percent each.10 But given their transferable across jobs and occupations and that dominance in the markets for online ads or books, help respond to changing labor market demands: some internet companies have considerable leverage higher-order cognitive, socioemotional, and technical over marketers and booksellers. This is similar to skills (fi gure 5.7). This multiplicity of skills has always credit card companies’ position with respect to retail- been important, but it is now essential. From Lebanon ers. Since many internet fi rms operate in two-sided to Peru to Vietnam, employers are looking for work- markets, where an online platform brings together ers who work well in teams and can solve problems, buyers and sellers, these internet intermediaries think critically, and present their work well to oth- could blur price signals in either of the two markets. ers.12 Yet in many countries, education systems fail Research by economists such as Jean Tirole shows to provide even basic literacy and numeracy. More that regulations in such industries must be carefully than half of 15-year-olds are functionally illiterate in tailored to ensure competition and not harm consum- middle-income countries such as Albania, Indonesia, ers.11 These are very challenging problems. Countries Jordan, Kazakhstan, Malaysia, and Peru.13 NATIONAL PRIORITIES 259

Figure 5.7 The types of skills needed in a modern economy

Cognitive Social and behavioral Technical

Literacy, numeracy, and Manual dexterity and the use Socioemotional skills and higher-order cognitive skills of methods, materials, tools, personality traits (for example, reasoning and instruments and creative thinking) Openness to experience, Technical skills developed through Raw problem-solving ability conscientiousness, postsecondary schooling or versus knowledge to extraversion, agreeability, training or acquired on the job solve problems and emotional stability

Verbal ability, numeracy, Self-regulation, grit, mind-set, Skills related to specific problem solving, memory, decision making, and occupations (for example, and mental speed interpersonal skills engineer, economist, IT specialist)

Source: WDR 2016 team, adapted from Pierre, Sanchez Puerta, and Valerio 2014. Note: IT = information technology.

Digital technologies accelerate the pace of labor manage some of the labor market disruptions that market changes, opening new opportunities but also technology creates, and to ensure that the benefi ts making skills obsolete more quickly. This calls for of digital technologies are widely shared. In emerg- more adaptability from individuals and institutions, ing digital countries, such as Nepal or Senegal, the a stronger link between educational and training skill base remains weak, with gaps in foundational institutions and the private sector, and policies that cognitive and socioemotional skills, as well as in promote lifelong learning. Since skill development basic digital literacy. Transitioning countries, includ- starts at birth and is lifelong, a life-cycle approach to ing Armenia, Sri Lanka, and Ukraine, have done learning is necessary.14 fairly well on foundational skills, but they face the These two labor market transformations—chang- challenge of keeping up with some of the new skill ing skill mix and rapid skill obsolescence—have pro- demands arising from digital technologies, especially found implications for the skill development agenda. higher-order skills. Transforming countries, such as While digital technologies affect the whole skill for- the Czech Republic and the Republic of Korea, are mation process by changing the role of teachers, how best positioned to think about more advanced techni- students learn, and the mechanisms to strengthen cal skills (both in ICT and in STEM), and, since many accountability in education and training (see sector of their citizens are also rapidly aging, also need to focus 2), priorities for skill development in the 21st prioritize lifelong learning (fi gure 5.8). century on three questions: Every country has multiple skill development systems. As international tests make clear, individ- • What are the policies and interventions needed to uals who perform best in less advanced economies provide current and future workers with skills for are often at par with peers in more advanced coun- a digital economy? tries, and there is a lot of inequality in outcomes • How do these priorities vary by each country’s within countries.15 So a simple sequencing of policy labor market challenges and skill base? reforms oversimplifi es what is needed in any coun- • How can digital technologies complement this try. Even transforming countries need to strengthen “analog” policy agenda? foundational skills, and emerging countries need to strengthen their provision of more advanced skills. But skill development is a cumulative process: it is Priorities for a skill development agenda in hard to develop technical and high-order cognitive the 21st century skills without basic literacy or a strong foundation Countries with strong skill development systems of socioemotional skills. Similarly, foundational and are best prepared to leverage digital technologies, to more complex cognitive and socioemotional skills 260 WORLD DEVELOPMENT REPORT 2016

Figure 5.8 Education that upgrades and socialization, especially among young children, skills also facilitates higher adoption of these technologies are here to stay (box 5.3). “Before digital technologies a child even starts primary school, she will be able to use her parents’ smartphone to learn her numbers 15 and letters, giving her a big head start,” the Gates Foundation notes.18 Online educational games for young children, with appropriate adult supervision, are increasingly used to develop foundational skills.

10 ScratchJr, for example, is an application aimed at teaching algorithm thinking and coding principles to kids aged 5−7 years through a simple drag and drop interface. Access to the internet, laptops, tablets, mobile 5 phones, digital whiteboards, and video-based instruc- tion are increasingly common in primary and second- ary education. Programs like One Laptop per Child

Quality-adjusted years of education (OLPC) are operating around the world (box 5.4). In most cases, however, there is no solid evidence base 0 of their effects on learning outcomes, and much more 0 0.2 0.4 0.6 0.8 1.0 careful evaluation of ICT initiatives in education Technology: is needed. But fi ve uses of technology in building Digital Adoption Index (People) foundational skills are promising: connecting teach- ers to content; making learning more personalized; Sources: WDR 2016 team, based on World Development Indicators (World Bank, various years) and the World Economic Forum Competitiveness Index reinforcing content learned in school; fostering col- (WEF, various dates). Data at http://bit.do/WDR2016-Fig5_8. laboration and increasing students’ engagement; and Note: The quality-adjusted years of education are constructed by multiplying promoting adult literacy (table 5.2). the average years of education for each country by the normalized WEF’s quality of education indicator; for more details, see Monroy-Taborda, Digital technologies can complement teachers and Moreno, and Santos, forthcoming. The Digital Adoption Index (People) is connect them to content. The BridgeIT model, started the simple average of two normalized indicators from the Gallup World Poll: mobile access at home and internet access at home. in 2004 and now used in around 10 countries, allows teachers using mobile phones to download video content, which is then connected to televisions in built early in life and throughout general education their classrooms. The videos are paired with learner- are the base for the building and updating of techni- centered lesson plans, accompanied by teacher train- cal skills in postsecondary education and training. ing in how to use the technology and incorporate it in the class.19 A separate initiative, the Bridge Interna- Priorities for emerging countries: Build tional Academies, goes even further. The largest chain foundational cognitive and socioemotional of preprimary and primary schools in the world, with skills and ensure basic ICT literacy more than 110,000 students in Kenya and Uganda, it The analog agenda for building foundational skills uses technology and data analysis to standardize and starts early, as early as 0–3 years of age, and goes scale education delivery. While the standardization beyond investment in education to that in health.16 of teaching remains controversial and is being rigor- These are the years when children become school- ously evaluated, it is an alternative way to make basic ready, their brains are the most sensitive to learning, education more affordable. The average family of a and the basis for future learning is laid. Families, indi- student in the Bridge Academy lives on US$1.60 a day viduals, school and training systems, and employers per person.20 will all play a role in building these skills throughout Digital technologies can offer access to high- the life cycle. It is also important to build these skills quality material where there are no teachers with before specializing too narrowly. Countries like the needed skills. In Uruguay, through video, English Poland have delayed tracking of students into voca- is taught to fi rst graders by teachers from the Philip- tional education, with positive results.17 pines. A pilot study found that videoconferencing and Digital technologies can help build these foun- laptops raised the children’s scores in English signifi - dational skills. While there are concerns about the cantly, as well as the English-language profi ciency of impact of digital technologies on cognitive capacities the Uruguayan teachers.21 NATIONAL PRIORITIES 261

Box 5.3 The impact of digital technologies on cognitive capacities and socialization

There are concerns about the impact of digital technolo- Yet the fear that new technologies lessen our ability to gies on cognitive capacities and socialization, especially function is nothing new; in Plato’s dialogue, The Phaedrus, among young children and adolescents, and some evi- Socrates worried that writing would diminish the ability dence backs these concerns. For example, studies show to engage in conversation. Parents and schools can do that using search engines decreases our memory. When things to lessen these risks. Strengthening self-regulation faced with diffi cult questions, people are primed to turn to in young children can mediate potentially negative eff ects their computers, and, when people expect to have future of internet use.c While it is true that using technology to access to information, they have lower rates of recall of do things for us that we no longer are doing for ourselves the information and enhanced recall of where to access it.a means that certain abilities can degenerate, it also means There are also concerns about internet addiction. A survey that we are freeing up cognitive energy for other things. of nearly 12,000 adolescents in 11 European countries found More research on the impact of digital technologies on a 4.4 percent prevalence of pathological internet use. An cognitive abilities, and how they can be managed, is an additional 14 percent had a milder addiction. People who important agenda for the future. exhibited problematic use were more likely to suff er from psychological problems, such as depression or .b

a. Sparrow, Liu, and Wegner 2011. b. Durkee and others 2012. c. Spada 2014.

Box 5.4 One Laptop per Child: Strengthening analog foundations and careful evaluation

Of all the educational initiatives that involve digital tech- distribution of laptops with some level of guidance, from nologies, the One Laptop per Child (OLPC) program (or providing training courses to using adaptive software for more generally, 1 to 1 computing initiatives) is arguably the weekly sessions, improved learning. These guided-use pro- most ambitious. Its main objective—empowering children grams had a positive, albeit small, impact of 0.17 standard by providing them with digital devices at an aff ordable deviations on average scores, against an insignifi cant eff ect price—has captured the attention of world leaders, media, for programs where use was not guided.d and academia.a Since the fi rst laptop distribution started in The success of digital devices in the classroom in Uruguay in 2007, OLPC programs have distributed more than improving academic performance hinges on complement- 2.4 million laptops around the globe.b In Uruguay, the program ing access to digital technologies with investments in was widely supported by children, parents, and school direc- teachers, and on embedding the technology with relevant tors, providing internet access to many low-income children.c content that can be integrated with traditional classroom Although all OLPC programs increased access to digital activities. More careful evaluations of pilot programs before technologies, their implementation and impacts on learning they go to scale are also important to better understand outcomes vary greatly. In a meta-analysis of 15 studies of whether the programs can produce positive results and are OLPC programs in fi ve countries (China, Colombia, Ecuador, cost-eff ective, as well as the specifi c changes needed to India, and Peru), only the programs that accompanied the improve their eff ectiveness.

a. Trucano 2012. b. http://one.laptop.org/map. c. Martinez, Alonso, and Diaz 2009. d. Arias Ortiz and Cristia 2014. 262 WORLD DEVELOPMENT REPORT 2016

Table 5.2 Emerging countries: A skill development agenda for a modern labor market

Pillars of an analog Priority policy agenda How digital technologies can complement Foundational cognitive • Early childhood • Complementing teachers and expanding access to and socioemotional development and school quality material (BridgeIT in the Philippines and skills readiness Tanzania; Khan Academy) • Making learning more personalized through adaptive curricula and assessments Digital literacy • Mainstream of • Reinforcing content learned in school, to consult with socioemotional skills teachers or learn new material (Eneza, Kenya) and basic digital literacy in teaching • Facilitating learning and collaboration and methods and increasing students’ engagement (Educopedia in assessments Brazil, game-based learning or online games) • Adult literacy • Promoting adult literacy (using SMS in Niger and ), in combination with e-entrepreneurship Source: WDR 2016 team. Note: SMS = short message service.

Digital tools also make it easier to tailor learning rooms in seven Latin American countries, between to each student and rapidly assess student progress. one-fi fth and one-quarter of total class time, a large Impact evaluations suggest that the successful appli- group of students (six or more in classrooms with cations of technology help introduce an appropriate average size of 25) is visibly not engaged in the class- curriculum or enable students to move through room. This is partly because most classes are boring: material at their own pace.22 Khan Academy is a good for about one-third of all time spent on teaching example (box 5.5). These types of tools can be particu- activities, teachers use the blackboard and nothing larly helpful in developing countries, where students else.25 Using digital technologies for collaboration often need to develop skills that their teachers lack or (box 5.6) or for game-based learning could improve do not teach.23 engagement and learning. At its simplest, textbook Digital technologies can reinforce learning. In material could be made more engaging: eLimu, Mumbai and Vadodara in India, a math reinforce- which has embedded the primary education curricula ment program, delivered through computer games, of Kenya in tablets, enriched the content with videos had large positive effects on children’s academic and other animations. achievement.24 Children were offered two hours of Yet introducing technology alone, without shared computer time per week, during which they improved teacher training and links to changes in played games that involved solving math problems. pedagogy, will not improve the learning process. The program increased math scores by 0.35 standard Digital technologies are most likely to have an impact deviations the fi rst year, and 0.47 the second year. when the focus is not on hardware and software but Similar uses of technology to reinforce learning are on how they contribute to learning.26 In Colombia, being applied around the world, often using sim- Computers for Education had little effect on student ple mobile phones. Eneza Education, in Kenya, is a test scores and other outcomes. Although students in virtual tutor that provides practice questions and the program schools were 30 percentage points more answers open questions via SMS (short message likely to report using a computer at school, they used service) from students on low-cost mobile phones. It it only for computer science.27 Similarly, in Ireland, has 39,000 active users, and more than 380,000 peo- teachers were positive about the usefulness of ICT ple have used it at least once. in secondary schools, yet most uses were outside the A fourth promising technology is facilitating col- classroom for preparing lessons, rather than exploit- laboration among teachers and students and making ing the potential of ICT to introduce innovative teach- learning more fun and effective through games and ing and learning practices.28 As with One Laptop per game-based learning. More often than not, students Child (see box 5.4), annual evaluations of Text2teach, are not engaged in school. In more than 15,000 class- the local version of BridgeIT, in the Philippines have NATIONAL PRIORITIES 263

Box 5.5 Khan Academy: A supplemental educational resource in and outside the classroom

Khan Academy is an online learning platform that provides Khan Academy in nine schools in the United States from free tutorial videos on disciplines ranging from elementary 2011 to 2013.b Khan Academy mainly supplemented teach- mathematics to computer programming. It attracts more ers’ own core instruction rather than replacing it. Forms of than 10 million unique users a month (35 percent from use varied: additional practice or remediation for students outside the United States), who have viewed 365 million who fell behind, enrichment activities for advanced stu- videos and solved more than 1.8 billion math problems.a It dents, and monitoring student progress. Khan Academy is an example of tools for the “fl ipped classroom,” in which increased engagement of students and the capacity of videos substitute for classroom instruction, and class hours teachers to support their students, but teachers found it are used for practice and discussion. diffi cult to fully integrate it with the core curriculum. Khan Academy is built on free lecture videos, adaptive A pilot in fi ve schools in Chile had similar results.c Khan learning practice problems, and personalized data. Lectures Academy improved student math skills, but was not “fl ip- are divided by subject, but users can select what to focus ping the classroom.” Moreover, the teachers believe that on, advance at their own pace, and receive instant feed- while Khan Academy improved the procedural skills, it was back. Following the video lectures, students can practice, not best for promoting deeper mathematics learning. and the system adjusts the diffi culty of practice problems Integrating tools like Khan Academy into education based on a student’s performance on the fi rst few prob- systems shares some of the challenges that other technol- lems. Students’ online performance is recorded, and stu- ogies in education face. Translation and adaptation to local dents, teachers, and parents can follow progress through contexts is costly. Khan Academy is partnering with other the personalized learning dashboards. organizations in several countries to do this. And stream- While Khan Academy is still used predominantly by ing videos requires high-speed connectivity, which is still individuals for informal study outside schools, its use in limited in many developing countries. Since the success of schools and other institutional settings is growing. There is the program is a function of teacher skills and engagement, not enough evidence to claim that it has been an unqual- school and government administrations should support ifi ed success, but the evidence points to its considerable teachers throughout the period of implementation.d potential. SRI International studied the implementation of

a. Murphy and others 2014. b. Murphy and others 2014. c. Light and Pierson 2014. d. Trucano 2015.

identifi ed the need to invest even more in teacher and training teachers in the use of the technology and training for further improvement.29 in how to incorporate it in the classroom. Training Even when used, technology does not substitute programs in developing countries often focus only for teachers. High-quality education continues to on basic computer literacy. In Singapore, by contrast, hinge on high-quality and motivated teachers. In both pre-service and in-service teacher training curric- India, the computer-assisted learning (CAL) program ula integrate pedagogy and ICT.31 With these invest- for math had very different impacts depending on ments, technologies can amplify teacher productivity how it was used. When the program supplemented and improve the quality of teaching. In systematic regular classes with one additional hour of CAL reviews of education interventions, technology-as- instruction every day after school, it raised math sisted learning, together with teacher training pro- scores by 0.28 standard deviation. But when it was a grams, have among the largest positive effects on substitute for regular instruction, scores fell by 0.57 learning outcomes.32 standard deviation.30 Outside formal education, digital technologies can Governments can play an important role in build- also boost adult literacy. In Afghanistan, Chad, and ing complementary investments to leverage educa- Mali, more than 60 percent of adults remain illiterate; tional technology. This is primarily about preparing in Niger, close to 85 percent.33 Digital technologies, 264 WORLD DEVELOPMENT REPORT 2016

Box 5.6 Using digital technologies to foster collaboration and learning: Rio de Janeiro’s Educopedia

In 2010, the Rio de Janeiro municipality launched Educo- of teachers that produces and assesses the adequacy pedia, a collaborative online platform of lessons open to and quality of online materials to reach out to colleague students and teachers from public schools. By off ering stim- teachers and provide hands-on support in using the ulating multimedia resources in the classroom, Educopedia’s new resources. In 2010, Rio municipality teachers used goals are to support teachers in creating and sharing teach- information and communication technology (ICT) only ing materials online and to increase students’ motivation to 1 percent of the time in the classroom; a year later, use had learn. The lessons online cover math, Portuguese, science, quadrupled. history, geography, English, music, and physical education, Today, Educopedia operates in all 700 Rio de Janeiro organized by grade and week of the curriculum. municipal schools and serves 680,000 students; 50 per- The modules generally blend videos and interactive cent of teachers report that they use the tool more than exercises for students with a clear structure and repeated once a week. The impact of Educopedia on student learning reinforcement questions. Each Educopedia module con- has not been evaluated yet, but the secretariat its sists of a lesson plan to help teachers structure the class; eff ects on the motivation of both teachers and students are supporting content such as PowerPoint presentations on positive. Educopedia was one of the pillars of education new material and texts, videos, and games; and interactive reforms in Rio de Janeiro, which between 2009 and 2012 resources such as a chat system, a digital library, quizzes, saw a 22 percent increase in its score in the Basic Education summaries, and sets of test questions. The materials are Development Index for middle schools. In a recent survey, projected on digital whiteboards. 80 percent of Rio de Janeiro municipal students agreed When the program was piloted, teacher take-up was that Educopedia contributes to their learning, particularly low. The secretariat responded by asking the network through the interactive exercises and educational games.

Sources: Bruns and Luque 2014; WDR 2016 team.

especially mobile phones, can promote basic numer- skills, including the ability of students to search for acy and literacy for these adults. In Niger, including information and separate high-quality sources from instruction on using simple mobile phones with voice low-quality ones. International experience offers and SMS capacity in a basic adult education curric- some important lessons (box 5.7). ulum substantially improved learning outcomes.34 In many developing countries, the greater chal- Similar programs in advanced settings, like Los Ange- lenge is to reach adults, often in low-literacy environ- les, have had similar effects, with technology actually ments. Lack of awareness of the potential usefulness improving literacy without the need for teachers.35 In of digital technologies remains an obstacle to adop- only four months, basic reading scores went up about tion.37 In some cases, digital technologies, especially fi ve years, and more advanced reading scores, about mobile phones, are adapted for the illiterate or those two and a half years. with little education through the use of graphics, symbols, and audio or video tutorials. In addition to Digital literacy is a new foundational skill technical training, which can be costly and which In a world where digital technologies are at the center many adults would not use, countries can raise of people’s personal and professional lives, digital lit- awareness and demonstrate some of the potential eracy—just like reading and writing—is a new foun- benefi ts of digital technologies—as in Turkey, where dational skill. Among children and youth, just giving a mobile unit goes to lagging regions and shows how access to these technologies can be enough for them to use government services online. to learn the basic use of devices. The evidence from Among youth and adults, digital literacy training One Laptop per Child (see box 5.4) and from Hole-in- can be combined with e-entrepreneurship training and the-Wall in India indicates as much.36 The main chal- seed funding to conduct business over the internet. lenge is to broaden the digital literacy agenda beyond Two areas of focus could be e-commerce and online the basic use of devices to focus more on information work. For e-commerce, the focus could be on how to NATIONAL PRIORITIES 265

Box 5.7 Emerging lessons from digital literacy programs

Digital literacy programs considered most successful have a lack of digital literacy or fear of using new methods some principles in common: is a constraint for teachers. As a result, programs are increasingly being combined with intense teacher train- • They are mainstreamed into the non-ICT (information and ing. eLimu—which has embedded the Kenya primary communication technology) curriculum. They emphasize education curricula in tablets and enriched the content ICT as a tool rather than a subject. Students learn how to with animations to make learning more engaging—had use digital technologies not in specialized and isolated to extend teacher training from 15 minutes to a full day. labs but as part of their learning other subjects. Finland • They go beyond ICT, into the beginnings of “computa- has gone for mainstreaming, which makes the digital tional thinking.” Such thinking refers to the problem- skills learned more useful in real life, but requires train- solving skills and techniques software engineers use to ing for all teachers. Training also aims at using internet write programs, especially those related to breaking a search eff ectively and safely, and understanding privacy problem into parts, pattern recognition, abstraction, and rules. algorithm design.a • They focus on teachers’ digital literacy. Evaluations • They are embedded in local content (chapter 4). This can of Khan Academy, the rollout in schools of broadband motivate learners by connecting them to information on in Ireland, eLimu and Eneza in Kenya, and the One issues relevant to them but also by reducing language Laptop per Child programs in Latin America show that barriers.

a. ISTE and CSTA 2011.

market yourself and be able to sell online. Countries apply it to the world around them in meaningful could explore partnerships with e-commerce plat- ways? forms. Alibaba, for example, offers training on how to A fi rst step is to modernize general education sell on its platform, and has signed agreements with curricula and teaching methods, shifting toward governments (such as the Republic of Korea) to do so. student-centered curricula that emphasize new For online work, the training could focus on intro- economy skills. Which qualities school systems ducing participants to online work, setting up their need in order to develop these skills is debated, but profi les, and establishing payment accounts. Nigeria there is agreement that it needs to start early. One is starting to provide this type of training.38 model is Montessori schools. Teachers trained in a Montessori perspective emphasize, from early child- Priorities for transitioning countries: hood onward, attention to the individual interests of Build new economy skills for careers, students. Classrooms are less hierarchical. Students not just for the fi rst job have substantial unstructured time to collaborate around shared interests and freedom to research a Improving advanced cognitive and topic of interest and present it to the class.39 While socioemotional skills requires rethinking rigorous evidence on this is limited, students in fi ve curricula and teaching methods Montessori middle schools in the United States were In addition to foundational skills, workers are being more intrinsically motivated and interested in their required to use more critical thinking and problem academic activities at school.40 Similarly, the Perry solving, communication, teamwork, and creativity. Preschool Program in the United States teaches kids These are general skills, which many traditional reading and math, but also to plan tasks, execute their education systems fail to impart, and can be applied plans, and review their work in groups. Although across occupations and jobs (fi gure 5.9 and table 5.3). based on a small sample, the program has been found How can teachers, administrators, and policy makers to boost socioemotional skills and to determine life create an environment that goes beyond rote learning outcomes of former students well into their 40s.41 and memorization, and where students really learn— Modern curricula should balance competency- where they think critically about information—and based and content-based learning, and be combined 266 WORLD DEVELOPMENT REPORT 2016

Figure 5.9 Even in advanced countries, youth are often unable to think critically and solve problems

Share of youth (aged 16–24) scoring below 2 in the Programme for the International Assessment of Adult Competencies in problem solving in technology-rich environments

70

60

50

40

30 Percent

20

10

0

Poland Japan Austria Ireland Estonia Canada Norway Sweden Finland Denmark Australia Germany Korea, Rep. United States Netherlands Slovak RepublicUnited Kingdom Czech Republic Source: Davalos and Santos, forthcoming, based on the OECD Programme for the International Assessment of Adult Competencies (PIAAC) website, http:// www.oecd.org/site/piaac. Data at http://bit.do/WDR2016-Fig5_9.

with teaching methods that stimulate and build in performer in international student assessments, Fin- critical thinking, problem solving, communication, land recently concluded a large reform of its preprimary teamwork, and creativity. The New School model in and basic education curricula. The aims are to develop Colombia (Escuela Nueva, in Spanish) has taken this schools as learning communities and to emphasize approach, putting stronger emphasis on group learn- the joy of learning and a collaborative atmosphere, as ing and problem solving than on memorization and well as promoting student autonomy in studying and copying. This model has informed the modernization in school life. In particular, there will be much focus of the curriculum in many countries: so far, it has on general transversal competencies and work across been piloted or has been already extended to around school subjects. The competencies will also be assessed 17 countries (box 5.8).42 as a part of subject assessment. The emphasis set on Countries like Brazil, Finland, the Republic of Korea, collaborative classroom practices will be brought about the former Yugoslav Republic of Macedonia, Mex- in multidisciplinary, phenomenon- and project-based ico, Peru, Singapore, and Vietnam are taking actions studies where several teachers may work with any to foster new economy skills.43 Despite being a top given number of students simultaneously.44

Table 5.3 Transitioning countries: A skill development agenda for a modern labor market

How digital technologies Priority Pillars of an analog policy agenda can complement Higher-order • Modern curriculum, teaching methods, • Directly promoting critical thinking, team- cognitive and and assessments with increased emphasis work, problem solving, and creativity socioemotional on critical thinking, problem solving, and (through online games, wikis, hackathons) skills socioemotional skills (Escuela Nueva model; Republic of Korea; Singapore) • Delivery of training in socioemotional skills (grit and mind-set interventions) • Modernize teacher training in-service and pre-service accordingly • Promoting collaboration (virtual exchanges) Source: WDR 2016 team. NATIONAL PRIORITIES 267

Box 5.8 Building new economy skills: Escuela Nueva in Colombia and Vietnam

The Escuela Nueva model started in Colombia in 1976 as children in these groups. The school calendar and evalua- an innovation in multigrade teaching, promoting active, tions are fl exible, so that children can meet learning goals participatory, and cooperative learning among primary at their own pace. school students. Today it serves 5 million students in 16 Evaluations of the model in Colombia and elsewhere countries, including Brazil, the Dominican Republic, Mexico, indicate that the program is fostering both cognitive and the Philippines, Uganda, and Vietnam. socioemotional skills. The model improves levels of Spanish The model is based on several innovations for improving and math in the third and fi fth grades, as well as self- teamwork and developing critical thinking. The curriculum esteem and the abilities to lead others in group tasks and is focused on a self-paced and self-directed learning guide. to work peacefully with others in a team. A recent impact Group learning is facilitated through the use of learning evaluation of Escuela Nueva’s fi rst two years in Vietnam guides that contain interactive exercises. Teachers are shows that the model helps children learn to work with trained in group management. Group learning is also fos- each other and develop communication and interpersonal tered by class arrangements and a modifi ed role of teach- skills. Improved cooperative learning skills also enable a ers as facilitators. Students are seated in clusters of four or student to obtain better results in math. fi ve, with teachers guiding, supervising, and evaluating the

Sources: WDR 2016 team based on Bodewig and others 2014; Colombia Aprende 2015; Forero-Pineda, Escobar-Rodriguez, and Molina 2006; Fundación Escuela Nueva 2015.

Singapore has also made large changes to its is an open debate. Even if taught separately, it is education system to adapt to the knowledge-based important to reinforce learning these modern skills economy. The main change was in 1997, with a move across subjects. A teacher could give a lecture on from an effi ciency-driven model to an ability-driven history with little class interaction, or could break model.45 The effi ciency-driven model, an engineer’s the class into small groups that would refl ect on the vision to education, had a top-down approach in reading material, prepare takeaways, and present designing, disseminating, and enforcing the national to the whole class, thus teaching history as well as curriculum; streaming students by competency lev- teamwork, empathy, communication skills, problem els to ensure that teachers were dealing with students solving, self-regulation, and self-esteem.47 Changes of similar levels; and creating clear but rigid paths to can also include introducing more open-ended ques- vocational and tertiary education. While this model tions in home assignments and in tests, more fl exible produced students who scored high on international classroom seating arrangements that allow work in math and science tests, it produced students who groups, having discussions involving the full class- were not critical thinkers and teachers who were not room, and creating spaces in classrooms and schools motivated.46 Under the new ability-driven model, for collaborative projects. The increased demands on schools have more autonomy over their curriculum teachers need to be accompanied by modernizing and develop programs to suit their students. More teacher training, both in- and pre-service. Teacher emphasis is placed on project work, introduced from training needs to focus on how to teach curriculum primary education onward, with a move from high- content and on how to impart socioemotional skills. stakes examinations to smaller assessments. The Countries are making progress in improving stu- government has also emphasized the use of ICT in dent assessments, critical for identifying strengths the curriculum. and weaknesses in the education system, designing Whether to streamline teaching new economy policy, and strengthening accountability. Yet current skills into the traditional curricula, as in Finland or assessments typically focus on testing information, Peru, or whether to teach them in targeted classes facts, or the ability to read or perform math opera- or interventions, as with grit (that is, perseverance tions. While important foundational skills, these skills in pursuing long-term goals) in FYR Macedonia, are often fairly routine and easily programmable, so 268 WORLD DEVELOPMENT REPORT 2016

student assessments can be further strengthened by tance barrier. The World Links for Development broadening their scope to better incorporate higher- Program has worked with ministries of education order skills. in more than 20 middle- and low-income countries Open-ended questions in tests can put greater to link classrooms to the internet but also to one emphasis on critical thinking and problem solving. another. European Schoolnet has fostered long- New economy skills can be monitored, as is tradi- term online connections among classrooms across tional academic learning, through mechanisms such borders.52 Social media can also be used to connect as the KIPP Report card, where socioemotional skills classrooms. are identifi ed, measured, and discussed with parents and students at evaluation time.48 At the international Priorities for transforming countries: level, the Organisation for Economic Co-operation Focus on advanced technical skills and and Development (OECD) and the World Bank have lifelong learning recently implemented new skill measurement assess- ments that aim at measuring, among adults, problem ICT skills and training in computational solving in technological environments, and socio- thinking emotional skills, respectively.49 L’Oréal, the French As in other countries, the skills agenda for transform- cosmetics company, asks job candidates in China ing countries needs to also strengthen foundational to answer three open-ended questions on mobile skills, and develop new economy skills, such as phones, and then uses “big data” and an algorithm to higher-order cognitive and socioemotional skills. mine the responses for cues on critical thinking and Because the skill base in transforming countries is socioemotional skills.50 stronger than in the emerging and transitioning Digital technologies can also foster higher-order countries, they also have to focus on more advanced cognitive and socioemotional skills—in at least three technical skills (table 5.4). ways: To address shortages in advanced ICT skills and improve competitiveness in a growing industry, • Directly promoting critical thinking, teamwork, problem countries are paying more attention to providing solving, and creativity. This could be done through advanced ICT skills training in their education sys- digital activities, such as programming. Scratch, a tem. Many fi rms report diffi culties in fi nding work- simple programming language for kids, can help ers with advanced ICT skills (chapter 2). Partly in develop abstract and critical thinking from an early response, coding, as a subject, has been introduced in age. Digital tools, such as wikis—online content- the general education curriculum in Estonia and the management systems that allow for collaborative United Kingdom. modifi cation, extension, or deletion of its content Incorporating advanced ICT skills in the general and structure—can promote discussion and com- curriculum can teach computational thinking in munication inside and outside the classroom. Many transforming countries. But not everyone needs to schools are now using hackathons, events where become a professional coder. Incorporating coding in teams collaboratively work on software projects, general education can strengthen not only ICT skills and which can provide a creative space that also but also critical thinking, especially if it is a vehicle to fosters problem solving. Incorporating learning teach logic and learning strategies for solving prob- games into classrooms (game-based learning) and lems, designing projects, and communicating ideas. applying the principles of gaming to education Done well, it can be not only about “learning to code” (gamifi cation) could also foster higher-order cogni- but about “coding to learn.”53 tive and socioemotional skills, such as abstraction, Training in advanced ICT skills can also be pro- reasoning, and teamwork, while bringing the power vided less systematically, and outside of the formal of play to education and engaging, inspiring, and education system. This could be the route for coun- immersing students in learning (box 5.9). tries with less mature education systems. For exam- • Developing and delivering training in socioemotional ple, advanced ICT skills could be offered in middle or skills. Grit (that is, perseverance in pursuing long- high schools through accredited massive open online term goals) and growth mind-set lessons,51 for exam- courses (MOOCs). Individuals can also learn coding ple, can be delivered via videos and over the internet. through specialized online platforms. Codecademy, Additional online modules that target other socioe- an online interactive platform that offers free coding motional skills could be part of teachers’ toolkits. classes in seven programming languages, has more • Supporting teamwork and communication skills by than 24 million users who have completed over 100 bringing together diverse teams and breaking the dis- million exercises.54 NATIONAL PRIORITIES 269

Box 5.9 Building modern skills: Game-based learning and “gamifying” education

Learning can be fun. Play can be a great conduit for learn- But most of today’s teachers are not equipped with ing and creativity because it taps into people’s intrinsic the skills to make games eff ective learning tools, such as motivation, satisfying needs of autonomy, competence, designing materials and environments for game-based and relatedness. When playing, many people achieve a learning, partly because they are not familiar with many state of “fl ow”—a state of concentration that activates the tools of the digital age.g Teachers need to be trained so brain and is considered optimal for new learning.a Applying that they understand both the potential and the limits of the principles of gaming—rules, goal achievement, pro- games for learning and fully understand the game, how gressive diffi culty, interaction and student control, uncer- to play, and how to use it as a teaching tool. Teachers can tain outcomes, and immediate feedback—to classroom work with parents in fi nding appropriate games and defi n- activities can teach both cognitive and socioemotional ing parameters for usage. Updated curricula, reformed skills. Some games apply these principles to teach basic assessments, and resources for learning games can all skills, such as math and vocabulary in games like King of help to shift classrooms to be focused on learning. Policy Math and FunEnglish, while others integrate learning into makers could also encourage more research and develop- the game. Dragonbox, Freddi Fish, and Guild Wars require ment to create successful learning games that are engag- creative thinking, problem solving, and building on existing ing, thoughtful, and immersing, and facilitate effi cient knowledge to succeed.b Requiring inference and problem learning. solving, rather than direct questioning for learning, they are The future of games and game-based learning extends more engaging and can lead to sustained learning.c beyond the classroom. Evoke, developed by the World Games can not only increase engagement but also Bank, uses social media tools and narrative approaches improve learning outcomes. A unique English program in commonly found in video games to empower young India can be implemented with diff erent technologies (a people to start solving urgent social, real-world problems. specially designed machine or games and activities based Food Force, created by the United Nations, teaches strategy on special fl ash cards). New methods yielded gains of about and networking by simulating aid agencies working in food 0.3 standard deviation in test scores.d A 2006 study of fi fth insecurity. The Facebook game Half the Sky highlights gen- graders found that playing math games was more eff ective der equality and raises money for female empowerment. than basic math drills at building math skills, as measured Games can also be tools for job training, especially when by performance on a standardized math exam.e And two real on-the-job training is dangerous; the U.S. military has science games, RiverCity, which emphasizes ecology and used video games to teach strategy, skills such as fi rst aid, biology, and Supercharged!, a program on electrostatics, and military rules and ethics.h improved learning outcomes over a traditional lecture by 15 to 18 percent and 8 percent, respectively.f

a. Jarvilehto 2014. b. Becker 2007; Prensky 2006; Jarvilehto 2014. c. McFarlane, Sparrowhawk, and Heald 2002. d. He and others 2007. e. Grabowski and Fengfeng 2007. f. Mayo 2009. g. Prensky 2006; Becker 2007; Jarvilehto 2014. h. Prensky 2007.

Countries with a local technology industry job placement rate of 90 percent.55 Similarly, iHub, one or where online work is common, already teach of the innovation and hacker spaces in Kenya, holds advanced ICT skills outside the education system. hacker labs for children and youth aged 10 to 16. In Kenya, NairoBits equips youth with knowledge Many developing countries are partnering with in web design, IT skills, creative multimedia, and the private sector to develop technical training in entrepreneurship. This technical training is combined advanced ICT skills after general education. Mexico with training in socioemotional skills to build self-con- First, a partnership with Cisco and Microsoft, targets fi dence. NairoBits-trained youth have secured employ- professionals and university students to facilitate ment in both formal and informal sectors and report a training and certifi cation in ICT. 270 WORLD DEVELOPMENT REPORT 2016

Table 5.4 Transforming countries: A skill development agenda for a modern labor market

How digital technologies Priority Pillars of an analog policy agenda can complement Advanced ICT • Strengthen collaboration between the • Coding for kids (Scratch) skills and STEM in private sector, governments, and general education educational institutions (Mexico First) • Coding among adults and outside the and beyond educational systems (NairoBits, Kenya; • Promote STEM education Codecademy) • Enhancing practical training (virtual labs and online work) Lifelong learning • Promote practical training in education • Increasing fl exibility and reach for adult learning (MOOCs) • Provide incentives for fi rms to provide on-the-job training Source: WDR 2016 team. Note: ICT = information and communication technology; MOOCs = massive open online courses; STEM = science, technology, engineering, and mathematics.

Strengthening education in science, with 18 percent of boys.57 Addressing gender stereo- technology, engineering, and mathematics types in school and at home can encourage girls to (STEM) believe that they have the required ability for STEM Improving STEM education has become a major goal and that signifi cant opportunities will emerge for not only in transforming countries but also in most them. Other actions include providing girls with rele- other countries. But good intentions are not always vant role models; targeting girls for recruitment into paired with concrete and effective actions. Address- the STEM fi elds early in their education, as Finland ing shortages and defi ciencies in the teaching of does; clearly establishing targets and incentives for STEM requires preparing and equipping teachers recruitment, retention, and graduation of women in properly and including STEM across the educational STEM fi elds; and working with employers to make system—not just in tertiary education. It also requires work settings gender-friendly. creatively involving the employers, more effectively connecting the teaching and research activities, and Promoting lifelong learning in rapidly establishing fi nancial incentives or compensatory aging societies mechanisms to make STEM education viable and In rapidly changing labor markets, the skills learned affordable to underrepresented groups. And it in formal education risk becoming obsolete. Contin- requires enabling institutions and governments to ued (re)training and upskilling can help workers stay share experiences and lessons in addressing the current, and ensure that their skills are complemen- STEM education defi cits. The World Bank–funded tary to new technological developments. Partnership for Skills in Applied Sciences, Engineer- A fi rst part of a lifelong learning agenda is ing and Technology brings together the expertise in ensuring that the skills students learn are useful for several Sub-Saharan countries with that in Brazil, entering the labor market and provide a strong base China, India, and the Republic of Korea. for continuing to learn on the job. Given the pace Women are much less likely than men to choose at which digital technologies are changing and how a STEM fi eld in tertiary education, in both advanced they are changing the labor market, students and and developing countries, and this is partly refl ected adults must learn how to learn. in occupational choices. In the United States, although Governments can: women fi ll close to half of all jobs nationwide, they hold fewer than 25 percent of STEM jobs, and for • Provide students with practical training and exposure to every dollar a man earns in STEM, a woman earns 86 the world of work prior to their graduation. In addition cents.56 Women with a STEM degree are more likely to traditional apprenticeships, technology can be to work in education or health care. useful. Virtual labs and games for sci- The gender gap in STEM fi elds starts early, both ence education and machine operations are spread- in the education system and at home. In OECD econ- ing and moving ever closer to the brick-and-mortar omies, at age 15, fewer than 5 percent of girls consider experience.58 Online work can also expose young careers in engineering and technology, compared students and trainees to the world of work. NATIONAL PRIORITIES 271

• Explore modular approaches to learning so that stu- least parts of their courses. The Dominican Repub- dents can move between education and work, as lic training institute offers a wide range of online in Denmark or in College to Careers in the United courses, including teacher training and inventory States in postsecondary education. In the latter, management, and reports particular interest from students complete modules with clear accredita- youth and individuals who are already working or tion and competencies acquired (often equivalent who have family responsibilities.62 MOOCs can also to “expertise” levels), which allow them to mone- be a tool for lifelong learning (box 5.10). tize the newly acquired skills in the labor market, • Promote on-the-job training. Traditional market fail- even before completing the full degree. Students ures limit on-the-job training63 so one common can then return to the community college and con- mechanism for promoting on-the-job training is to tinue to the next level of expertise. Nanodegrees or have institutional arrangements between employ- microdegrees, offered by MOOC platforms Uda- ers and employees. “Payback clauses” commit the city and Coursera in partnership with the private worker to stay at the fi rm for a specifi c amount of sector, take the modular approach to online-based time if training is provided. In apprenticeships, the education. worker shares the costs of training through lower • Strengthen links between the private sector and higher wages. The German apprenticeship system is a education and vocational systems. Ensuring that well-defi ned program that relies on a dual system graduates from the education system acquire job- of classroom training in vocational institutions and relevant cognitive, socioemotional, and technical training at the workplace. skills requires that fi rms, universities, and voca- A second common mechanism for on-the-job tional schools, and current and future students, training relies on fi nancial arrangements that become better connected. The German dual system internalize some of the externalities related to is a very formal and institutionalized way of fos- employer-provided training. In developing coun- tering these links, but there are other approaches. tries, this typically takes the form of training Chicago’s reforms of the city’s community colleges funds, usually through payroll taxes, as in Brazil, targeted curricula more explicitly to sectors with a Chile, Malaysia, and Mexico. Other schemes large presence in the region, including manufactur- compensate fi rms for the cost of training through ing and insurance. Specifi c components of these cur- levy exemptions, tax rebates, and cost reimburse- ricula were discussed with major employers in these ments. But as with other subsidies, these can be sectors to ensure relevance for the job market.59 ineffi cient and crowd out private expenditures on • Focus publicly fi nanced training on new economy skills training. It is, therefore, important to target the and ICT skills. New economy skills are likely to be subsidies carefully, trying to go beyond fi rm size, underprovided by fi rms because the benefi ts do not as in Malaysia’s Human Resources Development accrue solely to the fi rm.60 Several youth employ- Fund.64 In OECD countries, different cofi nancing ment programs, such as Jovenes in Latin America, savings and loan schemes match individual con- combine technical training with socioemotional tributions to contributions from employers and skills for disadvantaged groups, and many have governments. Individual learning accounts, learn- had an impact on employment and earnings.61 ing vouchers, and income-contingent repayment Vocational training institutes in developing loans are just some of the possible instruments countries are starting to use the internet to deliver at that could be used.65

Box 5.10 Massive open online courses (MOOCs): A promising tool for lifelong learning

MOOCs are a recent development in distance learning, are mostly free, with fees for getting certifi cations. Some characterized by three key aspects: open enrollment, popular courses’ enrollment reach over 100,000 students. online assessment, and an interactive forum. They sepa- Platforms such as Coursera.org, Udacity.com, and edX.org rate themselves from traditional distance learning tools host these online courses, facilitate online discussions, and by providing postsecondary education in large scale. They assess participants’ performances. In the fi rst two years

(Box continues next page) 272 WORLD DEVELOPMENT REPORT 2016

Box 5.10 Massive open online courses (MOOCs): A promising tool for lifelong learning (continued)

following its introduction in summer 2012, open courses to self-study are necessary characteristics to access and off ered by HarvardX and MITx, and hosted by edX, attracted complete MOOCs, but so are more basic elements such as an average of 1,300 new participants per course per day.a high-speed internet and digital skills. MOOCs are promising. People from anywhere around Yet, while their potential is large, there are questions the world can access postsecondary education from pres- about fi nancial sustainability, the value of online educa- tigious institutions with only a gadget and access to the tion, and the evaluation methods. Another study argues internet. Among adults who work and who have a certain that MOOCs are an improvement on noninteractive, online level of skills, online education can be a powerful force for courses, but that the current model is unsustainable. lifelong learning. These are individuals who would appre- Eventually, content will no longer be free, and star instruc- ciate the fl exibility (since they can combine it with work, tors will have to be paid like stars. And although they can which is one of the usual barriers for retraining). These are be an effi cient way for transferring content and reinforcing also workers who are likely to be best prepared for self- learning, especially in basic courses, the lack of face-to- guided learning. In addition, the courses provide access face interactions, the absence of individualized feedback to specialized topics that may not be available in the local from faculty, and the lack of access to complex infrastruc- economy. At the very least, MOOCs can complement offl ine ture such as labs, make MOOCs an unlikely competitor to learning. existing postsecondary education institutions, especially Coursera and edX have started to evaluate their pro- in the developed world.b Finally, in terms of evaluation, grams. One study fi nds that a slight majority of MOOC MOOCs are complementing multiple-choice and short- students are seeking certifi cation, and those opting for answer questions with peer-grading and peer-commenting fee-based, ID-verifi ed certifi cates have higher rates of com- systems, in which students help each other with assign- pletion. The study also fi nds that literacy and motivation ments and comment on each other’s work.

a. Ho and others 2015. b. Hoxby 2014.

Institutions: Connecting for Technology mixes appropriate to a capable and accountable institutions The policy agenda needs to be guided by the country government context, classifi ed here as emerging, transitioning, or Digital technologies have had limited impact on transforming depending on the extent to which gov- strengthening government capability and account- ernment uses digital technology and the quality of ability because of the misaligned incentives of policy political and administrative institutions (fi gure 5.10). makers and service providers—a gap between institu- Emerging countries are characterized by weak tions and technology. They have helped willing and institutions and, in fragile contexts, the institutions able governments serve their citizens better, but they may be failing. Politicians are often kept in power by a have not yet empowered citizens to make unwilling narrow set of elites who are largely unaccountable to governments more accountable. The policy agenda is the poor. The bureaucracy is largely patronage-based, thus to use digital technologies to strengthen insti- with little incentive to deliver services. These reform tutions by tailoring policies that are compatible with contexts are the most diffi cult, as the state is failing the incentives of politicians and providers. The policy not only to deliver services but also to perform the choices for different country contexts are informed most basic functions of maintaining security. The by the fi ndings in chapter 3 that for some services, priority in these countries is to lay the institutional digital technologies can substitute for weak institu- foundations so that governments can start perform- tions and be transformative—while for other services, ing basic functions. They should ensure that teachers digital technologies are only complementary and and doctors show up to work and are paid on time. require strong institutions to have an impact. They should curtail the outright theft of public funds NATIONAL PRIORITIES 273

Figure 5.10 Countries with more among them such fragile and confl ict-affected states accountable governments also adopt as Afghanistan, the Democratic Republic of Congo, more digital technologies Haiti, Somalia, the Republic of Yemen, and Zimbabwe. Transitioning countries have mixed institutions 2 characterized by pockets of pro-poor politicians and performance-oriented agencies. The priority for these countries should be to take advantage of this institu- 1 tional heterogeneity and to use digital technologies to strengthen government capability in the “islands of excellence” that exist, and through demonstration, 0 scale them up. The emphasis should also be on tar- geted transparency initiatives, those likely to be rele- vant to the daily lives of citizens and able to mobilize

−1 a broad coalition of supporters among businesses and other elites to align their incentives with those of willing politicians. Most developing countries fall

Quality of government institutions into this category. −2 In transforming contexts, governments are gener- ally accountable and capable, internet access is close to universal, many services are online, and many −3 back-offi ce systems are automated. One big challenge 0 0.2 0.4 0.6 0.8 1.0 in taking e-government to the next level is the rela- Technology: tively low use of online services by citizens, especially Digital Adoption Index (Governments) poor households. Within-government collaboration Sources: World Governance Indicators (World Bank, various years) and WDR across ministerial lines is weak, pushing up the trans- 2016 team. Data at http://bit.do/WDR2016-Fig5_10. action costs for citizens to interact with government. Note: The Digital Adoption Index (Governments) is the simple average of Another challenge is that government-citizen collabo- three normalized subindexes: core administrative systems, online public ser- vices, and digital identifi cation. Data for online public services are provided ration in policy making and service delivery is limited. by the UN’s Online Service Index. Data for core administrative systems and Policies should focus on deepening collaborative insti- digital identifi cation were compiled by the World Bank for this Report. tutions, both within government, through whole-of- government service delivery—and between govern- through transparency and accountability measures. ment and citizens, through participatory approaches, And they should strengthen alternative channels while instituting digital safeguards to protect privacy. of service delivery through nonstate providers. The Digital technologies can help in implementing focus, from the perspective of this Report, should be policies more effectively. The key is to align the in areas where digital technologies can be effective choice of digital technologies with their availability, substitutes for these weak initial institutional condi- skill requirements, and political incentives (table tions. Around 15 to 20 countries fall into this category, 5.5). While the focus here is on what to do, aligning

Table 5.5 A framework for policies: How to improve services in diff erent contexts

Emerging countries: Laying the Transitioning countries: Building Transforming countries: Deepening foundations for institutions capable and accountable institutions collaborative institutions • Improve informational services to • Strengthen government delivery • Improve collaboration across citizens systems government • Strengthen provider monitoring • Strengthen provider management • Enhance participatory policy and payment making • Get regular user feedback on • Establish population registers service quality • Scale up nonstate provision of • Increase transparency in priority services areas • Increase electoral accountability Source: WDR 2016 team. 274 WORLD DEVELOPMENT REPORT 2016

policies to institutions should also guide what not lack is an offi cial identity, which excludes them from to do. For example, the misalignment of policies and accessing many services, both public and private. So, political and administrative incentives is one of the a priority should be to build digital civil-identity sys- most common reasons for failed ICT interventions in tems to establish citizenship and become a platform government. As chapter 3 showed, many low-income for nonstate provision of services. countries have overinvested in low-impact adminis- The underlying policies can be grouped in fi ve - trative systems, resulting in fi scal waste. So, to use egories: improving informational services; strength- the popular metaphor, investing in “toilets, not the ening the monitoring and management of service internet” may indeed be advisable. Similarly, political providers and facilities; developing robust digital elites will have little interest in collaborative service population registers for citizen identity; strengthen- delivery or citizen engagement in clientelist systems, ing nonstate provision of services through for-profi t suggesting that an across-the-board push for open and not-for-profi t providers; and improving electoral data in emerging contexts is unlikely to be successful. accountability through better monitoring.

Emerging countries: Laying the Improve informational services to citizens foundations for institutions Mobile phones are delivering information to poor Given the persistent failure of many governments citizens in a variety of low-income countries with to deliver adequate services, the priority in emerging positive outcomes, and health applications are among contexts is to use digital technologies to begin build- the most promising uses of digital technology (see ing the foundations of government institutions and, sector focus 3). In Malawi, governments and non- where possible, to support alternative institutional governmental organizations send daily reminders channels for delivering services. The digital solutions to HIV-positive patients on their treatment schedule. should be relatively simple and low cost and should In the Democratic Republic of Congo, health call not require signifi cant business process reengineer- centers enable mothers to get answers to questions ing in government bureaucracies or interagency col- about their children’s health. And in Benin, Uganda, laboration. Where possible, they should substitute for and Zambia, mobile apps raise health awareness and ineffi cient government institutions. And to be incen- monitor disease outbreaks.66 Mobile phones have also tive compatible, they should deliver immediate and improved communication between governments and visible benefi ts for politicians without threatening the citizens for natural disasters and relief efforts—and elites who keep them in power. Examples of appropri- promoted literacy through daily text message stories ate technologies and complements appear in table 5.6. and teaching tips for teachers, with promising initial Although institutions are failing, mobile phones results (see sector focus 2). are ubiquitous, and many poor people in even the poorest countries have a cellphone. This almost Strengthen the monitoring and universal access opens many possibilities for mobile- management of service providers and phone applications to improve services and outcomes, facilities substituting for ineffective institutions in some Ghana, Niger, Pakistan, and Uganda have demon- services and functions. However, what poor people strated the effi cacy of mobile phones as a cost- effective monitoring technology to address absentee- Table 5.6 Emerging countries: ism among teachers and health workers. In principle, An agenda for laying institutional similar monitoring approaches can be applied to foundations and improving services address missing textbooks, drug stock-outs, and petty corruption. Scaling up from pilots to national pro- Technology Complements grams will, however, face bureaucratic resistance—to • Mobile phone–based • Civil society and be mitigated by focusing on rewards instead of sanc- informational services community involvement tions, such as fi nancial incentives and recognition • Mobile phone–based • Traditional media in the local community. A basic reward—or right—is monitoring and getting regularly paid for work, a problem in many management • Rewards and not sanctions low-income countries due to ineffi cient budget exe- • Digital population cution. In Haiti, three-quarters of school teachers sur- registers veyed had gone unpaid for months at a time.67 Mobile • Digital election payment platforms provide a low-cost solution, as monitoring in the Bridge International Academies in Kenya. If Source: WDR 2016 team. linked to digital population registers, they can also NATIONAL PRIORITIES 275 identify ghost workers, as in Nigeria, where a digi- monitoring capacity in government, and the collec- tal identifi cation scheme for civil servants removed tion and verifi cation of data to hold nonstate provid- about 60,000 fi ctitious workers from the government ers accountable. payroll, saving US$1 billion annually.68 Digital technologies can improve the impact of these schemes through better data collection, moni- Establish population registers toring, and dissemination of information on provider Digital population registers can establish citizen quality. Parents can make more informed decisions, identity and be leveraged later for a variety of applica- fi xing the market failures in private provision. Non- tions through appropriate credential verifi cation (see digital school report cards in Pakistan’s rural Punjab spotlight 4). The focus should be on developing the for example, improved parental information, lowered identity database and on the systems to ensure com- private school prices, and boosted school quality.71 pleteness and high quality. Only after the country has Digital technologies can make these choices easier developed harmonized identity registers can it legiti- through simpler versions of the school and health mately begin to tie e-services and issue the right cre- care provider rating systems that are now common- dentials to support them. In many cases, countries, place in the high-income countries. And govern- under vendor pressure, have prematurely procured ments, in the absence of parental choice, can better costly smart cards, which then remained unused as hold the private providers to account. the identity registers had not been developed fi rst. India focused on enrollment and unique identity and Improve electoral accountability launched the program without any smart cards or Digital technologies are improving both the sanctity credentials, just an Aadhaar number communicated of elections and providing citizens with meaningful to individuals. Now, more than fi ve years later, differ- and actionable information on government perfor- ent programs are issuing application-specifi c creden- mance. Although the number of electoral democ- tials linked to the Aadhaar framework and database. racies in poor countries has increased over the past two decades, the integrity of elections in these new Scale up nonstate provision of services democracies is low. Over half of the elections over Citizens in many low-income countries send their the past decade had irregularities either in the run-up children to nonstate schools (for-profi t or not-for- to the election or on election day.72 Elections are profi t) and seek care from private health providers. well-suited for digitally enabled monitoring. As high- Nonstate provision raises questions of equity and profi le events that attract signifi cant international quality. These risks can be mitigated through regu- attention and scrutiny, improving electoral integrity lations, disclosures, and public-private partnerships, may be possible even in politically diffi cult emerging such as voucher programs and contracting out. country contexts. These programs, if implemented well, can be highly Digital technologies can reduce election violence, effective. In an educational scheme for marginal as in Kenya and Mozambique, and uncover fraud in communities in rural Pakistan, the government paid vote-counting, as in Afghanistan. Digital identifi ca- the private provider a per-child subsidy, increasing tion is being increasingly used to register voters. In primary school enrollments and boosting test scores Pakistan for example, ahead of the 2013 parliamen- by 30 percentage points.69 These programs can also be tary elections, the digital identity database was used compatible with the interests of even clientelist pol- to clean the electoral rolls, leading to the removal of iticians, as they are likely to be supported by impor- 37 million voters with either no, invalid, or duplicate tant stakeholders like the business community and identities, and the addition of 36 million new voters, private service providers. mostly young and poor, who had valid identifi cation.73 Nonstate provision theoretically relies on the Similarly, in the 2015 presidential election in Nigeria, power of the market to solve accountability failures biometric identifi cation was used for the fi rst time in ways that public provision cannot. But in practice, to enroll 68 million voters and to eliminate 4 million parents may lack the choice of alternative providers duplicate identities (see spotlight 4). Despite these suc- or the information on provider quality to “vote with cesses though, biometric identifi cation is not without their feet” and hold nonstate providers accountable. its risks in emerging countries. Simpler, lower-cost The impact of low-cost private schools on student monitoring technologies like cellphones that require learning is generally positive, but in some cases they fewer institutional complements may be preferable.74 can be even worse than their public counterparts.70 Digital technologies can also improve electoral Performance agreements between governments and accountability by exposing corruption and abuse nonstate providers require some contracting and of offi ce, thereby better enabling voters to sanction 276 WORLD DEVELOPMENT REPORT 2016

lawbreaking politicians. Municipal audits in Brazil investments through complementary policies and and Mexico show that targeted digital transparency strengthen performance orientation in bureaucracies. initiatives can provide salient and credible informa- tion on corruption and on the quality of candidates Strengthen government delivery systems that is easy to understand and attribute to individual These countries should continue to expand e-gov- politicians. Widely disseminated before an election, ernment, particularly digital identifi cation systems such information can infl uence voters’ decisions. But and the business and citizen-facing services like such initiatives are contingent on a supportive legal online registration of businesses, e-fi ling of taxes, framework—such as right to information laws or e-procurement, and citizen service centers. But these disclosures of confl icts of interest and assets—or on investments should be conditional on enacting the independent supreme audit institutions that may be complementary reforms of regulatory changes, absent in these country settings. Civil society advo- improving interdepartmental and interagency coop- cacy, in partnership with traditional media, is neces- eration, and streamlining procedures. Digital technol- sary to uncover abuse and to make this information ogies can strengthen project management through available and understandable to voters. better monitoring of the different stages of the proj- ect cycle. By making procurement and contract mon- Transitioning countries: Building capable itoring more transparent, they can also give agencies and accountable institutions fl exibility in negotiating contracts with vendors, relaxing the procurement rigidities that cause many By introducing an automated complaint ICT projects to fail (box 5.11). Examples of appropriate management system we took a noose and put it technologies and complements appear in table 5.7. around our own necks. We are now accountable! —A manager of the Nairobi water utility Institutionalize user feedback on service quality Countries transitioning digitally have invested in Incorporating citizen feedback into policy maker- the automation of core government administration, to-provider management routines can be a powerful such as digital identifi cation, fi nancial and sector- mechanism for tackling petty corruption and improv- specifi c management information systems, and on ing services, as in the Dominican Republic, Kenya, government-to-citizen and government-to-business Nigeria, and Pakistan (chapter 3). It works especially services. They can increase the impact of these well for private goods and services that are easy to

Box 5.11 Increasing the impact of e-government systems

Better project management: The U.S. Offi ce of Management is not abused. The European Union allows agencies to negoti- and Budget in 2009 introduced the IT [Information ate with multiple bidders and award the contract to the ven- Technology] Dashboard, a public website that provides dor with the best revised bid. The greater discretion would detailed performance information on major IT investments not be possible without the greater monitoring, accountabil- by the federal government. That information is then used ity, and trust aff orded through the e-procurement system. to review problem projects at monthly TechStat meetings between the federal Chief Information Offi cer and the Public-private partnerships: A number of middle-income respective agencies. The Government Accountability Offi ce countries have implemented e-government projects found that these reforms improved the transparency and through public-private partnerships, minimizing the oversight of government IT spending. risks of failure associated with misaligned incentives and limited government capacity. Such partnerships are fea- Flexible and transparent IT procurement: E-government sible for revenue-raising services, for managing citizen projects often fail because of rigid procurement rules. But service centers, as in Brazil and India, and for government making procurement transparent can ensure that discretion e-procurement systems, as in the Indian state of Karnataka.

Sources: U.S. GAO 2014; Kenny 2014; Krishna 2015, for the WDR 2016. NATIONAL PRIORITIES 277

Table 5.7 Transitioning countries: monitoring technologies can assure citizens that their An agenda for building capable contributions are going to their chosen causes. Give- and accountable institutions and Directly, a Silicon Valley–based philanthropy, enables improving services well-off individuals all over the world to contribute funds directly to poor benefi ciaries in developing Technology Complements countries. The private sector has also been much more • Digital ID, G2G, G2C, • Streamlining of successful in getting citizens to create internet con- and G2B e-government procedures and e-procurement tent for public causes. Wikipedia, Yelp, Trip Advisor, systems • Improved and Amazon product reviews provide lessons for solv- interdepartmental ing collective action problems and engaging citizens • Digital performance cooperation management in public service delivery improvements. • Regulatory reforms • Targeted digital Strengthen provider management through transparency initiatives • Public-private partnerships for regular small-stakes monitoring • Digital platforms for fee-based services Most services and functions in government—teach- citizen feedback and ing, curative health, policy making, and manage- participation ment—require considerable discretion from workers Source: WDR 2016 team. and produce outputs and outcomes that are diffi cult Note: G2B = government-to-business; G2C = government-to-citizen; G2G = government-to-government. to monitor. Improving these services requires strong institutions. Digital technologies can only augment and not substitute for institutions. Even though mea- monitor, like drivers’ licenses, property transfer and suring performance is diffi cult for these services and registration, and water and electricity connections, functions, the processes of setting goals, of regularly since users have both the incentive and the ability communicating and discussing them with staff, and to provide feedback. Politicians are likely to support of associating small rewards, such as public recogni- it because it yields immediate and visible service tion, for teams and individuals in the achievement improvements without threatening elites and vested of even imprecisely measured goals can motivate interests. Similar approaches can monitor service workers to perform better (box 5.12). A survey of gov- providers, facilities, assets, and public spending. ernment workers in the Philippines found that a per- Digital platforms can also raise resources from citi- formance incentive scheme triggered improvements zens to fund infrastructure, hire contract teachers, and in management practices through goal setting and purchase drugs and educational materials, and digital improved teamwork.75 And a mobile phone–enabled

Box 5.12 Regular, small-stakes monitoring

Regular monitoring with small rewards or sanctions can be example, in Hawaii, revoking probation when probationers more eff ective than irregular monitoring with large rewards failed a drug test was for years an ineff ective strategy, or sanctions. The logic of this proposition is motivated by given the large numbers of violators and the impracticality the economic theory of crime that postulates that the incen- of sending them to prison for several years. But raising P tives for criminals to commit a crime are a function of the and lowering C—regular, randomized drug tests and jail probability (P) of getting caught multiplied by the severity sentences of two days—deterred parole violations because of the punishment (C). While standard economic analysis the punishment strategy became more credible. presumes that what matters is only the product P x C, the These fi ndings are generally applicable to other ser- evidence suggests that raising the severity of punishment vices. Digital technologies reduce the cost of monitoring in response to low probabilities of catching criminals is not and thus make it more feasible to give regular rewards and credible because many law enforcement offi cials are loathe sanctions. If the rewards and sanctions are small, they can to impose draconian punishments for smaller off enses. For be feasibly imposed, improving performance.

Source: Romer 2013. 278 WORLD DEVELOPMENT REPORT 2016

monitoring scheme increased the motivation of rural or on the quality of services that are easy to monitor, teachers in part by showing that offi cials in the far- like electricity and water. away ministry of education cared about their work and were looking out for them.76 Transforming countries: Deepening Digital technologies can improve goal setting and collaborative institutions within-government communication between policy Many transforming countries have improved service makers, managers, and providers. Many private orga- delivery, but only in some areas. Their services are nizations are introducing short “pulse” staff surveys not yet suffi ciently integrated, holding them back to get rapid responses from employees on particular from more ambitious interventions such as greater issues, to gauge employee motivation, and to moti- government-citizen collaboration in policy making vate workers through “employee of the month” public and service delivery. recognition schemes, and “gamifi cation” techniques like assigning points, leaderboards, badges, and other Increase citizen use of government services small rewards for the achievement of particular activ- through integrated whole-of-government ities.77 The ubiquity of mobile phones makes such digital solutions within-government feedback channels quite feasible The gap between the provision of e-services and even in low-income countries, and can certainly be their use is wide—even in digitally advanced coun- applied in the pockets of excellence in transitioning tries—and it is not closing. In Australia, Canada, contexts. While the measurement of performance and New Zealand, a majority of survey respondents and the feedback is subjective and imprecise, it is prefer traditional channels like phones for routine the regular communication and monitoring tied transactions such as paying taxes or registering chil- to the small-stakes rewards that improve worker dren for day care.79 This low use reveals the waste in motivation. e-government investments. Part of the explanation lies in less friendly experiences. Federal government Reinforce targeted transparency initiatives websites in the United States provide a poorer user Digital technologies can provide a new lease on life experience than comparable private ones.80 Of more for transparency initiatives such as the Open Govern- than 400 applications that the Korean government ment Partnership, an international platform for com- offers over smartphones, none meet the criteria of a mitted domestic reformers. But the initiatives should seamless, cross-agency, or cross-functional one-stop be targeted and of value to citizens and of interest to service.81 the private sector and other important stakeholders. Increasing use will require better integration of Consider the transparency of government contract- e-government across government, which requires ing. Governments annually spend over US$9 trillion breaking down agency silos and improving data shar- on procurement, at high risk of corruption, both in bid- ing across government. In the private sector, digitally ding and during contract execution.78 E-procurement enabled improvements in customer service, a core is technically easy to implement, yet, as described in differentiator in the service industry, are conditional chapter 3, developing countries have invested less in on high levels of shared knowledge in the fi rm.82 Esto- such systems than in the more complex budget and nia’s X-Road data-sharing protocol shows how data treasury systems—or have done so without the com- integration can work within the constraints of the plementary reforms to deliver results. E-procurement ministry and agency structure of government (box also has the potential to galvanize substantial support 5.13). Data integration is also essential for effective from the business community by reducing barriers to open data platforms that the private sector can use to entry—since it expands the market to new entrants, develop applications that citizens want. Examples of including international fi rms—and by opening politi- appropriate technologies and complements appear in cal space for reform. table 5.8. Digital technologies can also expand the range of actors that generate information, breaking the Expand participatory policy making and state’s monopoly in information provision. Because service delivery such information can be skewed toward the digitally Increasing use of digital technologies is also condi- savvy, it should be used where representativeness is tional on a better understanding of citizens’ needs and not a requirement and the risks of manipulation are thus on simplifi cation and customization. Behavioral small, as with real-time information on emergencies research has shown that individuals are “cognitive NATIONAL PRIORITIES 279

Box 5.13 Estonia’s X-Road

Estonia’s X-Road is an internet-based e-government queries were automated system-to-system exchanges. system that enables participating institutions, including The remaining one-third, about 113 million human queries, private business, to communicate and exchange data.a refl ects enormous demand for e-services from a population It serves as a platform for application development by of only 1.3 million. providing numerous common services to users, including The system’s main strength is that it is decentralized. query design, query tracking, and data visualization. Its Participating institutions retain ownership of their data, open design is protected by digital authentication, multi- but can share it or access other institutions’ data as neces- level authorization, log monitoring, and encrypted data sary. Estonia’s Public Information Act prohibits institutions transfers. This collective process improves the user expe- from requesting user information already stored in a data rience and motivates state institutions to develop digital repository connected to the X-Road.b Thus the system’s services and people to tap into digital government services. architecture—coupled with complementary policies—has The common goal is to shift activities from the physical reduced the need for repetitive data entry, increased gov- world to the much more effi cient digital realm. ernment effi ciency, and reduced costs to users. If e-services X-Road’s utility is evident in its exponential growth. The are assumed to yield 30 minutes in time saved per inter- system launched in 2003 with 10 participating institutions. action (for the service provider and the citizen) relative to By 2013, almost 900 had joined—70 percent are national predigital physical interaction, the number of applications or local government agencies, and the remainder private in 2014 implies a savings of more than 7 million work days fi rms. The annual number of queries through X-Road rose a year—5.4 work days for each citizen. from half a million to 340 million. In 2014, two-thirds of

Source: Vassil 2015, for the WDR 2016. a. See the Information System Authority’s website at https://www.ria.ee/x-road/. b. https://www.riigiteataja.ee/en/eli/522122014002/consolide.

Table 5.8 Transforming countries: simplifi cation and customization in their e-services to An agenda for deepening collaborative foster greater collaboration with citizens. institutions and improving services Technology Complements Digital safeguards • Integrated whole-of- • Breaking down government digital solutions of silos Reforms to strengthen the analog complements can ensure a high social and economic return from digital • Participatory policy making • Learning from the and service delivery private sector investments. But downside risks remain. Large-scale Source: WDR 2016 team. collection of identifi able information creates privacy and security concerns. Automation changes work in misers” who engage in motivated reasoning. Their ways that challenge existing social protections and decision making is often automatic, and they are reveal the inadequacy of existing labor laws. And more likely to search for and absorb information that scale economies create antitrust concerns. Digital confi rms their priors rather than process new infor- safeguards that mitigate these risks become more mation that does not.83 The private sector understands important as the digital transformation proceeds (fi g- this and simplifi es and customizes the user experi- ure 5.11). Competition and cybersafety are discussed ence to engage customers. Amazon’s search tool not elsewhere in this Report. This section focuses on only displays results the customer is looking for but another important safeguard for the digital economy, also recommends other products based on the cus- namely changes to social protection, taxation policy, tomer’s browsing history and the purchasing trends and labor market institutions. of other customers. Transforming countries can learn Digital technologies are accelerating job cre- from the private sector and embed these principles of ation and job destruction, requiring a more fl exible 280 WORLD DEVELOPMENT REPORT 2016

Figure 5.11 Digital safeguards in the WDR’s framework stringent hiring and fi ring practices.87 There are also implications for minimum wages, especially DIGITAL their growth path, if workers who earn close to the TECHNOLOGIES minimum wage are also in routine occupations sus- ceptible to automation. Tax systems that rely heavily on taxing labor also make labor more expensive in relation to capital and could become unsustainable if wage employment falls. In-work benefi ts, such as credits in the United States and the United Information Informalization and Winner-take-all Kingdom, are one mechanism for policy makers to deluge job displacement economy make labor taxes more progressive while improving incentives to work and hire.88 Digital technologies themselves can reduce non- PRIVACY SOCIAL ANTITRUST wage labor costs related to enforcement. In Brazil, POLICIES PROTECTION POLICIES the Annual Social Information report (digital records from social security with information about all work- Source: WDR 2016 team. ers, their wages, their occupations, and the types of fi rms they work for) monitors compliance with the Apprentice Law and, increasingly, other labor laws.89 workforce that responds to changing labor market Oman has a Worker Protection Scheme that allows for demands. They are also facilitating new forms of non- monitoring wage payments. Digital technologies can wage employment and work arrangements which, further reduce enforcement costs by shifting from while adding fl exibility to fi rms and workers, can also top-down to bottom-up accountability, empowering erode traditional employer-employee relationships workers and unions to convey complaints and vio- (chapter 2). This transformation of the world of work lations and to resolve confl icts. In the United States, requires a rethink of social protection, tax systems, workers can anonymously fi le online complaints and and labor market institutions. In incipient and tran- requests for inspections of their workplace if they sitioning countries, this agenda is less urgent, but believe there is a serious hazard or labor law violation; it will be important to build systems appropriate they can get information about their rights.90 They for the 21st century and not to blindly adopt models can also check online whether their employers have designed by advanced countries for an industrial paid their social security contributions. These tools era. increasing worker agency have the potential to bal- ance the loss of bargaining power for workers in non- Support technology adoption and level traditional work arrangements, such as independent the playing fi eld between workers and contractors in the gig economy, where many workers technology are freelancers or work online. Labor regulations, labor taxation and institutions need to support, rather than impede, technology Balance the relaxation of job protections adoption by fi rms. They should also avoid making with stronger workers’ protections labor unduly expensive, especially for workers who independent of work contracts compete more directly with digital technologies. Independent contracting, casual work, freelancing, Technology, by accelerating the pace of change in the and other new forms of work in online labor markets labor market, can make existing labor regulations and the sharing economy challenge the foundations obsolete and delay fi rms’ adjustments.84 And tight of most social protection and tax systems. In most labor regulations often apply to low-skilled workers,85 countries, social insurance schemes—for pensions, increasing the incentives for fi rms to substitute low- unemployment, and health—are tied to a (formal) job, skilled workers with labor-saving technology, accen- fi nanced through payroll taxes levied in the formal tuating the skill-bias of technological change. This sector. appears to be the case globally.86 The evolving nature of work increases the need Countries with unduly protective labor markets to delink social insurance from the labor contract. could be slowing their digital transformation, and Bolivia, Chile, and Costa Rica have been grappling distorting fi rms’ decisions to automate work. This with providing social insurance to noncontributing, is likely to affect countries that have particularly nonwage, informal workers for a long time. Their NATIONAL PRIORITIES 281 experiences suggest that all individuals should be the two pillars of their digital strategy. For details, see registered in the same social insurance system, Rwanda’s ICT strategy at http://www.rdb.rw/uploads regardless of where they work, with subsidies for the /tx_sbdownloader/NICI_III.pdf. poor or low-wage earners, and with fi nancing coming 6. See Hanna 2015. 7. See http://www.plugintheworld.com/mobisol/. from general revenues.91 8. Many developing countries adopt the competition The labor market disruptions that accompany frameworks of more advanced economies, but fail technological change increase the demands on active to implement them adequately. labor market policies and social assistance systems. 9. Fines imposed by competition authorities remain The disruptions are likely to be greatest for workers symbolic in several countries. In Armenia, Kenya, in routine occupations. Some of them are going to and Central American countries, the current struc- need intermediation and retraining services to fi nd ture of fi nes and sanctions is not conducive to deter- new jobs. Others, if they have diffi culties transition- ring anticompetitive conduct. In Armenia, maxi- ing, may need social assistance. Recent technological mum fi nes for price-fi xing cartels are 1.5 percent of changes are bringing to the fore discussions of a guar- the average turnover of the 100 largest taxpayers, compared with a best practice of about 10 percent of anteed basic income, especially in more advanced a fi rm’s turnover in other countries. countries. And giving workers a stake in digital cap- 10. According to data from the market-research fi rm ital could diversify workers’ assets and reduce their IDC. costs of displacements—through pension funds, 11. Rochet and Tirole 2006. mutual funds, or even more directly in fi rms active in 12. World Bank 2012a; Bodewig and others 2014; World the digital economy.92 Bank 2011; Cunningham and Villasenor 2014. A solution is to protect workers rather than jobs, 13. Functional illiteracy is defi ned as the proportion of and to level the playing fi eld for regulations and tax- exam takers who score below a level 2 in the PISA ation across work contracts. If workers are protected reading test. WDR 2016 team, based on OECD PISA outside their labor contract, regulations and taxes can 2012 scores. 14. Guerra, Modecki, and Cunningham 2014; Shonkoff be more lax across the board. A fi rst step is to do away and Phillips 2000; Almlund and others 2011; Cunha, with regulations that almost prohibit fl exible work Heckman, and Schennach 2010; Cunha and Heck- arrangements. In Montenegro contracts for part- man 2007. time employment cannot be less than 10 hours per 15. STEP household surveys and OECD’s Programme week.93 Another step is to reform tax systems that tax for International Student Assessment (PISA). part-time work at higher per hour rates than full-time 16. Guerra, Modecki, and Cunningham 2014; Bodewig work. In Serbia, the reference wage (determining and others 2014. a minimum social contribution) is not adjusted for 17. Arias and others 2014; Sondergaard and others 2012. hours worked, so that social contributions are dispro- 18. Gates Foundation 2015. 19. http://www.text2teach.org.ph/?page_id=2. portionately high for part-time workers.94 Reforms 20. http://www.bridgeinternationalacademies.com are also necessary in working time arrangements.95 /company/about/. 21. http://www.britishcouncil.co/en/about/english Notes -education-solutions/success-stories/remote-in -service-teacher-training-english. 1. See, for example, https://e-estonia.com. 22. Kremer and Holla 2009; and Kremer, Brannen, and 2. Decker and others 2014. Glennerster 2013. 3. See Atkinson and Miller 2015. 23. Bruns and Luque 2014. 4. See OECD 2015. 24. Banerjee and others 2007. 5. The fi rst stage (2000–05) of Rwanda’s National 25. Bruns and Luque 2014. Information and Communication Infrastructure 26. Bruns and Luque 2014. policy prepared the groundwork for the ICT sec- 27. Barrera-Osorio and Linden 2009. tor. The second phase (2006–10) concentrated on 28. Devitt, Lyons, and McCoy 2014. enhancing ICT infrastructure. The third phase 29. http://www.text2teach.org.ph/wp-content/uploads (2011–15) focused on improving service delivery. /2012/06/Phase3-by-third-party.pdf. The fi nal phase (2016–20) is expected to concentrate 30. Linden 2008. on the enhancement of skills and development of 31. Bruns and Luque 2014. the private sector and community. Instead of fi rst 32. McEwan 2013; Kremer, Brannen, and Glennerster dealing with the supply-side issues and then the 2013. demand-side issues, developing countries would be 33. World Development Indicators (World Bank, vari- better off by working to simultaneously strengthen ous years). 282 WORLD DEVELOPMENT REPORT 2016

34. Aker, Ksoll, and Lybbert 2012. ber 22, 2014. http://www.news24.com/Africa/News 35. Ksoll and others 2014. /Tanzania-orders-probe-into-ghost-government 36. In a slum in New Dehli, a freely accessible computer -workers-20150329. was put up for use in a hole in a wall. With no prior 69. Barrera-Osorio and others 2013. experience, the children learned to use the computer 70. Tooley and Longfi eld 2015. on their own (see http://www.hole-in-the-wall.com 71. Andrabi, Das, and Khwaja 2014. /Beginnings.html). 72. Bishop and Hoeffl er 2014. 37. Research ICT Africa surveys (various years), dis- 73. Malik 2014. cussed in chapter 2. 74. Gelb and Clark 2013. 38. Kuek and others, forthcoming. 75. World Bank 2015. 39. King and Rogers 2014. 76. Aker and Ksoll 2015. 40. Rathunde and Csikszentmihalyi 2005. 77. Banuri 2015. 41. Heckman and others 2010. 78. Kenny 2014. 42. For more information on the New School model, 79. Reddick and Turner 2012. see http://www.escuelanueva.org/portal/en/escuela 80. Morgeson and Mithas 2009. -nueva-model.html. 81. Eom and Kim 2014. 43. In FYR Macedonia, for example, the government is 82. Ray, Muhanna, and Barney 2005. starting a pilot introducing growth mind-set and 83. World Bank 2015. grit training in primary schools; see Bodewig and 84. Alesina, Battisti, and Zeira 2015. others (2014) for a discussion on ongoing reforms in 85. World Bank 2012b; Hamermesh 2014; Kuddo 2015. Vietnam. 86. Alesina, Battisti, and Zeira 2015; O’Mahoney and Van 44. Halinen 2015. Ark 2003. 45. The “Thinking Schools, Learning Nation” concept 87. Kuddo 2015, based on World Economic Forum’s was articulated by the Singaporean prime minister executive opinion survey on hiring and fi ring in a speech on June 2, 1997. http://www.moe.gov.sg practices. /media/speeches/1997/020697.htm. 88. Chetty and Saez 2009; Blundell and Brewer 2000; 46. Goh and Gopinathan 2008. Blundell, Duncan, and Meghir 2002; Hotz and 47. Guerra, Modecki, and Cunningham 2014. Scholz 2001; Trampe 2007. 48. Guerra, Modecki, and Cunningham 2014. 89. Silva, Almeida, and Strokova 2014. 49. See http://www.oecd.org/site/piaac/ and Pierre, 90. https://www.osha.gov/as/opa/worker/complain Sanchez Puerta, and Valerio 2014. .html. 50. http://www.bbc.com/news/world-asia-china 91. Ribe, Robalino, and Walker 2010. -31617597. 92. Freeman 2015. 51. See Duckworth and others 2007, for grit; and Dweck 93. Arias and others 2014. 2006, for growth mind-set. 94. Arias and others 2014. 52. Trucano 2015. 95. Kuddo 2015. 53. Resnik 2013. 54. http://www.codecademy.com/. 55. http://www.nairobits.com/. References 56. Beede 2011. 57. OECD PISA 2006. Adelman, Melissa, Moussa Blimpo, David K. Evans, 58. Waldrop 2013. Atabanam Simbou, and Noah Yarrow. 2015. “Enhanced 59. For more information on College to Careers, the Teacher Monitoring for Student Results in Haiti.” fl agship program of Chicago’s reform efforts, see Working paper, World Bank, Washington, DC. http://www.ccc.edu/menu/Pages/college-to-careers Aker, Jenny C., and Christopher Ksoll. 2015. “Call Me Edu- .aspx. cated: Evidence from a Mobile Monitoring Experi- 60. Almeida, Behrman, and Robalino 2012. ment in Niger.” Working Paper 406, Center for Global 61. Almeida, Behrman, and Robalino 2012. Development, Washington, DC. 62. WDR 2016 team interview with INFOTEP (April Aker, Jenny C., Christopher Ksoll, and Travis J. Lybbert. 2015). 2012. “Can Mobile Phones Improve Learning? Evi- 63. Almeida, Behrman, and Robalino 2012. dence from a Field Experiment in Niger.” American 64. Almeida, Behrman, and Robalino 2012. Economic Journal: Applied Economics 4 (4): 94–120. 65. OECD 2004. Aker, Jenny C., and Isaac M. Mbiti. 2010. “Mobile Phones 66. Aker and Mbiti 2010. and Economic Development in Africa.” Journal of Eco- 67. Adelman and others 2015. nomic Perspectives 24 (3): 207–32. 68. Statements of then-minister of fi nance Dr. Ngozi Alesina, Alberto, Michele Battisti, and Joseph Zeira. 2015. Okkonjo-Iweala, as reported by the media on Octo- “Technology and Labor Regulations: Theory and Evi- NATIONAL PRIORITIES 283

dence.” NBER Working Paper 20841, National Bureau Bishop, Sylvia, and Anke Hoeffl er. 2014. “Free and of Economic Research, Cambridge, MA. Fair Elections: A New Database.” Working Paper Almeida, Rita, , and David Robalino. 2012. WPS/2014-14, Centre for the Study of African Econo- The Right Skills for the Job? Human Development Per- mies, Oxford, U.K. spectives. Washington, DC: World Bank. Blundell, Richard, and Mike Brewer. 2000. “Tax and Almlund, Mathilde, Angela Duckworth, James Heckman, Benefi t Reform in the UK: Making Work Pay.” Com- and Tim D. Kautz. 2011. “Personality Psychology and ment paper from host country expert: U.K. Draft Economics.” In Handbook of the Economics of Education, peer review for the European Commission, Brussels, edited by Machin Hanushek and Ludger Woess- November 6. mann, 1–181. Amsterdam: Elsevier. Blundell, Richard, Alan Duncan, and Costas Meghir. Andrabi, Tahir, Jishnu Das, and Asim Khwaja. 2014. 2002. “Evaluating the Working Families .” “Report Cards: The Impact of Providing School and Social Policy Monitoring Network background paper, Child Test Scores in Educational Markets.” Faculty Inter-American Development Bank, Washington, DC. research working paper, Harvard Kennedy School, Bodewig, Christian, Reena Badiani-Magnusson, Kevin Cambridge, MA. Macdonald, David Newhouse, and Jan Rutkowski. Arias, Omar, Carolina Sanchez-Paramo, Maria Davalos, 2014. Skilling Up Vietnam: Preparing the Workforce for Indhira Santos, Erwin Tiongson, Carola Gruen, Nata- a Modern Market Economy. Washington, DC: World scha de Andrade, Gady Saiovici, and Cesar Cancho. Bank. 2014. Back to Work: Growing with Jobs in Europe and Cen- Bruns, Barbara, and Javier Luque. 2014. Great Teachers: tral Asia. Washington, DC: World Bank. How to Raise Student Learning in Latin America and the Arias Ortiz, Elena, and Julián Cristia. 2014. “The IDB and Caribbean. Washington, DC: World Bank. Technology in Education: How to Promote Effec- Chetty, Raj, and Emmanuel Saez. 2009. “Teaching the tive Programs?” IDB Technical Note IDB-TN-670, Tax Code: Earnings Responses to an Experiment Inter-American Development Bank, Washington, with EITC Recipients.” NBER Working Paper 14836, DC. National Bureau of Economic Research, Cambridge, Atkinson, Robert D., and Ben Miller. 2015. “A Policymak- MA. er’s Guide to Spurring ICT Adoption.” Information Colombia Aprende [Colombia Learns]. 2015. “Descripción Technology and Innovation Foundation, Washing- del modelo [Model description].” Colombia Aprende ton, DC. [Colombia learns]: La red de conocimiento [The Autor, David. 2014. “Polanyi’s Paradox and the Shape of knowledge network]. http://www.colombiaaprende Employment Growth.” NBER Working Paper 20485, .edu.co/html/mediateca/1607/article-89868.html. National Bureau of Economic Research, Cambridge, Cunha, , and James Heckman. 2007. “The Technol- MA. Banerjee, Abhijit, Shawn Cole, Esther Dufl o, and Leigh ogy of Skill Formation.” American Economic Review 97 Linden. 2007. “Remedying Education: Evidence from (2): 31–47. Two Randomized Experiments in India.” Quarterly Cunha, Flavio, James Heckman, and Susanne Schennach. Journal of Economics 122 (3): 1235–64. 2010. “Estimating the Technology of Cognitive and Banuri, Sheheryar. 2015. “ICT and Public Service.” Back- Noncognitive Skill Formation.” Econometrica 78 (3): ground paper for the World Development Report 2016, 883–931. World Bank, Washington, DC. Cunningham, Wendy, and Paula Villasenor. 2014. Barrera-Osorio, Felipe, David Blakeslee, Matthew “Employer Voices, Employer Demands, and Impli- Hoover, Leigh Linden, and Dhushyanth Raju. 2013. cations for Public Skills Development Policy.” Policy “Expanding Educational Opportunities in Remote Research Working Paper 6853, World Bank, Wash- Parts of the World: Evidence from an RCT of a ington, DC. Public-Private Partnership from Pakistan.” enGender Davalos, Maria, and Indhira Santos. Forthcoming. Impact (database), World Bank, Washington, DC. Addressing the Youth Employment Challenge in Europe and Barrera-Osorio, Felipe, and Leigh Linden. 2009. “The Use Central Asia. Washington, DC: World Bank. and Misuse of Computers in Education: Evidence Decker, Ryan, John Haltiwanger, Ron S. Jarmin, and Jav- from a Randomized Experiment in Colombia.” Policy ier Miranda. 2014. “The Secular Decline in Business Research Working Paper, Impact Evaluation Series Dynamism in the U.S.” Working paper, University of 4836, World Bank, Washington, DC. Maryland, College Park. Becker, Katrin. 2007. “Digital Game-Based Learning, Devitt, Niamh, Sean Lyons, and Selina McCoy. 2014. “Per- Once Removed: Teaching Teachers.” British Journal of ceived Benefi ts and Barriers to the Use of High-Speed Educational Technology 38 (3): 478–88. Broadband in Ireland’s Second-Level Schools.” MPRA Beede, David. 2011. “Women in STEM: A Gender Gap to Paper 60185, University Library of Munich. Innovation: Executive Summary.” Economics and Di Castri, Simone, and Lara Gidvani. 2014. “Enabling Statistics Administration Issue Brief No. 04–11, U.S. Mobile Money Policies in Tanzania: A Test and Learn Department of Commerce, Washington, DC. Approach to Enabling Market-Led Digital Financial 284 WORLD DEVELOPMENT REPORT 2016

Services.” Mobile Money for the Unbanked, GSMA Labor, Bonn, Germany. http://wol.iza.org/articles (Groupe Speciale Mobile Association), London. /do-labor-costs-affect-companies-demand-for-labor Duckworth, Angela, Christopher Peterson, Michael Mat- .pdf. thews, and Dennis Kelly. 2007. “Grit: Perseverance Hanna, Nagy. 2015. “Transforming to a Networked and Passion for Long-Term Goals.” Journal of Personal- Society: Guide for Policy Makers.” http://www ity and Social Psychology 92 (6): 1087–1101. .ericsson.com/us/res/docs/2014/transforming-to-a Durkee, Tony, and others. 2012. “Prevalence of Patho- -networked-society.pdf. logical Internet Use among Adolescents in Europe: Hawkins, Robert J. 2002. “Ten Lessons for ICT and Edu- Demographic and Social Factors.” Addiction 107 (12): cation in the Developing World.” In The Global Infor- 2210–22. mation Technology Report 2001–02: Readiness for the Net- Dweck, Carol. 2006. : The New Psychology of Success. worked World, edited by G. Kirkman, 38–44. Oxford, New York: Ballantine Books. U.K.: Oxford University Press. Eom, Seok-Jin, and Jun Houng Kim. 2014. “The Adop- He, Fang, Leigh L. Linden, and Margaret MacLeod. tion of Public Smartphone Applications in Korea: 2007. “Helping Teach What Teachers Don’t Know: Empirical Analysis on Maturity Level and Infl uential An Assessment of the Pratham English Language Factors.” Government Information Quarterly 31 (June): Program.” Unpublished report, available at http:// S26–S36. www.cid.harvard.edu/neudc07/docs/neudc07_s6_ Forero-Pineda, C., D. Escobar-Rodriguez, and D. Molina. p02_he.pdf. 2006. Escuela Nueva’s Impact on the Peaceful Social Inter- Heckman, James J., Seong Hyeok Moon, Pinto, action of Children in Colombia. Education for All and Peter A. Savelyev, and Adam Yavitz. 2010. “The Rate of Multigrade Teaching: Challenges and Opportunities. Return to the HighScope Perry Preschool Program.” Dordrecht, Netherlands: Springer. Journal of 94 (1): 114–28. Freeman, Richard B. 2015. “Who Owns the Robots Rules Ho, A. D., I. Chuang, J. Reich, C. Coleman, J. Whitehill, the World.” IZA World of Labor, Bonn, Germany. C. Northcutt, J. J. Williams, J. Hansen, G. Lopez, and Fundación Escuela Nueva [New School Foundation]. R. Petersen. 2015. “HarvardX and MITx: Two Years of 2015. “Escuela Nueva en el mundo [New School in the Open Online Courses.” HarvardX Working Paper 10, world].” http://www.escuelanueva.org/portal/es Cambridge, MA. /modelo-escuela-nueva/escuela-nueva-en-el-mundo Hotz, Joseph V., and John Karl Scholz. 2001. “The Earned .html. Income Tax Credit.” NBER Working Paper 8078, Gates Foundation. 2015. “Annual Letter: Our Big Bet National Bureau of Economic Research, Cambridge, for the Future.” Bill and Melinda Gates Foundation, MA. Seattle, WA. http://www.gatesnotes.com/2015 Hoxby, Caroline. 2014. “The Economics of Online Post- -annual-letter?WT.mc_id=01_21_2015_DO_GFO secondary Education: MOOCs, Nonselective Educa- _domain_0_00&page=0&lang=en. tion, and Highly Selective Education.” NBER Working Gelb, Alan, and Julia Clark. 2013. “Identifi cation for Devel- Paper 19816, National Bureau of Economic Research, opment: The Biometrics Revolution.” Working Paper Cambridge, MA. 315, Center for Global Development, Washington, DC. ISTE (International Society for Technology in Education) Goh, Chor Boon, and S. Gopinathan. 2008. “Education in and CSTA (Computer Science Teachers Association). Singapore: Developments since 1965.” In An African 2011. “Operational Defi nition of Computational Exploration of the East Asian Education, edited by Birger Thinking for K–12 Education.” http://csta.acm.org Fredriksen and Jee Peng Tan, 80–108. Washington, /Curriculum/sub/CurrFiles/CompThinkingFlyer.pdf. DC: World Bank. Jarvilehto, L. 2014. Learning as Fun. Espoo, Finland: Rovio Grabowski, Barbara, and Kel Fengfeng. 2007. “Gameplay- Entertainment. ing for Maths Learning: Cooperative or Not?” British Kenny, Charles. 2014. “Publishing Government Con- Journal of Educational Technology 38 (2): 249–59. tracts: Addressing Concerns and Easing Implemen- Guerra, Nancy, Kathryn Modecki, and Wendy Cunning- tation.” Center for Global Development, Washington, ham. 2014. “Developing Socio-Emotional Skills for DC. the Labor Market: The PRACTICE Model.” Policy King, Elizabeth, and Halsey Rogers. 2014. “Intelligence, Research Working Paper 7123, World Bank, Washing- Personality and Creativity: Unleashing the Power of ton, DC. Intelligence and Personality to Build a Creative and Halinen, Irmeli. 2015. “What Is Going On in Finland? Innovative Economy.” Paper presented at the sympo- Curriculum Reform 2016.” Current Issues (blog), March sium “Achieving HOPE: Happiness of People through 25. http://www.oph.fi/english/current_issues/101/0 Education: Innovation in Korean Education for a /what_is_going_on_in_fi nland_curriculum_reform Creative Economy,” Seoul, Korea, November 4. _2016. Kremer, Michael, Conner Brannen, and Rachel Glenner- Hamermesh, Daniel. 2014. “Do Labor Costs Affect ster. 2013. “The Challenge of Education and Learning Companies’ Demand for Labor?” IZA World of in the Developing World.” Science 340 (6130): 297–300. NATIONAL PRIORITIES 285

Kremer, Michael, and Alaka Holla. 2009. “Improving Morgeson, Forest, and Sunil Mithas. 2009. “Does Education in the Developing World: What Have We E-Government Measure Up to E-Business? Compar- Learned from Randomized Evaluations?” Annual ing End User Perceptions of U.S. Federal Government Review of Economics One: 513–42. and E-Business Web Sites.” Public Administration Krishna, Gunjan. 2015. “Improving Service Delivery Review 69 (4): 740–52. through the Use of Digital Technologies: A Case Murphy, Robert, Larry Gallagher, Andrew E. Krumm, Jes- Study of Karnataka E-Procurement.” Background sica Mislevy, and Amy Hafter. 2014. “Research on the paper for the World Development Report 2016, World Use of Khan Academy in Schools: Implementation Bank, Washington, DC. Report.” SRI International, Princeton, NJ. Ksoll, Christopher, Jenny C. Aker, Danielle Miller, Karla C. OECD (Organisation for Economic Co-operation and Perez-Mendoza, and Susan L. Smalley. 2014. “Learn- Development). 2004. “Co-fi nancing Lifelong Learn- ing without Teachers? A Randomized Experiment of ing: Towards a Systemic Approach.” OECD, Paris. a Mobile Phone–Based Adult Education Program in ————. 2015. “Digital Economy Outlook.” OECD, Paris. Los Angeles.” Working Paper 368, Center for Global OECD PISA. 2006. “The Programme for International Development, Washington, DC. Student Assessment.” OECD, Paris. Kuddo, Arvo. 2015. “Labor Regulations and Institutions ————. 2012. “PISA 2012 Results in Focus: What 15-Year- for the New Economy.” Background paper for the Olds Know and What They Can Do with What They World Development Report 2016, World Bank, Washing- Know.” OECD, Paris. ton, DC. O’Mahoney, Mary, and Bart Van Ark. 2003. “EU Produc- Kuek, Siou Chew, Cecilia Paradi-Guilford, Toks Fayomi, tivity and Competitiveness: An Industry Perspective. Saori Imaizumi, and Panos Ipeirotis. Forthcoming. Can Europe Resume the Catching-Up Process?” Euro- The Global Opportunity in Online Outsourcing. Washing- pean Commission, Brussels. ton, DC: World Bank. Pierre, Gael, Maria Laura Sanchez Puerta, and Alexandria Light, Daniel, and Elizabeth Pierson. 2014. “Increasing Valerio. 2014. “STEP Skills Measurement Surveys: Innovative Tools for Assessing Skills.” Working Paper Student Engagement in Math: The Use of Khan Acad- 89729, World Bank, Washington, DC. emy in Chilean Classrooms.” International Journal of Plaza, Sonia, Seyed Reza Yousefi , and Dilip Ratha. 2015. Education and Development using Information and Com- “Technological Innovations and Remittance Costs.” munication Technology 10 (2): 103–19. Background paper for the World Development Report Linden, Leigh. 2008. “Complement or Substitute? The 2016, World Bank, Washington, DC. Effect of Technology on Student Achievement in Prensky, Marc. 2006. Don’t Bother Me Mom—I’m Learning. India.” Working paper, Columbia University, New St. Paul, MN: Paragon House. York. ————. 2007. Digital Game-Based Learning. St. Paul, MN: Malik, Tariq. 2014. “Technology in the Service of Devel- Paragon House. opment: The NADRA Story.” Center for Global Devel- Pritchett, Lant, Michael Woolcock, and Salimah Samji. opment, Washington, DC. Forthcoming. What Kind of Organization Capability Is Martinez, Ana, Serrana Alonso, and Diego Diaz. 2009. Needed? Monitoreo y evaluacion de impacto social del plan CEIBAL: Rathunde, Kevin, and Mihaly Csikszentmihalyi. 2005. Metodologia y primeros resultados a nivel nacional “Middle School Students’ Motivation and Quality [Monitoring and social impact evaluation of the of Experience: A Comparison of Montessori and CEIBAL plan: Methodology and fi rst national results]. Traditional School Environments.” American Journal of Plan CEIBAL, Montevideo, Uruguay. Education 111 (3): 341–71. Mayo, M. 2009. “Video Games: A Route to Large-Scale Ray, Gautam, Waleed A. Muhanna, and Jay B. Barney. STEM Education?” Science 323 (5910): 79–82. 2005. “Information Technology and the Performance McEwan, Patrick. 2013. “Improving Learning in Primary of the Customer Service Process: A Resource-Based Schools of Developing Countries: A Meta-Analysis Analysis.” Management Information Systems Quarterly of Randomized Experiments.” Review of Educational 29 (4): 625–52. Research 20 (10): 1–42. Reddick, Christopher, and Michael Turner. 2012. “Chan- McFarlane, A., A. Sparrowhawk, and Y. Heald. 2002. nel Choice and Public Service Delivery in Canada: “Report on the Educational Use of Games.” TEEM: Comparing E-Government to Traditional Service Department for Education and Skills, University of Delivery.” Government Information Quarterly 29 (1): 1–11. Bristol, Bristol, U.K. Research ICT Africa. Various years. Household survey. Monroy-Taborda, Sebastian, Martin Moreno, and Indhira Capetown, South Africa. http://www.researchictafrica Santos. Forthcoming. “Technology Use and Changing .net/home.php. Skills Demands: New Evidence from Developing Resnik, Mitchel. 2013. “Learn to Code, Code to Learn.” Countries.” Background note for the World Develop- edSurge Newsletters. https://www.edsurge.com/n ment Report 2016, World Bank, Washington, DC. /2013-05-08-learn-to-code-code-to-learn. 286 WORLD DEVELOPMENT REPORT 2016

Ribe, Helena, David Robalino, and Ian Walker. 2010. Help Improve Effi ciency and Effectiveness.” GAO, Achieving Eff ective Social Protection for All in Latin America Washington, DC. and the Caribbean. Washington, DC: World Bank. Vassil, Kristjan. 2015. “Estonian e-Government Ecosys- Rochet, J. C., and J. Tirole. 2006. “Two-Sided Markets: tem: Foundation, Applications, Outcomes.” Back- A Progress Report.” RAND Journal of Economics 37 (3): ground Paper for the World Development Report 2016, 645–67. World Bank, Washington, DC. Romer, Paul M. 2010. “What Parts of Globalization Mat- Waldrop, Mitchell. 2013. “Education Online: The Virtual ter for Catch-Up Growth?” American Economic Review Lab.” Nature 499: 268–70. Papers and Proceedings 100 (2): 94–98. WEF (World Economic Forum). Various years. Compet- ————. 2013. “Small Stakes, Good Measurement.” Urban- itiveness Index (database). WEF, Geneva, http:// ization Project (blog), July 2. http://urbanizationproject knoema.com/WFGCI2014/the-global-com .org/blog/small-stakes-good-measurement/. petitiveness-index-2014-2015-data-platform-2014. Shonkoff, Jack P., and A. Phillips. 2000. From WITS (Workshop on Information Technologies and Sys- Neurons to Neighborhoods: The Science of Early Childhood tems). 2014. “Value Creation from Innovative Tech- Development. Washington, DC: National Academy nologies.” Auckland, New Zealand. Press. World Bank. Various years. Doing Business (database). Silva, Joana, Rita Almeida, and Victoria Strokova. 2014. World Bank, Washington, DC, http://www.doing Sustaining Employment and Wage Gains in Brazil: A Skills business.org/. and Jobs Agenda. Directions in Development Series. ————. Various years. Enterprise Surveys (database). Washington, DC: World Bank. World Bank, Washington, DC, http://www.enter- Sondergaard, Lars, Mamta Murthi, Dina Abu-Ghaida, prisesurveys.org/. Christian Bodewig, and Jan Rutkowski. 2012. Skills, ————. Various years. Global E-Government Systems Not Just Diplomas: Managing Education for Results in (database). World Bank, Washington, DC, http://data Eastern Europe and Central Asia. Directions in Develop- .worldbank.org/data-catalog/pfm-systems-eservices ment Series. Washington, DC: World Bank. -dataset. Spada, Marcantonio. 2014. “An Overview of Problematic ————. Various years. ID4D (Identifi cation for Develop- Internet Use.” Addictive Behaviors 39 (1): 3–6. ment database). World Bank, Washington, DC, http:// Sparrow, Betsy, Jenny Liu, and Daniel Wegner. 2011. data.worldbank.org/data-catalog/id4d-dataset. “Google Effects on Memory: Cognitive Consequences ————. Various years. Logistics Performance Index of Having Information at Our Fingertips.” Science 5 (database). World Bank, Washington, DC, http://lpi. 333 (6043): 776–78. worldbank.org/. Tooley, James, and David Longfi eld. 2015. “Rejoinder ————. Various years. World Development Indicators to ‘Private Schools Rigorous Review Team’ on the (database). World Bank, Washington, DC, http:// Role and Impact of Private Schools in Developing data.worldbank.org/data-catalog/world-development Countries: A Response to the DFID-Commissioned -indicators. Rigorous Literature Review.” https://egwestcentre ————. Various years. World Governance Indicators .files.wordpress.com/2015/03/tooley-longfield (database). World Bank, Washington, DC, http://info -rejoinder.pdf. .worldbank.org/governance/wgi/index.aspx#home. Trampe, Paul. 2007. “The EITC Disincentive: The Effects ————. 2004. World Development Report 2004: Making on Hours Worked from the Phase-Out of the Earned Services Work for Poor People. Washington, DC: World Income Tax Credit.” Econ Journal Watch 4 (3): 308–20. Bank. Trucano, Michael. 2012. “Evaluating One Laptop per Child ————. 2011. World Development Report 2012: Gender Equal- (OLPC) in Peru.” EduTech (blog), March 23. http://blogs ity and Development. Washington, DC: World Bank. .worldbank.org/edutech/olpc-peru2. ————. 2012a. Lebanon, Good Jobs Needed: The Role of Macro, ———— . 2015. “Establishing and Connecting Leagues of Investment, Education, Labor and Social Protection Policies Innovative Schools around the World.” World Bank (MILES). Report of the Middle East and North Africa Blogs, March 10. http://blogs.worldbank.org/category Human Development Group. Washington, DC: /tags/european-schoolnet. World Bank. United Nations. 2014. “E-Government Survey 2014: ————. 2012b. World Development Report 2013: Jobs. Wash- E-Government for the Future We Want.” United ington, DC: World Bank. Nations, New York. ————. 2015. World Development Report 2015: Mind, Society, U.S. GAO (Government Accountability Offi ce). 2014. and Behavior. Washington, DC: World Bank. “Information Technology: Reform Initiatives Can

288 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 5

Energy

Reliable and affordable access to electricity services rural areas the surrounding communities lack access is fundamental to achieving the World Bank Group’s to electricity. Energy service companies in many parts twin goals: shared prosperity and elimination of of Africa have been oversizing the energy generation extreme poverty by 2030. Efforts to meet the Agenda unit of the tower and providing local consumers with 2030 Sustainable Development Goal to “ensure access access to electricity. The Mobile for Development to affordable, reliable, sustainable, and modern energy program of the Groupe Speciale Mobile Association for all” will also require increased investment, new (GSMA) explores such opportunities.4 and refi ned technologies and systems, and institu- To address some of the fi nancing challenges of tional reform. To help the world achieve this energy rural energy services, companies in Africa and South transformation, the effective use of information and Asia are leveraging ICT and sensors utilizing cellular communication technology (ICT), including data and networks through various fi nancing models.5 For advanced analytics, is already playing an essential role. example, in the same value chain all the way from electricity distribution to maintenance, ICT is being Changing energy business deployed through short message service (SMS) or voice interaction with clients to collect and analyze models with ICT mobile data, facilitate mobile payment systems, and Providing electricity to the approximately 1 billion utilize applications that are increasingly enabled by people without access to affordable and sustainable the internet. These innovative products and services electricity services presents an enormous challenge are facilitating a signifi cant increase in energy access and opportunity.1 In Sub-Saharan Africa, only 14 per- globally. In East Africa alone, they have reached 14 mil- cent of people have access to grid electricity; however, lion people in six years. The average annual market nearly 70 percent now have access to mobile phones.2 growth is 140 percent, according to market research 6 By the end of the decade, it is estimated that nearly by the World Bank Group’s Lighting Africa Initiative. 930 million people in Africa will own a mobile phone,3 In East Africa, falling costs of ICT and effi cient creating not only a huge demand for phone charging, appliances have enabled off-grid photovoltaic (PV) but facilitating a variety of energy and other services systems to provide more energy services at a lower through the use of mobile phones. price. Instead of relying on traditional means of col- A potential symbiotic relationship therefore exists lecting payments (with high transaction costs and between the development of information and commu- losses), electricity bills can now be paid by cellphone nication technologies and increasing energy access through mobile money services such as M-Pesa. This in Africa and other low-access areas. As an example, approach also generates large amounts of data on cell towers require a local power supply, but in many mobile money transfers, and has helped to establish credit histories for a previously underserved segment This sector focus was contributed by Todd Johnson, Anna Lerner, of the population. Energy supply companies combine and Karan Capoor. fi nancial information with geographic and census ENERGY 289 data to identify new markets and to differentiate 20 percent of peak demand.8 So far, such systems customers based on varying levels of service, thus have been introduced largely in advanced economies allowing them to tailor the service to the customer’s of the Organisation for Economic Co-operation and ability to pay. Development (OECD) and emerging economies. But ICTs have also been deployed to reduce risks for the development of low-cost sensors that will increas- suppliers and to facilitate after-sales service by tag- ingly be factory-installed in all electrical appliances ging energy systems (such as solar PV panels and bat- (televisions, refrigerators, fans, air conditioners) will teries) with sensors so that they can be tracked as they allow both the monitoring and control of electricity pass through distributors and are sold and installed. use by consumers and electricity suppliers.9 The sensors capture remote real-time data about the One of the biggest challenges for the power sector equipment, enabling the provider to monitor perfor- of the future is the need to “balance” electricity gen- mance and offer maintenance support. Companies eration and load on a particular grid. The growth of like M-Kopa and Off Grid Electric in East Africa variable is creating new challenges have call center agents that resolve most payment or for planners and power system operators, who must service-related queries, while remaining service rely on other sources of supply when the sun goes issues are addressed by a fl eet of technicians on down or the wind stops blowing. Predicting such motorcycles who are radio-dispatched. variations in energy production is challenging. In combination with (more predictable) demand vari- ICT, the smart grid, and ations, it can lead to deviations in power frequency and power system reliability issues, and increase the demand management need for grid balancing assets, such as fast-ramping gas power plants or electric storage.10 Research by Improved real-time information and automated con- the IEA shows that renewable energy costs can be trols are increasing the effi ciency of the electricity grid reduced signifi cantly through new forecasting and all over the world. The “smart grid”—which has come digital technologies that can help balance generation to defi ne a broad range of sensors, meters, and con- and load by monitoring and predicting the supply of trols enabled by information technology (IT), as well as variable energy resources.11 In countries and regions large-scale and real-time data collection—can enhance with growing shares of renewable energy such as Cal- the operational effi ciency of the electricity system by ifornia, Denmark, and Germany, electricity markets optimizing energy transactions.7 The integration of are now designed to match on a ICT into the grid can improve system security through minute-by-minute basis through sophisticated auc- more rapid analysis of service interruptions and pre- tions, regulatory mechanisms, and control systems diction of outages. Connected devices and software that rely on the internet.12 offer customers access to real-time data to help them manage their energy usage more effectively, while advanced meters and automatic controls provide Energy, open data, and the opportunities for energy effi ciency, such as automatic internet dimming of street lights, or turning off lights or air conditioning in unoccupied rooms and buildings. Having high-quality, easily accessible information Potentially one of the most revolutionary internet- on energy resources, demand, and usage is crucial to based innovations in the electricity sector is the ability supporting the formulation of government policies, to adjust consumer demand by providing new signals and can be a catalyst for commercial investment. and transparency. “Demand response” allows electric- Just as the existence of high-quality geological data ity consumption to be reduced during peak periods is an imperative for licensing and explo- based on agreed-upon reductions in power supply. ration, publicly available data on renewable energy For example, consumers agree to minor reductions resources is one of the fi rst steps a country can take in air-conditioning demand during summer peaks in to encourage investment. ICT facilitates this resource exchange for a rate cut. This helps increase system assessment and mapping through the use of super- reliability, reduce energy supply costs, and lower gen- computers to analyze years of historical meteorologi- eration investments due to the reduced need for oper- cal and satellite data, by facilitating the transmission ational reserves. The International Energy Agency of measurement data from fi eld-based instruments, (IEA) estimates the savings potential for demand and by supporting the wide dissemination of outputs response programs in competitive markets at 15 to through open data platforms. 290 WORLD DEVELOPMENT REPORT 2016

A recent trend in the United States is the provision synergies are maximized will be critical to meeting of anonymous energy use and performance data that the sustainable energy goals of the World Bank’s cli- can support energy effi ciency markets and services.13 ent countries. The industry-led Green Button Initiative in the United States is another voluntary industry data standard for Notes utilities and companies; it allows consumers to make their energy consumption data available to service 1. World Bank 2015. providers, which can help consumers lower their 2. Ericsson 2014. energy bills or “green” their energy supply through 3. Ericsson 2014. home energy effi ciency solutions or the design and 4. GSMA 2015. 5. Vogt 2015. fi nancing of renewable energy equipment. Since 6. http://www.lightingafrica.org. 2012, more than 50 utilities and electricity suppliers 7. MITei 2014. have signed on to the initiative, providing over 60 mil- 8. IEA 2014. 14 lion homes access to their detailed energy use data. 9. Spijker 2014. In India, a pilot project launched by Prayas Energy 10. Martinez-Romero and Hughes 2015. Power frequency Group, a nongovernment organization, has installed refers to the frequency of oscillations of alternating around 100 devices in volunteers’ homes to measure current in a power grid. the availability and quality of electricity on a continu- 11. IEA 2014. ous basis.15 The results provide a potentially powerful, 12. Martinot 2015. crowdsourced fl ow of information that can help con- 13. U.S. Department of Energy, “Open Energy Data.” sumers understand the causes and impacts of power 14. U.S. Department of Energy, “Green Button.” 15. Prayas Energy Group. shortages, and can help policy makers and utility 16. TREND Consortium Partners 2014. companies prioritize power system investments. 17. Koomey 2013. Given the rapid growth of the ICT industry, there has been an increased focus on how much energy the sector uses. Recent estimates of energy use by ICT— References split almost equally into energy use by communica- Ericsson. 2014. “Sub-Saharan Africa: Ericsson Mobility tion networks, computers, and data centers—show Report Appendix.” Ericsson, , Sweden. that the sector accounts for between 3 and 4 percent GSMA (Groupe Speciale Mobile Association). 2015. of total global electricity consumption, and that “From Fringe to Mainstream: The Role of Mobile in energy consumption increased around 7 percent per Energy Access.” http://www.gsma.com/mobilefor year between 2007 and 2012.16 While a breakdown of development/from-fringe-to-mainstream-the-role how much of this increased demand came from devel- -of-mobile-in-energy-access. oping countries is not readily available, this share is IEA (International Energy Agency). 2014. “Task 24: Clos- likely signifi cant, as more than 750 million new users ing the Loop–Behaviour Change in DSM: From The- ory to Policies and Practice.” IEA, Paris. have been added in China and India alone since 2000. Koomey, Jonathan. 2013. “Does Your iPhone Use as Much There are many ways to reduce the energy use from Electricity as a New Refrigerator? Not Even Close.” the ICT industry, including moving to faster systems. http://thinkprogress.org/climate/2013/08/25/2518361 For example, 2G networks (which are common in /iphone-electricity-refrigerator/. developing countries) require over 400 kilowatt-hours Martinez-Romero, S., and W. Hughes. 2015. “Bringing (kWh) of electricity per year to deliver 12 gigabytes Variable Energy Up to Scale.” World Bank, Washing- (GB) of data (1 GB per month), while a 3G network con- ton, DC. sumes less than one-tenth that amount (35 kWh), and Martinot, Eric. 2015. “How Is California Integrating and a 4G network uses one-sixtieth (7 kWh).17 In addition, Balancing Renewable Energy Today?” http://www the industry is exploring new technologies, includ- .martinot.info/renewables2050/how-is-california ing microtracing—a new, leaner technology for fi ber -integrating-and-balancing-renewable-energy-today. MITei (Massachusetts Institute of Technology: Energy access that uses less energy. However, many of these Initiative). 2014. “Utility of the Future Study.” MITei. gains are offset by ever rising volumes of data. https://mitei.mit.edu/research/utility-future-study. While it is too early to judge the overall impact of Prayas Energy Group. “Electricity Supply Monitoring digital technologies on the energy sector, especially Initiative: Watch Your Power Quality.” http://www in developing countries, it is clear that many of the .watchyourpower.org. trends described above will continue to affect energy Spijker, Arent van’t. 2014. The New Oil: Using Innovative Busi- consumers and suppliers. Ensuring that benefi ts and ness Models to Turn Data into Profi t. Technics Publications. ENERGY 291

TREND Consortium Partners. 2014. “Assessment of Vogt, Heidi. 2015. “Making Change: Mobile Pay in Power Consumption in ICT.” http://cordis.europa Africa.” Wall Street Journal, January 2. http://www .eu/docs/projects/cnect/0/257740/080/deliverables .wsj.com/articles/making-change-mobile-pay-in /001-trendd16fi nalwp1report.pdf. -africa-1420156199. U.S. Department of Energy. “Green Button.” http:// World Bank. 2015. “Progress toward Sustainable Energy: energy.gov/data/green-button. Global Tracking Framework 2015.” World Bank, ————. “Open Energy Data.” http://energy.gov/data/open Washington, DC. -energy-data. CHAPTER 6 Global cooperation

Data, spam, disease, poverty, commerce, rivers, shipping physical objects. Can trade and IPR rules and polluted air—all cross borders and touch many and procedures be modernized? people. And all can be better managed if people and Third is providing global public goods—including nations cooperate. The internet is both a subject of poverty reduction and environmental sustainability. cooperation and a new tool to facilitate cooperation These are complex and diffi cult challenges requiring in other realms. This chapter explores its potential for both cooperation and massive amounts of informa- boosting three forms of cooperation. tion. Can the advent of the internet make these quests First is governing the internet. The internet requires more inclusive, more informed, and more effi cient? technical coordination and harmonized standards to ensure its smooth operation worldwide. Wher- ever you are, when you click or tap on http://www Internet governance .worldbank.org/wdr2016, a message goes out from We reject: kings, presidents, and voting. We believe in: your device, initiating processes to discover which of rough consensus and running code. the millions of servers on earth corresponds to that —David Clark, computer scientist and internet pioneer1 address. Information is handed off from network to network and perhaps from satellite to cable to Wi-Fi, ending up on your screen. That nanomiracle depends From its beginnings, the internet has always been on the consensual and universal use of protocols for different from the networks that preceded it—tele- data transmission and on a legal system that deter- phone, radio, television, and cable. It was founded mines who has the right to use the “worldbank.org” as a research network, explicitly noncommercial for address. Less consensus exists about how to deal the fi rst few decades of its existence, and designed as with cross-border issues such as suppressing viruses a radically decentralized network. As a result, it has and spam, or defi ning and enforcing rights to pri- been governed more like a federation of networks, vacy, while ensuring cybersecurity and combating in some respects like a Republic of Users.2 But the cybercrime (see chapter 4). As technology and its uses pragmatic libertarian of the internet’s evolve rapidly together, who decides how the internet founders is under assault by the commercial and is structured and run? political interests of its other stakeholders, mainly Second is facilitating cross-border exchanges of goods large corporations and nation-states. The growing and services. During the last two decades or more, commercialization of user data by private businesses countries have worked out trade agreements to and mass surveillance by states, including many sov- collect duties and taxes and to enforce intellectual ereigns accustomed to greater state control over their property rights (IPRs): patents, copyrights, and trade- citizens, have gradually eroded the trust the internet marks. Many of these agreements are poorly adapted once enjoyed. This has given rise to widely disparate for the world of the internet because intellectual views on how to govern the internet. property can be duplicated and disseminated with The internet’s growing popularity has increased little or no cost, and purely digital goods and services the need to manage its worldwide operation. can now cross borders without customs inspections. Whereas less than 2 percent of phone calls cross The increasingly frictionless ease of transacting busi- borders, between 60 and 75 percent of internet traffi c ness can be stymied by 19th-century procedures for is international (depending on the country). Despite GLOBAL COOPERATION 293 its virtual qualities, it has an elaborate physical infra- two phases in 2003 (Geneva) and 2005 (Tunis). The structure that spans the globe. As a global resource, it summit confi rmed that “management of the internet requires some degree of international coordination to encompasses both technical and public policy issues function. Information fl owing through the internet and should involve all stakeholders.”7 As described has transboundary repercussions, raising other issues in its agenda, the groups of stakeholders with an for international debate and discussion. To address interest in internet governance include governments, these issues, the United Nations–authorized Work- private fi rms, civil society, intergovernmental and ing Group on Internet Governance defi ned internet international organizations, academia, and the tech- governance as “the development and application by nical community (box 6.1). governments, the private sector, and civil society, in The MSM label refl ects the diverse actors that their respective roles, of shared principles, norms, form the loose de facto framework for internet gover- rules, decision-making procedures, and programs nance. For example, one of these stakeholders is the that shape the evolution and use of the Internet.”3 Internet Engineering Task Force, a loose affi liation of engineers that is in charge of the protocols that run the How is the internet governed? internet. Another is ICANN, the Internet Corporation The main attributes of the global internet are that it is for Assigned Names and Numbers, a California not- distributed, decentralized, fl exible, multilayered, and for-profi t corporation that performs certain technical end-to-end.4 Not a monolith, it is instead a mosaic of functions on the internet—such as running root serv- separate but interrelated infrastructure, applications, ers (the computers that control traffi c on the internet) actors, and decisions. Consequently, its governance and administering the system of domain names (.com, is likewise not unitary but consists of an ecology of .org, .gov). But not all stakeholders feel equally repre- decisions and decision makers. Internet governance sented within the current governance framework. features are different from those in telecommunica- tions, characterized by a multilateral system. The internet unites people; its governance The governance framework for the internet is divides nations considerably more diverse than that of the telecom Politicians and regulators are increasingly concerned sector. The policy, legal, and regulatory framework for about their inputs into the future governance of the information and communication technologies (ICTs) internet.8 Their concerns may be justifi ed (fi gure 6.1): has shifted since the start of the millennium. Then it was primarily about the technical and economic • The rapid and unprecedented growth of the inter- regulation of more liberalized and competitive tele- net has meant that it has evolved from an American communications infrastructure and services envi- technology to a critical global infrastructure in a ronment—along an arc with the state at the center very short period. This has given rise to an unstated of the policy and regulatory piece and state-licensed power struggle between its traditional stakeholders operators orbiting around it. Now it is a much more and many new ones, especially some developing complex, multifarious, and loose amalgam of policies, country governments, which feel left out of the laws, and actors having their origins in constitutional discussion. rights, criminal justice, and technical and economic • The bulk of internet nonusers are in developing regulation. This new framework also involves non- countries—4.2 billion people, more than half the state actors, national, regional, and even multilateral world, have no internet access, and 5.1 billion are agents—each with an interest in the wide-ranging not on social media. Their interests, some argue, and diverse aspects of what has come to be known as are best promoted by their national governments. the internet governance framework. • Mass collection of user data by private businesses The prevailing model of internet governance con- and unwarranted surveillance by governments tinues to evolve to include its key stakeholders. The have eroded trust and fed growing skepticism multistakeholder model (MSM) refl ects the notion about the current governance arrangement. that better policies and outcomes can be achieved if • Increasing use of social media and unhindered those affected are part of a policy consultation pro- access to all content, which is generally thought to cess;5 it also refers to the range of actors running, be a positive development, can clash with local cul- administering, and governing the internet.6 The tural norms and social practices, prompting some term came into popular use as a result of the UN’s governments to advocate a more activist approach World Summit on the Information Society, held in to govern the internet. 294 WORLD DEVELOPMENT REPORT 2016

Box 6.1 Categories of stakeholders in internet governance

Individuals, organizations, and institutions with an interest Figure B6.1.1 maps these stakeholders according to their or stake in particular internet governance issues include: degree of localization or internationalization. Figure B6.1.1 Stakeholders in internet • States, which have policy authority for sovereign nations governance and are responsible for internet-related public policy issues at the global level • Private businesses, with expertise in technical and eco- International IETF ICANN ITU nomic fi elds W3C WIPO Regional COE • Civil society, especially at the community level Registrars CERTs Ministries • Intergovernmental organizations, particularly in facilitat- National Civil society Regulators ing the coordination of internet-related policy issues ISPs Operators Parliaments Company/local • International organizations, with important roles in developing internet-related technical standards and Individual Users relevant policies No government All government • Technical communities, such as the members of technical Degree of government involvement standard-setting bodies and other experts in computer science and engineering Source: Center for Democracy and Technology, Internet Governance • Academics, with a focus on and involvement in internet Forum, Athens. governance. Note: CERTs = computer emergency response teams; COE = Council of Europe; ICANN = Internet Corporation for Assigned Names and Numbers; IETF = Internet Engineering Task Force; ISPs = internet service providers; ITU = International Telecommunication Union; W3C = World Wide Web Consortium; WIPO = World Intellectual Property Organization.

Sources: World Summit on the Information Society 2005, Tunis Agenda for the Information Society, Document WSIS-05/TUNIS/DOC/6(Rev. 1)-E, 18 November: Paragraphs 31, 33, 35; Kummer 2013; Gasser, Budish, and West 2015.

Figure 6.1 Concerns that have fueled the debate on how the internet is governed

3. Privacy and surveillance concerns

2. Digital divide, with 4. Clash with most nonusers in local cultures developing countries and social practices

5. Nonalignment 1. Power struggle: Internet with national traditional vs. new governance policies and stakeholders regulations

Source: Dutton 2015, for the WDR 2016. GLOBAL COOPERATION 295

Table 6.1 Multistakeholderism or multilateralism

Dimension Multistakeholder Multilateral/intergovernmental Leading principle Collaborative leadership among stake- Sovereign right of governments to holders, with a commitment to resolving determine internet policy and regulation particular problems Representation of Direct engagement of private business National government agency represents stakeholders and industry, governments, bilateral and interests of all in bilateral and multilateral multilateral international institutions, civil treaties and agreements, anchored in advice society and academia, NGOs and consultation with all stakeholders Role of governments Governments are a key stakeholder, with National governments represent other legitimacy to make decisions interests in an intergovernmental entity Process • Bottom-up participatory • Top-down consultative • Horizontal across stakeholders • Hierarchical within states and through • Generally open and transparent international agreements and treaties • Intergovernmental negotiations, with open consultation Examples of ICANN, Internet Society, World Summit on ITU, UN, WIPO, WTO relevant bodies the Information Society, Internet Governance and processes Forum Source: Adapted from Bauer and Dutton 2015, for the WDR 2016. Note: ICANN = Internet Corporation for Assigned Names and Numbers; ITU = International Telecommunication Union; NGO = nongovernmental organization; UN = United Nations; WIPO = World Intellectual Property Organization; WTO = World Trade Organization.

• More countries are asserting national policy and More recently, a new approach to revive the multi- regulation over areas (mainly around content) that lateral model arose in an attempt by the membership they feel the existing governance structure is not of the International Telecommunication Union addressing.9 (ITU) to update the International Telecommunica- tion Regulations (ITRs) at the World Conference on As a result, support for the MSM is not unani- International Telecommunications (WCIT) in Dubai mous. Tension between the MSM and a multilateral in 2012. Measures aimed at regulating privacy and model (MLM) that puts governance responsibility free speech were introduced in a special codicil in predominantly in the hands of nation-states has the regulations. This resulted in a clear split between arisen many times since the World Summit embraced 89 countries that approved the new draft and 80 that MSM. Some governments worry about surrendering maintained reservations (map 6.1).12 Although it was sovereignty or asserting national jurisdiction over expected that these issues might have been taken internet activities within their borders. In September up at the ITU Plenipotentiary in Busan, Republic of 2011, China, the Russian Federation, Tajikistan, and Korea, in 2014, they were not and remain unresolved. Uzbekistan jointly submitted a draft resolution to A key point of contention for countries favoring a the UN General Assembly on an International Code multilateral approach is the U.S. government’s over- of Conduct for Information Security outlining their sight of IANA (Internet Assigned Numbers Author- vision of governing the internet.10 ity).13 The MLM would like to see an IANA transition The promoters of MSM argue that state control to an intergovernmental body, like the ITU. The of the internet would not leave space for the range MSM, especially the United States, insists on further of players currently involved in internet governance autonomy of ICANN and IANA. In March 2014, the and could pave the way to suppression of privacy U.S. Department of Commerce announced a process and rights of free expression—say, in the name of to relinquish its oversight over certain technical national security (table 6.1). Without the checks and functions performed by IANA to the multistakeholder balances available through the MSM, abuses might community—in shorthand, the IANA transition, orig- increase. The UN has asserted that rights already inally scheduled to be completed by September 2015.14 protected in the real world—the right to privacy A key condition of the IANA transition is that over- and the right to communicate (to receive and impart sight of the technical functions would not be ceded information)—are to be protected in the virtual to a government or group of governments, but to the world.11 multistakeholder community. 296 WORLD DEVELOPMENT REPORT 2016

Map 6.1 A divided world: Country positions on the International Telecommunication Regulations

Countries that approved the draft Countries that maintained reservations against the draft Not applicable

IBRD 41790 Source: International Telecommunication Union. Data at http://bit.do/WDR2016-Map6_1.

Protecting the interests of users is evolution.15 The MSM model is evolving, and that paramount is one of its great strengths. However, care should This inability of internet stakeholders to reach a be taken to ensure respect for principles of good consensus on future governance mechanisms could governance (including transparency, accountability, be costly. The lingering uncertainty is likely to affect inclusiveness, and representativeness); safeguard the private sector’s long-term investment planning. meaningful dialogue among stakeholders; empower Delay in agreeing on new protocols and standards stakeholders to participate effectively across groups; could slow the pace of innovation. Some have even and prevent capture by special interests.16 According to suggested that the internet is at risk of being split, so one recent survey in 24 countries, the MSM is also the that the world ends up with several local internets— preferred governance model of end users (fi gure 6.2). Google CEO Eric Schmidt calls it the splitnet—destroy- It is therefore likely that a multistakeholder format— ing the essential global character of the internet. one that addresses the concerns over the digital divide The new governance structure of the internet and trust and use of the internet; provides meaningful needs to accurately refl ect the plurality and the mul- forums for “left out” stakeholders; and balances uni- tifarious nature of internet actors and be fl exible versal community acceptance with appropriate local enough to accommodate its rapid and continuous fl exibility—could be the accepted way forward. GLOBAL COOPERATION 297

Figure 6.2 The multistakeholder model their peers in the rest of the world. Digital transac- of internet governance enjoys greater tions will continue to increase as the number of goods support than other options and services offered online increases. For example, with the introduction of online music stores in many Percentage of respondents who said they would trust the governance options below completely or somewhat developed countries in 2004, digital music started to gain a share of global music sales—from 2 percent in 60 2004 to 46 percent in 2014.17 Access to a global market 50 can be particularly advantageous to fi rms in small, island, and landlocked countries, as well as for coun- 40 tries with small populations, where the size of the 30 local market is often constraining growth. Percent 20 Digital trade is potentially global in scope, but bar- riers to digital integration prevent fi rms from reach- 10 ing appropriate scale (box 6.2). Issues such as trade 0 logistics and infrastructure, online payment systems, and trade barriers can be handled by national govern- body ments (chapter 5). But cross-border issues affecting United States Engineers and United Nations My government digital trade need international coordination. Indeed, A multistakeholder the internet’s ability to seamlessly deliver digital technical community Technology companies goods and services around the world is considerably Source: CIGI and Ipsos 2014. Data at http://bit.do/WDR2016-Fig6_2. impeded by various regulations that could be stream- lined through greater international cooperation. The two main cross-border issues are barriers to data fl ows Toward a global digital and uncoordinated intellectual property rights regime. market Removing barriers to cross-border One of the advantages of the internet is its ability data fl ow to deliver digital goods to a global market, uncon- The internet has revolutionized the way data are col- strained by national infrastructure—enabling devel- lected and shared. This in turn has increased economic oping country citizens to enjoy the same products as effi ciency and productivity, improving welfare and

Box 6.2 European Union: A fragmented market for digital products

Despite being a single market with a free fl ow of goods, to a geographic location. Consumers are directed to their services, and people for many decades, the European Union local websites, where they face diff erent prices from online (EU) still functions like a fragmented market for digital retailers based on their location. Content like video is also goods and services. Consumers and fi rms face diffi culties restricted within a single geographic location. in engaging in the digital economy. Firms within the EU also face many diffi culties in selling Consumers in the EU prefer to shop from online stores their goods and services online in other EU markets. Firms that are within their national borders. While 44 percent engaging in e-commerce face high and uneven cross- of consumers made an online purchase from a domestic border delivery charges, which are passed on to consumers. business in 2014, only 15 percent did so from a business in For example, Copenhagen, Denmark, and Malmo, Sweden, another EU country.a The consumers may have concerns are separated only by an 8-kilometer bridge, but a package about payment security, product quality, or the reliability sent from Copenhagen to Malmo costs €27 whereas a - of the sellers located in another country (fi gure B6.2.1). age sent from Malmo to Copenhagen costs €42.b Moreover, Moreover, consumers in the EU face diff erent prices for the fi rms face large costs to adapt to various national laws, and same online goods and services because of the practice believe that the costs outweigh the benefi ts of setting up a of geoblocking, in which services and prices are limited website (fi gure B6.2.2).

(Box continues next page) 298 WORLD DEVELOPMENT REPORT 2016

Box 6.2 European Union: A fragmented market for digital products (continued) Figure B6.2.1 Perceived barriers to buying over the internet in 2009 percent of individuals responding

30

25

20

15 Percent

10

5

0 ic m R ia ta d ly a ia g d ia ia o Y t a ds um n r rus v ain n rbia rkey al lan It a u lan onia u garia oa reeceM lan gi ustri n t tugal Sp ve nmark Se T r Poland ce r el rmanyA m i Lat France o Ireland Norway ul C G I B e Swedenthuania mbo Cyp F Es or l De B G Li Ro e P S Hungary Nethe Lux Czech Republ Slovak Republic United Kingd Macedonia, F Concerns about payment security Concerns about receiving or returning goods, complaints and redress

Source: Eurostat. Data at http://bit.do/WDR2016-FigB6_2_1.

Figure B6.2.2 Obstacles for enterprises not selling online in 2013 percent of enterprises responding

35

30

25

20

Percent 15

10

5

0

y g ia ic d R ia ta m s d ia d al a ia n n Y um l o en ru n n ri n Italy and nce ur any F i at a pain an a ra ustria ve m o lands M S el inl F A Latvia bo o Pola gd wed cela r lg sto F Norway er a, Belg thuaniaCr er n S Cyp I I Greece u E Hungarem Sl Republ Li th Roma Portug B x h G oni e Lu c d N ze ce C Ma United Ki Legal framework Cost greater than benefits

Source: Eurostat. Data at http://bit.do/WDR2016-FigB6_2_2.

a. European Commission (EC) (2015). b. Figures given by Andrus Ansip, the European Commission vice president for the digital single market. See http://www.politico.eu /article /cross-border-delivery-costs-pose-problems-for-small-businesses/. GLOBAL COOPERATION 299 raising standards of living. Barriers to data fl ows pre- tions, maintain an effi cient supply chain, and manage vent countries from benefi ting fully from the inter- human resources, production, and sales. The data can net, increasing costs and ineffi ciencies for fi rms and range from personal information about employees individuals. Firms require free fl ows of data across and customers to production and technical data. The national borders to operate internationally. This is emerging use of cloud computing and the “internet increasingly important with production being frag- of things” (see spotlight 6, “Six digital technologies to mented into global value chains and the digitization of watch”) will increase the need for freer data fl ows. In goods and services. Firms move data internationally a survey of U.S. fi rms in 2012, a large share of fi rms to control and coordinate their international opera- felt that data localization requirements and privacy

Figure 6.3 Perception of U.S. fi rms on barriers to data fl ows as obstacles to trade, 2012 percent of fi rms responding a. Requirements to locate data servers in-country

90

80

70

60

50

40 Percent 30

20

10

0 Large SME Large SME Large SME Large SME Large SME Large SME Large SME

Content Digital Finance Manufacturing Retail Other services Wholesale communications

b. Barriers to sharing information across borders 90

80

70

60

50

40 Percent 30

20

10

0 Large SME Large SME Large SME Large SME Large SME Large SME Large SME

Content Digital Finance Manufacturing Retail Other services Wholesale communications

Minor obstacle Substantial obstacle

Source: USITC 2014. Data at http://bit.do/WDR2016-Fig6_3. Note: Large fi rms are classifi ed as fi rms with more than 500 employees. SMEs (small and medium enterprises) are fi rms with more than 20 employees and less than 500 employees. 300 WORLD DEVELOPMENT REPORT 2016

requirements have emerged as substantial obstacles to on data fl owing back and forth as users retrieve and trade (fi gure 6.3). Many of these fi rms operate globally update information directly on the servers. Barriers and have business models that depend on free access to data fl ows will force fi rms to relocate tasks and to data about their users. operations, change their information technology (IT) In contrast to industry preferences for open data architecture, engage a different supplier, or discon- fl ows, many countries are beginning to insist that data tinue services to customers. These barriers disrupt should reside within their national borders—often two of the most important business trends facilitated referred to as “data nationalism”—and are imposing by the internet: the fragmentation of production into barriers on cross-border fl ows of data. So-called data global value chains, and the creation of offshore ser- localization requirements require fi rms to locate data vice hubs like the business-processing operations in servers and data centers within the country to store India or the Philippines. and process the information. Other countries have The estimated effects of barriers to cross-border strict regulations about data leaving their borders to data fl ows are signifi cant. An economic simulation protect privacy, refl ecting the preference of their citi- conducted by the European Centre for International zens, as users become more cautious and risk averse Political Economy (ECIPE) considers the proposed to how their data are being used for commercial gains or enacted legislation on data fl ows in six develop- (fi gure 6.4). Barriers to data fl ows can affect trade more ing countries and the EU and calculates the costs of than an increase in tariffs: they can prevent fi rms from regulations on businesses.19 The countries examined entering or operating in a sector, while tariff rates are proposing a mix of regulations in their legislative merely impose an additional cost to operations that packages that will impose stricter privacy regula- can be absorbed by the fi rm or passed on to customers. tions, such as the right to be forgotten and the con- Cross-border data fl ows, which are essential to sent to transfer data to a third party (table 6.2). The international trade, are a substantial portion of data study fi nds that these laws can reduce gross domestic traffi c, comprising about 16–25 percent of U.S. data product (GDP) by up to 1.7 percent, investments up to traffi c and 13–16 percent of European data traffi c.18 4.2 percent, and exports by 1.7 percent (fi gure 6.5).20 Cross-border data fl ows are likely to increase with The argument to remove all barriers to data fl ows, the increasing use of cloud computing, which relies however, is not as clear-cut as the arguments to remove barriers to cross-border trade and investment. Figure 6.4 A majority of respondents Government regulations to place data servers in the agree that their online data and country or to prevent information from being shared personal information should be across borders can stem from legitimate concerns physically stored on a secure server in about privacy and security for their citizens’ infor- their own country mation. Countries may believe that imposing data localization and privacy requirements will allow them to better protect the data in their country. But it is not Total agreeing always clear whether the data are more secure in a BRIC local data server or in a dedicated secure server over- Latin America seas. Moreover, data localization requirements might threaten the progress of new technologies that depend North America on unhindered cross-border data fl ows such as the APEC internet of things, “big data,” and cloud computing.

Middle East and Africa While countries may have legitimate reasons to impose barriers to data fl ows, these barriers should Europe not be a disguised method to impede trade and eco- G-8 countries nomic activities. By imposing barriers on data fl ows, countries may mistakenly believe that they can 020406080 100 encourage domestic data-driven sectors, like cloud Percent Strongly agree service providers. Such policies are akin to import Somewhat agree substitution strategies, which have had a mixed record. Moreover, the barriers should not treat fi rms Source: CIGI and Ipsos 2014. Data at http://bit.do/WDR2016-Fig6_4. differently according to their ownership, as that Note: APEC = Asia-Pacifi c Economic Cooperation; BRIC = Brazil, , India, China; G-8 = Canada, France, Germany, Italy, Japan, the Russian Federation, would reduce competition and slow down digital the United Kingdom, and the United States. adoption in the rest of the economy (chapter 1). GLOBAL COOPERATION 301

Table 6.2 Many countries have proposed comprehensive legislation on data fl ows

Korea, Proposed requirement Brazil China EU-28 India Indonesia Rep. Vietnam Data localization requirement No Yes No Partial Yes Partial Yes Consent required for data collection Yes Yes Yes Yes Yes Yes No Consent required for transfer to third parties Yes Yes No Yes No Yes No Right to review No No Yes Yes Yes Yes No Right to be forgotten Yes Yes Yes No No Yes Yes Breach notifi cation No Yes Yes No Yes Yes No Impact assessment No Yes Yes No No No No Data privacy offi cers No No Yes No No Yes No Sanctions for noncompliance Yes Yes Yes Yes Yes Yes No Government access required Yes No No Yes No No Yes Data retention requirement Yes No No Yes No No Yes Source: Bauer and others 2014. Note: EU-28 = current member countries in the European Union.

While these barriers remain, developing coun- Figure 6.5 Changes in GDP, investment, and exports tries participating in global markets should ensure due to regulatory restrictions on data fl ows that they are not excluded because of weak or absent Korea, regulations. At a minimum, countries should enact Brazil China EU-28 India Indonesia Rep. Vietnam data protection regulations following internationally 0 recognized principles so that multinational compa- –0.5 nies do not avoid the country because of uncertainty about compliance and trust in the handling of data.21 –1.0 For example, the EU allows personal data to leave its –1.5 borders if countries have “adequate” protection for –2.0 personal data. Australia allows data to leave its bor- ders only to those jurisdictions with substantially –2.5 similar levels of privacy protection. However, these –3.0 Percent change regulations continue to evolve, and countries must constantly adapt to such changes to reduce the uncer- –3.5 tainty and compliance costs on companies.22 –4.0 Recognizing the importance of cross-border data –4.5 fl ows, many countries have agreed on basic princi- GDP Investment Exports ples to coordinate their data regulations, as with the recently revised Privacy Guidelines and Declaration Source: Bauer and others 2014. Data at http://bit.do/WDR2016-Fig6_5. Note: The fi gure shows percentage changes according to using a GTAP model. EU-28 = on Cross-Border Data Flows by the Organisation of current member countries in the European Union; GTAP = Global Trade Analysis Project. Economic Co-operation and Development (OECD), or the Privacy Framework of the Asia-Pacifi c Economic Cooperation (APEC). Other efforts focus on ensuring Setting global standards on data exchange that companies that move data across borders have Standard setting can be another tool to reap digital rigorous internal policies to maintain the privacy and dividends. Data exchange standards, for instance, can protection of the data. In the absence of international greatly facilitate cross-border services and enable a coordination on privacy rules, other solutions are seamless fl ow of information (see spotlight 5, “The being sought that rely on companies’ internal policies data revolution”). Various ongoing initiatives look at to protect privacy, such as the EU’s Binding Corporate setting data exchange standards in certain sectors at Rules and APEC’s Cross-Border Privacy Rules. While the international level. The standard should also be these programs help facilitate cross-border data fl ows, open, and use publicly available platforms and soft- the rules may be too cumbersome for most fi rms and ware, so that it does not exclude fi rms. One example implicitly benefi t larger fi rms.23 is the GS1 system of standards used in transport and 302 WORLD DEVELOPMENT REPORT 2016

logistics. The GS1 assigns a global unique number markets. Firms must seek recognition and enforce- so that any fi rm can identify and share product data ment in each jurisdiction. Similarly, if a fi rm wants to with its trading partners. The number allows anyone use an IPR it does not own to produce a digital good to trace the origins of the product up the supply chain or service, it must apply for licenses in each country and track its movement downstream. in which it wants to sell. While there are international Data exchange standards are being promoted in treaties governing recognition and protection across the context of Open Data by international organiza- member countries, each jurisdiction has different tions. Examples include the Comprehensive Knowl- rules and requirements, depending on the type of IPR. edge Archive Network’s (CKAN) standards—already Navigating these IPR regimes is confusing and creates adopted by more than 40 governments and organiza- large administrative costs for fi rms. These costs can tions, such as the governments of the United Kingdom even prevent fi rms from entering new markets. and Uruguay and the EU—and the Open Data portals Firms operating in the global market face huge for Africa and Bermuda. Progress has also been made costs to apply and use IPRs across borders. For exam- in developing data exchange standards for geospatial ple, in order to protect an IPR, a separate patent appli- information (such as the EU INSPIRE directive) in the cation is needed in each country where protection is electricity sector, particularly for smart grids, and in sought. A patent application that is valid in the EU the health sector. costs at least fi ve times more than an application in Countries should consider joining ongoing the United States.26 The scattered nature of the IPR international efforts for standard setting for data regime can add complications for fi rms when they exchange, and expanding them to other areas, at least need to identify the rights holders and construct a for critical sectors. Data exchange standards, however, new business model. In particular, the fi rm offering are not a silver bullet; they need to be accompanied by a digital good or service over the internet needs to appropriate policies and actions to have some effect apply for a copyright license, patent, or trademark on the digital market. For e-freight, the International in each market it serves, and then must be ready to Air Transport Association (IATA) has been working on enforce it under each national regime. The complex- setting data exchange standards and has identifi ed ity and costs of doing so can discourage new market actions that need to be in place for these initiatives to entrants. Only larger fi rms will be able to bear those succeed.24 One is the requirement that public institu- costs and reap the benefi ts from the internet. tions (such as customs for e-freight) accept electronic Countries should make it easier for fi rms to documentation as an alternative to paper-based docu- protect their intellectual property rights, thus stim- mentation. Another is that all relevant actors need to ulating more innovation and creative sectors. The be involved and act in a coordinated fashion (such as regulation process to apply for intellectual property customs, security, and transportation). And third is an licenses should also be harmonized and streamlined. end-to-end supply chain vision and e-document, as The Patent Cooperation Treaty allows fi rms to fi le an well as common business processes that are aligned international patent application, but this just creates with international standard-setting bodies such as a standardized application for the fi rm that must still the World Customs Organization, the United Nations be submitted to national authorities. Going one step Centre for Trade Facilitation and Electronic Business, further would be preferable. For instance, under the and the International Civil Aviation Organization. World Intellectual Property Organization’s Madrid Protocol, fi rms need only register their trademark in Better coordination of intellectual any signatory country and it will be protected across property rights regime member countries. Similarly, it should be easier for The internet allows people to purchase products and fi rms to obtain licenses to certain IPRs that will allow access vast amounts of content and information, them to operate globally. The availability of an online whether protected by intellectual property rights (IPRs) licensing system for intellectual property rights can or not, leading to allegations of IPR infringements. IPR reduce the administrative burden to fi rms. infringement presents a risk to consumers who might National IPR regimes should extend beyond the purchase counterfeit goods, and a disincentive for laws defi ning and protecting the intellectual property fi rms and artists to create new products and content.25 (IP). Countries need a broader public sector IP strategy, National IPR regimes allow fi rms to protect and use which considers how policies and regulations in tax, their intellectual property. But having different IPR education, science, technology, and industrial policies regimes and multiple IPR registrations, in different can not only make it easier for fi rms to protect their countries creates problems for fi rms selling in global IPRs but also help to stimulate innovation and creative GLOBAL COOPERATION 303 sectors and benefi t the public interest. The availability succeed over the resistance of those who benefi t of an online licensing system for IPRs can reduce the when information is tightly held? administrative burden to fi rms. Such an online licens- • Mustering global information for global goods. Global ing system requires international coordination and an weather and water data are costly to gather but agreement among many countries. Countries can also now almost costless to distribute. They take on the allow the concept of “fair use” in intellectual property character of a global public good. Information on regulations that permits an individual or business to weather affecting mountain headwaters helps peo- use the IPR for purposes such as research and educa- ple prepare for fl oods or harvests half a continent tion, journalism, and library activities. away. New technologies make it easier to assemble and analyze this information. But who will pay for it? And can countries agree to share it with their Leveraging information for downstream and downwind neighbors? sustainable development Table 6.3 and fi gure 6.6 map these information Eliminating global poverty and heading off planetary needs onto the landscape of development spending. environmental threats are two of the world’s great- Some projects are characterized by geographic dis- est challenges. The costs of action are in the trillions persion, sensitivity to local environmental and social of dollars, making global cooperation essential. But context, and an inherent need for adaptive implemen- development assistance represents only 0.6 percent of tation. These activities could benefi t greatly from rich developing country income. How can donors, devel- and rapid feedback. Another class of funding involves opment agencies, nongovernmental organizations investments in big infrastructure systems or sup- (NGOs), and international organizations hope to make port for government policies. Here, fi ne-grained and a difference? They can use digital technologies to timely information—on everything from poverty to deploy funds and knowledge more effi ciently and to water consumption to service delivery—could sup- include more people in the process. They can do so in port better decisions on system design and operation. three ways: This section explores the potential scope of these ben- efi ts and the largely institutional challenges in attain- • Getting wired for feedback. Development projects are ing them. The discussion is unavoidably speculative. often very sensitive to context and have no standard blueprints. They could be more relevant if their Getting wired for feedback—more design and implementation drew on local knowl- inclusive, better, faster data to guide edge. Projects could be more successful if wired for development projects rapid feedback, allowing continuous improvement Feedback on what is working and what is not enables by implementers and more effective supervision by more nimble and effective implementation. In the funders. National-level policies could be fi ne-tuned past, feedback has been thin, slow, and unreliable. with early information on whether poor people Now the technical barriers to better feedback have were benefi ting. Digital technologies can make been overcome. But often organizational inertia pre- these feedback processes much cheaper and easier. vents the benefi ts of innovative technologies from But can stodgy organizations and their partners being achieved. Revamped structures and incentives become nimble users of feedback? point to possible ways forward. • Taking information to scale: information as an input. Decision makers from villages to capitals lack the Lack of feedback hampers development information they need to optimize their actions. eff orts Farmers could better manage their fi elds with Development agencies committed US$6.8 trillion in timely information about weather, pests, and mar- project assistance from 1947 to 2012. One would not kets. Ministries could better fi ght poverty with expect these projects to be universally successful. up-to-date information about where poor people Development is inherently complex, and develop- live and their constraints. Land markets and envi- ment agencies venture into areas that others would ronmental protection could be more effi cient with not; a high success rate would indicate unambitious accessible information about the condition and objectives. But there is room for improvement in the ownership of land. New technologies make it pos- performance of international foundations, donors, and sible to widely and cheaply distribute all this infor- lenders, whether bilateral or multilateral. About one- mation, supporting better outcomes. But can this third of the recent projects of international fi nancial 304 WORLD DEVELOPMENT REPORT 2016

Table 6.3 Information as feedback and as input, by expenditure type

Expenditure type Example Information as feedback Information as input Dispersed and intensive Rural primary health, Rapid feedback on service provision, n.a. services community-driven uptake, and impacts development Dispersed land and water Watershed management n.a. Rapid feedback on crop, water, soil, management infrastructure, and environmental conditions; information on weather, hydrological and market conditions Small-scale and dispersed Rural roads and sanitation Feedback on infrastructure n.a. infrastructure construction, use, condition, and maintenance Large-scale n.a. Optimizing design and operation of infrastructure power and water systems

Policy design Educational policy design n.a. Fine-grained, timely information on economic and social conditions for policy design Institutional Legal and judicial Feedback on institutional capacity n.a. development development and performance Budget support Development policy loans Timely and fi ne-grained feedback on Strategic planning for the Sustainable impacts of policy implementation on Development Goals income distribution, service delivery, environmental conditions Source: WDR 2016 team. Note: n.a. = not applicable.

institutions are deemed unsuccessful by the institu- outcome and the quality of its M&E is stunning. Fig- tions themselves (fi gure 6.7). What goes wrong? ure 6.8 (based on data from the World Bank, the only Critics paint development practice as infl exible, offi cial donor agency that publishes this informa- ineffective, and out of touch.27 Too much emphasis tion in this detail) shows that the few projects with on spending rather than on results, they say. Project high-quality M&E are almost all rated “satisfactory” designs are cooked up by development experts and or “highly satisfactory” in outcome; the larger number government offi cials, in this view, and often do not with “negligible” M&E are mostly rated “moderately refl ect local preferences, insights, or political realities. unsatisfactory” or worse. Evaluations suggest a strong They can be further hampered by lack of rigorous, causal link: a good M&E system supports clear for- systematic learning about the success or failure of mulation of objectives, and M&E detects and solves prior projects. Without reliable information on proj- problems in implementation.29 Despite this, an OECD ect effectiveness, fl ashy bad ideas go viral while subtle survey30 of 28 donors fi nds pervasive shortcomings in good ones languish. With such poor design, projects their M&E systems. A majority of the donors report can go off the rails. And if they do, there is often no incomplete results frameworks, diffi culty in tracking fast channel for feedback and course correction. outcomes, and a view by partners that a results frame- A combination of rigid design and poor feed- work is a formal requirement rather than a useful back might be particularly lethal for “complex” and tool. Feedback loops are reportedly lacking. NGOs context-sensitive projects, where workable solutions echo these views. Many donor staff and recipients see require experimentation.28 Consider health services, M&E as a burdensome “extractive industry,” generat- where success depends on changing the culturally ing information that satisfi es some distant funder’s rooted behavior of clients and staff—or watershed accountability requirements but does not serve the management, where social and environmental condi- needs of project managers or benefi ciaries. tions vary from one valley to the next. Worse: learning is slow. The multilateral develop- Yet there is a paradox: monitoring and evalua- ment banks have independent evaluation units for tion (M&E) improves performance, but agencies both accountability and for learning—an important under invest in it. The correlation between a project’s source of objective feedback. But these units do not GLOBAL COOPERATION 305

Figure 6.6 The evolution of development aid by information needs, 1973–2010

300

250

200

150

100

Commitment amount used (constant US$billion) amount used (constant Commitment 50

0 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Expenditure type Not determined Institutional and organizational development Trade finance Small-scale and dispersed infrastructure and industry Budget support Dispersed land and water management Large-scale infrastructure and industry Dispersed and intensive services Policy design

Source: WDR tabulation from AidData (http://AidData.org). Data at http://bit.do/WDR2016-Fig6_6. have resources to review every project in depth. They in average incomes, together with environmental also focus on after-the-fact evaluations to avoid con- improvements, and helped shape the revised national fl icts of interest. This slows feedback: at the World watershed guidelines. Bank, the average lag between project approval and Computer-assisted surveying—by mobile phone evaluation is 7.5 years. or by tablet-wielding fi eld agents—now enables much faster and more reliable assembly and use of informa- Technology enables feedback for agile tion on the progress of projects. A nascent example is implementation and learning the Social Observatory, a pilot that embeds a variety of New technology should make M&E even faster, feedback systems into a multibillion dollar program cheaper, and more effective. A poster-child example on rural livelihoods—precisely the kind of complex is the World Bank–backed Sujala project in Karna- program noted earlier. The Observatory fi elds rigor- taka, India, which employed a technology-intensive ous but slow randomized controlled trials to deter- M&E system operated by the Indian Space Research mine whether particular phases or elements of the Organisation.31 This combined remote-sensing imag- program had the desired impact. It supports internet- ery with household surveys and sophisticated data connected management information systems, wired systems for management and geographic informa- from the village up, that allow program managers at tion. Detecting that benefi ts were not being focused the district or state level to detect and correct anom- on the poorest groups, the project fi xed the problem. alies in performance. Perhaps most interesting, it It rigorously demonstrated a 24-percent increase supports a benefi ciary-designed survey, administered 306 WORLD DEVELOPMENT REPORT 2016

Figure 6.7 Proportion of international International Academy in Kenya monitors teacher fi nancial institution projects with and student performance, allowing management successful outcomes to intervene when teachers, students, or schools are underperforming (see chapter 3). Bus rapid transport World Bank systems can automatically monitor vehicle speed and ridership, tracking maintenance, capacity use, and ADB: Sovereign responses to weather. Health systems can track use and outcomes if patients and providers use digital ADB: Nonsovereign ID cards to log services. Remote sensing can track program impacts on agriculture, water use, deforesta- IFAD tion reduction, or . Ground sensors can track the sustainability of infrastructure in remote IFC and confl ict-affected areas, as in Afghanistan and

AfDB Pakistan. These systems offer rich, representative, actionable real-time information. EBRD Organizational inertia impedes agile, 0 1020304050607080 feedback-intensive approaches Percent These techniques are technically feasible, but are they

Sources: Based on most recent available reviews of independent evaluations organizationally feasible? One obstacle is a skills gap, or independent validations of self-assessments. Data at http://bit.do but this can be remedied through training. Obstacles /WDR2016-Fig6_7. posed by incentives are more formidable. First, donor Note: ADB = Asian Development Bank; AfDB = African Development Bank; EBRD = European Bank for Reconstruction and Development; IFAD = agencies, their staff, and their recipients are often International Fund for Agricultural Development; IFC = International Finance focused more on maximizing disbursements than on Corporation. “Outcome” is an index combining effi cacy, relevance, and effi ciency. “Success” is a rating of 4 or better on a 6-point scale (IFAD, IFC, optimizing results. Disbursements are more imme- World Bank); 3 or better on a 4-point scale (ADB, EBRD); or 2.5 or better on diately linked to organizational power and survival, a 4-point scale (AfDB). and have traditionally been much easier to track and manage. by locally recruited women using a tablet computer. Second, organizations cling to procedures that The annual survey, planned for 1 million respondents, assume that projects can be thoroughly planned in aims to provide actionable feedback at all levels from advance and in detail. According to an OECD survey, village to state (box 6.3). “Development practitioners report being constrained Feedback is facilitated for programs that are by rigid results frameworks that are not fl exible built around data platforms. For instance, the Bridge enough to include new sets of information to adapt to

Figure 6.8 High-quality M&E improves project outcomes

Modest M&E Substantial M&E Negligible M&E High-quality M&E

Outcome (independently rated) Highly satisfactory Moderately unsatisfactory Satisfactory Unsatisfactory Moderately satisfactory Highly unsatisfactory

Source: WDR 2016 team based on World Bank data. Data at http://bit.do/WDR2016-Fig6_8. Note: Colors code the proportion of projects by outcome. The size of the circle is proportional to the number of projects in the M&E category. M&E = monitoring and evaluation. GLOBAL COOPERATION 307

Box 6.3 The Social Observatory and P-tracking

“Do you decide on what types of clothes to wear based statewide use.) Second, it is administered by local staff , not on your own preferences?” That’s a question on a survey by some external contractor. Third, it informs higher-level instrument to assess whether Tamil Nadu’s Empowerment project management but also provides actionable informa- and Poverty Reduction Project (part of the Pudhu Vaazhvu tion at the village level. Project, or PVP) is actually having an impact on women’s Digital technologies help. The surveys are administered empowerment. For project benefi ciaries, it was a touch- using tablet-based software, and the data are pooled stone indicator of empowerment. through village-based internet connections. (These, by the That may be because it was crafted by a group of the way, also allow villagers to buy train tickets, pay electric women the project is designed for. bills, and check school exam scores without hours-long The Social Observatory seeks to build diagnostics and queueing or travel.) Automating the survey makes it easier feedback loops into the implementation of PVP and similar to train the local interviewers. It is diffi cult to imagine how livelihood projects in India. These are complex projects, the Social Observatory could realize its goal of scaling up operating in hundreds or thousands of locations. They have to 1 million respondents, and providing timely annual infor- multiple objectives and lots of moving parts: job training, mation, using traditional paper-based surveys. loans for small-scale income-generating activities ranging Most interesting, the Observatory tackled ways to make from dairy production to sanitary napkin manufacture, this massive information set digestible, a problem facing even self-defense classes. So it is hard to get a fi x on what all producers of data, big or small—and a particularly is going on without a lot of information. To provide use- challenging problem when many users are illiterate. The ful feedback, the Observatory deploys a variety of tools, Observatory commissioned an interactive graphical dis- including randomized controlled trials, internet-connected play that allows a village to compare its performance with management information systems, and “P-tracking,” the neighbors, catalyzing healthy competition. participatory approach to monitoring. Like so many promising ideas covered in this report, P-tracking builds on a few innovations. First, as the P-tracking is still in a demonstration phase. About 40,000 name suggests, the survey instrument was designed in interviews have been completed in a pilot district, and it a participatory fashion with a group of benefi ciaries to will be a big leap to cover the projected million. Another ensure that the information is seen as relevant. Indeed, the challenge will be to mainstream annual P-tracking data into women report that the acts of administering and respond- project implementation at all levels. Information supply is ing to the survey helped change perceptions. (The survey just half the battle. is not village-specifi c; after testing, it was standardized for

Source: WDR 2016 team; http://www.worldbank.org/en/programs/social-observatory.

changing contexts and emerging challenges.”32 Third, learning. Many private companies, and some devel- when funders—such as taxpayers—cannot directly opment agencies such as the Bangladeshi-based observe results, there can be a confl ict between account- development organization BRAC, already have cul- ability and learning. The OECD survey notes that tures that do this. For others, culture change will be “results data for accountability purposes tend to a challenge. Traditional and nontraditional agencies emphasize the positive,” getting in the way of learning. are experimenting with ways to make their business Ninety percent of the agencies surveyed reported diffi - more agile—ways that are untried and carry risks as culties in using results for accountability. In sum, as well as promise. seen also in chapter 3, rapid feedback may be technically One approach concentrates on inclusion—bring- feasible, but it will have limited impact unless there ing in feedback and ideas at the design stage so as is willingness to accept it and fl exibility to act on it. to combine global technical expertise with profound local insights and knowledge. Amplify, an innovation A focus on results and learning can shift of the U.K. Department for International Develop- incentives toward demand for information ment (DFID), tries to build collaborative networks of Thus the key to technology effectiveness is, once development practitioners to develop more inclusive again, in the analog world—in institutional struc- proposals for grant funding. It uses online crowd- tures and incentives that elevate feedback and sourcing to suggest and refi ne ideas, and facilitates 308 WORLD DEVELOPMENT REPORT 2016

collaborations among distant groups—such as a New phase. Those with demonstrated impact can get large York design study and a Nepalese NGO. grants for scaling up. Although Development Innova- Another approach to changing institutional incen- tion Ventures is not restricted to digital technologies, tives is by pursuing outcome-based aid. This shifts many of the supported concepts are based on ICTs or funder and client attention from disbursements to use ICTs for monitoring in the demonstration phase. results. Disbursements are made only after achiev- ICTs could thus speed the more reliable demonstra- ing pre-agreed outcomes. This should motivate both tion and dissemination of investment ideas. funders and recipients to rapidly track processes, To sum up, an alternative vision of development outputs, and results, introducing the possibility of practice starts with incentives. Project or program fl exibility in getting those results; a downside is staff get incentives and fl exibility to pursue results. shifting implementation risk to the recipients. The So do benefi ciaries. Digital technologies facilitate World Bank’s new Program for Results (P4R) uses rapid, real-time monitoring of outputs and outcome independent verifi cation of outputs and outcomes to results that matter, and provide this information motivate countries to beef up their own monitoring in a comprehensible form to those who can act on systems. In Vietnam, for instance, a P4R sanitation it—complemented by systems that assess impacts. project links disbursements to installations of house- Consistent with privacy considerations, these data hold toilets. The state audit agency sends out survey- streams would be made public, as open data. Any- ors equipped with smartphones to independently one—program managers, civil society, academics, or verify, with GPS-stamped photos, a sample of health auditors—can have access to this information. With ministry–reported installations. Beyond simply many eyes on the information and a profusion of incentivizing particular outcomes, this approach may analyses, accountability and learning are enhanced. catalyze a wholesale shift in P4R clients toward a cul- In this system of open-source evaluation, it is important ture of results, motivating them to invest in real-time to have an independent group assure integrity of the monitoring. data system. More disruptively, development impact bonds have been proposed as a way to use an outcome-based Taking information to scale approach to solve agencies’ constraints of infl exi- How can emerging opportunities to build internet- bility, risk aversion, and short time horizons. In this based services make development and environmen- relatively untried approach, agencies contract with tal management more effi cient and inclusive? These a private party or NGO for the achievement of spec- services reduce the cost of information as an input, ifi ed and verifi ed results, potentially over a period thereby reducing risks or improving decisions—for longer than most projects. The contractor has the individuals, for managers of infrastructure net- fl exibility to decide how to achieve those results— works, and for public authorities. The rationale for and could profi t or lose depending on performance. involvement by development or international agen- But the approach’s effectiveness will depend on how cies is fourfold. First is a characteristic feature of well specifi ed and reliably measured the indicators information services: many of these interventions are for making disbursements. Poorly specifi ed are expensive to set up and run, but can distribute proxies could produce perverse impacts, as could information at very low marginal cost. So fi nancing measurement systems subject to tampering. Digital becomes problematic. If the service is nearly costless technologies could make a difference by allowing to distribute, then it is effi cient to provide it to every- better measurement and independent verifi cation one who might benefi t. But the fi xed costs must be of the desired outcomes. They could also monitor fi nanced somehow. High fees might exclude poor unexpected social or environmental impacts. And people; low fees imperil sustainability; price discrim- they can be used to collect rich qualitative data as a ination might be diffi cult. And if the information can check against the potential bias of narrowly specifi ed be copied, it may be diffi cult to get anyone to pay for quantitative targets. it. So there is often an argument for public or donor Another way to motivate attention to results fi nance. Second, the information often underpins is by an evidence-based approach to fi nancing, as a public good, such as environmental protection or illustrated by the Development Innovation Ventures poverty reduction, reinforcing the rationale for pub- initiative of the U.S. Agency for International Devel- lic fi nance. Third, development agencies—including opment (USAID). Small grants support proof-of- NGOs and foundations—can play a distinctive role concept pilots. Ideas that have passed the fi rst stage not only through fi nance, but also via capacity build- are rigorously evaluated in a second-stage testing ing, and support for complementary investments GLOBAL COOPERATION 309 and policy reforms. Finally, there is a need for inter- For tropical maize in Africa, yields are typically national coordination to face planetary issues such only one- fi fth of the potential, and for other crops and as climate change. places in the developing world, yields rarely exceed What follows are some concrete examples of how 80 percent of the potential.33 Many factors account information and analytics might be able to boost effi - for this poor performance, but a lack of information— ciency at the individual and system level, with impli- about techniques, weather forecasts, and markets—is cations for the role of development agencies (box 6.4). a part of the gap. Traditional agricultural extension agencies have tried to fi ll the information gap, but Information as an input to individuals have depended on in-person farm visits and printed Poor people are often poor in critical information: material. Extension agents are typically assigned to about the benefi ts of hand-washing, or tomorrow’s serve an impossibly large number of farmers spread weather forecast, or the price their produce will over a logistically daunting stretch of countryside, fetch. When information becomes dramatically so information fl ows are low, and do not include cheaper, there is a prospect for welfare gains. Some customized information based on seasonal and daily of the clearest and most relevant examples to poverty weather forecasts. reduction are in agriculture, which will here serve Pilot programs are beginning to tap the poten- as an illustration of the wider potential across many tial of ICTs to reduce costs and increase the reach aspects of life and the economy. of extension efforts. Some initiatives in India show

Box 6.4 ICTs and the Sustainable Development Goals

In the year 2000, when the international development of 2015b—achieving “universal” coverage would imply community came together at the United Nations (UN) to a required growth rate of 51 percent a year, much faster approve the Millennium Development Goals (MDGs),a the than the LDCs have achieved since 2000. Aff ordability digital revolution was just beginning. At that time, there is also likely to be a challenge as only 1 of the 48 LDCs, were fewer than 1 billion mobile phones in use worldwide Bhutan, actually met the UN target in 2013.c Neverthe- and just 400 million internet users. In the intervening years less, LDC mobile coverage currently stands at close to to 2015, when the MDGs were reviewed, both indicators 70 subscriptions per 100 inhabitants. As those mobile have grown by more than sevenfold. Thus target 18—to phones are upgraded to smartphones, and as coverage make available the benefi ts of new technologies, including of 3G (third-generation) and 4G (fourth-generation) ICTs (information and communication technologies)—was signal expands, this target will become more attainable— one where progress was easy to show. although perhaps not by 2020. But with the new generation of Sustainable Develop- The real signifi cance of the internet for the SDGs is ment Goals (SDGs), adopted by the UN in 2015, the bar has likely to lie in helping to achieve other targets, such as been raised. As part of a broader goal 9, on infrastructure, target 3.9 on achieving universal health coverage, target industrialization, and innovation, the SDGs commit to “sig- 5b on promoting women’s empowerment, or target 10c on nifi cantly increase access to ICTs and strive to provide uni- reducing the transmission costs of migrant remittances to versal and aff ordable access to the internet in LDCs [least below 3 percent.d As noted, timely, fi ne-grained informa- developed countries] by 2020.” Given that many developed tion on households, the economy, and the environment can countries are already close to saturation in terms of mobile accelerate achievement of the SDGs. This information can penetration, it is logical to focus on internet access, and provide actionable feedback on where and why progress specifi cally on the needs of the LDCs. The two key words in is or is not being made. This is the essence of the “data the new target are “universal” and “aff ordable.” revolution,” called for by the High-Level Panel of Eminent Given the current low level of internet penetration Persons in their 2013 report, which kick-started discussion in the LDCs—just over 10 per 100 inhabitants by the end of the post-2015 development agenda.e

a. See http://www.un.org/millenniumgoals/. b. See ITU 2015. c. The target for aff ordability, as defi ned by the UN Broadband Commission, is a monthly price for entry-level broadband below 5 percent of gross national income (GNI) per capita: see http://www.internetsociety.org/map/global-internet-report/?gclid=CLv0tI35yMYCFQ2QHwodcWYGaw#aff ordability -mobile-broadband. d. See UN 2014. e. See UN 2013. 310 WORLD DEVELOPMENT REPORT 2016

Box 6.5 Digital Green: “How-to” videos for agriculture and health

Digital Green uses video cameras, battery-operated projec- The system automatically keeps track of which videos tors, and the internet to get information into the hands of are most likely to prompt adoption, so monitoring is an villagers in India and Africa. It works like this: Local peo- integral element of operations, not a burdensome add-on. ple fi lm local farmers demonstrating a useful agricultural An early self-evaluation reported a 90-percent reduction technique or health practice. These include preparation in the cost per adoption. A randomized controlled trial of seeds, enhancements of milk yield in cattle, and oral found that 16 percent of farmers exposed to Digital Green rehydration therapy for children suff ering from diarrhea. adopted a yield-increasing, water-conserving rice culti- The videos are pooled on the internet. Then local resource vation system, versus 10 percent in a control group. The people download, show, and explain seasonally appropri- program, in operation since 2008, reached 660,000 people ate videos to community assemblies. The premise is that in 7,645 villages by April 2015. the home-grown videos convey the technical details of the technique, and do it with the credibility of a neighbor speaking in a familiar dialect.

Sources: http://www.digitalgreen.org; Gandhi and others 2009; Vasilaky and others 2015.

mixed results.34 A program that sent voice advice to developed by the national rice association with sup- farmers was little used and had no impact. Another, port from the International Center for Tropical targeted at cotton farmers, shifted them to less harm- Research. Colombian rice yields have been falling as ful pesticides and more valuable crops. A program increased climate variability invalidates traditional that put smartphone-equipped extension agents in calendar-based rules of thumb on when to plant. The contact with experts improved farmer awareness of decision-support system merged historical data on extension services, but the impact on incomes was climate and harvests with current weather forecasts not measured. In Peru, however, a careful experiment to advise farmers on optimal planting decisions. In an showed that provision of market information via early application, the model correctly foresaw that cellphones boosted farmers’ selling prices by 13 per- crops would fail due to drought and advised farmers cent and increased the participation of subsistence not to waste time and money on planting. farmers in the markets.35 The effects were particularly Donor funds (including NGO funds) and technical marked for perishable goods and for risk-averse farm- inputs have been important in most of these initia- ers. A particularly interesting program—because its tives. Many fail to scale up. All face the challenge of “how-to” video approach could be applied in domains fi nding a sustainable business model. This may entail outside agriculture—is Digital Green (box 6.5). a combination of subscription sales and recurrent More ambitiously, a movement toward “precision public fi nancing. Assessment of the benefi ts of this agriculture” crunches big data to provide farming information could help in pricing decisions and in advice. At its most capital-intensive, an instrumented determining the role of public fi nance. tractor gathers georeferenced data on fi eld condi- tions and plant health. These data are combined with Information as an input to managing weather data and fed into a computer model of plant complex systems growth. The model then works through the tractor, Much of development involves managing complex fi ne-tuning the application of fertilizer and pesti- systems, such as power networks and urban transpor- cide as it traverses the fi eld. By curbing overuse of tation. With better information and analysis, these fertilizer, water, and pesticides, precision agriculture systems can be designed and managed for greater can improve the environment while boosting farm effi ciency. profi ts. Power utilities can be helped by information in a More affordable approaches are needed for the variety of ways (see sector focus 5, “Energy”). Devel- developing world. An emergent example is a system of oping countries, though short on power, nonetheless decision support for rice agriculture in Colombia see much of their electricity production lost to poor GLOBAL COOPERATION 311 maintenance and to theft. Data analytics can opti- digital registration of land titles reduces the cost of mize schedules for transformer maintenance and verifying ownership and inhibits fraud and corrup- replacement, and can pinpoint locations for attention tion. Thus land registration is a platform for building to reduce theft. Big data can also be used to enable a variety of valuable services. grids to assimilate the ebb and fl ow of power from As an example of the potential, a pilot land wind turbines and solar panels, and to adjust prices in regularization project in Rwanda used aerial pho- real time to shave off peaks in demand. Put together, tography and digital registration, and was found to these tweaks in system operation can reduce power double the rate of investment and maintenance of outages, reduce the need to invest in expensive peak soil conservation structures.37 The effect was partic- supply, improve the bankability of power, and reduce ularly marked for female-headed households. But environmental damages. Many of the same principles the gender impacts were complicated because of can be applied to water utilities. the way tenure regularization interacted with laws Urban transport is another complex system where and traditions. Married women improved their land newly cheap information unlocks possibilities for access, but those in informal unions were worse off. greater effi ciency. A good starting point is with the Girls’ and boys’ planned land inheritance was equally most basic foundation of planning: of the 25 largest boosted—except in female-headed households, where low- and lower-middle-income cities, 92 percent do sons gained an edge. Contrary to expectations, there not have complete maps of their transit networks.36 was no effect on access to credit. Overall, the pilot Compiling these maps used to be time-consuming and was suffi ciently successful that it is being expanded expensive. Recently, though, Manila developed and nationwide, to 10 million land parcels, at a cost of applied a mobile phone–based application to survey US$5 a parcel. and map routes, using an open-source data standard. Environmental management also depends on land The map powers a consumer trip-planning app and data. Since the 1970s, Brazil has used satellites to is being used by city planners to reduce redundant monitor deforestation in the Amazon. In 2003 the routes and plan a new mass transit corridor. In Seoul, government made a landmark decision to release the the transit agency used locational data from 3 billion raw data. This enabled NGOs such as IMAZON to nighttime phone messages to map travel patterns develop their own analyses, improving the informa- and design an optimal route for the immense city’s tion base and raising public awareness and pressure fi rst late-night bus service. Traditional household for environmental action. Starting in 2004, the Brazil- surveys would have been slower, more expensive, and ian government began using monthly data to guide less informative. Moving on from static mapping, an enforcement of forest and agricultural policy, with initiative in Cebu, the Philippines, was able to ingest demonstrable impacts on reducing deforestation.38 real-time taxi data to generate speed and congestion More recently, Brazil embarked on an ambitious maps for the entire city. This reduced the time to program to precisely map every rural property and analyze travel time for a bus corridor from two weeks use satellite observations to monitor compliance to two seconds. Ultimately cities may be able to use with forest conservation regulations. By September real-time data to design and implement policies that 2015, it had registered 2.1 million properties cover- increase transit accessibility, decrease travel time, ing 2.6 million square kilometers.39 Registry in this substitute for expensive road construction, and abate environmental cadastre is a condition for credit and congestion and pollution. other services, motivating compliance. It also sets up Rural land management, to be effi cient, depends a framework for trading forest conservation obliga- on accurate and accessible information about land tions, which reduces the cost and increases the ben- plots. Land sales and rural credit markets function efi ts of conservation. better when there is a comprehensive system of land Disaster risk management inherently demands a registration. Secure, registered land tenure also gives great deal of timely and location-specifi c information. farmers incentives to invest in their land and protects Over just the past few years, internet-enabled infor- them from illegal expropriation. ICTs have revolu- mation has begun to change the way in which donors tionized the logistic burden of mapping, recording, respond to emergencies, and is helping to increase and verifying landholdings—key elements of docu- disaster preparedness and resilience (box 6.6). menting land tenure. They do so in two ways. First, For all these systems’ management challenges, mapping boundaries with GPS is much faster, less big data offers plausible effi ciency gains. In many expensive, and less demanding of scarce skills than cases, realizing those gains involves overcoming traditional surveying. Second, open and transparent not just technical obstacles, but also political ones. 312 WORLD DEVELOPMENT REPORT 2016

Box 6.6 Disaster risk management

As climate becomes more volatile and populations swell in operations. Here again, crowdsourcing has revolutionized coastal, fl oodplain, and earthquake-prone areas, disaster risk response. After the devastating , a management (DRM) becomes ever more important. Annual spontaneous volunteer eff ort used the Ushahidi platform to disaster losses now average US$200 billion globally, and large aggregate and map status reports and requests for help from disasters can derail a country’s growth. Information and com- social media and text messages. Over 1,000 Haitian volun- munication technology (ICT) applications—especially those teers living abroad translated and geolocated the messages. involving crowdsourced, open data and open-source soft- And more than 700 volunteers used to ware—are rapidly being incorporated in each aspect of DRM: create new maps. These were subsequently updated, by Risk identifi cation and mitigation. The best way to deal more volunteers, with satellite and aerial imagery, to form the with disasters is to build resilience in advance, so that when basis for the postdisaster needs assessment. Technological the storm or fl ood or earthquake hits, lives are not lost and progress continues: after Vanuatu’s 2015 cyclone, drones damage to property is minimized. This requires extensive, were used to survey damage, providing much more detailed geographically precise information and planning. What damage information than is possible with aerial photography. locations are subject to natural disasters, with what proba- The move of these methods to open-source software bility? How sturdy are the buildings in those locations, and and open data is noteworthy, and has implications beyond who occupies them? Which ones need to be shored up and DRM. For capacity- and cash-constrained agencies in devel- how much will that cost? Detailed answers have typically oping countries, open-source software is much cheaper to not been available. Map data have been inadequate, often acquire than proprietary software, and avoids dependency inaccessible, and diffi cult to handle. But with revolutions on a single supplier. It allows high levels of customization in geographic information systems, remote sensing, and and, because it often adheres to open standards, easier crowdsourcing, that is changing. integration with other software. It encourages cross- In fl ood-prone Jakarta, Indonesia, for instance, the fertilization, community building, and code sharing with Disaster Management Agency asked the heads of the city’s others around the world; for an example, see geonode.org. 267 wards to map the location of critical infrastructure These advantages need to be weighed against the poten- using OpenStreetMaps, a tool that allows volunteers to cre- tially (but not necessarily) higher level of support off ered ate detailed digital maps that can be used without restric- by proprietary software. tion. More than 1 million buildings have been mapped. Similar considerations apply to open data. By opening InaSAFE, an open-source software program, was then used data, it is easier to fi nd and assemble the multiple layers to analyze potential impacts of fl oods, and helped in the of information—land cover, topography, population, infra- response to the 2013 and 2014 fl oods. More recently, peta- structure, hydrology—necessary for DRM. And, of course, jakarta.org exploited the city’s high rate of Twitter usage to open data can be distributed to users free of charge. crowd-map fl ood locations in real time during the 2014–15 Crowdsourced data are not without diffi culties. Accuracy monsoon season, helping citizens navigate the fl ooded city and completeness are issues. Both Haiti and Jakarta and augmenting the DRM agency’s information base. developed systems for confi rming the accuracy of reports. Early warning. Many lives have been lost for lack of Sustainability of volunteer engagement is a challenge, adequate warning of impending disaster. Now, widespread since DRM data need constant updating. Feeding back the access to mobile phones, combined with weather monitor- information to communities is one way to encourage this. ing systems, allows for timely and near-universal warnings. Of course, crowds are just one data source. A key to DRM For instance, the Philippines’s Project NOAH (Nationwide is the integration of many diff erent kinds of information. Operational Assessment of Hazards) uses real-time data Rio de Janeiro’s Center of Operations prepares the city to from rain gauges, water sensors, and radar to provide evac- deal with its many risks, including fl oods and landslides, by uation alarms. combining real-time information on rain, tides, transport, Damage assessment and response. In the chaos following and city services with 560 video cameras. It also serves to a disaster, it has been diffi cult to match relief supplies with coordinate the diverse government departments, and to need, and then to inventory damage and prioritize recovery feed information to the public and the media.

Sources: WDR 2016 team based on inputs from the Global Facility for Disaster Reduction and Recovery; Haklay and others 2014; Heinzelman and Waters 2010; Holderness and Turpin 2014; Matheus and Ribeiro 2014. GLOBAL COOPERATION 313

Powerful interest groups may stand to lose from sys- alone at the fi ne-grained level relevant to provinces tems effi ciency. These include consumers and utility or communities. employees who benefi t from electricity theft, oper- However, many countries lack the money to fund ators of unregulated minibuses, land-grabbers, and basic information systems, the organizational incen- fl outers of environmental regulations. Disaster risk tives to share information, and the human skills to management and climate adaptation require exten- supply and use data. International actors can help on sive coordination among ministries used to working all counts. Donors have supported keystone house- independently. Donors may be able to play a role in hold surveys, including the Living Standards Mea- supporting the complementary reforms necessary to surement Surveys and the Demographic and Health overcome these obstacles. Surveys. These have provided crucial information for decision making, with the added benefi t of facilitat- Information as an input to poverty ing international comparisons, thanks to standard- reduction eff orts ized defi nitions. Donors have also supported capacity In fi ghting poverty, as in any strategic endeavor, it building at national statistical organizations. But is useful to diagnose challenges and track progress. as noted, coverage is far from complete and timely. Timely, disaggregated poverty indicators could pro- A comprehensive program of statistical gap-fi lling mote accountability, boost motivation, and provide for the 77 low-income countries that belong to the the kind of relevant feedback described earlier in World Bank’s International Development Association this chapter, at the policy level. For instance, many (IDA) is estimated to cost about US$1 billion a year.43 proposed policy reforms have potentially large and Current offi cial development assistance runs around poorly understood distributional impacts, sometimes US$350 million, according to the report, while domes- leading to contentious gridlock. The ability to predict tic spending is not known. those impacts and then to verify them in practice Costs may be reduced, or effectiveness improved, could help to accelerate reforms. Up-to-date informa- through new technologies. For instance, computer- tion on the location, health status, and consumption aided surveying can reduce the cost of traditional patterns of poor people could help design and target household surveys by 60 percent.44 New opportuni- programs more effi ciently. For these reasons, a panel ties are arising to complement traditional censuses commissioned by the United Nations40 sees better data and surveys with big data that can provide up-to-date as an indispensable tool for achieving the Sustainable information at a fi ne geographical level. For instance, Development Goals (SDGs) (see spotlight 5, “The data cellphone call detail records are being used to impute revolution”). real-time measures of poverty—and internet data, to To date, however, development decisions and construct real-time infl ation or even GDP measures. priorities have often been made while fl ying blind. These techniques are not likely to replace traditional For instance, reform—a trillion- statistical sources. Quite the contrary: high-quality dollar-a-year issue with huge poverty and environ- censuses and surveys become even more valuable mental implications—is contentious in many coun- because they can adjust for biases in big data (see tries. Yet an authoritative global analysis of the pov- box 6.7). erty impacts of reform was forced to use household data that were, on average, more than eight years old.41 Mustering information for global public Consider this sobering picture of the state of knowl- goods edge about African growth and poverty.42 Just half of Many problems—climate change, ozone depletion, African countries compute GDP using current stan- air pollution, epidemics, fi nancial crises—are features dards; when Ghana adopted those standards, national of globally interconnected environmental, economic, income was found to be nearly two-thirds higher and social systems. Addressing them requires coordi- than previously thought. Only one-third of countries nated global actions. Setting priorities and targeting have two comparable measures of poverty over time, actions require global information. That information and even then the latest estimate is often years out is itself a global public good. As such, it is subject to of date. To make things worse, basic civil registra- the scale effects described earlier in this Report. tion systems cover only 45 percent of births in Sub- Perhaps the quintessential example is the Saharan Africa, making it diffi cult to track child mor- immensely valuable data on weather, water fl ows, tality (one of the starkest measures of poverty) or to and climate. One estimate of the benefi ts of upgrad- get an accurate measure of population. Thus it is dif- ing developing country weather systems to devel- fi cult to track progress even at the national level, let oped country standards is between US$4 billion 314 WORLD DEVELOPMENT REPORT 2016

Box 6.7 Can “big data” provide real-time, geographically detailed information on poverty?

Traditional surveys and newfangled big data sources have buys airtime, in what denominations, and the frequency of complementary strengths and weaknesses. Surveys work international calls. Hurdles include fi nding ways to ensure very hard at being accurate—for instance, precisely tallying data anonymity and convincing phone companies to share the value of everything a household buys or produces— data. But the big disadvantage, from a monitoring per- and making sure that everyone in the target population, spective, is that the data do not accurately represent the especially poor people, are covered. This makes it possible population. Many poor people do not have phones. And the to accurately measure changes over time. But even with relationship between income and airtime purchases will tablet-assisted surveying, traditional household survey data change over time as prices of service and phones change. will be months old before they are published. And sample One possible solution is to use repeated small “gold sizes of a few thousand do not allow much insight into what standard” surveys to calibrate the cellphone data. A hint is happening with subgroups or at particular locations. of the possibilities is given by a study that successfully rep- So researchers are trying to estimate income with licated the results of a populationwide opinion poll based cellphone use data, which cover a large proportion of the on a survey of Xbox gamers—a predominantly young, male population, are literally up to the minute, and can provide a group. The key was a sophisticated method of correcting snapshot of activity at the level of a village or urban ward. for the skewed nature of the sample using information on The estimates use information such as how often the user the age and sex distribution of the general population.a

a. Wang and others, forthcoming.

and US$36 billion a year.45 Data on river fl ows are publishes monthly deforestation maps that can spot essential for disaster risk planning and for planning clearings of as little as 25 hectares.48 On the horizon and operating hydropower and irrigation systems. are satellite data that might enable daily monitoring And detailed long-term data are essential for tracking at a level well below a hectare. Similar advances are and understanding climate change patterns. These being made in transboundary air pollution (see sector data are sorely underprovided. Because weather is focus 6, “Environmental management”) and surveil- a global system, every observation of temperature lance of emergent zoonotic diseases. and precipitation improves weather forecasts (and So there is a role for global fi nance in the supply climate models) at the transnational and even global of this kind of information, but that is only the begin- levels. But poor countries do not have a dense enough ning. There is a need also to build the willingness to network of observation stations even for their own share information, and the capacity to use it. purposes, let alone to contribute to the global good (see map 6.2). Another 4,000 to 5,000 stations are The future of global development needed in Africa alone.46 Part of the problem is a lack cooperation of funds and capacity to build and maintain stations. What’s the role of development agents and interna- Another is that many state-run hydrometeorological tional organizations in a world where their fi nancial agencies refrain from the free distribution of globally heft is miniscule? The answer set forth here is three- valuable data to garner modest revenues from selling fold: in concentrated efforts in the poorest countries; data points, not just in poor African countries but also in catalytic injections of ideas, backed with cash, in the majority of European countries.47 everywhere; and in spearheading global collective Once again, technology is making it easier to action. The data and technology revolutions are arriv- gather and analyze data of global signifi cance. For ing in time to bridge the gap between resources and instance, tropical deforestation is a global concern, ambition by amplifying the impacts of action and a major source of carbon emissions and biodiversity including more people in the formulation and execu- loss and an issue related in complex ways to rural tion of plans. But for this to work, development agents poverty. Yet until recently, data on tropical deforesta- must tackle policy constraints, internal and external. tion were spotty, inconsistent among countries, Start with the how of development operations. and woefully outdated. Now, Global Forest Watch With new technologies, development agencies can GLOBAL COOPERATION 315

Map 6.2 Availability of reports from weather stations

Proportion of time that weather stations report SYNOP (surface synoptic observations) data to the World Meteorological Organization

Report 90–100 percent of the time 3,678 Report 45–89 percent of the time 573 Report < 45 percent of the time 193 Silent station 444

IBRD 42013 Source: World Meteorological Organization SYNOP data for October 2013, https://www.wmo.int/pages/prog/www/ois/monitor/index_en.html.

be more inclusive by tapping the wisdom of benefi - tion. These services have fi xed setup costs in software ciaries in designing interventions. They can crank and data assembly, but can have near zero costs for up their effi ciency by using rapid feedback to refi ne distributing information. So the private sector will and improve their actions through trial and error, tend either to shy away from providing these services, and use open-source evaluation to align performance or to price them at a level that shuts out poor people management with learning. But these approaches who could benefi t. will not come easily in organizations that emphasize Donors can provide the funds and expertise to spending and outputs over results, have burdensome help get these services going. They can, for instance, structures for accountability, and see any failures support the assembly of critical data and the creation as damning rather than informative. If traditional of broadly applicable open-source software. There agencies cannot adapt, some of their business may is a real need, however, for coordination. A DFID be taken up by disruptive newcomers, such as Global- study found 70 unrelated mobile-for-health pilots in Giving, GiveDirectly, or development impact bonds Uganda, and over 1,800 mobile apps worldwide.49 (box 6.8). There’s another reason for development agents Next, the what. Previous chapters have pointed to to get involved: to help address the policy issues that many areas where external agencies could catalyze impede innovation, or to support complementary development—for instance, in supporting policy investments or capacity building. Thus, for instance, reforms that would open up internet access or new making progress on urban transit could require a approaches to service delivery in health. In this chap- combination of information systems, investments ter, the emphasis is on creating information services in infrastructure, and investments that address that help individuals and systems managers make opposition to regulation of minibuses or reduction in better decisions in ways that promote poverty reduc- on-street parking. 316 WORLD DEVELOPMENT REPORT 2016

Box 6.8 Disruptive approaches to development

Some radical reimaginings of development assistance regularly report audits), feeding this information back to involve digital technologies to eliminate some of the over- donors. GlobalGiving uses its platform to provide online head costs and procedures of traditional approaches. Give- capacity building to the grantees, and allows them to build a Directly uses satellite imagery, crowdsourced assistance, reputation based on performance and investments in learn- and machine learning to identify poor villages in Kenya.a ing. It plans to assist the grantees in tracking their results, The trick: using thatched roofs as an indicator of poverty. ultimately fostering benchmarking and mutual learning. Staff are dispatched to the villages to enroll the residents GiveDirectly and GlobalGiving thus sidestep much of the thatched houses. GiveDirectly then sends US$300, of the traditional costly apparatus for project selection, by mobile phone, to the enrollees. Overhead costs are just appraisal, and supervision. GiveDirectly achieves these 8 percent (for staff and the cost of foreign exchange and effi ciencies by focusing on a logistically simple interven- money transfers). A randomized controlled trial found a tion. This cannot substitute for interventions that involve 42-percent reduction in the number of days children go externalities or require collective action or large invest- without food, a 58-percent increase in assets, and a 38- ments. But it implicitly challenges interventions that have percent increase in revenue from self-employment.b more “moving parts”—can they achieve equal impact at GlobalGiving’s internet platform allows nongovern- equal cost? Global-Giving achieves effi ciencies by pushing mental organizations (NGOs) to post development project project appraisal, quality assurance, fi nancial controls, and proposals for funding by donors, large and small. Thus safeguards to the funders and to , it taps a more grassroots fi eld of ideas and implementers informed by a growing stream of data. Time will tell than traditional agencies. Its website keeps track of project whether these are inherently niche operations, or the pio- performance and NGO credibility (such as whether NGOs neers of a new way of doing development business.

a. Abelson, Varshney, and Sun 2014. b. Haushofer and Shapiro 2013.

Finally, external agents and international organi- could work with partners to expand the availability of zations can encourage the unlocking of existing data. open data. The principle would be that all supported Open data can be a powerful tool (see the discussion activities should generate open data by default, with of open data in the context of disaster risk manage- exceptions on the grounds of privacy and other ment in box 6.6). But too often, socially benefi cial data defi ned concerns. So, support for health and educa- are kept under wraps. Both the public and private sec- tion systems should ensure that these systems track tor are culprits. As noted, this is true of government- and disclose performance data to the extent possi- collected data on weather and river fl ows. In some ble, while preserving confi dentiality of individual transboundary river basins, upstream countries do records. This also holds for public-private partner- not share river fl ow data with downstream neighbors, ships (PPP), many of which currently do not furnish crimping fl ood preparedness and irrigation man- any data about their performance, and which may use agement. Basic data on poverty also suffer. Among client-generated data for commercial gain. Agencies the Sub-Saharan household surveys that underpin that advise on or fi nance concessions in power, water, poverty estimates and give insight into causes and transit, telecom, and other services could demon- correlates of income, education, and health out- strate to public authorities the advantages of making comes, fewer than half are available online and free of performance data open. Data could include public charge.50 For East Asia and the Pacifi c, the proportion scrutiny of the concessionaire, informative new is less than 20 percent. Private telecom companies data “mash-ups,” and the capacity for open-source closely hold, and use for commercial purposes, data evaluation of options for improved equity and effi - on cellphone calling patterns useful for transport ciency. Agencies could then help governments craft planning and poverty analysis. PPP contracts that recognize the rights of clients to Donors and development agencies can catalyze their own information and of the public to ano- the potential for open data in several ways. They nymized performance data. Donors, both traditional GLOBAL COOPERATION 317 and nontraditional, can set a good example by ensur- .org/english/bodies/hrcouncil/docs/17session/A.HRC ing that their own data are open—particularly, mete- .17.27_en.pdf. orological and hydrological data with transboundary The UN Human Rights Council issued a second benefi ts. They can support policy reforms in devel- report in April 2013 suggesting that nations’ laws on oping countries that fund data-producing agencies, surveillance were overbearing and recommending that states revise and align their surveillance laws to allowing them to reduce fees. comply with international human rights. See http:// Agencies or international organizations could also www.ohchr.org/Documents/HRBodies/HRCouncil explore the potential to create sectoral data standards. /RegularSession/Session23/A.HRC.23.40_EN.pdf. Open data are more useful when they can be consis- In June 2012, the UN Human Rights Council tently compared and aggregated across units. Data adopted a resolution preserving human rights on the standards also make it easier to develop apps and internet, affi rming that people have the same rights programs for use and analysis. The Group on Earth online that they have offl ine—in particular, freedom Observations, an international consortium, is doing of expression (A/HRC/20/L.13) (June 29, 2012). this for remote sensing data. Development agencies 12. The WCIT Final Acts are available at http://www.itu could work together to create standards for data on .int/en/wcit-12/Pages/default.aspx. For an analysis of the outcome, see http://www.cullen-international development impact or results. A beginning would be .com/asset/?location=/content/assets/regulatory to more thoroughly report results under the Interna- -intelligence/regulatory-news/wcit-12_post-mortem tional Aid Transparency Initiative standards. Some _culleninternational.pdf/wcit-12_post-mortem agencies report fi nancial transactions, but only 5 out _culleninternational.pdf. of 68 agencies publish machine-readable information 13. IANA, a subsidiary of ICANN, has traditionally per- on project targets and results—which would allow formed certain technical functions essential for the benchmarking and comparison across providers.51 smooth and secure running of the internet, includ- Finally, as noted, there are signifi cant fi nancing ing control of the internet protocol (IP) addressing needs for building the data that could guide the system (numbering resources, including IPv4 and pursuit of the SDGs, and the capacity to use those IPv6), managing IP protocols, and managing the root zone database for domain names (domain data. There could be opportunities to build reusable, name management). The U.S. government, through customizable open-source software and systems for the National Telecommunications and Informa- monitoring the SDGs, reducing duplication of efforts, tion Administration (NTIA), a branch of the U.S. saving costs—and improving lives. Department of Commerce, has had oversight over these functions through a series of agreements Notes with ICANN, fi rst formalized in 1998 under what was called a memorandum of understanding. The 1. David Clark’s often-quoted statement refers to the memorandum went through several iterations and work of the Internet Engineering Task Force (IETF). morphed into what was then called the Joint Project See https://www.ietf.org/tao.html. Agreement, which itself underwent one amend- 2. Wu 2010. ment. In 2009, it was replaced by a document called 3. World Summit on the Information Society 2005. the Affi rmation of Commitments. In each iteration 4. Much of the discussion in this section is derived of the agreement between NTIA and ICANN, and from Dutton (2015) for this Report. over time, NTIA has “loosened” its oversight of the 5. Sadowsky 2014, 82; UNESCO 2015. performance of these IANA functions by ICANN. 6. Portions of this section are adapted from Satola and This latest move by NTIA can be seen as part of a Kelly (2014). continuing evolution away from NTIA’s control and 7. See http://www.itu.int/wsis/index.html. oversight of IANA functions. 8. Dutton 2015 for this Report. 14. In August 2015, NTIA announced it would defer the 9. Dutton 2015 for this Report. deadline for considering a transition proposal for at 10. See, for example, Judy and Satola (2014). least a year. 11. In May 2011, UN Human Rights Council Special Rap- 15. In addition to endorsement of the MSM by the UN’s porteur Frank La Rue issued his report on the right World Summit on the Information Society, both the to freedom of expression on the internet, saying Organisation for Economic Co-operation and Devel- that states are increasingly censoring information opment and the European Commission have recon- online through arbitrary blocking or fi ltering of con- fi rmed their support. See European Commission (EC tent, criminalizing legitimate expression, imposing 2014) and OECD (2014b). intermediary liability, disconnecting users from 16. See European Commission (EC 2014). the internet, and not providing adequate protec- 17. It is important to note that the sale of music fi les tions to privacy and data. See http://www2.ohchr does not necessarily take place across borders. In 318 WORLD DEVELOPMENT REPORT 2016

many cases, the sale of digital goods across borders 30. OECD 2014a. is blocked by geographically limited IP regimes. This 31. IEG 2011. applies to other digital goods like e-books and videos. 32. OECD 2014a, 22. Instead, one should think about these cross-border 33. Lobell, Cassman, and Field 2009. transactions as transactions between fi rms in differ- 34. Nakasone, Torero, and Minten 2014. ent countries for the license to sell the content. See 35. Nakasone 2013. Nicholson and Noonan (2014) for a discussion. 36. World Bank 2015. 18. Mandel 2013. 37. Ali, Deininger, and Goldstein 2014. 19. Bauer and others 2014. 38. Assunção, Gandour, and Rocha 2013; Assunção and 20. Bauer and others (2014) calculate the percentage others 2013. changes to GDP, investment, and total exports as a 39. Government of Brazil 2015. result of the proposed legislation in a Global Trade 40. See UN 2013. Analysis Project (GTAP) model. The effects of the 41. del Granado, Coady, and Gillingham 2012. data regulations are included in the model as an 42. Devarajan 2013. increase in business costs, an increase in costs for 43. Espey 2015. data-driven services trade, a decrease in invest- 44. Espey 2015. ments, and a decrease in the effectiveness of data-re- 45. Hallegatte 2012. lated research and development. 46. Rogers and Tsirkunov 2013. 21. National Board of Trade 2015. 47. Rogers and Tsirkunov 2013. 22. For example, in October 2015, the European Court 48. See http://www.globalforestwatch.org. of Justice declared invalid the EU-U.S. “safe har- 49. Ranger, Chandler, and Arscott 2015. bor” rules regulating internet fi rms’ retention of 50. Demombynes and Sandefur 2014. Europeans’ data in the United States. An ongoing 51. Aid Transparency Index 2014. http://ati.publish case between Microsoft and the U.S. Department whatyoufund.org/fi ndingapproaches-to-publishing of Justice seeks to clarify whether U.S. government -results-information/. agencies can access e-mail messages stored in Microsoft servers in Europe. 23. For example, the EU’s Binding Corporate Rules References (BCRs) require fi rms to set up internal company pol- Abelson, Brian, Kush R. Varshney, and Joy Sun. 2014. icies and appoint personnel to handle cross-border “Targeting Direct Cash Transfers to the Extremely data transfers. These fi rms must go through a certi- Poor.” Paper presented at the 20th Association for fi cation process overseen by a data protection offi ce, Computing Machinery’s Special Interest Group on which involves audits and company visits. As of May Knowledge Discovery and Data Mining’s Interna- 2015, only 66 companies had obtained BCRs, and tional Conference on Knowledge Discovery and Data most of them are large multinational companies. Mining, New York City, August 24–27. 24. IATA 2015. Aguiar, Luis, and Bertin Martens. 2013. “Digital Music 25. A study by Aguiar and Martens (2013) fi nds that illegal music downloads do not displace legal music Consumption on the Internet: Evidence from Click- sales. The individuals in the study would not have stream Data.” Working Paper 2013/04, Institute for purchased the music they downloaded illegally Prospective Technological Studies Digital Economy, if it was not available on the illegal downloading Seville, Spain. website. Ali, Daniel Ayalew, Klaus Deininger, and Markus Gold- 26. van Pottelsberghe de la Potterie and Mejer 2010. stein. 2014. “Environmental and Gender Impacts of 27. See Easterly 2006, 2014; Andrews, Pritchett, and Land Tenure Regularization in Africa: Pilot Evidence Woolcock 2013; Ramalingam 2013; Whittle 2013; from Rwanda.” Journal of Development Economics 110: Booth and Unsworth 2014. 262–75. 28. Andrews, Pritchett, and Woolcock 2013; Whittle 2013. Andrews, Matt, Lant Pritchett, and Michael Woolcock. 29. The correlation shown in fi gure 6.8 could potentially 2013. “Escaping Capability Traps through Problem- have a spurious component, if the ex post M&E Driven Iterative Adaptation (PDIA).” World Develop- rating infl uences the outcome rating. However, the ment 51: 234–44. correlation holds when the independent M&E rating Assunção, Juliano, Clarissa Gandour, and Romero Rocha. is related to the self-rating of outcome. Moreover, a 2013. “DETERring Deforestation in the Brazilian review of 195 in-depth project evaluations looked at Amazon: Environmental Monitoring and Law positive and negative mentions of the specifi c roles Enforcement.” Technical report, Climate Policy Initia- of problem-solving, frequency of supervisory visits, tive, PUC-Rio. and M&E design quality. It found them to be highly Assunção, Juliano, Clarissa Gandour, Romero Rocha, and associated, in magnitude and statistical signifi cance, Rudi Rocha. 2013. “Does Credit Affect Deforestation? with better and worse project outcomes. Evidence from a Rural Credit Policy in the Brazilian GLOBAL COOPERATION 319

Amazon.” Technical report, Climate Policy Initiative, Agricultural Extension.” Information Technologies and PUC-Rio. International Development 5 (1): 1–15. Bauer, Johannes, and Bill Dutton. 2015. “Addressing the Gasser, Urs, Ryan Budish, and Sarah Myers West. 2015. Cybersecurity Paradox: Economic and Cultural Chal- “Multistakeholder as Governance Groups: Observa- lenges to an Open and Global Internet.” Background tions from Case Studies.” Berkman Center Research paper for the World Development Report 2016, World Publication 2015-1. http://papers.ssrn.com/sol3 Bank, Washington, DC. /papers.cfm?abstract_id=2549270. Bauer, Matthias, Hosuk Lee-Makiyama, Erik Van der Haklay, Mordechai, Vyron Antoniou, Sofi a Basiouka, Marcel, and Bert Verschelde. 2014. “The Costs of Data Robert Soden, and Peter Mooney. 2014. “Crowd- Localization: Friendly Fire on Economic Recovery.” sourced Geographic Information Use in Govern- ECIPE Occasional Paper 3/2014, European Centre for ment.” Report prepared for the World Bank, Global International Political Economy, Brussels. Facility for Disaster Reduction and Recovery, London. Booth, David, and Sue Unsworth. 2014. “Politically Smart, Hallegatte, Stéphane. 2012. “A Cost-Effective Solution Locally Led Development.” Discussion paper, Over- to Reduce Disaster Losses in Developing Countries: seas Development Institute, London. Hydro-Meteorological Services, Early Warning, and Brazil, Government of. 2015. “Cadastro ambiental rural: Evacuation.” Policy Research Working Paper 6058, Boletim informativo Setembro de 2015.” http://www World Bank, Washington, DC. .florestal.gov.br/download-document/2739-boletim Haushofer, Johannes, and Jeremy Shapiro. 2013. House- -informativo-setembro-de-2015. hold Response to Income Changes: Impacts of Uncondi- CIGI (Centre for International Governance Innova- tional Cash Transfers. Cambridge, MA: Abdul Latif tion), and Ipsos. 2014. “Global Survey on Internet Jameel Poverty Action Lab. Security and Trust.” https://www.cigionline.org Heinzelman, Jessica, and Carol Waters. 2010. Crowdsourc- /internet-survey. ing Crisis Information in Disaster-Aff ected Haiti. Special del Granado, Francisco Javier Arze, David Coady, and Report 252. Washington, DC: U.S. Institute of Peace. Robert Gillingham. 2012. “The Unequal Benefi ts of Holderness, Tomas, and Etienne Turpin. 2014. “Assessing Fuel Subsidies: A Review of Evidence for Developing the Role of Social Media for Civic Co-management Countries.” World Development 40 (11): 2234–48. during Monsoon Flooding in Jakarta, Indonesia.” Demombynes, G., and J. Sandefur. 2014. “Costing a Data Smart Infrastructure Facility, University of Wollon- Revolution.” Working Paper 383, Center for Global gong, New South Wales, Australia. Development, Washington, DC. IATA (International Air Transport Association). 2015. Devarajan, Shantayanan. 2013. “Africa’s Statistical Trag- “E-Freight Fundamentals.” IATA presentation. http:// edy.” Review of Income and Wealth 59 (S1): S9–S15. www.iata.org/whatwedo/cargo/e/efreight/Pages Dutton, William. 2015. “Multi-Stakeholder Internet /index.aspx. Governance.” Background paper for the World Devel- IEG (Independent Evaluation Group). 2011. “Project opment Report 2016, World Bank, Washington, DC. Performance Assessment Report: India—Cluster Easterly, William. 2006. The White Man’s Burden: Why the Assessment of Forestry and Watershed Development West’s Eff orts to Aid the Rest Have Done So Much Ill and So Activities.” World Bank, Washington, DC. Little Good. New York: Penguin Books. ITU (International Telecommunication Union). 2015. ————. 2014. The Tyranny of Experts: How the Fight against “Facts & Figures.” http://www.itu.int/en/ITU-D Global Poverty Suppressed Individual Rights. New York: /Statistics/Documents/facts/ICTFactsFigures2015.pdf. Perseus Books Group. Judy, Henry, and David Satola. 2014. “Business Interests EC (European Commission). Various years. Eurostat under Attack in Cyberspace: Is International Regula- (database). European Commission, Brussels, http:// tion the Right Response?” Business Law Today. http:// ec.europa.eu/eurostat/data/database. www.americanbar.org/publications/blt/2011/12/02 ————. 2014. “The Future of the Multistakeholder _judy.html. Model.” European Commission, Brussels. Kummer, Markus. 2013. “Multistakeholder Cooper- ————. 2015. “Digital Single Market Factsheet.” http:// ation: Refl ections on the Emergence of a New ec.europa.eu/priorities/digital-single-market/docs Phraseology in International Cooperation.” Inter- /dsm-factsheet_en.pdf. net Society Public Policy (blog). http://www.internet Espey, Jessica. 2015. Data for Development: A Needs society.org/blog/2013/05/multistakeholder Assessment for SDG Monitoring and Statistical -cooperation-reflections-emergence-new Capacity Development. Paris: Sustainable Devel- -phraseology-international. opment Solutions Network. http://unsdsn.org Kurbalija, Jovan. 2015. An Introduction to Internet Gover- /resources/publications/a-needs-assessment-for-sdg nance. 6th ed. Geneva: DiploFoundation. -monitoring-and-statistical-capacity-development/. Lobell, David B., Kenneth G. Cassman, and Christopher Gandhi, Rikin, Rajesh Veeraraghavan, Kentaro Toy- B. Field. 2009. “Crop Yield Gaps: Their Importance, ama, and Vanaja Ramprasad. 2009. “Digital Green: Magnitudes, and Causes.” Annual Review of Environ- Participatory Video and Mediated Instruction for ment and Resources 34: 179–204. 320 WORLD DEVELOPMENT REPORT 2016

Mandel, Michael. 2013. “Data, Trade and Growth.” Work- Satola, David, and Michael Kelly. 2014. “Internet Gov- ing paper, Progressive Policy Institute, Washington, ernance Overview: How Snowden Didn’t Really DC. Change Anything After All.” Business Law Today (No- Matheus, Ricardo, and Manuella Maia Ribeiro. 2014. vember). http://www.americanbar.org/publications “Case Study: Open Government Data in Rio de /blt/2014/11/01_satola.html. Janeiro City.” http://www.opendataresearch.org UN (United Nations). 2013. A New Global Partnership: /projects/2013/cities. Eradicate Poverty and Transform Economies through Nakasone, Eduardo. 2013. “The Role of Price Information Sustainable Development. New York: United Nations in Agricultural Markets: Experimental Evidence from Publications. Rural Peru.” Paper presented at the Agricultural & ————. 2014. Open Working Group Proposal for Sustain- Applied Economics Association’s 2013 AAEA & CAES able Development Goals. New York: United Nations Joint Annual Meeting, Washington, DC, August 4–6. Publications. Nakasone, Eduardo, Maximo Torero, and Bart Minten. UNESCO (UN Educational, Scientifi c and Cultural Orga- 2014. “The Power of Information: The ICT Revolu- nization). 2015. “Keystones to Foster Inclusive Knowl- tion in Agricultural Development.” Annual Review of edge Societies: Access to Information and Knowledge, Resource Economics 6 (1): 533–50. Freedom of Expression, Privacy, and Ethics on a Global National Board of Trade. 2015. “No Transfer, No Pro- Internet.” http://www.unesco.org/new/fileadmin duction: A Report on Cross-Border Data Transfers, /MULTIMEDIA/HQ/CI/CI/pdf/internet_draft_study Global Value Chains, and the Production of Goods.” .pdf. Kommerskollegium (Swedish Board of Trade). http:// USITC (U.S. International Trade Commission). 2014. www.kommers.se/In-English/Publications/2015 Digital Trade in the U.S. and the Global Economies, Part 2. /No-Transfer-No-Production/. Washington, DC: USITC. Nicholson, Jessica, and Ryan Noonan. 2014. “Digital van Pottelsberghe de la Potterie, Bruno, and Malwina Economy and Cross-Border Trade: The Value of Dig- Mejer. 2010. “The London Agreement and the Cost itally Deliverable Services.” ESA Issue Brief 01-14, U.S. of Patenting in Europe.” European Journal of Law and Department of Commerce, Economic and Statistics Economics 29: 211–37. Administration. Vasilaky, Kathryn, Kentaro Toyama, Tushi Baul, Mohak OECD (Organisation for Economic Co-operation and Mangal, and Urmi Bhattacharya. 2015. “Learning Development). 2014a. Measuring and Managing Results Digitally: Evaluating the Impact of Farmer Training in Development Cooperation: A Review of Challenges and via Mediated Videos.” Columbia University Earth Practices among DAC Members and Observers. Paris: Institute, New York. OECD. Wang, Wei, David Rothschild, Sharad Goel, and Andrew ————. 2014b. Principles for Internet Policy Making. Paris: Gelman. Forthcoming. “Forecasting Elections with OECD. http://www.oecd.org/sti/ieconomy/oecd-prin Non-representative Polls.” International Journal of ciples-for-internet-policy-making.pdf. Forecasting. Ramalingam, Ben. 2013. Aid on the Edge of Chaos. London: Whittle, Dennis. 2013. “How Feedback Loops Can Oxford University Press. Improve Aid (and Maybe Governance).” Center for Ranger, Pippa, Julia Chandler, and Beatrice Arscott. 2015. Global Development, Washington, DC. DFID Review of Digital in Development Programmes. Lon- World Bank. 2015. Asia: Urban Transport and ICT Capacity don: Department for International Development. Building. Washington, DC: World Bank. Rogers, David P., and Vladimir V. Tsirkunov. 2013. Weather World Summit on the Information Society. 2005. “Tunis and Climate Resilience: Eff ective Preparedness through Agenda for the Information Society (paragraph 33).” National Meteorological and Hydrological Services. Wash- http://www.itu.int/net/wsis/docs2/tunis/off/6rev1 ington, DC: World Bank. .html. Sadowsky, George. 2014. Accelerating Development Using Wu, Tim. 2010. Is Internet Exceptionalism Dead? The Next the Web: Empowering Poor and Marginalized Populations. Digital Decade: Essays on the Future of the Internet. Wash- Washington, DC: World Wide Web Foundation. ington, DC: Tech Freedom.

322 WORLD DEVELOPMENT REPORT 2016

SECTOR FOCUS 6

Environmental management

The internet is transforming how policy makers and The benefi ts of the internet citizens meet the challenge of environmental man- agement. New technologies have greatly improved in managing air quality monitoring of environmental quality and helped free Enabling comprehensive monitoring data from the obscurity of closed government offi ces, Countries now have new tools to monitor the con- instead making them instantly accessible. This is hav- centration of ambient pollutants, the fi rst critical ing profound effects. Not only are these technologies step in devising an air quality management plan. improving the ability of public authorities to monitor Ground-level monitoring is common in high-income pollution, but they are also pushing the boundaries for countries and some middle-income countries such civic engagement. This is happening in many areas of as China, but there are large parts of the globe where environmental policy making, including forest mon- monitoring of PM is rare. Barriers include the high itoring, water quality assessments, identifi cation of 2.5 cost of installing and maintaining monitoring net- natural hazard risk, and air quality management—the works and a shortage of technicians and experts with emphasis of this sector focus. specialized skills. Air pollution is a growing threat to health the world New satellite-based sensors measure pollution over. Exposure to particulates with a diameter of less concentrations from space.2 In combination with than 2.5 micrometers (PM )—which are capable of 2.5 models of particulate movement through the atmo- penetrating deep into the lungs—increased in all sphere, satellite data supplement information pro- regions between 1998 and 2012, with the exception of vided by sparse ground-level monitors to produce North America and Europe. Health risks include heart estimates of pollution concentrations over larger disease, stroke, lung cancer, and respiratory infec- areas (map F6.1). In the United States, public agencies tions. Exposure to outdoor PM pollution was the 2.5 are investigating using satellite-based measurements ninth leading cause of death and disability worldwide to enhance air quality forecasts for areas where gaps in 2010, accounting for 3.2 million deaths (6 percent exist in the monitoring network. Remotely sensed of global mortality). There is also growing evidence data will prove especially helpful for regions where that PM pollution could harm cognitive functions 2.5 there is currently very little on-the-ground monitor- and contribute to diseases such as Alzheimer’s and ing, such as in Africa. Parkinson’s. The economic costs of deaths and debili- Personal mobile technologies could also improve tating illnesses related to PM pollution, measured by pollution monitoring. One California startup has reduced labor productivity, amount to 4 percent of created a wearable monitor primarily intended 1 GDP for some developing countries. for developing countries such as China, India, and Mexico. The keychain-size device, called Clarity and This sector focus was contributed by Chris Sall and Urvashi costing between US$50 and US$75, uses an optical Narain. sensor to measure concentrations of PM2.5, nitrogen ENVIRONMENTAL MANAGEMENT 323

Map F6.1 Satellite estimates of average PM2.5 concentrations provide global coverage, 2010

Oslo Helsinkiliki Moscow LondLoLdLondon Berlini PariPariss BudaBudapestdaapestpestst

Chicago Romeome Istanbulstanbu MadMadridd idd Newew York BeijingBeBeiji SeoulSeou Washington,W D.C. AAtheAthens Tokyo Los Angeles Casablanca CaiCairCairoo Neweww OsakaO DeDelhilhii Shanghaih MiamiMi i DbDDubaib i Riyadhh Mexicoo CCityity MbiMumbai BangkokBangkok ManilaMil CaracasCarac

Bogotá LagoLagos Kuala Lumpurmpupur SingSingaporegapore Nairobiirobi

Jakarta

Mean annual PM2.5 micrograms/cubic meter 0–10 São Paulo Rioio de 11–15 Janeiro JohaJohannesburge bur 16–25 Buenosueno SydSydney Santiago AireAires Cape 26–35 Town 36–50 MMelbourneelbourne Above 50

IBRD 42014 Source: Data from Brauer and others 2012 for the Global Burden of Disease study (Lim and others 2012).

Note: Includes only anthropogenic emissions; concentrations of dust and sea salt are not shown. PM2.5 = particulates with a diameter of less than 2.5 micrometers.

oxides, ammonia, and volatile organic compounds. of occasionally sampling gases in smokestacks. Such Data will be gathered and analyzed in the cloud so data support schemes and other that users can view real-time, crowdsourced maps of market-based policies to reduce pollution. An example pollution in their neighborhood or city. is the successful sulfur dioxide (SO2) permit trading

system in the United States, which nearly halved SO2 Improving policy design emissions within a year and saved industries at least Singapore is one of the fi rst cities to pioneer the use US$255 million in annual compliance costs. In India, of “big data” to help fi ght congestion and thus reduce regulators in the states of Gujarat, Maharashtra, and pollution from idling vehicles. The city is testing Tamil Nadu are installing low-cost sensor technol- new satellite positioning technologies to improve ogies at industrial facilities for a new pilot trading its existing Electronic Road Pricing (ERP) scheme. scheme for PM pollution. The system tracks the exact locations of vehicles and measures their distance traveled on congested road- Empowering citizens ways. By pinpointing traffi c congestion, ERP tariffs Publicly accessible data on pollution concentration can be automatically adjusted to individual driving levels enable citizens to take preventive measures to behavior, providing drivers with incentives to fi nd reduce their personal risk of exposure and pressure less congested routes. governments to enforce existing emission standards. The AIRNow program of the U.S. Environmental Empowering regulators Protection Agency (U.S. EPA) provides hourly, real- New, lower-cost sensors and information technolo- time maps and forecasts of air quality. The program gies are reducing the cost to regulators of enforcing illustrates recent changes in data policies. In the early limits on emissions from sources such as factories, 1990s, local, state, and federal agencies collected air farms, and energy producers. Sensors can be placed quality data from monitoring stations and sent the within facilities, at fence lines, or in downwind data to the U.S. EPA for processing. Concerns over data communities to continuously monitor emissions quality, potential misuse of data, and limited technical and transmit data via the internet to regulators. This capabilities meant that data were released with long makes it possible to gauge actual pollution loads delays. Today, AIRNow gathers real-time data from (such as kilograms of PM emitted) over time, instead 1,300 ozone and 300 PM monitoring sites, performs 324 WORLD DEVELOPMENT REPORT 2016

automated data quality checks, and provides hourly environmental standards are being violated. There are maps and forecasts of local, regional, and national already some signs of success. On several occasions air quality. Cloud-based computing and mapping, in 2014, authorities in the city of Yantai, Shandong combined with new mobile platforms, make local air Province, publicly responded to social media posts that quality reports and forecasts accessible to everyone, enterprises in the area were violating emissions limits everywhere. And open data standards allow devel- and resolved the complaints within a few days. opers and public interest groups to pull information Smart public policy can help make the most of from AIRNow into their own web services. new technologies for monitoring and reducing Public information disclosure programs that mon- pollution. Apart from stronger requirements for itor and publicize the environmental performance of public disclosure of monitoring data, governments fi rms empower communities to pressure otherwise play an important role in setting standards for data weak regulatory institutions to enforce environ- collection, reporting, and sharing; offering guidance mental standards. These programs now rely on the on best practices for use of technologies; and rating internet to improve the free fl ow of information and the performance of devices used by consumers. To reduce the cost of data collection and dissemination. keep up with the pace of technological change, it is Online registries that compile information on pollut- vital for governments to regularly review policies and ant releases by industrial facilities and other sources standards in recognition of new data services. Such have been set up in at least 20 nations, including in policies for encouraging the innovative use of infor- Japan, the Republic of Korea, Mexico, the United mation technology may form an important part of a States, and the countries of the European Union. Dis- broader air quality management strategy. closure programs that rate the environmental perfor- mance of fi rms have also been implemented in about 60 countries. Ratings programs in countries such Notes as India and Indonesia have had a signifi cant, mea- 1. Global Commission on the Economy and Climate surable impact in reducing pollution by the worst- 2014. performing plants. 2. van Donkelaar and others 2015. China offers an example of how information tech- 3. Li 2011. nologies can leverage mandatory public disclosure programs and expand citizen engagement.3 Public References disclosure of environmental monitoring data in China has expanded rapidly in the last few years in what Brauer, M., M. Amman, R. T. Burnett, A. Cohen, F. Den- could be characterized as a virtuous cycle made pos- tener, M. Ezzati, S. B. Henderson, M. Krzyzanowski, sible by information technologies. The attention gen- R. V. Martin, R. Van Dingenen, A. van Donkelaar, erated by Tweets of air quality data from unoffi cial and G. D. Thurston. 2012. “Exposure Assessment for monitoring sites in Beijing and other cities prompted Estimation of the Global Burden of Disease Attribut- able to Outdoor Air Pollution.” Environmental Science & programs by public agencies to build offi cial moni- Technology 46. doi: 10.1021/es2025752. toring systems and make the data publicly available Global Commission on the Economy and Climate. 2014. in real time. In 2013, the Ministry of Environmental Better Growth, Better Climate. New Climate Economy. Protection issued rules mandating the disclosure of http://2014.newclimateeconomy.report. real-time pollution data by key enterprises and local Li, Wanxin X. 2011. “Self-Motivated versus Forced Dis- authorities. closure of Environmental Information in China: The next year, the Institute of Public and Environ- A Comparative Case Study of the Pilot Disclosure mental Affairs (IPE) in Beijing, a public advocacy group, Programmes.” China Quarterly 206. doi: 10.1017 released a free mobile app to put these data directly in /S0305741011000294. the hands of the public. Users of the Pollution Map Lim, S. S., and others. 2012. “A Comparative Risk Assess- (wuran ditu) app can see current emission levels of ment of Burden of Disease and Injury Attributable to 67 Risk Factors and Risk Factor Clusters in 21 Regions, local factories and other entities and ascertain whether 1990−2010: A Systematic Analysis for the Global Bur- the exceed legal limits. They can share this den of Disease Study 2010.” Lancet 380: 2224−60. information on popular social media apps and submit van Donkelaar, A., R. V. Martin, M. Brauer, and B. L. Boys. reports and photos of companies that are violating 2015. “Use of Satellite Observations for Long-Term emission limits. According to IPE’s founder, Ma Jun, Exposure Assessment of Global Concentrations of the app gives citizens the hard data they need to put Fine Particulate Matter.” Environmental Health Perspec- pressure on enterprises and local authorities when tives 123 (2): 135–43. doi: 10.1289/ehp.1408646.

326 WORLD DEVELOPMENT REPORT 2016

ENABLING DIGITAL DEVELOPMENT Six digital technologies to watch

This Report looks at what countries can do to get perform current 4G networks by providing data at a more out of the internet and other digital technolo- speed several hundred gigabits per second (Gbit/s). gies. Although it looks ahead to a world in which the In 2015, researchers at the University of Surrey’s 5G internet is universally available and affordable, the Innovation Centre (5GIC) managed one terabit per analysis largely takes the technology for granted. But second (Tbit/s) during their speed tests—many thou- technological change is continuous and frequently sands of times faster than current data connections.4 disruptive. This spotlight examines a range of tech- Accommodating 5G will require using parts of the nologies, identifi ed in the technology forecasting lit- spectrum that have not previously been considered erature that promise to be far-reaching in their impact commercially useful, in particular above 3 gigahertz on development. 1 They are at different stages of the (GHz), and in the millimeter band that stretches from “hype” cycle (fi gure S6.1). Like the internet, they are 30–300 GHz. It will also require new kinds of anten- 5 likely to be encountered fi rst in the developed world, nae. Developments in smartphones have increased but will spread quickly to the developing world, where capabilities about 25 percent per year in the past their impact may be even greater.2 fi ve years. Even though mobile itself is hardly a new technology, it is its combination with the internet SPOTLIGHT 6 that makes it a disruptive force and one of the tech- Fifth-generation (5G) nologies with potentially the greatest impact for the mobile phones developing world. Developing countries will need to closely follow developments taking place in 5G and The development of cellular mobile phones from the start preparing for its eventual rollout. earliest incarnation of commercial service in the late 1970s has proceeded in generations. First-generation (1G) analog cellphones were replaced with 2G digital Artifi cial intelligence phones, starting in Finland in 1991, while 3G phones, Defi nitions of artifi cial intelligence (AI) differ which provide faster internet access, were launched at widely, but generally refer to computer systems that the opening match of the 2002 FIFA World Cup in the can perform tasks that normally require human intel- Republic of Korea. By the end of June 2015, there were ligence—including visual and speech recognition, 2.33 billion 3G mobile subscriptions worldwide, and a decision making, and language translation. Faster further 757 million subscribers to data-optimized 4G, computing, “big data,” and better algorithms have 3 or Long Term Evolution (LTE), technology. helped propel recent breakthroughs in AI.6 Algo- 5G wireless networks are the next generation of rithms are now better able to recognize language and mobile networks. 5G networks are expected to out- images, for example, thanks to the availability of huge Contributed by Wajeeha Ahmad, Hallie Applebaum, Naomi torrents of data from interconnected phones, tablets, Halewood, Arturo Muente-Kunigami, Marcela Sabino, Randeep and computers. Firms like Narrative Science have Sudan, and Darshan Yadunath. automated the writing of fi nancial reports using AI. jobs suchaslegal analysts,fi humans andmachines,aswell as alossoftraditional ing innewopportunitiesforcollaboration between risk management,andcompliance. banking andinsurance,inareas likecustomerservice, farming, andoptimizedresource application;andin nostics; inagriculture,withcropplanning,precision with personalizedlearning;inhealth,deepdiag- benefi virtually everysectorrelevanttodevelopment.The can provideimportantinsightsandgeneratevaluein ognizing thepotentialrisksofAI,suchtechnologies Gates haveweighedinonthedangersofAI.Whilerec- naries likeElonMusk,StephenHawking,andBill algorithm toitsboardofdirectors! Deep KnowledgeVentures,hasevenappointedan venture capitalfundbasedinHongKongSAR,China, increasingly usedinpersonalandbusinesscontexts.A bilities likeApple’sSiriandMicrosoft’sCortanaare advice. Virtualassistantswithvoicerecognitioncapa- assistance todoctors,providingcustomizedmedical IBM’s WatsoncomputerusesAItoprovidediagnostic Source: S6.1 Figure potentially an“existentialrisk”formankind. book onsuperintelligence,whichconsidersAItobe humanity itself.AnexampleisNickBostrom’s2014 intelligence, andbecomingathreattothefutureof cerns aboutmachineintelligenceovertakinghuman Rising expectations Advances inAIwillprovetobe disruptive,result- Rapid advancesinAIhavealsoresultedcon- WDR 2016 team, adapted from Gartner 2015. Gartner from adapted team, 2016 WDR ts ofAIare beginningtobeseenineducation, 5G mobile 5G Innovation trigger intelligence Artificial Gartner “hype” cycle applied to selected digital technologies digital toselected applied cycle “hype” Gartner Robotics Autonomous vehicles Peak of inflated expectations nancial andsports of things Internet 7

3D printing Consumer Consumer disillusionment 8 Lumi- Trough of Trough Time nosticians, andfi reporters, onlinemarketers,anesthesiologists,diag- agents. customer requestspreviouslyhandledbycallcenter virtual assistantcapableofhandlingmanyroutine ish BankBBVA,forexample,hasintroducedLola,a that cansubstituteforhumanworkers.TheSpan- sophisticated naturallanguageprocessingsystems oping countriescouldbeaffectedbyincreasingly number ofcallcenterspreviouslyoffshoredtodevel- sold forpersonalanddomesticuse. to 1.5million,witharound4.7million servicerobots operational industrialrobotsatthe endof2014wasup and sensingcapabilities.Theworldwidestockof through theircomputingpower,precision,strength, tion, andexoskeletons).Robotscanprovidebenefi elderly, agriculture,fl and servicerobots(logistics,medicine,assistingthe tive, chemical,rubber,plastics,andfoodindustries) generally dividedintoindustrialrobots(automo- systems thatcanautomaticallyhandletasks,be Robotics, whichreferstomachinesormechanical Robotics intelligence. They aremorecompact,adaptable, and ing moreadvancedandgaining senses,dexterity,and cally difficult ordangerous jobs,theyarenowbecom- While robotshavebeenprimarily usedinphysi- 10 enlightenment Slope of 3D printing Business SIX DIGITAL TECHNOLOGIES TO WATCH nancialanalysts. oor-cleaning, civilconstruc- productivity Plateau of Plateau 9 11 Similarly,alarge

ts

SPOTLIGHT 6 327 328 WORLD DEVELOPMENT REPORT 2016

intelligent, have manipulation capabilities, and are and even providing internet service in remote areas. increasingly able to work alongside humans. Eventu- Rwanda plans to be home to the world’s fi rst drone ally they may displace or augment humans, primarily airport, or “droneport,” to facilitate the delivery of in low-skilled areas like manufacturing, cleaning, and medical and emergency supplies, quickly and cost- maintenance, but also in high-tech fi elds, like surgery effectively, across geographical barriers.16 and prosthetics.12 A number of leading information The introduction of AVs is likely to be gradual, technology (IT) companies have made investments in with many cars and planes already incorporating ele- robotics and are driving development of smart robots. ments of assistive technology. The impact on jobs will Amazon acquired Kiva Systems and is using Kiva ultimately be a function of price (self-driving cars are robots for order fulfi llment. Google acquired Boston currently prohibitively expensive), legislation (will Dynamics and several other robotics companies. they always require a human with manual override?), Demand for industrial robotics is driven by the and time. desire to reduce labor costs and by the need for accu- racy in undertaking repetitive processes. Robots are Internet of things not paid, they do not get sick, and they can work as long as there is power. They can also take on danger- The “internet of things” (IoT) refers to the intercon- ous or risky tasks, such as detecting landmines. The nection of objects to internet infrastructure through take-up of robots is expected to rise as a function of embedded computing devices, such as radio fre- their falling costs and growing intelligence, including quency identifi cation (RFID) chips and sensors. IoT in developing countries. China is the world’s biggest products can be classifi ed into fi ve broad categories: importer of industrial robots. Guangdong province wearable devices, smart homes, smart cities, envi- recently announced a program to automate 80 per- ronmental sensors, and business applications.17 Cisco cent of its factories by 2020, by substituting human estimates that by 2020, 50 billion devices and objects labor with robots.13 Robots have been deployed in the will be connected to the internet. Democratic Republic of Congo, in Kinshasa, to man- IoT is quickly redefi ning service delivery and age traffi c.14 unlocking opportunities in multiple areas. Smart fi tness sensors and trackers are transforming health Autonomous vehicles care and improving personal fi tness and health. Embedded sensors accurately relay moisture, air and Autonomous vehicles (AV), or self-driving cars, attract water pollution levels, and resource levels, allowing major research spending from car companies as well for closer monitoring of environmental problems. as internet fi rms. Their proponents argue that they Factories and supply chains use smart sensors to

SPOTLIGHT 6 will reduce road accidents (for instance, through improve the effi ciency of manufacturing and dis- lane-keeping systems, auto-parking, and cruise tribution of goods. Globally, there has been a rise in control), ease congestion, reduce fuel consumption, spaces where people can gather to build and learn improve the mobility of the elderly and disabled, and with electronics, software, and digital fabrication. free up commuting time for other tasks.15 But they Known as makerspaces, these spaces have democra- also threaten the jobs of millions of people currently tized access to tools and empowered participants to employed as drivers. They also raise complex legal build and learn on their own. issues, including for liability insurance, and onboard One of the key applications of IoT is in combating networked computers would be at risk of hacking. climate change and its effects. Farms in developing The European project SARTRE is piloting the concept countries can use intelligent sensors to monitor soil of “autonomous car platoons,” which allows multiple conditions and guide autonomous irrigation systems. vehicles to drive autonomously within meters of one Smart traffi c synchronization systems in cities save another at highway speeds, guided by a professional on travel time and fuel consumption. Countries such pilot vehicle. This approach is expected to reduce as Singapore are deploying smart networks that use fuel consumption and emissions by up to 20 percent, global positioning systems (GPS), sensor informa- improve road safety, and reduce traffi c congestion. tion from monitoring cameras, and other sources to Drones (unmanned aerial vehicles, and a special- sense population movement, ease traffi c congestion, ized type of AV) are growing in popularity as prices and re-route traffi c in the case of special events and fall. They have many potential applications, including emergencies. police work, assisting the disabled, home delivery, Some experts believe that the IoT will mark a new farming, entertainment, safety, wildlife conservation, stage of the internet’s development, since it has the SIX DIGITAL TECHNOLOGIES TO WATCH 329

Box S6.1 Using cellphones for medical diagnosis, thanks to 3D printing

Doctors in developing countries often lack the necessary and diagnose cancer, nervous system disorders such as equipment to diagnose diseases, but they do have smart- Alzheimer’s disease, and drug resistance to infectious phones. Dr. Aydogan Ozcan, an engineering professor diseases.a Similarly, a 15-year-old girl from Cupertino, at the University of California Los Angeles (UCLA), has California, has designed a spirometer, a device that can created software and hardware that can convert smart- diagnose respiratory illnesses such as chronic obstructive phones into microscopes and diagnostic machines. He pulmonary disease and asthma. She designed this tool uses 3D printing to create a fl uorescent microscope device using open hardware and 3D printing, at a fraction of the that can be attached to the phone. The device includes an cost of a commercial spirometer.b app that connects the smartphone to a server at UCLA and transmits raw images of DNA molecules to detect a. http://newsroom.ucla.edu/releases/ucla-engineers-fi rst-to-detect-and-measure-individual-dna-molecules-using-smartphone-microscope. b. https://cogito.cty.jhu.edu/43295/maya-varma-fi nal-device-and-app-design/.

potential to revolutionize the way people live, work, The revolutionary aspect of 3D printing lies in its interact, and learn. However, there are still signifi cant digital nature: physical objects become digital infor- barriers to full commercialization of IoT, such as the mation that can be remixed, reformulated, improved, fragmented landscape of standardization, which is and shared. However, desktop 3D printers are still rel- preventing interoperability; and the relatively high atively expensive, and use between 50 and 100 times cost of embedded devices. The maker movement more energy per unit of weight than conventional offers a possible solution for the standardization chal- injection molding. Other concerns include potentially lenge, empowering individuals to adjust devices to fi t negative uses such as gun and drug printing, and the local context. There are also signifi cant privacy infringement of intellectual property.19 However, as and security concerns. As more devices are connected the performance improves and the cost of the printers to networks, hacking unsecure devices could have and their inputs decline, 3D printers could be adopted repercussions that far exceed the damage posed by more widely (box S6.1). The most relevant immediate conventional security threats. progress may be in industrial printers for highly cus- tomizable objects that are relatively expensive and SPOTLIGHT 6 require replicable results. This is most suited to the 3D printing construction industry, where 3D printed buildings can provide lower-cost housing solutions. 3D printing, a process whereby machines can print 3D printing is expanding in developing countries. objects from digital fi les or scans, consists of adding In Uganda, for example, the technology is used to successive layers of material to make a three- create 3D-printed prosthetic limbs. The startup Proto- dimensional (3D) object. This technology has transfor- print, a social enterprise in Pune, India, helps waste mational potential for manufacturing, since it enables pickers turn the waste plastics they collect into 3D users to create smaller batches of highly customizable printer fi lament. iLab Haiti has used 3D printing to products at declining prices. In recent years, 3D print- design basic medical supplies for local clinics (such as ing has advanced to printing of body parts (titanium umbilical clamps, fi nger splints, and casts) and pro- jaws, spines), exoskeletons, rocket parts, and even duce them using on-demand manufacturing.20 food.18 As prices have fallen, consumer-oriented devices have appeared on the market in recent years, allowing individuals to make three-dimensional solid objects Notes locally, often using a computer-assisted design (CAD) 1. Sources used in the analysis of technology trends fi le that can be downloaded from the internet. The include Gens and IDC Predictions 2015 Team; “ink” used in the printer is usually plastic, but other Deloitte Consulting 2015; McKinsey Global Institute materials—including epoxy resins, silver, titanium, 2013; and Meeker 2015. steel, and wax—are also available. 2. Muente-Kunigami and others 2015. 330 WORLD DEVELOPMENT REPORT 2016

3. GSMA databases, https://gsmaintelligence.com/. Bugmann, Guido, Mel Siegel, and Rachel Burcin. 2011. “A 4. http://www.bbc.com/news/technology-31622297. Role for Robotics in Sustainable Development?” IEEE 5. One gigahertz represents 1 billion cycles per second; (Institute of Electrical and Electronics Engineers) see http://spectrum.ieee.org/telecom/wireless/smart Africon 2011, September 13–15, Livingstone, Zambia. -antennas-could-open-up-new-spectrum-for-5g. http://www.tech.plym.ac.uk/soc/staff/GuidBugm 6. http://www.wired.com/2014/10/future-of-artifi cial /pub/SustainableRobotics_Africon_2011.pdf. -intelligence/. Deloitte Consulting. 2015. “Tech Trends 2015: The Fusion 7. http://www.bbc.com/news/technology-27426942. of Business and IT.” http://www2.deloitte.com/global 8. Bostrom 2014. /en/pages/technology/articles/tech-trends.html. 9. http://fortune.com/2015/02/25/5-jobs-that-robots Dotz, Dara A. 2015. “A Pilot of Printing 3D Medical -already-are-taking/. Devices in Haiti.” In Technologies for Development: 10. http://cloudlendinginc.com/lending-blog/will-banks What Is Essential? edited by Silvia Hostetller, Eileen -survive-the-era-of-digital-disruption/. Hazboun, and Jean-Claude Bolay, chapter 4. Springer 11. International Federation of Robotics 2015. International Publishing. 12. Bugmann, Siegel, and Burcin 2011. Gartner. 2015. “Hype Cycle for Emerging Technologies.” 13. http://www.china-briefing.com/news/2015/04/22 https://www.gartner.com/doc/3100227. /automation-chinas-labor-force.html. Gens, Frank, and IDC Predictions 2015 Team. 2015. 14. http://www.theguardian.com/travel/2015/mar/05 “IDC Predictions 2015: Accelerating Innovation and /robocops-being-used-as-traffic-police-in Growth on the Third Platform.” http://www.idc.com -democratic-republic-of-congo. /getdoc.jsp?containerId=252700. 15. http://www.economist.com/blogs/economist International Federation of Robotics. 2015. “World - -explains/2013/04/economist-explains-how-self ics: Industrial Robotics 2015 and Service Robotics -driving-car-works-driverless. 2015.” http://www.ifr.org/service-robots/statistics/. 16. http://www.theguardian.com/technology/2015/sep McKinsey Global Institute. 2013. “Disruptive Technolo- /30/rwanda-chosen-for-worlds-first-drone-port-to gies: Advances That Will Transform Life, Business, -deliver-medical-supplies. and the Global Economy.” McKinsey & Company. 17. Perera, Liu, and Jayawardena 2015. http://www.mckinsey.com/insights/business 18. http://www.digitaltrends.com/cool-tech/3d-food _technology/disruptive_technologies. -printers-how-they-could-change-what-you-eat/. Meeker, Mary. 2015. “Internet Trends 2015.” http://www 19. http://www.techrepublic.com/article/the-dark-side .kpcb.com/internet-trends. -of-3d-printing-10-things-to-watch/, accessed Muente-Kunigami, Arturo, Naomi Halewood, Marcel March 2014. Sabino, and Hallie Applebaum. 2015. “The Future 20. Dotz 2015. of Digital Technologies.” Background paper for the World Development Report 2016. World Bank, Washing- ton, DC.

SPOTLIGHT 6 References Perera, Charith, Chi Harold Liu, and Srimal Jayawardena. Bostrom, Nick. 2014. Superintelligence: Paths, Dangers, Strat- 2015. “The Emerging Internet of Things Marketplace egies. Oxford, U.K.: Oxford University Press. from an Industrial Perspective: A Survey.” IEEE Trans- actions on Emerging Topics in Computing. doi: 10.1109 /TETC.2015.2390034.

ECO-AUDIT Environmental Benefi ts Statement Th e World Bank Group is committed to reducing its environmental footprint. In support of this commitment, the Publishing and Knowledge Division leverages electronic publishing options and print-on-demand technology, which is located in regional hubs worldwide. Together, these initiatives enable print runs to be lowered and shipping distances decreased, resulting in reduced paper consumption, chemical use, greenhouse gas emissions, and waste. Th e Publishing and Knowledge Division follows the recommended stan- dards for paper use set by the Green Press Initiative. Th e majority of our books are printed on Forest Stewardship Council (FSC)–certifi ed paper, with nearly all containing 50–100 percent recycled content. Th e recycled berfi in our book paper is either unbleached or bleached using totally chlorine free (TCF), processed chlorine free (PCF), or enhanced elemental chlorine free (EECF) processes. More information about the Bank’s environmental can be found at http://crinfo.worldbank.org/wbcrinfo/node/4.