Poverty and Shared Prosperity 2016: Taking on Inequality

Home , Economic stability, Poverty in North Korea, Poverty in Romania, Prosperity

© 2016 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: 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 findings, in- terpretations, and conclusions expressed in this work do not necessarily reflect 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 information 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. Note: The term “country,” used interchangeably with “economy,” does not imply political independence, but refers to any territory for which authorities report separate social or economic statistics. 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 specifically reserved. Rights and Permissions

This work is available under the Creative Commons 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. Poverty and Shared Prosperity 2016: Taking on Inequality. Washington, DC: World Bank. doi:10.1596/978-1-4648-0958-3. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create 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, ﬁgures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights @worldbank.org. ISBN (paper): 978-1-4648-0958-3 ISBN (electronic): 978-1-4648-0979-8 DOI: 10.1596/978-1-4648-0958-3 Cover design: Patricia Hord Graphik Design Library of Congress Cataloging-in-Publication Data has been applied for. Contents

Foreword ix Acknowledgments xi About the Core Team and the Contributors xiii Abbreviations xvii Overview 1 Notes 19 References 20 1 Poverty and Shared Prosperity: Setting the Stage 23 Concepts, measurement, and data 23 A special focus in 2016: taking on inequality 29 Notes 32 References 33 2 Global Poverty 35 Monitoring global poverty 35 A proﬁle of the poor in the developing world 42 Annex 2A Historical global and regional poverty estimates 46 Annex 2B Technical note: global poverty measurement using 2013 data 47 Notes 50 References 51 3 Shared Prosperity 53 Shared prosperity: where we stand 53 The bottom 40 in relative terms: the shared prosperity premium 56 Incomes of the bottom 40 and the top 10: the Palma premium 57 Who are the bottom 40? 57 Is the poverty goal attainable at current levels of growth and shared prosperity? 59 Conclusions: continued progress, but no room for complacency 62

v Annex 3A Shared prosperity estimates based on the latest surveys, by country, circa 2008–13 64 Notes 66 References 67 4 Inequality 69 Inequality matters 69 Separating fact from myth: what is the evidence on inequality? 75 Concluding remarks 87 Annex 4A Data construction 89 Global inequality database 89 Database of within-country Gini indexes 89 Notes 91 References 95 5 Reductions in Inequality: A Country Perspective 101 Introduction 102 Brazil: multiple policies aligned to redress record inequality 103 Cambodia: new earning opportunities emerging from impressive growth 107 Mali: a vulnerable economy favored by the vagaries of agriculture 110 Peru: equalizing investment despite informality and low human capital 114 Tanzania: sharing prosperity in the midst of diversiﬁcation 117 Concluding remarks: learning from country experiences 120 Notes 122 References 124 6 Reductions in Inequality: A Policy Perspective 129 Introduction 130 Early childhood development and nutrition 131 Health care and education 135 Conditional cash transfers 140 Rural infrastructure 145 Taxation 148 Concluding remarks 152 Notes 154 References 159

vi POVERTY AND SHARED PROSPERITY 2016 Boxes 1.1 The Welfare Aggregate: Income 4.2 Perceptions of Inequality in the versus Consumption 25 Middle East and North Africa 73 2.1 Measuring Global Poverty and 4.3 Absolute versus Relative Inequality 78 Purchasing Power Parities 37 4.4 Comparison of Levels and Trends in 2.2 Projecting Poverty Rates 43 Income and Consumption Inequality 79 3.1 The Bottom 40 versus the Poor 59 4.5 Comparing Trends in Inequality: 3.2 Simulating Poverty Trajectories 61 Household Surveys and 4.1 Cross-Country Studies of the Effect Administrative Records 80 of Inequality on Growth 71 6.1 Tax Reform and Fiscal Consolidation 150

Figures O.1 Distribution of the Extreme Poor, 2.3 Regional and World Trends, Number the Nonpoor, the Bottom 40, of the Extreme Poor, 1990–2013 38 and the Top 60, 2013 2 2.4 Regional and World Trends, O.2 World and Regional Trends, Poverty Extreme Poverty Headcount Ratio, Headcount Ratio, 1990–2013 5 1990–2013 39 O.3 Trends in the Global Poverty 2.5 Poverty Headcount Ratios and Headcount Ratio and the Number Number of the Poor, by Country of the Global Poor, 1990–2013 5 Income, 2013 40 O.4 Where Are the Global Poor Living? 2.6 Poverty Headcount Ratios, Top 10 The Global Poor, by Region, 2013 6 Countries, 2013 41 O.5 Profile of the Poor, by Characteristics 2.7 Number of the Poor, Top 10 and Region, 2013 6 Countries, 2013 41 O.6 Age Profile of the Poor, 2013 7 2.8 Trends in the Poverty Gap and the O.7 Shared Prosperity, 83 Countries, Global Headcount Ratio, 1990–2013 42 2008–13 8 2.9 Profile of the Poor, by Characteristics O.8 Boosting Shared Prosperity and and Region, 2013 44 Ending Poverty, 10-Year Scenario, 2.10 Profile of the Extreme and Moderate 2013–30 9 Poor, by Selected Characteristics, O.9 Global Inequality, 1820–2010 9 2013 44 O.10 Global Inequality, 1988–2013 10 2.11 Age Profile of the Poor, 2013 45 O.11 Average Within-Country Inequality, 3.1 Shared Prosperity, 83 Countries, 1988–2013 10 2008–13 55 O.12 Trends in the Average National Gini, 3.2 The Bottom 40, Brazil, India, by Region, 1988–2013 11 and the United States, Circa 2013 58 O.13 The National Income Share of the B3.1.1 Distribution of the Extreme Poor, Richest 1 Percent, Selected Economies 12 the Nonpoor, the Bottom 40, O.14 Available Country Poverty Estimates, and the Top 60, 2013 59 Number, by Region and Year 19 3.3 Income Group Composition, the 1.1 Available Country Poverty Estimates, Bottom 40, Selected Countries, by Region and Year 28 Circa 2013 60 1.2 Availability of Poverty Data, by All 3.4 Boosting Shared Prosperity and Possible 10-Year Periods, 1990–2013 28 Ending Poverty, 2013–30 62 1.3 Double and Triple Dips in Growth, 4.1 Growth of the Bottom 40 versus Selected Economies, 2006–15 30 Growth at the Mean, 2008–13 72 2.1 The Global Poverty Headcount Ratio B4.2.1 Actual versus Anticipated Feelings of and the Number of the Extreme Poor, Well-Being, Middle East and North 1990–2013 36 Africa 74 2.2 Where Are the Global Poor Living? 4.2 The Top 1 Percent Income Share, The Global Poor, by Region, 2013 37 Selected Economies 76

CONTENTS vii 4.3 Global Income Inequality, 5.6 Contributions of Growth and 1820–2010 76 Redistribution Effects to Poverty 4.4 Long-Run Changes in the Gini Reduction, Cambodia, 2008–12 108 Index, Selected Developing 5.7 Trends in the Gini Index, Mali, Countries, 1980–2014 77 2001–10 111 B4.3.1 Comparing Absolute and Relative 5.8 Growth Incidence Curve, Mali, Gains across the Distribution 78 2001–10 112 B4.4.1 Levels of Income and Consumption, 5.9 Contributions of Growth and Gini Indexes, 2013 79 Redistribution Effects to Poverty B4.4.2 Trends in Income and Consumption, Reduction, Mali, 2006–09 112 Gini Indexes, Circa 2008–13 79 5.10 Trends in the Gini Index, Peru, B4.5.1 Comparison of Top Incomes and the 2004–14 115 Gini Index, Brazil, 2006–12 80 5.11 Growth Incidence Curve, Peru, 4.5 Global Inequality, 1988–2013 81 2004–14 115 4.6 Average Within-Country Inequality, 5.12 Contributions of Growth and 1988–2013 83 Redistribution Effects to Poverty 4.7 Trends in the Average Gini, by Reduction, Peru, 2004–14 115 Region, 1988–2013 83 5.13 Trends in the Gini Index, Tanzania, 4.8 The Gini Index, 101 Countries, 2001–12 117 2013 84 5.14 Growth Incidence Curve, Tanzania, 4.9 Distribution of the Gini Index, 2007–12 118 2013 85 5.15 Contributions of Growth and 4.10 Trends in the Within-Country Gini Redistribution Effects to Poverty Index, 1993–2013 86 Reduction, Tanzania, 2007–12 118 5.1 Trends in the Gini Index, Brazil, 6.1 The Mental Development of 1981–2014 104 Stunted Children, Jamaica, 1986–87 132 5.2 Growth Incidence Curve, Brazil, 6.2 Median Coverage, Selected Health 2004–14 104 Care Interventions, by Wealth Quintile, Low- and Middle-Income 5.3 Contributions of Growth and Countries, Circa 2005–13 136 Redistribution Effects to Poverty Reduction, Brazil, 2004–14 105 6.3 Mathematics Scores, by Household Income Level, Selected Countries, 5.4 Trends in the Gini Index, Cambodia, Circa 2007–11 139 2007–13 107 6.4 Simulated Gini Point Reduction 5.5 Growth Incidence Curve, Cambodia, in the Gini Index Attributable 2012–13 108 to CCTs, Circa 2013 141

Tables O.1 World and Regional Poverty 3A.1 Shared Prosperity Estimates, Estimates, 2013 4 Circa 2008–13 64 2.1 World and Regional Poverty 4.1 Countries with an Increasing or Estimates, 2013 36 Decreasing Gini Index and the 2A.1 Historical Trends, World Extreme Average Gini 86 Poverty Estimates, 1990–2013 46 4A.1 Population Coverage of the Data 2A.2 Historical Trends, Regional Poverty Used in the Global Inequality Headcount Ratios, 1990–2013 46 Estimates 89 2A.3 Historical Trends, Number of 4A.2 Population Coverage of the Data Extreme Poor, by Region, Used in the Analysis 90 1990–2013 46 5.1 Annualized per Capita GDP 2B.1 Poverty Estimates Based on China’s Growth and Reductions in Old Survey Methodology, 2013 49 Inequality, Selected Countries 103 3.1 Shared Prosperity, Circa 2008–13 54 5.2 Typology of Poverty and Inequality Levels, Selected Countries, Circa 2013 103 viii POVERTY AND SHARED PROSPERITY 2016 Foreword

The World Bank Group’s goals are clear—we are committed to ending extreme poverty by 2030 and boosting shared prosperity of the bottom 40 percent of populations in every country. If we are to reach our goals, it’s crucial to report on both the progress and the barriers to improving people’s lives. We have created the Poverty and Shared Prosperity annual flagship series to do just that, providing the latest and most accurate statistics and analysis on extreme poverty and shared prosperity. As we work toward the end-poverty goal in 2030, it’s important to remember that the developing world has made unprecedented progress in reducing extreme poverty. Since 1990, nearly 1.1 billion people have lifted themselves out of extreme poverty. In areas ranging from child survival to primary school enrollments, the improvements to people’s lives have advanced with a momentum that few could have imagined when the World Bank was founded more than 70 years ago. But today we face a powerful threat to progress around the world: Inequality. High income inequality is hardly new in human history. But today, inequality is constraining national economies and destabilizing global collaboration in ways that put humanity’s most critical achievements and aspirations at risk. This includes the goal of ending extreme poverty by 2030. That is why this first Poverty and Shared Prosperity report took a deeper look at inequality—making the case for action by explaining the benefits for countries in closing persistent gaps. More equal countries tend to have healthier people and be more economically efficient than highly unequal countries. And countries that invest smartly in reducing inequality today are likely to see more prolonged economic growth than those that don’t. Less inequality can benefit the vast majority of the world’s population. The last part of this report describes the successful strategies that many countries are already using to fight inequality. World Bank Group economists have conducted a comprehensive review of policies that can raise the incomes of the poor, analyzed a vast body of evidence, and singled out some of the policies that are well known to work best. Their results offer policy options that can be relevant for most countries in the world. Whether you’re a government leader, an entrepreneur, an activist, or a frontline service provider, my hope is that this report will inform your decisions and inspire you to make your actions count. Thank you for your work to build a fairer, more equal, and more prosperous future for all.

Jim Yong Kim President, World Bank Group ix

Acknowledgments

Dedicated to the loving memory of Martin Andres Cuesta Lopez.

This report has been prepared by a team led by José Cuesta and Mario Negre and comprising Timm Bönke, Soumya Chattopadhyay, Shaohua Chen, Will Durbin, María Eugenia Genoni, Aparajita Goyal, Christoph Lakner, Terra Lawson-Remer, Maura K. Leary, Renzo Massari, Jose Montes, David Newhouse, Stace Nicholson, Espen Beer Prydz, Maika Schmidt, and Ani Silwal. The work has been carried out under the general supervision of Francisco H. G. Ferreira and Ana Revenga and the guidance of Kaushik Basu and Jan Walliser. World Bank President Jim Yong Kim was an invaluable source of encouragement to the team. Kathleen Beegle, Branko Milanovic´, Ambar Narayan, and Sudhir Shetty served as peer reviewers. Robert Zimmermann edited the document, and Susan Graham was the production editor. The report’s publishing was supervised by Patricia Katayama. Venkat Gopal- akrishnan, Phil Hay, Mary D. Lewis, and Mikael Ello Reventar contributed to dissemination. Additional support was provided by Anna Regina Rillo Bonfield, Pamela Gaye C. Gunio, Estella Malayika, Nelly Obias, and Clara Serraino. The team would like to thank François Bourguignon, Branko Milanovic´, and Matthew Wai-Poi for sharing data. The team would also like to acknowledge the following people for insightful discussions, including Omar Arias, João Pedro Azevedo, Benu Bidani, Andrés Castañeda, Luc Christiansen, Andrew Dabalen, Klaus Deininger, Dung Doan, Martin C. Evans, Deon Filmer, Alan Fuchs, Emanuela Galasso, Xavier Gine, Stephane Hallegatte, Ruth Hill, Leora Klapper, Jose R. Lopez-Calix, Carolina Mejia-Mantilla, Rinku Murgai, Luis F. López-Calva, Minh Cong Nguyen, Pedro Olinto, Maria Beatriz Orlando, Carlos Rodríguez- Castelán, Halsey Rogers, Julie Rozenberg, Carlos Silva-Jauregui, Emmanuel Skoufias, Pablo Suárez-Becerra, Hiroki Uematsu, Adam Wagstaff, and Yukata Yoshino. Comments to previous versions of the report were provided by Pedro Alba, Maurizio Bussolo, Jose Familiar Calderon, Anna Chytla, Amit Dar, Augusto de la Torre, Shanta Devarajan, Marianne Fay, Erik Feyen, Norbert Fiess, Lisa Finneran, Haisan Fu, Ejaz Ghani, Aart Kraay, Cyril Muller, Mamta Murthi, Alberto Ninio, Martin Rama, Sheila Redzepi, Jose G. Reis, Joanna Silva, Philip Schellekens, Richard Scobey, Radwan Shaban, Nikola Spatafora, Hans Timmer, Yvonne Tsikata, Laura Tuck, Jos Verbeek, and Xiaoqing Yu. The report benefited from substantial support from the German Development Institute/ Deutsches Institut für Entwicklungspolitik (DIE).

About the Core Team and the Contributors

The Core Team José Cuesta, co-director of the report, is a World Bank senior economist with a PhD in economics from Oxford University. He is also an affiliated professor at Georgetown Uni- versity’s McCourt School of Public Policy. Cuesta was previously an assistant professor in development economics at the Institute of Social Studies in The Hague. He also worked as a research economist and social sector specialist for the Inter-American De- velopment Bank and as an economist for the United Nations Development Programme in Honduras. His research interests revolve around poverty and conflict economics, specifically the distributive analysis of social policies, intrahousehold allocation, social protection, and labor distortions. He also studies the interaction among poverty, conflict, and culture. A Spanish national, Cuesta has experience in countries in Africa, Asia, and Latin America. He is currently an associate editor for the European Journal of Development Research and the Journal of Economic Policy Reform, and the editor of the World Bank’s quarterly Food Price Watch.

Mario Negre, co-director of the report, is a senior economist in the World Bank Develop- ment Economics Research Group, where he is seconded by the German Development Institute. He is a nonresidential research fellow at Maastricht School of Management. He has worked at the European Parliament, ﬁrst as an adviser to the chairman of the Develop- ment Committee and then for all external relations committees. Since 2012, he has been a senior researcher at the German Development Institute. His ﬁelds of specialization are pro- poor growth, inclusiveness, inequality, and poverty measurement, as well as development cooperation policy, particularly in Europe. Mario holds a BSc in physics from the University of Barcelona, an MA in development policies from the University of Bremen, and a PhD in development economics from the Jawaharlal Nehru University, India.

Soumya Chattopadhyay is a research fellow in the Growth, Poverty, and Inequality Pro- gramme at the Overseas Development Institute. His research interests include assessing the impact on subjective well-being of macro conditions and policy interventions using household surveys, identifying the vulnerable and the marginalized, and the issues revolving around infrastructure investment and service delivery. Previously, he worked at the World Bank. He was a senior research associate in the Global Economy and Development Pro- gram at the Brookings Institution, and taught at the School of Public Policy at the University of Maryland. Soumya holds a BA in economics from the University of Delhi and the Univer- sity of Cambridge, a masters in public management, and a PhD in international economic policy from the University of Maryland.

xiii Shaohua Chen is a lead statistician in the Development Economics Research Group of the World Bank. Her main research interests over the past 20 years have been on poverty and inequality measurement. She has managed the global poverty monitoring and online com- putational tool PovcalNet at the World Bank since 1991. She is also responsible for the measurement and projection of global poverty for the major reports of the World Bank, such as World Development Indicators and the Global Monitoring Report. Before joining the World Bank, Shaohua was a lecturer at Huazhong University of Science and Technology. Her research ﬁndings have been published in major economic and statistical journals, including the Journal of Development Economics, the Journal of Public Economics, The Quarterly Journal of Economics, and The Review of Economics and Statistics. She received her MSc in statistical computing from the American University.

María Eugenia Genoni is a senior economist at the Poverty and Equity Global Practice at the World Bank. Her ﬁelds of specialization are survey design, poverty and inequality, migration, and risk management. At the World Bank, she has led the poverty and equity programs in Bolivia and Peru. She has also contributed to the poverty work in Argentina and Central America and to the Regional Gender Impact Evaluation Initiative. Before joining the World Bank, María worked in the research department at the Inter-American Development Bank and the Ministry of Finance of the Province of Buenos Aires in Argentina. She holds a PhD in economics from Duke University.

Christoph Lakner is an economist in the Development Research Group at the World Bank (Poverty and Inequality Team). He previously worked in the World Bank’s Poverty Practice on Poverty and Inequality issues in Argentina. His research interests include inequality, poverty, and labor markets in developing countries. In particular, he has been working on global inequality, the relationship between inequality of opportunity and growth, the implications of regional price differences for inequality, and the income composition of top incomes. He holds a BA, MPhil, and DPhil in economics from the University of Oxford.

Maura K. Leary is the communications lead for the World Bank Group’s Poverty and Equity Global Practice, where she manages strategic communications and outreach on poverty reduction, equity, shared prosperity, and inequality. From 2011 to 2013, she was the partnerships and communications specialist on the Gender and Development team. Prior to joining the World Bank in 2011, she worked at George Washington University and Tufts University, managing academic programs in France and the United States for high school, university, graduate, and adult students. She holds a BA in French from Connecticut College and a master’s degree in international affairs from the Elliott School of International Affairs at the George Washington University.

José Montes is a data scientist in the Poverty and Equity Global Practice of the World Bank. He has been working for more than 10 years in poverty and inequality measurement and helping national statistics ofﬁces improve the quality, consistency, and documentation of their household surveys. Since 2013, he has been part of Europe and Central Asia Team Statistics Development, and since 2016, part of the core team of the Global Team Statistics Development. Prior to the World Bank, he worked at the Inter-American Development Bank as a household survey specialist in the Poverty and Advisory Unit. He was also a lecturer at the Universidad del Pacíﬁco. José holds a BA in economics from the Universidad del Pacíf- ico and an MSc in economics and an MSc in statistics from Texas A&M University.

Espen Beer Prydz is an economist in the World Bank’s Development Economics Research Group. His research interests include issues of poverty, inequality, and survey methods. He has worked in Cambodia, Indonesia, and South Sudan on poverty, social protection, and economic policy. Prior to joining the World Bank, Espen undertook research on poverty, xiv POVERTY AND SHARED PROSPERITY 2016 labor markets, and gender with the Development Centre of the Organisation for Economic Co-operation and Development and the Abdul Latif Jameel Poverty Action Lab. Prydz is a Norwegian national who holds a BS from the London School of Economics and an MPA in international development from the John F. Kennedy School of Government at Harvard University.

Maika Schmidt is a consultant in the World Bank’s Development Economics Research Group. Her research interests are poverty, inequality, and pro-poor growth with a focus on measurement issues, speciﬁcally multidimensional indicators. Maika has worked for the Deutsche Gesellschaft für Internationale Zusammenarbeit. She holds a BSc in economics from the University of Mannheim, a master’s from the Barcelona Graduate School of Eco- nomics, and a master of research from Pompeu Fabra University, Barcelona. She is currently pursuing her PhD in development economics from the University of Sussex.

Ani Silwal is a consultant in the World Bank’s Development Economics Research Group. He holds a BA from Swarthmore College, an MSc from the University of Maryland, and a PhD in economics from the University of Sussex. He has also worked on international migration and remittances. His research interest is the constraints that households face in escaping poverty.

The Contributors Timm Bönke is assistant professor of public economics at the School of Business and Economics at Freie Universität Berlin. His research interests revolve mainly around inequality and the distribution of income and wealth, the inclusiveness of growth, the design and incentives of state welfare institutions, and redistribution and insurance through tax-benefit systems. Since 2015, he has been associate editor of the Journal of Income Distribution and scientific board member of the Research Institute Economics of Inequality at Vienna University of Economics and Business. In addition, he is a regular reviewer for scientific foundations and acts as a scientific and political consultant. Timm holds a masters and a PhD in economics from Freie Universität Berlin.

Will Durbin studied ethics, politics, and economics as an undergraduate at Yale University and international development for a master’s degree at the Woodrow Wilson School of Public and International Affairs, Princeton University. He previously worked on climate change and environmental policy and currently focuses on poverty reduction.

Aparajita Goyal is a senior economist in the Poverty and Equity Global Practice of the World Bank. Her work focuses on microeconomic issues of development, with an empha- sis on technological innovation in agriculture, access to markets, and intellectual property rights. Her research has been published in leading academic journals, such as American Economic Review, Journal of Human Resources, and Journal of Development Economics, and has been featured in the popular press, including Frontline, Economist, and Wall Street Journal. Within the World Bank, she has previously worked in the Development Economics Research Group, Ofﬁce of the Chief Economist for the Latin America and Caribbean Region, and Agriculture Global Practice after joining the Young Professionals Program. She holds a BA in economics from St. Stephen’s College, University of Delhi, India, an MSc from the London School of Economics, and a PhD in economics from the University of Maryland.

Renzo Massari is a consultant in the World Bank’s Development Economics Research Group. His research interests include developing and applying econometric and machine learning methods to poverty and inequality measurement, the use of big data in development, imputation methods, and survey methodology. Previously, he worked at the Labor

ABOUT THE CORE TEAM AND THE CONTRIBUTORS xv and Social Protection Unit at the Inter-American Development Bank and as a consultant in empirical regulatory and policy issues in Peru and in the United States. He holds a certiﬁca- tion in data science and a PhD in economics from Duke University.

Terra Lawson-Remer’s work addresses the determinants and consequences of sustainable development, poverty and inequality, and social and economic rights fulﬁllment within and across generations. Her expertise includes health care, education, international economic law, legal empowerment, natural resource governance, global trade and transna- tional investment, extractive industries, democratic transitions, civil society, property rights, and business and human rights. She is the author of dozens of academic books and articles, including, most recently, Fulﬁlling Social and Economic Rights (with Sakiko Fukuda-Parr and Susan Randolph), published by Oxford University Press. Terra earned her BA in ethics, politics, and economics from Yale University; her JD from New York University School of Law, where she was a Dean’s Merit Scholar; and her PhD in political economy from the Law and Society Institute, New York University.

David Newhouse is a senior economist in the Poverty and Equity Global Practice. He currently leads the Bank’s engagement on poverty in Sri Lanka, as well as on projects that aim to understand the nature of global poverty and incorporate satellite imagery into poverty measurement. He was formerly a labor economist in the Social Protection and Labor Practice, where he helped lead efforts to analyze labor markets and the policy response in the wake of the 2008 ﬁnancial crisis. He ﬁrst joined the Bank in August 2007 and co-led the Indonesian Jobs Report. Previously, he worked in the Fiscal Affairs Department of the International Monetary Fund providing policy advice on energy and food subsidies. David holds a PhD in economics from Cornell University. He has co-authored a book and several journal articles on a wide range of issues relating to labor, health, and education in developing countries.

Stace Nicholson is a senior program officer for international economic and financial affairs at the Japan International Cooperation Agency’s U.S. office. In this position, he conducts and oversees research that supports credit risk assessment of the agency’s concessional loan portfolio and serves as the agency’s junior liaison with multilateral financial institutions based in Washington. His work includes tracking macroeconomic developments across a range of emerging and frontier, or developing economies, facilitating research partnerships and project co-financing on an ad hoc basis, and reporting on development finance trends. Prior to joining the agency, Stace interned with the Ghana Center for Democratic Devel- opment and undertook field research in Uganda as a microfinance client assessment fellow for the Foundation for International Community Assistance. He is a summa cum laude graduate (political science) of Manchester College and holds a master’s degree in global finance, trade, and economic integration from the Josef Korbel School for International Studies, University of Denver.

xvi POVERTY AND SHARED PROSPERITY 2016 Abbreviations

BRICS Brazil, Russian Federation, India, China, and South Africa CCT conditional cash transfer CPI consumer price index ECD early childhood development EU European Union GDP gross domestic product IQ intelligence quotient OECD Organisation for Economic Co-operation and Development PPP purchasing power parity SEDLAC Socio-Economic Database for Latin America and the Caribbean UNESCO United Nations Educational, Scientiﬁc, and Cultural Organization VAT value added tax WDI World Development Indicators

Note: All dollar amounts are U.S. dollars (US$) unless otherwise indicated.

For a list of the 3-letter country codes used by the World Bank, please go to: https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country -and-lending-groups.

xvii

Overview

Two complementary goals to leave no one behind On April 20, 2013, the Board of Executive This complementarity has three import- Directors of the World Bank adopted two ant implications. First, by choosing these ambitious goals: end global extreme pov- two goals, the World Bank focuses squarely erty and promote shared prosperity in every on improving the welfare of the least well country in a sustainable way. This implies off across the world, effectively ensuring reducing the poverty headcount ratio from that everyone is part of a dynamic and in- 10.7 percent globally in 2013 to 3.0 percent clusive growth process, no matter the cir- by 2030 and fostering the growth in the in- cumstances, the country context, or the come or the consumption expenditure of time period. Second, monitoring the two the poorest 40 percent of the population goals separately is necessary to understand (the bottom 40) in each country. These two with precision the progress in achieving goals are part of a wider international devel- better living conditions among those most opment agenda and are intimately related to in need. Third, policy interventions that United Nation’s Sustainable Development reduce extreme poverty may or may not be Goals 1 and 10, respectively, which have effective in boosting shared prosperity if the been adopted by the global community. two groups—the poor and the bottom 40— Each goal has an intrinsic value on its are composed of distinct populations. own merits, but the two goals are also highly To understand more clearly the prog- complementary. Take the example of a ress toward the achievement of the goals, low-income Sub-Saharan African country the World Bank is launching the annual with a high poverty headcount ratio and an Poverty and Shared Prosperity report series, upper-middle-income country in Eastern which this report inaugurates. The report Europe or Latin America with low levels of series will inform a global audience com- extreme poverty, but rising concerns about prising development practitioners, policy inequality. Ending extreme poverty is espe- makers, researchers, advocates, and citizens cially relevant in the former, while expand- in general with the latest and most accu- ing shared prosperity is especially meaning- rate estimates on trends in global poverty ful in the latter. The complementarity of the and shared prosperity. Every year, it will two goals also derives from the composition update information on the global number of the world’s poor and bottom 40 popula- of the poor, the poverty headcount ratio tions. At a global scale, while 9 in every 10 of worldwide, the regions that have been more the extreme poor were among the national successful or that have been lagging in bottom 40 in 2013, only a quarter of the bot- advancing toward the goals, and the en- tom 40 were among the extreme poor (both hancements in monitoring and measuring cases refer to the orange area in ﬁgure O.1). poverty. In addition, it will feature a special

OVERVIEW 1 FIGURE O.1 Distribution of the Extreme Poor, the Nonpoor, the Bottom 40, and the Top 60, 2013

100

40 National percentile 30 20

0 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 Cumulative global polulation

Top 60, nonpoor Bottom 40, nonpoor Top 60, poor Bottom 40, poor

Source: Inspired by Beegle et al. 2014 and updated with 2013 data. Note: The figure has been constructed from vertical bars representing countries sorted in descending order by extreme poverty headcount ratio (from left to right). The width of each bar reflects the size of the national population. The figure thus illustrates the situation across the total global population.

focal theme. This year, the focal theme is These substantive considerations highlight inequality. the importance of directing attention to the problem of inequality. Inequality matters for There are other reasons too. Sustaining achieving the goals, the rapid progress in reducing poverty and boosting shared prosperity that has been but also for other reasons achieved over the last 25 years is at risk be- Despite decades of substantial progress in cause of the struggles across economies to boosting prosperity and reducing poverty, recover from the global financial crisis that the world continues to suffer from substan- started in 2008 and the subsequent slow- tial inequalities. For example, the poorest down in global growth. The goal of elimi- children are four times less likely than the nating extreme poverty by 2030—which is richest children to be enrolled in primary ed- likely to become more difficult as we ap- ucation across developing countries. Among proach more closely to it—might not be the estimated 780 million illiterate adults achieved without accelerated economic worldwide, nearly two-thirds are women. growth or reductions in within-country Poor people face higher risks of malnutri- inequalities, especially among those coun- tion and death in childhood and lower odds tries with large concentrations of the poor. of receiving key health care interventions.1 Generally speaking, poverty can be reduced Such inequalities are associated with through higher average growth, a narrow- high financial cost, affect economic growth, ing in inequality, or a combination of the and generate social and political burdens two.2 Achieving the same poverty reduction and barriers. But leveling the playing field during a slowdown in growth therefore reis also an issue of fairness and justice that quires a more equal income distribution. It resonates across societies on its own merits. follows that, to reach the goals, efforts to fos-

2 POVERTY AND SHARED PROSPERITY 2016 ter growth need to be complemented by eq- mean that such forms of inequality do not uity-enhancing policies and interventions. deserve attention. According to Oxfam, 62 Some level of inequality is desirable to individuals in 2015 had the same wealth as maintain an appropriate incentive struc- the bottom half of the world’s population; ture in the economy or simply because within the African continent, this statistic inequality also reflects different levels of is even more extreme.3 However, the report talent and effort among individuals. How- looks into inequality in income, in out- ever, the substantial inequality observed comes such as in health care and education, in the world today offers ample room for and inequality in opportunities. Income in- taking on inequality. Doing so without equality and unequal opportunities are in- compromising growth is not only possible, timately related. This report aims to dispel but can be beneficial for poverty reduction myths around income inequality. Reflecting and shared prosperity if done smartly. A on what has worked in addressing this pro- trade-off between efficiency and equity is found problem is key to taking on inequal- not inevitable. The evidence that equity- ity more successfully. enhancing interventions can also bolster The report makes four main contribu- economic growth and long-term prosperity tions. First, it presents the most recent num- is wide-ranging. To the extent that such in- bers on poverty, shared prosperity, and interventions interrupt the intergenerational equality. Second, it stresses the importance of reproduction of inequalities of opportu- inequality reduction in ending poverty and nity, they address the roots and drivers of boosting shared prosperity by 2030, partic- inequality, while laying the foundations for ularly in a context of weaker growth. Third, boosting shared prosperity and fostering it highlights the diversity of within-country long-term growth. Reducing inequalities inequality reduction episodes and synthe- of opportunity among individuals, econ- sizes the experiences of several countries omies, and regions may also be conducive and policies in addressing the roots of in- to political and societal stability and social equality without compromising economic cohesion. In more cohesive societies, threats growth. Along the way, the report shatters arising from extremism, political turmoil, some myths and sharpens our knowledge of and institutional fragility are less likely. what works in reducing inequalities. Finally, it also advocates for the need to expand and The key question the report improve data collection—availability, comparability, and quality—and rigorous evi- addresses: what can be dence on inequality impacts. This is essential done to take on inequality? for high-quality poverty and shared pros- This report addresses the issue of inequality perity monitoring and the policy decisions by documenting trends in inequality, iden- such an exercise ought to support. tifying recent country experiences in successfully reducing inequality and boosting Extreme poverty is shared prosperity, examining key lessons, shrinking worldwide, but is and synthesizing the evidence on public still widespread in Africa policies that lessen inequality by reducing poverty and promoting shared prosperity. In 2013, the year of the latest comprehen- Inequality exists in many dimensions, sive data on global poverty, 767 million and the question “inequality of what?” is people are estimated to have been living essential. The report focuses on inequalities below the international poverty line of in income or consumption expenditures, US$1.90 per person per day (table O.1). Al- but it also analyzes the deprivations among most 11 people in every 100 in the world, the extreme poor and the well-being of the or 10.7 percent of the global population, bottom 40. However, it does not address all were poor by this standard, about 1.7 per- types of inequality, for example, inequality centage points down from the global pov- related to ownership of assets. This does not erty headcount ratio in 2012. Although this

OVERVIEW 3 TABLE O.1 World and Regional Poverty Estimates, 2013

Headcount Poverty Squared poverty Poor Region ratio (%) gap (%) gap (%) (millions) East Asia and Paciﬁc 3.5 0.7 0.2 71.0 Eastern Europe and Central Asia 2.3 0.6 0.3 10.8 Latin America and the Caribbean 5.4 2.6 1.8 33.6 Middle East and North Africaa —— — — South Asia 15.1 2.8 0.8 256.2 Sub-Saharan Africa 41.0 15.9 8.4 388.7 Total, six regions 12.6 3.8 1.8 766.6 World 10.7 3.2 1.5 766.6

Source: Latest estimates based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch .worldbank.org/PovcalNet/. Note: Poverty is measured using the US$1.90-a-day 2011 purchasing power parity (PPP) poverty line. The six-region total includes all developing regions. World includes all developing regions, plus industrialized countries. Definitions of geographical regions are those of PovcalNet. — = not available. a. Estimates on the Middle East and North Africa are omitted because of data coverage and quality problems. The population coverage of available household surveys is too low; the share of the total regional population represented by the available surveys is below 40 percent. There are also issues in the application of the 2011 PPP U.S. dollar to the region. These issues revolve around the quality of the data in several countries experiencing severe political instability, breaks in the consumer price index (CPI) series, and measurement or comparability problems in specific household surveys. These caveats suggest that further methodological analyses and the availability of new household survey data are both needed before reliable and sufficiently precise estimates can be produced.

represented a noticeable decline, the pov- global extreme poverty headcount ratio erty rate remains unacceptably high given dropped steadily over this period. Despite the low standard of living implied by the more rapid demographic growth in poorer $1.90-a-day threshold. areas, the forceful trend in poverty reduc- The substantial decline is mostly ex- tion culminated with 114 million people plained by the lower number of the ex- lifting themselves out of extreme poverty in treme poor in two regions, East Asia and 2013 alone (in net terms). Pacific (71 million fewer poor) and South Asia (37 million fewer poor), that showed cuts in the extreme poverty headcount ratio The geography of global of 3.6 and 2.4 percentage points, respec- extreme poverty is tively. The former is explained in large part changing as poverty by lower estimates on China and Indonesia, declines whereas the decrease in South Asia is driven by India’s growth. The number of the poor As extreme poverty declines globally, the in Sub-Saharan Africa fell by only 4 million regional poverty profile has been chang- between 2012 and 2013, a 1.6 percentage ing. This is a direct result of uneven prog- point drop that leaves the headcount ratio ress, mainly at the expense of Sub-Saharan at a still high 41.0 percent. Eastern Europe Africa, which has the world’s largest head- and Central Asia’s headcount ratio shrank by count ratio (41.0 percent) and houses the about a quarter of a percentage point, down largest number of the poor (389 million), to 2.3 percent, while, in Latin America and more than all other regions combined. This the Caribbean, the ratio declined by 0.2 per- is a notable shift with respect to 1990, when centage points, to 5.4 percent (figure O.2). half of the poor were living in East Asia Both the extreme poverty headcount and Pacific, which, today, is home to only ratio and the total number of the extreme 9.3 percent of the global poor. South Asia poor have steadily declined worldwide since has another third of the poor, while Latin 1990 (figure O.3). The world had almost 1.1 America and the Caribbean, along with billion fewer poor in 2013 than in 1990, a Eastern Europe and Central Asia, complete period in which the world population grew the global count with 4.4 percent and 1.4 by almost 1.9 billion people. Overall, the percent, respectively (figure O.4).4

4 POVERTY AND SHARED PROSPERITY 2016 FIGURE O.2 World and Regional Trends, Poverty Headcount Ratio, 1990–2013

Poverty headcount ratio (%) 20

0 1990 1993 1996 1999 2002 2005 2008 2011 2013 East Asia and Paciﬁc South Asia Eastern Europe and Central Asia Sub-Saharan Africa Latin America and the Caribbean World Middle East and North Africa

Who are the poor? FIGURE O.3 Trends in the Global Poverty Headcount Ratio and the Number of the Global Poor, 1990–2013 Exploring the characteristics of the poor is 60 2,000 key to a better understanding of the circum- 1,850 1,855 stances and contexts surrounding poverty. 1,666 1,693 1,800 A large database of household surveys in 50 1,588 89 developing countries provides insights 1,600 into this issue by facilitating a demographic 1,328 1,400 40 1,206 profile of the poor at the US$1.90 pov- 35.0% erty line.5 This poverty profile reveals that 33.5% 1,078 1,200 28.8% 946 30 28.1% the global poor are predominantly rural, 25.3% 881 1,000 young, poorly educated, mostly employed 800 in the agricultural sector, and live in larger 20.4% 20 17.8% 767 households with more children. Indeed, 80 15.6% 600 Number of poor (millions) percent of the worldwide poor live in rural Poverty headcount ratio (%) 13.5% 400 10 areas; 64 percent work in agriculture; 44 12.4% percent are 14 years old or younger; and 39 10.7% 200 percent have no formal education at all. The 0 0 data also confirm wide regional variations in the distribution of the poor across these 1990 1993 1996 1999 2002 2005 2008 2010201120122013 characteristics (figure O.5). Poverty headcount ratio (% of people who live below $1.90) When looking at the incidence of pov- Number of people who live below $1.90 a day (right axis) erty across different population groups, Source: Latest estimates based on 2013 data using PovcalNet (online analysis tool), World Bank, Wash- poverty headcount ratios are more than ington, DC, http://iresearch.worldbank.org/PovcalNet/. three times higher among rural residents Note: Poverty is measured using the US$1.90-a-day 2011 PPP poverty line.

OVERVIEW 5 FIGURE O.4 Where Are the Global Poor than among urban dwellers: 18.2 percent Living? The Global Poor, by Region, 2013 versus 5.5 percent, respectively. Agricultural Share of global poor by region (%) workers are over four times more likely than people employed in other sectors of the 0.8% 9.3% economy to be poor. Educational attain- 1.4% ment is inversely correlated with poverty. A 4.4% small share of primary-school graduates are living in poverty: fewer than 8.0 percent of people who completed primary school, but not secondary school, are living below the 50.7% US$1.90 poverty line. Among individuals who have attended university, the share is 33.4% less than 1.5 percent.6 Similar differences are observed if poverty incidence is measured relative to the US$3.10-a-day poverty line. Age profiles confirm that children are East Asia and Pacific more likely than adults to be poor. Children South Asia under 18 account for half the global poor in Eastern Europe and Central Asia 2013, but less than a third of the sample pop- Sub-Saharan Africa ulation (32 percent) (figure O.6). Younger Latin America and the Caribbean Rest of the world children (ages 0–14) contribute especially heavily to the poverty headcount, much more Source: Latest estimates based on 2013 data using PovcalNet than their share in the world’s population. (online analysis tool), World Bank, Washington, DC, http:// iresearch.worldbank.org/PovcalNet/. Progress in boosting shared prosperity worldwide is FIGURE O.5 Profile of the Poor, by Characteristics and Region, 2013 uneven Shared prosperity is measured as the growth in the average income or consumption of Share of poor the bottom 40. The larger the growth rate in rural areas in the income of the bottom 40, the more quickly prosperity is shared with the most disadvantaged sectors in society. To the extent that greater economic Share of poor adults working in agriculture growth is associated with rising incomes among the poor and the bottom 40, more rapid growth will lead to greater shared prosperity and poverty reduction. Likewise, Share of poor 0–14 years old a more rapid increase in shared prosperity and in the narrowing of inequality typically accelerates the decline in poverty at any given rate of growth. Share of poor adults Progress on this indicator is examined with no education in this report using the latest information available on each country, currently circa 0 102030405060 70 80 90 2008–13. To take into account the share Percent of prosperity going to groups other than East Asia and Pacific South Asia the bottom 40, the report also monitors the Eastern Europe and Central Asia Sub-Saharan Africa shared prosperity premium, defined as the Latin America and the Carribean World difference between the growth in the in- Source: Castañeda et al. 2016. come of the bottom 40 and the growth in Note: Poverty is measured using the US$1.90-a-day 2011 PPP poverty line. income at the mean in each country. A pos-

6 POVERTY AND SHARED PROSPERITY 2016 FIGURE O.6 Age Profile of the Poor, 2013 a. The extremely poor b. Sample population

5.8% 10.3% 9.4% 15.9% 9.0%

8.6% 15.4%

44.0% 5.1%

12.9% 57.6%

6.0%

Children, ages 0–4 Children, ages 10–14 Adults, ages 18–59 Children, ages 5–9 Children, ages 15–17 Adults, ages 60 or more

Source: Newhouse et al. 2016.

itive premium indicates that the growth in the Caribbean, the income of the bottom the income or consumption of the bottom 40 grew by 8.0 percent in Paraguay, while 40 exceeds that of the mean, and by impli- in Honduras, income contracted by about cation, that of the rest of the population. A 2.5 percent annually during the same spell. higher or lower premium indicates the ex- A source of concern is the small value tent to which distributional changes favor of the shared prosperity premium. While the bottom 40 relative to the top 60. the average annualized growth in the in- The bottom 40 benefited from solid eco- come or consumption of the bottom 40 nomic growth in many countries in 2008– was 2.0 percent worldwide circa 2008–13 13. Overall, the bottom 40 in 60 of the 83 (a population-weighted 4.6 percent), the countries monitored experienced positive average shared prosperity premium was income growth, representing 67 percent of only 0.5 percentage points during the same the world’s population and 89 percent of the period (a population-weighted 0.4 percent- population represented by the surveys (fig- age point). Is this sufficient to expect large ure O.7). A total of 49 countries reported a reductions in inequality and poverty so as positive shared prosperity premium: income to achieve the World Bank goals by 2030? growth among the bottom 40 exceeded that of the mean (and therefore, that of the top A more rapid decline in 60). However, there is no room for compla- inequality is needed to cency: in 23 countries, the incomes of the bottom 40 declined during the period. end poverty There are wide regional differences in Figure O.8 makes it clear that the goal of shared prosperity and the shared prosperity ending poverty by 2030 cannot be reached premium. The best performers were in East at current levels of economic growth. It Asia and Pacific and in Latin America and shows the trajectory of the global poverty the Caribbean, while high-income indus- headcount ratio under various assumptions trialized countries performed the least well. about distributional changes and under the Greece, a high-income country, experienced assumption that every country will grow at an annualized contraction of 10.0 percent its rate of the last 10 years. These changes in the income of the bottom 40, while the are modeled by means of alternative shared Democratic Republic of Congo recorded prosperity premiums in each country. a rise of 9.6 percent. In Latin America and Thus, in the scenario of a premium labeled

OVERVIEW 7 FIGURE O.7 Shared Prosperity, 83 Countries, 2008–13

East Asia and Paciﬁc China Mongolia Cambodia Thailand Vietnam Indonesia Philippines Lao PDR

Eastern Europe and Central Asia Belarus Kazakhstan Russian Federation Slovak Republic Macedonia, FYR Moldova Georgia Ukraine Turkey Romania Poland Bulgaria Armenia Kyrgyz Republic Czech Republic Slovenia Albania Serbia Lithuania Hungary Estonia Montenegro Latvia Croatia Industrialized countries Norway Switzerland Sweden Finland Germany Belgium Austria France Netherlands United States Denmark Spain United Kingdom Portugal Luxembourg Cyprus Italy Iceland Ireland Greece

Latin America and the Caribbean Paraguay Ecuador Bolivia Brazil Colombia Peru Chile Uruguay Nicaragua Panama El Salvador Argentina Dominican Republic Costa Rica Mexico Honduras

Middle East and North Africa Iran, Islamic Rep. Iraq

South Asia Bhutan India Pakistan Sri Lanka

Sub-Saharan Africa Congo, Dem. Rep. Uganda Tanzania Congo, Rep. Togo Cameroon Mauritius Rwanda Senegal

–10–5 0 5 10 Annualized growth in mean income or consumption (%) Total population Bottom 40

Poverty headcount ratio (%) an historical exception to a 5 rising trend 3 Data since the 1990s show a substantial nar- 0 rowing in inequality in income or consumption worldwide, irrespective of residence. 200220042006200820102012201420162018202020222024202620282030 This is the first such reduction since the m = 0 m = 1 m = 2 industrial revolution (figure O.9). This un- m = –1 m = –2 2030 goal precedented decline occurred during a pe- Source: Updated results based on Lakner, Negre, and Prydz riod of increasing global integration. From 2014. Note: m = the assumed shared prosperity premium, that is, the 1820 to the 1990s, global inequality steadily growth in income or consumption among the bottom 40, minus rose. Then, the Gini index fell to 62.5 in 2013, the growth in income or consumption at the mean. Thus, for most markedly beginning in 2008, when the example, m = 2 indicates that the growth in income among the bottom 40 exceeds the growth in income at the mean by 2 per- Gini was 66.8 (the blue line in figure O.10). centage points in each country. Poverty is measured using the US$1.90-a-day 2011 PPP poverty line. FIGURE O.9 Global Inequality, 1820–2010 m = 1, the growth in the income of the bottom 40 in each country is assumed to ex- 70 ceed the growth rate in the mean by 1 percentage point. Meanwhile, in the scenario m = 0, growth is distributionally neu- 60 tral: the income of the bottom 40 and the mean grow at the same pace. Under these 50 scenarios, the poverty goal would only be Gini index reached if the shared prosperity premium is in excess of 1 percentage point, which is 40 double the simple average premium countries are able to achieve today (0.5 percentage points). Thus, income or consumption 30 1820 1870 1920 1970 1990 2010 needs to grow more quickly among the bottom 40 than at the mean, and at a more Source: Based on figure 1 (p. 27) of The Globalization of Inequal- rapid pace than today, especially in coun- ity by Francois Bourguignon (Princeton University Press 2015). Used with permission. tries with substantial numbers of the poor. Note: The discontinuity in the series represents the change in This is the analytical result of a set of the base year of the purchasing power parity (PPP) exchange simulations. In practice, this does not mean rates from 1990 to 2005. The figure uses GDP per capita in combination with distributional statistics from household surveys. that every country worldwide must improve Figure O.10 uses income (or consumption) per capita directly its income distribution to achieve the pov- from household surveys (in 2011 PPP exchange rates).

OVERVIEW 9 FIGURE O.10 Global Inequality, 1988–2013 This unprecedented drop in global inequal- 80 1.0 ity was driven by a convergence in average incomes across countries that was spurred by 70 rising incomes in populous countries such 0.8 as China and India. As a result, between- country inequality declined. In contrast, 60 within-country inequality, the other com- 0.6 80 76 74 72 70 65 ponent of global inequality, took on a greater role in global inequality (explaining 0.4 50 Gini index a third of the total variation) (ﬁgure O.10). Mean log deviation

0.2 40 Despite recent progress, 20 24 26 28 30 35 average within-country 0 30 inequality is greater now 1988 1993 1998 2003 2008 2013 than 25 years ago Within-country inequality Gini index (right axis) Between-country inequality The population-weighted Gini index cap-

Sources: Lakner and Milanovic´ 2016; Milanovic´ 2016; World Bank calculations based on PovcalNet (online tures within-country inequality relative to analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. the average person across the countries on Note: For each country, household income or consumption per capita is obtained directly from house- which data are available (figure O.11). This hold surveys and expressed in 2011 PPP exchange rates. Each country distribution is represented by 10 decile groups. The line (measured on the right axis) shows the level of the global Gini index. The height indicator rose steeply, by 6 points, from 34 of the bars indicates the level of global inequality as measured by GE(0) (the mean log deviation). The to 40 between 1988 and 1998. Since then, red bars show the corresponding level of population-weighted inequality within countries. The level of inequality has declined more moderately, between-country inequality, which captures differences in average incomes across countries, is shown by the yellow bars. The numbers in the bars refer to the relative contributions (in percent) of these two by almost 1 point, to a Gini of 39 in 2013. sources to total global inequality. Thus, within-country inequality for the average person in the world was wider in 2013 FIGURE O.11 Average Within-Country Inequality, 1988–2013 than 25 years previously. The population-weighted result on 42 within-country inequality is largely robust to other specifications, such as population- 40 unweighted estimates or estimates drawing on different country samples. As shown 38 in figure O.11, the unweighted Gini index of within-country inequality worldwide also rose during the 1990s, but by a smaller 36 amount than the population-weighted index. The simple average Gini increased by 34 around 5 points, from 36 in 1988 to 41 ten years later, before declining steadily thereaf-

Average within-country Gini index 32 ter, reaching 38 in 2013. The levels and trends in average inequality are quite different across regions, 30 although the most recent decline is broad- 1988 1993 1998 2003 2008 2013 based (ﬁgure O.12). Developing countries Unweighted Weighted tend to exhibit wider within-country in- Unweighted, balanced Weighted, balanced equality relative to developed countries. Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World Latin America and the Caribbean, as well Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/; WDI (World Development Indicators) (database), World Bank, Washington, DC, http://data.worldbank.org/data-catalog/world-development as Sub-Saharan Africa, stand out as high- -indicators. inequality regions. The former is also the Note: The solid lines show the trend in the average within-country Gini index with and without population region most successful in reducing inequal- weights in the full sample (an average of 109 countries per benchmark year). The dashed lines refer to the balanced sample, that is, using only the set of 41 countries on which data are available in every bench- ity. Sub-Saharan Africa has likewise steadily mark year. narrowed inequality since the early 1990s,

10 POVERTY AND SHARED PROSPERITY 2016 although this progress hides wide-ranging FIGURE O.12 Trends in the Average National Gini, by Region, 1988– variations within the continent.7 In Eastern 2013 Europe and Central Asia, average inequality 60 rose sharply after the fall of the Berlin Wall, but has since been on a declining trend. The average industrialized country saw an 50 increase in the Gini index from 30 to 33 between 1988 and 2008. In the five years leading up to 2013, average within-income in- 40 equality appears to have fallen in all regions except in the Middle East and North Africa and in South Asia. 30 Providing a simple explanation behind regional inequality trends is particularly challenging because the patterns may be Average within-country Gini index 20 distinctive and the drivers specific to the trends exhibited by countries within a region. Rather than providing a simplistic 10 1988 1993 1998 2003 2008 2013 explanation, it may therefore be useful to examine closely the country variations East Asia and Pacific Eastern Europe and Central Asia Latin America and the Caribbean Middle East and North Africa within regions to understand the extent South Asia Sub-Saharan Africa to which the common drivers behind in- Industrialized countries equality—gaps in human capital accumu- Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World lation, varying access to jobs and income- Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. generating opportunities, and government Note: The lines show the average within-country Gini index by region. It is the simple average in the full interventions to address market-based in- sample without weighting countries by population. Industrialized countries are a subset of high-income countries. See chapter 2, annex 2B, for the list of industrialized countries. equalities—are relevant in each country. Indeed, between 2008 and 2013, the number of countries experiencing declin- bution is available, such as Argentina; India; ing inequality was twice the number exhib- the Republic of Korea; South Africa; Taiwan, iting widening inequality. This shows that China; and the United States, the share of within-country inequality can widen or the top 1 percent in total income has been narrow. Despite the progress, stark inequal- increasing. In South Africa, the top income ities persist. For example, Haiti and South share roughly doubled over 20 years to lev- Africa are the most unequal countries in els comparable with those observed in the the world (for which data are available), United States (figure O.13). with a Gini exceeding 60 points in 2013. Another Sub-Saharan African country (Rwanda) and another seven Latin America Inequality reduction is not and Caribbean countries (Brazil, Chile, Co- limited to a few countries, lombia, Costa Rica, Honduras, Mexico, and settings, and policy choices Panama) make up the top 10 most unequal countries in the world, with Gini indexes in Some countries have performed remarkably excess of or close to 50. well in reducing inequality and boosting shared prosperity. Others have not. Among the constellation of policies that have been In many economies, implemented, what have been the key levers the income share of the in boosting shared prosperity and narrow- top income groups is ing inequality among countries? expanding The report focuses on the experiences of five low- and middle-income countries, In many economies in which information covering Asia, Latin America and the Carib- on the top 1 percent of the income distri- bean, and Sub-Saharan Africa. The countries

OVERVIEW 11 FIGURE O.13 The National Income Share of the Richest 1 Percent, Selected Economies a. Industralized economies since 1900 b. Developing economies since 1980

20 20

15 15

10 10 Income share of top 1 percent (%) 5 Income share of top 1 percent (%) 5

1900 1925 1950 1975 2000 2015 1980 1990 2000 2010 2013 United States Japan France South Africa Argentina India Korea, Rep. Taiwan, China

Source: Calculations based on data of WID (World Wealth and Income Database), Paris School of Economics, Paris, http://www .parisschoolofeconomics.eu/en/research/the-world-wealth-income-database/. Note: The income share excludes capital gains. These measures are typically derived from tax record data. For South Africa, the income share refers to adults.

analyzed are Brazil, Cambodia, Mali, Peru, also play a role in reducing inequalities. For and Tanzania. These are among the best per- example, the minimum wage and safety nets formers, showing good shared prosperity have been crucial in allowing Brazil to lessen premiums and strong records in narrowing inequality, while diversification from agri- income inequality and reducing extreme culture into light manufacturing and ser- poverty. They are also sufficiently diverse vices in Cambodia opened job opportunities to embody different development strategies to the poor. and historical circumstances. Overall, these country cases also high- The five countries exercised judicious light that success in reducing inequality macroeconomic management, appropriately and boosting shared prosperity in a given dealt with external shocks, and implemented period does not necessarily translate into more or less protracted and coherent eco- similar success on other economic, social, nomic and social sector reforms. They also or political fronts, nor into sustainable re- benefited from favorable external conditions ductions in inequality over time. Indeed, in the form of cheap and abundant interna- conflict emerged in Mali after the period tional credit, high commodity prices, and of inequality reduction, in large part be- booming trade. Decision making and the cause of protracted flaws in governance.8 context allowed rapid, sustainable, and in- The marked differences in the most recent clusive growth. The countries also highlight policy choices between Brazil and Peru on the importance of labor markets in trans- fiscal consolidation and the control of infla- lating economic growth into inequality re- tion largely explain the stark differences in duction by increasing job opportunities and their most recent growth patterns: gradual earnings, reintegrating individuals who have recovery in Peru, recession in Brazil. Mean- been excluded from economic opportuni- while, long-standing barriers constraining ties, and narrowing gaps across workers be- productivity and investments in agriculture cause of gender, residence, or sector of em- in Cambodia and an unfinished transition ployment. Notwithstanding these common to a market-based economy in Tanzania call factors, country-specific choices and eco- into question the sustainability of inequal- nomic developments—deliberate or not— ity reduction in these two countries.9

12 POVERTY AND SHARED PROSPERITY 2016 The common elements and country- placing the country among the most rap- specific peculiarities are summarized below. idly growing economies in the world. Poor Cambodians harnessed the opportunities Brazil, 2004–14: policies aligned created by this growth. They seized jobs to redress record inequality in labor-intensive industries and services, diversifying their incomes away from sub- In 1989, Brazil’s Gini index was 63, the sec- sistence agriculture and reaping higher re- ond highest in the world. However, the in- turns from traditional agricultural activi- comes of the less well off in Brazil surged be- ties. Annual consumption growth among tween 2004 and 2014 amid rapid economic the bottom 40 averaged 6.3 percent between growth. The Gini dropped to 51 in 2014, 2008 and 2013, twice the consumption while income growth among the bottom 40 growth of the top 60. averaged 6.8 percent a year, well above the A proliferation of employment opportu- average 4.5 percent among all Brazilians. nities followed expansions in the garment, Multiple drivers underlie Brazil’s suc- tourism, and real estate sectors.12 Relative cess. The 1988 Constitution laid the foun- to other sectors, wages in the garment in- dations for tackling historical inequalities by dustry tended to be higher and more stable, guaranteeing basic social rights such as free while the gender gap tended to be narrower. public education, free universal health care, Meanwhile, the agricultural sector’s vitality pensions, and social assistance. A macroeco- at a time of historically high international nomic framework established in the 1990s prices explains how farm incomes from allowed inflation to be curbed, promoted the paddy rice farming more than doubled prudent management of fiscal balances, and between 2004 and 2009.13 Indeed, rural created an enabling environment for poli- areas largely drove the country’s success in cies to address inequality. During the 2000s, inequality and poverty reduction. Nonethe- the boom in commodity prices generated less, obstacles are evident in the inadequate positive terms of trade. Macroeconomic sta- pace of job creation, given Cambodia’s bility, combined with this favorable external young demographic and structural con- context, propelled economic growth. Labor straints that weigh on leading sectors. market dynamics—including increasing wage premiums for the less skilled, more Mali, 2001–10: vagaries of formal jobs, and a rising minimum wage— agriculture rescue a weak and the expansion of social policies helped economy boost the incomes of the poor. These two factors accounted for approximately 80 per- Before the outbreak of conflict in the cent of the decline in inequality in 2003–13: country’s northern region in 2012, Mali 41 percent of the Gini decline in these years had made important strides in reducing stemmed from labor incomes, and 39 per- inequality. Between 2001 and 2010, GDP cent from nonlabor income sources such as growth averaged 5.7 percent a year. During government transfers.10 According to some the period, the Gini index fell 7 points. The estimates, Bolsa Família, Brazil’s flagship income of the bottom 40 grew, while the conditional cash transfer (CCT) program, mean contracted. alone explains between 10 percent and 15 Agriculture has been a key driver be- percent of the narrowing income inequality hind the improvement in living conditions observed in the 2000s.11 among the poor. Approximately 73 percent of Malians and 90 percent of the poor Cambodia, 2004–14: earning live in rural areas. For those involved in opportunities emerging from farming activities, own-account production growth typically does not permit self-sufficiency, and income has to be supplemented with Cambodia’s annual economic growth av- casual labor and private transfers. Higher eraged 7.8 percent between 2004 and 2014, cereal production in the 2000s benefited the

OVERVIEW 13 labor income of the poor by raising both mal workers, high labor force participation farm production and off-farm labor income rates, and low unemployment. through greater demand for wage labor by Challenges remain. Analysts question commercial cereal producers. In the latter the quality of public spending, notably in half of the first decade of the 2000s, while education. Despite significant gains in en- manufacturing was contracting, agricul- rollments, Peru lags comparator countries tural production, favored by good weather in international assessments of education conditions, boomed, resulting in reduced quality outcomes, such as student test scores. inequality.14 Since 2012, however, the con- This is a serious consideration because the flict in the north has put the brakes on the favorable external conditions that have progress of the previous decade. The crisis under pinned Peru’s growth have recently has disrupted education and health care begun to recede. Maintaining the impressive services in the north, and displaced popula- gains in a much less favorable environment tions are exerting pressure on service deliv- will require policy reforms that address the ery in the south. This resurgence of conflict limited productivity resulting from the low comes after two decades of relative stabil- quality of human capital and the high rates ity, including multiparty elections, and is of informality. associated with a long-term deterioration in governance, the expanding presence of Tanzania, 2004–14: sharing political pay-offs and co-optation, and an prosperity amid diversification army with limited capacity to face increasing security threats.15 Tanzania maintained robust and stable economic growth between 2004 and 2014, Peru, 2004–14: equalizing averaging 6.5 percent a year. The national growth through capital poverty headcount ratio fell from 34.4 per- investment cent in 2007 to 28.2 percent in 2012. The Gini index declined from around 39 to 36 The improvement in living conditions over the same period. Annual consumption among the poor and the bottom 40 in Peru growth among the bottom 40, at 3.4 per- has been remarkable. The Gini index fell cent, was more than three times the growth from 51 in 2004 to 44 in 2014, and poverty among the top 60, at 1.0 percent. rates dropped from 12 percent in 2004 to 3 Since the early 2000s, the country’s eco- percent in 2014. The outstanding growth of nomic expansion has been driven primarily the economy (6.6 percent annually during by rapidly growing sectors, especially com- the period) in a context of macroeconomic munications, financial services, and con- stability, favorable external conditions, and struction. However, the growth in these sec- important structural reforms was respon- tors has not been translated into substantive sible for this progress. In the early 2000s, improvements in the living conditions of prudent macroeconomic policies and high the poor, the less well educated, or rural commodity prices attracted foreign direct residents. After 2007, there was a surge in investment into the economy, particularly retail trade and manufacturing, particularly in the mining sector. Capital accumulation agroprocessing in products such as food, became the main driver of growth, account- beverages, and tobacco, which has allowed ing for more than two-thirds of total growth the inclusion of less highly skilled workers after 2001. The labor market was the main in the economy.17 Among policies explicitly pathway for the translation of the country’s aimed at rendering the income distribution impressive growth into less inequality and more equitable, the Tanzania Social Action poverty, explaining about 80 percent of the Fund stands out. It encompasses a CCT reduction in the Gini and three-quarters of program, public works, and a community the reduction in extreme poverty during the savings component that is expected to en- last decade.16 Critical to this success were a able the poorest segments of the popula- closing wage gap between formal and infor- tion to increase their savings and their in-

14 POVERTY AND SHARED PROSPERITY 2016 vestments in livestock and to become more Tanzania point to the need to realign fis- resilient.18 Despite this progress, much re- cal systems to produce a greater impact in mains to be done to trim regional dispar- reducing inequality. Infrastructure is ap- ities and expand access to basic services in parently still a significant obstacle in Cam- a context of rapid urbanization. Indeed, to- bodia, while in Mali, in addition to con- day’s economy is still characterized by a lack flict, dependence on external factors, from of competition in the private sector and the donor flows to the vagaries of weather, absence of growth, as well as a strongly reg- threaten sustained improvement. In Peru, ulated economic environment. the quality of education is below regional standards and represents a barrier to maintaining and enhancing economic produc- There are some common tivity should favorable external conditions building blocks behind disappear. successful inequality Countries willing to make the appropri- reductions ate policy choices are more likely to narrow inequality. Those that are not willing to The experiences of five countries cannot make these choices might continue to suffer supply precise policy prescriptions that the disadvantages of growing inequality. are valid everywhere and in all circumstances. However, they demonstrate that narrowing inequality and sharing prosper- Taking on inequality ity are possible in many settings, including involves human capital low- and middle-income countries; rural accumulation, income economies; more highly diversified, mod- generating opportunities, ern economies; and countries benefiting consumption smoothing, from external booms, but also countries facing unfavorable conditions, such as a and redistribution history of conflict or substantial, long-term The report assesses what we know about key inequality. The building blocks of success domestic policy interventions that are effec- have been prudent macroeconomic pol- tive in reducing inequality, the benefits they icies, strong growth, functioning labor generate, the choices that need to be made markets, and coherent domestic policies concerning their design and implementa- focusing on safety nets, human capital, and tion, and the trade-offs with which they are infrastructure. associated. It is not meant to provide an ex- As the building blocks get in place, haustive or comprehensive review of every many approaches to narrowing inequality intervention that could reduce inequality, are possible. However, sustaining this suc- nor does it seek to supply universal pre- cess may require similar approaches. The scriptions. Instead, it focuses on a few policy accumulation of good-quality human cap- areas on which a body of rigorous evidence ital, diversification in the income-earning allows lessons to be drawn with confidence. opportunities available to the poor, safety The policies, if well designed, have favorable nets capable of protecting the poorest from effects not only on inequality reduction, but risk, and enhanced infrastructure to con- also on poverty reduction without major nect lagging regions to economically more efficiency and equity trade-offs. The pol- vibrant ones are all potentially desirable icy areas are early childhood development approaches to sharing prosperity and re- (ECD), universal health care, universal ac- ducing inequality. cess to good-quality education, CCTs, in- Thus, the experiences of Cambodia, vestments in new or improved rural roads Mali, and Tanzania underscore the need and electrification, and taxation, mainly on to expand safety nets, which have not been personal income and consumption. sufficient to protect the poorest in these There are many pathways through which countries. The experiences of Brazil and policy interventions can affect inequality,

OVERVIEW 15 whether this effect is intended or unin- cent of the population, including 18 million tended. The impacts can be large or small, previously uninsured people.21 short term or lifelong, and they may narrow Recent assessments in developed and de- disparities in income, well-being, or oppor- veloping countries highlight the important tunity. For example, taxes can have direct consequences of successful experiences in and deliberate redistributive effects, reach- improving the quality of teaching. For ex- ing up to 20 points of the Gini index of ample, estimates in the United States indi- market incomes in some European Union cate that pupils taught by teachers who are (EU) economies.19 In contrast, investments at the 90th percentile in effectiveness are in rural roads and electrification influence able to learn 1.5 years’ worth of material in income generation opportunities, employ- a single academic year, while pupils taught ment, and even perceptions of gender roles. by teachers at the 10th percentile learn only Expanding ECD, health care coverage, and a half-year’s worth of material.22 Increased good-quality education often reduces cog- schooling has been linked to more produc- nitive, nutritional, and health status gaps, tive nonfarm activities in China, Ghana, thereby narrowing inequalities in human and Pakistan.23 capital development and future income In Bangladesh, the Shombob Pilot Pro- opportunities. By smoothing consumption gram reduced the incidence of wasting among the most deprived, especially during among 10- to 22-month-old infants by 40 shocks, CCTs help prevent the widening of percent.24 Mexico’s Prospera Program has inequality. helped lower infant mortality and maternal Evidence of the benefits of such interven- mortality by as much as 11 percent.25 The tions is encouraging. For example, in 1986, Nahouri Pilot Project in Burkina Faso is a Jamaican intervention sought to support credited with raising primary and secondary toddlers ages 9–24 months who suffered enrollment rates by 22 percent among boys.26 from stunting.20 The intervention consisted In Pakistan, CCTs made available only in of weekly visits to the households of the tod- favor of girls led to increases in enrollment in dlers by community health workers to teach the range of 11–13 percentage points.27 parenting skills aimed at fostering cognitive Also in Bangladesh, the Rural Develop- and socioemotional development among ment Program and the Rural Roads and the children. It also provided nutrition Markets Improvement and Maintenance supplements and psychosocial stimulation. Program have boosted employment and Researchers followed up among the partic- wages in agricultural and nonagricultural ipants 20 years after the intervention and activities, as well as aggregate harvest out- found that the groups of children receiving puts. Per capita annual spending across stimulation (with or without the nutrition households in the program areas has risen supplements) had, as adults, 25 percent by about 10 percent.28 In rural Vietnam, higher earnings than the control group. The school enrollment rates among children in greater earnings had allowed individuals households on the electricity grid were 9.0 in the stimulation program to enjoy liveli- percentage points higher among girls and hoods at a similar level as the members of 6.3 percentage points higher among boys a nonstunted comparison group, effectively relative to children in households not on eliminating the inequality in incomes be- the grid. Electrification was also associated tween the groups. with almost an extra year in the average Thailand’s Universal Coverage Scheme years of schooling among girls and an extra enhances equity by bringing a large unin- 0.13 year among boys.29 Similarly, access in sured population under the umbrella of a rural areas to telenovelas (television soap national insurance program, thereby greatly operas) resulted in lower fertility rates in reducing catastrophic health care payments Brazil, which may be related to the empow- and improving access to essential health erment of women through the imitation of services among the poor. Within a year of role models of emancipated women and the its launch, the scheme was covering 75 per- representation of smaller families.30

16 POVERTY AND SHARED PROSPERITY 2016 Such evidence demonstrates that inter- equalizing without compromising efficiency. ventions can be designed successfully in a Different choices in Chile and Mexico in variety of settings. Yet, the long road ahead recent tax reforms with the same objectives argues against any complacency and against led to different impacts. The ultrarich bore the fallacy of sweeping prescriptions. The the brunt of the income tax component of challenges and uncertainties are diverse and the reform in Chile, while in Mexico, the complicated, as follows: middle class also largely shared the cost of Despite progress, intolerable disparities in the reform.32 ECD programs are most effec- well-being still exist that concrete policy inter- tive if they are aimed at the first 1,000 days ventions could confront directly. In many low- of the lives of children, continue during and middle-income countries, preschool childhood, and integrate stimulation, par- enrollment rates among the poorest quin- enting, and nutrition components. In many tile are less than a third of the rates among contexts, incentivizing higher quality in the richest quintile. Mothers in the bottom teaching, while making social transfers con- 40 across developing countries are 50 per- ditional on school completion may have cent less likely to receive antenatal care. The a greater impact than constructing new poorest children are four times less likely schools. than the richest children to be enrolled in Avoid unexamined reliance on univer- primary education and systematically record sal prescriptions and unique models of suc- lower test scores than children in the richest cess. Evidence strongly suggests that the households. Among the estimated 780 mil- implementation of such prescriptions and lion illiterate adults worldwide, nearly two- models does not automatically ensure a thirds are women. Only one-quarter of the reduction in inequality. Nonetheless, some poorest quintile are covered by safety nets, initiatives are more likely than others to and the share is even smaller in Sub-Saharan generate inequality reductions and im- Africa and South Asia.31 provements in the well-being of the poor- Trade-offs in implementation should not est. For example, integrated interventions be overlooked because of excessive attention are more likely to succeed than isolated, to efficiency and equity trade-offs. Invest- monolithic interventions. Composition in- ments in ECD, universal health care, and fluences the degree of success. If CCTs are good-quality education have both equity combined with other safety net interven- and efficiency benefits given the current tions, such as transfers of productive assets, gaps in access. Connecting poor farmers to skills training, and access to credit and fi- urban markets can positively affect the in- nance, they have been shown to generate come of farm households as well as reduce wide-ranging benefits. Investments in rural their income gaps with the rest of the pop- roads that attract additional investments in ulation. In reducing inequality, many policy public services, such as electrification, ag- choices are less often restricted by an imbal- ricultural extension services, and enhanced ance in the equity-efficiency trade-off than water and sanitation services, improve not by an imbalance in the trade-offs between only the connectivity of people to eco- expanding the coverage of an intervention nomic opportunities, but also security, pro- and increasing the benefits, between en- ductivity, and the quality of services. Sim- hancing the quality of services and increas- plicity and flexibility often drive success. ing access to services through the construc- Thus, the ability of the safety net system in tion of facilities such as schools or clinics, the Philippines to scale up to reach hun- between expanding the coverage of electrifi- dreds of thousands of beneficiaries after cation in rural areas and ensuring program catastrophic events is in part explained by financial viability, between cash or in-kind the flexibility of the system in the face of resource transfers, and between condition- emergency situations. Exclusive and pro- ality and the lack of conditionality. longed breastfeeding is another example of The fine points of policy design absolutely a simple and extraordinarily cost-efficient matter in ensuring that interventions are intervention to improve ECD.33

OVERVIEW 17 Equalizing interventions are not a luxury keep electrification campaigns financially reserved for middle- and high-income coun- feasible, but this often means the poorest tries, nor an option only available during households must opt out.35 Policy design periods of prosperity. There are numerous needs to take such outcomes into account instances of the implementation of suc- up front and explicitly. cessful interventions in ECD, universal More knowledge! Despite the growing health care coverage, CCTs, investment in evidence on the impacts of policy interven- rural infrastructure, and redistributive tax tions, improving the evidence base on ini- schemes across low-income countries. This tiatives that successfully narrow inequality evidence should dispel the notion that only requires more investment in filling data gaps middle- and high-income countries can and enhancing the understanding of the spe- afford equalizing policies. Of course, con- cific pathways—whether intended or unin- text always matters: weak capacity, lack of tended—through which programs affect in- political will, restricted fiscal space, vulner- equality. For example, rigorous evaluations ability to external crises or climate change, have played a critical role in fine-tuning the internal conflict, and challenging geography design of CCTs and advocating for CCT are among the obstacles to the reduction of desirability. Monitoring ECD programs for inequality worldwide. These obstacles are decades has made the quantification of the not insurmountable, however. This is also long-term effects of such programs pos- the case during periods of crisis. Examples sible. Yet, the road ahead is still long and of CCTs integrated in safety nets that effec- steep. Especially important is the long-term tively protect the most vulnerable against generation of more microeconomic house- natural disasters demonstrate that a crisis is hold data, more compelling evidence on not an excuse for inaction, but an incentive the benefits of the integration of multiple for the adoption of equalizing interventions. interventions, and more information on the The poor must be able to participate in potential distributional effects of policy in- and benefit from interventions: good policy terventions aimed at addressing long-term choices benefit the poorest. Evidence on ECD challenges such as climate change.36 programs, initiatives to promote univer- Data need to allow for more compre- sal health care coverage, and efforts to fos- hensive monitoring of specific changes in ter good-quality teaching proves that the inequality, but also in poverty and shared most underprivileged children often bene- prosperity. Substantial efforts are required fit the most.34 Yet, this outcome should not to address the poor quality, comparability, be taken for granted. Thus, the more well and availability of data, especially in low- off households among the targeted pop- income countries. Figure O.14 shows the ulation, that is, households with children availability of poverty estimates by country with higher baseline levels of development and region. The availability is particularly and more well educated mothers, are typ- limited in Sub-Saharan Africa and in the ically more likely to send their children to Middle East and North Africa. This report preschool or to take part in parenting pro- makes a strong case for expanding the avail- grams. Many rural electrification initiatives ability of and access to data on inequality, are associated with high connection costs to poverty, and shared prosperity.

18 POVERTY AND SHARED PROSPERITY 2016 FIGURE O.14 Available Country Poverty Estimates, Number, by Region and Year

60 11 10 798 5 7 9 6 8 2 3 2 1 33 2 50 3 3 2 2 7 44 2 4 2 2 3 1 2 6 1 17 1 40 17 2 8 18 16 14 2 4 16 17 17 15 16 1 15 1 16 4 3 30 5 1 7 9 1 17 4 17 13 1 3 3 15 1 1 13 1 15 Number of poverty estimates 20 8 25 25 25 26 21 23 24 14 9 25 25 26 2 1 13 14 23 25 3 6 13 9 16 10 1 11 12 13 16 8 6 11 8 6 5 1 3 8 7 778 8 4 4 44 6 4 6 3 0 3 1 2 3 2 33332 3 3 2

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

East Asia and Paciﬁc Eastern Europe and Central Asia Latin America and the Caribbean Middle East and North Africa South Asia Sub-Saharan Africa

Source: Poverty and Equity Data (database), World Bank, Washington, DC, http://povertydata.worldbank.org/poverty/home/. Note: The presentation follows the deﬁnition of the developing world of Serajuddin et al. (2015), which includes 150 countries and territories in the early 1990s and 155 in 2013. High-income countries such as the original members of the Organisation for Economic Co-operation and Development (OECD), where extreme poverty is assumed to be zero, are not considered in this sample.

Notes sion on Global Poverty,” World Bank, Wash- ington, DC, http://www.worldbank.org/en 1. WHO (2015). /programs/commission-on-global-poverty. 2. For a formal decomposition, see Datt and 5. Poverty and Equity Data (database), World Ravallion (1992), among others. In the gen- Bank, Washington, DC, http://povertydata eral case, a reduction in inequality at a given .worldbank.org/poverty/home/. growth rate leads to a reduction in poverty 6. Castañeda et al. (2016) analyze the robust- according to most poverty measures. Ex- ness of these results by comparing different ceptions are, for instance, progressive dis- lineup methods and different ways to adjust tributional changes whereby some nonpoor welfare aggregates, weights, and poverty fall under the poverty line over time, thus lines. They find only minimal differences. increasing the headcount ratio. Even in that They also check for fixed effects and sensi- case, other poverty measures with higher tivity to missing data. The resulting demo- social welfare weights for lower percentiles graphic profile thus paints a robust picture tend to decrease. of global poverty. 3. See Oxfam (2016). Lakner (2015) estimates 7. Beegle et al. (2016) and Cornia (2014) doc- that the 10 wealthiest Africans own as much ument a bifurcation in inequality trends in as the poorest half of the continent. Sub-Saharan Africa, that is, within a set of 4. The countries classified as industrialized in African countries with at least two recent, this report are assumed to have zero poverty strictly comparable surveys, there is an even at the $1.90-a-day poverty line, an assump- split between countries with widening in- tion that may change in the future because equality and countries with narrowing in- of the World Bank’s implementation of the equality. The surveys are drawn from the report of the Commission on Global Poverty first decade of the 2000s. on global poverty estimation. See “Commis- 8. World Bank (2015a).

OVERVIEW 19 9. In Cambodia, these difficulties revolve 2014. “Challenges to Promoting Social In- around problems in land tenure, bottlenecks clusion of the Extreme Poor: Evidence from in fertilizer and seed markets, inadequate a Large Scale Experiment in Colombia.” IFS extension services and irrigation systems, Working Paper W14/33, Institute for Fiscal and, among farmers, the lack of savings and Studies, London. access to credit. In addition, recent agricul- ADB (Asian Development Bank). 2014. Cambo- tural productivity gains from an expansion dia: Diversifying beyond Garments and Tour- in the arable land under cultivation are not ism, Country Diagnostic Study. November. sustainable indefinitely and are thus unable Manila: Economics and Research Depart- to offset these problems. As a result, small- ment, ADB. holders are generally vulnerable to swings in Aggarwal, Shilpa. 2015. “Do Rural Roads Create the international prices for rice. In Tanzania, Pathways Out of Poverty? Evidence from the unfinished transition to a market-based India.” Working paper, Indian School of economy translates into a private sector Business, Hyderabad, India. characterized by a lack of competition, an Akresh, Richard, Damien de Walque, Harounan absence of growth, and a heavy regulatory Kazianga. 2013. “Cash Transfers and Child burden associated with the public sector. Schooling: Evidence from a Randomized See World Bank (2014, 2016a). Evaluation of the Role of Conditionality.” 10. World Bank (2016b). Policy Research Working Paper 6340, World 11. Barros et al. (2010); Osorio and Souza (2012). Bank, Washington, DC. 12. ADB (2014). Araujo, María Caridad, Pedro Carneiro, Yyannú 13. World Bank (2009, 2014). Cruz-Aguayo, and Norbert Schady. 2016. 14. Josz (2013). “Teacher Quality and Learning Outcomes in 15. World Bank (2015a). Kindergarten.” Quarterly Journal of Econom- 16. Genoni and Salazar (2015). ics 125 (1): 175–214. 17. World Bank (2015b). Asher, Sam, and Paul Novosad. 2016. “Market Ac- 18. World Bank (2016c). cess and Structural Transformation: Evi dence 19. Avram, Levy, and Sunderland (2014); De from Rural Roads in India.” Working paper Agostini, Paulus, and Tasseva (2015). (April 20), University of Oxford, Oxford. 20. Gertler et al. (2014); Grantham-McGregor Avram, Silvia, Horacio Levy, and Holly Suther- et al. (1991). land. 2014. “Income Redistribution in the 21. UNICEF (2016). European Union.” IZA Journal of European 22. Araujo et al. (2016). Labor Studies 3 (22): 1–29. 23. Fafchamps and Quisumbing (1999); Jolliffe Barros, Ricardo Paes de, Mirela De Carvalho, (1998); Yang (1997). Samuel Franco, and Rosane Mendonça. 2010. 24. Ferré and Sharif (2014). “Markets, the State, and the Dynamics of 25. Behrman and Hoddinott (2005); Gertler Inequality in Brazil.” In Declining Inequality (2004). in Latin America: A Decade of Progress?, ed- 26. Akresh, de Walque, and Kazianga (2013). ited by Luis F. López-Calva and Nora Lustig, 27. Fiszbein and Schady (2009). 134–74. New York: United Nations Develop- 28. Aggarwal (2015); Asher and Novosad (2016). ment Programme; Baltimore: Brookings In- 29. Khandker, Barnes, and Samad (2013). stitution Press. 30. La Ferrara, Chong, and Duryea (2012). 31. See the evidence presented in chapter 6. Beegle, Kathleen, Luc Christiaensen, Andrew 32. Abramovsky et al. (2014); World Bank Dabalen, and Isis Gaddis. 2016. Poverty in a (2016d). Rising Africa. Washington, DC: World Bank. 33. See the evidence presented in chapter 6. Beegle, Kathleen, Pedro Olinto, Carlos E. 34. See the evidence presented in chapter 6. Sobrado, Hiroki Uematsu, and Yeon Soo 35. See the evidence presented in chapter 6. Kim. 2014. “Ending Extreme Poverty and 36. Hallegate et al. (2016). Promoting Shared Prosperity: Could There Be Trade-Offs between These Two Goals?” Inequality in Focus 3 (1): 1–6. References Behrman, Jere R., and John Hoddinott. 2005. Abramovsky, Laura, Orazio Attanasio, Kai “Programme Evaluation with Unobserved Barron, Pedro Carneiro, and George Stoye. Heterogeneity and Selective Implementation:

20 POVERTY AND SHARED PROSPERITY 2016 The Mexican PROGRESA Impact on Child from PROGRESA’s Control Randomized Ex- Nutrition.” Oxford Bulletin of Economics and periment.” American Economic Review 94 (2): Statistics 67 (4): 547–69. 336–41. Bourguignon, François. 2015. The Globalization Gertler, Paul J., James J. Heckman, Rodrigo of Inequality. Translated by Thomas Scott- Pinto, Arianna Zanolini, Christel Vermeersch, Railton. Princeton, NJ: Princeton University Susan Walker, Susan M. Chang, and Sally Press. Grantham-McGregor. 2014. “Labor Market Castañeda, Andres, Dung Doan, David New- Returns to an Early Childhood Stimulation house, Minh C. Nguyen, Hiroki Uematsu, Intervention in Jamaica.” Science 344 (6187): João Pedro Azevedo, and Data for Goals 998–1001. Group. 2016. “Who Are the Poor in the De- Grantham-McGregor, Sally M., Christine A. veloping World?” Policy Research Working Powell, Susan P. Walker, and John H. Himes. Paper 7844, World Bank, Washington DC. 1991. “Nutritional Supplementation, Psycho- Cornia, Giovanni Andrea. 2014. “Income In- social Stimulation, and Mental Development equality Levels, Trends, and Determinants of Stunted Children: The Jamaican Study.” in Sub-Saharan Africa: An Overview of the Lancet 338 (8758): 1–5. Main Changes.” Università degli Studi di Hallegatte, Stephane, Mook Bangalore, Laura Firenze, Florence. Bonzanigo, Marianne Fay, Tomaro Kane, Datt, Gaurav, and Martin Ravallion. 1992. Ulf Narloch, Julie Rozenberg, David Treguer, “Growth and Redistribution Components of and Adrien Vogt-Schilb. 2016. Shock Waves: Changes in Poverty Measures: A Decomposi- Managing the Impacts of Climate Change on tion with Applications to Brazil and India in Poverty. Climate Change and Development Series. Washington, DC: World Bank. the 1980s.” Journal of Development Economics Jolliffe, Dean. 1998. “Skills, Schooling, and 38 (2): 275–95. Household Income in Ghana.” World Bank De Agostini, Paola, Alari Paulus, and Iva Val- Economic Review 12 (1): 81–104. entinova Tasseva. 2015. “The Effect of Tax- Josz, Christian. 2013. “Mali: Achieving Strong Beneﬁt Changes on the Income Distribution and Inclusive Growth with Macroeconomic in 2008–2014.” Euromod Working Paper Stability.” African Development 13/08, Inter- EM11/15, Institute for Social and Economic national Monetary Fund, Washington, DC. Research, University of Essex, Colchester, Khandker, Shahidur R., Douglas F. Barnes, and United Kingdom. Hussain A. Samad. 2013. “Welfare Impacts of Fafchamps, Marcel, and Agnes R. Quisumbing. Rural Electriﬁcation: A Panel Data Analysis 1999. “Human Capital, Productivity, and from Vietnam.” Economic Development and Labor Allocation in Rural Pakistan.” Journal Cultural Change 61 (3): 659–92. of Human Resources 34 (2): 369–406. La Ferrara, Eliana, Alberto Chong, and Suzanne Ferré, Céline, and Iffath Sharif. 2014. “Can Con- Duryea. 2012. “Soap Operas and Fertility: ditional Cash Transfers Improve Education Evidence from Brazil.” American Economic and Nutrition Outcomes for Poor Children Review 4 (4): 1–31. in Bangladesh? Evidence from a Pilot Project.” Lakner, Christoph. 2015. “The Ten Richest Af- Policy Research Working Paper 7077, World ricans Own as Much as the Poorest Half Bank, Washington, DC. of the Continent.” Let’s Talk Development Fiszbein, Ariel, and Norbert Schady. 2009. Con- (blog), March 11. http://blogs.worldbank.org ditional Cash Transfers: Reducing Present and /developmenttalk/ten-richest-africans-own Future Poverty. Washington, DC: World Bank. -much-poorest-halfcontinent. Genoni, María Eugenia, and Mateo Salazar. 2015. Lakner, Christoph, and Branko Milanovic´. 2016. “Steering toward Shared Prosperity in Peru.” “Global Income Distribution: From the Fall In Shared Prosperity and Poverty Eradication of the Berlin Wall to the Great Recession.” in Latin America and the Caribbean, edited World Bank Economic Review 30 (2): 203–32. by Louise Cord, María Eugenia Genoni, and Lakner, Christoph, Mario Negre, and Espen Beer Carlos Rodríguez-Castelán, 269–302. Wash- Prydz. 2014. “Twinning the Goals: How Can ington, DC: World Bank. Shared Prosperity Help to Reduce Global Gertler, Paul J. 2004. “Do Conditional Cash Poverty?” Policy Research Working Paper Transfers Improve Child Health? Evidence 7106, World Bank, Washington, DC.

OVERVIEW 21 Milanovic´, Branko. 2014. “All the Ginis, 1950– of Cambodia: Proposed Smallholder Agri- 2012 (updated in Autumn 2014).” Novem- culture and Social Protection Support Op- ber, World Bank, Washington, DC. http:// eration.” Report 48083-KH (June 19), World go.worldbank.org/9VCQW66LA0. Bank, Washington, DC. ———. 2016. Global Inequality: A New Approach ———. 2014. “Where Have All the Poor Gone? for the Age of Globalization. Cambridge, MA: Cambodia Poverty Assessment 2013.” Report Harvard University Press. ACS4545, World Bank, Washington, DC. Newhouse, David, Pablo Suarez-Becerra, Martin ———. 2015a. “Republic of Mali, Priorities C. Evans, and Data for Goals Group. 2016. for Ending Poverty and Boosting Shared “New Estimates of Extreme Poverty for Chil- Prosperity: Systematic Country Diagnostic dren.” Policy Research Working Paper 7845, (SCD).” Report 94191-ML (June 22), World World Bank, Washington, DC. Bank, Washington, DC. Osorio, Rafael Guerreiro, and Pedro H. G. ———. 2015b. “Tanzania Mainland: Poverty As- Ferreira de Souza. 2012. “O Bolsa Família de- sessment.” Report AUS6819 v2, World Bank, pois do Brasil Carinhoso: uma análise do po- Washington, DC. tencial de redução da pobreza extrema.” Insti- ———. 2016a. “Ngazi Ya Pili (To the Next tute for Applied Economic Research, Brasília. Level): United Republic of Tanzania Sys- Oxfam. 2016. “An Economy for the 1%: How tematic Country Diagnostics.” Regional Op- Privilege and Power in the Economy Drive erations Committee Decision Review Draft Extreme Inequality and How This Can Be (June 2), World Bank, Washington, DC. Stopped.” Oxfam Brieﬁng Paper 210, Oxfam ———. 2016b. Brazil Systematic Country Diag- GB, Oxford. nostic: Retaking the Path to Inclusion, Growth, Serajuddin, Umar, Hiroki Uematsu, Christina and Sustainability. Report 101431-BR, Wash- Wieser, Nobuo Yoshida, and Andrew Dabalen. ington, DC: World Bank. 2015. “Data Deprivation: Another Depriva- ———. 2016c. “International Development tion to End.” Policy Research Working Paper Association Project Paper, United Republic 7252, World Bank, Washington, DC. of Tanzania: Productive Social Safety Net UNICEF (United Nations Children’s Fund). (PSSN).” Report PAD1500 (May 20), World 2016. The State of the World’s Children 2016: Bank, Washington, DC. A Fair Chance for Every Child. June. New ———. 2016d. “Chile: Distributional Effects of York: UNICEF. the 2014 Tax Reform,” World Bank, Washing- WHO (World Health Organization). 2015. ton, DC. World Health Statistics. Geneva: WHO. Yang, Dennis. 1997. “Education and Off-Farm World Bank. 2009. “International Development Work.” Economic Development and Cultural Association Program Document, Kingdom Change 45 (3): 613–32.

22 POVERTY AND SHARED PROSPERITY 2016 1 Poverty and Shared Prosperity: Setting the Stage

Concepts, measurement, and data

The World Bank goals Measuring the World Bank goals On April 20, 2013, the Board of Executive Directors of the World Bank adopted two Poverty ambitious goals: end extreme poverty glob- The first World Bank goal, ending poverty, ally and promote shared prosperity in every is global, and measurement involves adding country in a sustainable way. up the number of poor people in all coun- Progress toward the first of these goals tries. Three specific components are critical is measured by monitoring the rate of ex- to this aggregation. The first component is treme poverty using the international pov- cost-of-living comparability across coun- erty standard. The World Bank set a target tries. The key information for ensuring this of reducing the poverty headcount ratio comparability is data to gauge purchasing from 12.4 percent globally in 2012 to 3.0 power parities (PPPs). The data are col- percent by 2030, and, to avoid overreliance lected through the International Compar- on efforts toward the end of the period, the ison Program, an independent worldwide institution set an interim target of 9.0 per- statistical partnership that gathers data on cent by 2020, consistent with an annual re- prices and expenditures within economies. duction in global extreme poverty of a half The comparison of incomes across house- 1 percentage point over 2012–20. holds in different countries entails the use of The second goal is not defined globally, exchange rates between local currencies, the but tracks progress at the level of countries. units of value used by household surveys. Despite not providing a single global target, However, the use of market exchange rates the shared prosperity goal is universal be- is not sufficient because market exchange cause it includes developed countries that rates do not accurately assess differences in do not contribute to the global poverty purchasing power and are inaccurate in cap- count, but are still monitored in terms of turing the costs faced by the poor. Despite shared prosperity. Progress on the shared the limitations stemming from the com- prosperity goal is measured by the growth plexity of a global exercise of this nature, the in the average income or consumption ex- PPPs fill this gap by allowing income com- penditure of the poorest 40 percent of the parisons in real terms, while accounting for population distribution (the bottom 40) differences in prices across countries. Com- within each country.2 This goal is not asso- paring prices across the world is a tricky en- ciated with a target in 2030, but it reflects deavor because most products are not sold the aim that every country should promote everywhere, or, even if they are, the quality the welfare of its least privileged citizens. may vary considerably country-by-country

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 23 and region-by-region. Moreover, an every- although a nonnegligible share of house- day product in one country may be a lux- holds in these countries report incomes ury item in another. Likewise, prices may below the international poverty line. The also vary widely within a single country; United States is a good example. According so, spatial price variations should also be to some estimates, 1 percent to 4 percent of taken into account where possible. Because the population there (measured by income) relative prices evolve, updates of PPPs may is living below a $2.00-a-day poverty line (in be necessary so that PPP exchange rates are 2005 prices).6 If welfare is measured by con- used reliably in price comparisons. sumption, however, high-income countries A second key component of the aggrega- typically report no poverty at all. This is be- tion of the poor is the definition of an inter- cause the poor in these countries generally national poverty line in PPP terms. Coun- have access to free public services and social tries determine their own poverty lines transfers that are not accounted for in the mostly by means of a basic needs standard income aggregate, but that ensure that the that is linked to a predefined consumption population’s consumption levels are above basket of essential goods and services, or the recognized poverty standards. Box 1.1 relative to an agreed position along the dis- discusses key issues in the use of incomes tribution of income or consumption (for and consumption as welfare aggregates. example, 60 percent of the median national household income). The World Bank uses Shared prosperity an international poverty line based on the To analyze both prosperity and equity dy- national thresholds of some of the poorest namically, the World Bank focuses, in the countries.3 To account for changes in rela- second goal, on the growth experienced by tive prices across countries over time, the people at the bottom of the income or con- 2011 PPP was adopted in 2015 as the stan- sumption distribution within a country. In dard. This involved an upward adjustment practice, this is defined as the growth in the of the international poverty line to US$1.90 average income or consumption of the bot- a day in income or consumption expendi- tom 40, which is measured through house- ture so as to preserve the real value of the hold surveys.7 Although the choice of the US$1.25-a-day line in 2005 PPP that was bottom 40 is somewhat arbitrary, the bot- the previous international line. Global pov- tom 40 has been a focus of poverty analysis erty incidence changes little as a result, and for quite some time.8 Indeed, the bottom 40 progress toward the global poverty target threshold reflects a commitment to focus for 2030 is unaffected.4 on the least well off regardless of the actual The third element in a proper aggrega- poverty rate. In cases where the national tion of the poverty count is the treatment of poverty headcount is above 40 percent, the the lack of reliable data on some economies shared prosperity indicator centers on the and a range of difficulties in the aggrega- poorest of the poor. tion of existing and available data. Hence, Why choose an income growth indica- for those countries lacking reliable income tor that focuses only on the bottom of the or consumption data or on which data are distribution? One of the most widely used not available for analysis, a regional average indicators of economic prosperity is growth poverty rate is used. In calculating this rate, in per capita gross domestic product (GDP). available and reliable evidence is gathered The economic health and development of on countries within the same region and countries are often judged on this basis. then extrapolated onto the countries on However, a per capita growth rate says lit- which data are unavailable.5 tle about how specific population groups These three ingredients permit the esti- benefit from economic progress. Similarly, mation of the global count of the poor. Yet, standard indicators of inequality, such as the global aggregation exercise rests on var- the Gini index, do not offer information on ious assumptions. Most significantly, zero the size of the pie being shared. poverty is assumed in most high-income The shared prosperity indicator therefore countries, that is, industrialized countries, represents a practical compromise between a

24 POVERTY AND SHARED PROSPERITY 2016 BOX 1.1 The Welfare Aggregate: Income versus Consumption

The global poverty headcount requires such as the frequent case of households that the number of the poor within a that declare zero income on a survey, country be measured by adding up the but exhibit a consumption level that poor based on a welfare aggregate is not zero. This may occur because obtained through household surveys. the households lack income during a In most countries, the aggregate of survey recall period, are dissaving, or are choice is per capita consumption. experiencing a spell of unemployment, Indeed, 75 percent of the countries in or because the consumption of home- the World Bank PovcalNet database— produced goods has not been correctly the ofﬁcial online repository of World measured. Bank poverty data—use this aggregate.a In contrast, consumption does not The countries in the database that normally vary as widely; it displays use incomes are mostly high-income a much smoother behavior. For this countries and Latin America and reason, consumption tends to be the Caribbean countries. preferred aggregate in measuring Are these two aggregates—income poverty in developing economies, which and consumption—the same? They typically depend more on agriculture and are not. Conceptually, income is a have a larger informal sector. measure of the potential set of all Despite these differences, both goods and services that an individual aggregates are used indistinctively in the or a household could obtain based on measurement of the World Bank goals their purchasing power. Meanwhile, to maximize the number of countries consumption represents a direct monitored. Although this creates issues measure of the goods and services that of comparability in the measurement the individual or household has actually of poverty, it allows the coverage of obtained. Therefore, consumption does the global goals to be expanded. The not capture opportunities, but realized distinction may be more problematic, outcomes that directly determine an however, in the analysis of inequality. individual or household’s well-being. This is a result of the fact that the In practice, income is generally more coverage of household surveys is volatile in the sense that it may be generally incomplete among top earners, inﬂuenced greatly by seasonal factors entrepreneurial and capital incomes are or by a lack of regularity, particularly in inadequately reported, and measures of agriculture and in the informal sector. It consumption often underestimate the also has other important shortcomings, living conditions of the rich.

a. PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org /PovcalNet/.

prosperity indicator that fails to capture in- matter the circumstances, the country con- equality and an equity indicator that fails to text, or the time period.9 capture growth. By focusing on the growth in incomes of the bottom 40, the shared Linking shared prosperity and prosperity indicator incorporates a measure inequality: the shared prosperity of prosperity and distributional dimensions. premium Furthermore, by choosing this indicator to Tracking the growth rate of the average in- measure one of its goals, the World Bank fo- come of the poorest two-ﬁfths of the pop- cuses squarely on improving the welfare of ulation is crucial for assessing the inclusive the least well off within every country across nature of growth. However, taking this mea- the world, ensuring that everyone is part of sure a step further and comparing it to the a dynamic and inclusive growth process, no performance of the mean provides import-

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 25 ant additional insights. Whether the income tended metric of growth and not an in- or consumption of the bottom 40 grows equality indicator. more quickly or more slowly than the aver- Rewriting equation 1.1, one may easily age determines whether or not the bottom appreciate that the shared prosperity pre- 40 disproportionately benefit from growth mium represents the change in the total in- relative to the average individual in society. come share of the bottom 40, as follows: If incomes among the bottom 40 grow at a g = g – g = the shared prosperity rate above the rate at the mean, this implies shareB40 40 mean premium (1.2) that the incomes of the bottom 40 are rising more rapidly than the incomes among the where gshareB40 is the growth in the income rest of the population, that is, the top 60. share of the bottom 40; g40 is the income This concept has been embodied in the growth of the bottom 40; and gmean is the shared prosperity premium, an indicator growth of the mean. that is defined as the difference between the While the shared prosperity indicator is growth in the incomes or the consumption useful in gauging the progress in achieving of the bottom 40 and the growth in the in- the ambition of leaving no one behind, a comes or consumption of the mean.10 positive value in the shared prosperity pre- The shared prosperity goal thus pos- mium ensures an extra focus on the partic- sesses an inequality dimension even though ular progress of those who are most in need. it is not an inequality indicator. This di- It shows that the relative growth in incomes mension can be readily assessed by com- or consumption among this population paring growth among the bottom 40 with segment is larger than that observed among growth at the mean. The difference is posi- the rest of the distribution. tive if the former grows more quickly than the latter and negative otherwise. The indi- Poverty, shared prosperity, cator can easily be used to provide a rough and the global goals assessment of the magnitude of changes in inequality. One may also decompose the The World Bank’s first goal—reducing the sources of the growth measured: the growth poverty rate to 3.0 percent by 2030—en- of the bottom 40 may derive from growth compasses the world. Yet, not all countries in the mean income (or consumption) of or regions or groups within countries can the overall population, or it may arise from be expected to reach the goal individually. changes in the share of overall income that However, an even more ambitious target, accrues to the bottom 40 according to the eradicating extreme poverty “for all people following accounting relationship:11 everywhere” by 2030, was adopted by the United Nations General Assembly in Sep- g40 = gmean + gshareB40, (1.1) tember 2015 as part of the first Sustainable

where g40 is the income growth of the bot- Development Goal. tom 40; gmean is the growth of the mean; and This slight difference in targets between gshareB40 is the growth in the income share of the World Bank’s goal and the Sustain- the bottom 40. able Development Goal—3 percent versus The growth in income or consump- zero—is not accidental. As countries have tion among the bottom 40 may therefore eradicated extreme poverty, the speed at be associated with participation in average which they have been able to eliminate growth, with a rising share of the bottom 40 the last pockets of poverty has varied con- in total income or consumption, or, ideally, siderably. Some countries have witnessed with both so that achievements through a slowdown in poverty reduction in the one channel are not offset by poor perfor- last stages of the process, such as Austra- mance in the other. Such a decomposition lia, Canada, the United Kingdom, and the synthesizes the distributional character of United States.12 Moreover, even if most the shared prosperity indicator, although countries signiﬁcantly reduce poverty by the indicator, strictly speaking, is an ex- 2030, the persistence of conﬂicts and disas-

26 POVERTY AND SHARED PROSPERITY 2016 ters means there will likely remain a small, may drive significant changes in the shared but significant number of people living in prosperity indicator.15 extreme poverty. In part for this reason, Amid such issues, the availability of the World Bank goal of reducing poverty good-quality household surveys is perhaps is restricted to narrowing the share of the the most critical need. Indeed, limited avail- world population living below the interna- ability or a complete lack of household sur- tional poverty line to less than 3.0 percent veys not only hampers efforts to monitor by 2030. poverty and shared prosperity, but also the The United Nations has also built on the capacity to design policies and interven- concept of shared prosperity for Sustainable tions that can reach those who need them Development Goal 10, on inequality. This the most. Household surveys therefore need development goal aims to “progressively to be undertaken with regularity, must be achieve and sustain income growth of the comparable from one round to another, bottom 40 percent of the population at a and must be of good quality. Meeting any of rate higher than the national average” by these requirements is not common practice 2030 (target 10.1).13 The United Nations is in many developing countries. using the shared prosperity premium, de- Figure 1.1 illustrates the status of house- fined above, to measure progress. hold survey availability among develop- In addition, in 2016, the G20 adopted ing countries by reporting the number the shared prosperity premium indicator of national poverty estimates available to monitor the progress of its members in each year. Household survey data availabil- making their economies more inclusive. ity has considerably improved over time. The global community, by putting these How ever, poverty and welfare estimates are issues at the forefront of policy making, has typically not available immediately after a an unprecedented opportunity to help push survey has been conducted because pro- governments and citizens into action to end cessing the data with confidence takes time. extreme poverty and promote more inclu- This is the cause of the precipitous drop sive societies. in available poverty estimates in 2014 and 2015: household surveys collected in those The importance of data in years have not yet yielded an official, val- tracking progress and spurring idated poverty estimate. Global poverty action estimates are therefore typically reported with a three-year lag, although ambitious Monitoring progress on the two World efforts are under way to ensure more fre- Bank goals involves a massive effort in data quent household surveys and hopefully collection and harmonization. Though not allow for more up-to-date reporting in the the only essential element, national house- future.16 hold surveys are mandatory in estimating To monitor the trends in poverty, sur- the proportion of people who are living in veys need to be conducted regularly, at least extreme poverty based on the international every three years or even more frequently. poverty line and in gauging the extent to Figure 1.2 shows the data availability during which the bottom 40 benefit from eco- all possible 10-year periods between 1990 nomic growth.14 and 2013. The progress is evident. The While measurement of the shared pros- number of countries on which there is no perity goal is not associated with specific poverty estimate over any 10-year period additional data requirements relative to the fell from 50 in 1990–99 to 21 in 2004–13. poverty goal, and while it does not rely on That the progress achieved so far is insuf- the use of PPPs, it does pose specific chal- ficient is equally evident. In 2004–13, 74 lenges. Thus, within countries, slight meth- countries still lacked the basic data required odological changes in survey questionnaires to monitor poverty and shared prosperity from round to round or in the composition over two points in time within a 10-year of the consumption or income indicator span. Indeed, these countries have no data

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 27 FIGURE 1.1 Available Country Poverty Estimates, by Region and Year 70

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

East Asia and Paciﬁc Eastern Europe and Central Asia Latin America and the Caribbean Middle East and North Africa South Asia Sub-Saharan Africa

Source: Poverty and Equity Data (database), World Bank, Washington, DC, http://povertydata.worldbank.org/poverty/home/. Note: The presentation follows the deﬁnition of the developing world employed by Serajuddin et al. (2015), which includes 150 countries and territories in the early 1990s and 155 in 2013. High-income countries such as the original members of the Organisation for Economic Co-operation and Development (OECD), where extreme poverty is assumed to be zero, are not considered in this sample.

FIGURE 1.2 Availability of Poverty Data, by All Possible 10-Year Periods, 1990–2013

160

140

120 63 62 62 66 68 67 75 76 81 80 82 81 84 84 81 100

80 37 42 47 44 60 45 51

Number of countries 46 52 49 50 48 51 49 50 53 40

50 46 20 42 41 38 33 30 25 23 24 24 22 22 21 21 0

1990–99 2000–09 2001–10 2002–11 2003–12 2004–13 1991–2000 1992–2001 1993–2002 1994–2003 1995–2004 1996–2005 1997–2006 1998–2007 1999–2008 Deprived (0 data points) Deprived (1 or 2 data points, more than 5 years apart) Non-deprived

Source: Poverty and Equity Data (database), World Bank, Washington, DC, http://povertydata.worldbank.org/poverty/home/.

28 POVERTY AND SHARED PROSPERITY 2016 points at all, have only one data point each, upgrade and expand the availability of ap- or have two data points, but more than five propriate data, particularly where these are years apart. This constitutes the definition needed the most. In 2015, the World Bank of a country that is data deprived, which is committed to supporting the poorest coun- the case of these 74 countries today. tries in undertaking at least one household It is also worrisome that data depriva- survey every three years and is now working tion is unequally distributed worldwide. closely with governments and partner insti- The data gaps within regions vary widely. tutions to broaden and improve data qual- In Eastern Europe and Central Asia, 28 of ity and availability in other crucial areas.19 30 countries have at least three data points in the most recent 10-year period, 2004–13. A special focus in 2016: At the other extreme, most Sub-Saharan taking on inequality African countries only have one survey for the entire 10-year span: a total of 35 of 48 Preceding the launch of the new World countries in Sub-Saharan Africa are data Bank goals by several years, World Develop- deprived. The Middle East and North Africa ment Report 2006: Equity and Development region and the East Asia and Pacific region described the extraordinary inequality of are also severely data deprived. In Latin opportunity across the world.20 Since 2006, America and the Caribbean, countries typ- substantial progress has been made in re- ically either have excellent survey coverage ducing poverty, expanding prosperity, com- (with a data point for almost every year) or bating hunger, enhancing education and no data points at all (often, in the Carib- health services, and lowering maternal and bean subregion). child mortality. Do the large inequalities Zero data points does not necessarily of 10 years ago persist today? What can be mean that no survey has been carried out. done to eliminate them? It may be that surveys are deemed insuffi- The World Bank devotes this first flag- cient in quality and comparability or that ship report on poverty and shared prosper- access to the data is not granted to the ity to addressing these questions. Although World Bank and its partners. The following the findings indicate that there has been statistics provide an illustration of the dire progress in establishing greater equality effects of one of these issues, comparability. across the globe, there is also no room for If one discounts the surveys in Sub-Saharan complacency: we continue to live in a world Africa that are not nationally representative, characterized by intolerable inequality of that were not conducted at similar times of opportunity, gender disparities, and depri- the year (to control for seasonality in con- vations, particularly in health, education, sumption patterns), and that use different safe water and sanitation, nutrition, and instruments in collection, the typical Afri- consumption. can country would have conducted only 1.6 The issue of the sustainability of pros- comparable surveys in the 22 years between perity is as relevant today as 10 years ago. 1990 and 2012.17 People worldwide are paying closer atten- Multiple challenges drive data depri- tion and taking and demanding action to vation. The funding of surveys is often ad tackle a broad range of inequalities more hoc and unpredictable, reflecting lack of effectively. Today, few stakeholders ques- commitment, competing priorities, poor tion that equality of opportunity should be coordination, or a combination of all these a central concern in the design and imple- factors. Weak technical capacity, weak gov- mentation of policies to foster development ernance in data collection, and a lack of and inclusive growth. globally accepted methodological standards Taking on inequality implies that pros- also pose severe challenges.18 perity must be shared meaningfully within The World Bank’s commitment to mon- developed and developing countries. The itoring extreme poverty and shared pros- less well off must be at center stage if so- perity therefore represents a critical call to cieties worldwide are to achieve the sta-

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 29 bility and well-being to which they aspire. are experiencing double or triple dips in Growth alone will not do the job; any growth. The first dip took place in 2009, type of growth will not suffice. Growth when the per capita GDP growth of these must be sustainable to achieve the maxi- economies plummeted to negative values or mum possible increase in living standards zero, followed by a strong rebound in 2010. among the less well-off. Growth then declined again in these econ- Reducing deprivations among the poor- omies, marking a second dip. In the United est, increasing their living standards, and States, the dip was short-lived because the leveling the playing field correspond to economy rebounded again in 2012, but notions of fairness and justice and reso- slowed once more in 2013 (the third dip). nate across societies on their own merits. In Europe, the rebound took longer and Nonetheless, continuing the notable prog- materialized in 2014. The BRICS economies ress in poverty reduction and shared pros- have been slowing since 2011. perity that took place over the first decade This report finds that the goal of elimi- of the 2000s is at risk because of the some- nating extreme poverty will not be achieved what lower growth today and the projected by 2030 without significant shifts in within- global slowdown. In the last few years, eco- country inequality. Ultimately, poverty nomic growth has been sluggish worldwide, reduction can occur through higher aver- a period typically described as the Great Re- age growth, a narrowing in inequality, or cession. Figure 1.3 illustrates the pattern of a combination of the two. So, if poverty slowing growth among the main economic reduction is to be achieved in a context of powerhouses, such as Brazil, the Russian slow growth, such as the current context, Federation, India, China, and South Africa more equitable income or consumption (together known as BRICS), the European distribution will be required. Union (EU), and the United States, in which Taking on inequality also matters for per capita GDP growth rates are exhibiting other reasons. It can bolster economic recurrent ups and downs. These economies growth if it is carried out smartly.21 There is no inevitable trade-off between efficiency and equity.22 For example, investing in early FIGURE 1.3 Double and Triple Dips in Growth, Selected Economies, childhood development (ECD) among un- 2006–15 derprivileged children can help prevent the emergence of inequalities in cognitive 150 development and health status and, by enabling a more successful accumulation 100 of human capital among individuals who 50 would otherwise lag throughout the rest of the life cycle, improve future earnings and 0 life chances. Trimming down inequalities in opportunity and outcomes among indi- –50 viduals, populations, and regions may also generate additional benefits through polit- –100 ical and social stability and social cohesion. This inaugural flagship report on poverty –150 and shared prosperity documents trends in –200 income inequality, identifies recent country experiences that have been successful –250 in reducing such inequalities, analyzes key Index of annualized GDP per capita growth rates (2006 = 100) 2006 20072008 2009 2010 2011 2012 2013 2014 2015 lessons drawn from these experiences, and synthesizes the rigorous evidence on public United States European Union BRICS policies that can shift inequality to bolster Source: WDI (World Development Indicators) (database), World Bank, Washington, DC, http://data.world poverty reduction and shared prosperity. bank.org/data-catalog/world-development-indicators. Note: The data on the European Union and BRICS are population-weighted averages of the respective Specifically, the report addresses the follow- national per capita GDP growth rates. ing questions:

30 POVERTY AND SHARED PROSPERITY 2016 • What is the latest evidence on the levels fusion on global trends, which can often be and evolution of global extreme poverty properly explained simply by selecting the and shared prosperity? appropriate inequality indicator, country, • Which countries and regions have been or time period for analysis. more successful in terms of progress to- Chapter 5 briefly discusses the driv- ward these goals and which are lagging? ers that explain recent notable successes in • What does the global context of lower boosting shared prosperity and reducing economic growth mean for the poorest inequality within countries. What have suc- people and countries? cessful countries done to reduce income in- • How can narrowing inequality contribute equality, while inequality has been widening to ending extreme poverty and improving in other economies? Among the constella- the welfare of the least well off? tion of policies that have been implemented, • What does the evidence show concerning this chapter identifies key levers in reduc- global and between- and within-country ing income inequality and boosting shared inequality trends? prosperity in selected countries. Country • What does the evidence tell us about how cases assess how macroeconomic manage- countries and interventions have success- ment, sectoral reforms, the expansion of fully reduced inequality? safety nets, responses to external shocks, and initial conditions or context all contribute To address these questions, the report to boost shared prosperity. Five countries is organized in six chapters, including this have been chosen based on their success in introduction. Chapter 2 presents the latest sharing prosperity and reducing inequality: numbers on global and regional poverty Brazil, Cambodia, Mali, Peru, and Tanza- using the international extreme poverty line nia. These countries have also been chosen of US$1.90 (2011 PPP). The chapter also with the aim of achieving diversity in terms discusses the geographical concentration of of region, income group, and distinctive poverty and the composition of global pov- contexts. erty, complementing the global headcount Chapter 6 explores the drivers of inequal- numbers with a profile of global poverty. ity reduction by examining interventions Chapter 3 presents the latest shared rather than country-specific experiences. prosperity data on the largest possible set of It reviews emerging as well as consolidated countries on which at least two comparable evidence on how interventions successfully data points are available circa 2008–13, that address the roots, drivers, and outcomes is, 83 countries covering 75 percent of the of inequality, while supporting economic world’s population. The analysis identifies growth and poverty reduction. The chap- the best and worst performers in terms of ter focuses on early childhood develop- the growth of the bottom 40. The chapter ment (including breastfeeding), universal also shows that, in the context of declining health care, quality teaching, conditional global and regional growth rates and an cash transfers (CCTs), investments in rural outlook of weaker growth, the continuation infrastructure (roads and electrification), of the progress achieved so far toward end- and taxation. The choice of policy inter- ing poverty by 2030 is at serious risk. Pov- ventions is highly selective and responds erty trajectories to 2030 are simulated under to the availability of rigorous evidence that several scenarios of growth and changes in supports reliable lessons and reflects some shared prosperity, providing insights on the of the policy areas highlighted by the coun- additional reductions in inequality required try narratives in chapter 5. The interven- to end poverty. tions chosen confirm the multiple pathways Chapter 4 focuses on income inequality. through which policies can successfully It presents the most recent trends on global promote equality. The chapter also provides and between- and within-country inequal- several general lessons useful in establishing ities and contrasts them with secular long- a basic agenda of interventions that over- term trends. The analysis separates fact come equity-efficiency and implementation from fiction and clarifies much of the con- trade-offs.

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 31 Notes and ensure geographical comparability; and national accounts and administrative data, 1. Basu (2013). which can be used, for example, to adjust for 2. Unless otherwise indicated, income under reported incomes or consumption. and consumption are henceforth used 15. Estimates of the growth in income or con- interchangeably. sumption of the bottom 40 are based on 3. The methodology was originally applied by comparisons at two points in time, which Ravallion, Chen, and Sangraula (2008), who may include years that do not correspond to established a first set of countries. the most recent survey in all countries. The 4. Ferreira et al. (2016). growth rates must therefore be annualized, 5. World Bank (2015a). and, although this makes them more readily 6. Chandy and Smith (2014). comparable, the operation is only performed 7. In this report, shared prosperity is estimated on rates in those countries on which more as the growth rate of real survey per capita than two surveys are available within the income or consumption among the bottom period under analysis. 40 over a five-year period that ends three 16. United Nations (2014). years before the publication of the report. 17. Beegle et al. (2016). Only countries with surveys meeting the 18. World Bank (2015a). following criteria are included: (a) the lat- 19. World Bank (2015b). est household survey for the country falls 20. World Bank (2005). between 2011 and 2015 and (b) the survey 21. Interventions aimed at narrowing inequality for the first year of the period is the nearest can also result in economic distortions, thus survey collected five years before the most affecting efficiency and, ultimately, growth. recent survey available; thus, only surveys 22. Interventions that reduce inequality without collected between three and seven years be- compromising economic growth are de- fore the most recent survey are considered in scribed here generically as equity-enhancing the inclusion or exclusion of countries. policies. World Development Report 2006: 8. Among the prominent references to the Equity and Development (World Bank 2005) bottom 40, see Kuznets (1955); McNamara calls them efficient redistribution policies. (1972). The change in the term in this report aims to 9. Saavedra-Chanduvi (2013). avoid any confusion around the notion of 10. Lakner, Negre, and Prydz (2014). redistribution. Redistribution may be strictly 11. Where gshareB40 denotes the growth rate of understood as removing resources from one the income share of the bottom 40. Equation individual or group and giving them to an- 1.1 follows from differentiating the follow- s other individual or group. However, not all m 40 ∗ m ing identity: 40 = , where s40 is the equity-enhancing policies must rely on re- 0.4 income share of the bottom 40, and m and distribution in this strict sense. Some may m 40 are the overall mean and the mean of achieve progressive distributional changes bottom 40, respectively. Equation 1.1 holds simply through legislation. For example, leg- in continuous time; as t becomes larger, it islation on the minimum wage does not in- becomes an approximation. Also see Dollar, volve any direct government redistribution Kleineberg and Kraay (2016); Rosenblatt of resources, but often nonetheless affects and McGavock (2013). the distribution of income. Moreover, the 12. World Bank (2015a). term redistribution is frequently used among 13. Sustainable Development Knowledge Plat- pundits to refer to anonymous changes in form (database), Department of Economic the distribution regardless of their cause. Ef- and Social Affairs, United Nations, New York, ficiency is also subject to alternative interpre- https://sustainabledevelopment.un.org/. tations. The arguments presented in this re- 14. The following are also essential: censuses, be- port that tackle inequality draw from an idea cause they make possible the design of the of efficiency developed by Kaldor and Hicks household surveys; price data, to adjust in the late 1930s whereby an increase in over- trends in income and consumption over time all incomes may be desirable even at the cost

32 POVERTY AND SHARED PROSPERITY 2016 of a few losers (Hicks 1939; Kaldor 1939). In McNamara, Robert S. 1972. “Annual Address by contrast, efficiency can also be understood in Robert S. McNamara, President of the Bank the Pareto (1906) sense that no one can be and Its Affiliates.” Summary Proceedings, An- made more well off without hurting some- nual Meetings of the Boards of Governors, Re- one else. The efficient equity-enhancing ini- port 53408, 16–31. Washington, DC: World tiatives discussed in this report are based on Bank. the concepts of Kaldor and Hicks. Pareto, Vilfredo. 1906. Manuale di economia politica: Con una introduzione alla scienza References sociale. Milan: Società editrice libraria. Ravallion, Martin, Shaohua Chen, and Prem Basu, Kaushik. 2013. “Shared Prosperity and the Sangraula. 2008. “Dollar a Day Revisited.” Mitigation of Poverty: In Practice and Pre- Policy Research Working Paper 4620, World cept.” Policy Research Working Paper 6700, Bank, Washington, DC. World Bank, Washington, DC. Rosenblatt, David, and Tamara J. McGavock. Beegle, Kathleen, Luc Christiaensen, Andrew 2013. “A Note on the Simple Algebra of the Dabalen, and Isis Gaddis. 2016. Poverty in a Shared Prosperity Indicator.” Policy Research Rising Africa. Africa Poverty Report. Washing- Working Paper 6645, World Bank, Washing- ton, DC: World Bank. ton, DC. Chandy, Laurence, and Cory Smith. 2014. “How Saavedra-Chanduvi, Jaime. 2013. “Why Didn’t Poor are America’s Poorest? U.S. $2 a Day the World Bank Make Reducing Inequality Poverty in a Global Context.” Policy Paper One of Its Goals?” Let’s Talk Development 2014–03, Global Economy and Development (blog), September 23, https://blogs.world at Brookings, Brookings Institution, Wash- bank.org/developmenttalk/why-didn-t-world ington, DC. -bank-make-reducing-inequality-one-its Dollar, David, Tatjana Kleineberg, and Aart Kraay. -goals. 2016. “Growth is Still Good for the Poor.” Serajuddin, Umar, Hiroki Uematsu, Christina European Economic Review 81 (C): 68–85. Wieser, Nobuo Yoshida, and Andrew Dabalen. Ferreira, Francisco H. G., Shaohua Chen, Andrew 2015. “Data Deprivation: Another Depriva- Dabalen, Yuri Dikhanov, Nada Hamadeh, tion to End.” Policy Research Working Paper Dean Jolliffe, Ambar Narayan, Espen Beer 7252, World Bank, Washington, DC. Prydz, Ana Revenga, Prem Sangraula, Umar United Nations. 2014. “A World That Counts: Serajuddin, and Nobuo Yoshida. 2016. “A Mobilising the Data Revolution for Sustain- Global Count of the Extreme Poor in 2012: able Development.” November, Independent Data Issues, Methodology, and Initial Results.” Expert Advisory Group on a Data Revolution Journal of Economic Inequality 14 (2): 141–72. for Sustainable Development, United Na- Hicks, John. 1939. “The Foundations of Wel- tions, New York. fare Economics.” Economic Journal 49 (196): World Bank. 2005. World Development Report 696–712. 2006: Equity and Development. Washington, Kaldor, Nicholas. 1939. “Welfare Propositions DC: World Bank; New York: Oxford Univer- in Economics and Interpersonal Compar- sity Press. isons of Utility.” Economic Journal 49 (195): ———. 2015a. A Measured Approach to Ending 549–52. Poverty and Boosting Shared Prosperity: Con- Kuznets, Simon. 1955. “Economic Growth and cepts, Data, and the Twin Goals. Policy Re- Income Inequality.” American Economic Re- search Report. Washington, DC: World Bank. view 45 (1): 1–28. ———. 2015b. “World Bank’s New End- Lakner, Christoph, Mario Negre, and Espen B. Poverty Tool: Surveys in Poorest Countries.” Prydz. 2014. “Twinning the Goals: How Can Press Release, October 15. http://www.world Shared Prosperity Help to Reduce Global bank.org/en/news/press-release/2015/10/15 Poverty?” Policy Research Working Paper /world-bank-new-end-poverty-tool-surveys 7106, World Bank, Washington, DC. -in-poorest-countries.

POVERTY AND SHARED PROSPERITY: SETTING THE STAGE 33

2 Global Poverty

Chapter 2 presents the latest data on global and regional poverty using the international extreme poverty line of US$1.90 (2011 purchasing power parity [PPP] U.S. dollars). The chapter discusses the geographical concentration of poverty and complements the global headcount ratio and data by providing a profile of global poverty. It also reflects on recent methodological changes and their consequences on global estimates. The global poverty estimate for 2013 is 10.7 percent of the world’s population, or 767 million people. This confirms the continuation of the rapid downward trend in the poverty headcount ratio since 1990 (an average of 1.1 percentage points per year). The reduction in 2013 is even greater than the average, with a decline in the headcount ratio of 1.7 percentage points. In absolute net terms, this represents 114 million fewer poor people in a single year. Much of the observed reduction was driven by remarkable progress in the East Asia and Pacific region (71 million fewer poor) and South Asia (37 million fewer poor). A significant change in the geography of poverty has meant that Sub-Saharan Africa was hosting more than half the world’s poor in 2013. This is despite the fact that the African subcontinent experienced progress in lowering both the headcount ratio (1.6 percentage points) and the number of the poor (4 million in 2012–13). However, these achievements are modest compared with reductions in East Asia and Pacific and in South Asia. Other regions with lower poverty rates and totals—notably, Eastern Europe and Central Asia, as well as Latin America and the Caribbean—saw marginal declines in 2012–13. The profile of the global poor shows they are predominantly rural, young, poorly educated, mostly employed in the agricultural sector, and living in larger households with more children.

Monitoring global poverty Ever since the first World Development cant progress in poverty reduction has been Report, World Development Report 1990: accompanied by important improvements Poverty, the share of people living on less in data availability, although substantial than US$1.90 per person per day has been gaps remain (see chapter 1). As a result, steadily declining.1 This has occurred at a the global poverty headcount in 2013 in- rapid average pace of 1.1 percentage points corporated survey-based estimates on 137 a year. Overall, the total number of poor countries.2 has also decreased steadily and dramati- In 2013, an estimated 767 million peo- cally throughout the period, except for the ple were living under the international pov- 1997–99 span of increasing poverty associ- erty line of US$1.90 a day (table 2.1). This ated with the Asian financial crisis. Signifi- means that almost 11 people in 100, or 10.7

GLOBAL POVERTY 35 TABLE 2.1 World and Regional Poverty Estimates, 2013

Headcount Poverty Squared poverty Poor Region ratio (%) gap (%) gap (%)a (millions) East Asia and Paciﬁc 3.5 0.7 0.2 71.0 Eastern Europe and Central Asia 2.3 0.6 0.3 10.8 Latin America and the Caribbean 5.4 2.6 1.8 33.6 Middle East and North Africab —— — — South Asia 15.1 2.8 0.8 256.2 Sub-Saharan Africa 41.0 15.9 8.4 388.7 Total, six regions 12.6 3.8 1.8 766.6 World 10.7 3.2 1.5 766.6

Sources: Annex 2A; most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. Note: Poverty is measured using the 2011 US$1.90-a-day PPP poverty line. The six-region total includes all developing regions. World includes all developing regions, plus industrialized countries. Definitions of geographical regions are those of PovcalNet. — = not available. a. The squared poverty gap attaches increasing weights to the poor the further below the poverty line they are. The higher the squared poverty gap, the greater the share of the poor reporting extremely low consumption levels. b. Estimates on the Middle East and North Africa are omitted because of data coverage and quality problems. The population coverage of available household surveys is too low; the share of the total regional population represented by the available surveys is below 40 percent. There are also issues in the application of the 2011 PPP U.S. dollar to the region. These issues revolve around the quality of the data in several countries experiencing severe political instability, breaks in the consumer price index (CPI) series, and measurement or comparability problems in specific household surveys. These caveats suggest that more methodological analyses and the availability of new household survey data are both needed before reliable and sufficiently precise estimates can be produced.

percent, were poor, 1.7 percentage points erty remains unacceptably high despite the lower than the global headcount ratio in recent progress. 2012. Given the low standard of living im- The poverty gap provides a measure of plied by the US$1.90-a-day threshold, pov- how far below the poverty line the poor in a given country or region fall. This gap is expressed as a share of the poverty line FIGURE 2.1 The Global Poverty Headcount Ratio and the Number of and represents the average distance to the the Extreme Poor, 1990–2013 poverty line among all the poor. While the 60 2,000 1,850 1,855 global poverty gap is small (3.2 percent), the poverty gap in Sub-Saharan Africa is 1,666 1,693 1,800 50 1,588 almost ﬁve times larger (15.9 percent). This 1,600 indicates that the region not only houses the 1,328 1,400 largest number of the poor of any region in 40 1,206 35.0% the world, but also that the region’s poor are, 33.5% 1,078 1,200 on average, living much further below the 28.8% 946 30 28.1% US$1.90-a-day extreme poverty threshold. 25.3% 881 1,000 Figure 2.1 depicts the steady decline in 800 20.4% the share and total number of the poor in 20 17.8% 767 15.6% 600 the world since 1990. Since 2002, in particu- Number of poor (millions) Poverty headcount ratio (%) lar, the global headcount ratio has followed 13.5% 400 10 12.4% a steady downward trajectory, showing 10.7% 200 no sign of slowing down even during the 0 0 global ﬁnancial crisis (2008–09). Despite the more rapid demographic growth in 1990 1993 1996 1999 2002 2005 2008 2010201120122013 poorer areas, this strong trend culminated Poverty headcount ratio (% of people who live below $1.90) in 114 million people lifting themselves out Number of people who live below $1.90 a day (right axis) of extreme poverty in 2013 alone. (See box 2.1 and annex 2A.)3 Sources: Annex 2A; most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. The global number of the poor has fallen Note: Poverty is measured using the 2011 US$1.90-a-day PPP poverty line. dramatically in only slightly more than two

36 POVERTY AND SHARED PROSPERITY 2016 BOX 2.1 Measuring Global Poverty and Purchasing Power Parities

Global and regional poverty numbers inﬂation rates. Part of the reason in and trends are affected by changes some of these countries is that when in relative prices across economies. the price data required for the PPP To account for these changes, the estimates were being collected in 2011, International Comparison Program the countries were undergoing an has produced PPPs using 2011 as the exceptional period of instability, which reference year.a A few countries, such inﬂuenced the quality of the exercise. as Bangladesh, Cabo Verde, Cambodia, For this report, the World Bank has the Arab Republic of Egypt, Jordan, Iraq, either adopted 2011 PPPs for some the Lao People’s Democratic Republic, of these countries after additional and the Republic of Yemen, were exploration and assessment of the initially regarded as outliers because quality of the data, or used PPPs based their 2011 PPPs deviated substantially on regression estimates if quality issues from expectations based on observed are a special concern.

a. See annex 2B and ICP (International Comparison Program) (database), World Bank, Washing- ton, DC, http://siteresources.worldbank.org/ICPEXT/Resources/ICP_2011.html.

decades (figure 2.1). There were almost 1.1 from over 1.8 billion to 767 million during billion fewer people living in poverty world- these years. This implies an average of almost wide in 2013 than in 1990, a period in which 50 million persons escaping poverty every the world population grew by almost 1.9 year in net terms, equivalent to the popula- billion. The total number of poor dropped tion of Colombia or the Republic of Korea. The progress was even more impressive in 2002–13, when an average of 75 million peo- FIGURE 2.2 Where Are the Global Poor ple, similar to the population of Germany or Living? The Global Poor, by Region, 2013 Turkey, moved out of poverty annually. Share of global poor by region (%) As extreme poverty has declined globally, the regional profile of poverty has shifted as 0.8% 9.3% a consequence of uneven progress. In 2013, 1.4% Sub-Saharan Africa accounted for more of 4.4% the poor—389 million people—than all other regions combined; the share of the region in the global total was 50.7 percent (see figure 2.2). This is a remarkable change 50.7% in the geography of global poverty during 33.4% the two decades since 1990, when half of the poor were living in East Asia and Pa- cific. Indeed, Sub-Saharan Africa first over- took East Asia and Pacific in 2005 and then South Asia in 2011 as the region with the East Asia and Pacific largest number of the poor worldwide. In South Asia Eastern Europe and Central Asia 2013, one-third of the global poor were liv- Sub-Saharan Africa ing in South Asia. The East Asia and Pacific Latin America and the Caribbean region was home to 9.3 percent, while the Rest of the world Latin America and Caribbean region and the Eastern Europe and Central Asia region Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, reported global poverty shares of 4.4 per- http://iresearch.worldbank.org/PovcalNet/. cent and 1.4 percent, respectively.

GLOBAL POVERTY 37 FIGURE 2.3 Regional and World Trends, Number of the Extreme Poor, 1990–2013 2,000

1,800

1,600

1,400

1,200

1,000

800

600

400 Number of extreme poor (millions) 200

0 1990 1993 1996 1999 2002 2005 2008 2011 2013

East Asia and Paciﬁc South Asia Eastern Europe and Central Asia Sub-Saharan Africa Latin America and the Caribbean World Middle East and North Africa

Most of the changing geography of The implication of the current geogra- global poverty arises from the lagging per- phy of global and regional poverty is that if formance of Sub-Saharan Africa in reduc- the goal of ending poverty is to be achieved, ing poverty relative to East Asia and Pacific most of any future decline will have to come and to South Asia. The growing number of from Sub-Saharan Africa, and to a lesser the poor in Sub-Saharan Africa stands out extent, South Asia. This is so because the as an exception (figure 2.3). The increase in future role of East Asia and Pacific in total Sub-Saharan Africa occurred despite sig- poverty reduction will narrow following the nificant reductions in the share of the poor rapid reductions already attained. In sum, in the regional population and substantial future progress toward the global poverty economic growth in the subcontinent (fig- goal is likely to taper off in coming years ure 2.4).4 As a result, the number of people if an acceleration in current poverty re- living below the international poverty line duction trends does not take place in Sub- in Sub-Saharan Africa has gradually ex- Saharan Africa. panded since the early 1990s—with the ex- The trends in regional poverty reduction ception of 2002–05—and peaked in 2010. help unpack the global trend in the head- Thereafter, the total number of the poor count ratio (see figure 2.4). In this case, in the region appears to have somewhat the East Asia and Pacific region showed a declined, from 399 million to 389 million strong performance. In the 23-year period by 2013. In absolute terms, the ability of up to 2013, the region managed to reduce economic progress to reduce poverty in its headcount ratio by a staggering 56.7 Sub-Saharan Africa has been partly offset percentage points, down to 3.5 percent, by population growth, and in many cases, close to the World Bank’s global target of an unequal distribution of the benefits of 3.0 percent by 2030. In the latest 2013 data the economic growth.5 round, the poverty rate in the Latin Amer-

38 POVERTY AND SHARED PROSPERITY 2016 FIGURE 2.4 Regional and World Trends, Extreme Poverty Headcount Ratio, 1990–2013

Poverty headcount ratio (%) 20

0 1990 1993 1996 1999 2002 2005 2008 2011 2013 East Asia and Paciﬁc South Asia Eastern Europe and Central Asia Sub-Saharan Africa Latin America and the Caribbean World Middle East and North Africa

Sources: Annex 2A; most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. Note: Poverty is measured using the 2011 US$1.90-a-day PPP poverty line. Breaks in the trends shown in the figure arise because of the lack of good-quality data. ica and Caribbean region (5.4 percent) ap- gion (71 million) and South Asia (37 mil- peared to taper off in 2008–13, following lion), which represent respective reductions the significant declines in 2002–08, with in the headcount of 3.6 and 2.4 percentage annual rates of reduction of 1.0 percent- points. The former is explained largely by age point in the latter, but only 0.3 in the the substantial contraction in the estimates former. The Eastern Europe and Central for China and Indonesia, whereas the latter Asia region maintained a slow, but steady is driven by India’s performance. (For a dis- decline in the headcount ratio, to around cussion of the methodological changes in 2 percent.6 The poverty estimates in South these countries, see annex 2B.) Meanwhile, Asia indicate substantial progress, particu- in Sub-Saharan Africa, the number of the larly in the five years up to 2013, when the poor declined by only 4 million, a 1.6 per- annual average reduction reached 2.9 percentage point change that leaves the average centage points. Indeed, the region, in which regional headcount ratio at 41.0 percent, almost one person in two was extremely which is still high. The share of the poor in poor only 25 years ago, reduced the share Eastern Europe and Central Asia decreased of poverty dramatically, to 15.1 percent, at by about a quarter of a percentage point, an average decline of 1.3 percentage points down to 2.3 percent, while in Latin America a year. and the Caribbean, the share was reduced In 2012–13, the changes in poverty were by 0.2 percentage points, leaving the 2013 remarkable. The world’s headcount ratio headcount ratio at 5.4 percent. fell from 12.4 percent to 10.7 percent, and The largest number of the global poor the number of the poor declined by 114 live in lower-middle-income countries (fig- million. This sharp drop is mostly explained ure 2.5). This is despite the greater aver- by the cuts in the number of the poor in age income and lower headcount ratios of two regions, the East Asia and Pacific re- these countries relative to low-income coun-

GLOBAL POVERTY 39 FIGURE 2.5 Poverty Headcount Ratios car, and Mozambique. The red bars super- and Number of the Poor, by Country imposed on the headcount ratios in ﬁgure Income, 2013 2.6 show the number of the poor. Among 45 500 the countries with the highest headcount ratios, at almost 60 percent and above, the 40 450 Democratic Republic of Congo is the coun- 35 400 try housing the largest number of poor by 350 far through a combination of an extremely 30 high poverty rate (75.9 percent) and a large 300 25 population, implying around 55 million 250 poor people. The high rates in Madagascar 20 200 (78.0 percent) and Mozambique (60.0 per- 15 cent) result in lower absolute poverty num-

150 Number of poor (millions) poor of Number

Poverty headcount ratio (%) bers, 17.9 million and 15.9 million, respec- 10 100 tively, because of the smaller populations of 5 50 these countries. Although over half the world’s poor 0 0 Low Lower Upper live in Sub-Saharan Africa, four of the top income middle middle 10 countries by the number of the poor income income are not in this region, namely, Bangladesh, Population-weighted poverty rate China, India, and Indonesia (see figure 2.7, Number of poor (right axis) where the red bars show the number of the poor). This is because, despite the relatively Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, low headcount ratios, these four countries http://iresearch.worldbank.org/PovcalNet/. have large populations. India is by far the Note: Poverty is measured using the 2011 US$1.90-a-day PPP country with the largest number of people poverty line. Countries are grouped into income categories following the 2016 classification of lower-, lower-middle, and living under the international US$1.90-a- upper-middle-income countries in PovcalNet. day poverty line, 224 million, more than 2.5 times as many as the 86 million in Nigeria, which has the second-largest population of tries. Around 439 million people in lower- the poor worldwide. Thus, Sub-Saharan Af- middle-income countries live below the rica has one in two of the poor worldwide, international poverty line (the right y-axis while India accounts for one in three (see in figure 2.5). This is 181 million more than table 2.1). in low-income countries. Driving this result Overall, in our sample, 243.5 million is the fact that some of the most populous people live in countries with poverty head- countries in the world, with large num- count ratios above 50 percent, while around bers of poor in absolute terms, are in this 356.0 million live in economies where the lower-middle-income category. More gen- ratio ranges from 30 percent to 50 percent. erally, fewer people live in low-income These figures are relevant in so far as a countries, which accounted for 8 percent higher share of poverty is often associated of the global population in 2013, compared with lower average income or consumption with 40 percent in lower-middle-income among the poor and therefore a wider pov- countries and 35 percent in upper-middle- erty gap. Thus, in Sub-Saharan Africa, the income countries. average consumption of the poor is US$1.16 The countries with the highest poverty a day (2011 PPP), which is US$0.74 below headcount ratios and the largest number of the international poverty line.7 The magni- the poor are not the same (figures 2.6 and tude of the poverty gap indicates how dif- 2.7). All countries in the first category— ficult the eradication of poverty will be in the highest poverty headcount ratios—are the near future as the World Bank goal of in Sub-Saharan Africa, and only three of eradicating extreme poverty comes within them appear in both categories, namely, the reach. The gap gives a sense of the urgency Democratic Republic of Congo, Madagas- and also the effort still required to lift up

40 POVERTY AND SHARED PROSPERITY 2016 FIGURE 2.6 Poverty Headcount Ratios, Top 10 Countries, 2013 90 60

80 50 70

60 40

50 30 40

30 20 Number of poor (millions)

Poverty headcount ratio (%) 20 10 10

0 0

Rwanda Zambia Malawi Burundi Mozambique South Sudan Madagascar Republic Guinea-Bissau Central African Congo, Dem. Rep. Poverty headcount ratio Number of poor (right axis)

FIGURE 2.7 Number of the Poor, Top 10 Countries, 2013

90 250

70 200

60 150 50

40 100 30 Number of poor (millions) Poverty headcount ratio (%) 20 50 10

0 0

China India Ethiopia Nigeria Tanzania Indonesia Mozambique Madagascar Bangladesh Congo, Dem. Rep. Poverty headcount ratio Number of poor (right axis)

Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http:// iresearch.worldbank.org/PovcalNet/. Note: Poverty is measured using the 2011 US$1.90-a-day PPP poverty line. people living below the poverty line, espe- count ratio shows a more rapid decline than cially those living far below it. the poverty gap. This difference in pace has Figure 2.8 shows the downward trends in been particularly marked since 2010. If this the global poverty gap and in the headcount difference persists, the downward trend in ratio over the last two decades. The head- the headcount ratio may begin to taper off

GLOBAL POVERTY 41 FIGURE 2.8 Trends in the Poverty Gap strong reminder that poverty eradication is and the Global Headcount Ratio, 1990–2013 well within reach (as also suggested by re- 40 cent projections; box 2.2) and that a better effort at sharing prosperity would be in- 35 strumental in increasing the speed at which 30 the goal is reached. Chapter 3 examines this challenge by providing an assessment of the 25 World Bank’s shared prosperity goal based 20 on the most recently available data.

Percent 15

10 A profile of the poor in the developing world 5 Exploring the characteristics of the poor is 0 key to a better understanding of the circum- 1990 1993 1996 1999 2002 2005 2008 2011 2013 stances and contexts surrounding poverty. Poverty gap at $1.90 A large database of household surveys from Poverty headcount ratio at $1.90 89 developing countries provides insights into this issue by facilitating demographic Sources: Annex 2A; most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, profiles of the poor at the US$1.90 per per- DC, http://iresearch.worldbank.org/PovcalNet/. son per day poverty line and at a higher line of US$3.10 per person.11 These poverty profiles reveal that the global poor are pre- as the poverty gap more slowly narrows. dominantly rural, young, poorly educated, This combination may indicate that it is mostly employed in the agricultural sector, the relatively less poor that are moving out and live in larger households with more of poverty. This may have important im- children.12 Figure 2.9 reports the share of plication in terms of slower future poverty the poor who live in rural areas (80 percent reductions. of the poor worldwide), work in agriculture Multiplying the income needed to lift (64 percent), are 14 years of age or younger every poor individual out of poverty by (44 percent), and have no formal education the total number of the poor in the world (39 percent). The data also confirm wide re- provides a rough indication of the order of gional variations in the distribution of the magnitude of the cost of ending poverty. poor across these characteristics. This is not meant to be a policy prescrip- Figure 2.10 displays the share of the poor tion or a judgment on specific current poli- at the US$1.90-a-day and the US$3.10-a- cies but a low-bound approximation of the day global poverty lines by urban or rural cost of achieving the poverty goal under residence, age category, employment in stylized and simplifying assumptions, key agriculture or nonagriculture, and educa- among them is that the income shortfall tional attainment. Poverty rates are more to the poverty line is a reasonable initial than three times higher among rural res- approximation to the cost of eliminating idents than among urban dwellers: 18.2 8 poverty. While this represented 1.0 per- percent versus 5.5 percent, respectively. cent of the world’s GDP in 1990, it would Agricultural workers are more than four have required less than 0.2 percent in 2013, times more likely to be poor relative to peo- which is 10.0 percent more than all the of- ple employed in other sectors of the econ- 9 ficial development assistance that year. It omy. Educational attainment is inversely is also about 50 percent of the tax revenue correlated with poverty. A small share of estimated to be annually lost through tax primary-school graduates are living in pov- 10 avoidance. The income shortfall relative erty: the share of people who completed to the poverty line among the worldwide primary school but not secondary school poor is therefore almost negligible in com- and who are living below the US$1.90 pov- parison with the global economy. This is a erty line is 8.0 percent. Among those who

42 POVERTY AND SHARED PROSPERITY 2016 BOX 2.2 Projecting Poverty Rates

In 2015, the World Bank published a growth more accurately: 0.72 for projected extreme poverty headcount China and 0.51 for India. Furthermore, ratio—9.6 percent of the world’s it is assumed that projected growth population for that year—using $1.90- was distributionally neutral between a-day poverty line at 2011 PPP. In 2016, 2013—the latest year for which poverty the analytical exercise was repeated is actually estimated, not projected— using the same methodology. The and 2016. This means that the gains projected global poverty headcount ratio of growth are assumed to have been for 2016 is 9.1 percent. equally distributed across population This projection is based on the latest groups in each country, for example, available household survey information among poor and nonpoor households. on each country and incorporates This methodological description assumptions that allow past highlights that poverty projections are consumption and income to be updated associated with additional challenges to current levels. Thus, the exercise relative to poverty estimates. This applied a national accounts–based increases the uncertainty around the growth rate to adjust the latest available accuracy of the projections. Producing survey mean income or consumption projections requires that consumption to 2016. The growth projection of per or income data derived from actual capita private consumption expenditure observations in every country, for was used across the board except for instance, from household surveys, the Sub-Saharan Africa region, where be adjusted to estimate more recent the growth projection of per capita GDP income or consumption. This is so was used instead. Elsewhere, projected because, in the projected year, 2016 in GDP growth was used only if growth this case, there is no available household projections for private consumption survey from which to draw estimates. expenditure were missing in a given Moreover, the growth rates used in this country. In four countries, both GDP exercise rely on projections rather than and private consumption expenditure observed, revised, and vetted rates for were missing; so the average growth 2016. The further ahead projections rate of per capita GDP during the three look relative to a reference year, the most recent years available was used. more the assumption that the past is a Empirically, it has been established that, good measure of the future becomes on average, household survey–based questionable. Thus, even the growth growth captures about 87 percent projections between 2013 and 2016 of national accounts–based growth.a do not fully capture global and regional As a consequence, in the application shifts, such as the weakening of of national accounts–based growth international commodity prices or the projections to update household survey– slower growth in Latin America and the based consumption or income means, Caribbean, in Sub-Saharan Africa, or, an adjustment factor of 0.87 was used more generally, among emerging market across the board, except for China and economies and low-income economies.b India, where alternative factors reﬂect For these reasons, poverty projections the discrepancies between national need to be viewed with extreme account– and household survey–based caution.

Source: World Bank calculations. a. Ravallion (2003). b. World Bank (2016).

have attended university, the share is 1.5 A complete demographic poverty proﬁle percent.13 Similar differences are observed should also include a gender dimension. if poverty incidence is measured relative to However, the global poverty database does the US$3.10-a-day poverty line. not contain information on the consump-

GLOBAL POVERTY 43 FIGURE 2.9 Profile of the Poor, by Characteristics and Region, 2013 tion or incomes of individuals within a household, only on the total and per capita household consumption or income. With- Share of poor out information on the intrahousehold al- in rural areas location and use of resources, the study of gender poverty trends and profiles must remain limited to differences in poverty Share of poor adults rates between woman-headed and man- working in agriculture headed households. Yet, woman-headed households may not be representative of the welfare of many women in develop- Share of poor ing countries. Women may head a house- 0–14 years old hold because they have the means to live independently or because men household members are absent but sending remit- Share of poor adults tances. Women in such households would with no education show lower poverty rates. There are, however, other contexts in which estimates of 0 102030405060 70 80 90 poverty headcount ratios may be higher in Percent gender terms. For instance, in situations of South Asia East Asia and Pacific national or local conflict, households led Eastern Europe and Central Asia Sub-Saharan Africa by widows may be a significant share of all Latin America and the Carribean World households and present disproportionally Source: Castañeda et al. 2016. high poverty rates. Using the characteristics Note: Poverty is measured using the 2011 US$1.90-a-day PPP poverty line. of household heads does not provide a reliable picture of gender gaps because it over- looks intrahousehold discrimination and FIGURE 2.10 Profile of the Extreme and Moderate Poor, by Selected uncooperative practices within households Characteristics, 2013 in sharing resources evenly and according to each member’s needs.14 Adults Meanwhile, age profiles confirm that Children children are more likely to be poor than adults.15 Children under 18 account for half Urban the global poor in 2013, but less than a third

Region Age of the sample population (32 percent) (ﬁg- Rural ure 2.11). Younger children contribute especially heavily to the poverty headcount, Nonagriculture much more than their share in the world’s sector Agriculture population. In Sub-Saharan Africa, children are much more likely than adults to be living Tertiary on less than US$1.90 a day, and almost half Secondary complete of all the poor on the subcontinent are 14 Primary or years of age or younger. The region also some secondary contributes the most to global child pov- Primary incomplete erty: 52 percent of the extremely poor chil- Education (adults 15+) Employment No schooling dren worldwide live in Sub-Saharan Africa. 0 102030405060 70 80 90 100 Among countries, the largest contributor is India, where 30 percent of the world’s poor- Share of poor (%) est children live. Global poor (living below $1.90 a day) Nonpoor These estimates of global child poverty People living between $1.90 and $3.10 a day are limited. As in the case of poverty among Sources: Castañeda et al. 2016; Newhouse et al. 2016. women, they do not account for the intra-

44 POVERTY AND SHARED PROSPERITY 2016 FIGURE 2.11 Age Profile of the Poor, 2013 a. The extremely poor b. Sample population

5.8% 10.3% 9.4% 15.9% 9.0%

8.6% 15.4%

44.0% 5.1%

12.9% 57.6%

6.0%

Children, ages 0–4 Children, ages 10–14 Adults, ages 18–59 Children, ages 5–9 Children, ages 15–17 Adults, ages 60 or more

Source: Newhouse et al. 2016.

household allocation and use of resources. adapted to account for the fact that the They also assume identical physical needs international per capita poverty standard across household members of different ages does not reflect differing individual needs.18 and ignore the scale economies associated While there is widespread agreement on the with larger households.16 Strong assump- benefit of making such adjustments, there tions are required to address intrahouse- is no agreement on how to do this properly. hold allocation issues with the information Nonetheless, recent evidence using alterna- currently available through household sur- tive equivalence scales, economies of scale, veys.17 Variations in physical need and scale and possible adjustments of the US$1.90 in- economies, however, are often addressed ternational poverty line confirms beyond a through the use of equivalence scales. More- doubt that children exhibit higher poverty over, the US$1.90-a-day threshold can be rates than adults.19

GLOBAL POVERTY 45 Annex 2A

Historical global and regional poverty estimates

TABLE 2A.1 Historical Trends, World Extreme Poverty Estimates, 1990–2013

Year Poverty line Headcount Poverty Squared poverty Poor Population (PPP US$/day) ratio (%) gap (%) gap (%) (millions) (millions) 1990 1.9 35.0 12.2 5.8 1,850.1 5,283.1 1993 1.9 33.5 11.6 5.5 1,855.4 5,537.8 1996 1.9 28.8 9.4 4.4 1,666.3 5,788.6 1999 1.9 28.1 9.2 4.3 1,692.9 6,034.9 2002 1.9 25.3 8.1 3.8 1,588.1 6,274.7 2005 1.9 20.4 6.2 2.9 1,327.5 6,514.0 2008 1.9 17.8 5.3 2.4 1,205.6 6,758.3 2010 1.9 15.6 4.6 2.1 1,077.5 6,923.7 2011 1.9 13.5 4.0 1.8 946.3 7,006.9 2012 1.9 12.4 3.7 1.7 880.9 7,089.5 2013 1.9 10.7 3.2 1.5 766.6 7,176.1

Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.

TABLE 2A.2 Historical Trends, Regional Poverty Headcount Ratios, 1990–2013 Percent

Region 1990 1993 1996 1999 2002 2005 2008 2010 2011 2012 2013 East Asia and Paciﬁc 60.2 52.4 39.4 37.2 29.0 18.4 14.9 11.1 8.4 7.1 3.5 Eastern Europe and Central Asia 4.0 6.9 7.8 8.1 6.3 5.1 3.3 3.0 2.7 2.6 2.3 Latin America and the Caribbean 15.8 14.4 14.2 13.9 13.0 9.8 7.1 6.5 6.0 5.6 5.4 Middle East and North Africa 6.0 5.6 4.8 3.0 2.8 South Asia 44.6 44.8 40.3 38.5 33.6 29.4 24.6 19.9 17.5 15.1 Sub-Saharan Africa 54.3 58.4 57.7 57.1 55.6 50.0 47.0 45.7 44.1 42.6 41.0 World 35.0 33.5 28.8 28.1 25.3 20.4 17.8 15.6 13.5 12.4 10.7

Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.

TABLE 2A.3 Historical Trends, Number of Extreme Poor, by Region, 1990–2013 Millions

Region 1990 1993 1996 1999 2002 2005 2008 2010 2011 2012 2013 East Asia and Paciﬁc 965.9 876.8 683.8 669.0 535.1 349.2 288.2 218.2 166.9 141.8 71.0 Eastern Europe and Central Asia 18.2 32.2 36.3 37.8 29.3 23.8 15.5 14.2 13.0 12.2 10.8 Latin America and the Caribbean 71.2 68.3 70.7 72.2 70.6 55.6 41.9 38.8 36.4 34.1 33.6 Middle East and North Africa 13.7 13.6 12.4 9.2 6.7 South Asia 505.0 541.5 517.0 552.4 508.3 464.7 400.3 327.9 293.3 256.2 Sub-Saharan Africa 276.1 323.1 346.1 371.3 391.3 381.5 389.1 399.1 395.7 393.1 388.7 World 1,850.1 1,855.4 1,666.3 1,692.9 1,588.1 1,327.5 1,205.6 1,077.5 946.3 880.9 766.6

Source: Most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.

46 POVERTY AND SHARED PROSPERITY 2016 Annex 2B

Technical note: global poverty measurement using 2013 data

The new 2016 round of global poverty mea- Changes in household survey sures based on the most recent available data data incorporates the following changes from the 2015 round of poverty measures. More than 35 new household surveys have This annex explains changes in the PPPs, been added to the World Bank’s global data- the household survey data, the consumer base, and over 100 other surveys have been price index (CPI), population data, and na- updated. See Global Consumption Data- tional accounts. It also deﬁnes the regions base, World Bank, Washington, DC, http:// used throughout the report. datatopics.worldbank.org/consumption/. The World Bank’s global poverty and Changes in purchasing power shared prosperity monitoring taskforce uses parities more than 1,200 household surveys from about 150 countries. See, for example, GDSP In the new 2016 round, the poverty mea- (Global Database of Shared Prosperity), sures for all countries are based on con- World Bank, Washington, DC, http://www sumption PPPs from the 2011 round of .worldbank.org/en/topic/poverty/brief/ data collection by the International Com- global-database-of-shared-prosperity. parison Program; see ICP (International Comparison Program) (database), World Bank, Washington, DC, http://siteresources. Changes in the CPI, population world bank.org/ICPEXT/Resources/ICP data, and national accounts _2011.html. The PPP exchange rates in- data clude benchmark countries where actual price surveys were conducted, as well as re- The CPI data used for global poverty gression-based PPP estimates where such estimation among 116 countries are taken surveys were not conducted. Details on the from the WDI database. See WDI (World regression model for the PPP estimation De velopment Indicators) (data base), World can be found in World Bank (2015). Bank, Washington, DC, http://data.world Changes in the countries that, in the bank.org/data-catalog/world-development previous 2015 round, still used 2005 PPPs -indicators. instead of 2011 PPPs include the follow- China and India use rural and urban ing: (1) the 2005 PPPs of Bangladesh, Cabo CPIs (as provided by the national statis- Verde, Cambodia, and Lao PDR are re- tics ofﬁces). Monthly CPIs are used by placed in this round by 2011 PPPs; (2) for 25 countries; most are in the Latin America countries in the International Comparison and the Caribbean region. Another seven Program region of West Asia, 2011 regres- countries—Bangladesh, Cambodia, Ghana, sion-based consumption PPPs are used in Iraq, Lao PDR, Malawi, and Tajikistan—use this round. the implied, that is, the expected CPI.

GLOBAL POVERTY 47 Population data have been updated as regression-based PPPs and 2011 PPPs, depart of the global poverty estimation exer- pending on the country). cise. Most population data differ from the data of the previous year. Total population Bangladesh’s poverty numbers changes in 21 countries are significant, that is, they range between 0.5 million and 4.5 A detailed assessment of price data in Ban- million. gladesh involving the Bangladesh Bureau of National accounts data have also been Statistics and the Asian Development Bank updated: per capita GDP, private consump- has recently determined that the price data tion, and expenditure data have all been is of good quality and that the 2011 PPP updated. reflects the purchasing power of the Ban- gladesh taka relative to the U.S. dollar more Middle East and North Africa accurately than the 2005 PPP does. There- region fore, the World Bank has adopted the 2011 US$1.90 a day PPP to measure the extreme As part of this new round of global pov- poverty index for Bangladesh in the context erty measurement, a detailed reassessment of global extreme poverty monitoring. Ac- of the 2011 PPPs has been conducted for cording to the new estimates, the propor- Egypt, Iraq, Jordan, and the Republic of tion of the extreme poor in Bangladesh was Yemen. It found that the coverage and qual- 18.5 percent in 2010, the year of the most ity of the 2011 PPP price data for most of recent household survey. This represents a these countries were hindered by the ex- significantly lower estimate than the pre- ceptional period of instability they faced at vious one for that year (43.3 percent). Yet, the time of the 2011 exercise of the Inter- the update of the historical poverty rates in national Comparison Program. Moreover, Bangladesh using the US$1.90 poverty line the poverty estimates resulting from using consistently shows the stable and sizable alternative regression-based PPPs still seem reduction in poverty since 1990 more ac- to underestimate poverty severely in these curately than the previous series using the economies, as well as in Lebanon and the 2005 $1.25-a-day PPP poverty line. Syrian Arab Republic (but not in West Bank and Gaza). In the Middle East and North Africa re- China’s 2013 survey gion, the exclusion of Egypt, Iraq, Jordan, A large part of the decline in poverty inci- and the Republic of Yemen and the lack of dence in East Asia and Pacific in this new recent data on Algeria and Syria imply that round based on 2013 data is attributable to the remaining countries account for only China and Indonesia. The 2013 household a third of the region’s population, below survey in China is the first integrated na- the 40 percent threshold of regional pop- tionwide household survey in that country. ulation coverage needed to report region- This means that it is not comparable with representative estimates. the previous household surveys, in which Adding to this low coverage is the fact rural and urban areas were sampled sep- that the failure to include data on Egypt, arately. In addition, the most significant Iraq, and the Republic of Yemen and the change in the 2013 national household sur- lack of recent data on Syria, which are likely vey relative to previous household surveys to face increasing poverty rates due to insta- was the inclusion of imputed rents into in- bility and civil conflicts, will seriously un- come and consumption aggregates for the derestimate regional poverty rates. first time. As a compromise between precision In 2012–13, China’s poverty rate based and coverage, the regional poverty to- on a US$1.90-a-day poverty line (in 2011 tals and headcount ratios are not reported PPP) declined by about 4 percentage points, for the Middle East and North Africa, but of which half, that is, about 2 percentage an estimate of the number of the poor points, can be traced to changes in the sur- is included in the global total (based on vey methodology. The actual poverty reduc-

48 POVERTY AND SHARED PROSPERITY 2016 TABLE 2B.1 Poverty Estimates Based on China’s Old Survey Methodology, 2013

Poverty line, Headcount Poverty Squared poverty Poor Coverage (PPP US$/day) ratio (%) gap (%) gap (%) (millions) East Asia and Paciﬁc 1.9 5.1 0.66 0.22 102.2 World 1.9 11.1 3.23 1.54 797.8

Sources: Annex 2A; most recent estimates, based on 2013 data using PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. tion not explained by these methodological Reference Period has also been used in changes was therefore 2 percentage points the collection of consumption data. The in 2012–13 (table 2.1 in the main text and methodology is closer to best international table 2B.1 above). practice. It relies on recall periods among The World Bank’s poverty estimates on respondents of 7, 30, and 365 days, depend- China are based on grouped distributions, ing on the items of consumption. If the con- which are often not as precise as direct sumption estimate derived from the latter estimates based on the full distribution of methodology had been used to estimate In- household income and consumption aggre- dia’s poverty rate, the result at the US$1.90 gates. In 2013, China’s poverty headcount poverty line would have been a substantially ratio under the US$1.90-a-day poverty line lower 12.4 percent in 2011/2012. The ap- was 2.2 percent using individual record plication of the methodology is still being data, as conﬁrmed by the National Bureau tested. Its adoption would eventually lead to of Statistics, while it was 1.9 percent based a substantial downward revision of the pov- on grouped data. erty numbers in India.

India’s poverty numbers Definition of geographical regions and industrialized For this report, as with many other coun- economies tries, the total poor population in India is based on estimates rather than actual num- In the past, PovcalNet used the World Bank’s bers provided through a household survey income classification back to 1990 to track collected in the year of reference, in this the Millennium Development Goals. Start- case, 2013. Such estimates are subject to a ing this round, a new regional geographical great deal of uncertainty, which typically classification is used. The income-coun- arises because of revisions of national ac- try categories within the six geographical counts in each country. This is also the case regions are (a) low- and middle-income in India, on which the estimates for 2013 countries and (b) countries eligible to re- and the adjusted historical series reflect ceive loans from the World Bank (such as the government’s periodic revisions of the Chile) and recently graduated countries growth in private consumption expenditure (such as Estonia). See “World Bank Country and the population. Notwithstanding the and Lending Groups,” World Bank, Wash- revisions, no methodological change un- ington, DC, https://datahelpdesk.world derpins the 2013 poverty estimates for India bank.org/knowledgebase/articles/906519 with respect to 2012. -world-bank-country-and-lending-groups. In addition, the poverty estimates for See also PovcalNet (online analysis tool), India at the global poverty line are histor- World Bank, Washington, DC, http:// ically based on the Uniform Reference Pe- iresearch.worldbank.org/PovcalNet/. riod consumption aggregate, which involves The rest of the high-income economies a 30-day recall among respondents in the are listed within the category of industri- recording of all items of consumption. For alized economies, including the following: 2011/2012, this implies a poverty rate of Andorra; Antigua and Barbuda; Aruba; 21.2 percent at the US$1.90 poverty line. Australia; Austria; The Bahamas; Bahrain; Since 2009, however, the Multiple Mixed Barbados; Belgium; Bermuda; British Vir-

GLOBAL POVERTY 49 gin Islands; Brunei Darussalam; Canada; Monaco; Netherlands; New Caledonia; New Cayman Islands; Channel Islands; Curacao; Zealand; Norway; Oman; Portugal; Qatar; Cyprus; Denmark; Finland; France; French Saint-Martin; Saudi Arabia; Singapore; Sint Guiana; French Polynesia; Germany; Gi- Maarten; Spain; St. Kitts and Nevis; Sweden; braltar; Greece; Greenland; Guadeloupe; Switzerland; Taiwan, China; Turks and Ca- Guam; Iceland; Ireland; Isle of Man; Israel; icos Islands; United Arab Emirates; United Italy; Japan; Korea; Kuwait; Liechtenstein; Kingdom; United States; and the U.S. Virgin Luxembourg; Macao SAR, China; Malta; Islands.

Notes 1. World Bank (1990). region were to continue their trends in the 2. The remarkable increase in the availability reduction of the number of the poor in the and use of household surveys has allowed an last decade, they may be close to eradicating enhancement in the coverage of global pov- extreme poverty soon. This will depend on erty data. During the early stages of global their success in addressing remaining pock- poverty measurement, much depended on ets of poverty. imputations based on a few countries. Over 7. With the exception of the Seychelles, the re- two decades later, the World Bank’s global gion’s poverty data are based on consumption. poverty and shared prosperity monitoring 8. The resources needed to help all the poor task relies on more than 1,200 household reach a standard of living at the poverty surveys from about 150 countries. line through policy interventions would 3. The calculation is in net terms. Of the 114 surely be greater than the aggregated in- million reduction in the number of the come shortfall to the poverty line if realistic poor, 31 million correspond to methodolog- assumptions about the administrative costs ical changes implemented in China’s 2013 involved in identifying the poor, targeting, national survey, which, for the first time, ensuring effectiveness, delivery, coordina- incorporated imputed rents and replaces the tion, use of country systems, and respond- previously separate rural and urban surveys. ing to political economy considerations are 4. This means that the number of the poor was taken into account. These costs would cer- growing more slowly than the total popula- tainly render the amount identified in the tion during this period. text insufficient to eradicate poverty even if 5. Bifurcation is the term sometimes used to the allocation was perfect. The analysis here describe the fact that half the countries in is therefore merely informative about orders Sub-Saharan Africa have done relatively well of magnitude in relative economic terms. in terms of distributional changes, while the 9. WDI (World Development Indicators) other half have not (Beegle et al. 2016). (database), World Bank, Washington, DC, 6. If the historical experiences of countries that http://data.worldbank.org/data-catalog have already eliminated poverty is a good /world-development-indicators. indication of the future, the pace of poverty 10. Zucman (2015). reduction in some regions should taper off. 11. This section draws largely on Castañeda Some countries that have eliminated pov- et al. (2016) and Newhouse et al. (2016). erty experienced poverty rates falling al- 12. Castañeda et al. (2016). most linearly to zero, while the rates in other 13. Castañeda et al. (2016) analyze the robust- countries declined more slowly. For exam- ness of these results by comparing different ple, Austria and Japan sustained a linear re- lineup methods and different ways to adjust duction in both the number and the share of welfare aggregates, weights, and poverty the poor as they neared poverty eradication. lines. They find only minimal differences. If the Latin America and Caribbean region They also check for fixed effects and sensi- and the Eastern Europe and Central Asia tivity to missing data. The resulting demo-

50 POVERTY AND SHARED PROSPERITY 2016 graphic proﬁle thus paints a robust picture mation, and an Application to Child Poverty of global poverty. in Malawi.” American Economic Review 103 14. Doss (2013); World Bank (2011). (1): 438–71. 15. This subsection draws on Newhouse et al. Lanjouw, Peter F., and Martin Ravallion. 1995. (2016). “Poverty and Household Size.” Economic 16. Lanjouw and Ravallion (1995). Journal 105 (433): 1415–34. 17. Dunbar, Lewbel, and Pendakur (2013). Newhouse, David, Pablo Suarez-Becerra, Martin 18. Batana, Bussolo, and Cockburn (2013); C. Evans, and Data for Goals Group. 2016. Ravallion (2015). “New Estimates of Extreme Poverty for Chil- 19. Newhouse et al. (2016). dren.” Policy Research Working Paper 7845, World Bank, Washington, DC. References Ravallion, Martin. 2003. “Measuring Aggregate Welfare in Developing Countries: How Well Batana, Yélé, Maurizio Bussolo, and John Cock- Do National Accounts and Surveys Agree?” burn. 2013. “Global Extreme Poverty Rates Review of Economics and Statistics 85 (3): for Children, Adults, and the Elderly.” Eco- 645–52. nomics Letters 120 (3): 405–07. ———. 2015. “On Testing the Scale Sensitivity Beegle, Kathleen, Luc Christiaensen, Andrew of Poverty Measures.” Economics Letters 137 Dabalen, and Isis Gaddis. 2016. Poverty in a (C): 88–90. Rising Africa. Washington, DC: World Bank. World Bank. 1990. World Development Report Castañeda, Andres, Dung Doan, David New- 1990: Poverty. Washington, DC: World Bank; house, Minh C. Nguyen, Hiroki Uematsu, New York: Oxford University Press. João Pedro Azevedo, and Data for Goals ———. 2011. World Development Report 2012: Group. 2016. “Who Are the Poor in the De- Gender Equality and Development. Washing- veloping World?” Policy Research Working ton, DC: World Bank. Paper 7844, World Bank, Washington DC. ———. 2015. Purchasing Power Parities and the Doss, Cheryl. 2013. “Intrahousehold Bargain- Real Size of World Economies: A Comprehen- ing and Resource Allocation in Developing sive Report of the 2011 International Compar- Countries.” World Bank Research Observer 28 ison Program. Washington, DC: World Bank. (1): 52–78. http://siteresources.worldbank.org/ICPEXT Dunbar, Geoffrey, Arthur Lewbel, and Krishna /Resources/ICP-2011-report.pdf. Pendakur. 2013. “Children’s Resources in Zucman, Gabriel, 2015. The Hidden Wealth of Collective Households: Identiﬁcation, Esti- Nations. Chicago: University of Chicago Press.

GLOBAL POVERTY 51

3 Shared Prosperity

This chapter reports on the latest progress achieved in the promotion of shared prosperity worldwide. Shared prosperity is measured as the growth in the income or consumption of the bottom 40 percent of the population in a country (the bottom 40). The larger the growth rate in the incomes of the bottom 40, the more quickly economic progress is shared with the poorer segments of society. Performance on this indicator is examined by country rather than globally. The latest data suggest that the bottom 40 beneﬁted from income growth in many countries in circa 2008–13 even though the period encompasses the global ﬁnancial crisis of 2008–09. Overall, the bottom 40 experienced positive income growth in 60 of the 83 countries monitored. This means that 89 percent of the population covered in the dataset resided in countries in which the income or consumption of the bottom 40 grew. A total of 49 countries reported a positive shared prosperity premium: the income growth among the bottom 40 exceeded that of the mean (and therefore, the income growth of the top 60). Nonetheless, there is no room for complacency. In 23 of the countries, the incomes of the bottom 40 declined, and, in 15 of the countries, the contraction in the income or consumption of the bottom 40 was larger than the corresponding contraction at the mean. In these countries, the living conditions deteriorated more quickly among the bottom 40 than among the rest of the population.

Shared prosperity: where we stand In 2016, based on data on the most up- contributed 8, 16, and 9 countries, respec- dated spell, circa 2008–13, the World Bank’s tively. In South Asia, 4 countries were cov- Global Database of Shared Prosperity re- ered, and, in the Middle East and North ported annualized growth rates among the Africa, 2. The population coverage also bottom 40 and the overall mean growth differed across regions. At the high end, the rate for 83 countries.1 Though the sample coverage was 94 percent in the East Asia and covered 75 percent of the world’s popu- Paciﬁc region and somewhat lower in East- lation in 2013, it included fewer than half ern Europe and Central Asia (89 percent), the world’s countries. The geographical South Asia (87 percent), and Latin America coverage across regions was not uniform. and the Caribbean (86 percent). The cover- Of the 83 countries, 24 belonged to a sin- age in the Middle East and North Africa was gle region, Eastern Europe and Central Asia, lower, at 32 percent, and the share of the while East Asia and Paciﬁc, Latin America population covered in Sub-Saharan Africa and the Caribbean, and Sub-Saharan Africa was substantially lower, at 23 percent.2 The

SHARED PROSPERITY 53 TABLE 3.1 Shared Prosperity, Circa 2008–13

Countries, Population- Countries, growth in Countries, Country Countries, weighted Palma Countries Population Population mean < 0 SP > 0 average SP premium average, SP premium, > 0 Region (number) (millions) (%) (number) (number) SP (%) > 0 (number) premium (pp)a (pp) (number)b East Asia and Paciﬁc 8 2006.2 94 0 8 5.0 7 0.7 7 Eastern Europe and Central Asia 24 479.1 89 10 15 1.5 12 0.3 14 Latin American and the Caribbean 16 622.0 86 3 15 4.1 12 1.4 12 Middle East and North Africa 2 350.1 32 1 2 1.8 1 2.7 1 South Asia 4 1,698.1 87 0 4 3.7 3 −0.4 2 Sub-Saharan Africa 9 948.3 23 1 8 2.7 4 0.6 5 Industrialized countries 20 1,072.4 68 10 8 −1.0 10 0.2 11 World 83 7,176.1 75 25 60 2.0 49 0.4 52

Source: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity. Note: SP = shared prosperity (growth in average income or consumption of bottom 40). pp = percentage point. Population coverage refers to 2013. a. Population-weighted shared prosperity premiums are relative to the covered population in each region or in the world. b. The Palma premium (p) is here deﬁned as the difference between the growth in the mean of the bottom 40 and the growth in the mean of the top decile (p ≡ g40 – gt10).

latter is a stark reminder of the data gaps growth rates of 2.7 percent in Sub-Saharan (see chapter 1). Africa, 1.8 percent in the Middle East and Though the period encompassed the North Africa, and 1.5 percent in Eastern global financial crisis of 2008–09, the growth Europe and Central Asia. The bottom 40 in in income of the bottom 40 was positive in 60 industrialized countries experienced an av- of the 83 countries, and negative in the other erage contraction of 1.0 percent of income. 23 in 2008–13 (table 3.1). Of the world’s pop- Figure 3.1 reports country-specific per- ulation, 67 percent (89 percent of the pop- formance. Countries are ranked by the size ulation captured by the Global Database of of the income growth of the bottom 40 (from Shared Prosperity) were living in countries largest positive growth to largest contrac- in which the income or the consumption of tion) within each region. For each country, the bottom 40 grew in 2008–13.3 This is a the growth in the income or consumption of substantial proportion, particularly because the bottom 40 is compared with the growth one-quarter of the world’s population is not in income or consumption of the mean. The covered by the dataset and could presum- Democratic Republic of Congo showed the ably increase the share.4 The simple average largest income growth rate among the bot- income or consumption growth among the tom 40, an annualized rate of 9.6 percent in bottom 40 across all countries monitored circa 2008–13. Five countries saw growth was 2.0 percent. If the worldwide estimate rates of income or consumption of the is calculated weighing each country’s popu- bottom 40 of 8.0 percent or above, namely, lation, the resulting average shared prosper- Belarus, China, Democratic Republic of ity is more than double the unweighted in- Congo, Mongolia, and Paraguay. A growth dicator, reaching an annualized 4.3 percent. rate in the income or consumption of the The strong performance of highly populated bottom 40 ranging between 4.0 percent and countries such as Brazil, China, India, and 8.0 percent was reported in 19 countries: Indonesia drive this result. Bhutan, Bolivia, Brazil, Cambodia, Chile, All regions report positive average in- Colombia, Ecuador, Georgia, Kazakhstan, come or consumption growth among the the former Yugoslav Republic of Macedo- bottom 40. East Asia and Pacific, Latin nia, Moldova, Nicaragua, Panama, Peru, the America and the Caribbean, and South Asia Russian Federation, the Slovak Republic, showed the best average growth perfor- Thailand, Uruguay, and Vietnam. This rank- mance among the bottom 40, with annual- ing suggests that large increases occurred ized rates of 5.0 percent, 4.1 percent, and 3.7 in all regions and among low-, middle-, and percent, respectively. They are followed by high-income countries.

54 POVERTY AND SHARED PROSPERITY 2016 FIGURE 3.1 Shared Prosperity, 83 Countries, 2008–13

East Asia and Paciﬁc China Mongolia Cambodia Thailand Vietnam Indonesia Philippines Lao PDR

Latin America and the Caribbean Paraguay Ecuador Bolivia Brazil Colombia Peru Chile Uruguay Nicaragua Panama El Salvador Argentina Dominican Republic Costa Rica Mexico Honduras

Middle East and North Africa Iran, Islamic Rep. Iraq

South Asia Bhutan India Pakistan Sri Lanka

Sub-Saharan Africa Congo, Dem. Rep. Uganda Tanzania Congo, Rep. Togo Cameroon Mauritius Rwanda Senegal

–10–5 0 5 10 Annualized growth in mean income or consumption (%) Total population Bottom 40

Source: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty /brief/global-database-of-shared-prosperity. Note: The data show the annualized growth in mean household per capita income or consumption. See table 3A.1, annex 3A. Greece experienced the largest contrac- the top 60 was 2.1 percent a year in Cam- tion in incomes among the bottom 40, an bodia and 5.3 percent in Cameroon. Such annualized 10.0 percent decline in 2008–13. differences produce contrasting effects on Other countries with large contractions in inequality trends. While, in Cambodia, the the income or consumption of the bottom Gini index fell by 4.4 points during the spell, 40, at 3.0 percent or more annually, were and it rose by 3.7 points in Cameroon.6 Croatia, Iceland, Ireland, Italy, and Latvia. A comparison of shared prosperity The negative performance reflects the prob- premiums across the countries on which lems of a period marked by the 2008–09 information is available—the monitored global financial crisis and its repercussions, sample of 83 countries—describes a gen- which are still widely felt in Western Europe erally positive picture. Of the entire sample and in Eastern Europe and Central Asia. In circa 2008–13, income growth among the Latin America, Honduras is the only coun- bottom 40 was greater than the growth of try showing a contraction (2.5 percent an- the mean in 49 countries. In the remaining nually). In 15 of the 23 countries exhibiting 34 countries, the bottom 40 fared less well negative growth in the income or consump- than the rest of the population. Globally, tion of the bottom 40, the contraction was the population-weighted shared prosperity greater than that of the mean (see figure 3.1). premium is positive, but small, at 0.4 percentage points, similar to the average across countries (0.5 percentage points). The bottom 40 in relative Around 3.5 billion people were living terms: the shared in countries in which the shared prosper- prosperity premium ity premium was positive in circa 2008–13. This represents 65 percent of the popu- The second World Bank goal, enhancing lation covered by the sample of countries shared prosperity, focuses on growth in monitored (5.4 billion people, or 75 per- the income or consumption of the bottom cent of the world population). In contrast, 40, but a comparison between this and the almost 1.9 billion people, or 35 percent of income or consumption growth of the en- the population covered, are living in coun- tire population, that is, the growth of the tries that experienced a negative premium. mean, supplies insights into how the gains China and India, which, together, represent of economic growth are shared across so- 37 percent of the world population and 49 ciety more generally. Such a comparison percent of the population covered in the indicates the extent to which distributional monitored sample, drive much of these ag- changes favor this group relative to the gregate results, with opposed shared pros- top 60. The additional growth represents a perity premiums of 0.6 and −0.5 percentage premium among the bottom 40 relative to points, respectively. the mean, the shared prosperity premium. Regionally, 7 of 8 countries in East Asia If the premium is positive, the bottom 40 and Pacific (almost 100 percent of the re- outperforms the average growth rate of gional population) and 12 of 16 countries total income and of the top 60. If the pre- in Latin America and the Caribbean (74 mium is negative, the overall growth rate percent) show positive shared prosperity and that of the top 60 exceed that of the premiums, while this is true of 3 of 4 coun- bottom 40.5 tries in South Asia (only 14 percent of the Consider the stark comparison between regional population because of India’s neg- Cambodia and Cameroon, two countries ative shared prosperity premium). Less en- with similar average consumption growth couraging are the results in Eastern Europe across the population, 3.9 percent and 3.7 and Central Asia, in industrialized coun- percent, respectively, circa 2008–13, but tries, and in Sub-Saharan Africa, where showing different growth values among the positive and negative shared prosperity pre- bottom 40: high in Cambodia, 6.5 percent, miums are observed in similar proportions but weak in Cameroon, 1.3 percent. This within each of the three regions. This is also implies that the growth in consumption of true of the Middle East and North Africa,

56 POVERTY AND SHARED PROSPERITY 2016 but this only includes a limited sample of 2 ically larger in absolute terms than the gain countries. of the average household in the bottom 40. The average population-weighted shared Of the 83 countries in the sample in prosperity premium is positive in all re- 2008–13, the income or consumption of gions, save South Asia, where India’s large the bottom 40 in 52 countries grew more population and negative premium heavily quickly relative to the top 10, while the op- influences the negative regional average. posite occurred in the remaining 31 coun- In the remaining regions, the shared pros- tries. The differences in the growth rates of perity premiums are positive: under 1 per- the bottom 40 and the top 10 are reported centage point in East Asia and Pacific (0.7), in the Palma premium column in table 3.1. Eastern Europe and Central Asia (0.3), and The larger number of countries with pos- Sub-Saharan Africa (0.6) and in excess of 1 itive Palma premiums suggests that more percentage point in Latin America and the countries experienced narrowing income Caribbean (1.4) and the Middle East and inequality than widening inequality over North Africa (2.7). Industrialized countries the period. An examination of the Gini have a meager shared prosperity premium index in a separate dataset—PovcalNet— of only 0.2 percentage points. confirms this (see chapter 4).8 These en- The share of countries that experienced couraging results were driven primarily a positive shared prosperity premium was by three regions, namely, East Asia and practically the same in circa 2008–13 as in Pacific, Latin America and the Caribbean, 2007–12. Around 60 percent of the coun- and South Asia. tries in the sample reported a positive pre- Care should be taken in interpreting mium in both rounds. This comparison re- these results, however, given the well-known quires caution, however. Possible changes in shortcomings in household survey data, the shares of countries with positive premi- from which the Palma premium is calcu- ums respond to changes both in the growth lated, namely, the high rates of survey non- performance of countries and in the com- response at the top; the underreporting of position of the sample of countries, and the incomes, particularly (but not only) capital impact of the latter can easily dominate that incomes; and the use of consumption (in- of the former. stead of incomes) in many countries, thus omitting the greater savings at the top.9 Incomes of the bottom 40 Who are the bottom 40? and the top 10: the Palma premium People in the bottom 40 differ in income or consumption across countries. They The shared prosperity premium fails to also differ in other dimensions of well- capture much of the variation in incomes being such as the educational performance that affects the topmost earners. To account of children, women’s access to health care for such intragroup differences, especially services, food insecurity and child stunting, among the top earners, an indicator com- access to safe water, and access to the Inter- paring income growth among the bottom net, as recently detailed in the World Bank 40 and the top 10 is used, the Palma pre- Global Monitoring Report 2015/16.10 mium.7 Similar to the shared prosperity pre- Moreover, the populations monitored mium, which compares the relative growth to construct the indicator of the growth in in income or consumption among the bot- income or consumption among the bottom 40 and the mean, the Palma premium tom 40 differ from the global extreme poor. reports the growth rate in income or con- Measured according to the global poverty sumption among the bottom 40, minus the line of US$1.90 (2011 purchasing power growth rate in the top 10. Even if the top 10 parity [PPP]) per person a day, all the poor achieves lower income growth than the bot- in, say, Brazil, China, Honduras, India, or tom 40, the income or consumption gain of South Africa have similar incomes below the average household in the top 10 is typ- the monetary threshold, that is, less than

SHARED PROSPERITY 57 FIGURE 3.2 The Bottom 40, Brazil, India, and the United States, the group of people on whom the second circa 2013 World Bank goal, the shared prosperity goal, 64,000 United States, 2013 focuses is not the same as the poor globally, on whom the ﬁrst goal focuses. The most recently updated estimates available show 32,000 Brazil, 2013 that, while 88.5 percent of the extreme poor 16,000 in the world are among the bottom 40 in their respective countries, around 11.5 per- 8,000 cent of the world’s extreme poor are not 4,000 India, 2011 captured by the national bottom 40 thresholds (box 3.1). This is because the incidence 2,000 of extreme poverty in the countries in which

(2011 PPP, log scale) (2011 PPP, these people live exceeds 40 percent. Con- 1,000 versely, 76 percent of the people who are among the bottom 40 in their countries are Annual income/consumption per capita 500 not among the extreme poor according to 250 the US$1.90-a-day global poverty line. 12345678910 The distinctions between the bottom 40 Deciles and the extreme poor are relevant in mon- United States 2013, bottom 40 itoring poverty and shared prosperity and India 2011, bottom 40 Brazil 2013, bottom 40 in policy design. Progress in the extreme Bottom 40, mean values poverty goal does not automatically mean Decile mean values progress in shared prosperity and vice versa. The people in the bottom 40 have different Sources: Updated from Lakner and Milanovic´ 2013; World Bank 2015. incomes across countries (figure 3.3). In some, such as the Democratic Republic of US$1.90 PPP. However, in the case of shared Congo and Togo, all the bottom 40 are also prosperity, income or consumption among among the extreme poor. In contrast, only a the bottom 40 may differ considerably fraction of people in the bottom 40 in Chile, across countries. Figure 3.2 illustrates this by Russia, and Turkey are also living below plotting the average income or consumption the international poverty line. A significant for each decile of the income distribution share of the bottom 40 in these countries in three countries at starkly different levels are living on five or more times the equiv- of development, namely, Brazil, India, and alent of the global poverty line. Meanwhile, the United States, in around 2013. Clearly, in other countries, the bottom 40 are liv- the bottom 40 in these three countries ex- ing on less than five times the global pov- perience different living standards. Thus, erty line. This is the case in Armenia, Bra- the mean incomes of the bottom 40 in Bra- zil, Costa Rica, the Kyrgyz Republic, Peru, zil are equivalent to the living standards of and Thailand, in which the shares of the the ninth richest decile in India. An analo- extreme poor among the bottom 40 are gous comparison holds between the United insignificant. In these countries, monitor- States and Brazil. Whereas the annual mean ing the two World Bank goals separately income among the bottom 40 in the United is necessary to understand with precision States is US$8,861 per person (in 2011 PPP), the progress in achieving better living con- the bottom 40 earn US$1,819 in Brazil and ditions among those most in need. In the US$664 in India, about 13 times less than in latter two groups of countries, policy inter- the United States. Only the top 10 in India ventions that reduce extreme poverty may earn sufficient average incomes to be part of or may not be effective in boosting shared the bottom 40 in the United States if that is prosperity if the two groups—the poor and where they had been located. the bottom 40—are composed of distinct An immediate consequence of this coun- populations. Chapters 5 and 6 examine try heterogeneity in absolute income or policy interventions that may boost shared consumption across the bottom 40 is that prosperity successfully.

58 POVERTY AND SHARED PROSPERITY 2016 BOX 3.1 The Bottom 40 versus the Poor

The overlap between the populations while 11.5 percent were among the top in the bottom 40 and the global poor 60 (yellow area). In 2013, the bottom 40 varies across countries. Figure B3.1.1 represented 2.9 billion people (red and illustrates how the extreme poor, blue areas). Of this group, 24.0 percent the bottom 40, and the top 60 were (red area) were among the extreme poor, distributed globally in 2013. The full and 76.0 percent (blue area) were not. area represents the world population. In This illustrates how the extreme poor 2013, the extreme poor (yellow and red are mostly situated within the bottom 40 areas) covered about 10.7 percent of the of their countries of residence. However, world population. Of the extreme poor, it also demonstrates that a large share of 88.5 percent were within the bottom 40 the bottom 40 across the world are not in their respective countries (red area), among the extreme poor.

FIGURE B3.1.1 Distribution of the Extreme Poor, the Nonpoor, the Bottom 40, and the Top 60, 2013

100

40 National percentile 30 20

0 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 Cumulative global polulation

Top 60, nonpoor Bottom 40, nonpoor Top 60, poor Bottom 40, poor

Is the poverty goal a pattern of slowing growth among the attainable at current levels main economic powerhouses, including of growth and shared Brazil, Russia, India, China, and South Af- rica (together known as the BRICS), the Eu- prosperity? ropean Union (EU), and the United States. Economic growth has been sluggish world- These economies are experiencing double wide since the global ﬁnancial crisis that or triple dips in growth. The ﬁrst dip took began in 2008. Figure 1.3 (chapter 1) shows place in 2009, when the per capita growth

SHARED PROSPERITY 59 FIGURE 3.3 Income Group Composition, the Bottom 40, Selected Countries, Circa 2013

Thailand Russian Federation Turkey Chile Kyrgyz Republic Costa Rica Armenia Peru Brazil Honduras India Togo Congo, Dem. Rep.

0 102030405060 70 80 90 100 Share of population (%) Extreme poor (< $1.90 a day) < 2x the poverty line < 5x the poverty line > 5x the poverty line

Source: Updated from World Bank 2015. Note: Countries are sorted by increasing poverty headcount ratio.

in gross domestic product (GDP) of these changes in distribution has been the main economies plummeted to negative values driver of poverty reduction. However, it is or zero, followed by a strong rebound in a mathematical fact that, in a context of a 2010. Growth then declined again, mark- slowdown in growth, a more equal distri- ing a second dip. In the United States, the bution will be required to achieve the same dip was short-lived because the economy poverty reduction. rebounded in 2012, but then dipped once This begs the question of whether the more in 2013 (the third dip). In Europe, the goal of ending poverty is still within reach rebound took longer and materialized in if such sluggish growth rates and inequal- 2014. Growth in the BRICS economies has ity levels were to persist. Simulations have been slowing since 2011. been conducted to address this question More recently, the worldwide growth (box 3.2). Figure 3.4 shows the results of outlook has been revised downward sev- simulations of the trajectory of the global eral times, especially among commodity- poverty headcount ratio under various as- exporting countries. For example, the World sumptions about the level of growth and Bank growth update in June 2016 reduced changes in the distribution, as measured by its estimate of global GDP growth for 2017 the shared prosperity premium (see above). from 3.1 percent to 2.8 percent relative This helps simulate the impact that changes to the forecast in January 2016 and from in the shared prosperity premium have on 3.2 percent to 2.4 percent for commodity- the global headcount ratio over time. As- exporting emerging and developing econ- suming, in these simulations, that every omies.11 Longer-term projections also sug- economy continues to grow at the rate of gest declining growth rates in 2010–30, well the 10 years leading up to 2013, the poverty below the 2000–10 average rates.12 These goal can only be reached with a positive downward revisions of economic growth shared prosperity premium (panel a, ﬁgure should not lead to doubts about the ability 3.4). This means that the bottom 40 needs of growth to reduce poverty and the com- to grow on average more quickly than the pelling evidence that growth rather than mean. The solid line in panel a, ﬁgure 3.4,

60 POVERTY AND SHARED PROSPERITY 2016 BOX 3.2 Simulating Poverty Trajectories

The simulations in figure 3.4 begin with imposed on all countries. A linear the most updated household surveys growth incidence curve is assumed so in 2013.a In the simulations up to 2030, as to specify the distributional change country mean income is assumed to fully, which, for simplicity, is imposed grow at a constant country-specific on all countries regardless of past rate based on historic growth rates performance. This is only one of the in the national accounts, adjusted for possible assumptions about the nature the observed differences between the of the pro-poor growth that could take national accounts and the growth shown place in the future among the infinite in household surveys. For the historic ways one might attain a given positive growth rates, two periods are presented shared prosperity premium consistent here: the 10 years and 20 years leading with a specific growth rate in the up to 2013. For country population, mean. The report website incorporates United Nations projections are used.b an interactive online tool that allows Different distributional changes defined users to produce their own simulations by the shared prosperity premium are independently by choosing a growth incorporated. This premium is defined rate for the mean across countries and as the difference between the growth for the shared prosperity premium. of the bottom 40 and the growth in The Stata program developed for these the mean, and it can be positive (the simulations is also publicly available on pro-poor scenario) or negative (the pro- the website. This program can be used rich scenario). The different headcount with both grouped and microdata and trajectories correspond to a single value accommodates a range of functional for the shared prosperity premium forms of the growth incidence curve.

Source: Lakner, Negre, and Prydz 2014. a. See Lakner, Negre, and Prydz (2014) for details on the methodology. These household surveys are also used in the estimation of the regional and global poverty headcount ratios reported in this chapter and in chapter 2. b. Health Nutrition and Population Statistics (database), World Bank, Washington, DC, http:// databank.worldbank.org/data/reports.aspx?source=health-nutrition-and-population-statistics.

shows the distribution-neutral scenario, to be clear that this is the analytical result which is the poverty trajectory without any from a set of simulations. However, in prac- distributional change, that is, if the growth tice, this does not mean that every country of the bottom 40 and the growth of the must improve the distribution of incomes mean are equal. Without distributional to achieve the poverty goal by 2030. The changes, more than 4 percent of the world’s result indicates that, under current average population are projected to be poor in 2030, growth trajectories, reductions in inequal- above the World Bank 3 percent goal. ity will be key to reaching the poverty goal Under a scenario whereby growth is more by 2030. This is so under speciﬁc assump- inclusive, the global poverty headcount is tions about how economic growth and simulated to reach less than 3 percent in the related distributive changes will occur 2030. This would be achieved through a up to 2030. For the poverty goal to be ac- shared prosperity premium of 1 percentage complished by 2030, improvements in the point, that is, a scenario whereby the growth distribution of incomes need to take place in the income of the bottom 40 exceeds, on among countries in which there are high average, the growth in the income of the numbers of poor, relatively wide inequality mean by 1 percentage point worldwide. If levels, and weak economic growth. the premium were 2 percentage points, the These simulations offer a good indica- goal of a 3 percent global headcount ratio tion of the challenges ahead. Under a sce- would be achieved by 2025. It is important nario based on the 10-year historic growth

SHARED PROSPERITY 61 FIGURE 3.4 Boosting Shared Prosperity and Ending Poverty, 2013–30

a. 10-year growth rates b. 20-year growth rates 12 12

10 10

8 8

6 6

4 4 3 3 2 2

Share of global population living < $1.90/day (%) 1 Share of global population living < $1.90/day (%) 1

2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030

m = 0 (distribution neutral) m = 1 (falling inequality) m = 2 (falling inequality) m = –1 (rising inequality) m = –2 (rising inequality) 2030 goal

Source: Updated results based on Lakner, Negre, and Prydz 2014. Note: m = the assumed shared prosperity premium, that is, the growth in income or consumption among the bottom 40, minus the growth in income or consumption at the mean. For example, m = 2 indicates that the growth in income among the bottom 40 exceeds the growth in income at the mean in each country by 2 percentage points.

rates, a shared prosperity premium of at 2030 if the average 20-year growth rates per- least 1 percentage point would be needed sist until 2030. to achieve the poverty goal by 2030. This is Based on these simulations, the goal of twice the average level of shared prosperity eliminating extreme poverty by 2030 can- premiums observed between 2008 and 2013 not be achieved unless prosperity is shared (0.5). This stresses the importance of accel- more quickly, growth rates are greater than erating the sharing of prosperity especially the 10- and 20-year historical averages used within countries having large numbers of in the simulations, or both. Nonetheless, the extreme poor. Pro-rich scenarios with this is entirely possible, particularly be- negative shared prosperity premiums, that cause these findings are simulations, not is, the income growth of the mean exceeds predictions.13 that of the bottom 40, show that the 3 percent goal in the global headcount ratio Conclusions: continued would not be reached by 2030. progress, but no room Panel b in figure 3.4 shows that the simulations based on the country-specific for complacency growth rates over the last 20 years before The diagnosis of shared prosperity in the 2013 (which, on average, are lower than countries on which data are available circa the 10-year historic growth rates) require 2008–13 is fairly favorable in terms of both an even greater shared prosperity premium the headline indicator of growth in income to achieve the poverty goal. For instance, or consumption among the bottom 40 and under the distribution-neutral growth path, the shared prosperity premium. In 60 of the the projected poverty headcount in 2030 83 countries on which it is possible to con- is over 5 percent with the 20-year growth struct these indicators, there has been prog- rates, compared with 4 percent with the last ress toward achieving shared prosperity, 10-year growth rates. Indeed, the average and, in 49 countries, the shared prosperity shared prosperity premium of 1 percentage premium has been positive. This means that point will not fulfill the World Bank goal by 89 percent and 65 percent of the population

62 POVERTY AND SHARED PROSPERITY 2016 covered in this sample are living in coun- data are population-weighted), the average tries showing positive advances in shared shared prosperity premium was only 0.5 prosperity or in the shared prosperity pre- percent during the same spell (or 0.4 per- mium, respectively. The Palma premium, cent if the data are population-weighted). which measures the difference between the Is this average in the premium sufficient to bottom 40 and the top 10 in average in- support large reductions in inequality and come growth, is positive in 52 of the 83 (or poverty so as to achieve the World Bank’s 67 percent of the population covered in the twin goals by 2030? Simulations in this sample). chapter suggest that greater prosperity shar- This suggests that the number of coun- ing will be required to end poverty by 2030. tries experiencing improvements in the Behind the global aggregates lie large living conditions of the bottom 40 and de- regional differences and large differences clining relative income inequality is greater within each region. Among high-income than the number of countries experiencing countries, Greece has experienced a contrac- deterioration in living conditions and rising tion of 10.0 percent a year in the incomes of inequality. This is confirmed through an ex- the bottom 40, while, in Eastern Europe and amination of the Gini index from a separate Central Asia, Belarus has reported a rise in dataset, PovcalNet (see chapter 4).14 This the incomes of the bottom 40 in excess of progress is all the more significant given an annualized 8.0 percent. In Latin Amer- that it has taken place in a period marked ica and the Caribbean, the incomes of the by the global financial crisis of 2008–09, bottom 40 contracted by more than 2.5 per- the magnitude of which had not been seen cent in Honduras, while, in Paraguay, they since the Great Depression of the 1930s. expanded by 8.0 percent annually. In Sub- There was also a crisis involving spikes in Saharan Africa, the annualized growth in food prices in 2008 and 2010 with profound consumption among the bottom 40 aver- effects on developing countries. aged 9.6 percent in the Democratic Repub- Nonetheless, it is a source of concern that lic of Congo during 2008–13. 40 percent of the population covered in the The results reported here are subject sample is living in countries where income to significant caveats on data quality, par- or consumption is growing more quickly ticularly the accuracy of the income and among the top 60 than among the bottom consumption data on the top of the distri- 40, and a 10th of the population covered in bution, which is known to be severely under- the sample is living in countries in which represented in the household surveys used income among the bottom 40 is contract- in this monitoring exercise. Information ing. This reinforces the need for prudential available in the World Wealth and Income macroeconomic policies to prevent such Database suggests that accounting for bet- large systemic crises, as well as policy inter information on top incomes (typically terventions that raise income-earning op- by accessing income tax data) can change portunities among the poor, enable their trend conclusions across countries (also human capital accumulation, and protect see box 4.5, chapter 4).15 Furthermore, re- them from risks (see chapters 5 and 6). garding data availability, it is worth stress- Another source of concern is the small ing that before the results of this monitor- value of shared prosperity. While the av- ing exercise can be truly global, around 25 erage growth of income and consumption percent of the world’s population will need among the bottom 40 worldwide was 2.0 to be included in the Global Database of percent circa 2008–13 (4.3 percent if the Shared Prosperity.16

SHARED PROSPERITY 63 Annex 3A

Shared prosperity estimates based on the latest surveys, by country, circa 2008–13

TABLE 3A.1 Shared Prosperity Estimates, Circa 2008–13

(Box continues next page)

64 POVERTY AND SHARED PROSPERITY 2016 TABLE 3A.1 Shared Prosperity Estimates, Circa 2008–13 (continued)

Mean consumption or income per capita (US$ a day PPP)c Annualized growth in mean consumption or income per capita (%)c, d Baseline Most recent year Country Perioda Typeb Bottom 40 Total population Bottom 40 Total population Bottom 40 Total population Kyrgyz Republic 2009–14 c 0.40 −1.09 3.08 5.57 3.14 5.28 Lao PDR 2007–12 c 1.53 2.24 1.90 3.84 2.05 4.29 Latvia 2007–12 i −3.04 −4.33 9.69 22.38 8.31 17.94 Lithuania 2007–12 i −1.77 −1.16 10.14 20.99 9.28 19.79 Luxembourg 2007–12 i −2.67 −0.54 38.29 72.80 33.44 70.85 Macedonia, FYR 2009–13 I 4.98 0.73 3.36 9.46 4.08 9.74 Mauritius 2006–12 c 0.76 0.86 5.31 11.02 5.54 11.56 Mexico 2010–14 i 0.66 0.96 3.42 10.29 3.51 10.69 Moldova 2009–14 c 4.84 1.32 4.33 8.76 5.48 9.35 Mongolia 2010–14 c 8.03 7.05 4.01 8.05 5.46 10.58 Montenegro 2009–14 c −2.72 −2.27 8.64 16.27 7.53 14.51 Netherlands 2007–12 i −0.01 −0.99 28.06 51.72 28.05 49.21 Nicaragua 2009–14 i 4.71 4.72 2.62 7.54 3.30 9.50 Norway 2007–12 i 3.17 2.39 33.37 58.45 39.00 65.77 Pakistan 2007–13 c 2.81 2.53 2.07 3.81 2.44 4.42 Panama 2009–14 i 4.14 3.58 4.83 17.38 5.91 20.72 Paraguay 2009–14 i 8.01 8.16 3.80 12.68 5.59 18.77 Peru 2009–14 i 5.78 3.11 3.71 11.96 4.91 13.94 Philippines 2006–12 i 1.71 1.22 2.17 6.42 2.40 6.91 Poland 2007–12 i 2.57 2.26 9.68 19.97 10.98 22.34 Portugal 2007–12 i −1.99 −2.14 12.89 27.97 11.65 25.11 Romania 2007–12 i 2.59 1.62 3.71 8.80 4.21 9.54 Russian Federation 2007–12 c 5.86 5.27 7.60 19.42 10.10 25.11 Rwanda 2010–13 c 0.04 −0.57 0.92 2.76 0.93 2.71 Senegal 2005–11 c −0.01 0.54 1.30 3.06 1.29 3.16 Serbia 2008–13 c −1.73 −1.13 7.60 13.44 6.96 12.70 Slovak Republic 2007–12 i 5.48 6.67 12.46 20.27 16.27 28.00 Slovenia 2007–12 i −0.84 −0.28 20.64 33.44 19.79 32.97 Spain 2007–12 i −1.32 0.00 17.14 36.25 16.04 36.25 Sri Lanka 2006–12 c 2.21 1.66 2.96 6.80 3.37 7.51 Sweden 2007–12 i 2.04 2.25 26.22 45.14 29.01 50.46 Switzerland 2007–12 i 2.43 0.93 30.49 63.18 34.38 66.19 Tanzaniaf 2007–11 c 3.36 1.42 1.05 2.49 1.23 2.67 Thailand 2008–13 c 4.89 3.47 5.15 12.45 6.54 14.77 Togo 2011–15 c 2.76 0.82 0.89 2.63 0.99 2.71 Turkey 2008–13 c 3.18 3.54 5.94 14.29 6.94 17.01 Uganda 2009–12 c 3.59 1.37 1.28 3.25 1.42 3.39 Ukraine 2009–14 c 3.93 3.29 6.51 10.74 7.89 12.63 United Kingdom 2007–12 i −1.67 −2.78 23.89 51.10 21.96 44.38 United States 2007–13 i −0.16 −0.43 — — — — Uruguay 2009–14 i 5.48 2.95 7.33 21.72 9.56 25.12 Vietnam 2010–14 c 4.51 2.00 3.29 7.61 3.93 8.24 Source: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity. Note: All estimates are in 2011 PPP U.S. dollars. — = not available. a. Refers to the years in which the underlying household survey data were collected. In cases in which the data collection period bridged two calendar years, the first year in which data were collected is reported. The range of years refers to two survey collections, the most recent survey within the range and the nearest survey collected five years before the most recent survey. For the final year, the most recent survey available between 2011 and 2015 is used. Only surveys collected between three and seven years before the most recent survey are considered for the earlier survey. b. Denotes whether the data reported is based on consumption (c) or income (i). Capital letters indicate that grouped data are used. c. Based on real mean per capita consumption or income measured at 2011 PPP using data in PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch .worldbank.org/PovcalNet/. On some countries, the means are not reported because of grouped or confidential data. d. The annualized growth rate is computed as (Mean in year 2/Mean in year 1)^(1/(Year 2 − Year 1)) − 1. e. Covers urban areas only. f. Ex ante evaluation of these surveys suggest that they are not comparable. However, the poverty assessment attempted to create a more comparable series and also applied additional methodological techniques to establish comparability and consistency among welfare aggregates.

SHARED PROSPERITY 65 Notes (2009–14), at 2.1 percentage points; yet, the bottom 40 gained only an extra US$1.38 a 1. GDSP (Global Database of Shared Prosper- day, while the population as a whole gained ity), World Bank, Washington, DC, http:// an extra US$3.35 on average. In Russia www.worldbank.org/en/topic/poverty/brief (2007–12), a small positive shared prosperity /global-database-of-shared-prosperity. premium translated into a gain of US$2.50 2. As a consequence, the results on shared among the bottom 40 and a gain of US$5.69 prosperity may be affected by selection bias at the national mean. The more unequal a if countries having poorer survey coverage country is initially, the less effective a posi- also perform relatively poorly. tive shared prosperity premium becomes in 3. GDSP (Global Database of Shared Prosper- offsetting inequality. A more unequal coun- ity), World Bank, Washington, DC, http:// try is also much more likely to have a positive www.worldbank.org/en/topic/poverty/brief shared prosperity premium. If the bottom /global-database-of-shared-prosperity. 40 are quite poor, then every extra US$0.01 4. The sample-based average may even under- will make a large difference in the growth estimate the true share of the population rate. Conversely, if the top of the popula- exhibiting positive growth in the income of tion distribution is quite rich, they will need the bottom 40 worldwide. This is so because much more extra income to make a dent in the dataset used for this analysis covers 75 the growth rate in income. A positive shared percent of the total population. To the ex- prosperity premium is a necessary first step tent that the remaining 25 percent includes in promoting equality, but it is rarely suffi- countries in which the incomes of the bot- cient alone to close the inequality gap. tom 40 grew, the reported average underesti- 6. Reported changes refer to 2008–12 in Cam- mates the true value. bodia and 2007–14 in Cameroon. 5. The shared prosperity premium is a relative 7. The premium is named after José Gabriel measure in the sense that it analyzes income Palma, a Chilean economist who has long gains relative to initial income. For an indi- been devoted to the study of inequality. The vidual with an initial income of US$1.00, a Palma premium is inspired by the Palma 10 percent increase in income means an extra ratio, but they are not identical. The latter US$0.10 in income, whereas, for an individ- is the ratio of the income share of the top ual with an income of US$1,000, a 10 percent 10 to that of the bottom 40, while the pre- increase corresponds to an extra US$100. A mium is defined as the difference in income positive shared prosperity premium means growth among these groups. See Cobham that the relative income or consumption and Sumner (2016); Palma (2016). growth rate among the bottom 40 is larger 8. PovcalNet (online analysis tool), World than the relative growth rate in the rest of Bank, Washington, DC, http://iresearch society. However, this does not necessarily .worldbank.org/PovcalNet/. mean that the bottom 40 gain more income 9. The issue of missing top incomes is dis- in absolute terms. If the initial inequality is cussed in more detail in chapter 4, box 4.5. wide, the bottom 40 may show above aver- The chapter also discusses evidence based age growth in incomes, but still gain less in on tax records in selected countries show- absolute terms than the average individual. ing that top income shares have increased Chile exhibited a positive shared prosperity sharply in recent years. However, this type premium in 2009–13. The incomes of the of evidence is not available in all countries. bottom 40 grew an average of 1.4 percentage An increasing income share among the top points a year more quickly than the national 1 percent is also not necessarily inconsistent average. However, this translated into an in- with a positive Palma premium, that is, with come gain per capita of only US$1.49 a day more rapid growth in the incomes of the among the bottom 40, whereas the entire bottom 40 than in the incomes of the top 10. population received an average extra income 10. World Bank (2016a). of US$3.54 a day. Growth in Brazil and Rus- 11. World Bank (2016b). sia followed a similar pattern. The shared 12. IMF (2015, 2016); OECD (2014); World prosperity premium in Brazil was positive Bank (2016b).

66 POVERTY AND SHARED PROSPERITY 2016 13. Relative to historic growth rates, the growth Hoy, Chris, and Emma Samman. 2015. “What rates over the last 10 years have been rather If Growth Had Been as Good for the Poor high, especially in developing countries. as Everyone Else?” Report (May), Overseas Furthermore, there are substantial down- Development Institute, London. side risks to any long-run growth projection, IMF (International Monetary Fund). 2015. for example, because of climate change World Economic Outlook, October 2015: Ad- (Hallegatte et al. 2016). justing to Lower Commodity Prices. World 14. PovcalNet (online analysis tool), World Economic and Financial Surveys Series. Bank, Washington, DC, http://iresearch Washington, DC: IMF. .worldbank.org/PovcalNet/. ———. 2016. “Subdued Demand, Diminished 15 WID (World Wealth and Income Database), Prospects.” World Economic Outlook Up- Paris School of Economics, Paris, http:// date, January 19, IMF, Washington, DC. www.parisschoolofeconomics.eu/en/research Lakner, Christoph, and Branko Milanovic´. 2013. /the-world-wealth-income-database/. See “Global Income Distribution: From the Fall Alvaredo et al. (2013). of the Berlin Wall to the Great Recession.” 16. GDSP (Global Database of Shared Prosper- Policy Research Working Paper 6719, World ity), World Bank, Washington, DC, http:// Bank, Washington, DC. www.worldbank.org/en/topic/poverty/brief Lakner, Christoph, Mario Negre, and Espen Beer /global-database-of-shared-prosperity. Prydz. 2014. “Twinning the Goals: How Can Shared Prosperity Help to Reduce Global References Poverty?” Policy Research Working Paper 7106, World Bank, Washington, DC. Alvaredo, Facundo, Anthony B. Atkinson, OECD (Organisation for Economic Co- Thomas Piketty, and Emmanuel Saez. 2013. operation and Development). 2014. OECD “The Top 1 Percent in International and His- Economic Outlook. Vol. 2014/2 (November). torical Perspective.” Journal of Economic Per- Paris: OECD. spectives 27 (3): 3–20. Palma, José Gabriel. 2016. “Measuring Income Beegle, Kathleen, Pedro Olinto, Carlos E. Sob- Inequality: Comments on ‘Do Nations Just rado, Hiroki Uematsu, and Yeon Soo Kim. Get the Inequality They Deserve? The “Palma 2014. “Ending Extreme Poverty and Pro- Ratio” Reexamined.’” In Regions and Regular- moting Shared Prosperity: Could There Be ities, vol. 2 of Inequality and Growth: Patterns Trade-Offs between These Two Goals?” In- and Policy, edited by Basu Kaushik and Joseph equality in Focus 3 (1): 1–6. E. Stiglitz, 35–97. International Economic As- Cobham, Alex, and Andy Sumner. 2016. “Is It All sociation Series. London: Palgrave Macmillan. About the Tails? The Palma Measure of In- World Bank. 2015. A Measured Approach to End- come Inequality.” Center for Global Develop- ing Poverty and Boosting Shared Prosperity: ment Working Paper 343, Center for Global Concepts, Data, and the Twin Goals. Policy Re- Development, Washington, DC. search Report. Washington, DC: World Bank. Hallegatte, Stephane, Mook Bangalore, Laura ———. 2016a. Global Monitoring Report Bonzanigo, Marianne Fay, Tomaro Kane, 2015/16: Development Goals in an Era of De- Ulf Narloch, Julie Rozenberg, David Treguer, mographic Change. Washington, DC: World and Adrien Vogt-Schilb. 2016. Shock Waves: Bank. Managing the Impacts of Climate Change on ———. 2016b. Global Economic Prospects, June Poverty. Climate Change and Development 2016: Divergences and Risks. Washington, DC: Series. Washington, DC: World Bank. World Bank.

SHARED PROSPERITY 67

4 Inequality

Chapter 4 first explains why inequality of outcomes—specifically, the inequality in income or consumption expenditure—matters both by itself and in the context of reducing poverty. Second, it documents trends in income inequality, distinguishing between global, between-, and within-country inequalities. In doing so, it relies on a recent and comprehensive sample of countries. Evidence points to several facts, some largely acknowledged, others likely to surprise. Fact number one: global inequality—the inequality among all citizens worldwide, regardless of country of residence—increased from the industrial revolution through the 1980s. Fact number two: an exceptional period of falling global inequality has been observed since the early 1990s, which has been especially marked since 2008. Fact number three: global inequality is wider today than it was in the 1820s, despite the recent reduction. Fact number four: most of the recent reduction in global inequality derives from the convergence of income among countries mostly because of the rapid growth in populous developing countries, notably China and India. Fact number five: the share of income going to the top 1 percent is known to have increased in many countries on which information is available. Fact number six: within-country inequality for the average country, that is, considering the country-specific trends in all countries, only started to narrow in 1998. Fact number seven: for every country showing an increase in the Gini index of more than one Gini point between 2008 and 2013, there are more than two other countries showing a reduction in the Gini by the same amount. And fact number eight: while inequality may certainly widen, it can also narrow. Despite all these facts, inequality remains unacceptably high: the Gini exceeds 50 in several countries, and, in Haiti and South Africa, it even exceeds 60.

Inequality matters The strong progress in poverty reduction a reduction in inequality, or a combination and shared prosperity that took place over of the two.1 So, to achieve the same poverty the ﬁrst decade of the 2000s is at risk be- reduction during a slowdown in growth, a cause of the global slowdown in growth. more equal distribution is required. Indeed, the World Bank goal of eliminating Inequality is the thematic focus of this extreme poverty by 2030 cannot be achieved year’s report. Addressing inequality because in the current global context without signif- it slows down poverty reduction implies an icant shifts in within-country inequalities. instrumental concern for inequality, that Poverty reduction in a country can typically is, addressing inequality is only a means to be decomposed into higher average growth, confront another problem, namely, abso-

INEQUALITY 69 lute poverty. This report examines inequal- erty rights by the poor, and if institutions ity because of this instrumental concern, are inequitable (disproportionally benefit- but also for intrinsic reasons. Individuals ing the wealthy through, for example, re- express concern with rising inequality, gressive subsidies). To the extent that this broadly defined. In fact, their perceptions of efficient redistribution breaks off the inter- increasing inequality—even though objec- generational reproduction of inequalities, tive measures of inequality declined—have it both addresses the roots and drivers of been argued to be one of the factors con- inequality and lays out the foundations for tributing to the Arab Spring (see box 4.2). long-term growth, which helps end pov- Furthermore, evidence from experimental erty and expand shared prosperity. Other- studies suggests that individuals do not act wise, societies may fall into inequality traps exclusively in a purely self-interested man- whereby children in disadvantaged fam- ner, but often reveal a deep-rooted con- ilies lack the same opportunities to attend cern for fairness in outcomes.2 Leveling the good-quality schools, end up earning less in playing field, that is, reducing inequality of the labor market, or continue to have less opportunity, therefore relates to notions of voice in the political process, and the cycle fairness and justice and resonates across so- of underachievement persists. cieties on its own merits.3 Evidence also suggests that there are Inequality in outcomes and inequality no inevitable trade-offs between efficiency of opportunity are intimately connected.4 and equity considerations. Cross-country While reasonable people might disagree studies of the effect of inequality on growth over the desirable level of inequality in have failed to produce evidence that sup- incomes or consumption expenditures or ports the existence of a trade-off in this whether a particular increase should be of case. Rather, they point to inconclusive concern, an important reason for caring results (box 4.1). In addition to these about inequality in outcomes today is that macrolevel studies, analysts have also used it leads to inequality of opportunity among microlevel evidence to examine equity- the next generation.5 If families have vastly efficiency trade-offs associated with specific different economic resources, some children policies, where they may certainly exist. For in some families will face an unfair start in example, taxation may create disincentives life, and public policy will have to make a on labor supply or drive economic activity great effort to overcome these differences in into informality. In Mexico, efficiency costs initial conditions. At the same time, reduc- in the form of lower productivity and lower ing inequality of opportunity today reduces growth in gross domestic product (GDP), inequality of outcomes tomorrow (see caused by high social security contribu- chapter 6). The focus is therefore not only tions from formal employment and gener- on preventing the transmission of poverty ous noncontributory social safety nets, are and income disparities across generations, estimated to be on the order of 0.9 percent but also on the transmission of unfair ad- and 1.4 percent of GDP, respectively.7 An- vantages to the next generation. other classic example is consumer subsidies. Furthermore, narrowing inequality—as While these subsidies are typically intended in the case of reducing poverty—is compat- to protect the consumption of the poor, ible with boosting economic growth. World they often end up disproportionally bene- Development Report 2006: Equity and De- fiting the rich, who consume more in abso- velopment provides a strong empirical un- lute terms than the poor. For instance, the derpinning to the claim that interventions International Monetary Fund reports that, that narrow inequality—whether intended in a sample of 32 countries in the Middle or not—can also be good for growth and East and North Africa, Sub-Saharan Africa, long-term prosperity.6 This type of inter- and Latin America and the Caribbean, the vention can boost economic efficiency if richest 20 percent of the population cap- markets are missing or imperfect, because tures more than six times as much of the they fail to provide credit opportunities, benefit of fuel subsidies as the poorest 20 insurance against shocks, or access to prop- percent.8

70 POVERTY AND SHARED PROSPERITY 2016 BOX 4.1 Cross-Country Studies of the Effect of Inequality on Growth

A large body of literature uses cross- datasets.d World Development Report country data to examine whether 2006 found many examples of specific inequality is bad for subsequent growth. policies that reduce inequality and are Conceptually, the effect of inequality also good for growth, but this sort of might go either way: if higher inequality evidence has not been confirmed with leads to the more rapid accumulation of macroeconomic data.e The earliest savings, it may spur growth; if it leads macrolevel papers used data for a single to suboptimal investment in education cross-section of countries and found a or in health care, it may have a negative negative effect of inequality on growth.f effect on growth. Later papers, which used panel data, Recent research by the International found positive effects of inequality on Monetary Fund has claimed that subsequent growth.g Yet other studies lower inequality in disposable income argued for a nonlinear relationship, which is associated with more rapid and could explain these unstable results.h more durable growth for a given level Motivated by the ambiguity of the of redistribution.a The robustness of effect of inequality on growth both the findings has been questioned, conceptually and empirically, a recent however, because of the presence of set of papers decomposes overall weak instruments in the econometric inequality into components that may technique, a pervasive issue in this be especially harmful to growth. In field of research.b After accounting for particular, it may be expected that weak instruments, the coefficients turn inequality of opportunity is harmful for insignificant; so, it is not possible to growth, while the effect of inequalities draw any conclusion about the effect of that arise because of differences in inequality on growth, whether positive effort may act in the opposite direction.i or negative. Furthermore, the analysis There is some evidence that inequality relies on a dataset that extensively of opportunity may be bad for growth, at imputes inequality measures across least subnationally. Across U.S. states, countries, thus warranting further inequality of opportunity is found to validation of estimates as more data are have a negative effect on growth.j A collected.c similar effect is found across Brazilian Previous studies have also found municipalities.k Across countries in inconclusive results using a wider the world, however, there is no robust range of econometric techniques and evidence of a negative effect.l

Source: Ferreira et al. 2014. a. Ostry, Berg, and Tsangarides (2014); also see Dabla-Norris et al. (2015). b. Kraay (2015) shows that confidence intervals consistent with weak instruments are much wider. In particular, they include a wide range of positive and negative values for the coefficient on inequality in the growth regression. Kraay examines a number of papers using similar econometric techniques and considers alternative instrument specifications. This methodological cri- tique is not unique to Ostry, Berg, and Tsangarides (2014), but applies to many studies using system generalized method of moments estimators. These estimators are frequently used in cross-country growth regressions. c. The dataset being used is the Standardized World Income Inequality Dataset (Solt 2016). Jenkins (2015) offers a detailed description of the dataset and the conceptual issues with the cross-country imputations. d. See the summary by Voitchovsky (2009). e. World Bank (2005). f. Alesina and Rodrik (1994); Persson and Tabellini (1994). g. Forbes (2000). h. Banerjee and Duflo (2003). i. The decomposition of inequality into components related to opportunities and effort, respectively, has been suggested by Roemer (1998). j. Marrero and Rodríguez (2013). k. Teyssier (2015). l. Using two global datasets, Ferreira et al. (2014) do not find support for the proposition that inequality of opportunity is bad for growth.

INEQUALITY 71 This focus on inequality is not meant access to contributive forms of insurance) to question the critical role of economic as additional pillars to support the role of growth in improving the living conditions economic growth in poverty reduction. His- of the poor and, particularly, the bottom toric evidence confirms that countries fol- 40 percent of the income or consump- lowing such strategies have been more suc- tion distribution (the bottom 40).9 Volu- cessful in achieving a rapid pace and wide minous evi dence confirms that commonly breadth in poverty reduction.11 Bangla- acknowledged good macroeconomic pol- desh, Brazil, China, Ethiopia, and Vietnam icies—control of inflation, promotion of are some examples providing a compelling competition policy to reduce economic illustration.12 concentration, trade openness, and macro- Empirical evidence using cross-country economic stability—and good governance income data—the most recent and compre- are the foundations of sustainable growth, hensive covering 121 countries between which, ultimately, is good for the poor and 1967 and 2011—concludes that the av- the bottom 40.10 Others have highlighted erage incomes of the bottom 40 within the importance of investing in human de- each country tend to grow at the same velopment (improving access, but also qual- pace as the average incomes in the respec- ity and the institutional delivery of basic tive country.13 Figure 4.1 illustrates this services) and insuring against risks (espe- empirical finding with the most recently cially among the poor and those without available data. Countries in which growth at the mean is large also show the largest growth in the incomes of the bottom 40. FIGURE 4.1 Growth of the Bottom 40 versus Growth at the Mean, Conversely, countries with negative growth 2008–13 rates in the mean also show declines in the 10 incomes of the bottom 40. Consistent with CHN ZAR these two findings, evidence suggests that ECU BLR average income growth appears uncor- KHM PER BOL related with changes in the share of the in- MKD BRA SVK comes of the bottom 40. Between 1967 and 5 RUS VNM 2011, the average income growth in coun- IRN TZA COG tries was 1.5 percent a year during a typical five-year period, while changes in the share LAO CMR of incomes of the bottom 40 were close to 0 zero. As a result, some estimates suggest that GBR DNK ESP average income growth explains as much LVA as three-quarters of the variation in in- ITA come growth of the bottom 40.14 As coun- IRL –5 tries grow more quickly, the growth of the bottom 40 may be expected to increase as

Annual growth in the mean of bottom 40 (%) well. Thus, growing the economy, boosting shared prosperity, and reducing poverty are 15 –10 GRC three absolutely compatible goals. Although economic growth is essen- −10 −5 0 5 10 tial in sustaining improvements in living Annual growth in the mean (%) standards at the low end of the income distribution, growth alone typically falls Source: Calculations based on household survey data in GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of short of delivering the maximum sharing -shared-prosperity. of prosperity. In the empirical evidence Note: The ﬁgure shows annualized growth rates in per capita household income or consumption expendi- cited above, a quarter of the variation in tures over circa 2008–13. The red line is a 45-degree line, that is, in economies along this line, the bottom 40 grew at the same rate as the total population. The bottom 40 in economies below the line experienced the growth of the incomes of the bottom slower growth relative to the overall population. 40 across countries derives from sources

72 POVERTY AND SHARED PROSPERITY 2016 other than economic growth. It is also clear Including the less privileged in the pro- in figure 4.1 that, in a substantial number cess of growth and investing in their human of countries, the growth in the income or capital may also be good for growth. For ex- consumption of the bottom 40 lies below ample, removing credit constraints on the the growth in the mean. In such countries, poorest and the bottom 40 and developing the growth of the top 60 exceeds that of the insurance mechanisms for them are ex- bottom 40, and increasing inequality may pected, respectively, to lead to higher growth emerge from this type of growth. This is through productivity gains and to limit the the case even in settings characterized by impacts of natural disasters on economies. rapid economic growth (and, thus, high Improving the human capital of the poor growth in mean incomes) and lower-than- by promoting early childhood development average growth in incomes among the (ECD) and good-quality education; invest- bottom 40, such as the Republic of Congo ing in infrastructure that connects, for ex- and the Lao People’s Democratic Republic. ample, smallholder farmers with markets; Consequently, regardless of the importance and enhancing the coverage and quality of of economic growth in boosting shared electricity services have been shown to have prosperity, a premium associated with pov- positive effects on economic growth, while erty reduction that arises from narrowing improving the living conditions of the poor inequality needs to be directly exploited (see chapters 5 and 6). Policy choices and through specific policy interventions. For economic growth from now until 2030 will instance, from 2008 to 2013, Cambodia continue to affect the pace of sharing pros- and Cameroon had similar growth rates in perity and ending poverty. If these choices the mean (3.9 and 3.7 percent, respec- are made smartly, and economic growth is tively), but the average incomes of the strong and sustainable, the twin goals of the bottom 40 grew at 6.5 percent in the World Bank will be within reach.16 former and only 1.3 percent in the latter There are still other reasons why in- (see figure 4.1). (The specific choices a equality matters for the twin goals.17 Re- few countries have made to boost shared ducing inequalities of opportunity and of prosperity successfully and the specific pol- outcomes among individuals, populations, icy interventions that are more conducive and regions is conducive to political and so- to narrowing inequality are addressed in cial stability as well as social cohesion. Yet, chapters 5 and 6.) this outcome is not certain, as the recent

BOX 4.2 Perceptions of Inequality in the Middle East and North Africa

Inequality is a multifaceted Inequality in the Middle East steady progress in the equitable phenomenon; yet, discussions and North Africa region presents distribution of the gains from about it are often restricted to an illustrative case of marked economic growth.b Low by income-wealth-consumption differences between subjective comparison with other developing metrics.a Increasingly, however, assessments and objective regions, the population share of evidence from the ﬁeld on measures. These differences help the extreme poor had declined subjective well-being has explain the conditions leading to further in almost all countries.c In demonstrated the importance the Arab Spring in early 2011 that the early 2000s, income inequality of individual perceptions in the traditional metrics of income or was also moderately low in analysis of inequality, for example, wealth inequality failed to capture. comparison with other developing on satisfaction with basic services, From 1950 through the 1990s, regions. The region achieved the governance, or economic mobility. countries in the region had made Millennium Development Goals in

(Box continues next page)

INEQUALITY 73 BOX 4.2 Perceptions of Inequality in the Middle East and North Africa (continued)

poverty reduction and access to FIGURE B4.2.1 Actual versus Anticipated Feelings of Well-Being, infrastructure services (especially Middle East and North Africa Internet connectivity and drinking water and sanitation), and was also 8 successful in reducing hunger and child and maternal mortality and in d raising school enrollments. 7 Nonetheless, the period was also accompanied by a deterioration in critical factors associated with life satisfaction. These factors 6 include declining trust in the governance of institutions; declining standards in public services and

Cantril scale of well-being 5 local and national government accountability; an erosion in law and order and the impartiality and independence of the judiciary; 4 a growing shortage of formal 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 sector jobs offering job security Actual Anticipated and benefits, especially among educated youth; and a growing Source: Chattopadhyay and Graham 2015, based on data in Gallup World Poll, Gallup, Washington, DC, http://www.gallup.com/services/170945/world-poll.aspx. sense of despondency whereby Note: Data refer to Algeria, Bahrain, Egypt, the Islamic Republic of Iran, Iraq, Israel, Jordan, Kuwait, people believe their opportunities Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, West Bank for success are shrinking, economic and Gaza, and the Republic of Yemen. The estimates are population weighted. The 0–10 Cantril Scale of mobility is becoming divorced from Well-Being (the Cantril Self-Anchoring Scale), a metric of self-reported individual well-being, is presented their efforts, and future generations to respondents as follows: “Imagine a ladder with steps numbered 0–10. Suppose 10 represents the best possible life for you, and 0 represents the worst possible life for you. On which step of the ladder would are less likely to enjoy any gains in you say you personally feel you stand today (current)” and “about five years from now (anticipated).” For e standards of living. additional details, see Gallup (2011a). Average life satisfaction in the region since the early 2000s has been declining.f It is below the distribution and how the distribution objective and subjective measures global average and lower than the is perceived is expanding in the of satisfaction have consequences. average expected for countries region. Thus, in the Arab Republic The three most important at a similar level of development. of Egypt during the early 2000s, motivations of the unrest behind Part of this effect may be caused people at lower incomes felt they the Arab Spring cited by more by the unhappy growth paradox, were more affluent than they than half of the respondents to whereby rapid economic growth actually were, but this perception recent Arab Barometer surveys is destabilizing in the short had reversed by 2008–09, when are a desire to enhance economic run, particularly in transition the same groups thought they were conditions, fight corruption and economies. According to this less well off than they actually cronyism, and promote social and paradox, improving macroeconomic were.h This sense of declining economic justice.j conditions and the anticipation of economic fortunes in the lower Is the experience in the region benefits from this growth lead to end of the income distribution was likely to be replicated elsewhere? disappointment when the level compounded by a rising sense of Evidence across countries of self-reported well-being fails vulnerability and dissatisfaction suggests that the association to meet the expectations (figure among middle groups in the income between perceptions of relative B4.2.1).g distribution above the bottom 40.i inequality and reported life The systematic gap between The rising dissatisfaction satisfaction is neither linear nor trends in the actual income and the widening gap between unidirectional. Individuals and

(Box continues next page)

74 POVERTY AND SHARED PROSPERITY 2016 BOX 4.2 Perceptions of Inequality in the Middle East and North Africa (continued)

societies differ in who they regard genuinely regard such disparities distribution of prosperity.m The as a reference group, and this as the result of fair returns within combination of shifting attitudes leads to differences in subjective a meritocratic system, whereas and growing frustration because of evaluation. In addition, people other societies consider the long-run unmet expectations may tend to misstate their own income same income disparities as constitute a serious threat to social and wealth.k Moreover, different unacceptable.l The Arab Spring and political stability, especially societies view the signiﬁcance demonstrates that attitudes toward if macroeconomic growth is of relative inequality differently. historical norms can change if a perceived to have no beneﬁts for Some countries are more tolerant broadening share of a population low- and middle-income population of greater income disparities if they yearns for a more equitable groups.

Sources: Arampatzi et al. 2015; Chattopadhyay and Graham 2015; Gallup 2011b; Gallup World Poll, Gallup, Washington, DC, http:// www.gallup.com/services/170945/world-poll.aspx; Graham 2016; Graham and Lora 2009; Hassine 2015; Iqbal and Kiendrebeogo 2014; Khouri and Myers 2015; Ncube and Hausken 2013; Verme et al. 2014; World Bank 2015, 2016a. a. World Bank (2016a). b. Hassine (2015); Ncube, Anyanwu, and Hausken (2013). c. World Bank (2015). d. Iqbal and Kiendrebeogo (2014). e. Arampatzi et al. (2015); Chattopadhyay and Graham (2015); Khouri and Myers (2015). f. Life satisfaction is often measured using the Cantril ladder of life metric (Cantril 1965). g. Graham and Lora (2009) coined the phrase “unhappy growth paradox” as a reﬂection of the declines in life satisfaction during periods of rapid economic growth that also results in drastic upheaval in old social and economic norms, while the new ones that replace them take time to stabilize. This pattern was evident in transition economies of Eastern Europe in the 1990s and, more recently, in China and India. h. Gallup (2011b). i. Gallup (2011b). j. Verme et al. (2014), using data in WVS (World Values Survey) (database), King’s College, Old Aberdeen, United Kingdom, http://www.worldvaluessurvey.org/wvs.jsp. See also Arab Barometer Public Opinion Survey Series (database), Inter-university Consortium for Political and Social Research, Institute for Social Research, University of Michigan, Ann Arbor, MI, http://www .icpsr.umich.edu/icpsrweb/ICPSR/series/508. k. van der Weide, Lakner, and Ianchovichina (2016). l. Arampatzi et al. (2015); Chattopadhyay and Graham (2015). m. Arampatzi et al. (2015) using 2012–14 data in Arab Barometer Public Opinion Survey Series (database), Inter-university Consortium for Political and Social Research, Institute for Social Research, University of Michigan, Ann Arbor, MI, http://www .icpsr.umich.edu/icpsrweb/ICPSR/series/508. experience of Brazil and the Arab Spring Separating fact from myth: indicate (box 4.2). Overall, however, threats what is the evidence on to achieving the World Bank goals arising from extremism, political turmoil, and in- inequality? stitutional fragility are less likely in more Reductions in inequality are important in- cohesive societies. In contrast, political intrinsically and because they are associated stability is more likely to emerge and more with reductions in absolute poverty and difﬁcult to eradicate in societies where greater sharing of prosperity. There are economic growth and social policies have also a lot of misconceptions about recent reduced poverty without addressing inter- changes in inequality. Some narratives sug- personal and regional disparities, whether gest there has been an unrelenting increase objectively measured or subjectively per- in inequality worldwide. Many will be fa- ceived. It is worth recalling that the Arab miliar with the evidence on top incomes Spring began in Tunisia, a country with a rising more quickly than average incomes in long and strong record of economic growth a number of countries.19 Figure 4.2, panel a, and poverty reduction, but also with per- shows that the income share of the richest 1 vasive regional disparities and a history of percent in the United States has been rising cronyism and corruption.18 steeply since the 1970s, after falling in the

INEQUALITY 75 FIGURE 4.2 The Top 1 Percent Income Share, Selected Economies

a. Industralized economies since 1900 b. Developing economies since 1980

20 20

15 15

10 10 Income share of top 1 percent (%) 5 Income share of top 1 percent (%) 5

1900 1925 1950 1975 2000 2015 1980 1990 2000 2010 2013 United States Japan France South Africa Argentina India Korea, Rep. Taiwan, China

Source: Calculations based on data of WID (World Wealth and Income Database), Paris School of Economics, Paris, http://www .parisschoolofeconomics.eu/en/research/the-world-wealth-income-database/. Note: The ﬁgure shows the share of national income (excluding capital gains) going to the richest 1 percent of national populations. These measures are typically derived from tax record data. For South Africa, the ﬁgure shows the top 1 percent income share among adults.

first half of the 20th century. In contrast, in FIGURE 4.3 Global Income Inequality, France and Japan, not only do the richest 1820–2010 control a much smaller share of national income, but their share has also risen much less. This is remarkable because all three 70 countries had similar top income shares at the beginning of the 20th century. The 60 long-run evidence on developing countries is more limited because of the lack of tax record data. Figure 4.2, panel b, shows the up- 50 Gini index ward trends in the income shares of the top 1 percent in selected developing economies since the 1980s. In South Africa, the top in- 40 come share roughly doubled over a period of 20 years and is comparable to the levels 30 observed in the United States. 1820 1870 1920 1970 1990 2010 Globally, there has been a long-term secular rise in interpersonal inequality. Figure Source: Based on figure 1 (p. 27) of The Globalization of Inequality by Francois Bourguignon (Princeton University Press 4.3 shows the global Gini index since 1820, 2015). Used with permission. when relevant data first became available. Note: The discontinuity in the series represents the change in The industrial revolution led to a world- the base year of the purchasing power parity (PPP) exchange rates from 1990 to 2005. The figure uses GDP per capita in wide divergence in incomes across coun- combination with distributional statistics from household tries, as today’s advanced economies began surveys. Figure 4.5 uses income (or consumption) per capita pulling away from others. However, the fig- directly from household surveys, expressed in 2011 PPP exchange rates. ure also shows that, in the late 1980s and early 1990s, the global Gini index began to fall. This coincided with a period of rapid According to household surveys, na- globalization and substantial growth in tional inequality measured by the Gini populous poor countries, such as China index also rose steeply in a number of de- and India. veloping countries. Figure 4.4 shows the

76 POVERTY AND SHARED PROSPERITY 2016 FIGURE 4.4 Long-Run Changes in the (such as wealth) or nonmonetary (such as Gini Index, Selected Developing Countries, health care and education) are ignored as are 1980–2014 the inequalities among subgroups (for example, by sex or ethnicity).23 Income or ex- 55 penditure is measured at the household level (thus ignoring inequality among household 50 members) and assigned to each individual on a per capita basis (thus avoiding the complication of accounting for economies 45 of scale in larger households). The analysis is conducted in two stages. First, estimates are Gini Index 40 reported for interpersonal global inequality. Second, the focus shifts to inequality within 35 countries. Only relative measures of inequality, typically the Gini index, are considered.24 30 This is only one perspective on inequality. Other dimensions are no less important 1980 1990 2000 2010 2014 and may be trending in different directions. Argentina Indonesia China India Thus, people’s perceptions about inequality may be different from actual trends (see box Sources: World Bank 2016b; PovcalNet (online analysis tool), 4.2). People might care about wealth in- World Bank, Washington, DC, http://iresearch.worldbank.org equality, inequalities in access to basic ser- /PovcalNet/. Note: The welfare aggregate in Argentina is income; in all other vices, or the inequality between urban and countries it is consumption. rural areas or across geographical regions. Furthermore, all the standard inequality measures used here are relative, but peo- Gini index of Argentina, China, India, and ple may be concerned with absolute differ- Indonesia, on which longer-run data are ences (box 4.3). While the results here are available. In Argentina and China, inequality compared with alternative datasets, such widened appreciably until the early 2000s, comparisons are difficult because inequal- while the rise in Indonesia began around ity measures may be different, for example, the same time. The increase in inequality in after the application of equivalence scales or India has been more muted and began in the the use of households instead of individu- second half of the 2000s. In contrast, during als as the unit of analysis.25 For instance, in the 2000s, inequality narrowed sharply in Argentina, the use of equivalence scales— Argentina and in some other Latin American adjusting consumption by size and age— countries.20 The drop was so pronounced reduces the Gini index from 41.9 to 39.4, that the inequality levels in Argentina and compared with the per capita estimates.26 China became comparable.21 Inequality has There are several reasons for interpreting been stabilizing in China and, to some ex- these results carefully.27 First, as with any tent, in Indonesia in recent years, though at a global analysis, the data coverage of coun- much higher level than 20 years ago. tries is incomplete. Despite considerable Against the context of these long-run de- progress, good-quality data are still missing velopments, the remainder of this chapter in various countries. These include many analyzes recent changes in inequality more fragile countries, as well as some countries systematically. In this report, inequality in in the Middle East, the African continent, disposable income or consumption expen- the Caribbean, and the Pacific. Annex 4A de- diture is measured among individuals.22 scribes the countries included in the analysis Disposable income is defined as net market and their shares in regional populations. For income (that is, after personal income taxes example, in 2013, the data covered about a and social security contributions have been third of the population in the Middle East deducted), plus any direct social transfers. and North Africa region and around half the Other outcome measures, whether monetary population in Sub-Saharan Africa.

INEQUALITY 77 BOX 4.3 Absolute versus Relative Inequality

Standard measures of inequality, unchanged. This may be associated result, the Gini index fell from 50.2 such as the Gini index, are relative. with quite different absolute to 42.3. In Uganda, consumption Relative measures of inequality gains, however, depending on the among the bottom nine deciles obey the scale invariance axiom, dispersion in incomes in the initial grew at almost the same rate, which says that an inequality distribution. while the consumption of the top measure ought to remain The red lines in ﬁgure B4.3.1 decile grew more slowly. Thus, the unchanged in the face of any show how average incomes or Gini index fell from 45.2 to 42.4. transformation that multiplies all consumption of deciles have grown The blue lines in the two panels incomes by the same constant, in Argentina (panel a) and Uganda show the absolute annual gains by such as a simple rescaling from (panel b) over the past 10 years. decile. In absolute terms, the richer euros to U.S. dollars. This implies In Argentina, incomes have grown deciles gained much more over the that, if all incomes grow at the much more quickly among the period, reversing the conclusions same rate, the Gini index remains poor than among the rich. As a based on relative gains.

FIGURE B4.3.1 Comparing Absolute and Relative Gains across the Distribution a. Argentina (2004–13) b. Uganda (2002–12) 12 600 5 140

10 500 120 4 100 8 400 3 80 6 300 2 60 4 200 40 Annualized growth rate (%)

Annualized growth rate (%) 1 2 100 20 Average annual gain (2011 US$ PPP) Average annual gain (2011 US$ PPP) 0 0 0 0 12345678910 12345678910 Decile Decile Growth Absolute gain

Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. Note: The welfare aggregate in Argentina is income, while consumption expenditure is used in Uganda. According to Beegle et al. (2016), the spell used for Uganda is based on two comparable surveys. The red line is a variant of the growth incidence curve showing annualized growth rates of average income or consumption by decile group. The blue line shows absolute gains per year in 2011 PPP U.S. dollars.

Relative measures of inequality income gaps often carry absolute caring about relative measures and are conceptually appealing connotations. Thus, in experimental caring about absolute measures.a because, for instance, they allow studies, university students An analysis of absolute differences inequality and economic growth to in Germany, Israel, the United can, in any case, provide a be analyzed separately. However, Kingdom, and the United States are complementary perspective.b perceptions about widening approximately evenly split between

a. Ravallion (2016). b. For instance, Atkinson and Brandolini (2010) argue that global inequality analyses in particular need to consider both absolute and relative differences.

Second, for the sake of consistency, the shared prosperity, respectively. This means analysis relies on the same welfare aggre- that the analysis tends to use consumption gate to measure inequality that is used in expenditure for most developing countries chapters 2 and 3 to estimate poverty and and income for the industrialized countries

78 POVERTY AND SHARED PROSPERITY 2016 BOX 4.4 Comparison of Levels and Trends in Income and Consumption Inequality

In a number of countries in income-based Gini indexes. But shared prosperity spells introduced Eastern Europe and Central Asia, rankings remain somewhat similar. in chapter 3 (from around 2008 to both income- and consumption- For example, Georgia and Turkey 2013), the ﬁgure plots the change based Gini indexes are available for are the two most unequal countries in the Gini index for the two the same years. Figure B4.4.1 plots in the sample, but Poland moves up welfare aggregates. While there are the income Gini (left axis) against from rank 8 to rank 3 if consumption some large differences (notably, the consumption Gini (right axis) is used instead of income. Romania), both welfare aggregates for all the countries where such a Figure B4.4.2 addresses the point in the same direction, and comparison is possible for 2013. It issue of whether inequality trends one aggregate does not produce a is clear that consumption-based Gini are different if consumption is used consistently smaller change than indexes are considerably lower than instead of income. For the set of the other.

FIGURE B4.4.1 Levels of Income and Consumption, FIGURE B4.4.2 Trends in Income and Gini Indexes, 2013 Consumption, Gini Indexes, Circa 2008–13

45 45 Albania Georgia Turkey Armenia Armenia Kyrgyz Republic Belarus Turkey 40 Georgia Moldova Kazakhstan Romania 38 Serbia Kyrgyz Republic

Poland Moldova 34 34 Romania Poland Serbia Armenia Gini index of income 30 30 Turkey Kazakhstan Serbia Kyrgyz Republic Moldova 2 Romania −6 −4 −2 0 Belarus Gini index of consumption expenditure Change in Gini index 26 Kazakhstan Ukraine 25 Change in income−based Gini Belarus Ukraine Change in consumption−based Gini

Source: Calculations based on data from the ECAPOV database harmonization as of April 2016, Europe and Central Asia Team for Statistical Development, World Bank, Washington, DC. Note: In Poland and Romania, the World Bank uses income surveys for poverty monitoring. All other countries in the two figures use consumption expenditure.

and for Latin America.28 Given that the Europe and Central Asia, where it is possi- marginal propensity to consume declines ble to measure Gini indexes for both welfare with income, or that savings increase with aggregates in the same year. The ranking of income, consumption expenditure tends to countries by the Gini index is somewhat be more equally distributed than income. robust to whether consumption or income Compared with an analysis using income is used, although there are some notable distributions exclusively, the data presented exceptions. However, this only considers here would therefore systematically under- the ranking within a region, not across the state inequality in some countries. While world, and it is not obvious that this rela- this is a serious issue, an income distribu- tionship would also hold in other regions. tion cannot be inferred reliably from an ex- It is important to keep this in mind when penditure distribution.29 comparing inequality levels between, for Box 4.4 assesses the effects of using in- example, Latin America and the Caribbean come or consumption surveys in Eastern (mostly income surveys) and Sub-Saharan

INEQUALITY 79 Africa (mostly consumption surveys). Box at the top).30 Finally, consumption surveys 4.4 also shows that time trends are similar tend to understate true living standards at between the two welfare aggregates. the top because consumption declines with A third reason for caution is that the cov- income or because expenditure on durables erage of household surveys within countries (which are more important at the top) is is also typically incomplete at the top tail. poorly measured.31 While this is not an issue in measuring pov- Thus, it is likely that the household sur- erty, it is important in measuring inequality veys used in this chapter understate the accurately (see chapter 1), as well as in other level of inequality. While the evidence from areas of public policy such as ﬁscal policy. administrative records remains seriously Survey enumerators typically face difﬁcul- limited in developing countries, comple- ties interviewing the richest households, mentary evidence suggests that top in- and, if they do conduct interviews, the rich comes might have been rising more quickly. may understate their incomes. Furthermore, Household surveys may therefore also un- the coverage of the income surveys that are derestimate the trend in inequality. For ex- used for some emerging economies is in- ample, the labor share has been declining in complete in terms of entrepreneurial and many countries, while billionaire wealth on capital incomes (important income sources rich lists has been growing rapidly.32 Box 4.5

BOX 4.5 Comparing Trends in Inequality: Household Surveys and Administrative Records

Administrative data on top incomes, FIGURE B4.5.1 Comparison of Top typically based on tax records, have Incomes and the Gini Index, Brazil, become available for a sizable number of 2006–12 rich countries.a In developing countries, the availability of these data is more 28 58 limited, and, where they are available, data quality may be problematic given the absence of broad income taxes 26 56 in many developing countries and the incomplete taxation of capital incomes. 24 54 Figure B4.5.1 compares the income

share of the top 1 percent and the Gini Gini index index of Brazil between 2006 and 2012. 22 52 While the Gini index has been falling steadily over this period, by almost 4 Income share of top 1 percent (%) Gini points, the top income share has 20 50 been on an upward trend. Part of this 2006 2007 2008 2009 2010 2011 2012 divergence may derive from deﬁnitional Top 1 percent differences, such as the use of tax units Gini index (right axis) versus households or of gross incomes versus disposable incomes. However, Source: World Bank compilation based on Souza, Medeiros, this evidence may also suggest that the and Castro 2015; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch Gini index misses important changes in .worldbank.org/PovcalNet/. the top tail; particularly that top incomes Note: The red line shows the trend in the income share have grown more quickly than mean of the top 1 percent of individuals from tax record data. incomes. The two data series rely on different units of analysis (tax units versus households) and different welfare aggregates (disposable versus taxable gross income).

a. Thanks to the pioneering work of Alvaredo et al. (2013). See WID (World Wealth and Income Database), Paris School of Economics, Paris, http://www.parisschoolofeconomics.eu/en/research /the-world-wealth-income-database/.

80 POVERTY AND SHARED PROSPERITY 2016 compares the trends in household survey– FIGURE 4.5 Global Inequality, 1988–2013 80 based measures of inequality with adminis- 1.0 trative data on top incomes in Brazil, where such a comparison is possible in recent 70 years. In other developing countries, such as 0.8 Argentina or South Africa, the movement in 33 the two measures has been rather similar. 60 0.6 80 76 74 72 70 65

Global inequality Gini index 0.4 50

However imperfect, this compilation of Mean log deviation household surveys remains the only source of distributional statistics on a large set of 0.2 40 countries. Global inequality, defined here as 20 24 26 28 30 35 the inequality in income among all persons 0 30 in the world irrespective of their country of 1988 1993 1998 2003 2008 2013 residence, is an aspect of inequality that is Within-country inequality Gini index (right axis) often overlooked in discussions that focus Between-country inequality on country-specific inequality.34 During a Sources: Lakner and Milanovic´ 2016a; Milanovic´ 2016; calculations based on PovcalNet (online analysis period of rapidly increasing global integra- tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. tion, some of the poorest economies were Note: For each country, household income or consumption per capita is obtained from household surveys growing rapidly, thus raising average living and expressed in 2011 PPP exchange rates. Each country distribution is represented by 10 decile groups. The line (measured on the right axis) shows the level of the global Gini index. The height of the standards, but many of the same coun- bars indicates the level of global inequality as measured by GE(0) (the mean log deviation). The red tries also experienced increasing inequality bars show the corresponding level of population-weighted inequality within countries. The level of within their borders. Global inequality cap- between-country inequality, which captures differences in average income across countries, is shown by the yellow bars. The numbers in the bars refer to the relative contributions (in percent) of these two tures the overall effect of both forces. sources to total global inequality. Global inequality has diminished for the first time since the industrial revolution. The global Gini index rose steadily by around 15 is to decompose global inequality into dif- Gini points between the 1820s and the early ferences within and between countries. This 1990s, but has declined since then (see fig- allows an understanding of how much of ure 4.3).35 While the various methodologies the change in global inequality is explained and inequality measures show disagreement by countries reducing the inequality among over the precise timing and magnitude of them (that is, reducing the differences in the decline, the decline since the middle of average incomes across countries) relative the last decade is confirmed across multiple to the reduction in inequality within each sources and appears robust.36 The estimates one of the countries (that is, the differences presented in figure 4.5 show a narrowing in in incomes within a country). This decom- global inequality between 1988 and 2013. position is shown by the bars in figure 4.5. The Gini index of the global distribution The total height of the bars captures global (represented by the blue line) fell from 69.7 inequality as measured by GE(0) (the mean in 1988 to 62.5 in 2013, most markedly since log deviation), while the red and yellow bars 2008 (when the global Gini index was 66.8). show the within and between contribu- Additional exercises confirm that these re- tions, respectively. In contrast to the Gini, sults are reasonably robust, despite the er- GE(0) is a bottom-sensitive inequality mea- rors to which the data are typically subject.37 sure that can be additively decomposed into Global inequality is at a much higher within- and between-country components. level than inequality within countries. Few This decomposability feature makes this in- countries have Gini indexes above 60 (see dicator appealing for this type of analysis. below). This is not surprising given that Between two-thirds and four-fifths of the global distribution includes everyone global inequality stems from differences in from the poorest Congolese to the richest average incomes across countries (between- Norwegian. Another way to illustrate this country inequality). The reduction in over-

INEQUALITY 81 all global inequality was mostly driven by pened to average within-country inequality a decline in this component, that is, aver- in the world. Whatever the approach, average incomes converged across countries.38 age inequality within countries increased This reflects the rapid growth in aver- in the 1990s, but has been falling in recent age incomes in populous countries such years according to most specifications. as China and India. These developments The population-weighted average Gini were counteracted to some extent by an (solid blue line in figure 4.6) captures the increase in within-country inequality, espe- change in within-country inequality for the cially in the 1990s. Between 2008 and 2013, average person in the world. Therefore, the within-country inequality stabilized or even individual remains the unit of analysis as in declined slightly, which, together with the the global inequality analysis (see above).42 strong convergence effect, led to a marked The population-weighted average Gini rose decline in global inequality. As a result of sharply between 1988 and 1998, by some these developments, within-country differ- 6 points, from 34 to 40. Over this period, ences explain an increasing share of global the average person in the world was thus inequality, as shown in figure 4.5. living in a country in which inequality was A mixed picture emerges if one exam- growing steeply. Since then, inequality has ines the separate regional distributions. declined slightly, by almost one point, but Inequality increased within most regions, remains at a higher level than 25 years ago. with the notable exception of Latin America Next, the analysis presents unweighted and the Caribbean. For instance, inequal- averages. These averages allow us to cap- ity among all Sub-Saharan Africans rose ture how different countries have fared over between 1993 and 2008, driven by widen- time in terms of reducing inequality, thus ing between-country inequality, although learning from their successes (see chapter within-country differences remain the 5 for a discussion of a few selected country dominant source (around 60 percent).39 In cases). Thus, this part of the analysis treats contrast, growing regional inequality within China and Honduras the same in calculat- East Asia and Pacific was driven by rising ing the average, regardless of the fact that within-country inequality.40 This also im- the population of China is much larger plies that most of the convergence observed than the population of Honduras. As shown at the global level has been between regions. by the solid red line in figure 4.6, the unweighted average Gini index also increased Within-country inequality during the 1990s, but by a smaller amount. The simple average worldwide increased by Most studies of inequality focus on within- around 5 points, from 36 in 1988 to 41 ten country inequality, which remains the years later, and declined thereafter, reaching level at which most policies operate. Our 38 in 2013.43 analysis of within-country inequality, the Because not every country conducts most comprehensive among recent stud- a household survey every five years, the ies, covers all available countries regardless country composition changes across these of region or income level, as compiled in five-year periods. To avoid such composi- PovcalNet.41 Described in more detail in the tional shifts, inequality trends are studied annex to this chapter, the analysis defines for a smaller sample of countries that have six benchmark years (in five-year intervals household surveys in each of the five-year between 1988 and 2013) to which country- periods considered. The dashed lines in fig- year observations within a two-year win- ure 4.6 repeat the unweighted and weighted dow on either side are grouped. This al- averages, but include the same set of 41 lows a more meaningful comparison across countries throughout the period.44 The countries over time, because many coun- average inequality in this set of countries tries lack data for every year. On average, follows a similar trend. Taken together, the there are 109 countries per benchmark year analysis suggests that, in the average coun- (table 4A.2 in the annex). Figure 4.6 offers try, inequality may have peaked in the late four ways of summarizing what has hap- 1990s and early half of the 2000s and has

82 POVERTY AND SHARED PROSPERITY 2016 declined in the latter half of the 2000s. In- FIGURE 4.6 Average Within-Country Inequality, 1988–2013 equality in this smaller sample is also wider 42 in 2013 than the levels 25 years previously. The levels and trends in average inequality are quite different across regions, 40 although the most recent decline is broadbased. Figure 4.7 shows the unweighted 38 average Gini index across seven regions. Within-country inequality tends to be 36 higher in developing countries than in developed countries, the latter grouped in this analysis as industrialized countries (a subset 34 of high-income countries).45 The highest levels of inequality are observed in Latin Average within-country Gini index 32 America and the Caribbean. Contributing to the intrinsically high levels of inequality, within-country inequality in Latin Amer- 30 1988 1993 1998 2003 2008 2013 ica and the Caribbean is measured using income-based surveys that are expected to Unweighted Weighted Unweighted, balanced Weighted, balanced show higher levels of inequality. Latin America and the Caribbean stands Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World out as a region that has been successful in Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/; WDI (World Development Indicators) (database), World Bank, Washington, DC, http://data.worldbank.org/data-catalog/world-development narrowing inequality in the last 10 to 15 -indicators (see annex 4A). years, also driving the decline in the global Note: The solid lines show the trend in the average within-country Gini index with and without population 46 weights in the full sample (an average 109 countries per benchmark year). The dashed lines refer to the average. However, these declines occurred balanced sample, that is, using only the set of 41 countries on which data are available in every bench- after a prolonged increase during the 1980s mark year. (not shown) and 1990s, such that, by 2012, the average Gini in the region had returned FIGURE 4.7 Trends in the Average Gini, by Region, 1988–2013 to the level of the early 1980s.47 Hence, the long-run progress in the reduction of in- 60 equality in Latin America and the Carib- bean has been limited. Furthermore, the 50 downward trend has slowed, and inequality recently stagnated.48 The average Gini in Sub-Saharan Africa 40 has declined steadily since the early 1990s, but continues to be the second highest after 49 the Gini in Latin America. In Eastern Eu- 30 rope and Central Asia, average inequality rose sharply after the fall of the Berlin Wall,

but has since been on a declining trend Average within-country Gini index 20 (see ﬁgure 4.7).50 Similarly, inequality rose sharply during the transition to a market economy in some East Asian countries. The 10 average industrialized country experienced 1988 1993 1998 2003 2008 2013 an increase in the Gini index from 30 in 1988 East Asia and Paciﬁc Eastern Europe and Central Asia to 33 in 2008. From 2008 to 2013, average in- Latin America and the Caribbean Middle East and North Africa equality appears to have fallen in all regions South Asia Sub-Saharan Africa Industrialized countries except South Asia and the Middle East and North Africa, where data are limited.51 Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World Providing a simple explanation for Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/ (see annex 4A). Note: The lines show the average within-country Gini index by region. It is the simple average in the full these regional inequality trends is particu- sample without weighting countries by population. Industrialized countries is a subset of high-income larly challenging because countries within countries. See chapter 2, annex 2B, for the list of industrialized countries.

INEQUALITY 83 FIGURE 4.8 The Gini Index, 101 Countries, 2013 a region may exhibit distinctive trends and specific drivers behind the trends. In- Ukraine − Slovenia + Consumption survey Norway − stead of providing a simplistic explana- Slovak Republic + Income survey Czech Republic . tion, it is more useful to look closely at the Kazakhstan − Average Gini for all countries Belarus . country variations within regions and to Kosovo Iceland − understand how the common drivers of Finland − Sweden . inequality––such as gaps in human capi- Belgium − Netherlands − tal accumulation, differences in access to Moldova − Kyrgyz Republic − jobs and income-generating opportunities, Albania − Serbia . and government interventions to address Denmark + Iraq . Germany − market-based inequalities such as taxes and Tajikistan Austria . transfers––are relevant in each country. The Hungary + Pakistan − remainder of this chapter looks at the vari- Cambodia − Armenia + ations in within-country inequality in each Switzerland − Mongolia − region, while chapter 5 focuses on selected Montenegro . Mauritania − countries that have successfully reduced in- Croatia . Ireland . equality, and chapter 6 centers on specific United Kingdom − Poland − interventions that have been shown to re- France . Estonia + duce inequality and poverty without major Ethiopia Guinea Bosnia and Herzegovina efficiency and equity trade-offs in several Niger Sierra Leone countries around the world. Cyprus + Luxembourg + The bars in figure 4.8 show the level of Romania − Lithuania . the Gini index across the 101 countries Italy + India on which data are available for 2013. The Burkina Faso Latvia − figure highlights whether a country uses Mauritius . Spain + income or consumption expenditure, con- Bulgaria . Portugal . firming that most of the high-inequality Greece + Iran, Islamic Rep. − Vietnam − countries use income surveys. The most Tanzania − Thailand − unequal country in the world is South Af- Lao PDR + Bhutan . rica, followed by Haiti, each of which has Sri Lanka − Indonesia a Gini index in excess of 60. Another Sub- Georgia − Turkey + Saharan African country (Rwanda) and Senegal + Uganda − seven other Latin America and Caribbean United States . Russian Federation . countries (Brazil, Chile, Colombia, Costa Uruguay − Congo, Dem. Rep. Rica, Honduras, Mexico, and Panama) China . Argentina − Micronesia, Fed. Sts. make up the top 10 most unequal countries Madagascar Philippines − in the world. All the most equal countries Zimbabwe Chad are in the group of industrialized countries Benin El Salvador − or in Eastern Europe and Central Asia. Djibouti Peru − More broadly, all Latin America and Togo + Cameroon + Caribbean countries have Gini indexes in Seychelles Nicaragua + excess of 40, and the Gini in a third of those Dominican Republic − Ecuador − countries is above 50 (figure 4.9). There is Bolivia − Paraguay + almost no overlap between Latin Amer- Guatemala − Congo, Rep. Mexico + ica and the Caribbean and the two other Costa Rica . Rwanda . regions that primarily use income sur- Chile − Panama − veys, namely, the industrialized countries Brazil − Colombia − and Eastern Europe and Central Asia. The Honduras . Haiti other region that exhibits high inequality South Africa − is Sub-Saharan Africa, especially the south- 52 20 30 40 50 60 ern countries. More than half the African Gini index countries in the sample have Gini indexes in Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World excess of 40. The reported measures of in- Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/ (see annex 4A). Note: Countries are sorted by the Gini index. The red line shows the unweighted average Gini index in 2013. “+” = increase in the Gini > 1 Gini point, 2008–13. “–” = decline in the Gini > 1 Gini point. “.” = change in the Gini within 1 Gini point. See the detailed discussion in the text around table 4.1. equality in Sub-Saharan Africa use predom- FIGURE 4.9 Distribution of the Gini Index, 2013 inantly consumption expenditure; income- based inequality statistics would likely be East Asia and Pacific 9 surveys higher. Eastern Europe and Central Asia 26 surveys Short- and long-run trends in within-country inequality Latin America and the Caribbean 18 surveys

Results based on an analysis of trends are Middle East and North Africa 3 surveys often sensitive to the beginning and end points. The trends chosen in this chapter South Asia 4 surveys begin around 1993 given the limited data availability in the developing world (espe- Sub-Saharan Africa 21 surveys cially Africa) before that year. This choice might underestimate the rise in inequal- Industrialized countries 20 surveys ity. For example, the surge in inequality in many Eastern European countries follow- World 101 surveys ing the fall of the Berlin Wall had already occurred. During the 1980s, inequality rose 0 20406080 100 steeply in some rich countries, such as the Percent Netherlands, the United Kingdom, and the Gini below 30 Gini between 40 and 50 53 United States. For these reasons, two sets Gini between 30 and 40 Gini above 50 of country-level spells are analyzed (as discussed in annex 4A). First, the long-run Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/ (see annex 4A). spells include all countries that have an ob- Note: The figure shows that 2013 data were available on 101 countries. Of these, 8 had a Gini index servation around 1993 and 2008, as long as exceeding 50; 6 were in Latin America and the Caribbean, and 2 in Sub-Saharan Africa. Some 31 had a the welfare aggregate (income or consump- Gini index between 40 and 50: 12 in Latin America and the Caribbean, 11 in Sub-Saharan Africa, 3 each in East Asia and Pacific and Eastern Europe and Central Asia, and 1 each in the Middle East and North tion) is the same. Second, the sample of Africa and in the industrialized countries. Some 43 countries had a Gini index between 30 and 40: 12 each short-run trends from 2008 to 2013 is based in Eastern Europe and Central Asia and in the industrialized countries, 8 in Sub-Saharan Africa, 6 in East on the strictly comparable shared prosperity Asia and Pacific, 4 in South Asia, and 1 in the Middle East and North Africa. Some 19 countries had a Gini index below 30: 11 in Eastern Europe and Central Asia, 7 in the industrialized countries, and 1 in the spells used in chapter 3. Comparisons need Middle East and North Africa. to be drawn carefully because the regional composition of the two sets of spells is different. Thus, the set of short-run spells in- smaller) sample of countries. The majority cludes fewer countries, especially in South of countries appear to fall below the line, Asia and Sub-Saharan Africa. This reflects that is, they show a declining Gini index, and issues of data comparability and availability, the decline in Latin America and the Ca- which are discussed in detail in a recent re- ribbean appears even more pronounced. A port on Sub-Saharan Africa.54 few country examples are highlighted in Figure 4.10 plots the final year against figure 4.10, panel b, indicating that decreases initial year Gini indexes for the long-run in inequality have been observed across all (panel a) and short-run (panel b) spells. levels of inequality (high and low), income Countries above the line experienced in- groups (low- and middle-income catego- creasing inequality, whereas countries below ries), and regions. the line saw a decline. Between 1993 and Table 4.1 summarizes within-country 2008, large falls are observed. This might trends in inequality by region more system- indicate successful reductions in inequality, atically. The sample in the 15 years between but also some problems with survey com- 1993 and 2008 includes 91 countries, 42 of parability, especially in Sub-Saharan Africa. which showed increasing inequality; 39 had The other major region with declining in- a declining Gini index; and 10 showed no equality appears to have been Latin America significant changes, that is, changes below 1 and the Caribbean. The changes between Gini point in either direction.55 Hence, the 2008 and 2013 were smaller for the (also number of countries with rising inequality

INEQUALITY 85 FIGURE 4.10 Trends in the Within-Country Gini Index, 1993–2013

a. 1993–2008 b. 2008–13

70 70

ZAF ZAF 60 60 COL BRA PAN LSO PRY GNB BOL 50 50 BRA URY ECU BOL NIC TGO ECU PER LKA RUS 40 TZA 40 VNM LVA GRC VNM MRT TZA Gini index (around 2013) Gini index (around 2008) KHM JOR MLI DEU KGZ PAK KHM 30 DNK 30 IRQ DEU SVK UKR MDA SVN SVN UKR 20 20

20 30 40 50 60 70 20 30 40 50 60 70 Gini index (around 1993) Gini index (around 2008)

East Asia and Paciﬁc Latin America and the Caribbean South Asia Industrialized countries Eastern Europe and Central Asia Middle East and North Africa Sub-Saharan Africa

Source: Calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/ (see annex 4A). Note: Economies along the red (45-degree) line experienced no change in inequality. In economies below (above) the line, inequality narrowed (widened).

TABLE 4.1 Countries with an Increasing or Decreasing Gini Index and the Average Gini Number

Long-run trend (1993–2008) Short-run trend (2008–13) Number of countries: Mean Gini Number of countries: Mean Gini ↑ +/– pp ↓ Total 1993 2008 ↑ +/– pp ↓ Total 2008 2013 East Asia and Paciﬁc 5 1 3 9 37.8 39.1 1 1 5 7 39.2 37.3 Eastern Europe and Central Asia 5 2 6 13 33.9 32.5 6 8 9 23 31.9 31.4 Latin America and the Caribbean 8 0 11 19 49.0 47.0 3 2 12 17 49.7 48.0 Middle East and North Africa 1 1 3 5 39.8 36.4 0 1 1 2 35.3 33.4 South Asia 3 0 1 4 31.0 34.5 0 1 2 3 36.7 36.2 Sub-Saharan Africa 8 2 10 20 47.6 45.1 3 2 4 9 44.1 43.8 Industrialized countries 12 4 5 21 31.4 32.6 6 6 8 20 32.0 31.8 World 42 10 39 91 40.1 39.3 19 21 41 81 37.9 37.1 Source: World Bank calculations based on data in Milanovic´ 2014; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/ (see annex 4A). Note: Increases and decreases refer to changes that are greater than 1 Gini point in absolute value. The unweighted average Gini index is estimated over the sample of 91 countries (left panel) and 81 countries (right panel).

was slightly higher than the number of coun- (thus, barely signiﬁcant). This is because, tries with falling inequality. Yet, the average among countries with narrowing inequality, Gini among these 91 countries actually de- the Gini index fell by a larger amount than clined by 0.8 Gini points, from 40.1 to 39.3 it increased in the rising countries. However,

86 POVERTY AND SHARED PROSPERITY 2016 inequality widened more rapidly in larger Concluding remarks countries. The populous countries where the Gini index increased sharply include This chapter has presented empirical evi- Bangladesh (5 points), China (7 points), and dence on trends in income inequality using Indonesia (5 points).56 the latest available data from household During the long-term spell, 1993–2008, surveys. The number of countries and the Latin America and Caribbean region household surveys analyzed makes this stands out because of falling Gini indexes: exercise the most current and complete 58 percent of the countries in the region assessment of within-country inequality showed a decline, and the average Gini worldwide. Yet, the analysis has limita- declined by 2 points (3.5 points with pop- tions, namely, the changing and incomplete ulation weights). Meanwhile, many East country composition of the sample and Asia and Pacific countries (56 percent of the likely underreporting of top incomes the countries) and South Asian countries in household surveys. These caveats aside, (75 percent) in the sample saw inequality there was a steady reduction in global in- rising. The average Gini in East Asia and equality, defined as inequality among all Pacific and in South Asia rose by 1.3 points individuals in the world, between 1988 and (5.7 with weights) and 3.5 points (2.8 with 2013. This was driven by the strong con- weights), respectively. In the industrialized vergence in average incomes across coun- countries, the average Gini increased, and tries, that is, by reductions in the inequality inequality increased in over half the coun- between countries. This fall in between- tries. The Sub-Saharan African countries country inequality coincided with a period are roughly evenly split into increasing and of rapid globalization and high growth, no- decreasing inequality, with a decline on tably in some populous developing coun- average.57 The conclusions for the Middle tries, such as China and India. Furthermore, East and North Africa region need to be in- the most marked reduction in global interpreted carefully because of limited data equality occurred between 2008 and 2013, availability in the region. straddling the period of the global financial The evidence suggests that there was a crisis and the subsequent slowdown in the shift toward declining inequality between global economy. 2008 and 2013. However, the interpreta- This reduction in global inequality con- tion of the analysis needs to be cautious stitutes a recent shift in an otherwise long- because the global financial crisis occurred term secular rise in interpersonal inequality during this period. The number of coun- that began during the industrial revolution tries with decreasing inequality in this more in the 1820s, which is as far back as the data recent spell is more than double the num- allow, and that lasted through the 1980s. ber of the countries with rising inequality. Furthermore, during the recent exceptional The Gini index fell by more than 1 point in period of global inequality reduction, the 41 of 81 countries (51 percent of the sam- share of the incomes of the top 1 percent ple of countries). On average, the Gini fell has increased in a number of countries on marginally by 0.8 points during this period. which information is available. This is true Populous countries with falling Gini in- in developed and emerging economies as dexes include Brazil (−2.4 points), Pakistan long-term evidence from Argentina, India, (−1.2), and Vietnam (−5.1). This develop- the Republic of Korea, South Africa, and the ment toward narrowing inequality can be United States confirms. Clearly, there are observed across all regions. Hence, relative no grounds for complacency. In fact, even to the earlier period, the change has been after the recent decline, global inequality is most pronounced in East Asia and Pacific wider than the within-country inequality and in South Asia, where inequality had observed in most countries in the world. increased steeply between 1993 and 2008. The average country is more unequal Latin America and the Caribbean made the today than 25 years ago. Within-country biggest contribution to declining inequality, inequality as a whole—that is, consider- accounting for almost half the total decline. ing together the trends of all countries on

INEQUALITY 87 which evidence is available—only started to ects a larger decline in the incomes of the narrow in the last decade, after peaking in bottom 40 than the expected decline in the the 1990s. Inequality remains unacceptably average incomes of the entire population, high in many countries around the world. a consistent result across several scenarios Developing countries tend to exhibit higher of climate change projections, economic levels of inequality than developed coun- growth, and adoption of adaptive policy tries. Latin America and the Caribbean, interventions.58 This is because this group along with Sub-Saharan Africa, stand out as has lower-quality assets, less access to pro- historically high-inequality regions. But it is tection mechanisms, and is more vulnerable precisely the Latin American region that has to the negative effects of climate change on been more successful in reducing inequality agricultural productivity, weather shocks, than any other region. food prices, and diseases. Finally, signs of Another critical finding of the country a backlash against the free movement of analysis is that, for every country in which goods, people, and ideas have emerged in inequality widened by more than 1 Gini a number of countries. Even though it is point (19 of 81 countries) in recent years, entirely speculative, such a backlash could it narrowed in more than two countries by hurt the poor if it were to reverse the in- over 1 point (41 out of 81 countries). More equality reductions observed since 2008 at than a third of the population in the sample a global level. While globalization has cer- were living in a country in which the Gini tainly produced both winners and losers, it had fallen by more than 1 point. is highly unlikely that undoing it will lead While this is good news, overoptimism is to sustaining or increasing the pace of the out of place. There should be no presump- recent reduction in global inequality into tion that these favorable and exceptional the long run. recent trends will continue. The growth Beyond global processes, a country’s in- slowdown in developing and emerging come inequality is also a choice. Domestic economies, if it becomes protracted, is policy choices explain, to a large extent, re- likely to delay further reductions in global cent within-country inequality reductions, inequality. Closely related is the secular often combining solid macroeconomic decline in commodity prices. (The earlier management with coherent sectoral policies. boom in commodity prices had funded (Chapter 5 examines the policy choices that many equity-enhancing public investments have led to a reduction in inequality and an within countries.) Climate change is pro- expansion in shared prosperity in selected jected to have significant negative distribu- countries. Chapter 6 assesses specific policy tional as well as poverty effects both globally interventions that have had a demonstrated and within countries. A recent analysis proj- effect in reducing inequalities.)

88 POVERTY AND SHARED PROSPERITY 2016 Annex 4A

Data construction

Global inequality database The regions with the lowest coverage are the Middle East and North Africa, Sub-Saharan The estimates of global interpersonal in- Africa, and Other Asia. This is relevant be- equality (see figure 4.5) are primarily based cause changes in the sample composition on Lakner and Milanovic´ (2016a), who could have important effects on the level of cover the period 1988 to 2008. Here, their global inequality. However, the decline in data have been converted into 2011 pur- global inequality between 2008 and 2013 is chasing power parity (PPP) exchange rates, robust to a number of modifications, such and their estimates have been updated to as using a balanced sample of countries 2013 using data from PovcalNet and Mi- throughout. lanovic´ (2016). The data update is constructed in exactly the same way as the Lak- Database of within-country ner and Milanovic´ (2016a) dataset: surveys need to be within two years of a benchmark Gini indexes year; consecutive surveys need to be at least The primary source of the country-level three and no more than seven years apart; Gini data is the October 2016 release of and the welfare aggregate (income or con- PovcalNet.60 For most countries, PovcalNet sumption expenditure) must remain the computes these statistics directly from mi- same for a country over time.59 Table 4A.1 crodata.61 To increase the geographic cov- shows the share of the global or regional erage, data from the All the Ginis database population that is covered by the surveys (Milanovic´ 2014) are also used.62 Because included in the database. The regional defi- not every country holds a household survey nitions in table 4A.1 follow Lakner and Mi- annually, observations are grouped into six lanovic´ (2016a) and are different from the benchmark years from 1988 to 2013 in five- regions used in this chapter. In 2013, 80 per- year intervals. If data for a benchmark year cent of the global population was covered. are not available, the nearest year within a

TABLE 4A.1 Population Coverage of the Data Used in the Global Inequality Estimates

Benchmark year Region 1988 1993 1998 2003 2008 2013 Mean World 81 92 92 94 94 80 89 (Number of surveys) (74) (114) (119) (137) (139) (103) (114)

Mature economies 95 99 99 96 98 86 96 China 100 100 100 100 100 100 100 India 100 100 100 100 100 100 100 Other Asia 75 86 89 89 90 69 83 Middle East and North Africa 61 69 64 68 70 19 59 Sub-Saharan Africa 28 73 68 80 86 40 62 Latin America and the Caribbean 89 94 96 97 95 97 95 Russian Federation, Central Asia, Southeastern Europe 22 80 86 100 90 88 78

Sources: Lakner and Milanovic´ 2016a; Milanovic´ 2016; World Bank calculations based on PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch .worldbank.org/PovcalNet/ Note: The cells in the table show the share of the global or regional population (in percent) accounted for by the surveys included in the database. The last column is the simple average over the benchmark years from 1988 to 2013. The regions are deﬁned in Lakner and Milanovic´ (2016a).

INEQUALITY 89 TABLE 4A.2 Population Coverage of the Data Used in the Analysis Share of regional population covered by data (%)

Eastern Latin America Middle East East Asia Europe and and the and North South Sub-Saharan Industrialized World and Paciﬁc Central Asia Caribbean Africa Asia Africa countries A. Full sample 1988 (73 countries) 79 90 93 91 42 96 10 75 1993 (102 countries) 88 95 87 93 76 97 68 77 1998 (106 countries) 71 95 82 95 70 22 71 75 2003 (135 countries) 91 95 99 94 77 98 77 78 2008 (136 countries) 92 96 93 95 72 98 70 95 2013 (101 countries) 80 94 90 92 32 87 52 69 B. Sub-samples Balanceda (41 countries) 46 88 14 79 22 11 4 66 Long-run trendsa (91 countries) 84 95 83 87 62 97 53 76 Short-run trendsa (81 countries) 54 81 87 90 32 12 21 69

Sources: WDI (World Development Indicators) (database), World Bank, Washington, DC, http://data.worldbank.org/data-catalog/world-development-indicators; World Economic Outlook Database, International Monetary Fund, Washington, DC, https://www.imf.org/external/pubs/ft/weo/2016/01/weodata/index.aspx. Note: The cells in the table show the share of the global or regional population (in percent) accounted for by the surveys included in the database. For the balanced and trend samples, the population coverage refers to the ﬁnal year. a. For the balanced and trend samples, the population coverage refers to the ﬁnal year.

two-year window around the benchmark 4A.2. These samples are different from those year is chosen.63 These estimates do not described above because, if a country is to be extrapolate or interpolate, and they mix included, it needs to be observed in the ini- income and consumption surveys, as dis- tial and final year of the period considered.66 cussed in the main text. Furthermore, only countries that have the Table 4A.2 shows the population cov- same welfare measure (income or consump- erage of the resulting database by region.64 tion) in both these years are included.67 The The regional definition used throughout long-run trend sample is thus a subset of this chapter follows the World Bank’s geo- the benchmark-year sample. The short-run graphical classification, except for the in- sample is based on the World Bank Global dustrialized countries category, which is Database of Shared Prosperity (see chapter a subcategory of the World Bank’s high- 3).68 This database includes countries with income countries group.65 surveys around 2008 and 2013 in which In the main sample, as many countries as welfare aggregates satisfy a high standard of possible are used in each benchmark year. comparability.69 The subset of 75 countries This implies, however, that the country for which the initial and final years fit the sample may change from year to year, which benchmark years 2008 and 2013 is then se- could lead to spurious changes arising from lected.70 In particular, the initial year has to such sample attrition. To abstract from be between 2006 and 2010 and the final year such changes, results are also presented for between 2011 and 2015. Where possible, ad- the sample of countries that are present in ditional countries from last year’s version every benchmark year between 1988 and of the shared prosperity database (covering 2013 (see the balanced sample row in table 2007–12 on average) are included, as long as 4A.2). Countries included in this sample they fit these requirements. This produces also need to have the same welfare metric the sample of 81 countries’ short-run spells. (income or consumption) in all years. There exists a large number of alternative Figure 4.10 in the main text compares sources for data on Gini indexes (or other changes in inequality over two periods: distributional statistics). As reviewed in de- 1993–2008 (long run) and 2008–13 (short tail in a recent special issue of the Journal run). The population coverage of these sam- of Economic Inequality (Ferreira, Lustig, and ples is shown in the last two rows of table Teles 2015), these databases cover distribu-

90 POVERTY AND SHARED PROSPERITY 2016 tional statistics that are calculated directly forts, or talents of the individual. (Chapter from microdata, compiled from secondary 6 discusses interventions that have been suc- data, or imputed. PovcalNet is the only da- cessful in equalizing opportunities.) tabase that uses (almost exclusively) mi- 4. According to World Development Report 2006: crodata and that covers all countries in the Equity and Development, while “outcomes world. Other microdata-based sources have matter, we are concerned with them mainly a more limited geographical coverage. For for their influence on absolute deprivation and instance, Eurostat, the Luxembourg Income their role in shaping opportunities” (World Study, and the Organisation for Economic Bank 2005, 3). Brunori, Ferreira, and Peragine Co-operation and Development (OECD) (2013) find that, across countries, measures focus on high-income countries, while the of inequality of opportunity are positively Socio-Economic Database for Latin Amer- correlated with measures of inequality of out- ica and the Caribbean (SEDLAC) covers comes. They also find a negative correlation only countries in Latin America and the between inequality of opportunity and inter- Caribbean. The global coverage of Povcal- generational mobility. Similarly, Corak (2013) Net comes at a cost of lower comparabil- shows a negative correlation between inequality, such as in the use of both income and ity of outcomes and mobility. consumption surveys. PovcalNet is used 5. Atkinson (2015). here because it is based on microdata (as 6. World Bank (2005). Interventions that reduce opposed to, for example, the World Income inequality without compromising economic Inequality Database) and because of its growth are described here generically as global coverage. equity-enhancing policies (see chapter 1). In a background paper for this report, the World Development Report 2006: Equity and results discussed here are compared with the Development calls these “efficient redistribu- other microdata-based sources mentioned tion” policies (see World Bank 2005, 74). above. Comparisons need to be done care- 7. Levy (2008). fully because these sources may use differ- 8. Clements et al. (2015). ent equivalence scales, welfare aggregates, or 9. Dollar, Kleineberg, and Kraay (2016) and surveys.71 While the within-country trends Dollar and Kraay (2002) document the are quite similar, the levels of inequality can central importance of growth for a range be quite different. However, the ranking of of social welfare functions, including, for countries by level of inequality is reasonably robust across the data sources. instance, the growth of the bottom 20 and the bottom 40 or the growth of the income share of the bottom 20. Notes 10. The evidence is exhaustively reviewed by the 1. As discussed in the Overview, this decom- Commission on Growth and Development position was formalized by Datt and Raval- (2008). lion (1992). This does not imply that every 11. For example, see Gill, Revenga, and Zeballos inequality-reducing transfer would re- (2016). duce the poverty headcount ratio. For in- 12. For example, as discussed in Gill, Revenga, stance, a pro-poor transfer among the poor and Zeballos (2016), China’s economic would leave the poverty headcount ratio growth has been critical to poverty reduction unchanged. since the onset of China’s deep economic 2. Clark and D’Ambrosio (2015); World Bank transformation in 1978, first based on ag- (2005). ricultural productivity gains and then on 3. Leveling the playing field aims at removing export orientation. China’s levels of human obstacles to a decent life that individuals capital were already acceptable at the start of may face because of circumstances fixed at the reforms, and its traditional safety net has birth, such as sex, belonging to a particu- been transformed into a modern social assis- lar ethnic group or race, or family religious tance system. environment. These circumstances are, to 13. Dollar, Kleineberg, and Kraay (2015, 2016); a large extent, external to the decisions, ef- World Bank (2016a).

INEQUALITY 91 14. Dollar, Kleineberg, and Kraay (2015, 2016); are beyond the scope of this chapter. The World Bank (2016a). World Bank has worked extensively on mea- 15. Bourguignon (2004) described this poverty- suring and monitoring horizontal inequali- growth-inequality triangle more than 10 ties, for example, through the calculation of years ago. A sizable literature discusses the the human opportunity index. This index relationships among poverty reduction, measures access to basic services and cor- growth, and inequality. Analogous to the rects the observed access rate by the degree evidence presented here, this literature im- of intergroup inequality. The relevant roles plies that changes in inequality are, on aver- played by race and ethnicity, gender, age, lo- age, uncorrelated with economic growth, as cation, and socioeconomic status of house- reviewed by Ferreira (2012). holds have been accounted for and com- 16. World Bank (2016a). pared in Latin America and the Caribbean, 17. Reducing inequality is not always associated Sub-Saharan Africa, and the Middle East with good outcomes. For example, the re- and North Africa. See Barros et al. (2009); duction of inequality observed in the Mid- Dabalen et al. (2015); Krishnan et al. (2016). dle East and North Africa region coexisted 24. A relative measure obeys the scale invari- with an increasing sense of frustration and ance axiom (see box 4.3). Among relative resentment drawing from redistribution measures, the analysis here does not involve without voice, a shortage of quality jobs in tests for robustness to alternative measures the formal sector, poor-quality public ser- (or Lorenz dominance more generally), but vices, and a lack of government account- mostly relies on the Gini index. ability. A lesson from this experience is that 25. Detailed robustness checks are presented how inequality is reduced matters: a system in Lakner and Silwal (2016). See also the of generalized subsidies conceived to reduce special issue of the Journal of Economic In- poverty and inequality did not overcome the equality that reviews the various inequality rising frustration because of waning oppor- databases (Ferreira, Lustig, and Teles 2015). tunities. See World Bank (2015). 26. SEDLAC (Socio-Economic Database for 18. World Bank (2014). Latin America and the Caribbean), Center 19. For example, see Atkinson and Piketty for Distributive, Labor, and Social Studies, (2007, 2010). Facultad de Ciencias Económicas, Universi- 20. See, for instance, López-Calva and Lustig dad Nacional de La Plata, La Plata, Argentina; (2010). Equity Lab, Team for Statistical Develop- 21. This is remarkable given that the house- ment, World Bank, Washington, DC, http:// hold survey in Argentina uses income, while sedlac.econo.unlp.edu.ar/eng/statistics.php. China uses consumption expenditure. In- 27. As summarized by Beegle et al. (2016), other equality in income tends to be greater than aspects that could make surveys incompa- inequality in consumption. rable include changes in survey design (for 22. This section uses household surveys from instance, urban or national coverage), im- around the world, relying primarily on plementation (such as effects of seasonality), PovcalNet, an online tool that allows users to or the questionnaires (for example, a recall access the World Bank’s repository of house- period for consumption expenditure). Al- hold surveys. It is designed mainly to allow though PovcalNet enforces some degree of users to replicate World Bank estimates comparability across countries, differences of absolute poverty, but it also provides might exist in terms of these aspects or the distributional statistics such as the Gini type of spatial price adjustment. In addition index. See PovcalNet (online analysis tool), to the potential nonresponse at the top tail, World Bank, Washington, DC, http://iresearch surveys might exclude the poorest living in .worldbank.org/PovcalNet/. remote regions or poorly measure their in- 23. This chapter concentrates on vertical incomes; for instance, Meyer et al. (2015) dis- equality, that is, comparing the rich and the cuss missing transfer incomes in the United poor. A horizontal perspective might look States. at differences among people based on sex, 28. See the explanation of the twin goals in the ethnicity, location of residence, or age. These Overview and in chapter 1. For ease of expo-

92 POVERTY AND SHARED PROSPERITY 2016 sition, we tend to refer to income and con- cline between 1988 and 2013 is robust to a sumption interchangeably in this chapter, number of robustness checks (Lakner and unless otherwise indicated. Milanovic´ 2016b). First, there exists Lorenz 29. See Anand and Segal (2015). Some authors, dominance; so, the assessment is robust to such as Alvaredo and Gasparini (2015) in alternative measures of inequality (Atkinson their recent review of inequality trends in 1970). Second, the decline is similar if the developing countries, use an additive or same set of countries is used throughout, in- multiplicative adjustment factor to reduce stead of the largest possible sample in every the income-based Gini indexes in Latin year. This is important because the popula- America. Regarding definitional differences tion coverage declines in 2013, especially in in general (such as gross versus net income, Sub-Saharan Africa (see annex 4A). Further- or equivalence scales), Atkinson and Bran- more, as shown by Lakner and Milanovic´ dolini (2001) caution against simple across- (2016a), the fall between 2003 and 2008 is the-board adjustments. Furthermore, any even more dramatic if an adjustment for the such adjustment would be inconsistent with underreporting of top incomes is included. the World Bank approach to measuring pov- 38. This convergence effect is greater if GDP erty and shared prosperity. per capita is used instead of the means from 30. Alvaredo and Gasparini (2015). household surveys. As a result, studies that 31. Aguiar and Bils (2015). adjust to GDP per capita, such as Bour- 32. Karabarbounis and Neiman (2014). guignon (2015), find a decline in global in- 33. Alvaredo and Gasparini (2015). equality that is more rapid than the results 34. The perspective of global inequality adopted presented here. here corresponds to Concept 3 in the Mila- 39. Jirasavetakul and Lakner (2016). novic´ (2005) taxonomy. 40. China’s growth has pushed down between- 35. Bourguignon (2015). country inequality in the region, while the 36. As reviewed by Anand and Segal (2015), the increase in inequality in China has pushed main methodological differences concern up the within-country component. (1) the use of GDP per capita or average 41. These are not the first results on average household survey income, (2) the adjust- within-country inequality. For instance, ment for differences between income and World Bank (2005) and Ferreira and Raval- consumption surveys, and (3) different PPP lion (2009) follow a similar structure of dis- exchange rates. The methodology used in cussing the context of global inequality be- figure 4.5, which is described in more de- fore focusing on within-country inequality. tail in Lakner and Milanovic´ (2016a), is What sets our analysis apart from the recent consistent with the approach to global pov- literature is that it covers all available coun- erty measurement (see chapter 2). It uses tries regardless of region or income level. household survey income or consumption In contrast, Alvaredo and Gasparini (2015), expressed in 2011 PPP U.S. dollars without who also use PovcalNet data, only cover adjusting for differences between income developing countries and adjust for differ- and consumption surveys. Given the data ences in inequality levels between income limitations in the early years, each country and consumption surveys. Morelli, Smeed- distribution is approximated by 10 deciles. ing, and Thompson (2015) are also closely This tends to underestimate within-country related, but they cover only rich and (some) (and thus global) inequality, but the differ- middle-income countries. See PovcalNet ence is small (Anand and Segal 2015). In a (online analysis tool), World Bank, Wash- slightly different version of the global distri- ington, DC, http://iresearch.worldbank.org bution, the shift from percentiles to deciles /PovcalNet/. reduces the global Gini by around 0.5 points. 42. In fact, this estimate can be directly com- 37. These errors refer to both sampling- and pared with the within-country inequality nonsampling-related sources, which analysts part of the global decomposition (figure can do little to correct once the surveys are 4.5), although there exist small differences. collected. However, while no estimates of However, these differences are unlikely to sampling uncertainty are available, the de- be significant. Furthermore, they may arise

INEQUALITY 93 from multiple sources, such as differences been discussed by Beegle et al. (2016) and in the inequality measure (mean log devi- Cornia (2014). Milanovic´ (2003) argues that ation vs. Gini index), the country cover- African inequality is too high given the rela- age, and the use of decile groups vs. the full tively widely shared land ownership. microdata. 53. Morelli, Smeeding, and Thompson (2015) 43. Owing to the equal treatment of countries report trends in the Gini index of equivalent in the unweighted analysis, the impact on household income drawn from data of the the worldwide trend of the rapid increase in Organisation for Economic Co-operation inequality in large countries, notably, China, and Development (OECD) and national is less than the impact on the weighted esti- statistical offices. mates of average within-country inequality. 54. See Beegle et al. (2016). 44. In addition, for a country to be included in 55. As a rough adjustment for sampling errors, the balanced sample, it needs to have the we ignore changes within 1 point. With a same welfare measure (income or consump- lack of confidence intervals, any such cut- tion) throughout, as explained in more de- off value is essentially arbitrary. We follow tail in annex 4A. Alvaredo and Gasparini (2015), who use 1 45. See chapter 2, annex 2B, for the list of indus- point, which, they argue, is typically signifi- trialized countries. cant in the SEDLAC surveys (also see Statis- 46. The decline in inequality in the region is tics Canada, referenced by Atkinson 2003). supported by the expansion in real hourly Atkinson, Rainwater, and Smeeding (1995) earnings among the bottom of the wage dis- refer to a change in the Gini index between tribution and, to a lesser extent, the middle 1 and 2 points as a modest increase. Other au- part of the earnings distribution. The de- thors have used +/− 1 percent or +/− 2 per- cline in wage inequality in Latin America cent (Burniaux et al. 1998; Smeeding 2000), has been closely associated with a reversal which would correspond to 0.4 Gini points in the college/primary education premium and 0.8 points, respectively, evaluated at the and in the urban-rural earnings gap, cou- average within-country Gini of 40 in our pled with a steady drop (which accelerated sample. This suggests that our cutoff value is markedly beginning in the first decade of the relatively conservative. However, even if such 2000s) in the high school/primary education changes may be statistically significant, they premium as well as a decline in the experi- are not necessarily economically significant, ence premium across all age-groups. See for which Atkinson (2003) proposes a 3 point Rodríguez-Castelán et al. (2016). cutoff (in his sample of OECD countries). 47. Székely and Mendoza (2015). Without any adjustment for sampling errors, 48. Cord et al. (2016); Gasparini, Cruces, and inequality increases (falls) in 46 (45) coun- Tornarolli (2016). tries in the long-run spells, and in 30 (51) 49. Regional averages can mask substantial vari- countries in the short-run spells. ability across countries. This is also notable 56. As a result, the population-weighted aver- in the case of Sub-Saharan Africa, where age Gini in this sample increased by 2.1 countries are almost evenly split across ris- points, from 37.1 to 39.2, between 1993 and ing and falling inequality in recent years (see 2008. Beegle et al. 2016; Cornia 2014). 57. However, there might be issues with survey 50. Milanovic´ and Ersado (2010). comparability. Cornia (2014) also docu- 51. One reason the trends in regional averages mented this bifurcation of inequality trends need to be interpreted carefully is the change in Sub-Saharan Africa. Recently, Beegle et al. in country composition from one bench- (2016) find that a set of African countries mark year to the next, especially in regions with at least two strictly comparable and with many countries that do not have sur- recent surveys is split evenly into rising and veys every five years. This explains some of falling inequality. Their surveys are drawn the large changes in East Asia and South Asia from the 2000s and, so, concern a slightly illustrated in figure 4.7. different period relative to the trends in this 52. This has been noted by Ferreira and Raval- chapter. lion (2009). Africa’s wide inequality has also 58. Hallegatte et al. (2016).

94 POVERTY AND SHARED PROSPERITY 2016 59. Lakner and Milanovic´ (2016a) explain the https://datahelpdesk.worldbank.org/knowl data construction in more detail. In the 2013 edgebase/articles/906519-world-bank-coun benchmark year, one exception has been try-and-lending-groups. See chap ter 2, annex made to achieve sufficient population cov- 2B, for the list of industrialized countries. erage: the 2009/10 and 2011/12 surveys in 66. This is also different from the balanced sam- India are only two years apart. ple, in which countries need to be present in 60. These data are publicly available from all years in between. PovcalNet (online analysis tool), World Bank, 67. For countries for which the welfare measure Washington, DC, http://iresearch.worldbank changes, the initial and final years (within .org/PovcalNet/. They are supplemented the two-year window defined by the bench- with data from the earlier releases of mark years) are changed where possible. For PovcalNet where possible. these countries, some of the observations in- 61. For a small number of countries on which cluded in the trend sample are thus slightly microdata are not available (for example, different from the ones included in the orig- China), PovcalNet continues to use grouped inal cross-sectional benchmark-year sam- data, in combination with a parametric ple. For instance, this applies to Bulgaria, Lorenz curve. Kazakhstan, and Ukraine. 62. In the final database of benchmark years, 68. For the data and documentation, see GDSP about 9 percent of observations are drawn (Global Database of Shared Prosperity), from the All the Ginis database, especially World Bank, Washington, DC, http://www during the earlier years. While the All the .worldbank.org/en/topic/poverty/brief Ginis database includes a mix of primary and /global-database-of-shared-prosperity. secondary sources, it is used here only for 69. These comparability standards are stricter the observations that are based on the Lux- than for the long-run sample, for which it is embourg Income Study. These observations only required that the welfare aggregate be are calculated directly from microdata using either income or consumption. Beegle et al. per capita household disposable income (2016) discuss the issues around comparing among individuals (Milanovic´ 2014), which consumption aggregates over time. is consistent with PovcalNet. For the data- 70. Another requirement is that an estimate of base, see All the Ginis (dataset), World Bank, the Gini index be available in PovcalNet for Washington, DC, http://go.worldbank.org the same year and welfare aggregate. The /9VCQW66LA0. World Bank Global Database of Shared 63. Among surveys that are equidistant to the Prosperity reports estimates of shared pros- benchmark year, the more recent survey is perity, but no Gini indexes. preferred. For a few countries, the years that 71. See Lakner and Silwal (2016). For example, were used for a particular benchmark year the Luxembourg Income Study and OECD were changed so that the countries have the use the square root of household size, while, same welfare aggregate throughout and can in Eurostat’s scale, the weight varies with the thus be included in the balanced subsample. age of the household member. 64. Population data are obtained from World Development Indicators, supplemented References with data in World Economic Outlook for any missing countries. See World Economic Aguiar, Mark, and Mark Bils. 2015. “Has Con- Outlook Database, International Monetary sumption Inequality Mirrored Income In- Fund, Washington, DC, https://www.imf equality.” American Economic Review 105 (9): .org/external/pubs/ft/weo/2016/01/weodata 2725–56. /index.aspx; WDI (World Development Indi- Alesina, Alberto, and Dani Rodrik. 1994, “Dis- cators) (database), World Bank, Washington, tributive Politics and Economic Growth.” DC, http://data.worldbank.org/data-catalog Quarterly Journal of Economics 109 (2): /world-development-indicators. 465–90. 65. The World Bank geographical classification is Alvaredo, Facundo, Anthony B. Atkinson, available at “World Bank Country and Lend- Thomas Piketty, and Emmanuel Saez. 2013. ing Groups,” World Bank, Washington, DC, “The Top 1 Percent in International and His-

INEQUALITY 95 torical Perspective.” Journal of Economic Per- Say?” Journal of Economic Growth 8 (3): spectives 27 (3): 3–20. 267–99. Alvaredo, Facundo, and Leonardo Gasparini. Barros, Ricardo Paes de, Francisco H. G. Ferreira, 2015. “Recent Trends in Inequality and Pov- José R. Molinas Vega, and Jaime Saavedra- erty in Developing Countries.” In Handbook Chanduvi. 2009. Measuring Inequality of of Income Distribution, vol. 2A, edited by Opportunities in Latin America and Carib- Anthony B. Atkinson and François Bour- bean. With Mirela de Carvalho, Samuel guignon, 697–805. Handbooks in Economics. Franco, Samuel Freije-Rodríguez, and Amsterdam: North-Holland. Jérémie Gignoux. Latin American Develop- Anand, Sudhir, and Paul Segal. 2015. “The ment Forum Series. Washington, DC: World Global Distribution of Income.” In Hand- Bank; New York: Palgrave Macmillan. book of Income Distribution, vol. 2A, edited Beegle, Kathleen, Luc Christiaensen, Andrew by Anthony B. Atkinson and François Bour- Dabalen, and Isis Gaddis. 2016. Poverty in a guignon, 937–79. Handbooks in Economics. Rising Africa. Washington, DC: World Bank. Amsterdam: North-Holland. Bourguignon, François. 2004. “The Poverty- Arampatzi, Efstratia, Martijn Burger, Elena Ian- Growth-Inequality Triangle.” Working Paper chovichina, Tina Röhricht, and Ruut Veen- 125, Indian Council for Research on Interna- hoven. 2015. “Unhappy Development: Dis- tional Economic Relations, New Delhi. satisfaction with Life in the Wake of the Arab ———. 2015. The Globalization of Inequality. Translated by Thomas Scott-Railton. Prince- Spring.” Policy Research Working Paper 7488, ton, NJ: Princeton University Press. World Bank, Washington, DC. Brunori, Paolo, Francisco H. G. Ferreira, and Atkinson, Anthony B. 1970. “On the Measure- Vito Peragine. 2013. “Inequality of Oppor- ment of Inequality.” Journal of Economic tunity, Income Inequality, and Economic Theory 2 (3): 244–63. Mobility: Some International Comparisons.” ———. 2003. “Income Inequality in OECD Policy Research Working Paper 6304, World Countries: Data and Explanations.” CESifo Bank, Washington, DC. Economic Studies 49 (4): 479–513. Burniaux, Jean-Marc, Thai-Thanh Dang, Doug- ———. 2015. Inequality: What Can Be Done? las Fore, Michael F. Förster, Marco Mira Cambridge, MA: Harvard University Press. d’Ercole, and Howard Oxley. 1998. “Income Atkinson, Anthony B., and Andrea Brandolini. Distribution and Poverty in Selected OECD 2001. “Promise and Pitfalls in the Use of Countries.” OECD Economics Department ‘Secondary’ Data-Sets: Income Inequality in Working Paper 189 (March), Organisation OECD Countries as a Case Study.” Journal of for Economic Co-operation and Develop- Economic Literature 39 (3): 771–99. ment, Paris. ———. 2010. “On Analyzing the World Dis- Cantril, Hadley. 1965. The Pattern of Human tribution of Income.” World Bank Economic Concerns. New Brunswick, NJ: Rutgers Uni- Review 24 (1): 1–37. versity Press. Atkinson, Anthony B., and Thomas Piketty, eds. Chattopadhyay, Soumya, and Carol L. Graham. 2007. Top Incomes over the Twentieth Century: 2015. “From Arab Spring to Fall: Some In- A Contrast between Continental European sights from the Economics of Well-Being.” and English-Speaking Countries. New York: Brookings Institution, Washington, DC. Oxford University Press. Clark, Andrew E., and Conchita D’Ambrosio. ———, eds. 2010. Top Incomes: A Global Per- 2015. “Attitudes to Income Inequality: Ex- spective. New York: Oxford University Press. perimental and Survey Evidence.” In Hand- Atkinson, Anthony B., Lee Rainwater, and Timo- book of Income Distribution, vol. 2A, edited thy M. Smeeding. 1995. “Income Distribution by Anthony B. Atkinson and François Bour- in Advanced Economies: Evidence from the guignon, 1127–1208. Handbooks in Eco- Luxembourg Income Study (LIS).” Luxem- nomics. Amsterdam: North-Holland. bourg Income Study Working Paper 120, LIS Clements, Benedict, Ruud de Mooij, Sanjeev Cross-National Data Center, Luxembourg. Gupta, and Michael Keen, eds. 2015. Inequal- Banerjee, Abhijit, and Esther Duﬂo. 2003. “In- ity and Fiscal Policy. Washington, DC: Inter- equality and Growth: What Can the Data national Monetary Fund.

96 POVERTY AND SHARED PROSPERITY 2016 Commission on Growth and Development. 2008. N. Jefferson, 427–62. New York: Oxford Uni- The Growth Report: Strategies for Sustained versity Press. Growth and Inclusive Development. Washing- Ferreira, Francisco H. G., Christoph Lakner, ton, DC: World Bank. https://openknowledge María Ana Lugo, and Berk Özler. 2014. “In- .worldbank.org/handle/10986/6507. equality of Opportunity and Economic Corak, Miles. 2013. “Income Inequality, Equality Growth: A Cross-Country Analysis.” Policy of Opportunity, and Intergenerational Mo- Research Working Paper 6915, World Bank, bility. Journal of Economic Perspectives 27 (3): Washington, DC. 79–102. Ferreira, Francisco H. G., Nora Lustig, and Dan- Cord, Louise, Oscar Barriga Cabanillas, Leonardo iel C. Teles. 2015. “Appraising Cross-National Lucchetti, Carlos Rodríguez-Castelán, Lili- Income Inequality Databases: An Introduc- ana D. Sousa, and Daniel Valderrama, 2016. tion.” Journal of Economic Inequality 13 (4): “Inequality Stagnation in Latin America in 497–526. the Aftermath of the Global Financial Crisis.” Ferreira, Francisco H. G., and Martin Ravallion, Review of Development Economics May 19. 2009. “Poverty and Inequality: The Global Cornia, Giovanni Andrea. 2014. “Income In- Context.” In The Oxford Handbook of Eco- equality Levels, Trends, and Determinants nomic Inequality, edited by Wiemer Salverda, in Sub-Saharan Africa: An Overview of the Brian Nolan, and Timothy M. Smeeding, Main Changes.” Università degli Studi di 599–638. New York: Oxford University Press. Firenze, Florence. Forbes, Kristin. 2000. “A Reassessment of the Re- Dabalen, Andrew, Ambar Narayan, Jaime lationship between Inequality and Growth.” Saavedra-Chanduvi, and Alejandro Hoyos American Economic Review 90 (4): 869–87. Suarez. 2015. Do African Children Have an Gallup. 2011a. “Understanding How Gallup Equal Chance? A Human Opportunity Report Uses the Cantril Scale.” Latest News. Gallup, for Sub-Saharan Africa. With Ana Abras and Washington, DC. http://www.gallup.com/poll Sailesh Tiwari. Directions in Development: /122453/understanding-gallup-uses-cantril Poverty Series. Washington, DC: World Bank. -scale.aspx. Dabla-Norris, Era, Kalpana Kochhar, Frantisek ———. 2011b. “Egypt from Tahrir to Transi- Ricka, Nujin Suphaphiphat, and Evridiki tion.” Latest News. Abu Dhabi Gallup Center, Tsounta. 2015. “Causes and Consequences United Arab Emirates. http://www.gallup.com of Income Inequality: A Global Perspec- /poll/157046/egypt-tahrir-transition.aspx. tive.” IMF Staff Discussion Note SDN/15/13 Gasparini, Leonardo, Guillermo Cruces, and (June), International Monetary Fund, Wash- Leopoldo Tornarolli. 2016. “Chronicle of a ington, DC. Deceleration Foretold: Income inequality in Datt, Gaurav, and Martin Ravallion. 1992. Latin America in the 2010s.” Documento de “Growth and Redistribution Components of Trabajo 198 (May), Center for Distributive, Changes in Poverty Measures: A Decomposi- Labor, and Social Studies, Facultad de Cien- tion with Applications to Brazil and India in cias Económicas, Universidad Nacional de La the 1980s.” Journal of Development Economics Plata, La Plata, Argentina. 38 (2): 275–95. Gill, Indermit, Ana Revenga, and Christian Dollar, David, Tatjana Kleineberg, and Aart Zeballos. 2016. Grow, Invest, Insure: A Game Kraay. 2015. “Growth, Inequality, and Social Plan to End Poverty. Washington, DC: World Welfare: Cross-Country Evidence.” Economic Bank. Policy 30 (82): 335–77. Graham, Carol L. 2016. Happiness for All? Un- ———. 2016. “Growth is Still Good for the equal Lives and Unequal Hopes in the Land of Poor.” European Economic Review 81: 68–85. the Dream. Princeton, NJ: Princeton Univer- Dollar, David, and Aart Kraay. 2002. “Growth sity Press. Is Good for the Poor.” Journal of Economic Graham, Carol L., and Eduardo Lora. 2009. Par- Growth 7 (3): 195–225. adox and Perceptions: Quality of Life in Latin Ferreira, Francisco H. G. 2012. “Distributions in America. Washington, DC: Brookings Institu- Motion: Economic Growth, Inequality, and tion Press. Poverty Dynamics.” In The Oxford Handbook Hallegatte, Stephane, Mook Bangalore, Laura of the Economics of Poverty, edited by Philip Bonzanigo, Marianne Fay, Tomaro Kane,

INEQUALITY 97 Ulf Narloch, Julie Rozenberg, David Treguer, Lakner, Christoph, and Ani Rudra Silwal. 2016. and Adrien Vogt-Schilb. 2016. Shock Waves: “Inequality Increasing Everywhere? Conﬂict- Managing the Impacts of Climate Change on ing Evidence from an Updated Global Data- Poverty. Climate Change and Development base of Household Surveys.” World Bank, Series. Washington, DC: World Bank. Washington, DC. Hassine, Nadia Belhaj. 2015. “Economic Inequal- Levy, Santiago. 2008. Good Intentions, Bad Out- ity in the Arab Region.” World Development comes: Social Policy, Informality, and Eco- 66: 532-56. nomic Growth in Mexico. Washington, DC: Iqbal, Farrukh, and Youssouf Kiendrebeogo. Brookings Institution Press. 2014. “The Reduction of Child Mortality in López-Calva, Luis F., and Nora Lustig, eds. 2010. the Middle East and North Africa: A Success Declining Inequality in Latin America: A De- Story.” Policy Research Working Paper 7023, cade of Progress? New York: United Nations World Bank, Washington, DC. Development Programme; Washington, DC: Jenkins, Stephen P. 2015. “World Income In- Brookings Institution. equality Databases: An Assessment of WIID Marrero, Gustavo A., and Juan Gabriel Rodrí- and SWIID.” Journal of Economic Inequality guez. 2013. “Inequality of Opportunity and 13 (4): 629–71. Growth.” Journal of Development Economics Jirasavetakul, La-Bhus Fah, and Christoph 104: 107–22. Lakner. 2016. “The Distribution of Con- Meyer, Bruce D., Wallace K. C. Mok, and sumption Expenditure in Sub-Saharan Af- James X. Sullivan. 2015. “Household Surveys rica: The Inequality among All Africans.” in Crisis.” Journal of Economic Perspectives 29 Policy Research Working Paper 7557, World (4): 199–226. Bank, Washington, DC. Milanovic´, Branko. 2003, “Is Inequality in Africa Karabarbounis, Loukas, and Brent Neiman. Really Different?” Policy Research Working 2014. “The Global Decline of the Labor Paper 3169, World Bank, Washington, DC. Share.” Quarterly Journal of Economics 129 ———. 2005. Worlds Apart: Measuring Interna- (1): 61–103. tional and Global Inequality. Princeton, NJ: Khouri, Rami, and Joanne J. Myers. 2015. “Per- Princeton University Press. spectives from Inside a Tumultuous Middle ———. 2014. “All the Ginis, 1950–2012 (up- East: Syria-Iraq-ISIS-Russia and Iran.” Pub- dated in Autumn 2014).” November, World lic Affairs, Global Ethics Forum TV Series Bank, Washington, DC. http://go.worldbank (November 18), Carnegie Council for Ethics .org/9VCQW66LA0. in International Affairs, New York. http:// ———. 2016. Global Inequality: A New Approach www.carnegiecouncil.org/studio/multime- for the Age of Globalization. Cambridge, MA: dia/20151118/index.html. Harvard University Press. Kraay, Aart. 2015. “Weak Instruments in Milanovic´, Branko, and Ersado, Lire, 2010. “Re- Growth Regressions: Implications for Re- form and Inequality during the Transition: cent Cross-Country Evidence on Inequality An Analysis Using Panel Household Survey and Growth.” Policy Research Working Paper Data, 1990–2006.” Working Paper 2010–62, 7494, World Bank, Washington, DC. World Institute for Development Economics Krishnan, Nandini, Gabriel Lara Ibarra, Ambar Research, Helsinki. Narayan, Sailesh Tiwari, and Tara Vishwa- Morelli, Salvatore, Timothy Smeeding, and Jef- nath. 2016. Uneven Odds, Unequal Outcomes: frey Thompson. 2015. “Post-1970 Trends in Inequality of Opportunity in the Middle East within-Country Inequality and Poverty: Rich and North Africa. Directions in Development: and Middle-Income Countries.” In Hand- Poverty Series. Washington, DC: World Bank. book of Income Distribution, vol. 2A, edited Lakner, Christoph, and Branko Milanovic´. by Anthony B. Atkinson and François Bour- 2016a. “Global Income Distribution: From guignon, 593–696. Handbooks in Economics. the Fall of the Berlin Wall to the Great Re- Amsterdam: North-Holland. cession.” World Bank Economic Review 30 (2): Ncube, Mthuli, John Anyanwu, and Kjell 203–32. Hausken. 2013. “Inequality, Economic ———. 2016b. “Global Inequality after the Great Growth, and Poverty in the Middle East and Recession.” World Bank, Washington, DC. North Africa (MENA).” Working Paper 193

98 POVERTY AND SHARED PROSPERITY 2016 (December), African Development Bank, Here to Stay?” Journal of Human Develop- Tunis, Tunisia. ment and Capabilities 16 (3): 397–419. Ostry, Jonathan, Andrew Berg, and Charalambos Teyssier, Geoffrey. 2015. “Inequality of Opportu- Tsangarides. 2014. “Redistribution, Inequal- nity and Growth: An Empirical Investigation ity, and Growth.” IMF Staff Discussion Note in Brazil.” Paper presented at the sixth meet- SDN/14/02 (April), International Monetary ing of the Society for the Study of Economic Fund, Washington, DC. Inequality, Université du Luxembourg, Cam- Persson, Torsten, and Guido Tabellini. 1994. “Is pus Kirchberg, Luxembourg, July 13–15. Inequality Harmful for Growth?” American van der Weide, Roy, Christoph Lakner, and Elena Economic Review 84 (3): 600–21. Ianchovichina. 2016. “Is Inequality Under- Ravallion, Martin. 2016. The Economics of Pov- estimated in Egypt? Evidence from House erty: History, Measurement, and Policy. New Prices.” Policy Research Working Paper 7727, York: Oxford University Press. World Bank, Washington, DC. Rodríguez-Castelán, Carlos, Luis F. López-Calva, Verme, Paolo, Branko Milanovic´, Sherine Al- Nora Lustig, and Daniel Valderrama. 2016. Shawarby, Sahar El Tawila, May Gadallah, “Understanding the Dynamics of Labor In- and Enas Ali A. El-Majeed. 2014. Inside In- come Inequality in Latin America.” Policy equality in the Arab Republic of Egypt: Facts Research Working Paper 7795, World Bank, and Perceptions across People, Time, and Washington, DC. Space. World Bank Study Series. Washington, Roemer, John E. 1998. Equality of Opportunity. DC: World Bank. Cambridge, MA: Harvard University Press. Voitchovsky, Sarah. 2009. “Inequality and Eco- SEDLAC (Socio-Economic Database for Latin nomic Growth.” In The Oxford Handbook of America and the Caribbean), Center for Dis- Economic Inequality, edited by Wiemer Sal- tributive, Labor, and Social Studies, Univer- verda, Brian Nolan, and Timothy M. Smeed- sidad de La Plata, La Plata, Argentina; World ing, 549–74. New York: Oxford University Bank, Washington, DC (accessed June 14), Press. http://sedlac.econo.unlp.edu.ar/eng/index World Bank. 2005. World Development Report .php. 2006: Equity and Development. Washington, Smeeding, Timothy M., 2000 “Changing In- DC: World Bank; New York: Oxford Univer- come Inequality in OECD Countries: Up- sity Press. dated Results from the Luxembourg Income ———. 2014. The Unﬁnished Revolution: Bring- Study (LIS).” In The Personal Distribution of ing Opportunity, Good Jobs, and Greater Income in an International Perspective, edited Wealth to All Tunisians. Development Policy by Richard Hauser and Irene Becker, 205–24. Review (May). Washington, DC: World Bank. Berlin: Springer-Verlag. ———. 2015. “Inequality, Uprisings, and Con- Solt, Frederick. 2016. “Standardizing the World ﬂict in the Arab World.” MENA Economic Income Inequality Database.” Social Science Monitor (October), World Bank, Washington, Quarterly 90 (2): 231–42. DC. Souza, Pedro H. G. Ferreira de, Marcelo Medeiros, ———. 2016a. Global Monitoring Report 2015/16: and Fabio A. Castro. 2015. “‘Top Incomes in Development Goals in an Era of Demographic Brazil: Preliminary Results.” Economics Bulle- Change. Washington, DC: World Bank. tin 35 (2): 998–1004. ———. 2016b. Indonesia’s Rising Divide: Why Székely, Miguel, and Pamela Mendoza. 2015. “Is Inequality Is Rising, Why It Matters, and What the Decline in Inequality in Latin America Can Be Done. Jakarta: World Bank.

INEQUALITY 99

5 Reductions in Inequality: A Country Perspective

This chapter describes the key drivers behind the remarkable progress achieved by selected countries in boosting shared prosperity, narrowing inequality, and reducing poverty. These countries—Brazil, Cambodia, Mali, Peru, and Tanzania—are diverse in terms of geographical location, income status, development trajectory, and historical background. The heterogeneity of their experiences facilitates an examination of how environment, macroeconomic policies, sectoral strategies, and the management of external shocks interact to produce successful outcomes. The countries all exercised cautious macroeconomic management, appropriately dealt with external shocks, and implemented extensive and orderly economic and social sector reforms. They also benefited from a global context that was particularly favorable, with abundant and cheap credit in international markets, booming trade, and high commodity prices. These steps and the context allowed rapid, sustainable, and inclusive growth. They also highlight the importance of labor markets in translating economic growth into an expansion in opportunities by creating jobs and increasing earnings, promoting the integration of individuals otherwise excluded from economic and social advancement, and reducing gaps among workers owing to gender, residence, or sector. Country-specific choices also play a role in rolling back inequalities. Thus, the minimum wage and safety nets have played a substantial role in Brazil, while the shift from agriculture toward light manufacturing and services in Cambodia opened up employment for the poor. The examples confirm that good macroeconomic policies are essential to fostering shared prosperity. Rapid, sustained economic growth and a favorable context render the effort to establish equality easier. Coherent domestic policies, fiscal space, and improvement in the functioning of labor markets help undermine inequality. While the specific drivers are likely to diverge across countries, sustaining inequality reduction into the future may require some of the same elements: regular investments in human capital accumulation, enhanced productivity, economic diversification, spending on infrastructure to link lagging regions, and the framing of adequate safety nets. These country experiences also demonstrate that success in narrowing inequality does not necessarily translate into success on other economic, political, and social fronts. For example, conflict has recently resurged in Mali after two decades of relative stability. Despite some diversification in the economy, the absence of growth and competition continue to characterize the private sector in Tanzania. And recent decisions regarding the control of fiscal balances and inflation in Brazil have contributed to the recession in that economy as the global context has become less favorable.

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 101 Introduction been well in excess of the average growth across the population in each country. Ac- Since the early 1990s, income inequality has cording to the latest available information, been declining globally. There are, however, the growth in the bottom 40 exceeds that of large regional differences. The Latin Amer- the mean, ranging from about 2 percentage ica and Caribbean region, which has shown points in Tanzania to almost 5 percentage historically high levels of income inequal- points in Mali (one of the few countries in ity, has witnessed more progress. Overall, the world with negative average growth at country experiences confirm that inequal- the mean), while incomes among the bot- ity can widen as well as narrow. This begs tom 40 expanded at robust, positive rates. a fundamental question: what have suc- These achievements place these five coun- cessful countries done to cut into income tries among the top performers in boost- inequality that other economies have not ing shared prosperity (see chapter 3). As done? Among the constellation of policies expected from the links between shared that have been implemented, this chapter prosperity and distributional changes, most identifies the key levers in a few selected of the countries that have achieved large countries. It also stresses the importance premiums in shared prosperity also exhibit of the favorable global context during much a narrowing in income inequality as mea- of the period analyzed: low interest rates, sured by the Gini index. high international commodity prices, and Rather than an exhaustive and detailed booming international trade. Some coun- causal analysis, the country case studies tries also experienced favorable shocks that provide a modest and focused review of the were external to any policy decisions by gov- ways in which the nature of growth, policy, ernments, such as good weather spells and and context influence the reduction in in- the redirection of trade flows in Mali from equality and the growth among the bottom neighboring countries affected by conflict. 40. The case studies benefit from existing This chapter assesses how good macro- academic research, international and local economic management, sectoral reform, analytical work, and the evidence of proj- the strengthening of safety nets, responses ect and operational assessments. Evidence to external shocks, and initial conditions is provided by both partial and economy- all contribute to the effort to trim away at wide general equilibrium analyses. It is inequality and support shared prosper- from multiple sources, such as the wealth of ity. The countries have been chosen with a country reports on poverty and social im- view to extracting relevant lessons across pact analysis that examine the distributive a wide variety of successful experiences. effects of policy reform; systematic country They are therefore among the top perform- diagnostic reports that establish the key pri- ers in promoting shared prosperity, less- orities of countries in the effort to achieve ening inequality, and combating extreme the twin World Bank goals of ending pov- poverty during the periods analyzed. To erty and boosting shared prosperity; the reflect the diversity of experiences, the se- Commitment to Equity Initiative, which lection also highlights regional differences; offers a thorough assessment of the con- covers low- and middle-income-countries; tribution to equality of markets, taxation, includes countries at distinct stages of de- and social spending; and decomposition velopment, for example, agrarian and mod- analyses of trends in poverty and inequality ern economies; and is sensitive to special reduction.1 historical contexts, such as conflict or high While the five countries cannot fully rep- and pervasive levels of inequality. resent a global scale, they portray wide- The countries selected are Brazil, Cam- ranging diversity. They are in three regions: bodia, Mali, Peru, and Tanzania. These Asia, Latin America and the Caribbean, countries have had great success in shar- and Sub-Saharan Africa. They are low- or ing prosperity. The rates of income growth middle-income-countries. Two, Brazil and among the bottom 40 percent of the in- Peru, are predominantly urban societies, come distribution (the bottom 40) have while the populations of the other three

102 POVERTY AND SHARED PROSPERITY 2016 countries are largely concentrated in rural TABLE 5.1 Annualized per Capita GDP Growth and Reductions in areas. In terms of economic growth, the ex- Inequality, Selected Countries periences are wide-ranging as well. Most of Per capita GDP growth, the countries showed solid annual per cap- Country Years annualized, % Gini reduction, points ita growth in gross domestic product (GDP) Brazil 2004–14 2.4 5.5 during the period of study: Brazil, 2004–14; Cambodia 2007–13 3.4 11.0 Cambodia, 2007–13; Mali, 2001–10; Peru, Mali 2001–10 1.5 6.9 2004–14; and Tanzania, 2007–12. However, Peru 2004–14 4.8 7.1 while robust, per capita GDP growth has Tanzania 2007–12 3.0 2.7 been more modest in Brazil and Mali, with Sources: Tabulations of Equity Lab, Team for Statistical Development, World Bank, Washington, DC, based annual rates of 2.3 percent in Brazil and 1.2 on data in SEDLAC (Socio-Economic Database for Latin America and the Caribbean); “Measuring Inequal- percent in Mali. Annual per capita GDP ity,” World Bank, Washington, DC, http://go.worldbank.org/3SLYUTVY00; WDI (World Development Indi- growth in the other three countries exceeds cators) (database), World Bank, Washington, DC, http://data.worldbank.org/data-catalog /world-development-indicators; PovcalNet (online analysis tool), World Bank, Washington, DC, http:// these rates: 4.7 percent in Peru, 3.4 percent iresearch.worldbank.org/PovcalNet/. in Cambodia, and 3.0 percent in Tanzania during their respective spells. The experience of these countries in re- TABLE 5.2 Typology of Poverty and Inequality Levels, Selected ducing poverty and inequality is remark- Countries, Circa 2013 able. Poverty reductions have been substan- Level of inequality Low to moderate poverty, < 10% High poverty, > 10% tial in all five countries, but especially in Brazil, Cambodia, and Peru. The reduction Low to moderate, Gini < 30 Cambodia Brazil Tanzania in income inequality has also been consid- High, Gini > 30 erable in Brazil, Cambodia, Mali, and Peru, Peru Mali though more modest in Tanzania (table Note: Poverty is defined at the international extreme poverty line of US$1.90 a day in 2011 purchasing 5.1). Hence, in these countries, large pov- power parity (PPP) U.S. dollars. Brazil and Peru’s poverty and inequality measures are based on income, while the remaining countries use consumption-based measures. Income inequality is considered high if erty declines have typically been coupled the Gini index exceeds 30, while poverty is considered high if the incidence is 10 percent or greater. with substantive inequality reductions. Not- withstanding these favorable trends, the countries differ in the levels of poverty and generalizations. Instead, these country as- inequality. The two Latin American coun- sessments provide succinct narratives on tries showed low poverty and wide inequal- the extent to which shared factors, distinct ity; Cambodia had low poverty and narrow features, and specific contexts determine inequality; and the two African countries recent successes in boosting shared pros- experienced high poverty and wide inequal- perity and reducing inequality in selected ity (table 5.2). countries. The selected country cases also provide a relevant lesson: success in reducing inequal- Brazil: multiple policies ity does not automatically translate into a aligned to redress record similar success on other economic, political, inequality or social fronts. For example, the deterioration of governance in Mali that brewed Inequality reduction for two decades culminated in conflict in 2012, thus putting a brake on the reductions Brazil is historically known for its high and achieved during the previous decade. In pervasive inequality in incomes, in access to fact, inequality reductions can be short- basic services, such as education and health lived. While achieving substantive reduc- care, and in other measures of well-being. In tions in inequality is an accomplishment, 1989, Brazil’s Gini index was 63 and ranked sustaining these reductions is a superior feat. second highest in the world.2 However, be- The assessment of country experiences ginning in the mid-1990s, these stubborn does not seek to provide precise equalizing levels of inequality started to cede. Mirror- policy prescriptions. The small number of ing the regional trend in Latin America and countries reviewed is a sound argument the Caribbean, the Gini reached 51 in 2014, against any attempt at deriving sweeping 19 percent lower than in 1989 (figure 5.1).

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 103 FIGURE 5.1 Trends in the Gini Index, Brazil, 1981–2014 FIGURE 5.2 Growth Incidence Curve, Brazil, 2004–14 64 8 62 7 60 6

58 5

56 4 Gini index

3 54

Annual growth rate (%) 2 52 1 50 1981199020051999 1993 19841987 2008 2002 1996 2011 2014 0 5203550658095 Brazil Latin America and the Caribbean Income percentile Sources: Tabulations of Equity Lab, Team for Statistical Development, World Bank, Washington, DC, based Annualized growth of percentile on data in the SEDLAC database; WDI (World Development Indicators) (database), World Bank, Washing- Average growth ton, DC, http://data.worldbank.org/data-catalog/world-development-indicators; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. Source: Tabulations of Equity Lab, Team for Statistical Devel- Note: The data are based on a regional data harmonization effort that increases cross-country com- opment, World Bank, Washington, DC, based on data in the parability and may differ from official statistics reported by governments and national statistical offices. SEDLAC database. The welfare indicator used to compute the Gini is household per capita income. The Gini ranges from 0 (perfect equality) to 100 (perfect inequality). The Latin America and Caribbean aggregate is based on 17 countries in the region on which microdata are available. The Gini index of the Latin America and Caribbean region is computed based on pooled country-specific data previously collapsed into 8,000 individual incomes of top earners, confirms percentiles. In cases where data are unavailable for a given country in a given year, values have been interpolated by projecting incomes for that year based on GDP growth and assuming no changes in the the decline in the Gini, but at a more mod- 3 distribution of incomes in the interpolated year. Dotted lines cover years in which data are not available est pace. or present quality issues. The marked drop in income inequality helped translate economic growth into The incomes of the less well off surged large poverty reductions. Between 2004 and between 2004 and 2014 amid rapid eco- 2014, 26.5 million Brazilians exited pov- nomic growth. Figure 5.2 shows that the erty.4 While 22 in every 100 people were average annual growth in the incomes of living on an income of less than R$140 a the bottom 10 doubles that of the top 10. month in 2004, this was only true of 7 in The World Bank’s indicator of shared pros- 100 Brazilians 10 years later.5 The share of perity—the growth rate of household in- the population living on less than US$1.90 come per capita of the bottom 40—high- a day (in 2011 purchasing power parity lights how Brazil’s growth in the recent past [PPP] U.S. dollars) fell from 11.0 percent has disproportionally benefited the poorest to 3.7 percent during the period. Decom- households. Between 2004 and 2014, the position exercises conclude that about 60 income growth among the bottom 40 aver- percent of this reduction was caused by the aged 6.8 percent a year, compared with 4.5 average increase in incomes among Brazil- percent for the average Brazilian. Over the ian households (the growth effect), while same period, incomes among the bottom the remaining 40 percent can be attributed 40 in Brazil rose at the second fastest rate in to improvements in income distribution the region, only surpassed by Bolivia. This among Brazilians (figure 5.3). suggests that most of the decline in overall Despite these achievements, the country inequality occurred because of a reduction is still highly unequal. In 2014, the bottom in inequality at the bottom of the income 40 held approximately 12 percent of total distribution. Alternative evidence drawing income, while the top 20 held 56 percent.6 on tax records, which are more effective That year, Brazil’s Gini index of 51, while than household surveys at capturing the notably lower than 10 years previously, was

104 POVERTY AND SHARED PROSPERITY 2016 FIGURE 5.3 Contributions of Growth for basic education (Lei e Diretrizes de and Redistribution Effects to Poverty Bases) and national curriculum guidelines. Reduction, Brazil, 2004–14 Subsequent education policies were aimed Poverty $1.90 (2011 PPP) at decentralizing the school system, reduc- 0 ing the costs of education for poor children, and measuring and monitoring results.7 –1.0 A new macro framework in the 1990s created an enabling environment for in- –2.0 –4.5 equality reduction. First, the introduction –3.0 of the Real Plan in 1994 allowed the high levels of inflation to be cut that had been –4.0 an important contributor to increasing inequality in previous years.8 In the late 1990s, –5.0 the adoption of inflation targets, floating –6.0 –2.9 exchange rates, and a more prudent fiscal policy supported by the Fiscal Responsibil- –7.0 ity Law in 2000 created a context of macro-

Change in poverty headcount (percentage points) economic stability that prevailed during the –8.0 Growth Redistribution next decade. During the 2000s, the boom in com- Source: Tabulations of Equity Lab, Team for Statistical Devel- modity prices generated positive terms of opment, World Bank, Washington, DC, based on data in the SEDLAC database. trade to the advantage of Brazil as a major Note: The data have been calculated using a Shapley approach commodity exporter. The resulting macro- (Shorrocks 2013) that decomposes the change in the US$1.90- economic stability, combined with the fa- a-day poverty rate (2011 PPP U.S. dollars) into two components: the growth in household per capita income keeping distribution vorable external context, propelled Brazil’s constant (the growth effect) and the change in the distribution economic growth during the 2000s, which of income keeping average income constant (the redistribution accelerated the decline in inequality. The effect) as developed by Datt and Ravallion (1992). dynamics of the labor market and the expansion of social policies boosted the in- the third highest in the region, and Brazil comes of the poor.9 According to World was still the 15th most unequal country in Bank estimates, these two factors accounted the world. for the bulk of the decline in inequality, approximately 80 percent, between 2003 and What drives the reductions in 2013. Of the decline in the Gini index be- inequality? tween these two years, 41 percent was accounted for by labor incomes, and 39 per- The 1988 Constitution, which was adopted cent by nonlabor income sources such as following the reinstatement of democracy government transfers.10 in 1985, aimed to address the country’s his- In terms of the labor market, a fall in torical inequalities by guaranteeing basic the wage gap between skilled and unskilled social rights, such as free public educa- workers explains a great deal of the decrease tion, free universal health care, pensions, in labor income inequality. Various factors and social assistance. Policies launched in drove the narrowing in the wage gap. The the 1990s laid the foundations for the in- large expansion in access to education led equality declines observed years later. For to a significant increase in the relative sup- instance, the creation of the unified health ply of skilled workers. Between 1995 and system (Sistema Único de Saúde) allowed 2010, the average years of schooling among substantial progress in the level and equity adults above 25 years of age rose 56 percent, of health outcomes. In addition, education to 7.2 years. More than 4 in 10 workers in reforms were fundamental in closing edu- 2010 had 11 or more years of formal education gaps and producing a more highly cation, twice the share in the mid-1990s.11 skilled labor force. In 1996, the government In addition, positive external conditions fos- introduced the first broad legal framework tered an expansion in domestic consump-

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 105 tion, favoring activities, especially in non- Bolsa Família (family grant), Brazil’s flagship tradable sectors, such as construction and conditional cash transfer (CCT) program, services, that employed relatively less well has had a considerable equalizing impact. skilled workers. Furthermore, there was an Between 2004 and 2014, the number of ben- increase in formal sector jobs, accompanied eficiaries rose from 16 million to 56 million, by rises in the minimum wage.12 Though the reaching about a quarter of the country’s situation might change under more somber population. Bolsa Família alone explains economic conditions and a less favorable between 10 percent and 15 percent of the external context in the future, increases in reduction in income inequality observed the minimum wage during the first decade in the 2000s.16 Other targeted transfers, in- of the 2000s did not create large distortions dexed to a growing minimum wage, such in the labor market. Indeed, there were sub- as the Beneficio de Prestacao Continuada stantial reductions in the wage gaps across (continuous cash benefit), a transfer to the urban and rural areas, across regions, be- elderly and disabled, were equalizing as well. tween men and women, and between whites According to recent estimates, direct and nonwhites with similar educational taxes and transfers have reduced the Gini attainment and experience. A recent study index of market incomes by approximately estimates that the fall in gender, race, and 6 percent.17 This is a large reduction by spatial wage inequality and the growth in Latin American standards, but a paltry formal sector employment explain about 60 contribution compared with the capac- percent of Brazil’s drop in labor income in- ity of the fiscal systems of member coun- equality between 1995 and 2012.13 tries of the Organisation for Economic Moreover, during the first decade of the Co-operation and Development (OECD) 2000s, sufficient fiscal space was created for to reduce market income inequalities by up the government to finance large expansions to 33 percent.18 The in-kind benefits pro- in access to basic services and in social ex- vided through free public education and penditures. For example, access to electric- health care systems are progressive, except ity was almost universal by 2014 in large for tertiary education, while contributory part because of the rural electrification pro- pensions are clearly regressive.19 Consider- gram Luz para Todos (light for all) that pro- ing the size of the fiscal system in Brazil, the vided coverage to 15.2 million people be- impact of the system on inequality is small. ginning in 2004. Though still low, the share About three-quarters of the direct transfer of households among the bottom 40 with benefits (including contributory pensions a toilet connected to the sewerage network and unemployment benefits) go to the non- rose from 33 percent to 43 percent between poor, and the heavy reliance on indirect 2004 and 2013. Primary-school enrollment taxes acts against the inequality reduction was also close to universal.14 Nonetheless, derived from direct taxes. In addition, be- the quality of services and uneven service cause important government transfers are coverage constitute a growing challenge. indexed to it, increases in the minimum Children’s access to quality services still de- wage are associated with a large fiscal cost. pends on the economic and social circum- Estimates suggest that public expenditures stances into which children are born, which rise by R$350 million on a yearly basis for indicates that educational opportunities are every R$1 increase in the minimum wage.20 not equal in Brazil. For instance, parental Thus, despite the contribution of the fiscal educational attainment, incomes, and oc- system to reducing inequality, there is still cupation explained about 80 percent of the significant room for improving the system’s inequality in children’s mathematics test impact and efficiency. scores in 2012; these tests measure a basic In sum, building from important early ability to use mathematics in real-world reforms, macroeconomic stability, economic situations.15 growth, ample fiscal resources, and favorable Targeted government transfers also con- external conditions, Brazil has been able to tributed to the improvement in the living change course toward a more equitable soci- conditions of the poorest. The expansion of ety. The progress in the last decade in terms

106 POVERTY AND SHARED PROSPERITY 2016 of poverty and inequality is undeniable. FIGURE 5.4 Trends in the Gini Index, However, the shift in the global economy, Cambodia, 2007–13 the end of the commodity price boom, and 40 a number of policy choices (including, notably, the weakening of fiscal discipline) cur- 37.0 rently threaten Brazil’s past success in reduc- 35.5 ing income inequality. Reigniting sustainable growth by increasing investment and productivity—including nontradable services produced by low-skilled workers—will be an important precondition for sustaining 30 job creation and boosting the earnings of the Gini index poor and vulnerable. Otherwise, the decline 26.0 in income inequality in Brazil––associated with decisions on minimum wages, social transfers, and shifts in labor demand––runs the risk of being short-lived in the absence of more rapid productivity growth and fa- 20 vorable external conditions.21 Furthermore, 2004 2007 2013 in a context of much tighter fiscal resources, Source: Calculations based on data in “Measuring Inequality,” there is a growing need to improve the pro- World Bank, Washington, DC, http://go.worldbank.org gressivity and efficiency of taxes and trans- /3SLYUTVY00; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. fers. Expanding the access to basic services— Note: The numbers presented here are based on a regional data thereby closing the remaining gaps—and harmonization effort that increases cross-country comparability improving the quality of these services con- and may differ from official statistics reported by governments and national statistical offices. The welfare indicator used to stitute an immediate priority in the effort to compute the Gini is total household per capita consumption. maintain previous successes. The Gini ranges from 0 (perfect equality) to 100 (perfect inequality). Cambodia: new earning opportunities emerging from impressive growth reduction in inequality was evident in both urban and rural settings, although at a dif- Inequality reduction ferent pace: more rapid in urban areas other than Phnom Penh and in rural areas and Cambodia has made significant progress slower in Phnom Penh.23 in countering inequality over much of the Consumption has grown more rap- first decade of the 2000s. This is the result idly among less well off Cambodians than of large segments of Cambodian society, in- among the more well off. Annual consump- cluding rural populations, benefiting from tion growth among the bottom 40 averaged the country’s impressive economic growth. 6.3 percent between 2008 and 2013, well Annual economic growth averaged 7.8 per- above the 3.7 percent average consumption cent between 2004 and 2014, ranking among growth across the population and twice the the top 15 most rapidly growing economies rate among the top 60. During the last two in the world. GDP per capita increased years on which data are available, that is, fourfold, from US$253 in 1993 to around 2012 and 2013, the growth of consumption US$1,090 in 2014, and Cambodia is pro- among the poorest households surpassed jected to become a lower-middle-income that of richer households significantly country by 2017.22 Poor Cambodians have (figure 5.5). Indeed, consumption growth harnessed the opportunities being made among the richest households—the top 33 available to them by economic growth. As a percent—was below average during these result, the Gini index fell appreciably, from two years. This was so everywhere except 37 in 2007 to 26 in 2013, the latest year for in Phnom Penh, where the growth of the which data are available (figure 5.4). This poorest 20 percent was negative.24

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 107 FIGURE 5.5 Growth Incidence Curve, FIGURE 5.6 Contributions of Growth Cambodia, 2012–13 and Redistribution Effects to Poverty Reduction, Cambodia, 2008–12 15 US$1.90 (2011 PPP) 0 –1 –2 10 –3 –4 –7.9 –5 –6 –7 –8 5 –9

Annual growth rate (%) –10 –11 –12 –7.9 –13 0 –14 020406080 100 –15 –16 Consumption percentile –17

Annualized growth of percentile Change in poverty headcount (percentage points) –18 Average growth –19 –20 Source: “Measuring Inequality,” World Bank, Washington, Growth Redistribution DC, http://go.worldbank.org/3SLYUTVY00; PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch Source: World Bank calculations. .worldbank.org/PovcalNet/. Note: The data have been calculated using a Shapley approach (Shorrocks 2013) that decomposes the change in the US$1.90- a-day poverty rate (2011 PPP U.S. dollars) into two components: the growth in household per capita income, keeping distribution Poverty reduction likewise followed an constant (the growth effect), and the change in the distribution impressive path during the last decade, an of income, keeping average income constant (the redistribution achievement few could have foreseen upon effect), as developed by Datt and Ravallion (1992). the country’s emergence in the 1990s from a past of conflict. The proportion of the population living below the global poverty policy making, have generally gone hand in line declined significantly, from 21 percent hand with a favorable regional and global to 5 percent, between 2008 and 2012.25 The environment. In the recent past, the govern- rural poverty rate dropped precipitously, ment has maintained the credibility of the coming close to the rates in the capital macroeconomic policy making that it had and other urban areas. According to World built up prior to the global financial crisis Bank estimates, the average growth in con- and recession in 2008–09. It has adopted a sumption across households is responsible series of countercyclical measures meant for about half the poverty reduction (the to offset the negative effects of downturns. growth effect) that took place in Cambodia Also, it has gradually moved to restrain between 2008 and 2012 (figure 5.6). The fiscal deficits by mobilizing additional rev- other half is explained by the unequal in- enue and limiting current and capital ex- crease in consumption across households, penditures. After peaking at 25 percent in that is, by a consumption growth pattern 2008, inflation was modest and even fell that favored the less well off (the redistribu- to roughly 2 percent throughout 2014 and tion effect). 2015 as a result of low food and oil prices. In addition, official development assistance What is behind the reductions to finance major infrastructure projects has in inequality? helped keep the debt under control that might otherwise have soared.26 In Cambodia, stable domestic macroeco- The country’s brisk growth since the nomic conditions, supported by prudent middle years of the first decade of the 2000s

108 POVERTY AND SHARED PROSPERITY 2016 has been largely driven by garment and than doubled between 2004 and 2009, in apparel exports, tourism, real estate, and large part because of rising prices and, to a construction amid a transition away from lesser extent, because of expanding produc- agriculture.27 A proliferation of employ- tivity in the smaller agricultural labor force. ment opportunities, particularly in salaried The government has supported the sector or wage positions, has followed the expan- by, for instance, increasing investments in sion of these sectors. Paid employees consti- rural roads and working to improve seed tuted 41 percent of the workforce in 2013, distribution, strengthen oversight on the up from 22.6 percent in 2004, in large part quality and price of imported fertilizers, and because of the availability of more of these support the operation and maintenance of jobs in the capital.28 The garment and ap- irrigation schemes.30 The abandonment of parel industry contributed to this observed commodity price controls and related tax increase in employment, especially among levies arguably had a larger role in helping the poor. Now accounting for fully 80 per- the poor.31 cent of Cambodia’s total exports, garments Increasing nonfarm self-employment and related apparel employ the bulk of the and rising wages have also been key features country’s manufacturing labor force. Be- of the strides in reducing rural inequality. cause production in the sector is heavily The trend among the rural poor to diversify oriented toward exports and because of the household incomes away from agriculture prevalence of foreign direct investment in has been rising recently. More well paid em- the sector, wages are higher and more sta- ployment supported the well-being of rural ble in garment jobs relative to other jobs. households with little or no landholdings New research indicates that garment jobs during the recent agricultural commodity improve the well-being of the bottom 40 price boom.32 Average per capita incomes insofar as these households are more in- from daily wage labor among rural resi- clined to experience consumption gains dents rose 9.5 percent annually in 2004–09, from participating in this sector, suffer less and, as of 2013, wages and salaries repre- from food insufficiency, and report higher sented 43 percent of total rural household school enrollment rates. There have also income.33 been increased agricultural investments by Notwithstanding the positive effects of rural households that are receiving remit- the nonagricultural sectors of the economy tances from nonresident members working on employment and labor incomes, job in the garment industry. This is a reflection growth in these sectors has been insuffi- of the fact that this sector predominantly cient to absorb new labor market entrants, employs young women migrants from rural reflecting in part the country’s young demo- areas.29 Furthermore, the garment sector is graphic. The seemingly sagging competi- characterized by a much lower gender wage tiveness of the garment industry since 2012, gap than other sectors and is an import- reflected in the greater real wage growth rel- ant reason why Cambodia has been able ative to productivity growth, is also grounds to incorporate women into the productive for concern.34 Several long-standing barriers economy. The services sector—encompass- constraining productivity and investments ing communication, hospitality, trade, and in agriculture are particularly detrimental to transport—has also improved the employ- smallholder farmers. These revolve around ment prospects of the poor. problems in land tenure, bottlenecks in fer- Although its aggregate economic im- tilizer and seed markets, inadequate exten- portance has been declining, agriculture sion services and irrigation systems, and, has also contributed to the narrowing in among farmers, the lack of savings and ac- inequality in rural areas over the last sev- cess to credit.35 As a result, smallholders are eral years. The fall in inequality and poverty generally vulnerable to swings in the inter- in the midst of the global financial crisis is national prices for rice. partly explained by the agricultural sector’s Upgraded infrastructure and social pro- vitality at that time. Farm incomes from ag- tection are urgent in light of the constraints. ricultural crops, mainly paddy rice, more Enhanced capacity in electricity generation

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 109 and distribution, transport (especially rural narrowing in inequality and reductions of roads), irrigation, and information and poverty are evident in the slowing pace of communication technology are often cited job creation, which is troublesome, espe- as areas of infrastructure in need of expan- cially given Cambodia’s young demographic sion or upgrading. In social protection, the and the structural constraints that weigh on government has strived to extend coverage leading sectors. Scaling up policy efforts beyond formal employment–based schemes and investments in infrastructure and rural by scaling up the national household tar- development, together with social protec- geting system and strengthening relevant tion, would help eliminate bottlenecks and governance arrangements in an attempt to sustain the success already achieved. safeguard the well-being of the informally employed. Chief among those efforts is the Mali: a vulnerable economy launching of a health equity fund to extend favored by the vagaries of the coverage of health care (see chapter 6). agriculture As of 2013, the funds covered more than 2.5 million people in 51 of Cambodia’s 81 Inequality reduction districts. More recently, the country is undertaking a pilot cash transfer project fo- In 2014, Mali’s gross national income per cused on mother and child health and nu- person was US$650, placing the coun- trition. However, there are major problems try among the 15 poorest countries in the in coverage and the size of benefits in so- world.38 According to the human devel- cial assistance programs. Only 2 percent of opment index, which embodies life ex- the poorest 20 percent of Cambodians re- pectancy, education, and income indica- ceive any form of assistance through social tors, the country ranks 179th among 188 safety nets, and the transfers they receive countries.39 About half the population is only total 2 percent of poverty-threshold living in extreme poverty according to the expenditures per intended beneficiary.36 US$1.90 per person per day line; two-thirds This adds to the insufficient levels of pub- of the adult population is illiterate; and life lic spending on health care and education, expectancy at birth is 58 years.40 The econ- which, though rising recently, are rather low omy is undiversified, and the population by international comparison, at approxi- largely depends on rainfed agriculture. The mately 4 percent of GDP combined.37 country is thus highly vulnerable to changes In sum, inequality in Cambodia has in international prices and weather-related declined steadily in the midst of sustained shocks. The conflict and security crisis that and rapid economic growth and the ex- originated in the country’s northern region panding dynamism of sectors that employ in 2012 has added to the current challenges large numbers of relatively less well skilled involved in reducing poverty and boosting workers, of which the country has an abun- shared prosperity. dance. Women in the bottom 40 have lev- The resurgence of conflict put an end eraged their occupations in the production to two decades of political stability. Both of garments, Cambodia’s primary export, to the stability and its later demise are attrib- improve the well-being of their households. utable to an unsustainable practice of the The service sector, the growth of which is constant redistribution of positions and linked to increasing internal demand and resources among a small elite, tendencies tourism, has also facilitated the integration to co-opt any potential opposition through of greater numbers of Cambodians into the payoffs, corruption, nepotism, and impu- labor market. In rural areas, higher agricul- nity. The approach, known as the politics tural wages, growing off-farm employment, of consensus, has undermined the creation and earnings from rice farming have all of a professional bureaucracy, indepen- had an appreciable impact on inequality. dent oversight institutions, and a proper Indeed, rural areas have largely driven the judiciary and weakened the army.41 There- overall trends in inequality and poverty fore, the reduction of inequality took place reduction. Even so, obstacles to additional during a period of protracted flawed gov-

110 POVERTY AND SHARED PROSPERITY 2016 FIGURE 5.7 Trends in the Gini Index, ant dimensions such as nutrition and asset Mali, 2001–10 ownership. For instance, the prevalence of stunting among under-5-year-olds declined 40 39.9 from 43 percent in 2001 to 28 percent in 38.9 2010. In addition, mobile phone ownership surged from nearly zero to 67 percent; household ownership of refrigerators doubled, from 5 percent to 11 percent, while television ownership almost tripled, to 37 percent, in the same period.44 35 The progress in reducing poverty and

Gini index inequality reﬂects the more rapid consump- 33.0 tion growth of the poorest households relative to the consumption growth of richer households (ﬁgure 5.8, panel a). The poorest 10 percent showed the largest increases in consumption growth during the period, 30 and households in the top 25 percent ac- 2001 2007 2011 tually reduced their consumption between

Source: PovcalNet (online analysis tool), World Bank, Wash- 2001 and 2010. Decomposition exercises ington, DC, http://iresearch.worldbank.org/PovcalNet/. that estimate the contribution of growth Note: The numbers presented here are based on a regional data and redistribution effects to the change harmonization effort that increases cross-country comparability and may differ from official statistics reported by governments in poverty indicate that about 82 percent and national statistical offices. The welfare indicator used to of the poverty decline observed between compute the Gini is total household per capita consumption. 2001 and 2010 originated in changes in the The Gini ranges from 0 (perfect equality) to 100 (perfect inequality). distribution of consumption across households (the redistribution effect), while the remaining 18 percent arose because of the ernance, which may have affected the pace effect of growth, that is, the average increase of inequality reduction, but did not prevent in consumption experienced by Malian it. Meanwhile, the narrowing of inequality households.45 during the past decade was not sufficient to Behind this equalizing pattern, there avert conflict. were important differences across the first Before the outbreak of conflict, Mali had decade of the 2000s. During the first part of made important strides in reducing poverty the decade, the gains in per capita consump- and improving well-being. Between 2001 tion were lower in magnitude, but more and 2010, GDP growth averaged 5.7 percent widespread across households (figure 5.8, a year, while GDP per capita growth aver- panel b). This allowed more households to aged 1.5 percent.42 During the period, the exit poverty during this period with respect growth in the economy was largely inclusive to the second half: indeed, about 85 percent and translated into substantial poverty and of the poverty decline that occurred during inequality declines. Thus, the reduction in the decade took place in the first five years. inequality was sizable, underscored by a 7 Consumption growth among the poor and point decrease in the Gini index between nonpoor was similar during the period, 2001 and 2010 (figure 5.7).43 The share of which meant that, paradoxically, despite the Malians living on less than US$1.90 a day manifest poverty reduction, there was not a fell from 57.9 percent to 49.3 percent, al- substantive narrowing in inequality. In con- though the decline was not sufficient to trast, during the second part of the decade, reduce the number of the poor, which in- the poorest households saw, for reasons creased to about 360,000 because of the explained below, their consumption grow country’s high population growth. The im- more quickly than the consumption of richer provements in monetary well-being were households, which, in fact, declined during also accompanied by gains in other import- the period (figure 5.8, panel c). Because of

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 111 FIGURE 5.8 Growth Incidence Curve, Mali, 2001–10 FIGURE 5.9 Contributions of Growth and Redistribution Effects to Poverty a. 2001–10 5 Reduction, Mali, 2006–09

4 Poverty $1.90 (2011 PPP) 0.5 3 0.4 2 0

1 –0.5 0 Annual growth rate (%) –1.5 –1 –1

–2 2 1426385062748698 –1.5 Consumption percentile Change in poverty headcount (percentage points) b. 2001–06 –2 3 Growth Redistribution

Source: World Bank calculations. 3 Note: The data have been calculated using a Shapley approach (Shorrocks 2013) that decomposes the change in the US$1.90- a-day poverty rate (2011 PPP U.S. dollars) into two components: 2 the growth in household per capita income, keeping distribution constant (the growth effect), and the change in the distribution 2 of income, keeping average income constant (the redistribution effect), as developed by Datt and Ravallion (1992). 1 Annual growth rate (%) 1 these distinctive changes in consumption across households, inequality declined (a 0 5.9 point reduction in the Gini index), while 2 1426385062748698the poverty reduction in the second half of Consumption percentile the decade was meager. Indeed, a decomposition of the changes in poverty in 2006–09 c. 2001–10 reveals that the paltry reduction of poverty 12 in this period is linked to large distributional 10 changes (the redistribution effect) that were 8 partially offset by a growth effect in the op- 6 posite direction (ﬁgure 5.9). 4 2 What is behind the reductions 0 in inequality? –2 During the 1980s and 1990s, the govern- –4 Annual growth rate (%) ment of Mali implemented a series of –6 structural reforms aimed at shifting from –8 a state-controlled economy toward a more –10 market-oriented system. The reforms in- 2 1426385062748698cluded price and trade liberalization, tax Consumption percentile reform, and legal and regulatory reforms. Annualized growth of percentile Average growth In addition, the transition to democracy Source: Tabulations based on SSAPOV database harmonization, Sub-Saharan Africa Team for Statistical that had started in the 1990s was followed Development, World Bank, Washington, DC. by two decades of relative political stability

112 POVERTY AND SHARED PROSPERITY 2016 that allowed the expansion of access to basic Mauritania that depended on income from services and improvements in development relatives working abroad, mainly in France.46 outcomes. The distinct patterns of inequality and Relative to previous decades, the 2000s poverty reduction in each half of the first benefited from less volatility in GDP growth. decade of the 2000s described above depend One contributor to the reduced fluctua- on the changing role of nonagricultural sections was the expansion in cereal produc- tors as contributors to growth. In 2001–06, tion, explained by the liberalization of pro- the industry and service sectors and the ducer prices, more open cereal markets, agricultural sector grew at a similar pace. and persistently good weather in the second Hence, the growth of household consump- half of the decade. Improvements in cereal tion was fairly widespread so that a vast ma- production were also supported by water jority of Malians benefited from growth. In management and irrigation policies and by contrast, during 2006–10, manufacturing sustained input subsidies. Subsidies helped contracted, while agricultural production, expand access and predictability in the sup- favored by good weather, boomed. The dy- ply of fertilizers, and this contributed to the namics of the expansion of the agricultural expansion of maize and rice production sector did not translate into economy-wide over the decade. gains, but helped raise the consumption of Because of a surge in its growth, agri- the poorest households, which were mostly culture has been a key driver behind the working in agriculture. The boom in agri- improvement in living conditions among culture thus improved the well-being of the poor Malians. Approximately 73 percent of extreme poor relative to the rest of the pop- Malians and 90 percent of the poor live in ulation, thereby narrowing inequality.47 rural areas. Most of the poor are net buyers Public spending has been aligned with of food. Among those involved in farming the government’s poverty reduction strat- activities, own-account production does egy and focused on agriculture and educa- not permit self-sufficiency, and income has tion. During the decade, there were import- to be supplemented with casual labor and ant achievements in education and in access private transfers. The increased production to basic services. For instance, between in cereals benefited the labor income of the 2001 and 2011, the gross enrollment ratio poor by raising both farm production and in primary education rose by about 28 per- off-farm labor income through a higher de- centage points, reaching about 80 percent. mand for wage labor by commercial cereal Electricity coverage expanded from 8 per- producers. The increased supply of cereals cent to 34 percent in these same years. In also helped lower local cereal prices and addition, access to clean drinking water in- enhance the purchasing power of the poor. creased from 69 percent to 81 percent. Yet, Gains in household consumption during the coverage and quality of basic services the last decade were visible in regions that still lagged significantly in rural areas. Pub- grow rice (the most important cash crop) lic spending in Mali is sensitive to economic and maize and sorghum (two of the main shocks as might be expected in an economy consumption crops). highly exposed to weather shocks. This is Another contributor to the higher off- compounded by the fact that infrastructure farm incomes of the poor was the expansion investments in health care and education in artisanal gold mining in some regions of largely depend on donor funding. Target- the country. In addition, the region border- ing remains a challenge. Electricity and fuel ing Senegal benefited from the redirection subsidies mostly benefit the urban nonpoor, of trade following the 2002 crisis in Côte while fertilizer subsidies support the more d’Ivoire and the improved infrastructure well off farmers. In 2010, only 3 percent connecting Bamako, the Malian capital, and of households in the poorest consumption Senegal. Another contributing factor was quintile had access to electricity, compared remittances. Between 2001 and 2010, there with 62 percent of the households in the was a threefold rise in per capita remittances richest quintile. It is also estimated that the among households in border areas with top 20 benefited from about three-quarters

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 113 of the resources allocated through the pub- among the 15 highest rates in the world. lic education system.48 High international mineral prices; rising The redistributive role of social trans- incomes, mainly in labor-intensive services, fers has been limited. Most of the country’s commerce, and agriculture; increasing pub- safety nets are concentrated on food support lic transfers; and prudent fiscal policies all through subsidized prices and price stabili- contributed to improving the distribution zation. Cash transfers are implemented at of incomes, especially at the bottom of the a limited scale, and labor-intensive public distribution. Poverty reduction has been re- works programs have been introduced in remarkable, from poverty rates of 59 percent sponse to crises. In 2009, around 0.5 percent in 2004 to 23 percent in 2014 (measured of GDP was spent on social safety nets, a low using the national poverty line). Extreme amount considering that about one person poverty also fell by a significant margin, in two is counted among the extreme poor, some 12 percentage points, to 4.3 percent and one person in four is food insecure.49 in 2014. This implies that about 9 million Overall, the progress in reducing in- Peruvians—30 percent of the population— equality and poverty in Mali in recent years exited from poverty, while 3.2 million es- has been remarkable. In part, this is the re- caped extreme poverty (all in net terms). sult of deliberate policies and political sta- The share of the population living on less bility and, in part, to good luck in the form than US$1.90 a day (2011 PPP) fell from 12 of favorable weather and positive external percent to 3 percent during the period. events that have offset negative shocks. The Gini index fell from 51 in 2004 to However, there are signs of a recent stall in 44 in 2014, more impressive than the re- rainfall trends and a progressive increase in duction observed in the Latin America and temperatures (by 0.8° Celsius since 1975) as Caribbean region as a whole (figure 5.10). a consequence of global warming that will However, income inequality has not nar- keep the economy highly exposed to cli- rowed at the same pace in urban and rural mate hazards.50 Moreover, since 2012, the areas, with more limited declines in rural conflict in the north of the country has put settings. Inequality in rural areas, tradi- the brakes on the progress of the previous tionally at lower levels than in urban areas, decade. The crisis has disrupted education surpassed urban income inequality in 2008 and health care services in the north, and and has remained wider ever since.51 Illus- displaced populations are exerting pressure trative is the fact that the incidence of stunt- on service delivery in the south. Restoring ing among under-5-year-olds was 20 per- security and the realization of the peace centage points higher in rural areas than in agreement are necessary steps to supporting urban areas despite substantial reductions economic growth and sustaining the im- in stunting overall. provement of living conditions, particularly The bottom 40 experienced substantially among those affected by the conflict. Cou- larger income growth than the top 60. Be- pled with this priority, enhanced service de- tween 2004 and 2014, the growth rate of livery is still a challenge in sharing prosper- real income per capita among the bottom ity and reducing poverty. 40 was 6.5 percent, compared to 4.1 percent among the overall population.52 The growth Peru: equalizing investment incidence curve shows that the largest in- despite informality and low come growth between 2004 and 2014 took human capital place among the poorest households. It also shows that the annual income growth rates Inequality reduction were highest among the poorest 25 percent of households and declined among more Between 2001 and 2014, Peru’s unprece- well off households. Households in the dented growth led to an increase in real per bottom 25 experienced annualized growth capita income by a factor of almost two. rates of about 7 percent between 2004 and The economy expanded during the period 2013, while households in the top 10 saw at an average rate of 5.5 percent a year, increases of 3 percent or less (figure 5.11).

114 POVERTY AND SHARED PROSPERITY 2016 FIGURE 5.10 Trends in the Gini Index, FIGURE 5.12 Contributions of Growth Peru, 2004–14 and Redistribution Effects to Poverty 56 Reduction, Peru, 2004–14 54 Poverty $1.90 (2011 PPP) 0 52 –1 50 –2 48 –3 –5.5

Gini index 46 –4 44 –5 42 –6 40 –7 2004 2006 2008 2010 2012 2014 –3.5 Peru Latin America and the Caribbean –8

Source: Tabulations of Equity Lab, Team for Statistical Devel- –9

opment, World Bank, Washington, DC, based on data in the Change in poverty headcount (percentage points) SEDLAC database. –10 Note: The numbers presented here are based on a regional data Growth Redistribution harmonization effort that increases cross-country comparability and may differ from official statistics reported by governments Source: World Bank calculations. and national statistical offices. The welfare indicator used to Note: The data have been calculated using a Shapley approach compute the Gini is total household per capita income. The (Shorrocks 2013) that decomposes the change in the US$1.90- Gini ranges from 0 (perfect equality) to 100 (perfect inequality). a-day poverty rate (2011 PPP U.S. dollars) into two components: The aggregate for Latin America and the Caribbean is based the growth in household per capita income, keeping distribution on 17 countries in the region on which microdata are available. constant (the growth effect), and the change in the distribution The Gini index of the Latin American and Caribbean region of income, keeping average income constant (the redistribution is computed based on pooled country-specific data that have effect), as developed by Datt and Ravallion (1992). been collapsed into 8,000 percentiles. In cases where data are unavailable for a given country in a given year, values have been interpolated by projecting incomes for that year based on GDP growth and assuming no changes in the distribution of The contributions of the growth of the incomes in the interpolated year. economy and improvements in the distribution differ considerably. The World Bank’s decomposition of the drivers of FIGURE 5.11 Growth Incidence Curve, poverty reduction suggests that the growth Peru, 2004–14 effect explains about 61 percent of the re- 8 duction in poverty, while the remaining 39 percent is explained by the improvements 7 in the distribution of income across house- 6 holds (figure 5.12).

5 What is behind the reduction in 4 inequality? 3 The remarkable improvement in the living 2 Annual growth rate (%) conditions of the poor and the bottom 40 1 results from the outstanding growth of the 0 economy in a context of macroeconomic 2010 30 40 50 60 70 80 90 100 stability, favorable external conditions, and Income percentile important structural reforms. The structural Annualized growth of percentile reforms implemented in the 1990s laid the Average growth foundations for the economic recovery observed since the early 2000s. These reforms Source: Tabulations of Equity Lab, Team for Statistical Devel- opment, World Bank, Washington, DC, based on data in the included trade and ﬁnancial liberalization, SEDLAC database. a more ﬂexible exchange rate regime, and

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 115 the privatization of state-owned enterprises. sector doubled during the period.57 Yet, re- The Central Bank was granted more auton- strictive labor regulations, particularly those omy to strengthen monetary policy, while related to dismissals and nonwage labor the National Customs and Tax Administra- costs, contributed to the persistently high tion Superintendence was created to sup- rates in informality, at around 70 percent, port fiscal policies. one of the highest rates in Latin America. In the early 2000s, prudent macroeco- This evidence for Peru confirms that a nomic policies and high commodity prices unique factor cannot explain the trends in brought foreign direct investment into Peru, labor income inequality. Recent analyses in particularly into the mining sector, boosting Peru and elsewhere in the region compel- private investment. This capital accumula- lingly show that other factors drive wage tion reached 26 percent of GDP in 2014, up inequality, although the evidence on the from 18 percent in 2000, and was the main effects is not always conclusive. These driv- driver of growth, accounting for more than ers refer, for example, to the following: the two-thirds of total growth since 2001.53 The quality of education, shifts in the returns economic growth and high urbanization to experience, the obsolescence of skills, rates also boosted the demand for services within-industry wage gaps reflecting the (such as commerce, hotels and restaurants, heterogeneity of firms, and the degree of and transport), thereby expanding employ- compliance with the minimum wage. These ment in the urban informal sector. In rural are some of the many aspects underpinning areas, agricultural productivity rose because labor inequality trends.58 of the better connectivity in the sierra re- Social spending played a limited role in gion (highlands) following roadbuilding, addressing inequality and poverty during the introduction of mobile phones, and the the decade. Recent estimates suggest that expanding industrial production of agro- public transfers are responsible for less than exports in the costa (coastal) region. Non- 10 percent of the poverty reduction during traditional agricultural exports have ex- the last decade.59 One important reason for perienced substantial growth in the last 15 this limited role is the low level of spend- years. They include asparagus, grapes, man- ing. In 2013, social protection spending was goes, paprika, and quinoa. The expansion US$375 per person, approximately a third of employment in urban informal services of Mexico’s spending and half of Colom- and higher agricultural productivity have bia’s. Indeed, in Latin America, the level improved labor incomes, which have fueled of social spending in Peru exceeded only growth by raising domestic consumption.54 the social protection spending in Bolivia The labor market was the main contrib- and Paraguay.60 Social spending was gen- utor to the reduction in poverty and in- erally progressive. The introduction of the equality in the last decade, explaining about country’s flagship CCT, Juntos (together), three-quarters of the reduction in extreme in 2005, the adoption of results-based bud- poverty and 80 percent of the reduction geting beginning in 2007, and the creation in the Gini.55 The expansion in the educa- of the Ministry of Social Inclusion in 2011 tional attainment of the labor force, which have all contributed to rendering social pro- resulted in the more equal distribution of tection more well targeted and effective. educational attainment, contributed to the Analysts question the quality of public decline in labor income inequality.56 More- spending in Peru.61 One manifest case is ed- over, the labor market was characterized by ucation. Despite significant gains in prepri- high participation rates and low unemploy- mary and secondary enrollments, Peru lags ment rates. There was a large narrowing in comparator countries in scores on the Pro- the wage gap between formal and informal gram for International Student Assessment workers, leading to reduced wage inequality. of the OECD. In 2013, 75 percent of students The median monthly wage among informal at level 5 in mathematics in Peru did not workers doubled, while formal sector wages possess the standard expected basic math- grew by 55 percent between 2004 and 2014. ematics skills of a level-2 student.62 Public The share of the employed in the formal education spending increased during the

116 POVERTY AND SHARED PROSPERITY 2016 decade, but education expenditures—2.6 The favorable performance in economic percent of GDP—are still low by interna- growth and poverty reduction was accom- tional standards. Aiming to improve the panied by narrowing inequality, although quality of secondary education, the govern- the levels in Tanzania are low relative to ment has been implementing a merit-based the Sub-Saharan African region and much system for all tenured teachers since 2012. lower than in Latin America and the Ca- The quality of human capital is crucial ribbean, the world’s most unequal region.64 because the favorable conditions that have The Gini index declined from around 39 to underpinned growth in Peru have recently 36 between 2007 and 2012 (ﬁgure 5.13).65 started to recede. The external environment Evidence on the shared prosperity indica- is shifting after a sharp decline in com- tor suggests that inequality reductions were modity prices over the past few years. This mainly driven by a larger increase in the con- spotlights the need to boost the aggregate sumption accruing to the bottom quintiles. productivity of the economy to expand the The annual consumption growth among the demand for well-paid formal sector jobs. bottom 40, at 3.4 percent, was well above In sum, the reduction in poverty and in- three times the growth among the top 60, equality in Peru in the last decade took place at 1.0 percent. The growth incidence curve because of more favorable external condi- between 2007 and 2012 shows that the larg- tions, increased internal demand, improved est relative increase in consumption took educational attainment, advances in rural place among the poorest 20 percent, while connectivity, and the performance of la- growth was more moderate among other bor-intensive sectors, leading to higher, more middle-income groups and negative among equal labor incomes. Educational upgrad- the top 15 (ﬁgure 5.14). However, because ing was equalizing, though large challenges this growth took place from a low base, remain in terms of quality. To a much lesser the absolute gains were modest: it translated extent, safety nets contributed to the shrinking in inequality. Maintaining these impressive gains in a much less favorable environ- FIGURE 5.13 Trends in the Gini Index, ment will require that the Peruvian economy Tanzania, 2001–12 sustain policy reforms addressing the main 40 structural weaknesses, the limited productiv- 38.8 ity resulting from the low quality of human 38.5 capital, and the high rates of informality.

Tanzania: sharing prosperity in the midst of 35.8 diversification 35 Gini index Inequality reduction Tanzania maintained robust and stable economic growth between 2004 and 2014, averaging 6.5 percent a year. After a long period of stagnation, the poverty headcount declined from 34.4 percent in 2007 to 28.2 30 2001 2007 2012 percent in 2012, while extreme poverty fell from 11.7 percent to 9.7 percent.63 The re- Source: PovcalNet (online analysis tool), World Bank, Wash- duction in poverty appears more substan- ington, DC, http://iresearch.worldbank.org/PovcalNet/. Note: The numbers presented here are based on a regional data tial if one uses the international poverty line harmonization effort that increases cross-country comparability of US$1.90 per person per day. Based on and may differ from official statistics reported by governments this measure, the headcount ratio dropped and national statistical offices. The welfare indicator used to compute the Gini is total household per capita consumption. from 59.9 percent to 48.8 percent between The Gini ranges from 0 (perfect equality) to 100 (perfect 2007 and 2012. inequality).

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 117 FIGURE 5.14 Growth Incidence Curve, FIGURE 5.15 Contributions of Growth Tanzania, 2007–12 and Redistribution Effects to Poverty 10 Reduction, Tanzania, 2007–12 Poverty $1.00 (2005 PPP) 8 0

6 –1 –2.5

4 –2

2 –3

0 Annual growth rate (%) –4 –3.7 –2 –5 –4 0 20406080100 –6 Comsumption percentile Annualized growth of percentile Change in poverty headcount (percentage points) –7 Growth Redistribution Average growth Source: World Bank 2015c using the 2007 and 2011/12 house- Source: World Bank 2015c based on data of the 2007 and hold budget surveys. 2011/12 household budget surveys. Note: The data have been calculated using a Shapley approach (Shorrocks 2013) that decomposes the change in the poverty rate using a national poverty line of T Sh 36,482 per adult per to an additional consumption value of only month, equivalent to a US$1.00-a-day poverty rate (2005 PPP T Sh 4,300 per adult per month among U.S. dollars), into two components: the growth in household per capita income, keeping distribution constant (the growth effect), the poorest quintile, equivalent to approxi- and the change in the distribution of income, keeping average mately 10 percent of the cost of basic con- income constant (the redistribution effect), as developed by Datt sumption needs (less than US$3.00). and Ravallion (1992). A decomposition of the reduction in poverty at the national level between an increase in mean household consumption accelerated economic growth, benefits that (the growth effect) and changes in the dis- continued into the 2000s.67 The surge in tribution of household consumption (the growth in the early 2000s is partly attribut- redistribution effect) between 2007 and able to the pursuit of reforms through the 2012 suggests that distributional changes 1990s that sought to increase exports, use are more important than growth changes scarce foreign exchange more efficiently, in explaining the reduction.66 Household reduce the involvement of the public sec- consumption growth contributed around tor in commercial activities, and liberalize 40 percent (or 2.5 percentage points) of the market for goods and services.68 Steady Tanzania’s poverty reduction, while the nar- economic growth was also based on pru- rowing in inequality contributed 60 percent dent macroeconomic policies to control (3.7 percentage points) (figure 5.15). inflation and restore stability in exchange rates: achievements that were nonetheless What is behind the reductions challenged by, for example, a high inflation in inequality? episode in 2011–12. In addition, the transition from a socialist to a market-based Since independence in 1962, the govern- economy gave a further, significant boost to ment of Tanzania has undertaken a grad- domestic production. However, this transi- ual and partial transition from a state-led tion to a market-based economy is not fully development strategy to a market-based completed: the private sector is still charac- economy. Following the debt crisis in 1986, terized by high levels of market concentra- it undertook structural reforms that opened tion and informality and low levels of pro- up the economy to foreign investors and ductivity, mainly because of weak market

118 POVERTY AND SHARED PROSPERITY 2016 institutions and a weak business environ- from agriculture toward wage employment ment. The country is still facing persistent opportunities amid a process of urbaniza- state control over the market, particularly in tion within the country. agriculture.69 In addition, advances in the ownership The economic expansion has been driven of land and livestock, more-intensive activ- since the early 2000s primarily by rapidly ity among agricultural enterprises, and ris- growing sectors, particularly communica- ing nonfarm employment have contributed tion, financial services, and construction. to the better living conditions among the However, growth in these sectors did not poorest. From 2007 to 2012, the enhance- translate into substantive improvements in ments in household endowments were more the living conditions of the poor, the less significant among the bottom 40. The same well educated, or rural residents relative to is true of human development outcomes in the more highly skilled residing in urban the prevalence of malnutrition, infant and areas, notably, Dar es Salaam. This explains, maternal mortality, the female literacy rate, in part, a more limited ability of economic secondary-school enrollments, and access growth to reduce poverty during the early to clean drinking water. Nonetheless, much years of the first decade of the 2000s. more needs to be done to reduce regional Thus, in 2001–07, consumption growth was disparities and to expand the access to basic greater among the top 60 than the bottom services; coverage rates are still especially 40, and growth rates were negative among low in the supply of electricity and in im- the poorest 10 percent of the population. proved sanitation services. Since 2007, there has been a surge in basic Although rigorous evaluations have not metal industries, such as construction ma- yet been conducted to confirm the impact, terials, and in retail trade and manufactur- the explicit commitment of the government ing, particularly agroprocessing in food, to adopt policies unambiguously aimed at beverage, and tobacco products. The expan- reducing poverty is consistent with render- sion of these sectors has allowed greater in- ing the distribution of income more equi- clusion in the economy among low-skilled table. This commitment has been realized workers, thereby boosting the incomes of through two programs: (1) MKUKUTA II, the poorer segments of the population. a national strategy for growth and poverty Despite the increasing importance of reduction designed to boost the access of manufacturing and services and despite the the poor to basic services, including health plans for the future, the economy has his- care, primary education, and water and san- torically depended heavily on agriculture, itation, and to mobilize financial resources, which accounts for around 33 percent of and (2) the Tanzania Social Action Fund, GDP, provides approximately 47 percent which encompasses a CCT program, public of exports, and employs about 70 percent works, and a community savings compo- of the workforce. Over two-thirds of the nent. Although expected to be rolled out population lives in rural areas, and an esti- among up to 275,000 households, Tanza- mated 98 percent of rural households en- nia’s nascent CCT program is reportedly gage in agricultural activities. The sector aimed at targeting the poor effectively.71 thus plays a key role in supplying food and The poorest households use resources from generating incomes among the rural popu- this program to increase savings and invest lation. It has also proven highly strategic in in livestock, an indication that the pro- the industrialization process through the in- gram helps households reduce risk and im- puts that it generates for the industrial sec- prove livelihoods rather than merely raise tor (65 percent) and the food it produces for consumption.72 the urban and industrial labor force.70 Agri- From a fiscal point of view, Tanzania culture has also been subject to greater di- tends to redistribute more than anticipated versification toward higher-value cash crops based on its relatively low income level. (cotton, cashews, tea, and coffee), along with Nonetheless, the fiscal system contributes an increase in land productivity, in part be- little to reducing inequality. This is largely cause poor households have been shifting because several components of the fiscal

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 119 system impact the distribution of incomes icant reductions in inequality under partic- in different directions. Direct taxes, in-kind ular circumstances. Despite the small num- benefits in education (except in tertiary ber of experiences reviewed in this chapter, education), and cash transfers are progres- there is no dearth of countries that have sive. Some favorable features among other- recently narrowed income inequality. The wise regressive fiscal tools also contribute bottom 40 enjoyed income or consumption to improving the equalizing capacity of the growth rates exceeding the income growth fiscal system. For example, indirect taxes of the mean between 2008 and 2013 in 49 are more progressive in Tanzania than in countries. In recent years, reducing inequal- many Sub-Saharan African countries; bev- ity is thus not a rare feat even in challeng- erage and communication excise taxes are ing contexts marked by international crises, progressive; and fertilizer subsidies are close local instability and conflict, and unfavor- to neutral.73 Yet, electricity subsidies and able initial conditions such as substantial tertiary education are regressive, and other poverty and inequality. programs that are progressive in many The countries reviewed, without excep- countries, such as health care and food as- tion, have made serious commitments to sistance, are only neutral in Tanzania. establishing a stable macroeconomic envi- In sum, Tanzania has achieved a signif- ronment and reducing inflation. They have icant improvement in the reduction of in- also benefitted from favorable global con- come inequality and enhancement in the ditions in terms of low interest rates, high living conditions of the bottom quintiles of commodity prices, and booming interna- the population despite low income levels. tional trade. In some cases, good macro- The increasing diversification of the econ- economic policy has been associated with omy into labor-intensive manufacturing the appropriate management of an external and commerce and greater agricultural pro- commodity boom, as in Brazil and Peru. In ductivity help explain part of this success. other cases, as in Mali and Tanzania, trans- The diversification is being realized within formations away from an outdated eco- a framework of macroeconomic stability nomic model have been implemented be- and decades of liberalization, although the fore prudent macroeconomic management private sector remains unproductive and emerged. That sound macroeconomic pol- uncompetitive. The new Five Year Devel- icy is good for equality is hardly surprising opment Plan 2016–21 is designed to ac- given the comprehensive evidence linking celerate structural transformation to turn such policy with poverty reduction.74 Our Tanzania into a semi-industrialized econ- small but diverse sample of countries sup- omy, increasingly urbanized, with room for ports this conclusion across different spe- further rural diversification and secondary cific contexts and challenges. town development to promote the reduc- The contribution of economic growth to tion of regional disparities within the coun- shared prosperity and inequality reduction try. Strategic investments in infrastructure is critical. All country cases report vigorous and basic services such as electricity and and, in some cases, such as Cambodia, Peru, improved sanitation are also expected to and Tanzania, impressive average growth promote economic growth and shared rates during the period of analysis. The ev- prosperity. idence in each of these countries confirms that rising average income or consumption Concluding remarks: explains a great deal of the improvement in learning from country the living conditions of the poor. More than 60 percent of the decline in poverty in Bra- experiences zil and Peru has been attributed to a growth A handful of country experiences cannot be effect, that is, a rise in average household representative at a global scale or sufficient incomes, while a more moderate 20 percent to provide policy prescriptions valid around contribution is found in Mali (although the world. They are, however, illustrative of growth there is found to contribute to rais- the practices and conditions behind signif- ing poverty).

120 POVERTY AND SHARED PROSPERITY 2016 The functioning labor markets have economy. The economic gains generated been instrumental in all the countries under by the sudden reorientation of trade flows analysis in helping translate economic following conflict in a neighboring country growth and sound macro economic policies largely fell outside the responsibility of Ma- into a narrowing in inequality and an in- lian policy makers. However, strengthening crease in shared prosperity. The experiences the country’s fiscal position, preventing a of these countries show that labor markets climb in inflation, promoting productive can help reduce inequalities by expanding investments and economic diversification, job opportunities across emerging mar- weaving together more effective safety nets, kets, as in Cambodia, or traditional sectors, combating corruption and elite capture fol- as in Mali and Tanzania; sharing earning lowing such lucky gains do fall under the di- opportunities with individuals previously rect responsibility of policy makers. Coun- excluded from growth, such as the low- tries more willing to undertake such policy skilled and women in Tanzania and Cam- choices are also those more likely to sustain bodia; and narrowing earnings gaps across inequality reductions. Those that do not workers, such as between men and women make these choices will experience short- in Cambodia, between economic sectors in lived and fragile reductions in inequality. Mali, between formal and informal jobs in Thus, the marked differences in the most re- Brazil and Peru, and between urban and cent policy choices between Brazil and Peru rural residents in Peru. on fiscal consolidation, the control of in- These five successful cases also highlight flation, and investments largely explain the the importance of aligning macroeconomic stark differences in the most recent growth and fiscal policies with sound sector-specific patterns in these countries (gradual recov- interventions and public investments. Thus, ery in Peru, recession in Brazil), and the out- improving the coverage and quality of ed- look for poverty and inequality (positive in ucation, expanding the coverage of public Peru, pessimistic in Brazil).75 health care, enhancing market connectivity, Beyond the constants of good macroeco- and establishing or scaling up safety nets to nomic policy, strong growth, functioning protect vulnerable populations recur in the labor markets, and coherent policies, the country narratives. Although not all such levers of inequality reduction may also vary interventions took place simultaneously in considerably across countries. In Brazil, each country, the lesson is that the coher- several policies combined led to narrowing ence of policy choices in macroeconomic inequality, including reforms in health care and sectoral dimensions is critical to en- and education, raising the minimum wage abling steady reductions in inequality. Thus, and creating formal jobs, and a remarkable appropriate policies and investments were expansion in safety nets. The role of safety identified and implemented in low-income nets has been more marked in Brazil than Cambodia, Mali, and Tanzania during pe- in any other country. In contrast, impressive riods of economic diversification and in economic growth in Cambodia has been Brazil and Peru during periods of booming linked to the emergence of garment manu- commodity prices. facturing, tourism, and construction. These This does not mean every policy decision sectors opened employment opportunities in these countries was appropriate. In fact, among the less highly skilled, thus diver- these country experiences show that good sifying the incomes of the poor away from luck—in the form of favorable external subsistence agriculture. In Mali, which is an events unexpectedly benefiting a country— agrarian economy, the expansion in cereal cannot be a substitute for policy coherence. production has been key to reductions in Positive external shocks may magnify local inequality, along with increased remittances efforts to reduce inequality. The surge in and new trading opportunities with neigh- commodity prices favored the economies boring countries. In Tanzania, diversifica- of Brazil and Peru, while the rise in the in- tion in agriculture (toward cash crops) and ternational rice prices clearly helped to raise in the rest of the economy (toward manu- the inclusive capacity of the Cambodian factures and retail commerce) explains a

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 121 good deal of the success. In Peru, rising do- try. While Brazil is marked by historically mestic demand owing to substantial foreign record inequalities, conflict has affected investment in mining has expanded the em- Cambodia and, recently, Mali. Indeed, Mali ployment opportunities in the urban infor- shows that, although inequality can be re- mal sector, while a dynamic agriculture has duced during periods of weak governance, diversified into new exports. Earnings gaps such achievements may not be sustained or narrowed between formal and informal long-lived. Rapid population growth is also workers following economic growth, as in a dominant feature in Mali, which, along the case of Brazil and Cambodia. with Tanzania, is a low-income country. In While the experiences reviewed in this Peru, the extent of informality in the econ- chapter suggest that reducing inequality can omy is among the largest in the region. This be achieved in many different ways if appro- chapter thus provides suggestive insights priate building blocks are put in place, sus- on policy interventions that, among a myr- taining this success may require decisions iad of alternatives, have worked in specific that are similar to those outlined here. Thus, contexts. the five country cases suggest that there are still common policy interventions likely to Notes sustain inequality reduction in developing countries. For example, the situation in 1. See Azevedo, Inchauste, and Sanfelice countries such as Cambodia, Mali, and Tan- (2013); Azevedo, Sanfelice, and Nguyen zania underscores the need to expand safety (2012); Barros et al. (2006); Datt and Rav- nets, which are currently not large enough allion (1992); Inchauste et al. (2014); Lustig to protect the poorest. The Brazilian experi- (2015) for methodological discussions of ence, a renowned example of safety net ex- the Commitment to Equity Initiative and pansion, shows that the entire fiscal system the decompositions. can be designed to have a greater impact in 2. Ferreira, Leite, and Litchfield (2008). reducing inequality. This is also the case in 3. Using data from the national household Tanzania. Improving competition and pro- surveys and tax records to construct a full ductivity in the private sector is also a prior- individual income distribution, Souza, Me- ity in that country. Infrastructure is report- deiros, and Castro (2015) show that the Gini edly a huge obstacle in Cambodia, while, index fell by about 1 point between 2006 in Mali, dependence on external factors, and 2012, compared with a fall of around 3 from donor flows to the vagaries of weather, points if only household per capita income threaten sustained improvement. In Peru, from the survey is used. the quality of education is below regional 4. The estimate is based on a national poverty standards and would represent an obsta- line. However, Brazil does not have an offi- cle to maintaining or improving economic cial poverty line. The R$140 per capita per productivity should more favorable exter- month moderate poverty line is derived nal conditions disappear. The accumulation from the administrative poverty lines de- of quality human capital, diversification in fined for the Bolsa Família Program and the income-earning opportunities among the Brasil Sem Miséria Program. poor, safety nets that are sufficiently large 5. World Bank (2016a). to protect the poorest, and improved infra- 6. Tabulations of Equity Lab, Team for Statis- structure to connect lagging regions to eco- tical Development, World Bank, Washing- nomic growth opportunities are all desir- ton, DC, based on household survey data, able strategies to sustain success in boosting data in the SEDLAC database, and data de- shared prosperity and narrowing inequality. rived through PovcalNet (online analysis However, caution needs to be exerted to tool), World Bank, Washington, DC, http:// turn these findings into prescriptions with iresearch.worldbank.org/PovcalNet/. The in- international validity. This is not only be- formation has been updated as of April 2016. cause of the small number of cases analyzed, 7. López-Calva and Rocha (2012); World Bank but also because of the large differences in (2016a). the contexts and challenges in each coun- 8. Ferreira, Leite, and Litchfield (2008).

122 POVERTY AND SHARED PROSPERITY 2016 9. Azevedo et al. (2013); Barros et al. (2010); percent and 27 percent, respectively, while, Ferreira, Firpo, and Messina (2014); Lustig, in Phnom Penh, the index fell by 19 percent López-Calva, and Ortiz-Juárez (2013). between 2007 and 2013. 10. World Bank (2016a). 24. World Bank (2016b). 11. López-Calva and Rocha (2012). 25. The proportion of the population living 12. The widening in formality was supported below the national poverty line declined by by institutional policies aiming to enforce nearly half, to 23.9 percent, from 2007 to labor legislation (Ferreira, Firpo, and Mes- 2009 alone and continued to decline there- sina 2014). In particular, the government ex- after, although more moderately. In rural panded the role of the Brazilian Public Pros- areas, poverty was reduced by half between ecutor’s Office (Ministério Público) and the 2007 and 2011, from 51.8 percent to 23.7 Ministry of Labor and Employment. In ad- percent. World Bank (2014a). dition, reforms to improve the efficiency of 26. IMF (2015). labor inspections were implemented in 1995. 27. Agriculture’s share in GDP was eclipsed by Corseuil, Almeida, and Carneiro (2012) pro- industry’s share for the first time in 2007, vide evidence from 1996–2006 that estab- which was also the year that investment first lishes a causal relationship between the re- surpassed the flows of official development forms and increases in formal employment. assistance to Cambodia. 13. Ferreira, Firpo, and Messina (2014). 28. ADB (2014a). 14. World Bank (2016a). 29. Mejia-Mantilla and Woldemichael (2016). 15. Equity Lab tabulations based on house- 30. In addition, in 2010, the government un- hold survey data and data in the SEDLAC veiled an official rice policy that set out to database. increase annual milled rice exports to a mil- 16. Barros et al. (2010); Osorio and Souza lion tons by 2015, a target that went unful- (2012). filled (World Bank 2009). 17. This means that the income inequality gen- 31. World Bank (2014a). erated by the functioning of the markets— 32. ADB (2014b). Average per capita income estimated at a Gini index of 55 in 2009— from nonfarm self-employment jumped 63 was reduced by 6 percent, or 3.3 percentage percent between 2004 and 2009, and daily points, after direct taxes and transfers are rural wages more than doubled between taken into account. Higgins and Pereira 2007 and 2012. (2014); Higgins et al. (2013). 33. World Bank (2014a, 2016b). 18. Clements et al. (2015); Immervoll et al. 34. IMF (2015). (2006). The 33 percent reduction is equiv- 35. Recent agricultural productivity gains from alent to a reduction of about 14 percentage an expansion in the arable land under cul- points in the average market income Gini in tivation are not sustainable indefinitely and the OECD. are thus unable to offset these problems. 19. Higgins and Pereira (2014); Higgins et al. 36. ADB (2014b); ASPIRE (Atlas of Social Pro- (2013). tection Indicators of Resilience and Equity) 20. Carneiro (2006). (database), World Bank, Washington, DC, 21. Recent analyses also include other factors in http://datatopics.worldbank.org/aspire/. the discussion of trends in labor inequality in 37. Kanbur, Rhee, and Zhuang (2014). Brazil, such as degradation of tertiary educa- 38. Using the Atlas method, in current U.S. dol- tion, obsolescence of skills, the heterogeneity lars; see WDI (World Development Indica- of firms within industries, skill-biased tech- tors) (database), World Bank, Washington, nological change, and the role of unions. The DC, http://data.worldbank.org/data-catalog interplay of all factors in final wage inequality /world-development-indicators. is complex, and there does not seem to exist 39. UNDP (2014). a strong consensus on the contribution of 40. Data for 2014, WDI (World Development these factors. See Silva and Messina (2016). Indicators) (database), World Bank, Wash- 22. World Bank (2016b). ington, DC, http://data.worldbank.org/data 23. Other urban and rural areas experienced -catalog/world-development-indicators. the greatest declines in the Gini index, by 35 41. World Bank (2015a).

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 123 42. World Bank (2015b, 100). 65. There are also issues of comparability 43. Beegle et al. (2016) have questioned the among surveys in Tanzania stemming from comparability of surveys in Mali. However, changes in the survey design and method- after a careful analysis of the robustness of ological improvements implemented during welfare estimates using the same household the 2011/12 household budget survey. The surveys, the World Bank (2013) has con- World Bank (2015c) addresses these issues cluded that the results are robust to alterna- using different methods, including the re- tive choices of poverty lines accounting for evaluation of the consumption aggregates comparability issues. for the 2007 household budget survey using 44. World Bank (2015a). the same approach as in 2011/12, as well as 45. World Bank (2015b). nonparametric and parametric imputation 46. World Bank (2015a, 2015b). procedures. The various adjustment meth- 47. Josz (2013). ods support the main results presented here. 48. World Bank (2015a, 2015b). 66. World Bank (2016d). 49. Cherrier, del Ninno, and Razmara (2011). 67. Edwards (2012). 50. World Bank (2015a). 68. Robinson, Gaertner, and Papageorgiou 51. Vakis, Rigolini, and Lucchetti (2015). (2011). 52. World Bank (2016c). 69. World Bank (2016d). 53. WDI (World Development Indicators (data- 70. World Bank (2015c). base), World Bank, Washington, DC, http:// 71. World Bank (2016e). The estimate includes data.worldbank.org/data-catalog/world beneficiaries of public works programs and -development-indicators. the expansion of cash transfers. 54. World Bank (2016c). 72. World Bank (2015c). 55. Genoni and Salazar (2015). 73. Younger, Myamba, and Mdadila (2016). 56. Jaramillo and Saavedra-Chanduvi (2011). 74. Commission on Growth and Development 57. World Bank (2016c). (2008). 58. Rodríguez-Castelán et al (2016); Silva and 75. World Bank (2016f). Messina (2016). 59. Inchauste et al. (2014). References 60. World Bank (2016c). 61. Jaramillo (2013); Lustig et al. (2012). ADB (Asian Development Bank). 2014a. Cambo- 62. Genoni and Salazar (2015). dia: Diversifying beyond Garments and Tour- 63. The poverty figures are for Tanzania main- ism, Country Diagnostic Study. November. land only and come from 2007 and 2011/12 Manila: Economics and Research Depart- Household Budget Survey data. They are ment, ADB. estimated using, respectively, the national ———. 2014b. “Cambodia: Country Poverty basic needs poverty line of T Sh 36,482 per Analysis 2014.” ADB, Manila. adult per month and the national food pov- Azevedo, João Pedro, María Eugenia Dávalos, erty line of T Sh 26,085 per adult per month. Carolina Díaz-Bonilla, Bernardo Atuesta, and See World Bank (2016d). Raul Andres Castañeda. 2013. “Fifteen Years of 64. Tanzania’s Gini index in 2012 (35.8) was Inequality in Latin America: How Have Labor lower than the 45.1 Sub-Saharan African Markets Helped?” Policy Research Working average and, among East African countries, Paper 6384, World Bank, Washington, DC. it was below that of Burundi (46.0), Kenya Azevedo, João Pedro, Gabriela Inchauste, and (47.7), Uganda (44.3), and Rwanda (50.8). Viviane Sanfelice. 2013. “Growth without It was only slightly higher than the index of Reductions in Inequality.” Background paper, Ethiopia (33.6). It was on par with levels of Poverty and Gender Unit, World Bank, Wash- inequality in South Asia and East Asia, which ington, DC. were grouped around 38.4, and significantly Azevedo, João Pedro, Viviane Sanfelice, and lower relative to indexes in Latin America, Minh Cong Nguyen. 2012. “Shapley Decom- such as Bolivia, Brazil, and Mexico, where the position by Components of a Welfare Mea- levels of inequality ranged from 47.0 to 55.2 sure.” Unpublished working paper, World (World Bank 2015c). Bank, Washington, DC.

124 POVERTY AND SHARED PROSPERITY 2016 Barros, Ricardo Paes de, Mirela De Carvalho, Fall of Brazilian Inequality: 1981–2004.” Samuel Franco, and Rosane Mendonça. Macro economic Dynamics 12 (S2): 199–230. 2006. “Uma Análise das Principais Causas da Ferreira, Francisco H. G., Sergio P. Firpo, and Queda Recente na Desigualdade de Renda Julian Messina. 2014. “A More Level Playing Brasileira.” Revista Econômica 8 (1): 117–47. Field? Explaining the Decline in Earnings In- ———. 2010. “Markets, the State, and the Dy- equality in Brazil, 1995–2012.” IRIBA Work- namics of Inequality in Brazil.” In Declin- ing Paper 12 (September), International Re- ing Inequality in Latin America: A Decade of search Initiative on Brazil and Africa, School Progress?, edited by Luis F. López-Calva and of Environment, Education, and Develop- Nora Lustig, 134–74. New York: United Na- ment, University of Manchester, Manchester, tions Development Programme; Baltimore: United Kingdom. Brookings Institution Press. Genoni, María Eugenia, and Mateo Salazar. 2015. Beegle, Kathleen, Luc Christiaensen, Andrew “Steering toward Shared Prosperity in Peru.” Dabalen, and Isis Gaddis. 2016. Poverty in a In Shared Prosperity and Poverty Eradication Rising Africa. Africa Poverty Report. Wash- in Latin America and the Caribbean, edited ington, DC: World Bank. by Louise Cord, María Eugenia Genoni, and Carneiro, Francisco Galrão. 2006. “The Effects of Carlos Rodríguez-Castelán, 269–302. Wash- the Minimum Wage on the Brazilian Labor ington, DC: World Bank. Market.” In The Third Dimension of Labor Higgins, Sean, Nora Lustig, Whitney Ruble, and Markets: Demand, Supply, and Institutions in Timothy M. Smeeding. 2013. “Comparing Brazil, edited by Francisco Galrão Carneiro, the Incidence of Taxes and Social Spend- Indermit S. Gill, and Ricardo Paes de Barros, ing in Brazil and the United States.” CEQ 113–44. New York: Nova Science Publishers. Working Paper 16, Commitment to Equity, Cherrier, Cécile, Carlo del Ninno, and Setareh Inter-American Dialogue, Washington, DC; Razmara. 2011. “Mali Social Safety Nets.” So- Center for Inter-American Policy and Re- cial Protection and Labor Discussion Paper search and Department of Economics, Tu- 1412, World Bank, Washington, DC. lane University, New Orleans. Clements, Benedict, Ruud de Mooij, Sanjeev Higgins, Sean, and Claudiney Pereira. 2014. Gupta, and Michael Keen, eds. 2015. Inequal- “The Effects of Brazil’s Taxation and Social ity and Fiscal Policy. Washington, DC: Inter- Spending on the Distribution of Household national Monetary Fund. Income.” In “Analyzing the Redistributive Corseuil, Carlos Henrique L., Rita Almeida, and Impact of Taxes and Transfers in Latin Amer- Pedro Carneiro. 2012. “Inspeção do trabalho ica,” edited by Nora Lustig, Carola Pessino, e evolução do emprego formal no Brasil.” and John Scott, special issue, Public Finance Texto para Discussão TD 1688, Institute for Review 42 (3): 346–67. Applied Economic Research, Brasília. IMF (International Monetary Fund). 2015. Commission on Growth and Development. 2008. “Cambodia: Staff Report for the 2015 Arti- The Growth Report: Strategies for Sustained cle IV Consultation.” IMF Country Report Growth and Inclusive Development. Wash- 15/307 (October 1), IMF, Washington, DC. ington, DC: World Bank. https://openknowl Immervoll, Herwig, Horacio Levy, José Ricardo edge.worldbank.org/handle/10986/6507. Nogueira, Cathal O’Donoghue, and Rozane Datt, Gaurav, and Martin Ravallion. 1992. Bezerra de Siqueira. 2009. “The Impact of “Growth and Redistribution Components of Brazil’s Tax-Beneﬁt System on Inequality and Changes in Poverty Measures: A Decomposi- Poverty.” In Poverty, Inequality, and Policy in tion with Applications to Brazil and India in Latin America, edited by Stephan Klasen and the 1980s.” Journal of Development Economics Felicitas Nowak-Lehmann, 271–301. CESifo 38 (2): 275–95. Seminar Series. Cambridge, MA: MIT Press. Edwards, Sebastian. 2012. “Is Tanzania a Success Inchauste, Gabriela, João Pedro Azevedo, B. Story? A Long Term Analysis.” NBER Work- Essama-Nssah, Sergio Olivieri, Trang Van ing Paper 17764 (January), National Bureau Nguyen, Jaime Saavedra-Chanduvi, and Her- of Economic Research, Cambridge, MA. nan Winkler. 2014. Understanding Changes in Ferreira, Francisco H. G., Phillippe George Leite, Poverty. Directions in Development: Poverty and Julie A. Litchﬁeld. 2008. “The Rise and Series. Washington, DC: World Bank.

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 125 Jaramillo, Miguel. 2013. “The Incidence of Social Mejia-Mantilla, Carolina, and Martha Tesfaye Spending and Taxes in Peru.” CEQ Working Woldemichael. 2016. “To Sew or Not to Sew? Paper 9 (January), Commitment to Equity, Assessing the Welfare Effects of the Garment Inter-American Dialogue, Washington, DC; Industry in Cambodia.” World Bank, Wash- Center for Inter-American Policy and Re- ington, DC. search and Department of Economics, Tulane Osorio, Rafael Guerreiro, and Pedro H. G. University, New Orleans. Ferreira de Souza. 2012. “O Bolsa Família de- Jaramillo, Miguel, and Jaime Saavedra-Chanduvi. pois do Brasil Carinhoso: uma análise do po- 2011. “Menos desiguales: la distribución del tencial de redução da pobreza extrema.” Insti- ingreso luego de las reformas estructurales.” tute for Applied Economic Research, Brasília. Documento de investigación 59, Grupo de Robinson, David O., Matthew Gaertner, and Analisis para el Desarrollo, Lima, Peru. Chris Papageorgiou. 2011. “Tanzania: Growth Josz, Christian. 2013. “Mali: Achieving Strong Acceleration and Increased Public Spending and Inclusive Growth with Macroeconomic with Macroeconomic Stability.” In Yes Africa Stability.” African Development 13/08, Inter- Can: Success Stories from a Dynamic Conti- national Monetary Fund, Washington, DC. nent, edited by Punam Chuhan-Pole and Kanbur, Ravi, Changyong Rhee, and Juzhong Manka Angwafo, 21–50. Washington, DC: Zhuang, eds. 2014. Inequality in Asia and the World Bank. Pacific: Trends, Drivers, and Policy Implica- Rodríguez-Castelán, Carlos, Luis F. López-Calva, tions. Manila: Asian Development Bank; New Nora Lustig, and Daniel Valderrama. 2016. York: Routledge. “Understanding the Dynamics of Labor In- López-Calva, Luis F., and Sonia Rocha. 2012. come Inequality in Latin America.” Policy “Exiting Belindia? Lesson from the Recent Research Working Paper 7795, World Bank, Decline in Income Inequality in Brazil.” Re- Washington, DC. port 70155, Poverty, Gender, and Equity SEDLAC (Socio-Economic Database for Latin Unit, Latin America and Caribbean Region, America and the Caribbean), Center for Dis- World Bank, Washington, DC. tributive, Labor, and Social Studies, Univer- Lustig, Nora. 2015. “Inequality and Fiscal Re- sidad de La Plata, La Plata, Argentina; World distribution in Middle Income Countries: Bank, Washington, DC, http://sedlac.econo Brazil, Chile, Colombia, Indonesia, Mexico, .unlp.edu.ar/eng/index.php. Peru and South Africa.” CEQ Working Paper Shorrocks, Anthony. 2013. “Decomposition Pro- 31 (July), Commitment to Equity, Inter- cedures for Distributional Analysis: A Uni- American Dialogue, Washington, DC; Center for Inter-American Policy and Research and fied Framework Based on the Shapley Value.” Department of Economics, Tulane Univer- Journal of Income Inequality 11 (1): 99–126. sity, New Orleans. Silva, Joana, and Julián Messina. 2016. “Wage Lustig, Nora, Luis F. López-Calva, and Eduardo Inequality in Latin America: Trends and Puz- Ortiz-Juárez. 2013. “Declining Inequality zles.” Unpublished regional study, Office of the in Latin America in the 2000s: The Cases of Chief Economist, Latin America and Carib- Argentina, Brazil, and Mexico.” World Devel- bean Region, World Bank, Washington, DC. opment 44 (C): 129–41. Souza, Pedro H. G. Ferreira de, Marcelo Me- Lustig, Nora, George Gray Molina, Sean Higgins, deiros, and Fabio Avila de Castro. 2015. “‘Top Miguel Jaramillo, Wilson Jiménez, Veronica Incomes in Brazil: Preliminary Results.” Eco- Paz, Claudiney Pereira, Carola Pessino, John nomics Bulletin 35 (2): 998–1004. Scott, and Ernesto Yañez. 2012. “The Impact UNDP (United Nations Development Pro- of Taxes and Social Spending on Inequal- gramme). 2014. Human Development Report ity and Poverty in Argentina, Bolivia, Brazil, 2014; Sustaining Human Progress: Reducing Mexico and Peru: A Synthesis of Results.” Vulnerabilities and Building Resilience. New CEQ Working Paper 3, Commitment to Eq- York: UNDP. uity, Inter-American Dialogue, Washington, Vakis, Renos, Jamele Rigolini, and Leonardo Luc- DC; Center for Inter-American Policy and Re- chetti. 2015. “Overview; Left Behind: Chronic search and Department of Economics, Tulane Poverty in Latin America and the Caribbean.” University, New Orleans. World Bank: Washington, DC.

126 POVERTY AND SHARED PROSPERITY 2016 World Bank. 2009. “International Development ———. 2016a. Brazil Systematic Country Diag- Association Program Document, Kingdom nostic: Retaking the Path to Inclusion, Growth, of Cambodia: Proposed Smallholder Agri- and Sustainability. Report 101431-BR, Wash- culture and Social Protection Support Op- ington, DC: World Bank. eration.” Report 48083-KH (June 19), World ———. 2016b. “Cambodia: Poverty Reduction Bank, Washington, DC. and Equity in 2013.” World Bank, Washing- ———. 2013. “Mali: Poverty and Gender Notes.” ton, DC Report 77752-ML (May 31), World Bank, ———. 2016c. “Peru Systematic Country Diag- Washington, DC. nostic: Concept Note.” World Bank, Wash- ———. 2014a. “Where Have All the Poor Gone? ington, DC. Cambodia Poverty Assessment 2013.” Report ———. 2016d. “Tanzania Systematic Coun- ACS4545, World Bank, Washington, DC. try Diagnostic: Concept Note.” World Bank, ———. 2014b. “Social Gains in the Balance: A Washington, DC. Fiscal Policy Challenge for Latin America and ———. 2016e. “International Development the Caribbean.” LAC Poverty and Labor Brief, Association Project Paper, United Republic Report 85162 rev (February), Latin America of Tanzania: Productive Social Safety Net and Caribbean Region, World Bank, Wash- (PSSN).” Report PAD1500 (May 20), World ington, DC. Bank, Washington, DC. ———. 2015a. “Republic of Mali, Priorities for ———. 2016f. “Macro and Poverty Outlook: Ending Poverty and Boosting Shared Prosper- Spring Meetings Edition, 2016.” World Bank, ity: Systematic Country Diagnostic (SCD).” Washington, DC. Report 94191-ML (June 22), World Bank, Younger, Stephen D., Flora Myamba, and Washington, DC. Kenneth Mdadila. 2016. “Fiscal Incidence in ———. 2015b. “The Geography of Poverty in Tanzania.” CEQ Working Paper 36, Commit- Mali.” Report 88880-ML (April 23), World ment to Equity, Inter-American Dialogue, Bank, Washington, DC. Washington, DC; Center for Inter-American ———. 2015c. “Tanzania Mainland: Poverty As- Policy and Research and Department of Eco- sessment.” Report AUS6819 v2, World Bank, nomics, Tulane University, New Orleans. Washington, DC.

REDUCTIONS IN INEQUALITY: A COUNTRY PERSPECTIVE 127

6 Reductions in Inequality: A Policy Perspective

This chapter discusses what we know about key domestic policy interventions that are effective in reducing inequality, the benefits they generate, the choices that need to be made concerning their design and implementation, and the trade-offs with which they are associated. The chapter does not provide an exhaustive or comprehensive review of every intervention that could reduce inequality, nor does it supply universal prescriptions. Instead, it focuses on a few policy areas on which there is a sufficient body of rigorous evidence to draw out useful lessons with confidence: early childhood development (ECD), including breastfeeding; universal health care; good-quality education; conditional cash transfers (CCTs); investments in rural infrastructure; and taxation. The several lessons that are revealed through the examination are supported by consolidated and emerging evidence and appear to be generally valid. Despite progress, disparities persist today that are affecting well-being, the access to basic services, and current and future economic opportunities among the poor. Pathways to reduce inequality are many, from narrowing gaps in income generation opportunities to reducing the potential for inequalities in human capital development before they emerge; smoothing consumption among the most deprived, especially during shocks; and redistributing in favor of the poor. Although many interventions do not have equalizing objectives, they are associated with equalizing outcomes. Improved competition and economic efficiency are compatible with reducing inequality. The fine points of policy design absolutely matter in determining the extent to which policy interventions lead to reductions in inequality without compromising efficiency. Therefore, in the design of equalizing policies, implementation trade-offs should not be overlooked because of too much attention to efficiency and equity trade-offs. Nonetheless, universal prescriptions and unique models are problematic. Evidence indicates the value of general principles and the analysis of lessons learned. Integrated, simple, and flexible interventions are more likely to succeed than individual interventions, and the composition of interventions dictates the extent of success. Equalizing interventions are not a luxury exclusive to middle- and high-income countries, nor an option only during periods of prosperity. Good interventions are possible in all settings and at all times, including among low-income countries and during crises, and they may disproportionally favor the poorest households. To achieve this, the most deprived and the most vulnerable must be involved in the interventions. Knowledge about appropriate policies and interventions has increased significantly in recent decades, but more microeconomic data and better analysis of these data are needed.

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 129 Introduction specific interventions discussed are ECD, including breastfeeding; universal health Countries can reduce inequality. This is care coverage; good-quality education; cash true of low- and middle-income countries transfers, mostly conditional transfers; in- as well as high-income countries. Success- vestments in rural infrastructure, specif- ful country experiences confirm that good ically, rural roads and electrification; and macroeconomic management, solid and income and consumption taxes. steady growth, functioning labor markets, The selection means that other policy and coherent space for the play of domestic interventions with potential effects on in- policy lead to reductions in inequality (see equality are not included in the assessment. chapter 5). This is not intended to deny the potential Chapter 6 focuses on what is known equalizing effects of these other interven- about domestic policy interventions that tions; the evidence of the equalizing effects are effective in reducing inequality, the ben- is merely less compelling, is currently being efits they create, the choices that are neces- collected, or a general consensus is absent sary in designing and implementing them, on how to frame the policies to reduce in- and the trade-offs that tend to emerge. The equality while reaching other objectives. trade-offs involve choices between more eq- This is the reason interventions revolving uity and more economic efficiency, but also around land redistribution or financial along many other dimensions, for example, inclusion, measures to adapt to climate deciding who might benefit from an inter- change, or steps to promote technology and vention and who may not, who should pay, innovation are not included, for example.1 and whether to promote greater generosity Even the highly selective choice of in- or lower costs. terventions demonstrates that inequality The focus is on a narrow group of inter- can be successfully confronted from mul- ventions within the coherent space of do- tiple angles. Pathways to reduce inequality mestic policies identified in chapter 5 that are many, from narrowing gaps in income can successfully reduce inequality without generation opportunities to reducing the compromising efficiency. There is no at- potential for inequalities in human capital tempt to be exhaustive or to offer a com- development to emerge; smoothing con- prehensive review of every policy that may sumption among the most deprived, espe- narrow inequality or enhance the scope for cially during shocks; and redistributing in equality of opportunity. Voluminous re- favor of the poor. The interventions selected views of sectoral interventions already exist in this chapter cover each of these pathways. that provide relevant in-depth assessments. One approach to analyzing the evidence There is much less analysis available on the consists of following the effects of inter- capacity of certain policies to reduce in- ventions during an individual’s life cycle, equality and the pathways through which which helps determine the short- and long- this may be accomplished. term equalizing impacts of interventions. The policy areas examined in this chap- Thus, breastfeeding during the first year ter are those the evaluation in chapter of life and ECD interventions during the 5 shows to be relevant in the effort to re- first 1,000 days of life set the foundations duce inequality, alongside macroeconomic for greater opportunities later by prevent- management, steady growth, and functioning nutritional and cognitive development ing labor markets. These policy areas are gaps among the poorest children. They are human capital accumulation, investments predistribution interventions, that is, they in infrastructure, and progressive fiscal address potential inequalities early on, be- policies. Within these policy domains, this fore the inequalities emerge. By prevent- chapter tightens the focus to selected in- ing or reducing nutritional and cognitive terventions on which a body of rigorous development gaps, which have long-run evidence has been accumulated, including consequences on health, education, and evaluation evidence, thereby allowing les- future earnings, the interventions also level sons to be drawn out with confidence. The the playing field and help equalize oppor-

130 POVERTY AND SHARED PROSPERITY 2016 tunities throughout the life of an individ- pathways, and implementation issues most ual. Universal health care coverage and relevant to each of these interventions. good-quality education increase the human The assessment offers analysis of multiple capital of children, youth, and adults. Given evidence-based lessons on successful equal- the daunting inequalities in health care izing interventions across low- and middle- and education prevailing today, achieving income countries, regions, and contexts. universal health care coverage and good- quality education for all requires that those Early childhood who are currently excluded from health care and quality education benefit from inter- development and nutrition ventions. Raising the human capital of the ECD promotes physical, socioemotional, excluded also helps level the future playing language, and cognitive development dur- field. In contrast, CCTs and taxation have ing a child’s early years. Breastfeeding and short-term effects on income distribution complementary feeding are often integrated if they, respectively, smooth consumption in ECD interventions. Investing in such in- among the most deprived and redistribute terventions shapes an individual’s educa- income from the rich to the poor. The con- tional attainment, health, social behavior, struction or improvement of rural roads and earnings in adulthood.3 The first 1,000 likewise has immediate effects on the gen- days of life are especially important: nutri- eration of income opportunities by expand- tional deficiencies and cognitive underde- ing the market connectivity of, for instance, velopment during this period are associ- isolated smallholder farmers. In so far as ated with later cognitive delays and lower these investments allow the poorest com- academic achievement. Reducing inequality munities to increase and diversify incomes, through ECD therefore reduces inequalities the strategy should also have an impact on in ability, educational achievement, health inequality. status, and expected adult earnings, gains Another approach through which one that are carried over throughout an individ- may analyze how policy interventions af- ual’s life. fect inequality is by way of the functions The current inequalities in ECD are of the policies. To end poverty, policies can stark. Thus, worldwide, poor children enjoy grow the incomes of the poorest; boost far less access to ECD programs than their their human capital, thereby reducing the richer peers. Data show that, in 21 of 27 low- gaps between them and other population or middle-income countries, preschool en- groups; and protect them from unequaliz- rollment rates among the poorest quintile ing shocks, that is, shocks having dispropor- are less than a third of the rates among the tionate consequences on the poorest relative richest quintile.4 These disadvantages are to less vulnerable population groups.2 For compounded by the fact that poor children example, investment in rural infrastructure also have less access to adequate nutrition, supports economic growth and other phys- health care services, basic water and sani- ical investments; ECD, universal health care tation infrastructure, and childcare, which coverage, and good-quality education con- make them more prone to inadequate de- stitute investments in human capital; and velopment. As a result, nearly one under- cash transfers allow individuals to be pro- 5-year-old in four worldwide still suffers tected against various consumption shocks from stunting (low height for age), a sign of and other risks. Policy choices in taxation chronic malnutrition and abbreviated child may affect economic growth as well as the development.5 availability of household resources to invest Unlike some policy interventions, ECD in human capital and access other protec- programs are not associated with equity- tion mechanisms. efficiency trade-offs.6 Investments in ECD Organizing the policy discussion around interventions typically have economic bene- such analytical frameworks allows the fol- fits among individuals as well as society. For lowing sections to briefly describe the ex- example, earnings at ages 28–40 among par- isting evidence on the benefits, impacts, ticipants in the HighScope Perry Preschool

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 131 Study in the 1960s in the United States (de- FIGURE 6.1 The Mental Development of tailed below) were 28 percent higher than Stunted Children, Jamaica, 1986–87 7 the earnings among the control group. Such 115 individual beneﬁts represent about a third to a half of the total payoffs from such inter- 110 ventions, once social beneﬁts—the reduced public spending required to address grade 105 repetition, social assistance transfers, and 8 crime—are taken into account. On breast- 100 feeding, a recent meta-analysis estimates the economic impact of exclusive breastfeeding 95

during a prolonged period (until age 12 Development quotient months or later) at US$302 billion, or 0.5 90 percent of the world’s gross income.9 Given the broad nature of ECD inter- 85 ventions, those interventions reported to Birth 6 12 18 24 have the largest benefits and associated with Age (months) the most compelling supporting evidence Children of normal height of such benefits are examined here: par- Stimulation and nutritional supplement enting skills, preschool, breastfeeding, and Control group 10 nutrition. Source: Grantham-McGregor et al. 1991. Note: Development quotient = age of the group into which test scores place the child, divided by the child’s chronological age Parenting skills and multiplied by 100. Because of the preeminent role the home environment plays in the early development of a more rapid rate than undersized children children, many ECD programs concentrate who were not participating (figure 6.1).12 on parenting skills to promote greater cogni- Researchers followed up among the partic- tive stimulation among children and higher- ipants 20 years after the intervention and quality parent-child interactions. These profound that the groups receiving stimula- grams generally involve one or more of three tion (with or without the nutrition supple- initiatives: home visits, group sessions, and ment) had 25 percent higher earnings than clinic appointments. Research suggests all the control group. This increase in earnings these programs are largely effective. had allowed those in the stimulation pro- The most well known program of this gram to catch up completely with the kind in the developing world is a Jamaican nonstunted comparison group, effectively intervention that began in 1986 and lasted reducing the inequality of incomes across two years. The Jamaican study targeted tod- the groups. Meanwhile, the nutrition-only dlers ages 9–24 months who suffered from group showed no statistically significant stunting. The intervention consisted of difference with the control group.13 weekly visits by community health workers Building on the Jamaican success, other to teach parenting skills aimed at fostering countries have adapted the program to local cognitive and socioemotional develop- conditions. The results have been widely ment.11 It also provided nutrition supple- positive. For instance, a recent randomized ments and psychosocial stimulation. The controlled trial involving 1,400 children in sample was divided into four groups: one Colombia that enlisted local mothers to group received stimulation only; one re- make home visits found substantial ben- ceived nutrition only; one received both; efits in children’s cognitive and language and one received neither. Undersized chil- development.14 The program achieved the dren benefiting from psycho social stimu- most success among children with higher lation and nutrition interventions caught baseline levels of development and among up with normally sized children 18 months children whose mothers had higher levels of after the start of the program, developing at schooling. In Bangladesh, another random-

132 POVERTY AND SHARED PROSPERITY 2016 ized controlled trial involved weekly group pression. The estimated premium in lifetime meetings, coupled with home visits.15 The net benefits per program participant relative program led to significant benefits in men- to the control group was US$78,000.21 Other tal development, and children in the pro- preschool programs evaluated in the United gram were happier, more cooperative, more States, including the Abecedarian Project vocal, and more responsive. A program in and state programs, suggest that the aver- Ecuador found significant improvements age benefit-to-cost ratio was 6:1–7:1, that is, in language, memory, and fine motor skills, US$6.00–US$7.00 per US$1.00 invested in while a quasi-experiment in Brazil found preschool.22 significant benefits in mental and psycho- While no longitudinal study has been motor development in a project involving conducted in the developing world over group workshops among mothers, along such a long period, a number of studies with 10 home visits promoting play.16 A have investigated short-term impacts. In randomized evaluation in Antigua, Jamaica, Mozambique, a Save the Children pilot and St. Lucia studied the impact of a com- program begun in 2008 in 30 villages in bination of home visits and a package of rural Gaza Province has shown positive instructional videos and materials shown to results, all for an investment of less than mothers while they waited to see nurses in US$3 per child per month. A rigorous eval- clinics. The study found substantial benefits uation shows that the program improves to child cognitive development and the par- outcomes among children in cognitive, enting knowledge of mothers.17 socioemotional, and fine-motor development; increases the chances that children Preschool will be in primary school and at their age- appropriate grade; and improves self- A growing body of evidence demonstrates reported parental behavior in early stimula- that cognitive and socioemotional gaps be- tion and discipline.23 A program in Argen- come apparent even before children enter tina also demonstrates the positive effects of primary school. By the age of 4, children preschool. Between 1993 and 1999, the gov- in poorer households typically show lower ernment built classrooms so an additional levels of cognitive and language develop- 175,000 children could attend preschool. A ment with respect to their peers.18 The most study in 2006 found that children attending well known programs addressing such early one year of these preschools had higher test cognitive and emotional delays through scores and better school behavior, including preschool education are two studies in the greater attention, effort, and class participa- United States that followed participants tion and better discipline.24 for decades and demonstrated the bene- Improving preschool quality is another fits of preschool.19 Initiated in 1962, the avenue to enhancing outcomes among the HighScope Perry Preschool Study randomly most disadvantaged children. Quality refers assigned 3- and 4-year-olds in low-income to several aspects of learning, from teacher- households to preschool or no preschool student interaction, curriculum selection, groups, with follow-ups at ages 19, 27, and professional development, and staff accred- 40. By age 40, adults who had been in the itation to classroom size and learning mate- preschool group had higher earnings, were rials. In general, higher quality leads to bet- more likely to be employed and to have ter learning outcomes, as demonstrated, for graduated from high school, and were less example, in Bangladesh and several coun- likely to have been arrested, spent time in tries in East Africa.25 Evidence on preschools jail, or used drugs.20 The other well-known as well as primary and secondary education intervention was launched in the early 1980s (see the next section) points to the experi- at the Chicago Child-Parent Center. By age ence and incentives of teachers rather than 24, the program participants showed higher teacher training as the most effective driv- high school completion rates, higher rates of ers of improved learning.26 For example, in four-year college attendance, lower rates of Uruguay, among children of mothers with incarceration, and a lower incidence of de- low educational attainment, those who at-

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 133 tended preschool were 27 percentage points ucation (0.9 years), and higher monthly in- more likely to remain in school at age 15 comes (R$341, about a third of the current than a control group of children who did minimum wage) relative to people who had not participate in preschool. Among chil- been breastfed for less than a month during dren of more well educated mothers, the a corresponding period in the past.35 A large corresponding effect of preschool was only randomized trial in Belarus estimated a sys- 8 percentage points.27 tematic higher mean of about 6 IQ points Richer households are more likely to among treated children age 6 who had been send their children to preschool, but the breastfed throughout the first 12 months of children of poor households enjoy bigger life relative to a control group of nonbreast- benefits if they attend preschool. A survey fed children.36 of 52 countries in the early 2000s found a Furthermore, breastfeeding is one of the strong correlation between preschool at- few positive health behaviors that is more tendance and wealth and an even stronger prevalent among poor women than among correlation with mother’s education.28 Lon- rich women in low- and middle-income gitudinal studies in the United States and countries. This suggests that promoting some nonexperimental studies in the devel- breastfeeding can reduce cognitive, health, oping world show that children with low so- and future income gaps between rich and cioeconomic status or whose mothers have poor children. For example, the program low educational attainment enjoy greater evaluated in Belarus randomly assigned ma- benefits from preschool than more affluent ternity hospitals to encourage breastfeed- students.29 In Nepal, Nicaragua, and Uru- ing using the World Health Organization– guay, preschool benefits poor children more United Nations Children’s Fund guidelines, than their wealthier counterparts.30 while similar hospitals served as the control. The intervention led to much higher Breastfeeding and nutrition breastfeeding rates and lower diarrhea rates among infants.37 Alive and Thrive, a breast- The World Health Organization recom- feeding program launched in Bangladesh, mends that mothers start breastfeeding Ethiopia, and Vietnam, combines advocacy, within an hour of birth and exclusively community mobilization, and mass media breastfeed for the first six months of life.31 to encourage exclusive breastfeeding. Since Exclusive breastfeeding is correlated with 2010, the share of infants under 6 months lower child mortality; increasing the rate of age who are exclusively breastfed has of exclusive breastfeeding to 90 percent increased from 49 percent to 86 percent worldwide would prevent up to 13 percent in places that received the comprehen- of child deaths.32 Nonbreastfed infants face sive communication intervention package. eight times more risk of death than infants Thus, in Vietnam, exclusive breastfeeding benefiting from exclusive breastfeeding in tripled in areas where interpersonal coun- the first 12 months of life. These risks are seling services in health facilities supported larger among girls than boys.33 A meta- mass media campaigns. analysis of breastfeeding in low- and mid- However, many women in the develop- dle- income countries estimates that breast- ing world do not visit health facilities for feeding avoids about half of all cases of prenatal care or to give birth, especially diarrhea and a third of respiratory infec- among the poorest quintiles, raising the tions among infants.34 Prolonged, exclusive need for community- or home-based in- breastfeeding also improves children’s cog- terventions.38 Evidence from such interven- nitive development. A study in Brazil re- tions in Bangladesh, Burkina Faso, India, ports the substantial effects of breastfeeding Mexico, South Africa, and Uganda shows on intelligence, educational attainment, and success in initiating and extending the du- adult earnings. After 30 years, participants ration of breastfeeding.39 For example, Ban- who were breastfed for 12 months or more gladesh’s large-scale community-based nu- showed higher intelligence quotients (IQs) trition program, Shouhardo II, reportedly (about 4 points higher), more years of ed- reduced the prevalence of stunting among

134 POVERTY AND SHARED PROSPERITY 2016 under-5-year-olds from 62 percent at the reported above, has been attributed to the time of the project’s inception to 49 percent combination of nutrition-specific maternal only four years later. Other countries, Bra- and child interventions with other inter- zil chief among them, have engaged for de- ventions designed to empower women, pro- cades in large-scale multidimensional strat- mote livelihoods, and improve the health egies of behavioral change (in Brazil, as part environment of households. of the National Program for the Promotion of Breastfeeding) with the same sort of pos- Health care and education itive results described above.40 Some programs also deliver comple- Achieving universal health care mentary feeding. For example, a protein supplement program in Guatemala found The world has seen significant and sus- that, four decades after the intervention, tained improvements in health care ac- individuals who were not stunted at age 3 cess and health outcomes, but inequalities went on to enjoy levels of consumption 66 in access and outcomes associated with percent higher relative to individuals who wealth, sex, and location are still pervasive. were stunted. Beneficiaries also completed From 1990 to 2011, the annual number of more schooling, had higher cognitive skills, deaths among under-5-year-olds fell from earned higher wages, and were more likely 12.0 million to 6.9 million worldwide. The to be employed in higher-paying skilled annual global number of maternal deaths labor and white-collar jobs. Women in the and the maternal mortality ratio fell by sample had fewer pregnancies and faced less nearly 50 percent, short of the two-thirds risk of miscarriages and stillbirths.41 reduction at which the Millennium Devel- Despite this evidence on successful nu- opment Goals aimed. Over the second half trition-only programs, nutrition programs of the 20th century, life expectancy rates in- that include stimulation are generally creased twice as quickly in middle-income more effective.42 Similar to the results in countries such as China and Mexico as in the Jamaican home visit study mentioned high-income countries.45 The prevalence of above, a program in rural Vietnam shows stunting has fallen in 65 percent of coun- that infants who receive stimulation and tries worldwide, while antenatal care cover- nutrition do better than children who only age has increased in an even higher share of receive stimulation.43 The stimulation had countries, 78 percent.46 During the last 25 a greater impact on stunted children, sug- years, the rate of decline in adult mortality gesting that the most disadvantaged chil- across low-income countries has been more dren may benefit most from integrated ap- rapid among women than men.47 proaches. Nutrition programs that include Inequalities in health care are still stark. components other than stimulation have Poor people face higher risks of malnutri- also been found to be successful. In Burkina tion and death in childhood and lower odds Faso, the Enhanced Homestead Food Pro- of receiving key health care interventions. duction program combined nutrition in- In low- and middle-income countries, only terventions during the first 1,000 days of 23 percent of households in the poorest life with home gardening, small animal quintile have access to improved sanitation, production, and behavioral change com- compared with 71 percent in the richest munication components over two years. An quintile (figure 6.2).48 Antenatal care cover- impact evaluation of the program reports a age declines steeply in poorer populations; significant reduction in anemia (14.6 per- the median coverage is less than 50 percent cent) and diarrhea (15.9 percent) among among the poorest quintile of households, children ages 3–12 months, a reduction in compared with a median of 83 percent in wasting (8.8 percent) in such children, and the richest quintile. Skilled attendance at increases in dietary diversity and the intake birth, an important indicator of the ro- of nutrient-rich foods among all age ben- bustness of health service delivery, presents eficiaries.44 The success of the Shouhardo a similar pattern. Immunization coverage community-based program in Bangladesh, and access to improved drinking water

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 135 FIGURE 6.2 Median Coverage, Selected Health Care Interventions, across different contexts, from farm workers by Wealth Quintile, Low- and Middle-Income Countries, Circa 2005–13 in Indonesia to factory workers in China.52 For example, in an experimental study in Antenatal care Indonesia, nonanemic workers were found coverage to be 20 percent more productive as mea- Skilled birth sured by kilograms of latex collected per attendance day. A different experimental study in In- donesia found that the income gap between Access to improved men beneficiaries of iron supplements and drinking water sources nonbeneficiaries had increased by about 20 percent after four months of the inter- Access to improved vention (only 6 percent in favor of women sanitation beneficiaries). In China, a randomized con- 0204060 80 100 trolled trial of women cotton mill work- Median coverage (%) ers found a 17 percent rise in productivity among women who received 12 weeks of Quintile 1 (poorest) Quintile 5 (richest) iron supplementation relative to a control Source: “The World Health Organization’s Infant Feeding Recommendation,” World Health Organization, group.53 Geneva (accessed June 2016), http://www.who.int/nutrition/topics/infantfeeding_recommendation/en/. Better health improves educational per- Note: Results based on a sample of 83 countries using 2005–13 data or data on the closest available year in STATcompiler (DHS Program STATcompiler) (database), ICF International, Rockville, MD, http://www formance among children, because health- .statcompiler.com/; MICS (Multiple Indicator Cluster Surveys) (database), United Nations Children’s Fund, ier children are more likely to attend school New York, http://mics.unicef.org/. and have greater cognitive capacity for learning. For example, child-deworming show narrower disparities. Rural areas have inter ventions in Kenya significantly reduce lower median health care access than urban sickness and improve school attendance areas across low- and middle-income coun- among school-age children.54 Among girls tries.49 Health care inequalities also have a younger than 13 and all boys, school partic- regional dimension: one-third of the coun- ipation rose by 9.3 percent in the first year tries worldwide that are not making prog- after deworming relative to the compari- ress in reducing under-5 mortality and ex- son group. The Kenya study estimated the panding immunization are in Sub-Saharan long-run increase in adult incomes deriving Africa and South Asia.50 from mass deworming at 17 percent, simi- Reducing such health inequalities is not lar to average estimates elsewhere in Africa only fair, it also promotes improvement in that deworming could boost income by 24 the well-being of the poorest, enables their percent.55 accumulation of human capital, increases The 2013 Lancet Commission esti- their future earnings opportunities, and, mated the costs and benefits of scaling up as seen in the case of ECD, reduces income core health interventions in 34 low-income gaps with respect to the more well off. In- countries and 48 lower-middle-income deed, poor health among the less well off countries.56 Feasible scenarios involving the has important economic and distributional narrowing of disparities affecting the poor, impacts besides the obvious implications rural, and ethnic minority populations for health status. Disparities in health out- would prevent about 10 million deaths by comes affect the accumulation of human 2035, relative to a scenario of stagnant in- capital among poor children and adults. vestments in these countries.57 Other esti- Improved health raises income through mates suggest that expanding the coverage multiple channels.51 Thus, better health raises of key existing interventions would cut the individual productivity because healthier preventable deaths associated with pneu- workers are more productive and exhibit monia and diarrhea by two-thirds, at a total lower rates of absenteeism. Studies have cost of US$6.7 billion.58 Similarly, expand- found strong causal relationships between ing access through 10 proven interventions, levels of nutrition and worker productivity ranging from the treatment of acute malnu-

136 POVERTY AND SHARED PROSPERITY 2016 trition to vitamin A and zinc supplemen- cilities to serve patients.63 As of 2013, health tation, would avert 900,000 deaths among funds covered more than 2.5 million people under-5-year-olds in the 34 countries with in 51 of Cambodia’s 81 districts, supporting the highest under-5 mortality rates, at a cost more than a million health center consulta- of US$9.6 billion.59 tions. Between 2000 and 2015, the under-5 Because of such vast benefits in terms of mortality rate in Cambodia fell from 108 to the lives saved, progress toward universal 29 deaths per 1,000 live births, one of the health care constitutes the most promising most rapid rates of decline in the world.64 and fair strategy to reduce health inequal- In Kwara State, one of the poorest states ities, raise the human capital of the poor, in Nigeria, a community-based health in- and contribute to increasing future earn- surance pilot scheme that was conducted ings and narrowing income gaps simulta- in partnership with the national health in- neously. Achieving universal health care re- surance system and private providers is requires the delivery of timely health services ported to have increased the use of health to those who need them, but who are cur- care by 90 percent among beneficiaries of rently outside any form of health care be- the program. It raised their use of mod- cause of their inability to pay, geographical ern health care providers and facilities, cut distance to care providers, cultural and gen- health expenditures by half among benefi- der norms, or citizenship status. Universal ciaries, and increased their awareness of the health care also implies financial protection importance of health status. Services were against catastrophic or impoverishing ex- financed by contributions from the state penditures among those who receive health government, the national health insurance services. Reaching underserved populations system, and payments from community inevitably requires narrowing existing cov- insurance programs. Beneficiaries incurred erage gaps and supplying affordable ser- meager copremiums of US$0.14 per person vices. In practice, to reach universal cover- per year without having to make other out- age, health coverage among the poorest 20 of-pocket payments.65 percent of a population must expand much In Rwanda, the national health insur- more quickly than the coverage among the ance program, Mutuelle de Santé, currently more well off.60 covers about 90 percent of the population There are multiple examples of substan- and provides free coverage for the extreme tial progress toward universal health care poor.66 Out-of-pocket spending fell from 28 among low- and middle-income countries. percent to 12 percent of total health expen- Thailand’s Universal Coverage Scheme en- ditures during the program’s first decade. hances equity by bringing a large uninsured These and many other experiences con- population under the umbrella of a national firm that there is no unique model of suc- program, greatly reducing catastrophic cess in universal health care.67 For example, health payments among the poor, and im- direct public provision networks in China, proving access to essential health services.61 Colombia, Mexico, and Thailand effec- Within a year of its launch, the scheme was tively cover everyone not covered by exist- covering 75 percent of the population, in- ing social health insurance mechanisms. cluding 18 million previously uninsured Brazil and Costa Rica have unified govern- people.62 ment-run health insurance and the public In Cambodia, efforts to achieve more provision network into a single health sys- comprehensive access to health services are tem aimed at covering everyone.68 Most articulated through health equity funds. of these countries have defined an explicit The funds are multistakeholder initiatives in benefit package, legally mandated in some which nongovernmental organizations re- cases, as in Colombia and Thailand, while imburse public health facilities for treating others simply guarantee a minimum pack- poor patients, largely eliminating prohibi- age of services, as in Chile. Other countries tive fees and improving the quality of care have expanded coverage to specific popu- by supplying cash incentives for staff and fa- lation groups or specific types of interven-

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 137 tions as a way of reducing coverage gaps. quality within each stage of education. Thus, Indonesia, Tunisia, Turkey, and Viet- Countries made definite progress toward nam have expanded programs to poor pop- these goals. For example, worldwide, there ulations, while programs in Argentina, Ethi- was a 64 percent rise in enrollment in pre- opia, India, Kenya, and Peru have focused primary education, and 80 million more exclusively on maternal and child health children are now enrolled in school rela- among the poor. Another Indian program tive to a few years ago. Yet, only a third of focuses on inpatient care for the poor, while countries met all the goals by 2015. Indeed, Jamaica has focused on the affordable pro- UNESCO data show that 58 million children vision of medicines for all.69 of primary-school age and 63 million chil- Successful experiences must typically dren of lower-secondary-school age are cur- overcome some trade-offs before univer- rently not in school.72 At least 250 million sal health care is effectively achieved. Chief children of primary-school age either fail among these is the trade-off between service to advance to grade 4 or do not achieve the coverage and financial protection. For ex- minimum learning targets in a given year. In ample, despite substantive progress, Ethio- India, 47 percent of children in grade 5 were pia has achieved a high level of financial pro- unable to read a second-grade text; in Peru, tection, but a low level of service coverage. half of grade 2 pupils could not read at all.73 Brazil and Colombia perform well in service Poor-quality education has a strong coverage, but do worse in financial protec- socioeconomic dimension. The poorest tion than Mexico and South Africa. Peru children are four times less likely than the and Vietnam are at similar levels of univer- richest children to receive primary educa- sal coverage, but Peru dominates in the cov- tion. Among the estimated 780 million il- erage of prevention services, while Vietnam literate adults worldwide, nearly two-thirds dominates in treatment. In addressing such are women. (In Sub-Saharan Africa, half of trade-offs, countries choose different inter- all women are reportedly illiterate.) Within ventions depending on various factors such countries, certain population groups—the as the access to and availability of financing poor, women, rural residents—face greater resources, political economy considerations hurdles in gaining access to quality educa- among opposing interest groups, a govern- tion. Large differences in learning outcomes ment’s willingness or lack of willingness to measured through test scores among chil- launch universal health care–inspired re- dren are correlated with household incomes forms, local technical capacity, and the ev- across the developing world. Children in idence available to set appropriate priorities the poorest households systematically score in the coverage, beneficiaries, and benefits below children in the richest households targeted by universal health care strategies.70 in developing countries; the gap exceeds 50 percentage points in countries such as Shifting the focus from rising Gabon, Peru, and South Africa (figure 6.3). enrollments to achieving For example, there is no reason poor education for all children in Gabon or India should perform, on average, three or four times less well in Enrollments of children and adolescents mathematics than their richer peers. Allow- in education systems have grown glob- ing such educational disparities to prevail ally, but universal access to education reis unjust and inefficient. The disparities ex- mains an elusive goal, and progress toward acerbate existing inequalities in knowledge, good-quality education is uneven. In 2000, skills, employability, and future economic United Nations Educational, Scientific, and prospects. Disparities in the access to and Cultural Organization (UNESCO) mem- quality of education are a major driver of ber countries adopted the target of achiev- income inequality within countries world- ing Education for All by 2015.71 The effort wide.74 Such disparities are known to re- consisted in reaching measurable goals in sult in persistent, intergenerational poverty promoting gender parity and education gaps because the lack of education among

138 POVERTY AND SHARED PROSPERITY 2016 segments of a society feed into economic FIGURE 6.3 Mathematics Scores, by Household Income Level, and political inequalities, driving differ- Selected Countries, Circa 2007–11 ences in life chances and opportunities.75 Yemen, Rep. For example, higher scores on international Côte d’Ivoire assessments of reading and mathematics Mauritania among students are associated with ap- Kuwait preciably higher annual per capita growth El Salvador rates in gross domestic product (GDP).76 Malawi Cross-country comparisons of educational Philippines achievement and aggregate growth rates India show that an increase of one standard de- Morocco viation in student reading and mathematics Colombia scores—roughly equivalent to improving a Congo, Dem. Rep. country’s performance from the median to Chad the top 15 percent—is associated with an Gabon increase of 2 percentage points in annual Qatar 77 GDP per capita growth. In India, farmers South Africa who exhibit a higher level of skills are able to Peru adapt more effectively to new technologies, Argentina including the use of mobile phones to make Botswana more well informed decisions, and regions Cuba characterized by better rates of schooling Tanzania also show higher rates in the adoption of Costa Rica newer farming techniques and technolo- Australia gies.78 Schooling has likewise been linked United States to more productive nonfarm activities in Netherlands China, Ghana, and Pakistan.79 More well ed- Germany ucated parents also enjoy better health and Korea, Rep. greater ability to cope with economic down- 020406080 100 80 turns. Consequently, developing countries Percent characterized by relatively poor and unequal Quintile 1 (poorest) Quintile 5 (richest) levels of access to schooling and inadequate quality in education suffer a persistent Source: World Bank calculations based on data in WIDE (World Inequality Database on Education), United Nations Educational, Scientiﬁc, and Cultural Organization, Paris, http://www.education-inequalities.org/. handicap in growth prospects. Note: The ﬁgure indicates the share of primary-school students who satisfy international test standards Among the wide-ranging interventions in mathematics and who are in households in the bottom wealth quintile or in the top wealth quintile. that have sought to address the failings of Countries are sorted according to the maximum test score in mathematics reported for any quintile. existing education systems, those attempt- ing to improve the quality of education by focusing on learning, knowledge, and mar- worth.81 Quality teaching also helps instill ketable skill development merit special at- in children skills and behaviors that—in tention. Recent assessments in developed addition to enabling learning—are re- and developing countries highlight success- warded by labor markets.82 These include ful experiences with improving teaching attentiveness, memory, self-control, and the quality, a critical contributor to increases ability to shift attention among competing in learning achievement as measured by tasks, all developed early in life and proven test scores. For example, estimates for the to be affected by teaching quality.83 There is United States indicate that, during an aca- also evidence on the impact of teachers on demic year, pupils taught by teachers who long-term outcomes, such as the probability are at the 90th percentile in effectiveness, of college attendance, teenage pregnancies, are able to learn 1.5 years’ worth of mate- and future earnings. This evidence, com- rial, whereas those taught by teachers at ing from Ecuador and the United States, the 10th percentile learn only a half-year’s strongly suggests that the way children are

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 139 taught affects their future earning trajecto- rural locations in different world regions. ries, as well as overall income gaps.84 For example, interventions that focus on An effective way to improve teacher qual- improvements in physical facilities, books, ity is through financial incentives rewarding other teaching materials, and school man- teacher attendance and better pedagogy. agement, such as the Programa para Abatir el Quality teaching depends on interactions Rezago Educativo in the poorest rural regions with children in areas such as emotional of Mexico, or the supplementary classes support, classroom organization, and in- and library initiatives among children in struction.85 A program for teachers in rural the Pratham Shishuvachan Program in the Kenya involving scholarships and teacher slums of urban Mumbai, improved the test incentives improved student test scores, al- scores of poor, chronically disadvantaged though these benefits were only temporary. urban and rural students.90 Similarly, in- In an intervention in rural India, salary increasing the duration of the school day had centives among teachers helped improve a positive effect on scholastic achievement student test scores, but the impact was and graduation levels in some Latin Amer- greater if the incentives went to individual ican countries, particularly among the de- teachers rather than collectively. There is also prived sections of the society. Evidence indi- evidence in rural India that student atten- cates that expanding compulsory schooling dance increases if incentives are introduced has had a substantial effect on educational to reduce teacher absenteeism and foster the outcomes in Norway, the United Kingdom, monitoring of teacher attendance.86 and the United States.91 Furthermore, finan- In Ecuador, children assigned to rookie cial incentives to enroll and remain in school teachers—less than three years of experi- such as CCTs, subsidies, and scholarships ence—learned less than children taught by have been shown to produce greater gains in experienced teachers (after controlling for scholastic achievements among poorer and teacher and classroom characteristics). The disadvantaged students (see below). same study notes that teacher IQs and per- All such interventions show great prom- sonality do not have any significant influ- ise to realize education for all and reduce ence on the differences in learning among learning gaps among the poorest children. students. Instead, the quality of interactions The same is true of emphasizing the mea- was found to positively correlate with higher surement of educational achievement based test scores and children’s attention, self- on learning, and not merely coverage, even control, and memory skills.87 This is consis- in countries with a chronic shortage of tent with evidence from the United King- physical inputs. This calls for greater use of dom gathered by the Education Endowment robust and consistent data across countries Foundation.88 The evidence suggests that and across time using sources such as the strategies aimed at raising teaching quality Trends in International Mathematics and by promoting greater engagement with pu- Science Study and the Programme for Inter- pils and a more open intellectual environ- national Student Assessment for aggregate ment able to foster collaborative peer learn- comparisons and survey-based randomized ing are more effective than additional formal controlled trials for rigorous microevalua- teacher training. In settings as distinctive as tions. Current outstanding efforts to bench- India, Kenya, and the United States, teacher mark and monitor progress in learning certification, tenure, and the type of contract achievement and socioeconomic correlates do not seem to make a difference in chil- include the World Bank’s Systems Approach dren’s learning.89 for Better Education Results and UNESCO’s Evaluations (typically involving random- Global Education Monitoring Report and ized controlled trials) confirm that numer- World Inequality Database on Education.92 ous interventions besides teacher quality are correlated with improvements in test Conditional cash transfers scores. The gains have been achieved across diverse educational levels, class sizes, income CCTs serve multiple purposes. The first groups, and settings, including urban and among them is providing the severely de-

140 POVERTY AND SHARED PROSPERITY 2016 prived with the incomes necessary to meet FIGURE 6.4 Simulated Gini Point basic needs. CCTs also protect the poor Reduction in the Gini Index Attributable to against the impacts of income shocks asso- CCTs, Circa 2013 ciated with seasonal income variations, loss of family breadwinners, famines, or adverse Stipend for 93 Primary Students economic shocks. CCTs smooth the con- (Bangladesh) sumption of the poor in the face of such shocks. Finally, CCTs enable households to Familias en Accion take up investments that they would not (Colombia) otherwise take up, from their children’s education to the purchase of livestock or other Bolsa Família productive assets.94 As a result of their multi- (Brazil) ple functions, CCTs can reduce poverty considerably. Even though it is not an explicit Prospera objective of CCTs, by targeting additional (Mexico) incomes on most needy households and preventing household consumption to fall 4P appreciably during periods of stress, CCTs (Philippines) may have equalizing effects in the short run. When they enable investments in ECD and 0 0.5 1.0 1.5 2.0 2.5 nutrition, education, health care, or pro- Reduction in Gini index ductive assets, CCTs contribute to leveling Source: Estimates based on data in ASPIRE (Atlas of Social the playing field through greater access to Protection Indicators of Resilience and Equity) (database), World basic services today and increasing income Bank, Washington, DC, http://datatopics.worldbank.org/aspire/. opportunities tomorrow.95 World Bank sim- Note: The Gini index of income distribution is measured assuming the absence of transfer programs (pretransfer wel- ulations using the ASPIRE database suggest fare distribution). Specifically, the Gini inequality reduction is reductions in the Gini index by between 0.2 computed as follows: inequality pretransfer, less inequality and 2.3 points across the five largest CCT post-transfer, divided by inequality pretransfer assuming no programs worldwide (figure 6.4). change in behavior. CCTs have been designed and implemented with specific and focused policy 36 months. The program also reduced the objectives, such as Bolsa Escola in Brazil, probability of child stunting—by about a which targets education. Others, such as the sixth among that age-group—and led to a 12 Prospera Program in Mexico (the rebranded percent reduction in the incidence of illness Oportunidades Program) and Bolsa Família among participants ages 0 to 5 years.97 Nic- in Brazil, have broad impacts spanning mul- aragua’s Red de Protección Social Program tiple dimensions, such as children’s educa- also increased children’s height, by 5.5 per- tion and health and household consump- centage points among stunted children, 1.7 tion expenditures. times more quickly than the rate of annual CCTs have generally shown positive re- improvement nationally between 1998 and sults at improving child development and 2001.98 In Lesotho, an unconditional cash nutritional outcomes. In Bangladesh, the transfer scheme has been associated with a Shombob Pilot Program—conditional on 16 percentage point drop in the risk of mal- mandated regular growth monitoring of nourishment among beneficiary children children and the participation of mothers relative to similar children not receiving the in monthly nutrition-related sessions—sig- transfers.99 In Ecuador, cash transfers re- nificantly reduced the incidence of wasting portedly increased the height of preschool among 10- to 22-month-old infants by 40 children in the poorest households.100 percent.96 Mexico’s Prospera Program has in- CCTs have been used to promote the creased child growth at the equivalent of 16 adoption of health care services by helping percent in mean growth rate per year (cor- to defray the cost to households otherwise responding to 1 centimeter) among infants unable to bear the expenses and by encour- who received treatment between ages 12 and aging healthy lifestyle habits that would

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 141 not emerge without behavioral condition- mance at school and grade completion. The ing. These CCTs focus on the demand side results indicate substantive overall benefits of health care services, that is, patients, by accruing from the programs. CCTs have suc- seeking to increase their use of preventive cessfully raised school enrollments among health care through regular checkups and recipient children worldwide. For example, vaccinations.101 Evidence on such CCTs sug- the Nahouri Pilot Project—a two-year cash gests multiple benefits. They can be quite transfer program in Burkina Faso that had effective at increasing antenatal visits, child- both conditional and unconditional compo- births with skilled health care professionals nents—is credited with raising primary and in attendance, delivery at health facilities, secondary enrollment rates by 22 percent and the incidence of tetanus toxoid vaccina- among boys.109 Similarly, Chile Solidario tions among mothers.102 For example, Mex- boosted preschool enrollments by 4–5 per- ico’s Prospera Program has helped reduce centage points and increased the probability infant mortality and maternal mortality by among all children ages 6–14 of enrollment as much as 11 percent.103 The benefits are in school by 7 percentage points.110 In Cam- not restricted to children: adult beneficiaries bodia, CCTs have raised secondary-school (ages 18–50) had 17 percent fewer sick days attendance by 26 percentage points, while, and 22 percent fewer days in bed because in Malawi, the Philippines, and Zimbabwe, of illness.104 Women benefiting from Peru’s the effects on secondary-school attendance Juntos cash transfer are 91 percent more are—albeit more modest—still significant, likely to be attended by a doctor at childbirth between 5 and 10 percentage points. Evi- compared with women who are not in the dence on the effects of CCTs where eligibil- program.105 Antenatal and postnatal care ity is limited to only girls, such as in Bangla- have been shown to improve, along with desh and Pakistan, also demonstrates large facility-based deliveries, in Indonesia and rises in enrollments in the range of 11–13 the Philippines.106 Evidence on the Program percentage points. Evidence from Mexi- Keluarga Harapan, the first household-based co’s Prospera indicates that such interven- CCT in Indonesia, which was implemented tions have helped improve enrollment rates in 2007, shows greater usage of primary mostly by reducing the incidence of drop- health care services among all eligible house- ping out in the critical transition between holds in areas covered by the program. Ante- the primary and lower-secondary levels and natal visits by pregnant women rose by over between the lower-secondary and higher- 7 percentage points, the share of assisted secondary levels.111 deliveries (by midwives, nurses, or doctors, A few studies have also found gains in or at medical facilities) increased by approx- cognitive ability among children who have imately 5 percentage points.107 In India, an been beneficiaries of CCTs in Ecuador and impact evaluation of the Janani Suraksha Nicaragua, reporting significant increases Yojana CCT estimates that the expansion in language skills and personal-behavioral in the use of health facilities attributed to skills even after only brief exposure to the the program resulted in 4.1 fewer perina- programs. Evidence also points to greater tal deaths per 1,000 pregnancies, 2.4 fewer benefits among such children in poorer neonatal deaths per 1,000 live births, and an households.112 In Malawi, enrollments, test increase of 9.1 percentage points in the pro- scores in English, and the probability of re- portion of fully vaccinated children.108 maining in school among girls in grades 5–8 CCTs are also policy instruments affect- are reportedly larger—relative to enrolled ing the access to and quality of education. girls—among recent dropouts rejoining The most common forms of demand-side school, who were, on average, poorer and interventions involve transfers that finan- had lower baseline cognitive skills.113 In Ec- cially incentivize poor households to enroll uador, on account of Bono de Desarrollo their children and keep them in school. Also, Humano, significant improvement in cog- CCTs provide conditional stipends based on nitive skills was reported among children school performance, continued attendance, from the poorest quartile of the household and graduation that seek to improve perfor- income distribution compared with more

142 POVERTY AND SHARED PROSPERITY 2016 well off children as a result of the interven- and poor targeting. Fragmentation occurs if tion. In Nicaragua, the Atención a Crisis a collection of typically small, unconnected Program resulted in cognitive skill improve- programs target the same demographic ments similar to those in Ecuador. groups, regions, and vulnerabilities without Regarding the unintended effects of coordination or cost-benefit considerations. CCTs, recent evaluations of programs in Multiple ministries often have responsibil- Brazil, Chile, Honduras, Mexico, Nicaragua, ity for the implementation of the programs, and the Philippines fail to demonstrate re- making coordination cumbersome and reductions in the labor market participation quiring an institutional vitality that is rare of beneficiaries relative to nonbeneficiaries in many low-income countries. However, or increases in the consumption of alcohol this does not need to be the case. A recent or tobacco or in gambling.114 In the Philip- encouraging example of an effort to tackle pines, a steep decline in the consumption fragmentation is Vietnam’s commitment of alcohol of 39 percentage points has been to broaden coverage, expand the profile of documented among beneficiary households new beneficiaries, and integrate the mul- relative to control groups.115 The evidence is tiple objectives of its several cash transfers beginning to show that these programs may programs targeted on poor rural regions also have positive effects in reducing inter- that, until recently, were not covered by ac- personal violence and street crime in Ecua- tive labor market programs.121 dor and Brazil, respectively.116 Regarding Poor targeting occurs if CCT programs fertility, evidence suggests that these pro- fail to reach part of the target population. grams have little or no impact. For example, Poor targeting is often a reflection of weak no effects were found on the total fertility governance and limited administrative ca- rate among beneficiaries of Prospera in pacity. Context also plays a critical role. The Mexico, and only a slight rise in the fertility functioning of markets, seasonality, house- rate of 2–4 percentage points was revealed hold preferences, and community dynamics among eligible households in Honduras.117 all affect the capacity of the programs to deliver benefits effectively to priority bene- 122 Trade-offs in design and ficiaries. For example, cash transfers are implementation often preferred over in-kind transfers in locations with well-functioning food mar- CCT programs frequently identify eligi- kets, where recipients can exercise their ble recipients geographically and through expanded budgetary options. In contrast, means testing based on qualifying criteria in-kind transfers are preferred in contexts to rationalize the use of scarce resources and where the market prices for food are volatile improve program targeting efficiency. Typi- and public distribution systems are reliable. cally, CCT programs incur low administra- Another example is the frequent preference tive costs. The largest five CCT programs of women for in-kind transfers in social and in the world spend between 0.09 percent household contexts where they have less and 0.44 percent of GDP.118 However, the control over the cash. low costs mask a fundamental trade-off be- Numerous design factors also determine tween the size of the benefits and coverage. the success of CCT programs, such as the Most recent data indicate that only about size of the transfer, the nature of the condi- a third of households in the poorest quin- tionality, or the characteristics of the target tile are covered by CCTs worldwide.119 The population. Increasing the size of the trans- average benefit is also small in most coun- fer often raises the impact of the program. tries, especially in low-income countries. For example, one study concludes that dou- The average transfer in the five largest CCT bling the amount of the transfer in Brazil’s programs worldwide is about 15 percent of Bolsa Escola program would halve the per- the average household consumption of the centage of children in poor households who poorest quintile.120 do not attend school.123 In Cambodia, an Part of this trade-off is explained by analysis of a program that delivers transfers fragmentation in program interventions to students based on poverty status shows

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 143 that the impacts are not linearly correlated Several general lessons on implementa- with the size of the transfers. Thus, the dif- tion may be drawn from recent CCT pro- ferences in student enrollment associated grams. First, transfer programs of all types with no stipend relative to a stipend of benefit from the adoption of technological US$45 were large, but there was no observ- innovations. Technology aids in improving able difference in outcomes between a sti- targeting. This was the case of the biometric pend of US$45 and a stipend of US$60.124 smart cards used in the Targeted Public Dis- In Malawi, the variation in the size of the tribution System in Andhra Pradesh, India. transfer to the parents of adolescent girls in Electronic cash disbursements based on grades 5–8 did not cause differences in en- smart cards and mobile banking platforms rollment rates or literacy test scores. How- have resulted in lower transaction costs and ever the proportion of cash given directly to reduced opportunities for corruption and schoolgirls was associated with significantly other losses. For example, the electronic improved school attendance and progress if payment of a social transfer in Niger cuts conditional on school attendance.125 the overall travel time required to collect There is broad evidence that condition- cash transfers by three-quarters.129 In South ality determines the impact of a program. Africa, the cost of disbursing social grants Analyses have found that school enrollment using smart cards is a third of the cost in- was significantly lower among those house- volved in cash disbursements.130 In Argen- holds in Ecuador and Mexico that believed tina, electronic payments for Plan Jefes, a their cash transfers were unconditional, large national antipoverty program, virtu- although the benefits were conditional on ally eliminated the reporting of kickbacks, school attendance.126 Simulations in Bra- which stood at 4 percent of the payments zil conclude that CCTs would not have had when these were made in cash. The ease any impact on school enrollment if they had of accessing program benefits and the im- been delivered as unconditional transfers, proved targeting have boosted the adoption while in Mexico, the simulated impact of of transfer programs even in low-income unconditional transfers on educational at- countries.131 tainment would have been only 20 percent Second, CCT programs are enhanced of that of conditional transfers.127 However, by constant monitoring, evaluation, and this does not imply that conditionality inev- adjustment.132 This is particularly the case itably has an impact. The Malawi program in the identification of beneficiaries. Eligi- evaluation shows that, relative to the con- bility requirements must be governed by trol group of girls, the significant impacts of transparent rules, frequently fine-tuned, the transfers on school enrollment and test and flexible, especially if the portfolio of scores, the probability of early marriage, and transfers is rich or integrated with other pregnancy did not vary between the benefi- interventions or CCTs as part of crisis re- ciaries of the conditional and unconditional sponse strategies. Increasingly, rigorous transfers associated with the program. evaluations, comprising randomized con- The focus on CCTs is not intended to trolled trials, are being used to assess the indicate that unconditional cash transfers direct and spillover effects of interventions exert insignificant impacts on inequality or integrated within broader safety nets. This the living conditions of the poorest. Under is the case of a recent experimental eval- some circumstances, such as frail service uation of multiple interventions involv- delivery or onerous targeting, they may be ing cash transfers, transfers of productive superior to CCTs. Evidence on uncondi- assets, technical skills training, nutrition tional cash transfers in Burkina Faso, Kenya, and hygiene programs, and access to bank Lesotho, and Zambia shows remarkable im- accounts in Ethiopia, Ghana, Honduras, pacts on educational attainment, food secu- India, Pakistan, and Peru.133 Successful rity, agricultural production, and stimulus CCT programs are associated not only to local economies. Indeed, 40 of 48 coun- with efficient beneficiary identification tries in Sub-Saharan Africa apply these sorts and targeting, but also precise evaluations of transfers.128 of transfer effectiveness. This is the case

144 POVERTY AND SHARED PROSPERITY 2016 of programs in Brazil, Chile, Mexico, and, and remote rural areas to obtain higher more recently, Ethiopia and the Philippines. prices for their outputs and pay lower prices Part of the success of emerging CCT pro- for inputs and consumer goods. Improved grams integrated within other safety net rural roads also eliminate or reduce barriers interventions is flexibility. Thus, the ability to the reallocation of labor away from agri- of safety nets in Ethiopia (the Productive culture and contribute to the development Safety Net Program) and the Philippines of new local markets.138 Roads may lessen (the Pantawid Pamiliya Pilipino Program) credit constraints by helping increase the to reach thousands of new beneficiaries value of land and, therefore, its collateral (millions in the case of Ethiopia) after cata- price, thus reinforcing the positive effects strophic events indicates that the coordina- that cheaper transportation also has on the tion of cash transfers, emergency response, adoption of fertilizers and new technol- and postdisaster reconstruction is possible ogy. The incentive to invest in physical and and effective in protecting the poor from human capital has reportedly increased as natural disasters. This was the case of the new off-farm income-generating opportu- 4P program in the Philippines, which, in a nities emerge because of new or improved situation of national calamity, was effective rural roads. These fresh opportunities may after the conditionality was voided tempo- affect the perceived returns to education, rarily, and the CCTs were delivered without including the returns to the education of compliance verification requirements for women and girls. Women also benefit di- the duration of the emergency.134 rectly if they are able to participate in construction, rehabilitation, and maintenance Rural infrastructure activities, typically within a context of public works programs aimed at rural roads. Investing in rural roads Although the vast majority of these programs do not aim directly to reduce in- New or improved rural roads reduce trans- equality, but to promote connectivity and portation costs, facilitate the relocation improve the living conditions of the poor, of labor, foster the diversification of live- they become equalizing because they dis- lihoods, enhance access to markets and proportionally benefit the poorest. This oc- services, and promote human capital in- curs when, for instance, roads benefit rural vestments. Indeed, good transportation in- residents with the least agricultural wealth, frastructure has been widely acknowledged such as smallholding farmers and landless to facilitate growth, poverty reduction, workers; when they help diversify earning and income equality.135 Empirical estimates activities among those who would otherwise across the globe suggest that, along with spend their time in low-productivity activi- communication and power infrastructure, ties or household chores, such as women; or the quantity and quality of transportation when they contribute to reducing discrim- infrastructure are positively correlated with ination by, for example, allowing low caste growth and negatively correlated with in- villagers to abandon agriculture and search come inequality.136 While the links between for more well paid activities.139 A Bangla- infrastructure, growth, and equality are desh study suggests that the benefits of the generally recognized across various types of two rural roads programs discussed below infrastructure, one type—roads—is partic- are greater among the poorest quartile and ularly relevant in rural areas. Of the world’s are insignificant among the well off in the rural population, about a third—one bil- affected villages.140 In Georgia, the poor lion people—still live in settlements that are benefited more from an expansion in off- each more than 2 kilometers from the near- farm employment among women, while est paved road.137 the nonpoor benefited more from increased New or improved all-weather roads re- nonagricultural employment and greater duce transportation costs and the time access to medical services.141 needed to reach markets. This enhanced The US$37 billion Pradhan Mantri Gram access to markets allows farmers in poor Sadak Yojana (the prime minister’s rural

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 145 roads scheme), a large rural road construc- project villages, while declining in control tion program in India, has provided paved villages; and waiting times for the arrival of roads for more than 110 million people, or ambulance services were reduced.145 half the unconnected rural population in Competition is a critical factor in ensur- the country at the onset of the program in ing that benefits reach the poorest. Invest- 2001. According to a quasi-experimental ments in roads lead to higher transport vol- analysis, the benefits in newly connected umes and lower transport fees only if there districts include reduction in the levels and is competition among providers. If com- dispersion of food prices.142 Food secu- petition does not exist or is not promoted, rity improved through diet diversification. more transport opportunities and falling Households began spending more on pro- travel times may still emerge, but the poor cessed and perishable foods and consuming are less likely to benefit from the improved more nonperishable staples. The cultivated connectivity because they cannot afford to area treated with fertilizers expanded by 3 change their travel patterns (for example, percent, mostly for food rather than cash traveling to other villages in search of eco- crops, while 10 percent fewer households nomic opportunities) or because service reported agriculture as the main source of quality is not enhanced. Evidence from case income. Household earnings rose an aver- studies in Asia confirms that the promotion age of 8 percent entirely because of wage of competition following the expansion of labor. Enrollment rates among children ages rural roads disproportionally benefits the 5–14 increased among both boys and girls, poorest.146 A study among villages in Indo- although enrollments among the 14–20 nesia, the Philippines, and Sri Lanka finds age-group declined. This suggests that rural that, relative to control villages, villagers roads boosted education returns among who benefited from the expansion or reha- younger children and employment oppor- bilitation of rural roads and the promotion tunities among young people. of competition in transportation traveled In Vietnam, the Rural Transport Proj- more frequently to buy provisions, for em- ect I improved 5,000 kilometers of rural ployment and business, and to obtain doc- roads. The benefits—typically observed two uments, and they also required travel times years after the completion of the program— that were shorter by about two-thirds.147 included a 10 percent rise in the share of Evidence from community-based proj- communes participating in new markets ects in Georgia, including the rehabilitation relative to a control group of communes of schools, roads, and water supply systems that were not part of the project, a 20 per- illustrate that the equalizing impacts of in- cent growth in private businesses involved vestments in roads in poor, isolated, and in services, such as tailoring and hairdress- less densely populated rural settlements are ing, and a 15 percent rise in primary-school multiplied if the investments are coupled completion rates.143 with expansions in schools, medical facili- In Bangladesh, the Rural Development ties, banks, agricultural extension services, Program and the Rural Roads and Markets water and sanitation, and electrification. Improvement and Maintenance Program Interventions must also avoid unintended have boosted employment and wages in ag- negative effects on the poor such as an in- ricultural and nonagricultural activities, as crease in child labor or trading in contra- well as aggregate harvest outputs. Per capita band.148 Thus, rural road improvements annual spending across households in the that are accompanied by good maintenance projected program areas has risen by about plans help encourage poor households to 10 percent.144 invest in alternative livelihoods, which they In Georgia, the share of villages hosting might not undertake if they believe the new small nonagricultural enterprises increased all-weather roads will not be maintained. in project villages more than in control Women-inclusive practices in the rehabili- villages; employment and wages among tation and maintenance of roads, the inclu- women in off-farm activities also rose in sion of girls in new schooling opportuni-

146 POVERTY AND SHARED PROSPERITY 2016 ties, and the recognition of equal property ity grid were 9.0 percentage points higher rights for women that promote increases among girls and 6.3 percentage points in the prices for land need be part of rural higher among boys.152 Electrification was road improvement plans. also associated with more average years of schooling by almost a year among girls and Electrification 0.13 years among boys. Rural electrification can have positive Several studies find that access to electric- community effects. For example, street ity boosts household incomes by expanding lighting improves security; clinics with ac- labor supply and fostering a shift from farm cess to electricity can stay open longer and labor to formal employment. Thus, electri- provide cold chains for vaccines; and access fication in rural communities in Guatemala to electricity can help reduce absenteeism and South Africa has led to a 9 percentage among health workers and teachers. In Paki- point rise in women’s employment that stan, absenteeism among teachers in public cannot be attributed to greater demand schools that have electricity is half the rate because there has been no comparable in- in other schools. This is particularly rele- crease in men’s employment.149 In India, vant for the women teachers required for household electrification has been found the education of girls.153 Moreover, elec- to raise labor supply by about 16 days a trification programs can have large and year among men and 6 days a year among immediate health benefits by decreasing women. The increase in labor among men indoor air pollution in poor households, is attributable to a shift from casual wage typically by allowing kerosene and firewood work and leisure activities, while, among to be replaced as energy sources for lighting women, it is attributable to an expansion in or cooking. This leads to large reductions casual work.150 in respiratory infections among under- Electrification can also generate addi- 5-year-olds and can potentially lessen the tional household income by making small risk of lung cancer among household mem- home-based businesses economically via- bers. Two years after the baseline in a recent ble or more productive. Small enterprises electrification program in northern El Salva- and informal sector businesses often suffer dor, particle pollution concentration was an most from the erratic supply of electricity average 67 percent lower among households by public utilities, because they cannot af- that had been randomly encouraged to elec- ford generators to produce their own elec- trify relative to those that had not.154 The tricity and insulate themselves from power change is driven by less kerosene use. These outages. Stable supplies of electricity often various factors contribute to improving make business activities, including access human capital formation among the poorest to export markets, more productive and and therefore have the potential to level the economically viable. Indeed, evidence from playing field and reduce future inequality. rural Vietnam shows that households con- Electrification can also promote gender nected to the electricity grid are nine times equality by freeing up women’s time from more likely to be involved in home produc- household chores, such as the collection tion than households without such a con- of firewood, and raising women’s employ- nection.151 Incomes from nonfarm activi- ment. There is also evidence that access to ties among the former rose an estimated 29 electricity increases the income controlled percent as a result of electrification. by women through formal employment The availability of electric lighting also and the creation of small woman-run en- provides additional opportunities to study terprises. In South Africa, a rural electricity and is associated with greater school atten- project led to a large expansion in the use dance and school completion rates, espe- of electric lighting and cooking and steady cially among girls. The study in Vietnam reductions in wood-fueled cooking, as well reports that school enrollment rates among as a 9.5 percentage point rise in women’s children in households on the electric- employment.155 One of the most important

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 147 uses of electricity after lighting is for televi- services need to be financed. Taxes there- sion, which can potentially improve health fore constitute an essential component of education, reduce fertility, and challenge any successful strategy for guaranteeing entrenched perceptions of gender roles.156 equal opportunity. Taxes also raise the rev- In rural India, access to cable television may enues needed to finance interventions such result in less tolerance for spousal abuse, as noncontributory pensions or housing less preference for sons rather than daugh- subsidies that have deliberate distributional ters, and a greater likelihood young girls will objectives. Some historical examples of tax attend school.157 Similarly, access in rural and benefit reforms aimed at sharing the areas to telenovelas (television soap operas) gains of economic growth and enhancing resulted in higher divorce and lower fertility equity are the 1990 tax reform in Chile im- rates in Brazil, which may be related to the mediately after the country’s return to de- empowerment of women through the imita- mocracy and the 1994–99 tax reform of the tion of role models of emancipated women Katz Commission in postapartheid South and the presentation of smaller families.158 Africa. More recently, European Union (EU) The potential benefits of electrification countries have embarked on comprehensive are enormous. The International Energy fiscal consolidations based on clear equity Agency estimates that there were 1.2 billion considerations in response to the 2008–09 people without access to electricity in 2013, financial crisis.159 or 17 percent of the global population. Many Taxes also have a redistributive role on more have access only to an insufficient their own, whether deliberate or unintended. or unreliable supply. Worldwide, approx- Taxes redistribute income in two ways. They imately 80 percent of people without elec- address the income inequality emerging tricity live in rural areas, and more than 95 from labor and capital markets by establish- percent live in Asia and Sub-Saharan Africa. ing different tax rates to balance the relative However, efforts to expand electrification contributions of individuals, households, in rural areas often face a trade-off between and firms in total revenue collection. This keeping electrification financially viable by is done in multiple ways, including impos- defraying the costs and reaching people who ing a tax rate that rises with earned incomes, are least able to pay. Households are often offering tax credits based on age or house- required to pay at least part of the connec- hold composition for individuals earning tion cost, which can be prohibitively high similar labor incomes, or exempting some for the poorest. Yet, there are several ways consumption goods from the value added affordability can be ensured. Bangladesh tax (VAT), while raising this tax on others and Brazil have progressive pricing schemes to reflect, for example, the differences in the that offer lower prices to households that consumers of basic goods and luxury goods. consume only small amounts of electric- Taxes also affect the generation of in- ity. South Africa offers a poverty tariff that comes in labor and capital markets. They provides poor households with 50 kilowatt influence the labor, savings, and investment hours of electricity per month at no cost. decisions of individuals and firms, thus exerting an impact on the market incomes of Taxation these individuals and firms. For instance, introducing high tax rates on labor and Choices in the composition and progres- capital earnings may discourage labor effort sivity of taxes determine the capacity of a among individuals and reduce the extent government to finance interventions that to which individuals and firms save and help redistribute income. Moreover, taxes invest. High social insurance contributions can be designed to ensure that inequal- and payroll taxes make formal work less at- ity narrows, while keeping efficiency costs tractive in highly informal economies with low. Programs that foster childhood devel- generous noncontributory pension benefits opment and adequate nutrition, improve such as Colombia and Mexico. To the ex- the access to and quality of education, or tent that taxes introduce distortions, they ensure broadbased or universal health care restrict economic growth and revenue col-

148 POVERTY AND SHARED PROSPERITY 2016 lection. In the case of Mexico, the efficiency tax, but also on the composition of taxes costs of taxation have been estimated at 0.9 that support the fiscal system. Conventional to 1.4 percentage points of GDP owing to wisdom suggests that the more tax systems lower labor productivity and GDP growth rely on indirect rather than direct taxes, the deriving from the fragmented social secu- less these systems are able to narrow in- rity system.160 Redistribution can thus en- equalities.162 This is because indirect taxes tail high efficiency costs. are typically more regressive than direct But this does not have to be. There are taxes.163 However, such generalizations are many ways the efficiency costs of taxes— sometimes deceptive.164 For example, while such as administrative and compliance indirect taxes widen inequality in many costs—can be kept to a minimum even middle-income countries, this is not true in if they are difficult to avoid. Taxes can be Mexico and Peru, where indirect taxes alone designed to encourage risk-taking behav- reduce market income inequalities, or in ior that, ultimately, may boost the returns Ghana, South Africa, and Sri Lanka, where on investment, or credit constraints can be indirect taxes are mostly neutral. In eight relaxed so poor households can invest on Sub-Saharan African countries, the VAT is health care and education. Likewise, there reportedly less unequally distributed than are many ways to achieve effective redistri- market incomes.165 bution, including broadening the tax base. Specific country experiences help clar- Personal income tax deductions can be ify how composition influences the redis- carefully designed to avoid those that un- tributive capacity of tax systems. Box 6.1 duly benefit the more well off and do not compares how the choice of taxes and the generate significant revenue gains, such as amount of progressivity in two recent tax deductions for mortgage interest or chari- reforms with similar objectives affect redis- table donations. Earned income tax credits tributive outcomes. It also describes how can be used instead of tax allowances to the design of fiscal consolidation measures favor labor force participation and engage- during periods of crisis may have particular ment in the formal sector because, to earn redistributive effects. the credits, taxpayers must be working in The redistributive role of taxes is poten- the formal sector. Corporate income tax tially great. In some cases, along with large, rates can be harmonized with personal in- well-targeted benefits, taxes can redress come tax rates to prevent distortions in market income inequalities dramatically. savings and investment decisions relative For example, together, taxes and transfers to hiring. Reduced VAT rates and exemp- redistribute important shares of market in- tions can be limited because more well off comes in the EU, reducing the Gini index households benefit the most from their of the market income inequalities of the 27 greater consumption, and these reductions member states by an average of 20 points.166 and exemptions erode the revenue base and, But this is not always the case, especially in ultimately, lower the financing available for developing countries. Indeed, the redistrib- social spending. Other measures that in- utive role of taxes and fiscal systems more crease the progressivity of the tax system generally is often rather limited. A recent while raising revenues include property and comprehensive study comparing the tax and inheritance taxes, which can also help limit benefit systems of 150 countries concludes intergenerational inequality. In developing that tax systems have had a limited, but countries, property taxes are underutilized, inequality-increasing impact since 1990.167 representing about 0.5 percent of GDP, and There are multiple reasons for the lim- declining at a time when the concentration ited redistributive effects of tax systems. and accumulation of wealth are on the rise The administrative systems of countries, worldwide.161 All these measures boost rev- especially low-income countries, are often enue collection in an equalizing manner, inadequately funded and staffed, and lack while keeping disincentives to a minimum. the autonomy, transparency, accountabil- The impact of taxes on inequality de- ity, and technology to operate in the face of pends on the degree of progressivity of each challenges such as a large informal sector

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 149 BOX 6.1 Tax Reform and Fiscal Consolidation

Equity-enhancing tax reforms in of expenditures for these goods in Chile Chile and Mexico was greater among the poor than among The recent tax reforms in Chile and other socioeconomic groups. The two Mexico in 2014 and 2010, respectively, countries converge in that the reduction confirm the importance of design in the expenditures on health-damaging and composition in determining the goods was fairly limited, less than 0.5 redistributive effects of taxes. With percent of total household expenditures. similar reform goals, the choices were different, and so were the impacts. The distributional effects of fiscal Both sets of reforms aimed at reducing consolidation fiscal deficits, raising revenue to finance Reinforcing efficiency and equity social spending, and enhancing tax considerations becomes more important equity, that is, ensuring that the tax during periods of crises or protracted reforms were progressive. Mexico fiscal consolidation when the pressures relied on a combination of increases in to cut spending and raise revenues the VAT and special taxes on alcohol grow along with the need to protect the and tobacco, as well as lottery tickets most vulnerable from the consequences and telecommunication services. of a crisis. However, tax decisions in Additionally, Mexico raised the three response to crises also have distributive top rates in the personal income tax consequences. In the short run, fiscal schedule. In Chile, the reform focused consolidation may lead to declines in mainly on corporate taxes. It also output, employment, and wages that increased special indirect taxes, but, in compound the initial effects of a crisis. contrast with Mexico, these taxes were These effects tend to widen inequalities not intended to boost revenues, but because unskilled workers constitute to reduce the consumption of alcohol, a larger share of wage earners, and tobacco, and beverages containing unemployment hits already vulnerable sugar. youth and women the hardest. Fiscal The two reforms were generally adjustment affects the level of taxes and progressive, that is, the more well spending and often their composition. off bore a higher share of the burden If, for example, these adjustments of the reform. However, there were rely on raising VAT rates, imposing substantive differences in the impacts. pension freezes, or cutting the benefits In Chile, the top 3 percent of the to families with children, they widen income distribution bore the burden of inequality, especially in a context of high the income tax reform. In Mexico, the unemployment and recession. effects of the changes in income tax Nonetheless, it is possible to design rates affected the top 40 percent of the fiscal responses to a crisis that are distribution. progressive and prevent further widening The effects of the special in inequality. For example, in a sample consumption taxes were opposite in of 27 episodes of fiscal adjustment in the two countries: the impacts were 19 middle-income countries between progressive in Mexico, but regressive 1971 and 2001, the Gini index either in Chile. In Mexico, the richer deciles remained virtually unchanged after ended up paying more in absolute and the adjustment or rose slightly if the relative terms than the poor. In Chile, adjustment coincided with a period of the bottom 40 percent of the income economic contraction.a In a context of distribution (the bottom 40) paid a falling oil and natural gas prices, Algeria higher share of their expenditures than and Morocco recently reduced the scope the more well off (though a smaller of their poorly targeted energy subsidies. absolute bill). This is because the share Some progressive features of Greece’s

(Box continues next page)

150 POVERTY AND SHARED PROSPERITY 2016 BOX 6.1 Tax Reform and Fiscal Consolidation (continued)

successive fiscal packages to combat Additional pension taxes were created the 2008–09 global financial crisis with separate, rising rates based on reportedly prevented additional surges the amount of the pensions received. in inequality. Thus, public sector wages Interest income taxes were raised. were capped, and special allowances These interventions did not prevent for civil servants were reduced to avoid Greece from suffering the largest fiscal widening the disparities arising because contraction in the EU between 2008 of transfers to comparatively more and 2014. However, they helped block well paid civil servants. The 13th and further increases in inequality. The 14th monthly salaries were eliminated richest 20 percent of Greek households among high-income workers. A lump have borne the brunt of the fiscal sum transfer was introduced in 2014 for consolidation, although the poorest low-income families and the vulnerable, household decile was hit the hardest by along with emergency property taxes. the VAT increases.

Sources: Abramovsky et al. 2014; Avram, Levy, and Sutherland 2014; Clements et al. 2015; De Agostini, Paulus, and Tasseva 2015; Fabrizio and Flamini 2015; Tsibouris et al. 2006; World Bank 2016c. a. IMF (2014). or conflict. This weak capacity prevents au- ciency reasons, Eastern European countries thorities from raising significant revenues, impose flat-rate income taxes, which render launching progressive schedules, avoiding personal income taxes less redistributive evasion, and ensuring compliance.168 than progressive personal income taxes at A limited redistributive role is sometimes the same level of revenue collection. the result of the small size of the pie, which While some analysts believe that more- does not allow significant redistribution. unequal countries redistribute more, this Even if poor countries were politically and belief is called into question by the large dif- administratively able to shift large shares of ferences in the level of redistribution across income from the rich to the poor, they would countries at similar levels of inequality.171 need prohibitive marginal income tax rates For example, Chile and Colombia share to end poverty and substantially narrow in- similarly high levels of market inequality, equality.169 However, a recent analysis shows but Chile redistributes income significantly that much of global poverty as defined by more than Colombia. Mexico and Peru also the US$1.90 per capita per day poverty line have similar levels of inequality, but only could be eliminated in developing countries Mexico redistributes income substantially. by reallocating regressive fossil-fuel energy Indonesia and the Russian Federation also subsidies and excessive military spending to show comparable, though much lower lev- cash transfers.170 els of inequality, but only Russia appears to Limited redistribution is not exclusively redistribute significantly.172 Redistribution related to low-income countries or, more is also limited in Ethiopia and Sri Lanka; generally, to a country’s level of develop- in South Africa, it is more appreciable.173 ment. Indeed, certain tax choices limit the The norm is that the size of the tax benefit redistributive capacity of tax systems in system is closely correlated with the redis- middle- and high-income countries. Thus, tributive effect. This is important because Brazil’s heavy reliance on consumption countries with large tax systems also tend taxes that include basic foodstuffs, which to provide large benefit transfers, which, represent a disproportionally large share ultimately, show more redistributive im- of the poor’s consumption budget, reduces pacts.174 This explains how Russia and the redistributive impact of the fiscal system South Africa achieve greater redistribution (chapter 5). For administrative and effi- through the fiscal system than Indonesia

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 151 and Sri Lanka. Regardless of the inherent methods for addressing inequality. Thus, ability of taxes to reduce inequality, choices investments in rural roads and electrifica- about progressivity, composition, and size tion influence income generation opportu- thus largely determine the equity effects.175 nities, employment, and even perceptions of gender roles. Expanding ECD, health care Concluding remarks coverage, and good-quality education can reduce cognitive, nutritional, and health The progress documented in this chapter status gaps, thereby narrowing inequalities suggests that policy interventions can be in human capital development and future designed to narrow inequality and improve income opportunities. By smoothing con- the living conditions of the poor in a variety sumption among the most deprived, espe- of settings. Yet, the long road ahead argues cially during shocks, CCTs help prevent the against any complacency and against the fal- widening of inequalities. lacy of sweeping prescriptions that promise Improved competition and economic ef- success regardless of the diverse challenges, ficiency are compatible with reducing in- contexts, and uncertainties. Meaningful les- equalities. The example of investments in sons nonetheless emerge from the evidence new roads highlights the importance of im- of the policy interventions examined in this proving competition in programs aimed at chapter. benefiting the poorest. Investments in roads Despite progress, intolerable disparities in help expand transport volume and lower well-being still exist that concrete policy in- fares only if there is competition among terventions could confront directly. In many transport providers. Otherwise, the benefits low- and middle-income countries, pre- among the poor are seriously compromised. school enrollment rates among the poorest The resources offered by CCTs can offset quintile are less than a third the rates among numerous economic inefficiencies that, for the richest quintile. Mothers in the bottom instance, impede the acquisition of pro- 40 across developing countries are 50 per- ductive assets among poor farmers (such cent less likely to receive antenatal care. The as obstacles in the access to credit markets) poorest children are four times less likely and hamper investments in education and than the richest children to be enrolled in health care. Examples of successful efforts primary education. These children system- to achieve universal health care include ini- atically record lower tests scores than chil- tiatives that incentivize health providers to dren in the richest households, and this gap offer more competitive services to people exceeds 50 percentage points in some devel- who are excluded. oping countries. Among the estimated 780 Trade-offs in implementation should not million illiterate adults worldwide, nearly be overlooked because of excessive attention two-thirds are women. Only one-quarter to efficiency and equity trade-offs. There are of the poorest quintile are covered by safety numerous examples of equalizing policies nets, and the share is even smaller in Sub- that do not compromise economic growth. Saharan Africa and South Asia. Given the current gaps in access, invest- The equalizing effect of interventions must ments in ECD, universal health care, and be judged by the impacts they achieve, not good-quality education have both equity those originally (un)intended. Most interven- and efficiency benefits. Connecting poor tions register impacts on inequality whether farmers to urban markets can affect the in- unintended or deliberate. These effects can comes of farm households as well as income be large or small or short term or lifelong, gaps among a population. Indeed, in the ef- and they can widen or narrow disparities in fort to reduce inequality, policy choices are incomes, well-being, and opportunity. They less often restricted by an imbalance in the occur over multiple pathways. For example, equity-efficiency trade-off than by an imbal- taxes have direct and deliberate redistribu- ance in the trade-off between expanding the tive effects, reaching up to 20 points of the coverage of an intervention and increasing Gini index of market incomes in some EU the benefits; in the trade-off between en- countries. However, there are many other hancing the quality of services and increas-

152 POVERTY AND SHARED PROSPERITY 2016 ing access to services through the construc- been shown to generate wide-ranging ben- tion of facilities such as schools or clinics; efits. Investments in rural roads that attract in the trade-off between expanding the additional investments in public services coverage of electrification in rural areas and such as electrification, agricultural exten- ensuring program financial viability; in the sion services, and enhanced water and san- trade-off between cash or in-kind resource itation services improve not only the con- transfers; and in the tradeoff between condi- nectivity of people to opportunities to meet tionality and the lack of conditionality. Eco- basic needs, but also security, productivity, nomic growth and good macroeconomic and quality of services. Yet, composition is management contribute to circumventing also important if various interventions are such implementation policy trade-offs by combined. For example, at a given level of providing resources, stability, and opportu- revenue, fiscal systems in Eastern European nities to adopt appropriate policies. countries that depend on flat-rate income The fine points of policy design absolutely taxes sacrifice the redistributive power of matter in ensuring that interventions are progressive personal income taxes, while equalizing but do not compromise efficiency. yielding the same level of revenue collection. Different choices in Chile and Mexico in tax Simplicity and flexibility often drive suc- reforms with the same objectives led to dif- cess. Thus, the ability of the safety net sys- ferent impacts. The ultrarich bore the brunt tem in the Philippines to scale up to reach of the income tax component of the reform hundreds of thousands of beneficiaries after in Chile, while, in Mexico, the middle class catastrophic events is in part related to the also largely shared the cost of the reform. flexibility of the system in the face of emer- ECD programs are most effective if they are gency situations. CCTs are delivered uncon- aimed at the first 1,000 days of the lives of ditionally, thereby allowing quick delivery children, continue during childhood, and and the scaling-up of the benefits. Exclusive integrate psychosocial stimulation, parent- and prolonged breastfeeding is another ex- ing, and nutrition components. In many ample of a simple and extraordinarily cost- contexts, incentivizing higher quality in efficient intervention to improve ECD. teaching, while making social transfers con- Equalizing interventions are not a luxury ditional on school completion may have a reserved for middle- and high-income coun- greater impact than simply constructing a tries, nor an option only available during new school. Defining the package of services periods of prosperity. This chapter provides that are provided, the level of user contribu- numerous examples of the implementation tions in the financing of interventions, and of successful interventions in ECD, univer- the composition of the target population sal health care coverage, good-quality teach- are all critical to successful reforms aimed ing, CCTs, investment in rural infrastruc- at achieving universal health care coverage. ture, and redistributive tax schemes across Avoid unexamined reliance on universal low-income countries that should dispel the prescriptions and unique models of success. notion that only middle-income countries Evidence strongly suggests that the imple- can afford equalizing policies. Of course, mentation of such prescriptions and mod- context always matters: weak capacity, els do not automatically ensure a reduction lack of political will, restricted fiscal space, in inequality. Nonetheless, some initiatives vulnerability to external crises or climate are more likely than others to generate in- change, internal conflict, and challenging equality reductions and improvements in geography are among the obstacles to the the well-being of the poorest. reduction of inequality worldwide. They are Integrated interventions are more likely not insurmountable, however. This is also to succeed than isolated, monolithic inter- the case during periods of crises. Examples ventions, but composition influences the of CCTs integrated in safety nets that effec- degree of success. If CCTs are combined tively protect the most vulnerable against with other safety net interventions such as natural disasters, choices in the way fiscal transfers of productive assets, skills training, consolidation occurred in Europe during and access to credit and finance, they have the 2008–09 financial crisis, and the reform

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 153 of energy subsidies in some North African erment and address criminal behavior, risky countries demonstrate that crisis is not an sexual behavior, and adolescent pregnancy excuse for failure, but an incentive for the beyond the effects on incomes, employ- adoption of equalizing interventions. ment, and service use that are normally an- The poor must be able to participate in alyzed. Yet, the road ahead is still long and and benefit from interventions: good policy steep. Especially important is the generation choices benefit the poorest. Evidence from of more microeconomic household data, ECD programs, initiatives to promote uni- long-term evaluations and panel data, more versal health care coverage, and efforts to compelling evidence on the benefits of the foster good-quality teaching shows that the integration of multiple interventions, and most underprivileged children often ben- more information on the potential distribu- efit the most. Yet, this outcome should not tional effects of policy interventions aimed be taken for granted. Thus, the more well at addressing long-term challenges such as off households among the targeted pop- climate change. ulation, that is, households with children with higher baseline levels of development Notes and more well educated mothers, are typically more likely to send their children to 1. On each one of these policy areas, the preschool or to take part in parenting pro- World Bank has recently published com- grams. Many rural electrification initiatives prehensive reviews, identified sectoral are associated with high connection costs to strategies, or conducted evaluations. keep electrification campaigns financially 2. The approach borrows from the frame- feasible, but this often means the poorest work of Gill, Revenga, and Zeballos (2016). households must opt out. Likewise, the 3. Duncan et al. (2007); Georgieff (2007); immediate winners in new or rehabilitated Glewwe, Jacoby, and King (2001); rural road programs in South Asia are re- Grantham-McGregor et al. (2007); Ha- portedly people who already operate trans- nushek and Woessmann (2008); Heckman, port vehicles or who can afford to invest Pinto, and Savelyev (2013); Naudeau et al. in small motorized vehicles. Policy design (2011a, 2011b); Vegas and Santibáñez needs to take such outcomes into account (2010); Walker et al. 2007; World Bank up front and explicitly. (2009). More knowledge! Despite the growing 4. These results refer to data on around 2005 evidence on the impacts of policy interven- (Alderman 2011). tions, improving the evidence base on ini- 5. Other forms of undernutrition among chil- tiatives that successfully narrow inequality dren—wasting (low weight for height) and requires more investment in filling data gaps underweight—are more prevalent among and enhancing the understanding of the the poor by a wide margin worldwide. See specific pathways through which programs UNICEF (2016); World Bank (2015a). affect inequality (whether intended or unin- 6. Alderman (2011). tended). For example, rigorous evaluations 7. These estimated income benefits refer to have played a critical role in fine-tuning the men participants. Among women, the earn- design of CCTs and advocacy for the desir- ings were 22 percent higher than the earn- ability of CCTs. Monitoring ECD programs ings among the control group. See Heck- for decades has made the quantification man et al. (2010). of the long-term effects of such programs 8. Barnett and Masse (2007); Heckman et al. possible. The widespread use of random- (2010); Temple, Reynolds, and White (2007). ized controlled trials in new policy areas 9. Gillespie et al. (2016). facilitates the more precise measurement of 10. Other common ECD interventions include causal effects of interventions. The formu- pregnancy- and birth-related programs, lation of new indicators on, for example, which are discussed in the section on health learning or socioemotional development care from the perspective of mothers. has expanded our knowledge of the impacts 11. Gertler et al. (2014). of programs to promote women’s empow- 12. Grantham-McGregor et al. (1991).

154 POVERTY AND SHARED PROSPERITY 2016 13. Gertler et al. (2014). 45. Jamison et al. (2013). 14. Attanasio et al. (2014). 46. Wagstaff, Bredenkamp, and Buisman 15. Hamadani et al. (2006), cited in Alderman (2014). (2011). 47. For example, in India and the Islamic Re- 16. Berlinksi and Schady (2015); Eickmann public of Iran, the annual rate of decline in et al. (2003); Rosero and Oosterbeek (2011). adult mortality has been more than 1.0 and 17. Berlinksi and Schady (2015); Chang et al. 3.5 percentage points higher, respectively, (2015). among women than men. See Jamison et 18. Berlinksi and Schady (2015); Naudeau al. (2013). et al. (2011a). 48. “The World Health Organization’s In- 19. Murphy at al. (2015). fant Feeding Recommendation,” World 20. Alderman (2011); Reynolds et al. (2007). Health Organization, Geneva (accessed 21. Reynolds et al. (2007). June 2016), http://www.who.int/nutrition 22. Reynolds et al. (2007). /topics/infantfeeding_recommendation/en/. 23. Martinez, Naudeau, and Pereira (2012). 49. “The World Health Organization’s In- 24. Berlinski, Galiani, and Gertler (2006). fant Feeding Recommendation,” World 25. Naudeau et al. (2011a). Health Organization, Geneva (accessed 26. Araujo et al. (2016); Britto, Yoshikawa, and June 2016), http://www.who.int/nutrition Boller (2011); Chetty, Friedman, and Rock- /topics/infantfeeding_recommendation/en/. off (2014); Engle et al. (2011). 50. Wagstaff, Bredenkamp, and Buisman 27. Berlinski, Galiani, and Manacorda (2008). (2014). 28. Alderman (2011). 51. Bloom and Fisk (2013); Jamison et al. (2013). 29. Nadeau et al. (2011a). 52. The studies use iron deficiency, caloric in- 30. Berlinksi and Schady (2015). take, or height as proxies for nutrition. See 31. WHO and UNICEF (2003). Alleyne and Cohen (2002). 32. Naudeau et al. (2011a); Papp (2014); “The 53. Basta et al. (1979); Li et al. (1994); Thomas World Health Organization’s Infant Feed- et al. (2006). ing Recommendation,” World Health Or- 54. Baird et al. (2015); Miguel and Kremer ganization, Geneva (accessed June 2016), (2004). http://www.who.int/nutrition/topics 55. See Bleakley (2007). In addition, attempts /infantfeeding_recommendation/en/. to quantify the aggregate benefits from 33. The results refer to Bangladesh, Brazil, The improved health care suggest that there Gambia, Ghana, Guinea-Bissau, India, Pa- are substantive gains in economic growth. kistan, Peru, the Philippines, and Senegal Thus, Jamison et al. (2013) report that re- (Sankar et al. 2015). The results disaggre- ductions in adult mortality might have ac- gated by boys and girls refer only to Brazil, counted for as much as 11 percent of the The Gambia, Ghana, Pakistan, the Philip- overall economic growth seen in low- and pines, and Senegal. See WHO (2000). middle-income countries in 1970–2000. 34. Horta and Victora (2013). These trends echo findings on developed 35. Victora et al. (2015). countries. For example, improvements 36. Kramer et al. (2008). in health and nutrition are estimated to 37. Der, Batty, and Deary (2008); Kramer et al. account for up to 30 percent of growth (2001, 2002, 2008); Oster (2015). in gross domestic product (GDP) in the 38. Berlinksi and Schady (2015). United Kingdom between 1780 and 1979. 39. Bhandari et al. (2003); Haider et al. (2000); 56. See Jamison et al. (2013). The interven- Morrow et al. (1999); Tylleskär et al. (2011). tions spanned reproductive, maternal, 40. Berlinksi and Schady (2015); Pérez- newborn, and child health care; pregnancy- Escamilla et al. (2012). related interventions (antenatal care, treat- 41. Hoddinott et al. (2011). ment of complications associated with 42. Black et al. (2008), cited in World Bank pregnancy, delivery interventions, and (2009). postpartum care); abortion and associated 43. Naudeau et al. (2011a). complications; family planning; diarrhea 44. Gillespie et al. (2016). management; pneumonia treatment; im-

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 155 munization; nutrition (breastfeeding and (2008, 2010); Lucas (1988); Rebelo (1991); supplementation); HIV prevention activities Romer (1990) ; Schultz (1961). (community mobilization and work with 77. Hanushek and Woessmann. (2010); World specific groups such as intravenous drug Bank (2011a). users and men who have sex with men); the 78. Mittal and Tripathi (2009); Rosenzweig management of opportunistic infections, and Foster (2010). care, and treatment; collaborative tubercu- 79. See Yang (1997) on China; Jolliffe (1998) losis–HIV treatment; malaria (treatment on Ghana; Fafchamps and Quisumbing with appropriate drugs for adults, children, (1999) on Pakistan. pregnant women, and those with severe ma- 80. See Gakidou et al. (2010) on the associa- laria; indoor residual spraying; long-lasting tion between education and good health; insecticide-treated mosquito nets; intermit- Frankensberg, Smith, and Thomas (2003) tent presumptive treatment in pregnancy); and Corbacho, Garcia-Escribano, and In- tuberculosis (diagnosis, care, and treatment chauste (2007) on the resilience to eco- of drug-sensitive tuberculosis; diagnosis, nomic shocks. care, and treatment of multidrug-resistant 81. Effectiveness in teaching is often measured tuberculosis); neglected tropical diseases; by the value added, that is, student test and community-directed interventions to scores. See Araujo et al. (2016). control lymphatic filariasis, onchocerciasis, 82. Araujo et al. (2016). schistosomiasis, trachoma, and soil-trans- 83. Heckman and Kautz (2012). mitted helminths. 84. Araujo et al. (2016); Bau and Das (2016); 57. See Jamison et al. (2013). Most of the in- Chetty, Friedman, and Rockoff (2014). cremental investment costs would be as- 85. Araujo et al. (2016). sociated with improving health care sys- 86. See Bau and Das (2016). Some evidence tems, accounting for 70 percent of all costs gathered in schools in the United States in the first 10 years and 60 percent in the suggests there is no causal link between second 10 years. Such investments would teacher incentives and student outcomes. lead to functional health care systems that See Fryer (2013); Springer et al. (2012). can tackle long-term health challenges, not 87. Teacher behavior is measured by Class- merely temporary improvements in infec- room Assessment Scoring System scores, tious disease or mortality rates. a protocol that measures teacher behav- 58. UNICEF (2016). ior in three domains: emotional support, 59. UNICEF (2016). classroom organization, and instruction 60. UNICEF (2016). support. Each domain captures different 61. UNICEF (2016). dimensions, such as positive or negative 62. UNICEF (2016). classroom atmosphere, teacher sensitiv- 63. UNICEF (2016). ity, and regard for the student perspective. 64. UNICEF (2016). Each dimension is given a score of 1–6, 65. Gustafsson-Wright and Schellekens (2013). from which a total score is obtained. (See 66. UNICEF (2016). Araujo et al. 2016.) The attention, self- 67. Reviewed in Wagstaff et al. (2016). control, and memory effects are distinct 68. Wagstaff et al. (2016). from the effects of classroom atmosphere 69. Wagstaff et al. (2016). and parental influence. 70. Wagstaff et al. (2016). 88. Higgins et al. (2015). See also Teaching and 71. UNESCO (2000). Learning Toolkit (database), Education 72. World Bank (2016a). Endowment Foundation, London, https:// 73. See ASER Center (2011); Crouch (2006); educationendowmentfoundation.org.uk Das, Pandey, and Zajonc (2006). /evidence/teaching-learning-toolkit/. 74. World Bank (2016a). 89. Dobbie and Fryer (2013); Duflo, Dupas, 75. World Bank (2005). and Kremer (2011); Kane and Staiger 76. See Becker (1962); Becker, Murphy, and (2008); Muralidharan and Sundararaman Tamura (1990); Hanushek and Woessmann (2011).

156 POVERTY AND SHARED PROSPERITY 2016 90. See He, Linden, and MacLeod (2009); 106. Orbeta et al. (2014). Lopez-Acevedo (1999). The findings on the 107. World Bank (2011b). program in Mexico also suggest that dou- 108. Carvalho et al. (2015); Lim et al. (2010). bling the resources allocated per student 109. Akresh, de Walque, and Kazianga. (2013). could overcome a 30 percent deficit in test 110. Galasso (2006). scores among rural students. However, a 111. Fizsbein and Schady (2009), citing evi- meta-study of 30 primary-school interven- dence from Schady and Araujo (2008). tions associated with physical infrastruc- 112. See Schady and Araujo (2008) for an eval- ture (textbooks, improved buildings) in de- uation of the Bono de Desarrollo in Ecua- veloping countries finds that only one-third dor; Macours, Schady, and Vakis (2008) on had a significant impact on test scores. See Atención a Crisis and Maluccio and Flores Kremer, Brannen, and Glennerster (2013). (2005) on Red de Protección Social (the 91. Angrist and Krueger (1991); Black, De- latter two in Nicaragua). The benefits of vereux, and Salvanes (2008); Oreopoulos the programs are limited to enrollments (2006a, 2006b). and additional years of schooling, with no 92. See Global Education Monitoring Report gains in the quality of learning, but some in (database), United Nations Educational, cognitive abilities. The lack of quality gains Scientific, and Cultural Organization, is also supported by the small increase in Paris, http://en.unesco.org/gem-report/; wages among students who have bene- SABER (Systems Approach for Better Ed- fited from CCT-aided school enrollments. ucation Results) (database), World Bank, Behrman, Parker, and Todd (2009) find Washington, DC, http://saber.worldbank that children exposed to the Prospera Pro- .org/index.cfm; WIDE (World Inequality gram (the rebranded Oportunidades Pro- Database on Education), United Nations gram) for two more years earn wages that Educational, Scientific, and Cultural Or- are about 2 percent higher than the wages ganization, Paris, http://www.education earned by other children. -inequalities.org/. 113. See Baird, McIntosh, and Özler (2009). If 93. Gill, Revenga, and Zeballos (2016). evaluations fail to find significant impacts 94. Gill, Revenga, and Zeballos (2016). on test scores, it is often argued that this 95. World Bank (2016b). might be the result of programs that bring 96. Ferré and Sharif (2014). students with less ability back to school. 97. Behrman and Hoddinott (2005); Gertler See Filmer and Schady (2009). (2004). 114. Evans and Popova (2014). 98. Maluccio and Flores (2005). 115. Chaudhury, Friedman, and Onishi (2014). 99. Pellerano et al. (2014). 116. Chioda, de Mello, and Soares (2012); Hid- 100. According to Paxson and Schady (2010), robo et al. (2012); Walker et al. (2011). physical outcomes that include improved 117. Stecklov et al. (2007). hemoglobin, height, and fine motor con- 118. Brazil: 0.44 percent of GDP (2011); Mexico: trol were 16 percent more likely among 0.22 percent (2010); the Philippines: 0.40 children in the bottom income quintile of percent (2013); Colombia: 0.35 percent households who were beneficiaries of the (2010); Bangladesh 0.09 percent (2009) program. In contrast, the program had no (World Bank 2015b). statistically significant impact on children 119. This represents the unweighted average of in higher-income household quintiles that 17 countries on which incidence analysis also received CCTs. is available and in which the CCTs are not 101. Ahmed and Morgan (2011). delivered through pilot programs. Bolivia 102. Glassman et al. (2013); Lagarde, Haines, and Uruguay, where CCTs cover about and Palmer (2009). two-thirds of the poorest quintile, are ex- 103. Fiszbein and Schady (2009), citing Hernán- ceptions. See World Bank (2015b). dez et al. (2005). 120. The value of the average benefit relative to 104. Gertler and Boyce (2001). the average income of the poorest house- 105. Perova and Vakis (2012). holds varies widely, from 2.8 percent in

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 157 Bangladesh to 22.5 percent in Mexico. See 139. Asher and Novosad (2016) referring to evi- World Bank (2015b). dence in India. 121. Robalino, Rawlings, and Walker (2012). 140. Khandker, Bakht, and Koolwal (2009). 122. See Gentilini (2014) for a detailed 141. Lokshin and Yemtsov (2005). discussion. 142. Aggarwal (2015); Asher and Novosad 123. See Bourguignon, Ferreira, and Leite (2016). (2003). Todd and Wolpin (2006) estimate 143. Mu and van de Walle (2011). that incremental increases in the size of 144. Khandker, Bakht, and Koolwal (2009). transfers in Mexico would have diminish- 145. Lokshin and Yemtsov (2009). ing effects on educational attainment. 146. See Hettige (2006). In other cases, safe- 124. Filmer and Schady (2009). guards are needed to ensure the more well 125. Baird, McIntosh, and Özler (2009). off in rural communities do not beneﬁt 126. de Brauw and Hoddinott (2008); Schady disproportionately from new roads. In case and Araujo (2008). studies in Indonesia, the Philippines, and 127. Bourguignon, Ferreira, and Leite (2003); Sri Lanka, the direct winners of new or im- Todd and Wolpin (2006). proved rural roads were people who already 128. World Bank (2015b). These impacts refer operated transport vehicles or who were to Burkina Faso’s Cash Transfers to Or- able to afford to invest in smaller motorized phans and Vulnerable Children Program, three-wheelers. The more well educated, Lesotho’s Child Grant Program, Kenya’s the more well informed, and people own- GiveDirect, and Zambia’s Child Grant ing bicycles were able to obtain better wage Program. opportunities because of the improved in- 129. Aker et al. (2013). frastructure. Relative to the extreme poor, 130. CGAP (2011). the more well off were twice as likely to 131. Aker et al. (2013); Muralidharan, Niehaus, start or expand small businesses following and Sukhtankar (2014); Omamo, Gentilini, the improvements. and Sandström (2010); Vincent and Cull 147. See Hettige (2006). (2011). 148. Hettige (2006). 132. Berlinksi and Schady (2015). 149. Dinkelman (2011); Grogan and Sadanand 133. Banerjee et al. (2015). (2013). 134. Hallegatte et al. (2016). 150. van de Walle et al. (2013). 135. Calderón and Chong (2004); Calderón and 151. Khandker, Barnes, and Samad (2013). Servén (2004, 2008); Estache, Foster, and 152. Khandker, Barnes, and Samad (2013). Wodon (2002); Ndulu (2006); Seneviratne 153. Ghuman and Llloyd (2007). and Sun (2013); World Bank (2005). 154. See Barron and Torero (2015). Particle 136. Based on a sample of 100 countries, pollution is measured according to the

Calderón and Servén (2008) attribute to amount of PM2.5—or fine inhalable parti- infrastructure in, respectively, South Asia cles with diameters generally 2.5 microm- and Sub-Saharan Africa 2.7 percentage eters or smaller—that is found in the air points and 0.7 percentage points of annual (definition of the U.S. Environmental Pro- GDP growth in 2001–05 and 6 points and tection Agency). 2 points of the reduction in the regional 155. Dinkelman (2011). Gini. 156. Clark (2008); Dinkelman (2011); Lipscomb, 137. World Bank (2015c). Mobarak, and Barham (2013); Moser 138. These benefits are not automatic, how- and Holland (1997); World Bank (2011a, ever. Asher and Novosad (2016) describe 2011c). a successful case of labor reallocation out 157. Jensen and Oster (2009). of agriculture in India, while Banerjee, 158. LaFerrara, Chong, and Duryea (2012). Duflo, and Qian (2012) present a less suc- 159. De Agostini, Paulus, and Tasseva (2015); cessful case of resource reallocation follow- Manuel (2002); Mideplan (1999); Vivian ing rural road construction in China, and (2006). Bryan, Chowdhury, and Mobarak (2014) 160. See Levy (2008). Cuesta and Oliveira do the same in Bangladesh. (2014) find similar effects in Colombia.

158 POVERTY AND SHARED PROSPERITY 2016 161. Some countries, such as Colombia, Na- a reduction of over 7 points in the market mibia, the Russian Federation, South Af- income Gini, is an exception among mod- rica, and Uruguay, collect more than 1 erate redistributive effects. Redistribution percent of GDP as recurrent property taxes is associated with a reduction in the market (Broadway, Chamberlain, and Emmerson income Gini of about 1–3 points in Ethio- 2010; de Ferranti et al. 2004; IMF 2014; pia, Jordan, and Sri Lanka; the reductions Martinez-Vasquez 2008; World Bank 2005). are larger in Georgia and Russia. See In- 162. Lustig (2015). chauste and Lustig (forthcoming); Lustig 163. Direct taxes cause the more well off to bear (2015); Younger, Osei-Assibey, and Op- the brunt, while indirect taxes cause the pong (2015). poor to bear a larger relative share of the 174. See IMF (2014). For example, personal burden, given that the poor spend a higher income taxation in developing countries share of their incomes on consumption. raises an average 1–3 percent of GDP, com- 164. Bird and Zolt (2003). pared with 9–11 percent of GDP among 165. Sahn and Younger (2000). advanced countries. 166. Yet, large country variations are observed 175. Country-by-country comparative results across the EU. Reductions in market in- need to viewed with caution, however. They equalities are large in Western Europe, are derived based on different methodolo- but much more limited in the Baltic States gies (incidence analysis versus ex ante simu- (Avram, Levy, and Sutherland 2014; De lations), the indicator analyzed (income or Agostini, Palaus, and Tasseva 2015). consumption), conventions (the treatment 167. See Martinez-Vasquez, Vulovic, and of pension income), and the categories in- Moreno-Dodson (2014). Taxes are respon- cluded (health care and education spend- sible for an average increase of 1.5 percent ing). This implies that the analyses cover in the Gini index of market incomes in the different (but always incomplete) segments country sample since 1990. of the ﬁscal system. See Inchauste and 168. Bird and Zolt (2003, 2008). Lustig (forthcoming). 169. According to the deﬁnition of Ravallion (2009), poor countries show consumption References per capita under US$2,000 a year (2005 purchasing power parity [PPP] U.S. dol- Abramovsky, Laura, Orazio Attanasio, Kai lars). However, the more well off developing Barron, Pedro Carneiro, and George Stoye. countries—countries with consumption 2014. “Challenges to Promoting Social In- per capita at US$4,000 per year—would re- clusion of the Extreme Poor: Evidence from quire little additional taxation on the rich a Large Scale Experiment in Colombia.” IFS to eliminate extreme poverty. (Additional Working Paper W14/33, Institute for Fiscal marginal tax rates between 1 percent and 6 Studies, London. percent would be required.) In each coun- Aggarwal, Shilpa. 2015. “Do Rural Roads Create try, a rich household is a household that, Pathways out of Poverty? Evidence from in the United States, would be categorized India.” Working paper, Indian School of Busi- as nonpoor. No economic distortions, be- ness, Hyderabad, India. havioral changes, or political economy or Ahmed, Shakil, and Chris Morgan. 2011. administrative considerations are taken “Demand-Side Financing for Maternal into account in the analysis, which includes Health Care: The Current State of Knowledge 90 countries on which there are data for on Design and Impact.” Issues Brief 1 (Sep- around the late 2000s. tember), Nossal Institute for Global Health, 170. Hoy and Sumner (2016). University of Melbourne, Melbourne. 171. Ostry, Berg, and Tsangarides (2014). Aker, Jenny, Rachid Boumnijel, Amanda 172. Inchauste and Lustig (forthcoming); Lustig McClelland, and Niall Tierney. 2013. “How (2015). Do Electronic Transfers Compare? Evidence 173. Redistributive effects range from less than from a Mobile Money Cash Transfer Experi- 1 point to over 4 points (in Brazil) of the ment in Niger.” Technical Report, Tufts Uni- Gini for market incomes. South Africa with versity, Medford, MA.

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 159 Akresh, Richard, Damien de Walque, Harounan reau of Economic Research, Cambridge, MA. Kazianga. 2013. “Cash Transfers and Child http://www.nber.org/papers/w21428. Schooling: Evidence from a Randomized Baird, Sarah, Craig McIntosh, and Berk Özler. Evaluation of the Role of Conditionality.” 2009. “Designing Cost-Effective Cash Trans- Policy Research Working Paper 6340, World fer Programs to Boost Schooling among Bank, Washington, DC. Young Women in Sub-Saharan Africa.” Policy Alderman, Harold, ed. 2011. No Small Matter: Research Working Paper 5090, World Bank, The Impact of Poverty, Shocks, and Human Washington, DC. Capital Investments in Early Childhood Devel- Banerjee, Abhijit, Esther Duflo, Nathanael Gold- opment. Human Development Perspectives berg, Dean Karlan, Robert Osei, William Series. Washington, DC: World Bank. Parienté, Jeremy Shapiro, Bram Thuysbaert, Alleyne, George A. O., and Daniel Cohen. 2002. and Christopher Udry. 2015. “A Multifaceted “Health, Economic Growth, and Poverty Program Causes Lasting Progress for the Very Reduction: The Report of Working Group 1 Poor: Evidence from Six Countries.” Science of the Commission on Macroeconomics 348 (6236): 772–89. and Health.” World Health Organization, Ge- Banerjee, Abhijit, Esther Duflo, and Nancy Qian. neva. http://whqlibdoc.who.int/publications 2012. “On the Road: Access to Transporta- /9241590092.pdf. tion Infrastructure and Economic Growth Angrist, Joshua D., and Alan B. Krueger. 1991. in China.” NBER Working Paper 17897, Na- “Does Compulsory School Attendance Affect tional Bureau of Economic Research, Cam- Schooling and Earnings?” Quarterly Journal bridge, MA. of Economics 106 (4): 979–1014. Barnett, W. Steven, Leonard N. Masse. 2007. Araujo, María Caridad, Pedro Carneiro, Yyannú “Comparative Benefit-Cost Analysis of the Cruz-Aguayo, and Norbert Schady. 2016. Abecedarian Program and Its Policy Implica- “Teacher Quality and Learning Outcomes in tions.” Economics of Education Review 26 (1): Kindergarten.” Quarterly Journal of Econom- 113–25. ics 125 (1): 175–214. Barron, Manuel, and Maximo Torero. 2015. ASER Center. 2011. “Annual Status of Education “Fixed Costs, Spillovers, and Adoption of Report (Rural) 2010.” Pratham Resource Cen- Electric Connections.” MPRA Paper 63804, ter, Mumbai. http://img.asercentre.org/docs University Library of Munich, Munich. /Publications/ASER%20Reports/ASER_2010 Basta, Samir S., Soekirman, Darwin Karyadi, /ASERReport2010.pdf. Nevin S. Scrimshaw. 1979. “Iron Deficiency Asher, Sam, and Paul Novosad. 2016. “Market Ac- Anemia and the Productivity of Adult Males cess and Structural Transformation: Evi dence in Indonesia.” American Journal of Clinical from Rural Roads in India.” Working paper Nutrition 32 (4): 916–25. (April 20), University of Oxford, Oxford. Bau, Natalie, and Jishnu Das. 2016. “The Mis- Attanasio, Orazio P., Camila Fernández, Emla allocation of Pay and Productivity in the O. A. Fitzsimons, Sally M. Grantham- Public Sector: Evidence from the Labor McGregor, Costas Meghir, and Marta Rubio- Market for Teachers.” Paper presented at the Codina. 2014. “Using the Infrastructure of a United Nations University–World Institute Conditional Cash Transfer Program to De- for Development Economics Research’s De- liver a Scalable Integrated Early Child De- velopment Conference, “Human Capital and velopment Program in Colombia: Cluster Growth,” Helsinki, June 6–7. Randomized Controlled Trial.” BMJ 349 (Sep- Becker, Gary S. 1962. “Investment in Human tember 29): g5785. Capital: A Theoretical Analysis.” Journal of Avram, Silvia, Horacio Levy, and Holly Suther- Political Economy 70 (5, part 2): 9–49. land. 2014. “Income Redistribution in the Becker, Gary S., Kevin M. Murphy, and Robert European Union.” IZA Journal of European Tamura. 1990. “Human Growth, Fertility, Labor Studies 3 (22): 1-29. and Economic Growth.” Journal of Political Baird, Sarah, Joan Hicks, Michael Kremer, and Economy 98 (5): S12–S37. Edward Miguel. 2015. “Worms at Work: Behrman, Jere R., and John Hoddinott. 2005. Long-Run Impacts of a Child Health Invest- “Programme Evaluation with Unobserved ment.” Working Paper 21428, National Bu- Heterogeneity and Selective Implementation:

160 POVERTY AND SHARED PROSPERITY 2016 The Mexican PROGRESA Impact on Child American South.” Quarterly Journal of Eco- Nutrition.” Oxford Bulletin of Economics and nomics 122 (1): 73–117. Statistics 67 (4): 547–69. Bloom, David E., and Günther Fink. 2013. “The Behrman, Jere R., Susan W. Parker, and Petra E. Economic Case for Devoting Public Re- Todd. 2009. “Schooling Impacts of Con- sources to Health.” PGDA Working Paper ditional Cash Transfers on Young Children: 104, Program on the Global Demography of Evidence from Mexico.” Economic Develop- Aging, Harvard University, Cambridge, MA. ment and Cultural Change 57 (3): 439–77. Bourguignon, Francois, Francisco Ferreira, and Berlinski, Samuel, Sebastian Galiani, and Paul Phillippe Leite. 2003. “Conditional Cash J. Gertler. 2006. “The Effect of Pre-primary Transfers, Schooling, and Child Labor: Micro- Education on Primary School Performance.” Simulating Brazil’s Bolsa Escola Program.” IFS Working Paper W06/04, Institute for Fis- World Bank Economic Review 17 (2): 229–54. cal Studies, London. Britto, Pia Rebello, Hirokazu Yoshikawa, and Berlinski, Samuel, Sebastian Galiani, and Marco Kimberly Boller. 2011. “Quality of Early Manacorda. 2008. “Giving Children a Bet- Childhood Development Programs in Global ter Start: Preschool Attendance and School- Contexts: Rationale for Investment, Con- Age Profiles.” Journal of Public Economics 92 ceptual Framework, and Implications for (5–6): 1416–40. Equity.” Society for Research in Child Develop- Berlinski, Samuel, and Norbert Schady, eds. ment Social Policy Report 25 (2): 1–30. 2015. The Early Years: Child Well-Being and Broadway, Robin, Emma Chamberlain, and Carl the Role of Public Policy. Development in Emmerson. 2010. “Taxation of Wealth and the Americas Series. Washington, DC: Inter- Wealth Transfers.” In Dimensions of Tax De- American Development Bank; New York: sign: The Mirrlees Review, edited by Stuart Palgrave Macmillan. Adam, Tim Besley, Richard Blundell, Steve Bhandari, Nita, Rajiv Bahl, Sarmila Mazumdar, Bond, Robert Chote, Malcolm Gammie, Paul Jose Martines, Robert E. Black, Maharaj K. Johnson, Gareth Myles, and James Poterba, Bhan, and Infant Feeding Study Group. 2003. 737–824. London: Institute for Fiscal Studies; “Effect of Community-Based Promotion of Oxford: Oxford University Press. Exclusive Breastfeeding on Diarrhoeal Illness Bryan, Gharad, Shyamal K. Chowdhury, and and Growth: A Cluster Randomised Con- Ahmed Mushfiq Mobarak. 2014. “Under- trolled Trial.” Lancet 361 (9367): 1418–23. investment in a Profitable Technology: The Bird, Richard, and Eric Zolt. 2003. “Introduc- Case of Seasonal Migration in Bangladesh.” tion to Tax Policy Design and Development.” Draft prepared for “A Course on Practical Econometrica 82 (5): 1671–1748. Issues of Tax Policy in Developing Coun- Calderón, César A., and Alberto Chong. 2004. tries,” World Bank, Washington, DC, April “Volume and Quality of Infrastructure and 28–May 1. the Distribution of Income: An Empirical In- ———. 2008. “Technology and Taxation in De- vestigation.” Review of Income and Wealth 50 veloping Countries: From Hand to Mouse.” (1): 87-106. National Tax Journal 61 (4): 791–821. Calderón, César A., and Luis Servén. 2004. Black, Robert E., Lindsay H. Allen, Zulfiqar A. “Trends in Infrastructure in Latin America.” Bhutta, Laura E. Caulfield, Mercedes de Policy Research Working Paper 3401, World Onis, Majid Ezzati, Colin Mathers, and Juan Bank, Washington, DC. Rivera. 2008. “Maternal and Child Undernu- ———. 2008. “Infrastructure and Economic trition: Global and Regional Exposures and Development in Sub-Saharan Africa.” Policy Health Consequences.” Lancet 371 (9608): Research Working Paper 4712, World Bank, 243–60. Washington, DC. Black, Sandra E., Paul J. Devereux, and Kjell G. Carvalho, Natalie, Naveen Thacker, Subodh S. Salvanes. 2011. “Too Young to Leave the Nest? Gupta, and Joshua A. Salomon. 2015. “More The Effects of School Starting Age.” Review of Evidence on the Impact of India’s Condi- Economics and Statistics 93 (2): 455–67. tional Cash Transfer Program, Janani Surak- Bleakley, Hoyt. 2007. “Disease and Development: sha Yojana: Quasi-Experimental Evaluation Evidence from Hookworm Eradication in the of the Effects on Childhood Immunization

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 161 and Other Reproductive and Child Health Labor Market: The Case of Colombia.” Jour- Outcomes.” PLoS ONE 9 (10): e109311. nal of Policy Modeling 36 (6): 1118–34. CGAP (Consultative Group to Assist the Poor). Das, Jishnu, Priyanka Pandey, and Tristan Za- 2011. “CGAP G2P Research Project: South jonc. 2006. “Learning Levels and Gaps in Pa- Africa Report.” CGAP, Washington, DC. kistan.” Policy Research Working Paper 4067, Chang, Susan M., Sally M. Grantham-McGregor, World Bank, Washington, DC. Christine A. Powell, Marcos Vera-Hernández, De Agostini, Paola, Alari Paulus, and Iva Florencia Lopez-Boo, Helen Baker- Valentinova Tasseva. 2015. “The Effect of Henningham, and Susan P. Walker. 2015 Tax-Benefit Changes on the Income Distri- “Integrating a Parenting Intervention with bution in 2008–2014.” Euromod Working Routine Primary Health Care: A Cluster Ran- Paper EM11/15, Institute for Social and Eco- domized Trial.” Pediatrics 136 (2): 272–80. nomic Research, University of Essex, Col- Chaudhury, Nazmul, Jed Friedman, and Junko chester, United Kingdom. Onishi. 2014. “Philippines Conditional Cash de Brauw, Alan, and John Hoddinott. 2008. Transfer Program: Impact Evaluation 2012.” “Must Conditional Cash Transfer Programs Report 75533-PH, World Bank, Washington, Be Conditioned to Be Effective? The Impact DC. of Conditioning Transfers on School Enroll- Chetty, Raj, John N. Friedman, and Jonah E. ment in Mexico.” IFPRI Discussion Paper Rockoff. 2014. “Measuring the Impacts of 757, International Food Policy Research In- Teachers: Teacher Value-Added and Student stitute, Washington, D.C. Outcomes in Adulthood.” American Eco- de Ferranti, David, Guillermo E. Perry, Francisco nomic Review 104 (9): 2593–2679. H. G. Ferreira, and Michael Walton. 2004. Chioda, Laura, João Manoel P. de Mello, and Inequality in Latin America: Breaking with Rodrigo Soares. 2012. “Spillovers from Con- History? World Bank Latin American and ditional Cash Transfer Programs: Bolsa Caribbean Studies Series. Washington, DC: Família and Crime in Urban Brazil.” World World Bank. Bank, Washington, DC. Der, Geoff, G. David Batty, and Ian J. Deary. Clarke, Ronald. 2008. “Improving Street Light- 2008. “Results from the PROBIT Breastfeed- ing to Reduce Crime in Residential Areas: ing Trial May Have Been Overinterpreted.” Problem-Oriented Guides for Police.” Re- Archives of General Psychiatry 65 (12): sponse Guides Series 8, Office of Commu- 1456–57. nity Oriented Policing Services, Center for Dinkelman, Taryn. 2011. “The Effects of Rural Problem-Oriented Policing, U.S. Department Electrification on Employment: New Evi- of Justice, Washington, DC. dence from South Africa.” American Economic Clements, Benedict, Ruud de Mooij, Sanjeev Review 101 (7): 3078–3108. Gupta, and Michael Keen, eds. 2015. Inequal- Dobbie, Will, and Roland G. Fryer, Jr. 2013. “Get- ity and Fiscal Policy. Washington, DC: Inter- ting Beneath the Veil of Effective Schools: Ev- national Monetary Fund. idence From New York City.” American Eco- Corbacho, Ana, Mercedes Garcia-Escribano, nomic Journal: Applied Economics 5 (4): 28–60. and Gabriela Inchauste. 2007. “Argentina: Duflo, Esther, Pascaline Dupas, and Michael Macroeconomic Crisis and Household Vul- Kremer. 2011. “Peer Effects, Teacher Incen- nerability.” Review of Development Economics tives, and the Impact of Tracking: Evidence 11 (1): 92–106. from a Randomized Evaluation in Kenya.” Crouch, Luis. 2006. “Education Sector: Stan- American Economic Review 101 (5): 1739–74. dards, Accountability, and Support.” In A Duncan, Greg J., Chantelle J. Dowsett, Amy New Social Contract for Peru: An Agenda for Claessens, Katherine Magnuson, Aletha C. Improving Education, Health Care, and the Huston, Pamela Klebanov, Linda S. Pagani Social Safety Net, edited by Daniel Cotlear, et al. 2007. “School Readiness and Later 71–106. World Bank Country Study Series. Achievement.” Developmental Psychology 43 Washington, DC: World Bank. (6): 1428–46. Cuesta, Jose, and Mauricio Olivera. 2014. “The Eickmann, Sophie H., Ana C. V. Lima, Miriam Impact of Social Security Reform on the Q. Guerra, Marilia C. Lima, Pedro I. C. Lira,

162 POVERTY AND SHARED PROSPERITY 2016 Sharon R. Huttly, and Ann Ashworth. 2003. Wealth, and Welfare.” Journal of Human Re- “Improved Cognitive and Motor Develop- sources 38 (2): 280–321. ment in a Community-Based Intervention of Fryer, Roland G. 2013. “Teacher Incentives and Psychosocial Stimulation in Northeast Bra- Student Achievement: Evidence from New zil.” Developmental Medicine and Child Neu- York City Public Schools.” Journal of Labor rology 45 (8): 536–41. Economics 31 (2): 373–407. Engle, Patrice L., Lia C. H. Fernald, Harold Gakidou, Emmanuela, Krycia Cowling, Rafael Alderman, Jere Behrman, Chloe O’Gara, Lozano, and Christopher J.L. Murray. 2010. Aisha Yousafzai, Meena Cabral de Mello, “Increased Educational Attainment and Its Melissa Hidrobo, Nurper Ulkuer, Ilgi Ertem, Effect on Child Mortality in 175 Countries Semil Iltus, and the Global Child Develop- between 1970 and 2009: A Systematic Anal- ment Steering Group. 2011. “Strategies for ysis.” Lancet 376 (9745): 959–74. Reducing Inequalities and Improving Devel- Galasso, Emanuela. 2006. “With Their Effort and opmental Outcomes for Young Children in One Opportunity: Alleviating Extreme Pov- Low-Income and Middle-Income Countries.” erty in Chile.” Unpublished working paper, Lancet 378 (9799): 1339–53. World Bank, . Washington, DC. Estache, Antonio, Vivien Foster, and Quentin Gentilini, Ugo. 2014. “Our Daily Bread: What Wodon. 2002. “Accounting for Poverty in Is the Evidence on Comparing Cash versus Infrastructure Reform: Learning from Latin Food Transfers?” Social Protection and Labor America’s Experience.” World Bank Institute Discussion Paper 1420, World Bank, Wash- Development Studies, World Bank, Washing- ington, DC. ton, DC. Georgieff, Michael. 2007. “Nutrition and the Evans, David, and Anna Popova. 2014. “Cash Developing Brain: Nutrient Priorities and Transfers and Temptation Goods: A Review Measurement.” American Journal for Clinical of Global Evidence.” Policy Research Working Nutrition 85 (2): 614s–620s. Paper 6886, World Bank, Washington, DC. Gertler, Paul J. 2004. “Do Conditional Cash Fabrizio, Stefania, and Valentina Flamini. 2015. Transfers Improve Child Health? Evidence “Fiscal Consolidation and Income Inequal- from PROGRESA’s Control Randomized Ex- ity.” In Inequality and Fiscal Policy, edited by periment.” American Economic Review 94 (2): Benedict Clements, Ruud de Mooij, Sanjeev 336–41. Gupta, and Michael Keen, 159–74. Washing- Gertler, Paul J., and Simone Boyce. 2001. “An ton, DC: International Monetary Fund. Experiment in Incentive-Based Welfare: The Fafchamps, Marcel, and Agnes R. Quisumbing. Impact of Progresa on Health in Mexico.” 1999. “Human Capital, Productivity, and Working paper, Haas School of Business, Labor Allocation in Rural Pakistan.” Journal University of California, Berkeley, CA. of Human Resources 34 (2): 369–406. Gertler, Paul J., James J. Heckman, Rodrigo Ferré, Céline, and Iffath Sharif. 2014. “Can Con- Pinto, Arianna Zanolini, Christel Vermeersch, ditional Cash Transfers Improve Education Susan Walker, Susan M. Chang, and Sally and Nutrition Outcomes for Poor Children Grantham-McGregor. 2014. “Labor Market in Bangladesh? Evidence from a Pilot Proj- Returns to an Early Childhood Stimulation ect.” Policy Research Working Paper 7077, Intervention in Jamaica.” Science 344 (6187): World Bank, Washington, DC. 998–1001. Filmer, Deon, and Norbert Schady. 2009. “Are Ghuman, Sharon, and Cynthia B. Lloyd. 2007. There Diminishing Returns to Transfer Size “Poverty, Gender, and Youth: Teacher Ab- in Conditional Cash Transfers?” Policy Re- sence as a Factor in Gender Inequalities in search Working Paper 4999, World Bank. Access to Primary Schooling in Rural Paki- Washington, DC. stan.” Population Council Working Paper 1, Fiszbein, Ariel, and Norbert Schady. 2009. Con- Population Council, New York. ditional Cash Transfers: Reducing Present and Gill, Indermit, Ana Revenga, and Christian Ze- Future Poverty. Washington, DC: World Bank. ballos. 2016. Grow, Invest, Insure: A Game Frankenberg, Elizabeth, James Smith, and Plan to End Poverty. Washington, DC: World Duncan Thomas. 2003. “Economic Shocks, Bank.

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 163 Gillespie, Stuart, Judith Hodge, Sivan Yosef, and and Adrien Vogt-Schilb. 2016. Shock Waves: Rajul Pandya-Lorch, eds. 2016. Nourish- Managing the Impacts of Climate Change on ing Millions: Stories of Change in Nutrition. Poverty. Climate Change and Development Washington, DC: International Food Policy Series. Washington, DC: World Bank. Research Institute. Hamadani, Jena D., Syed N. Huda, Fahmida Glassman, Amanda, Denizhan Duran, Lisa Khatun, and Sally M. Grantham-McGregor. Fleisher, Daniel Singer, Rachel Sturke, 2006. “Psychosocial Stimulation Improves Gustavo Angeles, Jodi Charles, et al. 2013. the Development of Undernourished Chil- “Impact of Conditional Cash Transfers on dren in Rural Bangladesh.” Journal of Nutri- Maternal and Newborn Health.” Journal of tion 136 (10): 2645–52. Health, Population, and Nutrition 31 (4 Suppl Hanushek, Eric A., and Ludger Woessmann. 2): S48–S66. 2008. “The Role of Cognitive Skills in Eco- Glewwe, Paul, Hanan G. Jacoby, and Elizabeth nomic Development.” Journal of Economic M. King. 2001. “Early Childhood Nutrition Literature 46 (3): 607–68. and Academic Achievement: A Longitudinal ———. 2010. “The Economics of International Study.” Journal of Public Economics 81 (3): Differences in Educational Achievement.” 345–68. NBER Working Paper 15949, National Bu- Grantham-McGregor, Sally M., Yin Bun Cheung, reau of Economic Research, Cambridge, MA. Santiago Cueto, Paul Glewwe, Linda Richer, He, Fang, Leigh L. Linden, and Margaret Barbara Trupp, and the International Child MacLeod. 2009. “A Better Way to Teach Chil- Development Steering Group. 2007. “Child dren to Read? Evidence from a Randomized Development in Developing Countries: De- Controlled Trial.” Unpublished working velopmental Potential in the First Five Years paper, Columbia University, New York. for Children in Developing Countries.” Lan- Heckman, James J., and Tim Kautz. 2012. “Hard cet 369 (9555): 60–70. Evidence on Soft Skills.” Labour Economics 19 Grantham-McGregor, Sally M., Christine A. (4): 451–64. Powell, Susan P. Walker, and John H. Himes. Heckman, James J., Seong Hyeok Moon, Rodrigo 1991. “Nutritional Supplementation, Psycho- Pinto, Peter A. Savelyev, and Adam Yavitz. social Stimulation, and Mental Development 2010. “The Rate of Return to the HighScope of Stunted Children: The Jamaican Study.” Perry Preschool Program.” Journal of Public Lancet 338 (8758): 1–5. Economics 94 (1–2): 114–28. Grogan, Louise, and Asha Sadanand. 2013. Heckman, James J., Rodrigo Pinto, and Peter “Rural Electrification and Employment in Savelyev. 2013. “Understanding the Mech- Poor Countries: Evidence from Nicaragua.” anisms through Which an Influential Early World Development 43: 252–65. Childhood Program Boosted Adult Out- Gustafsson-Wright, Emily, and Onno Schellekens. comes.” American Economic Review 103 (6): 2013. “Achieving Universal Health Coverage in 2052–86. Nigeria One State at a Time: A Public-Private Hernández, Bernardo, Dolores Ramírez, Hort- Partnership Community-Based Health Insur- ensia Moreno, and Nan Laird. 2005. “Evalu- ance Model.” Brooke Shearer Working Paper ación del Impacto de Oportunidades en la 2, Global Economy and Development at Mortalidad Materna e Infantil.” In Evaluación Brookings, Brookings Institution, Washing- Externa de Impacto del Programa de Desar- ton, DC. rollo Humano Oportunidades 2004 (External Haider, Rukhsana, Ann Ashworth, Iqbal Kabir, evaluation of the impact of the human de- and Sharon R.A. Huttly. 2000. “Effect of velopment program oportunidades 2004), Community-Based Peer Counsellors on Ex- edited by Bernardo Hernández Prado and clusive Breastfeeding Practices in Dhaka, Mauricio Hernández Ávila, 73–95. Cuer- Bangladesh: A Randomised Controlled Trial.” navaca, Morales, Mexico: National Institute Lancet 356 (9242): 1643–47. of Public Health. Hallegatte, Stephane, Mook Bangalore, Laura Hettige, Hemamala. 2006. “When Do Rural Bonzanigo, Marianne Fay, Tomaro Kane, Roads Benefit the Poor and How? An In- Ulf Narloch, Julie Rozenberg, David Treguer, depth Analysis Based on Case Studies.” Work-

164 POVERTY AND SHARED PROSPERITY 2016 ing paper, Operations Evaluation Depart- Jolliffe, Dean. 1998. “Skills, Schooling, and ment, Asian Development Bank, Manila. Household Income in Ghana.” World Bank Hidrobo, Melissa, John Hoddinott, Amy Mar- Economic Review 12 (1): 81–104. golies, Vanessa Moreira, and Amber Peter- Kane, Thomas J., and Douglas O. Staiger. 2008. man. 2012. “Impact Evaluation of Cash, Food “Estimating Teacher Impacts on Student Vouchers, and Food Transfers among Colom- Achievement: An Experimental Evaluation.” bian Refugees and Poor Ecuadorians in Carchi NBER Working Paper 14607 (December), and Sucumbíos.” Final Report, International National Bureau of Economic Research, Food Policy Research Institute, Washington, Cambridge, MA. DC; World Food Programme, Rome. Khandker, Shahidur R., Zaid Bakht, and Gayatri Higgins, Steve, Maria Katsipataki, Rob Coe, Koolwal. 2009. “The Poverty Impact of Rural Lee Elliot Major, Robbie Coleman, Peter Roads: Evidence from Bangladesh.” Economic Henderson, and Danielle Mason. 2015. “The Development and Cultural Change 57 (4): Sutton Trust–Education Endowment Foun- 685–722. dation Teaching and Learning Toolkit.” July, Khandker, Shahidur R., Douglas F. Barnes, and Education Endowment Foundation, London. Hussain A. Samad. 2013. “Welfare Impacts of https://educationendowmentfoundation.org Rural Electriﬁcation: A Panel Data Analysis .uk/evidence/about-the-toolkits/about-the from Vietnam.” Economic Development and -toolkits/. Cultural Change 61 (3): 659–92. Hoddinott John J., John A. Maluccio, Jere R. Khandker, Shahidur R., and Gayatri Koolwal. Behrman, Reynaldo Martorell, Paul Melgar, 2011. “Estimating the Long-Term Impacts of Agnes R. Quisumbing, Manuel Ramirez- Rural Roads: A Dynamic Panel Approach.” Zea, Aryeh D. Stein, and Kathryn M. Yount. Policy Research Working Paper 5867, World 2011. “The Consequences of Early Childhood Bank, Washington, DC. Growth Failure over the Life Course.” IFPRI Kramer, Michael S., Frances Aboud, Elena Discussion Paper 1073, International Food Mironova, Irina Vanilovich, Robert W. Platt, Policy Research Institute, Washington, DC. Lidia Matush, Sergei Igumnov, et al. 2008. Horta, Bernardo Lessa, and Cesar G. Victora. “Breastfeeding and Child Cognitive Develop- 2013. “Short-Term Effects of Breastfeeding: A ment: New Evidence from a Large Random- Systematic Review of the Beneﬁts of Breast- ized Trial.” Archives of General Psychiatry 65 feeding on Diarrhoea and Pneumonia Mor- (5): 578–84. tality.” World Health Organization, Geneva. Kramer, Michael S., Beverley Chalmers, Ellen D. Hoy, Chris, and Andy Sumner. 2016. “Global Pov- Hodnett, Zinaida Sevkovskaya, Irina Dzikov- erty and Inequality: Is There New Capacity ich, Stanley Shapiro, Jean-Paul Collet, et al. for Redistribution in Developing Countries?” 2001. “Promotion of Breastfeeding Interven- Working paper, University of Sydney, Sydney. tion Trial (PROBIT): A Randomized Trial in IMF (International Monetary Fund). 2014. “Fis- the Republic of Belarus.” Journal of the Amer- cal Policy and Income Inequality” IMF Policy ican Medical Association 285 (4): 413–20. Paper (January 23), IMF, Washington, DC. Kramer, Michael S., Tong Guo, Robert W. Inchauste, Gabriela, and Nora Lustig, eds. Forth- Platt, Stanley Shapiro, Jean-Paul Collet, coming. The Distributional Impact of Fiscal Beverley Chalmers, Ellen Hodnett, Zinaida Policy: Evidence from Developing Countries. Sevkovskaya, Irina Dzikovich, Irina Vani- Washington, DC: World Bank. lovich, and PROBIT (Promotion of Breast- Jamison, Dean, Lawrence H. Summers, George feeding Intervention Trial) Study Group. 2002. Alleyne, Kenneth J. Arrow, Seth Berkley, Agnes “Breastfeeding and Infant Growth: Biology or Binagwaho, Flavia Bustreo, et al. 2013. “Global Bias?” Pediatrics 110 (2): 343–47. Health 2035: A World Converging within a Kremer, Michael, Conner Brannen, and Rachel Generation.” Lancet 382 (9908): 1898–1955. Glennerster. 2013. “The Challenge of Educa- Jensen, Robert, and Emily Oster. 2009. “The tion and Learning in the Developing World.” Power of Cable TV: Cable Television and Science 340 (6130): 297–300. Women’s Status in India.” Quarterly Journal La Ferrara, Eliana, Alberto Chong, and Suzanne of Economics 124 (3): 1057–94. Duryea. 2012. “Soap Operas and Fertility:

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 165 Evidence from Brazil.” American Economic Macours, Karen, Norbert Schady, and Renos Review 4 (4): 1–31. Vakis. 2008. “Cash Transfers, Behavioral Lagarde, Mylene, Andy Haines, and Natasha Changes, and Cognitive Development in Palmer. 2009. “The Impact of Conditional Early Childhood: Evidence from a Random- Cash Transfers on Health Outcomes and ized Experiment.” Policy Research Working Use of Health Services in Low and Middle Paper 4759, World Bank, Washington, DC. Income Countries (Review).” Cochrane Data- Maluccio, John A., and Rafael Flores. 2005. “Im- base of Systematic Reviews 4 (October 7), pact Evaluation of a Conditional Cash Trans- Cochrane Collaboration, London. fer Program: The Nicaraguan Red de Pro- Levy, Santiago. 2008. Good Intentions, Bad Out- tección Social.” IFPRI Research Report 141, comes: Social Policy, Informality, and Eco- International Food Policy Research Institute, nomic Growth in Mexico. Washington, DC: Washington, DC. Brookings Institution Press. Manuel, Trevor. 2002. “The South African Tax Li Ruowei, Chen Xue-Cun, Yan Huai-Cheng, Reform Experience Since 1994.” Address pre- P. Deurenberg, L. Garby, and J. G. Hautvast. sented at the International Bar Association’s 1994. “Functional Consequences of Iron Sup- Annual Conference, Durban, South Africa, plementation in Iron-Deficient Female Cot- October 20–25. ton Mill Workers in Beijing, China.” American Martinez, Sebastian, Sophie Naudeau, and Vitor Journal of Clinical Nutrition 59 (4): 908–13. Pereira. 2012. “The Promise of Preschool in Lim, Stephen S., Lalit Dandona, Joseph A. Africa: A Randomized Impact Evaluation of Hoisington, Spencer L. James, Margaret C. Early Childhood Development in Rural Hogan, and Emmanuela Gakidou. 2010. “In- Mozambique.” February 14, Save the Chil- dia’s Janani Suraksha Yojana, A Conditional dren, Fairfield, CT; World Bank, Washington, Cash Transfer Programme to Increase Births DC. http://siteresources.worldbank.org in Health Facilities: An Impact Evaluation.” /INTAFRICA/Resources/The_Promise_of Lancet 375 (9730): 2009–23. _Preschool_in_Africa_ECD_REPORT.pdf. Lipscomb, Molly, A. Mushfiq Mobarak, and Martinez-Vazquez, Jorge. 2008. “The Impact of Tania Barham. 2013. “Development Effects Budgets on the Poor: Tax and Expenditure of Electrification: Evidence from the Topo- Benefit Incidence Analysis.” In Public Finance graphic Placement of Hydropower Plants in for Poverty Reduction: Concepts and Case Brazil.” American Economic Journal: Applied Studies from Africa and Latin America, edited Economics 5 (2): 200–31. by Blanca Moreno-Dodson and Quentin Lokshin, Michael, and Ruslan Yemtsov. 2005. Wodon, 113–62. Directions in Development: “Has Rural Infrastructure Rehabilitation in Poverty Series. Washington, DC: World Bank. Georgia Helped the Poor?” World Bank Eco- Martinez-Vazquez, Jorge, Violeta Vulovic, and nomic Review 19 (2): 311–33. Blanca Moreno-Dodson. 2014. “The Impact Lopez Acevedo, Gladys. 1999. “Learning Out- of Tax and Expenditure Policies on Income comes and School Cost-Effectiveness in Mex- Distribution: Evidence from a Large Panel of ico: The PARE Program.” Policy Research Countries.” Hacienda Publica Española 200 Working Paper 2128, World Bank, Washing- (4): 95–130. ton, DC. Mideplan (Chile, Ministry of Social Develop- Lucas, Robert E., Jr. 1988. “On the Mechanics of ment). 1999. “Pobreza y distribucion del in- Economic Development.” Journal of Mone- greso en Chile, 1990–1998.” Mideplan, Santiago. tary Economics 22: 3-42. Miguel, Edward, and Michael Kremer. 2004. Lustig, Nora. 2015. “Inequality and Fiscal Re- “Worms: Identifying Impacts on Education distribution in Middle Income Countries: and Health in the Presence of Treatment Ex- Brazil, Chile, Colombia, Indonesia, Mexico, ternalities.” Econometrica 72 (1): 159–217. Peru and South Africa.” CEQ Working Paper Mittal, Surabhi, and Gaurav Tripathi. 2009. “Role 31 (July), Commitment to Equity, Inter- of Mobile Phone Technology in Improving American Dialogue, Washington, DC; Center Small Farm Productivity.” Agricultural Eco- for Inter-American Policy and Research and nomics Research Review 22: 451–59. Department of Economics, Tulane Univer- Morrow, Ardythe L., M. Lourdes Guerrero, Jus- sity, New Orleans. tine Shults, Juan J. Calva, Chessa Lutter, Jane

166 POVERTY AND SHARED PROSPERITY 2016 Bravo, Guillermo Ruiz-Palacios, Robert C. Journal of African Economies 15 (supplement Morrow, and Frances D. Butterfoss. 1999. 2): 212–44. “Efﬁcacy of Home-Based Peer Counsel- Omamo, Steven, Ugo Gentilini, and Susanna ling to Promote Exclusive Breastfeeding: A Sandström, eds. 2010. Revolution: From Randomised Controlled Trial.” Lancet 353 Food Aid to Food Assistance; Innovations in (9160): 1226–31. Overcoming Hunger. Rome: World Food Moser, Caroline, and Jeremy Holland. 1997. Programme. “Urban Poverty and Violence in Jamaica.” Orbeta, Aniceto C., Jr., Arnelyn Abdon, Michael World Bank Latin American and Caribbean del Mundo, Melba Tutor, Marian Theresia Studies: Viewpoints, World Bank, Washing- Valera, and Damaris Yarcia. 2014. “Keeping ton, DC. Children Healthy and in School: Evaluating Mu, Ren, and Dominique van de Walle. 2011. the Pantawid Pamilya Using Regression Dis- “Rural Roads and Local Market Develop- continuity Design Second Wave Impact Eval- ment in Vietnam.” Journal of Development uation Results.” World Bank, Washington, DC. Studies 47 (5): 709–34. Oreopoulos, Philip. 2006a. “The Compelling Muralidharan, Karthik, Paul Niehaus, and Effects of Compulsory Schooling: Evidence Sandip Sukhtankar. 2014. “Payments Infra- from Canada.” Canadian Journal of Econom- structure and the Performance of Public Pro- ics 39 (1): 22–52. grams: Evidence from Biometric Smartcards Oreopoulos, Philip. 2006b. “Estimating Average in India.” NBER Working Paper 1999, Na- and Local Average Treatment Effects of Ed- tional Bureau of Economic Research, Cam- ucation When Compulsory Schooling Laws bridge, MA. Really Matter.” American Economic Review 96 Muralidharan, Karthik, and Venkatesh Sunda- (1): 152–75. raraman. 2011. “Teacher Performance Pay: Oster, Emily. 2015. “Everybody Calm Down Experimental Evidence from India.” Journal about Breastfeeding.” FiveThirtyEight, May of Political Economy 119 (1): 39–77. 20. http://ﬁvethirtyeight.com/features/every Murphy, Cristina B., Andressa K. Peres, Elaine C. body-calm-down-about-breastfeeding/. Zachi, Dora F. Ventura, Luciana Pagan-Neves, Ostry, Jonathan, Andrew Berg, and Charalambos Haydee F. Wertzner, and Elaine Schochat. Tsangarides. 2014. “Redistribution, Inequal- 2015. “Generalization of Sensory Auditory ity, and Growth.” IMF Staff Discussion Note, Learning to Top-Down Skills in a Random- International Monetary Fund, Washington, ized Controlled Trial.” Journal of the Ameri- DC. can Academy of Audiology 26 (1): 19–29. Papp, Lauren M. 2014. “Longitudinal Associa- Naudeau, Sophie, Naoko Kataoka, Alexandria tions between Breastfeeding and Observed Valerio, Michelle J. Neuman, and Leslie Ken- Mother-Child Interaction Qualities in Early nedy Elder. 2011a. Investing in Young Chil- Childhood.” Child: Care, Health, and Devel- dren: An Early Childhood Development Guide opment 40 (5): 740–46. for Policy Dialogue and Project Preparation. Paxson, Christina, and Norbert Schady. 2010. Directions in Development: Human Devel- “Does Money Matter? The Effects of Cash opment Series. Washington, DC: World Bank. Transfers on Child Development in Rural Ec- Naudeau, Sophie, Sebastian Martinez, Patrick uador.” Economic Development and Cultural Premand, and Deon Filmer. 2011b. “Cogni- Change 59 (1): 187–229. tive Development among Young Children in Pellerano, Luca, Marta Moratti, Maja Jakobsen, Low-Income Countries.” In No Small Matter: Mateˇj Bajgar, and Valentina Barca. 2014. “The The Impact of Poverty, Shocks, and Human Lesotho Child Grants Programme Impact Capital Investments in Early Childhood De- Evaluation: Follow-up Report.” United Na- velopment, edited by Harold Alderman, 9–50. tions Children’s Fund. Maseru, Lesotho. Human Development Perspectives Series. Pérez-Escamilla, Rafael, Leslie Curry, Dilpreet Washington, DC: World Bank. Minhas, Lauren Taylor, and Elizabeth Brad- Ndulu, Benno J. 2006. “Infrastructure, Regional ley. 2012. “Scaling Up of Breastfeeding Pro- Integration, and Growth in Sub-Saharan motion Programs in Low- and Middle- Africa: Dealing with the Disadvantages of Income Countries: The ‘Breastfeeding Gear’ Geography and Sovereign Fragmentation.” Model.” Advances in Nutrition 3 (6): 790–800.

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 167 Perova, Elizaveta, and Renos Vakis. 2012. “5 Years Schultz, Theodore W. 1961. “Investment in in Juntos: New Evidence on the Program’s Human Capital.” American Economic Review Short and Long-Term Impacts.” Economía 35 51 (1): 1–17. (69): 53–82. Seneviratne, Dulani, and Yan Sun. 2013. “In- Ravallion, Martin. 2009. “Do Poorer Countries frastructure and Income Distribution in Have Less Capacity for Redistribution?” ASEAN-5: What are the Links?” IMF Work- Policy Research Working Paper 5046, World ing Paper WP/13/11, International Monetary Bank, Washington, DC. Fund, Washington, DC. Rebelo, Sergio T. 1991. “Growth in Open Econ- Springer, Matthew G., John F. Pane, Vi-Nhuan omies.” Policy Research Working Paper 799, Le, Daniel F. McCaffrey, Susan Freeman World Bank, Washington, DC. Burns, Laura S. Hamilton, and Brian Stecher. Reynolds, Arthur J., Judy A. Temple, Suh-Ruu 2012. “Team Pay for Performance: Experi- Ou, Dylan L. Robertson, Joshua P. Mersky, mental Evidence from the Round Rock Pilot James W. Topitzes, and Michael D. Niles. Project on Team Incentives.” Educational 2007. “Effects of a School-Based, Early Child- Evaluation and Policy Analysis 34 (4): 367–90. hood Intervention on Adult Health and Stecklov, Guy, Paul Winters, Jessica Todd, and Well-Being: A 19-Year Follow-up of Low- Ferdinando Regalia. 2007. “Unintended Ef- Income Families.” Archives of Pediatric and fects of Poverty Programmes on Childbear- Adolescent Medicine 161 (8): 730–39. ing in Less Developed Countries: Experimen- Robalino, David A., Laura Rawlings, and Ian tal Evidence from Latin America.” Population Walker. 2012. “Building Social Protection and Studies 61 (2): 125–40. Labor Systems: Concepts and Operational Temple, Judy A., Arthur J. Reynolds, and Barry A. Implications.” Background paper, Social Pro- B. White. 2007. “Age 21 Cost-Beneﬁt Analysis tection and Labor Strategy 2012–22, World of the Title I Chicago Child-Parent Centers: Bank, Washington, DC. Technical Report.” May, Chicago Longitudinal Romer, Paul M. 1990. “Human Capital and Study, University of Minnesota, Minneapolis. Growth: Theory and Evidence.” Carnegie- Thomas, Duncan, Elizabeth Frankenberg, Jed Rochester Conference Series on Public Policy 32 Friedman, Jean-Pierre Habicht, Mohammed (1): 251–86. Hakimi, Nicholas Ingwersen, Jaswadi, et al. Rosenzweig, Mark R., and Andrew D. Foster. 2006. “Causal Effect of Health on Labor 2010. “Microeconomics of Technology Adop- Market Outcomes: Experimental Evidence.” tion.” Annual Review of Economics 2 (1): Working Paper 2006–07, California Center 395–424. for Population Research, University of Cali- Rosero, José, and Hessel Oosterbeek. 2011. fornia, Los Angeles, Los Angeles. “Trade-Offs between Different Early Child- Todd, Petra E., and Kenneth I. Wolpin. 2006. hood Interventions: Evidence from Ecua- “Assessing the Impact of a School Subsidy dor.” Tinbergen Institute Discussion Paper TI Program in Mexico: Using a Social Experi- 2011–102/3, Faculty of Economics and Busi- ment to Validate a Dynamic Behavioral Model ness, University of Amsterdam, and Tinber- of Child Schooling and Fertility.” American gen Institute, Amsterdam. Economic Review 96 (5): 1384–1417. Sahn, David, and Stephen Younger. 2000. “Expen- Tsibouris, George C., Mark A. Horton, Mark J. diture Incidence in Africa: Microeconomic Flanagan, and Wojciech S. Maliszewski. 2006. Evidence.” Fiscal Studies 21 (3): 329–47. “Experience with Large Fiscal Adjustments.” Sankar, Mari Jeeva, Bireshwar Sinha, Ranadip Occasional Paper 246, International Mone- Chowdhury, Nita Bhandari, Sunita Taneja, tary Fund, Washington, DC. Jose Martines, and Rajiv Bahl. 2015 “Optimal Tylleskär, Thorkild, Debra Jackson, Nicolas Breastfeeding Practices and Infant and Child Meda, Ingunn Marie S. Engebretsen, Mickey Mortality: A Systematic Review and Meta- Chopra, Abdoulaye Hama Diallo, Tanya analysis.” Acta Pædriatica 104 (S467): 3–13. Doherty, et al. 2011. “Exclusive Breastfeeding Schady, Norbert, and María Caridad Araujo. Promotion by Peer Counsellors in Sub- 2008. “Cash Transfers, Conditions, and Saharan Africa (PROMISE-EBF): A Cluster- School Enrollment in Ecuador.” Economia 8 Randomised Trial.” Lancet 378 (9789): (2): 43–70. 420–27.

168 POVERTY AND SHARED PROSPERITY 2016 UNESCO (United Nations Educational, Scien- Adult Competence and Reduces Violent Be- tific, and Cultural Organization). 2000. “The havior.” Pediatrics 127 (5): 849–57. Dakar Framework for Action.” UNESCO, Walker, Susan P., Theodore D Wachs, Julie Meeks Paris. http://unesdoc.unesco.org/images/0012 Gardner, Betsy Lozoff, Gail A Wasserman, /001211/121147e.pdf. Ernesto Pollitt, Julie A Carter, and the Inter- UNICEF (United Nations Children’s Fund). 2016. national Child Development Steering Group. The State of the World’s Children 2016: A Fair 2007. “Child Development: Risk Factors for Chance for Every Child. New York: UNICEF. Adverse Outcomes in Developing Countries.” van de Walle, Dominique, Martin Ravallion, Lancet 369 (9556): 145–57. Vibhuti Mendiratta, and Gayatri Koolwal. WHO (Collaborative Study Team on the Role 2013. “Long-Term Impacts of Household of Breastfeeding on the Prevention of Infant Electrification in Rural India.” Policy Re- Mortality, World Heath Organization). 2000. search Working Paper 6527, World Bank, “Effect of Breastfeeding on Infant and Child Washington, DC. Mortality Due to Infectious Diseases in Less Vegas, Emiliana, and Lucrecia Santibáñez. 2010. Developed Countries: A Pooled Analysis.” The Promise of Early Childhood Development Lancet 355 (9202): 451–55. in Latin America and the Caribbean. With WHO (World Health Organization) and Bénédicte Leroy de la Brière, Alejandro Ca- UNICEF (United Nations Children’s Fund). ballero, Julien Alexis Hautier, and Domenec 2003. “Global Strategy for Infant and Young Ruiz Devesa. Latin American Development Child Feeding.” WHO, Geneva. http://www Forum Series. Washington, DC: World Bank. .who.int/nutrition/publications/infantfeeding Victora, Cesar G., Bernardo Lessa Horta, Chris- /9241562218/en/. tian Loret de Mola, Luciana Quevedo, Ricardo World Bank. 2005. World Development Report Tavares Pinheiro, Denise P. Gigante, Helen 2006: Equity and Development. Washington, Gonçalves, and Fernando C. Barros. 2015. DC: World Bank; New York: Oxford Univer- “Association between Breastfeeding and Intel- sity Press. ligence, Educational Attainment, and Income ———. 2009. “Supplementing Nutrition in the at 30 Years of Age: A Prospective Birth Cohort Early Years: The Role of Early Childhood Study from Brazil. Lancet Global Health 3 (4): Stimulation to Maximize Nutritional Inputs.” e199–e205. Child and Youth Development Notes 3 (1), Vincent, Katharine, and Tracy Cull. 2011. “Cell World Bank, Washington, DC. Phones, Electronic Delivery Systems, and ———. 2011a. “Learning for All: Investing in Social Cash Transfers: Recent Evidence and People’s Knowledge and Skills to Promote Experiences from Africa.” International Social Development; World Bank Group Education Security Review 64 (1): 37–51. Strategy 2020.” World Bank, Washington, DC. Vivian, Robert. 2006. “Equality and Personal In- ———. 2011b. “Program Keluarga Harapan: come Tax: The Classical Economists and the Main Findings from the Impact Evaluation Katz Commission.” South African Journal of of Indonesia’s Pilot Household Conditional Economics 74 (1): 79–109. Cash Transfer Program.” World Bank, Jakarta. Wagstaff, Adam, Caryn Bredenkamp, and Lean- ———. 2011c. World Development Report 2011: der R. Buisman. 2014. “Progress on Global Conflict, Security, and Development. Washing- Health Goals: Are the Poor Being Left Be- ton, DC: World Bank. hind?” World Bank Research Observer 29 (2): ———. 2015a. World Bank Support to Early 137–62. Childhood Development: An Independent Wagstaff, Adam, Daniel Cotlear, Patrick Hoang- Evaluation. Washington, DC: World Bank. Vu Eozenou, and Leander R. Buisman. ———. 2015b. “The State of Social Safety Nets 2016. “Measuring Progress toward Univer- 2015.” World Bank, Washington, DC. sal Health Coverage: With an Application to ———. 2015c. “Achieving Results for Wom- 24 Developing Countries.” Oxford Review of en’s and Children’s Health: Progress Report Economic Policy 32 (1): 147–89. 2015.” World Bank, Washington, DC. Walker, Susan P., Susan Chang, Marcos Vera- ———. 2016a. “SABER: Equity and Inclusion Hernández, and Sally Grantham-McGregor. Brief, 2016.” World Bank, Washington, DC. 2011. “Early Childhood Stimulation Benefits http://wbgfiles.worldbank.org/documents

REDUCTIONS IN INEQUALITY: A POLICY PERSPECTIVE 169 /hdn/ed/saber/supporting_doc/brief/SABER Younger, Stephen D., Eric Osei-Assibey, and Felix _EAI_Brief.pdf. Oppong. 2015. “Fiscal Incidence in Ghana.” ———. 2016b. Global Monitoring Report CEQ Working Paper 35 (December), Com- 2015/16: Development Goals in an Era of De- mitment to Equity, Inter-American Dialogue, mographic Change. Washington, DC: World Washington, DC; Center for Inter-American Bank. Policy and Research and Department of Eco- ———. 2016c. Chile: Distributional Effects of nomics, Tulane University, New Orleans. the 2014 Tax Reform. Washington, DC: World Bank. Yang, Dennis. 1997. “Education and Off-Farm Work.” Economic Development and Cultural Change 45 (3): 613–32.

170 POVERTY AND SHARED PROSPERITY 2016

ECO-AUDIT Environmental Beneﬁts Statement

The World Bank Group is committed to reducing its environmental footprint. In support of this commitment, we leverage 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. We follow the recommended standards for paper use set by the Green Press Initiative. The majority of our books are printed on Forest Stewardship Council (FSC)–certiﬁed paper, with nearly all containing 50–100 percent recycled content. The recycled ﬁber 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 philosophy can be found at http://www.worldbank.org/corporateresponsibility.