The Making of a in Bangladesh

Edited by Edited by Xiaobo Zhang

Shahidur Rashid Rashid Shahidur

and for Aquaculture for Enablers, Impacts, and the Path Ahead Ahead Path and the in Bangladesh The Making of a of Making The

The Making of a in Bangladesh Enablers, Impacts, and the Rashid Blue Revolution Path Ahead for Aquaculture & Zhang About IFPRI The International Food Policy Research Institute (IFPRI), established in 1975, provides research-based policy solutions to sustainably reduce poverty and end hunger and malnutrition. IFPRI’s strategic research aims to fos- ter a climate-resilient and sustainable food supply; promote healthy diets and nutrition for all; build inclusive and efficient markets, trade systems, and food industries; transform agricultural and rural economies; and strengthen institutions and governance. Gender is integrated in all the Institute’s work. Partnerships, communications, capacity strengthening, and data and knowledge management are essential components to translate IFPRI’s research from action to impact. The Institute’s regional and country programs play a critical role in responding to demand for food policy research and in delivering holis- tic support for country-led development. IFPRI collaborates with partners around the world.

About IFPRI’s Peer Review Process IFPRI books are policy-relevant publications based on original and innova- tive research conducted at IFPRI. All manuscripts submitted for publication as IFPRI books undergo an extensive review procedure that is managed by IFPRI’s Publications Review Committee (PRC). Upon submission to the PRC, the manuscript is reviewed by a PRC member. Once the manuscript is considered ready for external review, the PRC submits it to at least two external reviewers who are chosen for their familiarity with the subject matter and the country setting. Upon receipt of these blind external peer reviews, the PRC provides the author with an editorial decision and, when necessary, instructions for revision based on the external reviews. The PRC reassesses the revised manuscript and makes a recommendation regarding publication to the director general of IFPRI. With the director general’s approval, the manuscript enters the editorial and production phase to become an IFPRI book.

The Making of a Blue Revolution in Bangladesh Enablers, Impacts, and the Path Ahead for Aquaculture

Edited by Shahidur Rashid and Xiaobo Zhang

A Peer-Reviewed Publication

This publication is licensed for use under a Creative Commons Attribution 4.0 International License (CC BY 4.0). Subject to attribution, you are free to share (copy and redistribute the material in any medium or format), adapt (remix, transform, and build upon the material) for any purpose, even commercially.

Third-party content: The International Food Policy Research Institute does not necessarily own each component of the content contained within the work. The International Food Policy Research Institute 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.

Recommended citation: Rashid, S., and X. Zhang, eds. 2019. The Making of a Blue Revolution in Bangladesh: Enablers, Impacts, and the Path Ahead for Aquaculture. Washington, DC: International Food Policy Research Institute. https://doi.org/10.2499/9780896293618.

This is a peer-reviewed publication. Any opinions expressed herein are those of the authors and are not necessarily representative of or endorsed by the International Food Policy Research Institute (IFPRI). The boundaries and names shown and the designations used on the maps do not imply official endorsement or acceptance by IFPRI.

International Food Policy Research Institute, 1201 Eye Street, NW, 12th floor, Washington, DC 20005 USA, Telephone: +1-202-862-5600, www.ifpri.org

ISBN: 978-0-89629-361-8 DOI: https://doi.org/10.2499/9780896293618

Library of Congress Cataloging-in-Publication Data may be found on page vi. CONTENTS

Tables and Figures vii Abbreviations and Acronyms xi Foreword xiii Acknowledgments xv

Chapter 1 Introduction 1 Shahidur Rashid

Chapter 2 Sector Overview and Study Design 19 Shahidur Rashid, Kaikaus Ahmad, and Gracie Rosenbach

Chapter 3 Value Chain Transformation 31 Ricardo Hernandez, Ben Belton, Thomas Reardon, Chaoran Hu, Xiaobo Zhang, and Akhter Ahmed

Chapter 4 Cluster-Based Aquaculture Growth 57 Xiaobo Zhang, Qingqing Chen, and Peixun Fang

Chapter 5 Welfare and Poverty Impacts of Aquaculture Growth 77 Shahidur Rashid, Nicholas Minot, and Solomon Lemma

Chapter 6 Future Scenarios (Projections to 2050) 103 Paul Dorosh and Andrew Comstock

Chapter 7 Summary and Implications 143 Shahidur Rashid and Xiaobo Zhang

Authors 153 Index 155 Library of Congress Cataloging-in-Publication Data Names: Rashid, Shahidur, editor. | Zhang, Xiaobo, 1966– editor. Title: The making of a blue revolution in Bangladesh : enablers, impacts, and the path ahead for aquaculture / edited by Shahidur Rashid and Xiaobo Zhang. Description: Washington, DC : International Food Policy Research Institute, 2019. | Includes bibliographical references and index. Identifiers: LCCN 2019019949 (print) | LCCN 2019981222 (ebook) | ISBN 9780896293618 (paperback) | ISBN 9780896293625 (ebook other) Subjects: LCSH: Sustainable aquaculture—Bangladesh. | Aquaculture— Environmental aspects—Bangladesh. | Aquaculture—Social aspects— Bangladesh. | Sustainable fisheries—Bangladesh. Classification: LCC SH136.S88 M35 2019 (print) | LCC SH136.S88 (ebook) | DDC 639.8028/6095492—dc23 LC record available at https://lccn.loc.gov/2019019949 LC ebook record available at https://lccn.loc.gov/2019981222

Cover Design: Jason Chow Project Manager: John Whitehead, IFPRI Book Layout: BookMatters TABLES AND FIGURES

Tables 1.1 Bangladesh: Historical production trends (thousand metric tons) 8 1.2 Bangladesh: Historical fish consumption trends 9 1.3 Bangladesh: Historical fish price trends 10 1.A1 Local, common English, and scientific fish names of selected varieties 17 2.1 Area and production of fish in Bangladesh, 2014–2015 21 2.2 Zones and sampled districts 27 3.1 Definitions of actor size, by actor type 33 3.2 Zone characteristics 34 3.3 Structural change in hatchery clusters over 10 years 37 3.4 Structural change in feed mill clusters over 10 years 38 3.5 Structural change in feed dealer clusters over 10 years 39 3.6 Structural change in fish farmer clusters over 10 years 40 3.7 Landholdings and tenancy by year and zone 41 3.8 Factor productivity in 2013 42 3.9 Structural change in rural fish trader clusters over 10 years 42 3.10 Disposal of fish farm harvest by final user type, 2013 46 3.11 Disposal of fish farm harvest by final user location, 2013 47 3.12 Aquaculture production by fish category, 2008 and 2013 50 3.13 Fish farm capital-to-labor ratio by year and zone 53 4.1 The spatial linkage of the fish supply chain 62

vii viii Tables and Figures

4.2 Principal component analysis 63 4.3 Clustering measures at the district level, 2003, 2008, and 2013 63 4.4 Fish farmers’ specialization in output and adoption of modern inputs, 2008 and 2013 66 4.5 OLS estimates on the relationship between clustering and output specialization 68 4.6 Regressions on the relationship between clustering and adoption of modern inputs 69 4.7 Horizontal cooperation among fish traders and feed dealers, 2013 70 4.8 Probit regressions on the relationship between clustering and cooperation 72 4.9 Probit regressions on the relationship between clustering and different types of cooperative behaviors 74 5.1 Trends in fish production in Bangladesh, 1983/1984– 2012/2013 80 5.2 Changes in annual per capita fish consumption (kilograms per person per year) in Bangladesh, 2000–2010 82 5.3 Net positions of households in aquaculture fish in Bangladesh 92 5.4 Impacts of aquaculture growth on household income 94 5.5 Impacts of aquaculture growth on poverty reduction 96 6.1 Bangladesh: Expenditure elasticities by fish production system (QUAIDS model estimates) 109 6.2 Bangladesh: Econometric estimates of price elasticities of demand for fish, 2000, 2005, and 2010 110 6.3 Model variables and parameters 113 6.4 Model simulation assumptions for fish productivity growth 115 6.5 Model simulation assumptions for population and income growth 116 6.6 Bangladesh fish production and prices: Simulation results 116 6.7 Bangladesh fish consumption: Simulation results (per capita consumption) 118 6.A1 Demographic variables used in the econometric estimation of household demand parameters, 2000, 2005, and 2010 126 Tables and Figures ix

6.A2 Share and price variables: Descriptive statistics, 2000, 2005, and 2010 128 6.A3 Bangladesh fish exports, 2000–2014 (thousand metric tons) 130 6.A4 Scientific names for all fish by category 130 6.A5 Own-price elasticity estimates using the Linear Expenditure System (LES) 131 6.A6 Cross-price elasticity estimates using the Linear Expenditure System (LES) 132 6.A6a Alternative specifications of expenditure elasticity parameters, 2000, 2005, and 2010 133 6.A6b Alternative specifications of price elasticity parameters, 2000, 2005, and 2010 134 6.A7 Model calibration 140 6.A8 Bangladesh fish production and prices: Simulation results (alternative parameters) 140 6.A9 Bangladesh fish consumption: Simulation results (alternative parameters) 141 6.A10 Bangladesh fish production projections to 2015 142

Figures 3.1 Average weekly consumption per capita of the 10 most consumed fish species in rural Bangladesh 49 4.1 Twenty sample districts in fish-clustering areas 59 4.2 Distributions of number of actors per 1,000 rural people in 2013, by districts 60 4.3 Degree of fish clustering of sample districts, 2013 65 4.4 Clustering versus cooperation among input dealers and traders 71 4.5 Clustering and accessibility to truck rental companies 75 5.1 Real prices (2010 = 100) of selected fish varieties in Bangladesh, 1986–2014 84 6.1 Simulation results: Bangladesh fish production (thousand metric tons) 119 6.A1 Model validation: Production by fish system (million metric tons) 125

ABBREVIATIONS AND ACRONYMS

ADB Asian Development Bank BBS Bangladesh Bureau of Statistics BDT Bangladeshi taka BFDC Bangladesh Fisheries Development Corporation BFRI Bangladesh Fisheries Research Institute BIHS Bangladesh Integrated Household Survey CAPI computer assisted personal interviews CPI Consumer Price Index DID difference in differences DoF Department of Fisheries EA enumeration area EU European Union FAO Food and Agriculture Organization of the United Nations GAP good aquaculture practices GDP gross domestic product gm grams GMP good manufacturing practices GoB Government of Bangladesh ha hectare HES Household Expenditure Survey HH household HHI Hirschman–Herfindahl Index HIES Household Income and Expenditure Survey ICLARM International Center for Living Aquatic Resources Management (WorldFish)

xi xii abbreviations and Acronyms

IFPRI International Food Policy Research Institute IMPS integrated multipurpose sample kg kilogram km kilometer LES Linear Expenditure System MoFL Ministry of Fisheries and Livestock MPC marginal propensity to consume MT metric ton MT/ha metric tons per hectare NBR net benefit ratio NFP National Fisheries Policy NFS National Fisheries Strategy NGO nongovernmental organization OLS ordinary least squares PCA principle component analysis PPS probability proportion to size PRSSP Policy Research and Strategy Support Program PSU primary sampling unit QUAIDS Quadratic Almost Ideal Demand System RNI recommended nutritional intake SIS small indigenous species SMA significant metropolitan area US$ US dollar USAID United States Agency for International Development VC value chain FOREWORD

Fisheries and aquaculture are becoming increasingly important subsectors within the national food system in many developing countries. This broad trend has important implications for food and nutrition security, global trade, and overall livelihoods. Thus, it is no surprise that fisheries feature prominently in many of the United Nations Sustainable Development Goals (SDGs) and have a very direct relationship to SDG 14 (Life Below Water). Given this context, IFPRI initiated several country case studies to better understand the fish value chain’s transformation in certain developing countries. Bangladesh has been one of the first developing countries to undergo such a transformation. In the past two decades, production of culture fish in the country has quadrupled, real prices of most common fish varieties have declined, and fish consumption has increased from about 13 kg in 2000 to over 20 kg in 2016. More importantly, as this book highlights, the consumption growth has occurred across the lines of gender, region, and income quintile, with poorer households benefiting more than other groups. Since fish is the single largest animal-sourced food item in the Bangladeshi diet, the growth in aquaculture and fisheries has played an important role in improving nutritional outcomes. The success of Bangladesh’s aquaculture sector is so remarkable that it is commonly dubbed a “Blue Revolution.” Drawing on modeling exercises and primary surveys, as well as analysis of secondary data, this book presents evidence on the drivers, impacts, and prospects for promoting aquaculture in Bangladesh. Results from the primary fish value chain suggest that the aquaculture sector has indeed experienced a dramatic transformation. There are clear

xiii xiv Foreword

indications of disintermediation (that is, fewer actors per unit of output in the value chain), which has reduced transaction costs; fish production has become increasingly clustered to take advantage of agglomeration effects; and the rate of adoption of modern fish varieties and improved farming practices has increased. Given increased investment, the projection analysis indicates that the country has prospects for further growth in fish production, with supply outpacing demand, which in turn is likely to benefit the poor. However, as the authors point out, more data collection and analysis need to be carried out to adequately account for factors such as sustainability and habitat degradation. At a broader level, making a Blue Revolution in Bangladesh through aquaculture is consistent with the country’s comparative advantage—that is, numerous ponds and abundant laborers. Just as the availability of cheap labor fueled the garment export boom, the more efficient use of water and people has triggered the boom in the aquaculture sector. This book provides a systematic assessment of the Blue Revolution, highlighting opportunities and challenges, which will shed light on many issues of economic transformation in developing countries. I believe that the book’s analytical framework and survey results will be valuable for policymakers and future researchers.

Shenggen Fan Director General ACKNOWLEDGMENTS

The conceptualization of this book’s research was initiated under IFPRI’s Policy Research and Strategy Support Program in Bangladesh, funded by the United States Agency for International Development. The initial idea was to produce a value chain report, but early field visits and stakeholders’ consultations indicated that additional research was needed to better understand the unprecedented transformation in the sector. Thus, IFPRI supported additional work through the CGIAR Research Programs on Policies, Institutions, and Markets (PIM) and Agriculture for Nutrition and Health (A4NH), the European Commission, and the Swiss Development Corporation. We are thankful to these agencies for their financial support. The editors are indebted to many colleagues for their intellectual insights and analytical support. We have benefited from discussions with Akhter Ahmed, Paul Dorosh, Nurul Islam, and Shahidur R. Khandker. Their insights have contributed greatly to improving the content and structure of this book. Our special thanks go to Maximo Torero, who was at the time a division director at IFPRI, for generously supporting the additional work. We express our gratitude to Syed Mahmudul Hoque, chairman of the Bangladesh Shrimp and Fish Foundation, for sharing his deep understanding of the fisheries sector in Bangladesh. Mr. Hoque’s insights on emerging challenges in the sector continue to inspire us to undertake future research on sustainable fisheries development. Key messages from the book have been presented in several conferences—notably the 30th International Conference of Agricultural Economists of the International Association of Agricultural Economists, held in Vancouver in 2018, and the Regional Strategic Analysis and Knowledge Support System for Asia (ReSAKSS-Asia) conference, held in Bangkok in

xv xvi Acknowledgments

December 2017. The editors have benefited from the comments of the partici- pants in those conferences. We owe a great deal to a talented pool of survey experts, research analysts, and copy editors for their support. The stacked value chain survey was administered by IFPRI’s longtime survey partner in Bangladesh, Data Analysis and Technical Assistance Limited (DATA). We are grateful to Zahidul Hasan, Mohammed Zobair, and other DATA staff members for successfully implementing the survey and conducting initial analysis. Solomon Lemma, Sarah McMullen, Qingqing Chen, and Redwan Rokon provided analytical support to the researchers in Washington; and Nusrat Hossain, Latiful Haque, and Arifeen Akter from IFPRI-Dhaka helped administer surveys and conduct analysis in Bangladesh. We are thankful to Gracie Rosenbach and Jenny Smart for their invaluable support in preparing the manuscript; and John Whitehead for overseeing the book’s production process. Finally, we are grateful to IFPRI’s Communications and Public Affairs Division and the Publications Review Committee (PRC) for efficiently managing the book’s production and peer review process. Two rounds of comments by three anonymous reviewers have greatly improved the quality of the manuscript. We would like to thank the PRC chair, Professor Gerald Shively, for efficiently managing the review and adding his own comments and sug- gestions on the manuscript. We are also grateful for the editorial help from Gracie Rosenbach and Jenny Smart during proofing. Any errors are of course the authors’ own. Chapter 1

INTRODUCTION

Shahidur Rashid

The Blue Revolution in the Food Security Debate In the last half of the 20th century, food policy in most Asian countries meant ensuring availability of cereals, mainly rice and wheat.1 The rationale for this cereal-centric policy is well understood. Agricultural productivity was low, the world market was volatile, and the national food security depended on, apart from Mother Nature, the relationship with the donor countries, which was not smooth because of ideological differences. By the early 1960s, feeding a rapidly growing population became a daunting challenge for the region’s countries. Many experts viewed these challenges as too big to handle. “Famine 1975” (Paddock and Paddock 1967), “lifeboat ethics” (Hardin 1974), and “triage” (Ehrlich 1971) were the labels commonly applied to these countries. Thanks to Green Revolution technology and concerted policy actions, none of the dire predictions turned out to be true. By the 1980s, countries in the region ensured cereal availability and began to enjoy overall economic growth (Rashid, Cummings, and Gulati 2007). However, just when the countries began to celebrate the success of the Green Revolution, a different kind of food security challenge surfaced. Several studies demonstrated that human nutrition, in terms of the consumption of various nutrients, was not responsive to the income growth that the countries were enjoying. One of the first powerful studies on the issue, Behrman and Deolalikar (1987), demonstrated that a 10 percent increase in income leads to only a 1.7 percent increase in total calorie consumption in India. For some nutrients, such as iron and calcium, the estimates were negative, implying that consumption of these nutrients declined with increases in income. Similar evidence began to emerge from other countries in Asia, including Bangladesh (Pitt 1983) and the Philippines (Bouis and Haddad 1992; Bouis 1994), and more studies on India (Behrman and Deolalikar 1990). The literature grew

1 It is believed that the term “Blue Revolution” was first coined in a 2003 article of the Economist magazine.

1 2 chapter 1

rapidly, and studies have appeared from more diverse countries such as Fiji (Gibson and Rozelle 2010) and Tanzania (Abdulai and Aubert 2004). While a more detailed discussion on the literature is beyond the scope of this chapter, the evidence of the low responsiveness of nutrition to income growth has been influential. It challenged the conventional wisdom—as well as the widely accepted policy prescription—that poverty was the root cause of malnutrition and hence income growth would solve the problems of both poverty and malnutrition.2 Thus, despite the success of the Green Revolution, this strand of literature called for rethinking development strategies in Asia. This was particularly true for Bangladesh. Haunted by the memories of famine and food price–related political instabilities, Bangladesh had historically placed more emphasis on increasing rice production than any other country in the region. By the late 1980s, these policies began to pay off—the Green Revolution started taking root, rice prices began falling in real terms, and the country enjoyed overall economic growth. However, all these successes did not translate into improvement in a key aspect of food security: nutritional well-being. One of the first studies presenting such evidence, Pitt (1983) demonstrated that the expenditure elasticity of demand for nutrients was < 1 for all nutrients for the poor households, and for seven out of nine nutrients among richer households. Subsequent studies examined the causes behind the nonresponsiveness of nutrients to income. Bouis and Novenario-Reese (1997) reported that the primary reason for micronutrient deficiencies was low- quality diets that resulted from (1) unaffordability of richer sources of nutrients, such as fish; (2) lack of nutritional knowledge; and (3) discrimination in intrahousehold food allocation.3 Fish is the single largest animal-sourced food item in Bangladeshi diets, accounting for more than 60 percent of the animal-sourced protein (FAO 2005) in an average daily diet.4 However, fish prices continued to rise from the mid-1980s through early 2000 (Chapter 5). Concerns about rising fish prices, in particular among policymakers, were widespread, as many studies expressed doubts whether this price trend could be reversed. In presenting his research on commercial vegetable and polyculture fish in Bangladesh, Bouis (2000,

2 For detail, see World Bank 1981, 59. 3 Of these three factors, only the first one can be addressed by improving market fundamentals. The latter two are related to information asymmetry and sociocultural factors, which require a different set of policies than addressing rising prices. 4 These estimates are most likely from the 1990s reflected in the HIES 2000 round. Since then, per capita annual consumption of fish has increased from 13 kilograms in 2000 to over 23 kilograms in 2016. Therefore the share of animal-sourced protein from fish in Bangladeshi diets is presumed to have gone up further. Introduction 3

486) contended that “although inflation-adjusted cereal prices in Bangladesh have fallen by 40% over the last 25 years (a remarkable achievement), real prices of lentils, vegetables, and animal products have increased by 25% to 50%. Real fish prices have perhaps doubled. Dietary quality for the poor may be declining over time due to these price effects.” The literature expressing concerns about the rising prices was centered around three core themes. First, there were concerns about promoting aquaculture, as experts predicted that such a policy would have negative consequences on the environment and the open water system (Minkin and Boyce 1994), which has historically been the source of fish for the poor. The second theme was that, even if modern aquaculture was promoted, it was unlikely to benefit the poor. The argument was that, since aquaculture requires land and capital, the poor would be left out, resulting in greater inequality. One of the CGIAR centers, the International Center for Living Aquatic Resources Management (ICLARM), expressed similar concerns.5 Using survey data from two upazilas (subdistricts) in the early 1990s, the study reported that, even though a large portion of the households in the survey sites were landless, all pond owners or operators owned land (Ahmed and Lorica 2002). Finally, the third theme of the earlier literature, experts were skeptical about the capacity of the smallholders in promoting modern aquaculture. In general, nonshrimp aquaculture was believed to be backward with little chance of meeting the growing demand. For instance, Lewis (1997) concluded that, despite all the attention and centrality of fish in Bangladeshi culture, existing aquaculture practices were inadequate to deal with the shortfall in fish demand. The central part of the earlier predictions—that is, rising fish prices—did not come true. Sectoral statistics and readily available secondary data demonstrate this. The key trends in the fisheries sector began to reverse in the early 2000s. Production of cultured fish more than doubled between 2000 and 2010, from about half a million metric tons to 1.3 million metric tons, with much of the growth coming from pond fisheries.6 The corresponding changes in the real prices and consumption have been unprecedented: the real price of common aquaculture fish, such as carp, declined by about 45 percent; and per capita annual fish consumption jumped from only 7.7 kilograms in 1980

5 “CGIAR” was originally the acronym for the Consultative Group on International Agricultural Research. In 2008, CGIAR redefined itself as a global partnership. To reflect this transformation and yet retain its roots, “CGIAR” was retained as a name. CGIAR is now a global research partnership for a food secure future. 6 Chapter 5 presents a detailed analysis of production and price trends. 4 chapter 1

(FAO 2014) to 13 kilograms in 2000, and to more than 18 kilograms in 2010.7 Other studies show even higher estimates. For instance, in a recently published report on the Household Income and Expenditure Survey (HIES), the Bangladesh Bureau of Statistics (BBS) estimated per capita annual consumption to be even higher, at 23 kilograms (BBS 2017). Similarly, based on the Bangladesh Integrated Household Survey (BIHS) of 2011–2012, Ahmed et al. (2013) reported fish consumption by rural households to be 23.50 kilograms per person per year.8 Since the volume of aquaculture fish export is small, growth in aquaculture benefited the country in many ways. In addition to economic value addition and employment generation, this has contributed to improving food and nutritional security in the country. For instance, estimates from the Bangladesh HIES suggest that fish consumption by the bottom quintile of the population increased by 57 percent between 2000 and 2010. Fish is the most important source of high-quality protein and essential fatty acids (Roos et al. 2007); it is the most frequently consumed animal-source food across all social strata in Bangladesh (Toufique and Belton 2014). Therefore the growth in consumption has far-reaching implications for the economic and nutritional well-being of the country’s overall population. Aquaculture in Bangladesh can thus serve as an excellent case study to help in understanding the role of fisheries in the food security debate. The central argument has been that nutrient-rich food will continue to be expensive and out of the reach of the poor. This implies that the aquaculture sector will remain traditional and nonresponsive to broader economic changes. However, quite contrary to this assumption, the recent growth in Bangladeshi aquaculture appears to have been triggered by supply responses to increased consumer demand for fish. There have been massive aggregate investments by the value chain actors, including farmers, feed millers, hatcheries, nurseries, input dealers, and all types of traders. The fish value chain in Bangladesh today is much different from what it was only a couple of decades ago (Chapter 3 details this value chain transformation). The next section discusses why a systematic assessment of this transformation, which has hitherto been lacking, is impor tant for a better understanding of its contribution to employment, income, and poverty reduction.

7 The statistic for 1980 is per capita supply, used to infer per capita consumption. 8 This is the sum of small and large fish consumption estimates presented in Table 6.5 in Ahmed et al. 2013. Introduction 5

Fish Production Systems There are three primary systems of fish production in Bangladesh: aquaculture, inland capture, and marine capture.9 Each system has its own unique trend and faces unique constraints. To provide context for the analysis throughout this book, we detail each system below.

Aquaculture There are a variety of aquaculture methods practiced in Bangladesh—from cage production to the use of floodplains. However, the dominant method by far is that of pond culture, accounting for nearly 86 percent of total aquaculture. The two main pond culture methods are “homestead pond culture” and “entrepreneurial pond culture.” Homestead farm culture developed from small ponds used by individual households to supplement consumption and sometimes income. Often not the primary source of income for a household, these homestead farms have accounted for increasingly large shares of income over the past few years. A major constraint for these types of ponds is low productivity, although opportunities to raise productivity through improved practices exist. Numerous development projects have been implemented in recent years, but to date the results have been mixed, with slow uptake of new technology, leading to relatively small productivity gains (Belton et al. 2011; Bloomer 2012). Entrepreneurial ponds have been started with the expressed intent of being a primary source of income. These ponds produce at much larger scales, requiring greater access to input markets and labor (Belton et al. 2011). Entrepreneurial ponds face constraints as well. Access to finance is one of these, as farmers find it difficult to reliably access loans and other sources of capital (Bloomer 2012). Although improvements in seed and input markets have contributed to recent growth in aquaculture, low access to inputs still represents a key constraint in further improvements. Competition for space and resources with agriculture is also a concern (FAO 2014).

Inland Capture Inland capture production covers the more traditional fishing systems involving the capture of wild fish from streams, rivers, and lakes. Inland fisheries often either do not use boats, or only use small, nonmotorized boats, or are

9 The material in this section has been generously provided by Dorosh and Comstock (see Chapter 6). 6 chapter 1

small-scale. The types of various freshwater fish (including barbs, tilapia, koi, and medium catfish) produced from inland capture have long been a staple in the Bangladeshi diet and are often preferred by the local population to fish produced via aquaculture (FAO 2014). However, production in this sector has clearly begun to lag that of aquaculture ponds. The main constraints facing inland capture are habitat loss (due to urban- ization and agricultural intensification), pollution leading to environmental stress, and overexploitation of resources (Belton et al. 2011). Recurring floods and natural disasters have also led to major losses of habitats. In particular, intense floods have recently degraded portions of traditional inland capture fisheries. These disasters are only expected to become more frequent due to climate change, to which Bangladesh is particularly vulnerable (Ghose 2014).

Marine Capture Finally, marine capture refers to all fish production coming out of marine fisheries. As with inland capture, marine capture is mostly dominated by small-scale fisheries (in this case, using boats). However, a semi-industrialized fishery sector and a small industrial sector (FAO 2014) also exist. A significant constraint facing marine fishing is that of overfishing. Exploitive fishing practices have hindered long- and short-term prospects, as have challenges in establishing co-management areas (Ghose 2014). Another constraint facing marine fishing, according to Belton et al. (2011), is the difficulty of meeting international standards for the products. Marine fish (especially shrimp) producers have struggled to maintain the quality standards demanded by most large importers of shrimp, including the United States. Maintaining these standards would provide a significant boon to the viability of marine production exports. In July 1997 the European Union (EU) banned imports of fish produced in Bangladesh (most Bangladesh fish exports are from marine fisheries). The ban was initiated as a result of EU inspections of Bangladeshi fish processing plants. The plants were found to be in serious violation of EU standards for seafood products and lacking in quality controls. Overall, the ban is estimated to have cost Bangladesh US$15 million in just five months. In the subsequent years, Bangladesh addressed the issues that led to the ban, and exports to the EU began to open up again. However, with ever-changing quality standards, it will continue to be an issue that must be consistently re-evaluated (Cato and Subasinge 2003). Introduction 7

Historical Trends Aquaculture has become increasingly prominent in the fish production mix of Bangladesh over the past 15 years. According to BBS data, aquaculture’s share of fish production increased from 30 percent to 47 percent from 2000 to 2015. In addition, the BBS HIES data shows a kilogram per capita increase of 3.3 to 7.2 from 2000 to 2010. Table 1.1 shows production levels for 2010, which are estimated from the shares of consumption of various types of freshwater fish from the 2010 HIES multiplied by the 2010 total freshwater fish production figures from BBS. Since BBS disaggregates production into only two categories (aquaculture and inland capture), growth rates for the three categories shown (aquaculture, mixed, and inland capture) differ from the BBS production figure growth rates. (Definitions of marine fish are consistent across both the BBS and the HIES.) Estimates of production for aquaculture and inland capture for 2000 were constructed using Dorosh and Comstock’s 2010 production estimates and the 2005–2010 BBS production growth rates for these categories; mixed system production was estimated as the residual. Figures for production of aquaculture and inland capture for 2000 to 2015 are estimated using the calculated figures for levels of production in 2000 and the historical growth rate of these categories from 2010 to 2015 as calculated from BBS production data. The table shows just how rapid aquaculture growth has been as compared with the three other fish categories. The fourth type of fish production system (called “mixed”) contains the fish consumed in the HIES data that we could not accurately account for being produced via aquaculture or inland capture because of the heterogeneous systems used to produce some types of fish.10 Aquaculture production has grown from 498,000 metric tons to over 1,700,000 metric tons in the 15 years shown in the table. Inland capture only grew from 369,000 metric tons to 414,000 metric tons. Mixed and marine categories grew more than did inland capture, but both did not achieve nearly the same levels of growth as that of aquaculture. These levels of growth are mirrored by the annual growth rates shown below them in the table. Aquaculture had an annual growth rate of nearly 9 percent for the 15-year period in question, while growth in inland capture was less than 1 percent.

10 The categorization of the fish was taken from Toufique and Belton 2014. The full breakdown is provided in Table 6.A4. Dorosh and Comstock’s categorization classifies shrimp as inland capture rather than marine. 8 chapter 1

Table 1.1 Bangladesh: Historical production trends (thousand metric tons)

Inland Year Aquaculture capture Mixed Marine Total 2000 498 369 460 334 1,661 2001 541 372 395 379 1,688 2002 588 375 512 415 1,890 2003 638 378 550 432 1,998 2004 693 381 573 455 2,102 2005 753 384 605 475 2,216 2006 818 387 645 480 2,329 2007 888 389 675 487 2,440 2008 964 392 709 498 2,563 2009 1,047 395 744 515 2,701 2010 1,138 398 846 517 2,899 2011 1,235 401 832 546 3,015 2012 1,342 404 987 579 3,312 2013 1,457 408 956 589 3,410 2014 1,583 411 945 595 3,534 2015 1,719 414 934 600 3,667 Annual percentage growth rates 2000–2005 8.6 0.8 5.6 7.3 5.9 2005–2010 8.6 0.8 6.9 1.7 5.5 2010–2015 8.6 0.8 2.0 3.0 4.8 2000–2015 8.6 0.8 4.8 4.0 5.4

Source: BBS (2000, 2005, and 2010) and Dorosh and Comstock’s calculations (Chapter 6). Note: Historical production trends were calculated by estimating the “mixed” category proportion based on HIES 2010 data, adjusting backward to 2000 using the 2005–2010 growth rate, and then projecting forward using the 2010–2015 growth rate.

Growth in marine fish production has likewise been slow, although growth accelerated slightly after 2010. The consumption data from the HIES (Table 1.2) mirrors this pattern of the increasing importance of aquaculture and the decreasing importance of inland capture. In both rural and urban areas, aquaculture fish consumption increased faster than fish consumption from other production systems, from 3.76 kilograms per capita in 2000 to 7.41 kilograms per capita in 2010 in urban areas. Meanwhile, per capita consumption of inland capture system Introduction 9

Table 1.2 Bangladesh: Historical fish consumption trends

2000 2005 2010 Rural Urban Rural Urban Rural Urban Kilograms per capita Aquaculture 3.21 3.76 5.12 5.79 7.10 7.41 Mixed 4.38 3.82 4.95 4.57 5.07 6.27 Inland capture 3.58 3.47 2.64 3.31 2.33 3.15 Marine 1.76 3.41 1.42 3.91 1.51 4.36 Total fish 12.92 14.46 14.14 17.57 16.01 21.19 Consumption value shares (%) Aquaculture 3.1 3.7 4.0 4.5 5.4 5.2 Mixed 4.0 3.6 3.9 3.8 4.2 4.8 Inland capture 3.5 3.5 2.4 3.0 2.0 2.6 Marine 2.1 3.5 1.8 3.6 2.0 4.4 Total fish 12.7 14.3 12.2 14.9 13.6 16.9

Source: Dorosh and Comstock’s calculations (Chapter 6) from BBS 2000, 2005, and 2010. Note: Per capita kilograms are per year. fish declined, from 3.47 to 3.15 kilograms per capita in urban areas over the same time period. In the other two categories, mixed sees some increase (to be expected as this category contains some aquaculture) and marine slightly declines in rural areas and slightly increases in urban areas (again, expected since it is somewhat of a luxury item). The shares of fish in total expenditures, turning to value shares, in 2000, for urban areas consumed a larger share of aquaculture relative to total consumption than did the rural areas. However, by 2010 the rural areas were con- suming a larger share of aquaculture. This is most likely due to the expanding production and lower prices (Table 1.3). Urban areas also see much higher rates of consumption for marine fish, which, per the HIES data, are the most expensive of the fish produced in Bangladesh. Finally, price indexes for the four types of fish production from 2000 to 2010 are presented in Table 1.3. These price indexes were calculated using weights based on household expenditure shares in the 2010 HIES. Dorosh and Comstock use the consumer price index (CPI) as a price deflator to convert all prices to real 2010 prices. For purposes of presentation, however, we rescale the indexes such that 2000 = 100. As shown, prices of fish from most of the production systems have increased over time, with real prices 10 chapter 1

Table 1.3 Bangladesh: Historical fish price trends

2010 Price indexes %Δ %Δ %Δ value 2000– 2005– 2000– Major group Fish name weights (%) 2000 2005 2010 2005 2010 2010 Aquaculture Rui/Katla/Mrigel/ 35 0.960 0.812 1.000 −19 23 4 Kalibaus Silver carp/Grass 30 0.954 0.777 1.000 −22 29 5 carp/Minor carp Pangas/Boaal/ 34 1.484 0.909 1.000 −9 10 −33 Bagair Total primarily agriculture 1.138 0.835 1.000 −17 20 −12 Mixed Magur/shing/Shin- 5 0.758 0.681 1.000 −32 47 32 gi/Lal Kholisha Koi 8 0.481 0.688 1.000 −31 45 108 Mala-kachi/Chala- 26 0.855 0.769 1.000 −23 30 17 chapila Bhadi Puti/ 60 0.851 0.765 1.000 −24 31 17 Tilapia/Nilotica Total mixed 0.816 0.755 1.000 −24 32 23 Inland capture Shol/Gojar/Taki 25 0.803 0.745 1.000 −25 34 25 Tengra/Pakal 12 0.695 0.694 1.000 −31 44 44 Nuna baila/Tepa 3 0.808 0.781 1.000 −22 28 24 Shrimp 40 0.649 0.669 1.000 −33 50 54 Other 19 0.748 0.725 1.000 −27 38 34 Total inland capture 0.718 0.706 1.000 −29 42 39 Marine Ilish 35 0.623 0.652 1.000 −35 53 61 Dried fish 51 0.693 0.693 1.000 −31 44 44 Sea fish 14 0.873 0.714 1.000 −29 40 15 Total marine 0.693 0.682 1.000 −32 47 44

Source: Dorosh and Comstock’s calculations (Chapter 6) from BBS 2000, 2005, and 2010. Fish names by type extracted from Toufique and Belton 2014. Note: Price indexes are in real terms where 2010 = 100. Scientific names for all fish are listed inTable 6.A4. of fish from mixed, inland capture, and marine systems rising by 23, 39, and 44 percent, respectively. Prices of fish produced in aquaculture systems declined by 12 percent, reflecting an increase in supply relative to demand.

The Blue Revolution, Its Drivers, and Implications Declining trends in the real price of fish, in the face of a growing population and increasing demand for fish, implies that there has been a transformation Introduction 11

in the aquaculture sector in Bangladesh. The factors behind the process of transformation can be broadly grouped into (1) improved technology, (2) reduced transactions costs, and (3) innovation in the value chain. Improved technology includes the introduction of modern fish varieties, improved farming practices (for example, use of modern inputs), as well as postproduction marketing practices. Until the early 1990s, there were practically no commercial pond fisheries in the country. For instance, although Bangladeshis have historically produced and consumed pangas catfish, commercial production of non-native pangas began only in 1993 and expanded rapidly, with total production reaching 300,000 metric tons by 2008.11 This rapid commercialization has a value proposition in that net returns per unit of land are much higher for new varieties of aquaculture fish than other commodities. Total production per hectare currently averages 40 metric tons of some high- yielding varieties (Belton et al. 2011), which is worth about US$52,000 at current prices, several times more than the revenue generated from 3 metric tons of rice produced in the same amount of land. Reduction in transaction costs has come through improved infrastructure, better access to information, and reduced marketing risks. Roads, access to telecommunication, and rural electrification have improved dramatically. The following statistics from the World Bank (2017) illustrate the improvement: Between 2000 and 2010, the decade in which aquaculture experienced the most growth, there was a large increase in total road length, rural households with electricity went from 20 to 50 percent, and cell phone ownership soared from 0.2 percent to about 75 percent. This has contributed to lowering the costs of fish trade. Just a couple of decades ago, both production and trading of fish embodied great risks. Farmers relied mostly on their luck when they took their catch to market, as they neither knew the price nor had the options to store their fish. There have been remarkable changes in the value chain as well. There are clear indications of disintermediation—that is, fewer actors per unit of output in the value chain. For instance, our survey shows that the number of traders has increased from 14,800 in 2000 to 31,300 in 2014 (Chapter 6), while culture fish production has increased from about 500,000 metric tons to almost 2 million metric tons. Instead of farmers selling to local retailers, markets have moved closer to the farmers, where agents of the processors and large wholesalers compete to buy fish in rural markets. Also, there is now more direct

11 Local English names of fish are used throughout this book. Table 1.A1 provides the local name, common English name, and scientific name of all fish mentioned in this chapter. 12 chapter 1

marketing from the production areas to the terminal markets. This process has both efficiency and equity implications. Disintermediation makes value chains more efficient because it decreases marketing costs and hence supplies the commodity at a lower price to the consumers. However, it has equity implications as small-scale traders are eliminated from the value chain. From a policy standpoint, it is important to understand the winners, losers, and net gains from the transformation. A robust analysis of these changes is important on several grounds. The Government of Bangladesh can benefit from such analysis in revising or refor- mulating fisheries policies. The country’s national fisheries policies are dated. Although new subsector policies have recently been adopted—related to shrimp culture (2014), the leasing of shrimp culture plots (2013), and the reg- ulation regarding the use of government-owned water bodies (2009)—the government still relies on a set of national fisheries policies that was adopted in 1998 (Bangladesh, DoF 2017). Development partners have invested quite heavily in the aquaculture in Bangladesh. Citing other studies, Toufique and Belton (2014) report that, in addition to large-scale sectoral investments, 10 donors invested roughly US$275 million during 1990–2003 on fisheries projects in Bangladesh. However, while there are ex post evaluations of these projects, the donors’ investments should be guided by the broad understanding of the sector’s transformation.

Research Questions and Analytical Approach This book focuses exclusively on the nonshrimp part of aquaculture in Bangladesh. The rationale for such a focus is twofold: (1) it is nearly all domestically consumed; and (2) it has important implications for poverty and food security. For decades, the word “aquaculture” was associated with shrimp, as shrimp had become an important export item in the 1980s and made frequent news headlines when importing countries put bans on it for food safety reasons. In fact, both the donors and the national institutes have paid more attention to shrimp than fish, even though the latter has far greater implications for food security and poverty. Consequently, this book does not address export-related issues of standards, food safety, and international trade. Instead the book focuses on the three broad aspects of nonshrimp aquaculture: (1) determinants of the value chain transformation; (2) poverty and food security impacts of the transformation; and (3) the medium-term prospects of aquaculture in Bangladesh. The following questions are at the core of the analysis presented throughout this book: Introduction 13

1. Has the aquaculture value chain been transforming? If yes, what are the enabling factors behind the transformation? 2. What explains the specialization and formation of fish clusters at various geographic locations of the country, and what are the resulting impacts? 3. What implications does the transformation have in terms of income distribution, poverty, and food security? What is the magnitude of the impacts in terms of poverty reduction? 4. What is the future growth potential of the subsector given the structural and technological changes on the horizon? Specially designed stacked value chain surveys were conducted to address the first two sets of questions (detail provided in Chapter 2). The framework used in Reardon et al. (2012) is followed to analyze the changes and transformation in the value chain. Both descriptive and econometric methods have been used for analyzing the determinants of fish cluster formations and their impacts on efficiency gains. The poverty and food security impacts are assessed by carrying out microsimulations, with parameter estimates generated from several rounds of nationally representative household surveys. Another piece of analysis presented in the book is a medium-term projection of demand and supply of fish in Bangladesh. A multimarket model is set up based on the parameters estimated from the nationally representative surveys, the Bangladesh HIES.

Organization of the Book Following the book’s introduction, Chapter 2 presents the sector overview and study design. Chapter 3 focuses on the value chain transformation by analyzing both mesolevel and microlevel data. Chapter 4 takes a unique approach in analyzing determinants and impacts of cluster formation, which is a relatively new trend in aquaculture. These two chapters address core research questions on enablers of the Blue Revolution. Chapter 5 evaluates the impacts of aquaculture growth on income distribution and poverty. Chapter 6 shows the midterm projection of demand and supply. The book ends with Chapter 7, which highlights summaries and policy implications. 14 chapter 1

References Abdulai, A., and D. Aubert. 2004. “Nonparametric and Parametric Analysis of Calorie Consumption in Tanzania.” Food Policy 29 (2): 113–129. Ahmed, A. U., K. Ahmad, V. Chou, R. Hernandez, P. Menon, F. Naeem, F. Naher, W. Quabili, E. Sraboni, B. Yu, and Z. Hassan. 2013. The Status of Food Security in the Feed the Future Zone and Other Regions of Bangladesh: Results from the 2011–2012 Bangladesh Integrated Household Survey. Washington, DC: International Food Policy Research Institute. Ahmed, M., and M. H. Lorica. 2002. “Improving Developing Country Food Security through Aquaculture Development—Lessons from Asia.” Food Policy 27 (2): 125–141. Bangladesh, DoF (Department of Fisheries). 2017. “Policies.” Accessed December 11, 2017. www .fisheries.gov.bd/site/view/policies/Policy. BBS (Bangladesh Bureau of Statistics). 2000. Household Income and Expenditure Survey 2000. Dhaka.

—. 2005. Household Income and Expenditure Survey 2005. Dhaka. —. 2010. Household Income and Expenditure Survey 2010. Dhaka. —. 2017. Preliminary Report on the Household Income and Expenditure Survey of 2016. Dhaka: Ministry of Planning, People’s Republic of Bangladesh. Behrman, J. R., and A. B. Deolalikar. 1987. “Will Developing Country Nutrition Improve with Income? A Case Study for Rural India.” Journal of Political Economy 95 (3): 492–507. —. 1990. “The Intra-Household Demand for Nutrients in Rural South India: Individual Estimates, Fixed Effects, and Permanent Income.” Journal of Human Resources 25 (4): 665–696. Belton, B., M. Karim, S. Thilsted, K. Murshed-E-Jahan, W. Collis, and M. Phillips. 2011. “Review of Aquaculture and Fish Consumption in Bangladesh.” In Studies and Reviews 2011- 53. Penang, Malaysia: The WorldFish Center. Bloomer, J. 2012. Homestead Aquaculture in Bangladesh: Current Status and Future Directions. London: King’s College London. Bouis, H. E. 1994. “The Effect of Income on Demand for Food in Poor Countries: Are Our Food Consumption Databases Giving Us Reliable Estimates?” Journal of Development Economics 44 (1): 199–226. —. 2000. “Commercial Vegetable and Polyculture Fish Production in Bangladesh: Their Impacts on Household Income and Dietary Quality.” Food and Nutrition Bulletin 21 (4): 482–487. Bouis, H. E., and L. J. Haddad. 1992. “Are Estimates of Calorie-Income Elasticities Too High? A Recalibration of the Plausible Range.” Journal of Development Economics 39: 333–364. Introduction 15

Bouis, H. E., and M. J. G. Novenario-Reese. 1997. The Determinants of Demand for Micronutrients: An Analysis of Rural Households in Bangladesh. FCND Discussion Paper 32. Washington, DC: International Food Policy Research Institute. Cato, J. C., and S. Subasinge. 2003. “Case Study: The Shrimp Export Industry in Bangladesh.” Food Safety in Food Security and Food Trade, Focus 10, Brief 9. Washington, DC: International Food Policy Research Institute.

Economist. 2003. “Fish Farming: The Promise of a Blue Revolution.” Accessed December 1, 2017. www.economist.com/node/1974103.

Ehrlich, P. R. 1971. The Population Bomb. New York: Ballantine Books. FAO (Food and Agriculture Organization of the United Nations). 2005. “Bangladesh National Aquaculture Sector Overview.” Accessed December 11, 2017. www.fao.org/fishery/country sector/naso_bangladesh/en. —. 2014. “Fishery and Aquaculture Country Profiles: Bangladesh.” Country Profile Fact Sheets. Rome.

Ghose, B. 2014. “Fisheries and Aquaculture in Bangladesh: Challenges and Opportunities.” Annals of Aquaculture and Research 1 (1): 1–5. Gibson, J., and S. Rozelle. 2010. “How Elastic Is Calorie Demand? Parametric, Nonparametric, and Semiparametric Results for Urban Papua New Guinea.” Journal of Development Studies 38 (6): 23–46.

Hardin, G. 1974. “Living on a Lifeboat.” Bioscience 24 (10): 561–568. Lewis, D. 1997. “Rethinking Aquaculture for Resource-Poor Farmers: Perspectives from Bangladesh.” Food Policy 22 (6): 533–546. Minkin, S. F., and J. K. Boyce. 1994. “Net Losses: ‘Development’ Drains the Fisheries of Bangladesh.” Amicus Journal 16 (3): 36–40. Paddock, P., and W. Paddock. 1967. Famine 1975! America’s Decision: Who Will Survive? Boston: Little, Brown and Company. Pitt, M. M. 1983. “Food Preferences and Nutrition in Rural Bangladesh.” Review of Economics and Statistics 65 (1): 105–114. Rashid, S., R. Cummings, and A. Gulati. 2007. “Grain Marketing Parastatals in Asia: Results from Six Case Studies.” World Development 35 (11): 1872–1888. Reardon, T., K. Chen, B. Minten, and L. Adriano. 2012. The Quiet Revolution in Staple Food Value Chains: Enter the Dragon, the Elephant, and the Tiger. Manila: Asian Development Bank; Washington, DC: International Food Policy Research Institute. 16 chapter 1

Roos, N., M. A. Wahab, C. Chamnan, and S. H. Thilsted. 2007. “The Role of Fish in Food-Based Strategies to Combat Vitamin A and Mineral Deficiencies in Developing Countries.” Journal of Nutrition 137 (4): 1106–1109. Toufique, K. A., and B. Belton. 2014. “Is Aquaculture Pro-Poor? Empirical Evidence of Impacts on Fish Consumption.” World Development 64: 600–620. Wikipedia. 2018. “List of Fishes in Bangladesh.” Accessed September 17, 2018. https://en.wikipedia .org/w/index.php?title=List_of_fishes_in_Bangladesh&oldid=848562826.

World Bank. 1981. World Development Report. Washington, DC. —. 2017. World Development Indicators. Washington, DC. Introduction 17

Annex

Table 1.A1 Local, common English, and scientific fish names of selected varieties

Local name Common English name Scientific name Bagair Dwarf goonch Bagarius bagarius Bhadi puti Pool barb Puntius sophore Boaal Wallago Wallago attu Gojar Great snakehead Channa marulius Grass carp Grass carp Ctenopharyngodon Idella Ilish Hilsa Tenualosa ilisha Kalibaus Orange-fin labeo Labeo calbasu Katla Indian carp/Indian katla Catla catla Koi Climbing perch Anabas testudineus Lal kholisha Dwarf gourami Colisa lalia Magur/shing Magur/shing Gagata youssoufi Mala-kachi/ Ganges River gizzard shad/ Gonialosa manmina/gudusia chala-chapila Indian River shad chapra Minor carp Minor carp Crossocheilus latius Mrigel Mrigal Cirrhinus cirrhosis Nilotica Nile tilapia Oreochromis niloticus niloticus Nuna bailla Nuna bailla Brachygobius nunus Pakal Eel Anguilliformes Pangas Yellowtail catfish Pangasius pangasius Rui Indian rui/Rohu Labeo rohita Shingi Stinging catfish Heteropneustes fossilis Shol Snakehead murrel Channa striata Silver carp Silver carp Hypophthalmichthys molitrix Taki Spotted snakehead Channa punctate Tilapia Indian tilapia Oreochromis mossambicus Tengra Tyangra Macrones vittalus Tepa Ocellated pufferfish Tetraodon cutcutia

Source: Wikipedia (2018).

Chapter 2

SECTOR OVERVIEW AND STUDY DESIGN

Shahidur Rashid, Kaikaus Ahmad, and Gracie Rosenbach

Sector Overview

Importance of the Sector The fisheries sector in Bangladesh is important in terms of both economic and food security perspectives. The sector accounts for about 4 percent of national gross domestic product (GDP), 23 percent of the agricultural GDP, and about 3 percent of total foreign exchange earnings (Bangladesh, DoF 2015). In terms of employment, the sector’s role in the economy is even larger. About 17.8 million Bangladeshis, including 1.4 million women, find jobs (full time and part time) in the sector (FAO 2016), which translates to about 11 percent of the total population and more than 23 percent of the working population.1 Fish also occupies an important place in Bangladeshi diets and culture—so much so that there is a Bengali (both Bangladesh and Indian West Bengal) adage that says mache bhate bangali (meaning “fish and rice is what makes a Bengali”). The role of fish in improving food security of the poor is even greater. Fishing is an important source of livelihood for the poor, and it is often their only source of protein. It is estimated that about 70 percent of the rural population engaged in fishing for subsistence at some point in the year (FAO 2014). The sector is also the second largest export earner, next to readymade gar- ments, equaling about 18 percent of GDP (Taslim and Haque 2011). More important, although the sector grew by about 6 percent over the past 10 years, there is still large potential for future growth in production and exports. The country’s Seventh Five Year Plan (2016–2020) has set five goals to this end— namely, increased production and export, increased conservation of aquatic diversity, enhanced coastal and marine fisheries, equitable income generation, and improved safety (Bangladesh, Ministry of Planning 2015). The first and

1 According to World Bank (2017), the dependency ratio is 51.4, which means the working population is 48.6 percent.

19 20 chapter 2

the fifth goals appear to have been particularly designed for export promotion objectives. These goals set the targets of increasing aquaculture production by 45 percent, increasing fish farmers income by 20 percent, and increasing export earnings to US$1.25 billion by 2020. In addition, the plan aspires to achieve good aquaculture practices (GAP) and good manufacturing practices (GMP) at all stages of the value chain to comply with the standards of international markets. These are ambitious targets that might not be met by 2020, but as discussed below, the country certainly has the potential to achieve them.

Opportunities and Challenges The fisheries sector in Bangladesh consists of three main subsectors—inland capture, inland culture (aquaculture), and marine. Historically, inland capture and marine used to be the dominant subsectors. In fact, aquaculture was the smallest subsector until the 1980s, accounting for only about 16 percent of total fish production. Things started changing rapidly by the turn of the century, and aquaculture became the largest among the three subsectors. In terms of volume, the average annual production of culture fish averaged only 178 thousand metric tons from 1983–1984 to 1992–1993, but jumped to more than 2 million metric tons by 2014, representing about 69 percent of total fish production in the country (Table 2.1).2 Disaggregated estimates of water area and production of fish provide important insights into the recent trend as well as future opportunities and challenges. The numbers in Table 2.1 clearly suggest that the main driver of growth in culture fisheries has been cultivation in ponds. In 2015, production from ponds accounted for more than 70 percent of inland culture and 40 percent of total fish production. Historically ponds have been an impor tant part of Bengali culture—they have served as both private and common property for bathing as well as production and consumption of fish. Three general styles of fish production exist: (1) extensive (traditional method with “no intentional nutritional inputs”), (2) semi-intensive (rely on natural food and supplementary feed), and (3) intensive (“depend on nutritionally complete diets added to the system”) (Edwards and Demaine 1998). Previously production systems were traditional and households never used improved seeds and feeds. Gradual moves toward intensification and commercialization that occurred toward the end of the last century triggered the growth in the subsector.

2 All statistics presented in Table 2.1 are the authors’ calculations based on Bangladesh, DoF data. Sector Overview and Study Design 21

Table 2.1 Area and production of fish in Bangladesh, 2014–2015

Water area Percentage of total Production in Shares (capture + Subsectors (hectares) (capture + culture) 2015 (metric tons) culture) (%) Inland capture Beel 114,161 2.4 92,678 2.8 Floodplain 2,692,964 57.2 730,210 22.1 Kaptai lake 68,800 1.5 8,645 0.3 River 853,863 18.1 174,878 5.3 Sundarban 177,700 3.8 17,580 0.5 Capture total 3,907,488 83.0 1,023,991 31.0 Inland culture Baor (Oxbow lakes) 5,488 0.1 223,582 6.8 Cage culture 10 0.0 1,969 0.1 Pen culture 8,326 0.2 16,084 0.5 Pond culture 377,968 8.0 1,610,875 48.8 Seasonal water body 133,330 2.8 201,280 6.1 Shrimp/prawn 275,583 5.9 223,582 6.8 Culture total 800,705 17.0 2,277,372 69 (Capture + culture) 4,708,193 100.0 3,301,363 100 Marine Industrial — 84,846 14.1 Artisanal — 515,000 85.9 Marine total — 599,846 100 Country Total 4,708,193.00 3,901,209.00

Source: Compiled from the Bangladesh, DoF (2015) and FAO (2016). Note: — = data not available.

However, there is large potential for further growth in pond culture in three important ways. First, the culture fisheries in Bangladesh are largely extensive (traditional, whereby fish feed entirely from the food web within the pond) or improved extensive (traditional with some supplemental feeding), with a limited number of semi-intensive or intensive systems (fish are dependent on the feed provided and water must be replenished at a high rate to maintain oxygen levels and remove waste).3 As a result, overall productivity is much lower relative to many Asian countries. For instance, per hectare shrimp production in Bangladesh is only about 786 kilograms, which is only about 26 percent of the 3 metric tons per hectare in both Vietnam and Thailand.

3 The definition of the farming system is obtained from EC (2012). 22 chapter 2

Productive varieties, such as pangas and tilapia, are growing but remain a small share of total area under aquaculture. For a handful of commercial farmers who use an intensive system for these varieties, productivity per hectare is reported to be 60–70 metric tons (Edwards and Hossain 2010), which is incredibly low when compared to more than 240 metric tons in Vietnam (Phuong et al. 2007). Second, with economic growth, rice consumption in Bangladesh has been declining in recent years. For instance, according to a recently published Household Income and Expenditure Survey (HIES) report, per capita daily rice consumption in Bangladesh has declined from 416 grams in 2010 to 367 grams in 2016, equivalent to about a 2.7 percent annual rate of decline. The reduction in rice consumption in rural Bangladesh is even higher—from 442 grams in 2010 to 386 grams in 2016, or about a 3.8 percent rate of decline— during the same period (BBS 2017). This trend implies that there will be opportunities to convert paddy land to pond culture or to diversify into other crops. Finally, despite the country’s huge water bodies, rivers, and coastline, cage culture and coastal aquaculture are practically nonexistent. If the barriers to exploiting this opportunity (for example, a sound regulatory environment for cage culture; access to credit; and availability of seed, feed, and other technology) are alleviated, and if their viability is enhanced, it will give an additional boost to aquaculture production. The fisheries sector in Bangladesh faces serious challenges. A list of these challenges included in the official reports of the Department of Fisheries (DoF) can be grouped into three broad categories: (1) productivity, (2) habitat degradation and negative externalities, and (3) institutional and regulatory challenges. The official sources identify productivity challenges mainly with the scarcity of quality seed, feed, and other inputs. However, productivity enhancement is also constrained by enforcement of property rights in common pool resources. For instance, 3.9 million hectares of the total of 4.7 million hectares, equivalent to 83 percent, are under capture fisheries. Most of these lands have common property elements and hence suffer from classic “tragedy of commons” problems. While regulations can limit open access, achieving full productivity potentials can be challenging. The challenges related to habitat degradation and other environmental consequences are highlighted in the Seventh Five Year Plan. In fact, the plan is to help at least 75 percent of the endangered inland water species in designated sanctu- aries reappear by 2020. Similarly, there are policies to restrict marine catches to help grow certain species, such as ilish, which was chronically declining until recently. Thanks to a government program called “Jatka [young ilish less Sector Overview and Study Design 23

than 10 inches in length] Preservation,” production figures in the past couple of years have seen growth. However, this is not true for other marine fisheries. The regulatory and institutional challenges are at all stages of the value chain—from quality input supply to ensuring food safety for the consumers. The regulatory constraints to input supply (seed and feed) are elaborated in FAO (2016), and the challenges of institutional capacity in extension, quality and safety assurance, as well as enforcement of law are articulated in various reports from the Ministry of Fisheries and Livestock.

Policy Environment The fisheries subsector policies in Bangladesh have evolved over many decades. Therefore, many rules, acts, and ordinances have been passed by the government over the years. This section presents an overview of the policies, governance, and emerging challenges and strategies in the subsector. Serious policy thinking regarding the fisheries sector in Bangladesh began immedi- ately after Bangladesh gained independence from Pakistan. Between 1950 and 1997, the Government of Bangladesh passed a total of 21 pieces of legislation. A quick review of these documents suggests that the basic act to regulate inland fisheries in Bangladesh is the Protection and Conservation of Fish Act (1950). This act went through several amendments in the subsequent decades. Two of the main amendments are The Protection and Conservation Ordinance (1982) and the Marine Fisheries Ordinance (1983). One striking feature of the fisheries legislation in Bangladesh is that there are no separate sections on aquaculture, although some of the provisions are relevant to the subsector (FAO 2016). For instance, the Protection and Conservation of Fish Rules include protection of carp species, prohibit certain activities, and stipu- late that licenses to catch fish can only be issued for the purposes of aquaculture development. While the country had a wide range of ordinances, rules, and acts, an integrated National Fisheries Policy (NFP) was adopted only in 1998. The document highlights a long list of rather ambitious policy objectives, ranging from promoting economic growth to restoring environmental balance. Given current governance and institutional structure, it is unclear how these objectives can be achieved. The NFP extends to all government organizations involved in fisheries (and to all water bodies used for fisheries), with unclear mandates for any of those public entities. For instance, Section 6 of the NFP presents the details of the policies related to inland closed water fisheries that include 17 different policy actions. Of these 17 action points, 4 are related to addressing 24 chapter 2

property rights, 3 are related to private-sector development, 2 are related to research, and the remainder fall broadly under training and extension. Thus the Ministry of Fisheries and Livestock and its implementing arm, the DoF, have a tall order to enact these laws. The DoF has the overall responsibility to both develop and regulate the fisheries sector. While it is supported by two other public entities—analytical support by the Bangladesh Fisheries Research Institute (BFRI) and industry development support by the Bangladesh Fisheries Development Corporation (BFDC)—delivering on the complex mandates entrusted under various policies and strategies is a difficult task. The governance and coordination challenges are obvious from the fact that key pieces of regulations related to feed and hatcheries as well as aquaculture medicinal protocol did not get passed until 2010–2011 and 2015–2016, respectively (Bangladesh, DoF 2016). New challenges continue to surface. In a recent report the DoF (Bangladesh, DoF 2015) highlights several challenges that have important implications for future growth, sustainability, food safety, and overall governance of the sector. For instance, one of the key challenges is ensuring quality inputs (for instance, seed, feed, and chemicals), which is likely to have serious implications for future growth (FAO 2015). Similarly, habitat degradation, overfishing, and expansion of coastal aquaculture are of concern. Unless these challenges are addressed quickly, they can have longer-term consequences to the environment and biodiversity. There are two challenges that directly link to the overall governance. The first is the poor institutional links among the stakeholders (Bangladesh, DoF 2015), which essentially implies that the exe- cution of the recently passed policies and regulations will be difficult. The other challenge is in data generation and management. While the DoF main- tains times series data on prices and production, it uses an old survey framework that was devised in 1983–1984 when aquaculture was in its infancy. Given all the changes in the sector, much richer data generation, management, and analysis needs to be instituted to formulate evidence-based policies that can tackle the emerging challenges effectively.

Study Design and Data

The Process The design of the study began with a reconnaissance trip and stakeholders’ consultation initiated under the Policy Research and Strategy Support Program (PRSSP) of the International Food Policy Research Institute Sector Overview and Study Design 25

(IFPRI) in Bangladesh. The initial plan was to produce a value chain report for the project. However, it became clear from early consultations that there is much to be analyzed to better understand the transformation in aquaculture in Bangladesh. The knowledge gaps in three areas became obvious. First, the review suggested that there was no systematic assessment, based on a large sample, of the aquaculture value chain in Bangladesh. Second, even though the NFP repeatedly highlighted the importance of poverty and food security for promoting aquaculture, to the best of our knowledge, there were no studies assessing the poverty and food security impacts. Finally, it also appeared important to assess prospects of further growth of the sector. Addressing these questions involved the compilation of a large volume of secondary data, use of existing nationally representative surveys and generation of data with special surveys, and the application of a mix of analytical methods, including econometrics, microsimulation, and multimarket models. Two large datasets have been used. The first is a specially designed stacked value chain survey as proposed in Reardon et al. (2012). This survey combined both mesodata and microdata and has been the basis of analysis on value chain transformation (Chapter 3) and cluster formation (Chapter 4). The analysis of poverty impacts, and demand and supply projections, however, had to be based on nationally representative surveys along with other secondary data. Therefore several rounds of the Bangladesh HIES have been used for these two sets of analysis. A central part of the study is the design and implemen- tation of the value chain survey. Brief descriptions of these surveys are provided below.

The Value Chain Survey Administering this survey involved developing a sampling framework that can capture a representation of all actors in the value chain (VC). To do this, a large volume of data was gathered for developing a sampling frame and site selections. Based on this initial work, two sets of surveys were conducted: (1) a microlevel survey of all key actors in the fish value chain, and (2) a community- level survey to gather mesolevel information. Briefly, the sample for the micro level survey was drawn with a purposive stratified random sampling method. The reason for doing a purposive sample is twofold: (1) fish production is concentrated in certain districts of the country, and (2) a nationally representative sample was neither necessary nor financially feasible. The sample was drawn from 20 districts that fell under four zones (clusters): East (Brahmanbaria, Chittagong, Comilla, Cox’s Bazar, Noakhali, and Sylhet districts); North (Bogra, Dinajpur, Gazipur, Mymensingh, 26 chapter 2

Narsingdi, and Natore districts); Southwest (Khulna, Satkhira, and Bagerhat districts); and South Center (Barisal, Bhola, Chandpur, Gopalganj, and Jessore districts). Cox’s Bazar was subsequently dropped from our analysis because we decided to focus on nonshrimp aquaculture in this report. The VC analysis also dropped the interviews with the shrimp farmers and traders from the southern districts. With all these considerations, sampling followed this approach: in each of the districts, a set of subdistricts (upazilas) were randomly selected using probability proportion to size (PPS), which resulted in selection of 102 upazilas in 20 districts (Table 2.2). All mouzas (sub- unit treated as primary sampling unit [PSU]) in each selected upazila were selected. Once the PSUs were selected, a census of fish farmers was conducted in each of them and 25 farmers were randomly selected per PSU (20 farmers, plus 5 replacements). The final farm household sample of 77 mouzas (PSU) is representative of 86 percent of the fish pond areas in the districts selected. In turn, the districts selected constitute 61 percent of all pond production in the country. The questionnaires for the survey of each value chain actor were designed to capture all the information necessary to carry out analysis on value chain transformation presented in Chapter 3. The process of questionnaire development involved consulting the IFPRI household, trader, and market surveys in other countries, and the Reardon et al. (2012) questionnaires for the staple value chain study. All questionnaires were programmed in CSPro (Census and Survey Processing System) by an IFPRI programmer to conduct the surveys with computer assisted personal interviews (CAPI) using Mirus tablets. The survey questionnaire for each of the value chain segments included questions that can be grouped into four broad categories: (1) demographic and business characteristics, (2) input supplies, (3) value addition, and (4) marketing of outputs.

The Household Income and Expenditure Survey The HIES is a nationally representative survey conducted by the Bangladesh Bureau of Statistics (BBS) in five-year intervals. The HIES generates official estimates on income, expenditure, consumption, and poverty situation. The first round of the HIES was conducted in 1973–1974 in the newly inde- pendent Bangladesh. Since then, including the latest survey in 2015, the BBS has successfully completed 16 rounds of surveys. Over time, the survey expanded to include additional modules to track many emerging indicators. For instance, the 2010 round of the HIES added four new submodules Sector Overview and Study Design 27

Table 2.2 Zones and sampled districts

Zone Districts East Brahmanbaria, Chittagong, Comilla, Cox’s Bazar, Noakhali, Sylhet North Bogra, Dinajpur, Gazipur, Mymensingh, Narsingdi, Natore Southwest Bagerhat, Khulna, Satkhira South Center Barisal, Bhola, Chandpur, Gopalganj, Jessore

Source: Authors’ compilation based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. to gather information on microcredit, migration and remittances, shocks and coping, and disability. A big scaling up of the survey occurred during the 2016 round of the HIES, as the government decided to generate many disaggregated estimates, and so the sample size almost quadrupled from 12,240 in 2010 to more than 46,000 in 2016. Unfortunately, these data are yet to be available. Therefore the analysis presented in this book—welfare implications (Chapter 5) and demand system estimates and future projections (Chapter 6)—are based on the earlier rounds. The HIES follows an elaborate sampling method. Broadly, until the 2010 round the sampling was based on a two-stage stratified random sampling technique, with samples drawn under a framework called Integrated Multipurpose Sample (IMPS). Developed on the basis of the Population and Housing Census 2001, the IMPS design consists of 1,000 PSUs throughout the country. Until the 2010 round there were 640 rural and 360 urban PSUs—defined as two or more contiguous enumeration areas (EA)—each comprised of around 200 households. In the first stage 612 out of a total 1,000 PSUs were drawn from 16 different strata (6 rural, 6 urban, and 4 significant metropolitan area [SMA] strata). In the second stage 20 households were selected from each of the rural, urban, and SMA PSUs. In the 2010 round a total of 12,240 households were sampled, of which 7,840 were rural and the rest resided in urban areas. Following the sample design, the survey is completed in one cal- endar year (for example, February 1, 2010, to January 31, 2011, in the case of the 2010 round). Thus the survey captures the seasonal variations in a cycle of income, expenditure, and consumption patterns. The survey period is divided into 18 terms, and within each term 34 PSUs are covered to collect data from a total of 680 sample households. In the HIES 2010, 12,240 households were selected, whereby 7,840 were from rural areas and 4,400 were from urban areas. 28 chapter 2

By combining analyses from various rounds of the HIES and from the primary data collected in the value chain surveys, this book seeks to fill the existing gaps in the literature on fish value chains in Bangladesh. This data provides the opportunity to review historical trends and their determinants, assess the welfare effects, and make projections for future trends and impacts to provide relevant quality policy recommendations to successfully update and complement the current policy environment.

References Bangladesh, DoF (Department of Fisheries). 2015. Annual Report 2015. Ministry of Fisheries and Livestock. Dhaka.

—. 2016. Annual Report 2016. Ministry of Fisheries and Livestock. Dhaka. Bangladesh, Ministry of Planning, Planning Commission. 2015. Seventh Five Year Plan (2016– 2020): Accelerating Growth, Empowering People. Dhaka. BBS (Bangladesh Bureau of Statistics). 2010. Household Income and Expenditure Survey 2010. Dhaka.

—. 2017. Preliminary Report on the Household Income and Expenditure Survey of 2016. Dhaka: Ministry of Planning, People’s Republic of Bangladesh. EC (European Commission). 2012. “Aquaculture Techniques.” Accessed August 21, 2018. https:// ec.europa.eu/fisheries/sites/fisheries/files/docs/body/2012-aquaculture-techniques_en.pdf.

Edwards, P., and H. Demaine. 1998. Rural Aquaculture: Overview and Framework for Country Reviews. Bangkok: Regional Office for Asia and the Pacific, Food and Agriculture Organization (FAO) of the United Nations. Edwards, P., and M. S. Hossain. 2010. “Bangladesh Seeks Export Markets for Striped Catfish.” Global Aquaculture Advocate: 58–60. FAO (Food and Agriculture Organization of the United Nations). 2014. “Fishery and Aquaculture Country Profiles: Bangladesh.” Country Profile Fact Sheets. Rome. —. 2015. Aquaculture Seed and Feed Production and Management in Bangladesh: Status, Issues and Constraints. Rome —. 2016. “Fisheries Statistics in Bangladesh: Issues, Challenges, and Plans.” Asia and Pacific Commission on Agricultural Statistics, twenty-sixth session, Thimphu, Bhutan. Phuong, N. R., L. X. Sinh, N. Q. Thinh, H. H. Chau, C. T. Anh, and N. M. Hau. 2007. “Economics of Aquaculture Feeding Practices: Viet Nam.” In Economics of Aquaculture Feeding Practices in Selected Asian Countries, edited by M. R. Hasan, 183–205. Rome: FAO. Sector Overview and Study Design 29

Reardon, T., K. Chen, B. Minten, and L. Adriano. 2012. The Quiet Revolution in Staple Food Value Chains: Enter the Dragon, the Elephant, and the Tiger. Manila: Asian Development Bank; Washington, DC: International Food Policy Research Institute. Taslim, M. A., and M. S. Haque. 2011. “Export Performance of Bangladesh: Global Recession and After.” IGC Working Paper. Dhaka: Bangladesh Foreign Trade Institute. World Bank. 2017. World Development Indicators. Washington, DC.

Chapter 3

VALUE CHAIN TRANSFORMATION

Ricardo Hernandez, Ben Belton, Thomas Reardon, Chaoran Hu, Xiaobo Zhang, and Akhter Ahmed

Introduction The majority of literature on aquaculture in Bangladesh focuses on “microso- cioeconomics” and “value chains” (VCs) and tends to have a static perspective. However, this approach is at odds with several important emerging trends (Ali 1997; Ali, Haque, and Belton 2013). First, aquaculture is growing fast in Asia. From 1984 to 2014, Bangladesh’s farmed fish jumped from 124,000 metric tons to 1.96 million metric tons, increasing by 1,580 percent. As a result, aquaculture now accounts for 55 percent of Bangladesh’s fish supply, up from just 16 percent three decades ago (Bangladesh, DoF 1994, 1997, 2006, 2015). Second, there has been a rapid shift from home consumption (from one’s own pond) to purchasing farmed fish from the market—consumers of farmed fish got 92 percent of it via purchase from the market in 2010 versus 79 percent in 2000 (data extracted from BBS 2012). This implies that “commercial aquaculture” (which we define simply as fish farming output that is sold, with no specification of the farm size) has moved to be far more important than subsistence fish farming. Third, there has been rapid diversification of farmed fish composition. This involved a shift from traditional carps to introduced species (tilapia and pangas) that lend themselves better than carp to intensification through higher stocking densities combined with the use of manufactured feeds. This is an example of what economics terms the “product cycle,” which has not been studied in Asia as an evolution in the market. Fourth, far less studied is a rapid transformation of the structure of domestic aquaculture VCs in Asia, shown by our survey results for Bangladesh. As the sector expanded, rapid commercialization and diversification of species occurred, and there was a proliferation of upstream and downstream VC actors and in some cases concentration among them. The great majority of these changes have been driven by small and medium-size enterprises. These changes can be categorized as “immanent

31 32 chapter 3

development” (Belton and Little 2011)—that is, development unplanned and undirected by government or NGOs—arising mainly from private household, firm, and community choices, driven by changes in demand, technology, communications, and infrastructure, and abetted by propitious policies. This can be contrasted with “interventionist development” (NGO projects, centralized planning by governments). The “quiet revolution” in agrifood systems in Asia—observed by Reardon et al. (2012) in rice and potatoes in Bangladesh, China, and India—is symptomatic of these broad processes of immanent development led by small farms and small off-farm enterprises. We argue that aquaculture in Bangladesh has experienced a similar quiet revolution. This chapter addresses these four trends as a confluence, with an emphasis on the latter one (structure and conduct change in the aquaculture VC in Bangladesh), with a focus on fish. We address two questions and thus impor tant gaps in knowledge about VC transformation. First, how is the domestic fish value chain restructuring? Second, how is the conduct of the segments changing in terms of product composition and technology? It is beyond the scope of this chapter to explore impacts on farmers or consumers of these VC changes—that is an agenda for further research. The chapter proceeds as follows. First, we outline the characteristics of the main geographical zones or clusters included in the study, where high concentrations of farms and other off-farm VC actors occur. Second, we address the structure and conduct changes in the various segments of the aquaculture VCs in these zones serving rural and urban markets. Third, we conclude with policy implications.

The Study Areas and Their Characteristics Using the VC surveys discussed in Chapter 2, we define the VC actors by size, in order to categorize and observe VC transformations (Table 3.1). Next we examine the descriptive statistics of these aquaculture regions over time, stratified by zones, to determine basic trends in fish farming production (Table 3.2). Table 3.2 shows selected characteristics of the four zones or clusters. Several points stand out. First, there is a fairly homogenous picture across zones in terms of general characteristics. This may be because all the areas identified as containing high densities of ponds are located in major lowland rice-growing areas with relatively easy access to the capital city, Dhaka. The study zones have broadly similar socioeconomic conditions, as compared with more peripheral and remote areas with less conducive geographies and ecology for aquaculture. Value Chain Transformation 33

Table 3.1 Definitions of actor size, by actor type

Actor Defining characteristic Size category Definition Hatcheries Total production area (ha) Small < 0.04 Medium 0.04–0.8 Large > 0.8

Feed mills Total metric tons of feed Small < 50 produced per month Medium 50 to 300 Large > 300

Input dealers Total metric tons of feed Small < 10 sold per month Medium 10 to 100 Large > 100

Farmers Total pond area (ha) Small < 0.2 Medium 0.2–0.8 Large > 0.8

Traders Total metric tons of fish Small < 1 traded per week Medium 1 to 5 Large > 5

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013.

The study zones are of similar size. They have population densities of around 1,400 persons per square kilometer, except for the Southwest (with 800 persons per square kilometer). The lower population density in the Southwest is a function of a large part of the land in its three districts being comprised of uninhabited mangrove forest. Income differences over zones are proxied by differences in monthly per capita expenditure, which is similar across the zones at roughly US$2,700 per year for a five-person household. Road density per square kilometer corresponds closely with differences in population density (lower in the Southwest, similar in other zones). The share of paved roads in total roads jumped dramatically over the 10 years from 2004 to 2014 in all zones, to more than 80 percent. The North, which had the highest share of paved roads in 2014, at 92 percent, also had the highest share in 2004, with 77 percent, indicating a historically well-developed transport infrastructure. 34 chapter 3

Table 3.2 Zone characteristics

South Item Southwest Center North East All Total area (km2) 8,710 8,492 13,752 15,809 46,763 Population density (inhabitants/km2) 757 1,354 1,388 1,445 1,302 Monthly per capita expenditure (BDT) 2,359 2,842 2,621 2,932 2,730 Road density (km of roads/km2) 2004 0.11 0.18 0.19 0.21 0.18 2009 0.12 0.19 0.21 0.20 0.19 2014 0.12 0.19 0.22 0.22 0.20 Share of paved roads in total roads (%) 2004 54 60 77 55 63 2009 79 84 92 78 84 2014 81 82 92 82 85 Aquaculture area (ha) 2004 168,493 38,614 39,795 113,994 360,896 2009 168,560 45,529 40,619 113,337 368,044 2014 220,223 116,809 126,667 111,794 575,493 Fish pond area (ha) 2004 11,815 33,570 39,758 68,037 153,180 2009 16,630 46,368 60,071 48,901 171,970 2014 24,247 44,257 60,110 71,493 200,107 Change in pond area (%) 105.2 31.8 51.2 5.1 30.6 Aquaculture production (MT) 2004 125,677 89,953 103,824 219,135 538,589 2009 140,289 114,416 121,491 231,098 607,294 2014 225,798 269,568 469,830 300,914 1,266,110 Fish production (MT) 2004 37,264 87,852 101,110 177,415 403,641 2009 32,313 149,843 292,211 156,821 631,188 2014 56,107 182,123 398,979 240,468 877,677 Change in fish production (%) 50.6 107.3 294.6 35.5 117.4

Source: BBS 2007, 2012, and 2016. Note: Aquaculture area is the total area of the two main culture systems—fish pond, shrimp/prawn farms—plus other minor production systems: pen and cage culture, culture-based fisheries in oxbow lakes baor( ) and seasonal floodplains. BDT = Bangladeshi taka. Value Chain Transformation 35

Second, as noted earlier, fish farming has developed rapidly in Bangladesh since the 1990s, accelerating during the 2000s. This is reflected in the table in the expansion of fish pond area, which grew by 31 percent across the four zones (fastest in the Southwest and North, with increases of 105 percent and 51 percent, respectively, and lowest in the East, which grew only 5 percent). Fish pond output increased more rapidly than pond area, indicating that intensification was taking place. Output rose 117 percent overall, with the greatest increase in the North (295 percent) and South Center regions (107 percent) and lowest in the East (36 percent). Third, fish production yields (per hectare of pond surface) in the zones varied considerably, with 2.3 metric tons per hectare in the Southwest, 3.4 in the East, 4 in the South Center, and 6.7 in the North. As a result of its high productivity, the North cluster (mainly Mymensingh) accounted for 45.5 percent of fish production in the four zones, from 30 percent of the pond area. Large differences in land (pond) yields across zones reflect differences in the technologies deployed, with the North being the most “advanced” zone, the “cradle” of intensive aquaculture in Bangladesh. This status is partly path dependent, reflecting a number of initial conditions, including (1) supe- rior road access to the capital city Dhaka (ADB 2005); (2) a history of commercially oriented “Green Revolution” rice farming in such districts as Bogra (Crow 2001); and (3) the location of key institutions such as the Bangladesh Fisheries Research Institute in Mymensingh district, which played an important role in the transfer of seed and production technologies for new species to well-connected farmers and hatcheries in the area (Belton and Little 2011).

Transformation of Structure and Conduct in the Value Chain This section is organized into three subsections. The first focuses on growth and concentration, the second focuses on commercialization and spatial elongation, and the third subsection focuses on technological cum product composition/product cycle change and patterns.

Growth and Concentration There has been rapid development and proliferation of the off-farm components of the fish value chain in the study zones. The combination of that plus the rapid rise in aquaculture farms in those areas is creating dense clusters of VC actors in these places. This has occurred upstream in the value chain (in hatcheries, feed milling, feed wholesale and retail, and farms) as well as 36 chapter 3

midstream and downstream (in transport, rural and urban wholesale markets [arat] and traders [aratdar], and retailing). In all segments astounding development of these enterprises—and acceleration of that development over the past 5 to 10 years—has taken place. Tables 3.3, 3.4, 3.5, 3.6, 3.7, 3.8, and 3.9 show structural change, proxied by a number of different actors and shares of the size strata in the total number. We discuss this segment by segment, from upstream to midstream.

HATCHERY SEGMENT RESTRUCTURING Over the four zones, our survey data show that there was a 207 percent increase in hatcheries over the 10 years (Table 3.3), with the rate of growth fastest in the Southwest (314 percent) and slowest in the South Center (150 percent). This rate of expansion exceeded that of either fish farm numbers (up 63 percent) or farm output (up 117 percent), thus suggesting a shift to purchased seed. Big hatcheries (≥ 0.8 hectares) accounted for 53 percent of hatchery area in 2014, a bit lower than 58 percent in 2004, showing slight deconcentration. The share of numbers of big hatcheries in all hatcheries over all zones together dropped a bit, from 15 percent in 2004 to 15 percent in 2014. But still, by 2014 the hatchery segment was concentrated: the big hatcheries had 19 percent of numbers but 53 percent of total hatchery area. The rapid increase in hatchery numbers, outstripping farm growth, and the tendency toward size deconcentration may indicate the spread of new small and medium-size hatcheries beyond original “core” clusters. The prepon- derance of hatcheries in northern Bangladesh reflects the emergence of Bogra and Mymensingh as major producers of seed, for historical reasons, as Bogra hatcheries export pangas seed to India in addition to serving the domestic market (Ali, Haque, and Belton 2013). Hatchery growth has been accompanied by rapid expansion of nurseries, particularly in nearby areas, which buy hatchlings or fry from hatcheries and raise them to fingerling size for sale to farms directly, via small traders (patil wallah), or over longer distances by larger agents. Overall (from the survey but not shown in the tables), 53 percent of the seed produced by hatcheries is sold to fingerling traders, and 44 percent direct to farmers and nurseries.

FEED MILL SEGMENT RESTRUCTURING For feed mills (Table 3.4), as with hatcheries, there is a high degree of spatial concentration in the North, where 62 percent of the country’s mills are located. Just under one-third of mills are located in the East, around Chittagong, another major industrial center, Bangladesh’s second city and main seaport. Value Chain Transformation 37

Table 3.3 Structural change in hatchery clusters over 10 years

Hatcheries Share in total number (%) Total number of small of medium of large

Zone 2004 2009 2014 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 7 14 22 0 0 0 88 86 91 13 14 9 South Center 80 102 120 7 7 5 67 65 75 25 28 20 North 177 281 410 49 50 54 33 32 31 18 18 14 East 104 142 209 37 39 39 43 43 46 20 18 15 All 368 539 761 44 46 51 37 35 35 19 18 15 Share in total production area (%) of small of medium of large 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 0 0 0 71 68 79 29 32 21 South Center 0 0 0 48 45 57 52 55 43 North 3 3 3 38 38 42 60 59 55 East 2 2 2 42 45 50 57 53 48 All 2 2 3 40 40 44 58 58 53

Source: Authors’ calculations based on the mesolevel component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2014.

The number of feed mills jumped even faster than that of hatcheries, reflecting the later introduction and adoption of feeds as compared to hatchery seed. Interviews conducted during our rapid reconnaissance indicated that there were seven to eight feed mills (defined as formal firms, not backyard feed operations on farms) in Bangladesh in 2003. The number increased 15-fold to about 100 mills by 2004. Table 3.4 indicates numbers of mills increasing by 268 percent over the period to 255 and rising fast in all zones. The feed mill segment is even more concentrated than the hatchery segment for two reasons. First, many have developed from the addition of lines by existing large domestic poultry feed firms, with a head start in the industry, from major investments by foreign companies. Second, there are economies of scale involved in sourcing raw materials, maintaining high utilization rates, and spreading fixed costs over a large volume. Big mills accounted for 67 percent of total feed production volume in 2014, although the share of volume had barely changed from 2004 (66 percent). 38 chapter 3

Table 3.4 Structural change in feed mill clusters over 10 years

Feed mills Share in total number (%) of small of medium of large Total number (< 50 MT) (50–300 MT) (> 300 MT)

Zone 2004 2009 2014 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 4 5 7 68 56 49 23 37 46 9 7 5 South Center 2 7 9 0 29 22 50 43 56 50 29 22 North 62 100 154 30 30 27 40 37 38 30 34 35 East 28 43 84 43 30 45 26 38 27 31 32 28 All 95 155 255 35 30 33 35 37 35 30 32 32 Share of total volume (%) of small of medium of large 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 17 13 11 41 59 71 42 28 18 South Center 0 3 3 28 36 48 72 61 49 North 4 3 3 33 29 29 64 68 69 East 6 3 6 23 30 26 71 66 68 All 4 3 4 30 30 29 66 66 67

Source: Authors’ calculations based on the mesolevel component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2014.

FEED DEALER SEGMENT RESTRUCTURING Input dealers (Table 3.5) are mainly feed dealers who distribute feed for mills. They are numerous, totaling 15,483 over the four zones, up from 7,690 in 2004 (hence a 2-fold growth versus a 1.6-fold growth of farmers). Like hatcheries and mills, input dealers are concentrated in the North (home to 56 percent of dealers but only 35 percent of farmers). But they are underrep- resented relative to farmers in the Southwest and the South Center, perhaps unsurprisingly given that fish yields (and thus by implication feed use) are lowest in these two zones. Some concentration in market share is present, with larger dealers having 54 percent of the traded volume in 2004. Their share dropped to 48 percent in 2014 as many new small dealers entered the scene.

FISH FARM SEGMENT STRUCTURE RESTRUCTURING Fish farmers are more evenly distributed spatially than other value chain actors over the four clusters (Table 3.6), with the North accounting for 36 percent and the Southwest and the South Center 29 percent each but the East only 9 percent. The total number of fish farmers across the four zones Value Chain Transformation 39

Table 3.5 Structural change in feed dealer clusters over 10 years

Feed dealers Share in total number (%) of small of medium of large Total number (< 10 MT) (10–100 MT) (> 100 MT)

Zone 2004 2009 2014 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 1,373 1,915 2,461 66 69 65 31 29 31 3 3 4 South Center 1,428 2,109 3,111 50 48 43 42 45 48 8 7 9 North 4,308 6,835 8,448 43 41 44 32 36 36 25 23 20 East 581 907 1,464 69 62 60 28 32 34 3 5 7 All 7,690 11,766 15,483 50 49 49 34 36 37 16 15 14 Share in total volume (%) of small of medium of large 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 13 15 12 68 67 64 19 19 23 South Center 7 7 5 61 67 63 32 27 31 North 4 4 4 31 35 38 65 61 58 East 15 11 9 65 61 58 20 28 32 All 6 5 6 41 44 46 54 51 48

Source: Authors’ calculations based on the mesolevel component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2014. grew 63 percent, from 1.08 million in 2004 to 1.76 million in 2014, but the relative share of farm numbers across zones changed little over this period. From the perspective of average household operated aquaculture land (pond surface), the average farm size changed little over the decade, even as the total population of them nearly doubled (Table 3.7). When the average area of aquaculture landholdings operated across zones between 2008 and 2013 is compared, the overall increase was moderate (7.2 percent). However, there was considerable variation between zones, with the North and the South Center registering the largest increases, up 19 percent from 0.3 hectares to 0.35 hectares, and 17 percent from 0.24 hectares to 0.28 hectares, respectively, with other zones registering little change. Fish farming households sampled in the stacked survey operated 0.29 hectares of ponds in 2008 and 0.31 hectares of ponds in 2013. Fish farmers in Bangladesh hold on average 0.86 hectares of land (fish and nonfish land combined). Therefore they are located in the second upper land quintile of farmers in the country, and they have approximately double the 0.45 hectares average 40 chapter 3

Table 3.6 Structural change in fish farmer clusters over 10 years

Fish farmers Share in total number (%) Total number of small of medium of large (’000s) (< 0.2 ha) (0.2–0.8 ha) (> 0.8 ha) Zone 2004 2009 2014 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 335 417 494 15 17 17 41 41 46 44 42 37 South Center 319 384 487 39 35 42 39 45 34 22 20 23 North 343 474 634 43 46 45 34 33 32 23 21 23 East 84 112 148 44 43 45 34 37 32 22 20 24 All 1,081 1,388 1,763 33 34 36 37 39 37 29 27 27 Share in total pond area (%) of small of medium of large 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 4 4 4 30 31 36 66 65 59 South Center 14 12 15 41 47 36 46 41 49 North 15 17 16 36 37 34 48 46 50 East 16 16 16 36 41 34 48 44 50 All 10 11 12 35 37 35 54 52 53

Source: Authors’ calculations based on the mesolevel component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2014. landholding of rice farmers (Ahmed et al. 2013). However, they are smaller than a typical fish farmer in other countries in the region, such as Myanmar and Thailand. Furthermore, the average 0.31 hectares of pond area per fish farm in our sample is five times the average 0.06 hectares area of “homestead ponds” reported by Belton and Azad (2012) in Bangladesh. As previously mentioned, the survey sampling purposively selected districts with high concentrations of fish farming, which our household survey analysis shows tend to be correlated with higher shares of commercial farms and lower shares of (subsistence) homestead fish ponds. Moreover, fish farming tends to be concentrated among the upper stra- tum of small farms. Figures extracted from Bangladesh Integrated Household Survey (BIHS) data show that 89 percent of the aquaculture households contributed just 25 percent of total production, while the top 2.4 percent of fish farming households accounted for 50 percent of total output. Our mesolevel survey data reflect this concentration. Larger farms constituted 27 percent of fish farms but had 53 percent of pond area in 2014, with little change in this Value Chain Transformation 41

Table 3.7 Landholdings and tenancy by year and zone

Southwest South Center North East All Zone 2008 2013 2008 2013 2008 2013 2008 2013 2008 2013 Total nonfish land 0.26 0.31 0.45 0.54 0.38 0.42 0.33 0.39 0.34 0.40 overall (ha/HH) (zeroes in average)

Total fish pond land (ha/ 0.59 0.64 0.24 0.28 0.30 0.35 0.49 0.48 0.43 0.46 HH) (zeroes in average) Total operated land (used, 0.51 0.52 0.18 0.21 0.17 0.22 0.20 0.20 0.29 0.31 including owned and rented-in lands) Self-owned 0.36 0.33 0.12 0.12 0.10 0.11 0.05 0.05 0.18 0.17 Joint-owned with 0.12 0.16 0.04 0.08 0.01 0.06 0.09 0.10 0.07 0.10 another HH Rented in 0.04 0.03 0.02 0.01 0.06 0.05 0.06 0.05 0.04 0.04 Rented out 0.07 0.08 0.02 0.03 0.02 0.02 0.03 0.03 0.04 0.04 Jointly owned, used by 0.01 0.03 0.04 0.04 0.10 0.11 0.25 0.24 0.10 0.10 other HH

Total land 0.85 0.95 0.69 0.82 0.68 0.77 0.82 0.86 0.77 0.86

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: HH = household. share since 2004. The share of area among farmers of other size categories also remained stable over this period, at around 35 percent for medium-size and 11 percent for small farms. This suggests that while many new producers have entered farming, there has been little if any consolidation into larger farm units. Finally, Table 3.8 shows yields of 3.35 metric tons per hectare on average in the sample in 2013. The North has far higher land yields than the rest of the zones. This helps to explain why the concentration of farms in the North is low relative to the high concentration in the North of hatcheries, feed mills, and input dealers. The North simply has a much more intensive production technology, heavy in external inputs supplied by these off-farm enterprises, and that intensification is reflected in its extraordinary yields relative to the rest of the fish farming clusters. Absolute yield growth was greatest in the North, up from 5 metric tons per hectare to 10 metric tons per hectare, while in relative terms the East grew faster from a lower base of 1.3 metric tons per hectare to 2.9 metric tons per hectare. The South Center region grew from 42 chapter 3

Table 3.8 Factor productivity in 2013

Southwest South Center North East All Observations in 2013 465 280 420 340 1,505 Total fish output (kg) per Labor day (own + hired) 9.0 7.2 32.7 20.8 18.9 Own labor day 10.3 10.5 42.5 23.8 23.9 Pond hectare (kg) 801 2,565 10,017 2,916 3,352 Capital (thousand BDT)a 5.3 9.8 25.2 10.2 13.0

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: a Capital is calculated as the total variable cost (labor, rent, purchased inputs, and so on), plus the annual amortization of quasi-fixed assets. BDT = Bangladeshi taka.

Table 3.9 Structural change in rural fish trader clusters over 10 years

Rural fish traders Share in total number (%) Total number of small of medium of large (hundreds) (< 1 MT) (1–5 MT) (> 5 MT)

Zone 2004 2009 2014 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 42 46 59 71 74 76 26 22 20 4 4 4 South Center 24 33 38 61 59 45 28 26 35 11 15 20 North 34 54 128 61 58 76 22 23 14 17 19 9 East 48 74 88 68 74 71 24 21 23 8 6 6 All 148 207 313 66 67 71 24 22 20 9 10 9 Share in total volume (%) of small of medium of large 2004 2009 2014 2004 2009 2014 2004 2009 2014 Southwest 21 23 23 45 41 37 34 36 40 South Center 11 9 5 30 24 24 59 67 71 North 9 7 17 19 18 19 73 75 64 East 15 20 18 32 33 35 53 47 48 All 13 13 15 30 26 26 57 61 58

2 metric tons per hectare to 2.6 metric tons per hectare, while the Southwest remained stagnant at 0.8 metric tons per hectare.

FISH WHOLESALE SEGMENT RESTRUCTURING The fish wholesale segment has expanded rapidly. This occurred with a proliferation, especially during the 2000s, of rural fish wholesale markets and an increase in fish wholesale markets in such cities as Dhaka. Fish trader numbers more than doubled across the four zones, from 14,800 in 2004 to 31,300 in 2014 (2.1 times versus 1.6 times for farmers) as shown in Table 3.9. The largest share of traders was in the North (41 percent) in 2014, which had only 23 percent of the traders in 2004. Trader startups followed the concentration of fish production; most village traders started about 10 to 15 years ago, over the same period as the beginning and development phase of the aquaculture boom.

Commercialization and Spatial Elongation In Bangladesh, mainly since 2005, the value chain for farmed fish has com- mercialized and “lengthened” geographically. This is a typical trend in transformation and modernization of food supply chains, with concomitant interprovince market integration and reduction of transaction costs. This happened, for example, in rice and potatoes in Asia over the past decade (Reardon et al. 2012). The commercialization of fish farming is itself dependent on the proliferation of services discussed above: the development of off-farm components of the value chain permits a division of labor wherein small farmers can specialize in pond operations and enjoy cost savings via economies of scale, economies of scope, and economies of agglomeration by relying on the upstream feed and seed purveyers and downstream wholesale and logistic services that themselves are specialized enterprises. Belton, Ahmed, and Jahan (2014) contend that the availability of these services facilitates the entry of smaller producers into commercial fish farming; Reardon et al. (2012) contend similarly for rice and potato sectors. We discuss the trends of commercialization and chain lengthening below, from upstream to midstream.

SEED COMMERCIALIZATION Farmers have shifted from trapping wild fish on their farms or buying locally available wild seed in the early 1990s (Ahmed, Rab, and Bimbao 1993) to stocking hatchery-produced seed in the 2000s. By 2011, 98 percent of fish seed was produced by private hatcheries (Belton et al. 2011). The shift to hatchery-produced seed resulted in a lengthening of the distances over which seed was traded, which evolved due to a mix of initial environmental and 44 chapter 3

institutional conditions and the location of sources of demand. These trends have given rise to a situation in which there is a correlation of the level of activity of the broad cluster and the co-location of hatcheries. There is strong spatial concentration of hatcheries in the North, which has more than half of the hatcheries in the four zones. But our survey also found that on average half of what hatcheries produce is sold to buyers outside of their own district (not shown in the tables). Hatcheries thus tend to be “shared” across districts and even zones.

FEED COMMERCIALIZATION The commercialization of aquaculture feeds and the geographical lengthening of that segment have occurred in lockstep. There has been a long-term shift from little use of feed of any type (Ahmed, Rab, and Bimbao 1993) to use of feed available on-farm (for example, cow manure, rice bran), to purchase of the latter, and increasingly to purchase of formulated pelleted feeds; 90 percent of the latter are made by medium- and large-scale commercial mills in 2015 (Mamun-Ur-Rashid et al. 2013). Large feed mills in the peri-urban industrial zone north of Dhaka, where most feedlot poultry farming occurs, distribute feed throughout the country. Mills are concentrated there for centralized acquisition of inputs and because (similar to other countries) many fish feed manufacturers originally produced poultry feeds before diversifying into fish feed by adding additional lines. Of the 25 largest poultry feed mills in Bangladesh, 18 also produce fish feeds (Khaleduzzaman and Khandaker 2009). The input acquisition supply chain for feed manufacture stretches over long distances. Most dried fish, one of the main ingredients in fish feed, is sourced from marine fisheries in coastal districts of Bangladesh but is increasingly also imported, as is soy (another key ingredient). Meat and bone meals, important protein sources for feed, are sourced from the European Union (Mamun-Ur-Rashid et al. 2013). Much of the equipment used in the value chain (for example, feed milling machines, vehicles, pumps, cold chain equipment) is imported, mainly from East and Southeast Asia, as are chemicals (Mamun-Ur-Rashid et al. 2013). As elsewhere in Asia, foreign expertise has played an important role in the development of hatchery and feed operations (Belton 2012). For instance, internationally led training received by Bangladeshi entrepreneurs has been important in the establishment of mono- sex tilapia hatcheries. Over time, dependence on these sources of information has lessened as technical knowledge has become more widely available within Bangladesh. Value Chain Transformation 45

FISH FARM COMMERCIALIZATION The shift from subsistence to commercial production in the fish sector occurred as initially fish were only home-consumed from the household pond, then increasingly sold into nearby markets, and then also marketed to more distant urban markets. These sequential changes have occurred rapidly. As recently as the early 1990s, Ahmed, Rab, and Bimbao (1993) observed that only a small fraction of total harvested farmed fish entered the market outside the local village. In contradiction to the traditional view of fish farming in Bangladesh as mainly subsistence oriented, the value chain survey of the farm segment shows that 75 percent of households who engage in fish farming sell fish. Strikingly, the share of farms with a marketed surplus even in these dense aquaculture clusters was only 57 percent just five years prior to this, indicating that extremely rapid commercialization occurred. The Southwest has the highest share of fish farming households marketing fish (88 percent), in line with shrimp and prawn production in that zone. Figures are around 70 to 75 percent for the South Center and the North, and 60 percent for the less advanced (in terms of fish farming) East. Table 3.10 shows the disposal of fish output by aquaculture households in 2013 by final user type (note that only 8 percent of the fish was home consumed). Interestingly, although yields differ significantly across zones, the marketed surplus rate does not—the home consumption share is 4 percent in the North zone and about 15 percent in the other zones. Moreover, in contrast to the common image of the rural fish market being dominated by small rural brokers, the market has shifted to rural sourcing by large wholesalers based in towns and secondary cities. Tables 3.10 and 3.11 show that about two-thirds of the marketed volume goes to large wholesalers; again, that differs between the North with 68 percent and the average of the other zones at about 54 percent. By contrast, local rural brokers have a mere 5 percent share of the market as shown in Table 3.11. Just over a decade earlier, fish farmers usually sold their fish to local traders or fish col- lectors (ADB 2005). Interestingly, this is the same market structure development that has occurred in rice and potatoes in Asia (Reardon et al. 2012).

RISE OF THE FISH TRADER SEGMENT TO URBAN AREAS Growth in sales of farm output has been accompanied by a proliferation of traders in the midstream segment of the chain. As urban demand has grown and the road network has developed (see Table 3.2), fish is increasingly sold by traders in the zones of production to Dhaka and from one division to another. The national Household Income and Expenditure Survey (HIES) (BBS 2012) 46 chapter 3

Table 3.10 Disposal of fish farm harvest by final user type, 2013

Southwest South Center North East All Observations in 2013 465 280 420 340 1,505 Share of farmers who grow 87.5 70.7 74.5 59.7 74.5 fish selling fish Farmer’s own consumption 15.4 18.5 4.2 12.2 8.4 Sales through different value chains Consumed by another farm 0.3 0.0 0.0 0.0 0.0 household Direct consumer 0.0 0.7 0.9 0.0 0.7 Retailer at traditional market 11.4 7.7 2.7 15.6 6.0 Assembler (collector) 5.6 2.8 8.1 14.7 8.4 Large wholesaler 49.2 54.5 68.1 56.4 62.8 Supplier (broker) 1.5 8.6 8.7 0.6 6.6 Supermarket 0.0 0.0 0.0 0.0 0.0 Auctioned 16.6 7.1 7.3 0.5 7.1 Others 0.1 0.0 0.0 0.0 0.0 Total 100.0 100.0 100.0 100.0 100.0

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: “Large wholesaler” refers to the top wholesalers in the market. shows that from 2000 to 2005 the share of fish consumed in urban areas rose from 29 percent to 42 percent. The conduct of the segment has also changed from the traditional image common in Asia that traders are advancing funds to farmers to lock in farmers in transactions. Our survey showed that none of the farmers received any cash advance from fish traders. That was confirmed by our trader survey. But the trader survey showed that around 40 percent of traders, both rural (operating from villages) and rural-urban (operating from secondary cities or towns), provide advances of working capital to other traders to secure supplies of fish, with an average loan duration of just under one month. Among rural and rural-urban traders, our survey shows that the great majority have stalls in rural and rural-urban wholesale markets. Most rural fish traders (63 percent) and rural-urban traders (79 percent) take a commission on the transaction of fish (rather than through arbitrage where they buy and then sell). None of the rural traders and few of the rural-urban traders surveyed owned trucks, and only 6 percent rented them, indicating their role as intermediaries who operate from a base and just link buyers and sellers, Value Chain Transformation 47

Table 3.11 Disposal of fish farm harvest by final user location, 2013

Southwest South Center North East All Observations in 2013 465 280 420 340 1,505 Farmer’s own consumption 15.4 18.5 4.2 12.2 8.4 Sales through different value chains Consumed by another farm household 0.3 0.0 0.0 0.0 0.0 Direct consumer 0.0 0.7 0.9 0.0 0.7 Retailer at local (village, union, upazila) 11.4 7.7 2.7 15.6 6.0 traditional market Assembler locally 5.6 2.6 5.5 14.7 6.7 Assembler in same district 0.0 0.2 2.5 0.0 1.7 Large wholesaler locally 46.3 46.7 64.4 31.8 55.5 Large wholesaler in same district 2.7 4.3 2.9 14.8 4.9 Large wholesaler in different district 0.1 3.5 0.8 9.7 2.3 Supplier (broker) locally 1.2 8.6 6.9 0.6 5.4 Supplier (broker) in same district 0.3 0.0 0.0 0.0 0.0 Supplier (broker) in different district 0.0 0.0 1.8 0.0 1.2 Supermarket 0.0 0.0 0.0 0.0 0.0 Auctioned locally 16.3 5.4 1.9 0.0 3.4 Auctioned in same district 0.1 1.1 5.3 0.0 3.6 Auctioned in different district 0.2 0.6 0.1 0.5 0.2 Others 0.1 0.0 0.0 0.0 0.0 Total 100.0 100.0 100.0 100.0 100.0

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. relying on hiring transporters (or having the farmer hire transporters) to deliver. The average monthly working capital of rural traders was a little under half that of rural-urban traders, as expected. Few traders (less than 2 percent) owned ice-making plants, and almost none owned cold storage. Only 31 percent of rural traders and 20 percent of rural-urban traders reported icing the fish. This likely reflects their role as commission agents, who rapidly broker sales between buyers and sellers, without taking possession of the fish themselves, with buyers usually assuming responsibility for procuring ice from ice suppliers or manufacturers. The low ice use rate is not for lack of access to ice firms: 80–90 percent of traders felt they had good access to ice firms. The domestic market demands whole fresh fish, with little if any value addition occurring. It is thus not observed that wholesalers, feed companies, or hatcheries process their own fish. 48 chapter 3

Technological cum Product Composition/Product Cycle Change and Patterns Important interlinked changes have occurred in the technologies and the product composition of farm production concurrent with the above structural changes in the value chain.

THE PRODUCT CYCLE The product cycle is a widely observed feature of product development in many sectors of the economy, which can be observed for a range of agricultural sectors, including fish and fruit, in a number of other countries. Sequentially, the five stages of the product cycle are (1) the local niche product stage; (2) the commodity stage, during which a local (or exotic) niche product is “commoditized” by production in large quantities, driving down costs, but with little product variety or quality differentiation; (3) the product differentiation stage, when the commodity becomes differentiated along the lines of several possible tangible and intangible attributes (for example, variety, quality, organic versus conventional, confined versus free range); (4) the commoditi- zation stage, where the differentiated products are themselves produced on larger scale and commoditized; and (5) the introduction of new niche or differentiated products. The cycle can continue indefinitely depending on the capacity of innovation in the sector and the market. We posit that the Bangladesh fish sector has followed a typical product cycle development path, facilitated by the linked technology changes along the value chain described above, although to date only the three stages can be dis- cerned. The first (and ongoing) technology change linked to the first product cycle step (moving from niche to commodity) is the shift from capture of wild fish stocks from open waters to their production in ponds under controlled conditions. This shift began in earnest during the 1980s, as ponds were increasingly used for aquaculture, primarily by stocking native carp species, an important component of inland capture fisheries at that time (Ali 1997). As the decades progressed, additional species were introduced and commoditized, such as exotic carps in the 1980s, pangas in the mid-1990s, and mono- sex Nile tilapia in the early 2000s (Ali, Haque, and Belton 2013; Belton and Little 2011). Later, in the late 2000s, many additional native fish that were formerly available only from domestic capture fisheries were incorporated into farm production. Data from BIHS depict the consumption side of the product cycle (Figure 3.1). After becoming commoditized during the 2000s, pangas and Value Chain Transformation 49

Figure 3.1 Average weekly consumption per capita of the 10 most consumed fish species in rural Bangladesh

50 Q1 Q2

Q3 Q4 40 Q5 ct r

er 30

20 couto ee 10 ere

0 Pangas *Silver Tilapia *Rui *PutiTaki Mrigel * Dry Koi *ShrimpIlish carp * ﬁsh

Source: Derived by authors from the BIHS 2011–2012 dataset (Ahmed 2013). Note: Q1 = expenditure quintile 1, and so on; fish marked with an asterisk are produced predominantly from aquaculture. tilapia are now, respectively, the first and third most consumed fish in rural areas of Bangladesh. Rui, an indigenous carp species that was among the first fish to be commoditized and traditionally the most important farmed fish, is now ranked fourth in terms of consumption. The recently commoditized koi now ranks eighth. This change has occurred quickly. Comparison with similar household survey data collected in 2006–2007 shows rui still as the most consumed fish at that time, with pangas third, tilapia ninth, and koi not yet produced in sufficient quantities to feature (Belton, van Asseldonk, and Thilsted 2014). Our farm survey also shows evidence of product cycle changes from the production side in each zone in 2008 compared with 2013 (Table 3.12). There is a significant increase in production of pangas, tilapia, and niche species in the North as well as niche species in the East. It is interesting that production of niche species has increased slightly in the study zones, but in the North increased 72 percent, and in the East increased 16 times, though from a low base. 50 chapter 3

Table 3.12 Aquaculture production by fish category, 2008 and 2013

Southwest South Center North East Total Zone 2008 2013 2008 2013 2008 2013 2008 2013 2008 2013 Production (kg/year) Carps 173 175 235 325 521 928 233 369 291 445 Tilapia 58 71 60 99 182 503 169 309 116 244 Pangas (yellowtail catfish) 0 0 66 105 544 1,358 103 207 180 401 Shrimp 129 134 7 8 0 0 12 7 48 51 Niche 46 45 58 69 186 408 18 285 172 200 Others 42 46 60 70 92 216 94 137 70 117 Total 448 470 485 675 1,525 3,414 630 1,314 876 1,458 Share of production (%) Carps 39 37 49 48 34 27 37 28 33 31 Tilapia 13 15 12 15 12 15 27 24 13 17 Pangas (yellowtail catfish) 0 0 14 16 36 40 16 16 21 28 Shrimp 29 28 1 1 0 0 2 1 5 3 Niche 10 9 12 10 12 12 3 22 20 14 Others 9 10 12 10 6 6 15 10 8 8

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013.

CAPITAL-LED INTENSIFICATION The conventional image of the pond-fish sector was once millions of backyard “homestead ponds,” used primarily for subsistence (home consumption). Dey, Bose, and Alam (2008) refer to aquaculture in Bangladesh as a “low- input activity for household consumption.” However, the small and medium- size commercial farms that now dominate aquaculture output are making a transition from the traditional production technologies toward intensification—first by labor and then by productive capital (such as formulated feed and medicines and some equipment such as aeration). The main technology changes observed in the survey are as follows.

Rapid increase of purchased seed and feed Use of wild fish seed has been replaced by use of hatchery produced fish seed. Jahan et al. (2015) report that fewer than 4 percent of farmers use fish seed from open-water sources. Seed stocking density has also increased. Our farm survey shows that nominal expenditure on fingerlings per hectare more than doubled between 2008 and 2013. Moreover, there has been a transition in zones surveyed from no use of feed, to use of feed inputs available on-farm, Value Chain Transformation 51

to use of handmade feeds made with ingredients purchased off-farm, to the use of purchased manufactured feeds, formulated to meet the complete nutritional requirements of the fish produced and, increasingly, from formulated sinking feed to floating formulated feed. The latter allows for greater production efficiencies via reduction of waste and higher digestibility if the pond is stocked with medium- and top-dwelling fish (Mamun-Ur-Rashid et al. 2013). Strikingly, our farm survey showed that 38 percent of farmers in 2013 used commercial pelleted feeds, up from 30 percent in 2008; both figures are surprisingly high and run counter to the common image of fish farming in Bangladesh as mostly extensive or semi-intensive (ADB 2005). Use of commercially manufactured pelleted feeds is significant because their use can increase fish growth rates, facilitating higher yields, and is consistent with product cycle–driven diversification into species of fish (for example, pangas, tilapia, koi) that require formulated diets to attain optimal growth (Mamun- Ur-Rashid et al. 2013). Our survey results show that use of other supplementary feeds usually associated with semi-intensive farming (for example, rice bran, mustard oilcake) was already widespread in 2008 (60 percent of farmers) but grew to 69 percent of farmers by 2013. This indicates a range of stages of intensification, with some farmers shifting from extensive to semi-intensive production and others “upgrading” to more intensive production with pelleted feeds.

Rapid increase in the use of chemicals Use of medicines and other chemicals—including lime, antibiotics, salt, fungi- cides, insecticides, and feed additives such as vitamins—have increased in line with higher stocking densities and feed use. This has occurred as the incidence of disease has increased and better management has been necessary to maintain water quality within the parameters required for fish survival and growth (Ali et al. 2016). Our farm survey shows that use of lime for pond preparation was already a widespread practice in 2008 (63 percent of farmers) and was adopted by 73 percent of farmers by 2013. Use of medicines and vitamins was less common (6 percent of farms in 2013), increasing slightly from 4 percent in 2008. Use of both these inputs was greatest in the North, reflecting its intensified technology, and lowest in the least commercial zone, the East.

Increase of use of hired labor Hired labor use has increased along with the overall need for labor. Whereas in the early 1990s, aquaculture used little household labor as compared to crop or livestock cultivation (Ahmed, Rab, and Bimbao 1993), commercial forms 52 chapter 3

of aquaculture in Bangladesh now generate higher average demand for hired labor per unit area of land than paddy cultivation, due in part to the long fish cropping cycles (Belton, Ahmed, and Jahan 2014) and more yield and input use to manage. Our farm survey data show that both total labor inputs and inputs of hired labor have intensified per unit of (pond) land. The total outlay per hectare for labor (family plus hired) in 2013 was 1.6 times what it was in 2008. The share of hired (nonfamily) labor in total labor (hired plus own labor) increased from 11 percent to 21 percent over the same period. Hiring labor was concentrated among a small subset of commercial farms. The share of farms hiring casual workers was 16 percent for pond preparation, 5 percent for stocking, 5 percent for the post-stocking/growing-out stage, and 6 percent for harvesting. The share of farms hiring salaried or permanent laborers was much lower still. Moreover, differences in total labor inputs and in hired labor varied markedly across zones, in line with patterns of intensification. The capital to labor ratio in 2008 was more than 50 percent higher in the Southwest and North than in the South Center and East (Table 3.13). By 2013 these differences are even more extreme, with capital-to-labor ratio in the Southwest and North more than 100 percent higher than in the South Center and East.

Rapid increase/investment in quasi-fixed capital (equipment) There was substantial investment by fish farmers in productive capital. Investments (at nominal dollar rates) in assets used for fish farming jumped by 235 percent over 2008–2013.1 The rate of investment was similar over all zones, with the exception of the South Center, which was somewhat slower. However, the rapid jump in productive fish-related capital holdings masks the fact that few fish farms had these. Pumps, owned by 14 percent of households, were by far the most frequently owned item. Investment growth was thus from a relatively concentrated base among the small and medium-size commercial farms. The dollar value of the stock of quasi-fixed capital inputs (the rate of investment in capital) on fish farms rose faster than labor flow (total use of labor, measured in value terms) from 2008 to 2013. Investment in agricultural equipment by fish farmers was reflected in productive capital stocks used for crop farming increasing by 185 percent. This too was similar across zones. Fish farmers also invested in livestock (increasing by 533 percent), while nonfarm productive assets climbed 730 percent and consumer durables by 240 percent. Even discounting these rates for inflation,

1 Assets used in fish farming were calculated by the summation of the value of the following common assets: pumps, generators, aerators, nets, weighing scales, boats, bicycles, motorcycles, pickup trucks, and trucks. Value Chain Transformation 53

Table 3.13 Fish farm capital-to-labor ratio by year and zone

Item Southwest South Center North East All Capital-to-labor ratio (2013) 1.52 0.39 0.63 0.48 0.75 Capital-to-labor ratio (2008) 0.96 0.33 0.47 0.32 0.51 Change, 2008 to 2013 (%) 58 18 34 50 47

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. these findings show that fish farmers were a vibrant capital-accumulating segment. In conclusion, from 2008 to 2013 there was a remarkable increase in use of external inputs by farms. Total outlay on external inputs (feed, fertilizer, chemicals, and so on) tripled per hectare. The rate of external input use did not differ a great deal across three of the four zones. Average expenditure on inputs was much greater in the North however, reflecting the intensive technologies used there. Despite the rapid increase in total outlay on external inputs, the composition of that outlay was stable. Feed remained central, accounting for 78–79 percent of external input costs in both 2008 and 2013. Chemical fertilizer stayed at 5 percent; fuel 7–8 percent; and pesticide and lime 5–6 percent.2 Fifty percent of farmers (with a lower share in the East) used chemical fertilizer in the prestocking phase, up sharply from 39 percent. Use of other inputs also increased: 75 percent of farms used lime for pond preparation in 2013, up from 63 percent, and a third used water pumps for filling the pond in 2013, up from one-fifth, both indicating greater attention to maintenance of good water quality. In combination, these trends all point to a broad shift away from “traditional” low input forms of production, with capital intensification in the farm sector occurring in step with rapid growth, diversification, and technological change in the feed and seed value chain segments.

Mirroring of Farm Capital–Led Intensification: Growth and Technological Change in the Input Supply Segments Capital intensification in the farm sector has occurred in step with rapid growth and diversification in upstream value chain segments, most importantly feed and seed. As noted previously, there has been a proliferation of feed mills but overall a concentration of production in larger mills. Larger feed mills

2 Fertilizer is used to induce growth of phytoplankton in the pond, which provides a natural feed for fish. 54 chapter 3

have multiple lines and differentiate feed types for different fish species and age groups by protein content, complementing the species differentiation taking place on farms and in hatcheries in line with the product cycle. The most recent shift in feed mill technology has come about through the use of extrusion machinery to produce feeds that float instead of sink, providing further product differentiation and offering efficiency gains to producers. Mamun-Ur- Rashid et al. (2013) found that from 2008 to 2012 production of formulated fish feeds almost tripled from 360,000 metric tons to an estimated 1 million metric tons. The share of floating feeds in total formulated feed production grew from less than 5 percent to close to 20 percent. The extrusion equipment used to manufacture floating feed represents a major investment. The size of mill and adoption of extrusion are closely correlated, and big mills dominate the floating feed supply. This factor may result in further concentration in the industry over time. The growing use of formulated feed has also seen rapid expansion in the number of fish feed dealers (both wholesalers and retailers), as noted previously, and veterinary chemical/medicine input retailers. Technological shifts have also occurred within the seed supply segment of the value chain. The product cycle is observed in hatcheries, mirroring farm production—the share of hatcheries reporting carps to be the most impor tant species sold fell from 59 percent to 45 percent over 2008 to 2013, while the share of hatcheries reporting pangas and tilapia (combined) as their most important species rose from 15 percent to 20 percent and the share of “niche” species jumped from 11 percent to 26 percent.

Conclusion The chapter has produced a single, powerful finding: the fish value chain in Bangladesh is growing and transforming rapidly, in all segments. The quiet revolution in the fish value chain is a domestic market revolution. Dynamism of the aquaculture value chain in Bangladesh was shown in two interlinked ways. First, there has been a tripling of volumes and actors in all the segments of the value chain since 2008. Also, there has been rapid capital deepening in the form of investments by hundreds of thousands of actors in the fish value chain. This is apparent in a great jump in feed use, investment in equipment and pond construction, and investments in mills, hatcheries, and vehicles. These investments have been made by, and provided opportunities for, a multitude of smallholder farmers as well as small and medium-size enterprises throughout the chain. Second, there has been diversification and Value Chain Transformation 55

specialization beyond carps into production of commercial species such as tilapia and pangas, which have raised yields and helped to move the fisheries sector along the product cycle. One important positive externality of this process has been a reduction in the price of farmed fish over time, making an important contribution to food security.

References ADB (Asian Development Bank). 2005. An Evaluation of Small-Scale Freshwater Rural Aquaculture Development for Poverty Reduction. Operations Evaluation Department. Manila. Ahmed, A. 2013. Bangladesh Integrated Household Survey (BIHS) 2011–2012. Washington, DC: International Food Policy Research Institute (IFPRI). Ahmed, A. U., K. Ahmad, V. Chou, R. Hernandez, P. Menon, F. Naeem, F. Naher, W. Quabili, E. Sraboni, B. Yu, and Z. Hassan. 2013. The Status of Food Security in the Feed the Future Zone, and Other Regions of Bangladesh: Results from the 2011–2012 Bangladesh Integrated Household Survey. Washington, DC: IFPRI. Ahmed, M., M. Abdur Rab, and M. P. Bimbao. 1993. Household Socioeconomics, Resource Use and Fish Marketing in Two Thanas of Bangladesh. Manila: International Center for Living Aquatic Resources Management (ICLARM).

Ali, H., M. M. Haque, and B. Belton. 2013. “Striped Catfish (Pangasianodon hypophthalmus, sauvage, 1878) Aquaculture in Bangladesh: An Overview.” Aquaculture Research 44 (6): 950–965. Ali, H., A. Rico, K. Murshed-E-Jahan, and B. Belton. 2016. “An Assessment of Chemical and Biological Product Use in Aquaculture in Bangladesh.” Aquaculture 454: 199–209. Ali, M. Y. 1997. Fish, Water and People: Reflections on Inland Openwater Fisheries Resources of Bangladesh. Dhaka: University Press Ltd. Bangladesh, DoF (Department of Fisheries). 1994. Fisheries Statistical Year Book of Bangladesh 1992– 1993. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock. —. 1997. Fisheries Statistical Year Book of Bangladesh 1995–1996. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

—. 2006. Fisheries Statistical Year Book of Bangladesh 2004–2005. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

—. 2015. Fisheries Statistical Year Book of Bangladesh 2013–2014. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

BBS (Bangladesh Bureau of Statistics). 2007. Statistical Yearbook of Bangladesh 2005. Dhaka. —. 2012. Statistical Yearbook of Bangladesh 2011. Dhaka. 56 chapter 3

—. 2016. Statistical Yearbook of Bangladesh 2015. Dhaka: Ministry of Planning, People’s Republic of Bangladesh. Belton, B. 2012. “Culture, Social Relations and Private Sector Development in the Thai and Vietnamese Fish Hatchery Sectors.” Asia Pacific Viewpoint 53 (2): 133–146. Belton, B., N. Ahmed, and K. M. Jahan. 2014. Aquaculture, Employment, Poverty, Food Security and Well-Being in Bangladesh: A Comparative Study. Penang, Malaysia: CGIAR Research Program on Aquatic Agricultural Systems. Belton, B., and A. Azad. 2012. “The Characteristics and Status of Pond Aquaculture in Bangladesh.” Aquaculture 358: 196–204. Belton, B., M. Karim, S. Thilsted, K. Murshed-E-Jahan, W. Collis, and M. Phillips. 2011. “Review of Aquaculture and Fish Consumption in Bangladesh.” Studies and Reviews 2011-53. Penang, Malaysia: The WorldFish Center. Belton, B., and D. C. Little. 2011. “Immanent and Interventionist Inland Asian Aquaculture Development and Its Outcomes.” Development Policy Review 29 (4): 459–484. Belton, B., I. J. M. van Asseldonk, and S. H. Thilsted. 2014. “Faltering Fisheries and Ascendant Aquaculture: Implications for Food and Nutrition Security in Bangladesh.” Food Policy 44: 77–87.

Crow, B. 2001. Markets, Class and Social Change: Trading Networks and Poverty in Rural South Asia. New York: Palgrave. Dey, M. M., M. L. Bose, and M. F. Alam. 2008. Recommendation Domains for Pond Aquaculture. Country Case Study: Development and Status of Freshwater Aquaculture in Bangladesh. Penang, Malaysia: WorldFish.

Jahan, K. M., B. Belton, H. Ali, G. C. Dhar, and I. Ara. 2015. Aquaculture Technologies in Bangladesh: An Assessment of Technical and Economic Performance and Producer Behavior. Penang, Malaysia: WorldFish. Khaleduzzaman, A. B. M., and Z. H. Khandaker. 2009. “Commercial Feed Production and Quality Control: Present Status and Future Prospects in Bangladesh.” In Proceedings of the Sixth International Poultry Show and Seminar. Dhaka: World Poultry Science Association, Bangladesh Branch.

Mamun-Ur-Rashid, M., B. Belton, M. Phillips, and K. A. Rosentrater. 2013. Improving Aquaculture Feed in Bangladesh: From Feed Ingredients to Farmer Profit to Safe Consumption. Penang, Malaysia: WorldFish.

CLUSTER-BASED AQUACULTURE GROWTH

Xiaobo Zhang, Qingqing Chen, and Peixun Fang

Introduction As shown in Chapter 3, fish production appears to be largely clustered and the number of fish farmers, feed traders, and fish traders have all experienced rapid growth since 2008, roughly in the same magnitude. The first objective of this chapter is to quantify the trend of clustering. Based on the fish value chain survey and mesolevel primary data, we show that fish production has indeed become clustered over time. When a large number of actors work on the same sector in a limited area, the competition is inherently intense. A question arises: Why do people still want to co-locate to work on similar businesses? The cluster must create some collective efficiency, which offsets the adverse effect on profit margin due to strong competition (Schmitz 1995). Better access to market, easy learning from others, and labor pooling are the three most noted features of positive externalities in clusters (Marshall 1920). In developing countries, clustering can help to alleviate entrepreneurs’ financial constraints, a major limiting factor to private sector development, by lowering capital barriers to enter and providing trade credit for operation (Ruan and Zhang 2009; Ali, Peerlings, and Zhang 2014). Apart from these positive externalities, the actors in clusters also share market information and tools (Humphrey and Schmitz 1998; Schmitz and Nadvi 1999; Felzensztein and Gimmon 2009).1 They not only compete with each other but also collaborate in many ways. However, the empirical studies on cooperation are much more limited when compared to the vast body of literature on competition. As stated in Schmitz and Nadvi (1999, 1508), “While inter-firm cooperation has often been highlighted as a central feature of the more successful contemporary industrial clusters, it has rarely been investigated systematically, let alone quantified.” The studies on cooperation in agricultural or aquaculture clusters are even more scant than in industrial clusters. The second objective of the chapter is to show that cooperation is rather

1 For example, weighing scales, crates, or barrels.

57 58 chapter 4

common in Bangladesh fish farming clusters. In areas with higher degrees of clustering, cooperation is more widespread. The third aim is to examine the impact of clustering on specialization. In industrial clusters the production process is often divided into many incremental steps, which are undertaken by dispersed producers. As a result, clusters are normally associated with a high degree of production specialization. However, we find that clustering has little to do with fish production specialization in Bangladesh. As in industrial clusters, we find that fish farmers adopt more modern inputs to boost yield than those in less clustered regions.

Surveys The fish value chain survey covers five major actors—hatcheries, feed mills, feed dealers, fish farming households, and traders. Sampling procedures for the fish value chain survey can be found in Chapter 2. The sampled 20 districts colored in green are plotted on the map in Figure 4.1. Figure 4.2 plots the spatial distribution of these actors. The mesolevel survey measures the inventory of actors along the fish value chain at the upazila and district levels in 2003, 2008, and 2013. In each mouza two to five key stakeholders were interviewed. The key stakeholders include community leaders, officials from the fishery department, feed dealers, and fish traders. Multiple interviews were conducted to triangulate information from different sources, check data consistency, and eliminate outliers.

Measuring Clusters A “cluster” is defined as “a geographically proximate group of interconnected companies (and associated institutions) in a particular field” (Porter 2000, 16). Ideally a clustering measure should embody both concentration and relatedness. Conventional measures of industrial agglomeration—such as the concentration index, Hirschman–Herfindahl Index (HHI), and Krugman Index—are often used as a proxy for clustering in the literature. While these indexes mainly capture one aspect of clustering, meaning the degree of concentration in a place, they largely ignore relatedness, another important aspect of clustering. To overcome this limitation, Long and Zhang (2012) use the proximity matrix of products to measure proximity between industries based on the idea of production space (Hidalgo et al. 2007). The matrix of product proximity is calculated by the revealed comparative advantage of export products. Since most cultured fish in Bangladesh are for domestic consumption, it Cluster-based Aquaculture Growth 59

Figure 4.1 Twenty sample districts in fish-clustering areas

Source: Derived by the authors from the BIHS 2011–2012 dataset (Ahmed 2013). Note: The green areas represent the sampled 20 districts using the sampling procedures in Chapter 2. is not possible to use the export-based product proximity approach to measure relatedness in the context of the fish supply chain in Bangladesh. Therefore the approach cannot be directly used for the measurement of clustering here. Instead, in this chapter we use an alternative simple approach to measure clustering. Concentration can be measured by the number of actors engaged in the fish supply chain. However, using absolute values will favor large areas, 60 chapter 4

Figure 4.2 Distributions of number of actors per 1,000 rural people in 2013, by districts

Fish farmers Feed dealers

Fish traders Feed mills

Hatcheries

Source: Derived by the authors from the BIHS 2011–2012 dataset (Ahmed 2013). Note: The darker the blue is, the higher the number of actors per 1,000 rural people, for each district. Concentrated areas of each actor are circled in red. Cluster-based Aquaculture Growth 61

which naturally have more people working in the fishery sector. To remedy this bias, we use the local population and area to normalize the number of actors and total pond area, respectively. As for relatedness, we are more con- cerned about the links among four major actors in the fish supply chain—fish farmers, hatcheries, feed dealers, and traders.2 Although the four actors form the value chain, feed dealers, fish farmers, and traders are more spatially concentrated than hatcheries. Subject to strict requirements of weather conditions and water quality, the location of hatcheries is crucially determined by natural conditions, such as rainfall, water quality, and temperature. Therefore they are not necessarily located near cultured fish ponds. As indicated in Table 4.1, 62 percent of hatcheries sell their fingerlings to other districts. Because there is a lower transportation cost for feeds compared to perishable fish, it is not so essential for feed mills to situate in fish production areas. In fact, 43 percent of feed dealers buy feed outside their own districts. Fish farmers, traders, and feed dealers tend to co-locate in clusters. As shown in Table 4.1, 97 percent of farmers sell their fish to regular traders in the same district, and 93 percent of feed dealers sell inputs to fish farmers in the same district. Because fish farmers, input suppliers, and fish traders tend to co-locate in the same place, we consider the three groups of actors in the measurement of clustering. We use principal component analysis (PCA) based on the following variables in 2008 from the mesolevel survey to generate a composite clustering measure: 1. Total number of fish farmers in each district normalized by rural population of the district 2. Total area of fish ponds in each district normalized by total area of the district 3. Total number of traders in each district normalized by rural population of the district 4. Total quantity of fish transactions in each district normalized by rural population of the district 5. Total number of feed dealers in each district normalized by rural population of the district 6. Total quantity of input transactions in each district normalized by rural population of the district

2 As shown in Chapter 3, feed mills are not necessarily located near the fish production areas. 62 chapter 4

Table 4.1 The spatial linkage of the fish supply chain

Same Same Same union upazila district Outside Same (other (other (other the village villages) unions) upazilas) district Percentage of hatcheries whose regular 2.96 10.78 15.36 9.70 61.19 buyers come from Percentage of farmers whose regular 27.96 25.84 35.28 7.49 3.43 buyers come from Percentage of traders whose regular suppliers or clients come from Suppliers 42.97 10.04 30.92 5.02 11.04 Clients 46.06 6.43 24.07 4.15 19.28 Percentage of feed dealers whose regular suppliers or clients come from Suppliers 14.08 5.34 11.17 25.73 42.69 Clients 40.29 16.02 29.61 7.28 6.80

Source: Authors’ compilation based on household/trader/feed dealer component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013.

Table 4.2 presents the PCA results. As shown in Table 4.2, only one component has an eigenvalue greater than 1, suggesting that a composite indicator based on the principal component can largely represent the above six variables. Essentially, the indicator is a weighted average of the six variables, largely reflecting the concentration and relatedness of the three actors in the fish supply chain. We use it as a measurement for clustering. It should be noted that this principle component measures the degree of clustering only in 2008. To examine the changing patterns of clustering over time, the following procedure is used. First, we normalize each variable using the mean and standard errors in 2008. Second, applying the same weights used in the composite indicator in 2008 to the normalized variables in 2003 and 2013, we obtain clustering measures in 2003 and 2013. Table 4.3 reports the clustering measures in 2003, 2008, and 2013 at the district level. It is apparent from Table 4.3 that all the districts exhibited an increase in clustering from 2003 to 2013. Figure 4.3 plots the clustering measure in 2013 at the district level on a map. The deeper the color of a district, the greater degree of clustering. As shown in Figure 4.3, the degree of clustering varies greatly across the 20 districts well known for fish farming. Some places (such as Mymensingh, Cluster-based Aquaculture Growth 63

Table 4.2 Principal component analysis

Component Eigenvalue Proportion 1st Component 4.24 0.71 2nd Component 0.94 0.16 3rd Component 0.46 0.08 4th Component 0.19 0.03 5th Component 0.14 0.02 6th Component 0.02 0.00 Variable Weight in the 1st component Number of farmers in the district in 2008 (normalized by rural population of the district) 0.43 Pond area of the district in 2008 (normalized by total area of the district) 0.42 Number of traders in the district in 2008 (normalized by rural population of the district) 0.31 Quantity of fish transactions in the district in 2008 (normalized by rural population of the district) 0.41 Number of feed dealers in the district in 2008 (normalized by rural population of the district) 0.46 Quantity of input transactions in the district in 2008 (normalized by rural population of the district) 0.41

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013.

Table 4.3 Clustering measures at the district level, 2003, 2008, and 2013

District Zone 2003 2008 2013 Mymensingh North 0.528 0.809 1.000 Natore North 0.078 0.122 0.192 Bogra North 0.057 0.106 0.173 Gazipur North 0.080 0.115 0.124 Dinajpur North 0.014 0.023 0.038 Narsingdi North 0.026 0.031 0.038

Khulna Southwest 0.421 0.542 0.683 Bagerhat Southwest 0.410 0.375 0.448 Jessore Southwest 0.255 0.326 0.396 Satkhira Southwest 0.200 0.280 0.372 (continued) 64 chapter 4

Table 4.3 Continued

District Zone 2003 2008 2013 Gopalganj South Center 0.413 0.484 0.646 Bhola South Center 0.114 0.263 0.560 Barisal South Center 0.080 0.111 0.158 Chandpur South Center 0.000 0.005 0.011

Brahmanbaria East 0.146 0.193 0.253 Noakhali East 0.088 0.118 0.169 Sylhet East 0.032 0.073 0.105 Cox’s Bazar East 0.058 0.066 0.090 Chittagong East 0.059 0.069 0.086 Comilla East 0.032 0.050 0.078

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: The clustering measure in 2008 is generated by principle component analysis on the six variables explained in the text. The measures in 2003 and 2013 are calculated based on the parameters generated from the principle component analysis based on data in 2008.

Khulna, and Gopalganj) figure prominently in clustering, in accordance with perceptions. Next we examine the association between clustering and a series of performance indicators.

Impact of Clustering

Fish Farmers’ Output Specialization and Modern Input Use In industrial production, clustering is normally associated with more specialization in production. Very few studies examined the relationship between clustering and specialization in aquaculture. The fish supply chain survey includes detailed information on the number of species and their quantities in 2008 and 2013. Using the information, we calculate the HHI to capture output specialization. The larger the value, the more specialization in output. We classify the 20 districts into high- and low-cluster regions based on the median value of the clustering measure in 2008. Table 4.4 reports HHI and the number of species based on the fish farmer survey in high- and low-cluster regions. In the high-clustering regions, both HHI and the number of species barely changed between 2008 and 2013. By comparison, in the low-clustering Cluster-based Aquaculture Growth 65

Figure 4.3 Degree of fish clustering of sample districts, 2013

Source: Derived by the authors from the BIHS 2011–2012 dataset (Ahmed 2013). Note: Only 20 sample districts are green, other districts are white. The darker the green, the greater the degree of clustering in this district. regions, the number of species significantly increased from 4.23 to 4.43. HHI also witnessed a noticeable drop in the low-clustering areas, indicating increasing diversification in fish variety. Overall, it seems clustering is not associated with more specialization. In clusters, feed dealers are readily available. Thereby it is much more convenient for fish farmers in clusters to purchase modern inputs (fuel, fertilizer, pesticide, lime, feed, and vitamins) than those outside clusters. As shown in Table 4.4, the fraction of farmers adopting modern inputs significantly increased from 2008 to 2013 in both high- and low-clustering regions. High-clustered regions outperform their low-clustered counterparts in their adoption rates of modern inputs by about 20 percent in both 2008 and 2013. 66 chapter 4

Table 4.4 Fish farmers’ specialization in output and adoption of modern inputs, 2008 and 2013

2008 2013 Difference High-clustering region HHI at household level 0.52 0.52 0.00 Number of species 4.35 4.44 0.09 Share of HH adopted any kinds of modern inputs (%) 88.93 93.96 5.03*** Unit cost of modern inputs (thousand BDT/acre) 6.13 13.59 7.46*** Share of costs on modern inputs over total costs (%) 10.75 12.05 1.30 Number of observations 551 762 Low-clustering region HHI at household level 0.53 0.52 −0.02* Number of species 4.23 4.43 0.20** Share of HH adopted any kinds of modern inputs (%) 67.37 74.87 7.50*** Unit cost of modern inputs (thousand BDT/acre) 5.87 10.81 4.94** Share of costs on modern inputs over total costs (%) 8.42 9.66 1.24 Number of observations 521 752 Difference between higher and lower clustering regions HHI at household level −0.01 0.00 0.02 Number of species 0.12 0.01 −0.11 Share of HH adopted any kinds of modern inputs (%) 21.56*** 19.09*** −2.47 Unit cost of modern inputs (thousand BDT/acre) 0.26 2.78 2.52 Share of costs on modern inputs over total costs (%) 2.33** 2.39*** 0.06 Number of observations 1,072 1,514

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: HHI stands for Herfindahl-Hirschman Index. High- and low-clustering regions are defined based on the median value of the clustering measure in 2008. *** p < 0.01; ** p < 0.05; * p < 0.1. BDT = Bangladeshi taka; HH = household.

Similarly, the unit cost of modern inputs (thousand Bangladeshi taka [BDT] per acre) also shows an apparent uptick between 2008 and 2013. However, the difference in the unit cost of modern inputs between high- and low-clustering regions is not significant in either 2008 or 2013. By contrast, the share of modern input cost in total cost in highly clustered regions is much higher than in less clustered areas. There is a possibility that the difference between years or between low- and high-clustered regions is driven by regional or year- specific factors. We use the approach of difference in differences (DID) to control for these factors. The results are shown in the last row in the bottom panel. Neither coefficient is statistically significant. Cluster-based Aquaculture Growth 67

The patterns revealed in Table 4.4 are just suggestive since the analysis does not control for any household characteristics. In Table 4.5 we run multivariate regressions to quantify the relationship between initial clustering in 2008 and output specialization at the farm level (HHI and number of species) in 2013. In the first parsimonious regression we include only the initial value of HHI in 2008, initial degree of clustering at the district level, and travel hours to the nearest big city by car at the household level. The coefficient for the clustering measure is not significantly different from zero. The coefficient for travel hours to the nearest big city by car is statistically significant and negative. This means that farmers in areas close to cities are more likely to produce fewer varieties than those in more remote places. In the second column we add more control variables at the household level. Only two household variables turn out to be significant. More recently established married household heads and ponds are positively associated with output diversification. After including these control variables, the variable for the clustering measure becomes significant at the 10 percent level. Columns 3 and 4 repeat the same regressions of the first two columns, with an exception that we replace the outcome variable with the number of species. Whether household characteristics are controlled for or not, the coefficient for the clustering variable is insignificant. Overall, the relationship between clustering and output specialization, if any, appears to be rather weak. The specifications in Table 4.6 echo those in Table 4.5 except that we use three variables related to modern input use (adoption of modern inputs, unit cost of modern inputs, and share of modern input cost in total input cost) as dependent variables. For each dependent variable, like Table 4.5, we report two specifications with the first one being parsimonious. Since some farmers do not use modern inputs at all, we use a Tobit regression to take the censoring of zero values into account when the outcome variable is the unit cost of modern inputs. As shown in the second row, the coefficient for the clustering measure in 2008 is statistically positive across all of the six regressions. Interestingly, travel hours to the nearest city is not significant in any of the six regressions. Clustering matters more to fish farmers’ modern input use than to distance to major cities.

Traders’ and Feed Dealers’ Cooperative Behavior Given that clusters are composed of numerous competitors working on the same trade, competition presumably must be intense, making cooperative behavior rare. However, cooperative behavior among competitors is widely observed. Table 4.7 displays the percentage of fish traders and feed dealers who 68 chapter 4

Table 4.5 OLS estimates on the relationship between clustering and output specialization

Number of Number of species species Variables HHI 2013 HHI 2013 (2013) (2013) HHI 2008 0.761*** 0.762*** 0.873*** 0.869*** (0.030) (0.030) (0.026) (0.027)

Clustering degree in 2008 0.030 0.038* 0.161 0.051 (0.022) (0.021) (0.216) (0.221)

Hours to nearest big city by car 0.005*** 0.005*** 0.004 0.001 (0.001) (0.001) (0.019) (0.018)

Household size −0.002 0.005 (0.001) (0.008)

Male household head (dummy) 0.008 0.128** (0.011) (0.060)

Household head being married −0.028* 0.090** (0.014) (0.034)

Household head education level 0.002 0.003 (0.002) (0.017)

Pond size (acre) 0.000 0.000 (0.000) (0.002)

Pond ownership (owned = 1) −0.006 0.106** (0.005) (0.042)

Pond years 0.001*** −0.002 (0.000) (0.002)

Adjusted R2 0.722 0.724 0.789 0.790 AIC −1,962.764 −1,963.469 2,280.482 2,282.220 Observations 1,072 1,072 1,072 1,072

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: Standard errors in parentheses are clustered at the district level. *** p < 0.01, ** p < 0.05, * p < 0.1. OLS = ordinary least squares; HHI = Hirschman–Herfindahl Index; R2 = R squared; AIC = Akaike information criterion. shared tools, transportation vehicles, labor, purchasing information, or selling information in 2013. About two-thirds of fish traders shared purchase or sale information; 63 percent of fish traders lent tools to each other; one-third of them pooled labor; and more than one-third cooperated in the use of vehicles. Fish traders in more clustered districts are more likely to share trading information than in less clustered areas. But the difference in sharing tools, transportation vehicles, and labor between high-clustered and low-clustered areas is not statistically significant. The second-to-last row of Table 4.7 shows Cluster-based Aquaculture Growth 69

Table 4.6 Regressions on the relationship between clustering and adoption of modern inputs

Cost of modern Share of the cost of Adoption of inputs per modern inputs in modern inputs 1,000 acres total input cost Variables (OLS) (Tobit) (OLS) Initial value of the 0.719*** 0.710*** 1.848*** 1.846*** 0.816*** 0.815*** dependent variable (0.041) (0.041) (0.139) (0.139) (0.070) (0.070) in 2008 Clustering degree 0.131** 0.097** 15.653*** 13.322*** 0.039** 0.035* in 2008 (0.053) (0.045) (5.881) (5.054) (0.018) (0.019) Hour to nearest big 0.001 0.000 0.115 0.122 0.001 0.001 city by car (0.004) (0.004) (0.405) (0.410) (0.001) (0.001) Household size −0.005 −0.939* −0.001 (0.003) (0.479) (0.001) Male household 0.024 3.077 −0.004 head (dummy) (0.025) (3.125) (0.010) Household head −0.012 −4.550 0.001 being married (0.024) (6.524) (0.015) Household head −0.000 1.307*** 0.003** education level (0.003) (0.492) (0.001) Pond size (acre) −0.000 0.004 −0.000 (0.001) (0.087) (0.000) Pond ownership 0.038** 2.309 0.008 (owned = 1) (0.016) (1.500) (0.007) Pond years −0.002 −0.175* 0.000 (0.001) (0.093) (0.000) Adjusted R2 0.651 0.654 0.161 0.163 0.722 0.722 (Pseudo R2 for Tobit) AIC −196.722 −198.263 8,480.377 8,472.316 −2,355.177 −2,346.976 Observations 1,072 1,072 1,072 1,072 1,072 1,072

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Note: District-level cluster standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. OLS = ordinary least squares; R2 = R squared; AIC = Akaike information criterion. the percentage of fish traders who displayed sharing at least one of the above five items. The percentage is high: 84 percent of them somewhat cooperated. The degree of cooperation is particularly high in more clustered districts (88 percent), compared with 79 percent in less clustered areas. Feed dealers also display cooperative behavior, although the percentage of cooperation is generally lower than that of fish traders. Feed dealers in highly clustered regions figure more prominently in cooperation. While 82 percent of feed dealers in high-clustered areas reported having cooperation with other 70 chapter 4

Table 4.7 Horizontal cooperation among fish traders and feed dealers, 2013

Fish traders Feed dealers High Low High Low Total clustering clustering Difference Total clustering clustering Difference Percentage of traders/dealers sharing the following with their competitors Tools 63.3 65.4 60.9 4.5 20.3 33.7 8.8 24.9*** Transport 36.3 33.8 39.1 −5.3 28.3 35.7 21.9 13.8** vehicles Labor 33.7 36.5 30.4 6.1 18.4 25.5 12.3 13.2** Purchase 67.3 74.8 58.7 16.1*** 56.1 77.6 37.7 39.9*** information Sale information 66.1 72.9 58.3 14.6*** 45.8 66.3 28.1 38.2*** Percentage of sharing at least one of the above items 83.9 88.0 79.1 8.9*** 63.7 81.6 48.2 33.4*** Number of 496 266 230 212 98 114 observations

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by IFPRI in 2013. Note: We classify the 20 districts into high- and low-cluster regions based on the median value of the clustering measure in 2008. *** p < 0.01, ** p < 0.05, * p < 0.1. feed dealers, the percentage drops to 48 percent in less clustered districts. Figure 4.4 plots the prevalence of cooperative behavior among feed dealers/ traders versus clustering at the district level measured in 2013. Apparently, in districts with a higher degree of clustering, feed dealers and traders exhibit a stronger tendency to share tools, vehicles, labor, or information with others than those in less clustered districts. Table 4.7 and Figure 4.4 offer only suggestive evidence on clustering and cooperation, as some factors that might affect both the development of fish farming clusters’ and traders’ cooperative behavior are not considered. For example, distances to big cities may matter to traders’ cooperative behavior. Facing greater demand from nearby cities, traders may have a limited capacity to meet it. Thereby they may be more likely to cooperate with other traders. In addition, traders’ personal characteristics and their social connections may also matter. We use the following probit model to control for these factors.

* yt = β0 + β1clusteringdt + β1Xt + β3t(+ μd) + εt

* yt ≥ 0, cooperationt = 1

* { yt < 0, cooperationt = 0 Cluster-based Aquaculture Growth 71

Figure 4.4 Clustering versus cooperation among input dealers and traders

1 1

0.9 0.9 Coef. = 0.06 Coef. = 0.42 0.8 0.8

0.7 0.7

0.6 0.6

0.5 R2 = 0.23 0.5 p-value = 0.00 0.4 0.4 R2 = 0.01 p-value = 0.00 0.3 0.3

0.2 0.2

0.1 0.1

0 0 ooerto rte o trer trct ee

ooerto rte o ee eer trct ee 00.2 0.40.6 0.81 00.2 0.40.6 0.81 uter e trct ee uter e trct ee

where clusteringdt is the district-level clustering measure in year t. We also use a clustering measure in 2008 as a robustness check. Xt is a set of control variables, including the travel time to the nearest big city, owners’ age, gender, education, religion, business experience, family ties in the same line of business, membership of related associations, firm size (the number of employees in high season), and types of business (selling feeds or fish). Table 4.8 presents the estimations of the above model. The dependent variable is a dummy indicating cooperative behavior. The first column starts with the most parsimonious specification with only a current clustering measure included. The coefficient is 0.54, significant at the 1 percent level. In the second column we add travel time to the nearest big city (Dhaka, Chittagong, or Khulna). The coefficient for the clustering variable drops to 0.45 and remains highly significant. The negative coefficient for the travel time variable suggests that traders close to big cities show a higher tendency of cooperation. In the third column we further include household characteristics. The coefficient for the clustering variable barely changes. One limitation of the first column is that standard errors are not clustered. In columns 4 and 5 we cluster the standard errors at the district and local levels. The magnitude of the estimate for clustering measuring is in the same order as in columns 2 and 3. However, the significance level drops. It is marginally significant in column 5 72 chapter 4

Table 4.8 Probit regressions on the relationship between clustering and cooperation

Determinants of cooperation (1) (2) (3) (4) (5) (6) Clustering measure in 2013 0.54*** 0.45*** 0.47*** 0.47 0.42 (0.13) (0.14) (0.15) (0.42) (0.27) Clustering measure in 2008 0.61* (0.32) Travel time to the nearest big city −0.05*** −0.06*** −0.06 −0.05 −0.04 (Dhaka, Chittagong, or Khulna, in (0.02) (0.02) (0.05) (0.04) (0.04) hours) by car Business type (“feed dealer” = 0, 0.22** 0.22 0.17 0.18 “fish traders” = 1) (0.10) (0.18) (0.21) (0.20) Age −0.02*** −0.02*** −0.02*** −0.02*** (0.00) (0.01) (0.01) (0.01) Muslim 0.13 0.13 0.15 0.13 (0.09) (0.17) (0.17) (0.17) Education year −0.04*** −0.04** −0.05** −0.04** (0.01) (0.02) (0.02) (0.02) Membership of trader association 0.48*** 0.48** 0.53** 0.54** (0.09) (0.24) (0.22) (0.22) Business experience (year) 0.01 0.01 0.00 0.01 (0.00) (0.01) (0.01) (0.01) Family tie in the same line of 0.10 0.10 0.13 0.10 business (0.09) (0.16) (0.16) (0.15) Firm size (number of employees −0.00 −0.00 −0.00 −0.00 in high season) (0.00) (0.00) (0.00) (0.00) Whether consider intra-group No No No Yes Yes Yes clustering (on district (on location (on location level) level) level) Pseudo R2 0.01 0.02 0.09 0.09 0.08 0.09 AIC 1,716.71 1,711.06 1,511.67 1,511.67 1,364.70 1,052.04 Observations 1,724 1,724 1,657 1,657 1,454 1,101

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by IFPRI in 2013. Note: The dependent variable is a dummy indicating whether a respondent has shown cooperative behavior as defined inTable 4.7 (shared at least one of the five items). *** p < 0.01, ** p < 0.05, * p < 0.1; R2 = R squared; AIC = Akaike information criterion. when standard errors are clustered at the local level. Because both the clustering measure and the cooperation variable are measured in 2013, it is hard to tease out reverse causality. In column 6 we replace the current clustering measure with a lagged clustering measure in 2008. The clustering measure is now as high as 0.61, significant at the 10 percent level. Overall, traders’ and feed dealers’ degree of cooperation is higher in more clustered regions. Interestingly, travel time to the nearest big city is no longer significant after using clustered standard errors in columns 4 through 6. Young traders show Cluster-based Aquaculture Growth 73

a higher likelihood of cooperation than their elderly counterparts. Members of business associations demonstrate a higher percentage of cooperation than nonmembers. On the contrary, religion, years in business, and family ties in the same line of business do not matter much for cooperative behavior. Table 4.9 further looks at the association between clustering and different types of cooperation. The specification is the same as the last column in Table 4.8. We use a clustering measure in 2008 to partly address the potential reverse causality problem. The coefficient for three types of cooperation in columns 2–4 (sharing labor, sharing purchase information, and sharing sale information) is positive and statistically significant at the 5 percent level at least. Of course, the positive coefficient just represents an association, not causal relationship. It is puzzling that the coefficient for sharing tools and vehicles as shown in column 1 is negative, although insignificant. This puzzle might be explained by the emergence of professional transport services in highly clustered areas. As demonstrated in Figure 4.5, the fraction of traders/feed dealers having access to professional truck rental service is positively related to the degree of clustering at the district level. The insignificant impact of clustering on sharing vehicles/tools might be explained by the emergence of professional transport services in highly clustered areas. In a more developed cluster, there are many traders and feed dealers who need transportation services. When the market is big enough, specialized truck rental companies become viable. When truck rental services are readily available, it is probably more convenient for traders/feed dealers to hire professional truck services than to share trucks among themselves, which involves some coordination costs.

Conclusion and Policy Implications Based on primary surveys, this chapter shows that fish farming became increasingly clustered in Bangladesh. In regions with a high degree of clustering, farmers use more modern inputs. In addition, we find that cooperative behavior is widespread in clusters. The topic of cooperation in clusters is not yet well studied in the literature despite its ubiquity. Cooperation in clusters likely reduces the operational cost of running a business, further attracting more entries to the sector. The model of cluster-based development, in particular the collaborative behavior in clusters, deserves more in-depth research. Clustering offers a way for entrepreneurs to operate their businesses in the absence of sound institutions, a defining feature in many developing countries. The literature has emphasized the Marshallian advantages of clustering, including access to market, knowledge spillover, and labor pooling. Apart 74 chapter 4

Table 4.9 Probit regressions on the relationship between clustering and different types of cooperative behaviors

(1) (2) (3) (4) Sharing tools Sharing Sharing purchase Sharing sale Determinants of cooperation and vehicles labor information information Clustering measure in 2008 −0.37 0.61** 0.81*** 0.63** (0.27) (0.25) (0.29) (0.30) Travel time to the nearest big city −0.07** −0.01 −0.08** −0.07* (Dhaka, Chittagong, or Khulna, in (0.03) (0.03) (0.04) (0.04) hours) by car Business type (“feed dealer” = 0, 0.60*** 0.50*** −0.09 0.35* “fish trader” = 1) (0.18) (0.17) (0.18) (0.19) Age −0.01 −0.00 −0.01** −0.01 (0.01) (0.01) (0.01) (0.01) Muslim 0.05 0.15 0.04 0.09 (0.16) (0.16) (0.18) (0.18) Education year −0.03** 0.04* −0.04** −0.02 (0.02) (0.02) (0.02) (0.02) Membership of trader 0.44*** 0.29** 0.53*** 0.26 association (0.15) (0.13) (0.18) (0.17) Business experience (year) 0.01 0.00 −0.00 −0.00 (0.01) (0.01) (0.01) (0.01) Family tie in the same line of 0.02 0.17 −0.08 −0.11 business (0.14) (0.15) (0.13) (0.12) Firm size (number of employees −0.00 −0.00 0.00 0.00 in high season) (0.00) (0.00) (0.00) (0.00) Pseudo R2 0.11 0.05 0.07 0.06 AIC 1,357.35 1,282.04 1,371.09 1,423.71 Observations 1,101 1,101 1,101 1,101

Source: Authors’ calculations based on the farm household component of the value chain survey conducted by IFPRI in 2013. Note: *** p < 0.01, ** p < 0.05, * p < 0.1. R2 = R squared; AIC = Akaike information criterion. from these advantages, this chapter shows that clustering is associated with more prevalent cooperative behavior, which likely results in better performance. The benefit of cooperation probably offsets the disadvantage of intensive competition along with cluster development. Otherwise fish farming in Bangladesh would not have become so rapidly clustered in such a short time. The model of cluster-based fish farming in Bangladesh may shed some lessons for other developing countries. One should bear in mind several drawbacks of this study. First, we did not examine the link between clustering with performance outcome variables, such as profit rate and total factor productivity. Second, the mechanisms Cluster-based Aquaculture Growth 75

Figure 4.5 Clustering and accessibility to truck rental companies

0.9 Coef. = 0.26 0.8

0.7 trct ee

0.6 0.5

0.4 ret co R2 = 0.28 0.3 p-value = 0.00 to truc 0.2

0.1

ccet 0 00.1 0.20.3 0.40.5 0.60.7 0.80.9 1 uter e trct ee

Source: Authors’ calculations based on the farm household component of the Bangladesh fish value chain survey conducted by the International Food Policy Research Institute in 2013. Clustering measures are calculated by authors based on methodology discussed in this chapter. Note: The accessibility to truck rental companies in a district is defined as the share of fish traders and feed dealers who rented truck services. behind the correlation of clustering and cooperating were not studied. It is not clear what drives cooperation. Third, the observed relationship between clustering and cooperative behavior is likely just association not causal. We leave these topics for future research.

References Ahmed, A. 2013. Bangladesh Integrated Household Survey (BIHS) 2011–2012. Washington, DC: International Food Policy Research Institute (IFPRI). Ali, M., J. Peerlings, and X. Zhang. 2014. “Clustering as an Organizational Response to Capital Market Inefficiency: Evidence from Microenterprises in Ethiopia.” Small Business Economy 43 (3): 697–709. Felzensztein, C., and E. Gimmon. 2009. “Social Networks and Marketing Cooperation in Entrepreneurial Clusters: An International Comparative Study.” Journal of International Entrepreneurship 7 (4): 281–291. Hidalgo, C. A., B. Klinger, A. L. Barabási, and R. Hausmann. 2007. “The Product Space Conditions the Development of Nations.” Science 317 (5837): 482–487. 76 chapter 4

Humphrey, J., and H. Schmitz. 1998. “Trust and Inter-Firm Relations in Developing and Transition Economies.” Journal of Development Studies 34 (4): 32–61. Long, C., and X. Zhang. 2012. “Patterns of China’s Industrialization: Concentration, Specialization, and Clustering.” China Economic Review 23 (3): 593–612. Marshall, A. 1920. Principles of Economics. London: Macmillan and Co., Ltd. Porter, M. E. 2000. “Location, Competition, and Economic Development: Local Clusters in a Global Economy.” Economic Development Quarterly 14 (1): 15–34. Ruan, J., and X. Zhang. 2009. “Finance and Cluster-Based Industrial Development in China.” Economic Development and Cultural Change 58 (1): 143–164. Schmitz, H. 1995. “Collective Efficiency: Growth Path for Small-Scale Industry.” Journal of Development Studies 31 (4): 529–566. Schmitz, H., and K. Nadvi. 1999. “Clustering and Industrialization: Introduction.” World Development 27 (9): 1503–1514. Chapter 5

WELFARE AND POVERTY IMPACTS OF AQUACULTURE GROWTH

Shahidur Rashid, Nicholas Minot, and Solomon Lemma

Introduction Aquaculture is one of the world’s fastest growing food-producing sectors, and its share in global fish consumption by humans is projected to grow to more than 60 percent by 2030 (FAO 2014). This growth is remarkable given that the sector was almost nonexistent in the 1950s and its share in total fish production remained below 20 percent until the early 1990s. The underlying implications of this trend are considered to be so significant that they are now commonly termed a “Blue Revolution,” and there are good reasons for using the term. Aquaculture holds the promise of meeting most of the world’s fish demand without ruining the environment (Economist 2003; Sachs 2007); aquaculture also will be able to help reduce poverty while improving food security and nutritional well-being.1 If aquaculture had stopped growing in 1980—that is, if growth in the world’s fish supply depended only on marine and inland capture fisheries—per capita annual fish availability in 2013 would have been only 14.0 kilograms, which is 17 percent lower than the availability in 1980 and about half of the actual availability of 26.8 kilograms in 2013. The consequences of such a scenario are easy to imagine: higher prices, lower consumption, and far greater pressure on marine and inland capture fisheries. The adverse consequences would have been particularly severe for the developing countries of Asia, where fish is an important part of the diet and where fish production and marketing provide the livelihoods for millions of poor households. Thanks to technological innovation and some deliberate policy actions, the world did not have to live through such a reality. As discussed in the Introduction, calculations from the Food and Agriculture Organization’s

1 There is a body of literature on the relationship between aquaculture and food security, poverty, and environment. For further reference, see Ahmed and Lorica (2002) on food security; Toufique and Belton (2014) and Belton, Haque, and Little (2012) on poverty impacts; and Naylor et al. (2000) and Pingali (2001) on environmental impacts.

77 78 chapter 5

(FAO) Aquastat database indicate that Asian countries increased their aquaculture production from 10.8 million metric tons in 1990 to 58.9 million metric tons in 2012, with developing countries accounting for about 86 percent of the total production. Although China has led the way in this growth, some other countries in Asia have also experienced impressive growth in aquaculture since the 1990s. For example, farmed fish production in Bangladesh has increased significantly, with growth in fish availability far exceeding growth in cereal availability during the Green Revolution; there has also been a recent emphasis specifically on aquaculture as opposed to marine or inland capture.2 This growth and structural change have contributed to increasing fish availability, reducing the real price of fish, and generating employment, with important implications for poverty reduction and nutritional well-being.3 Thus there has been increasing interest in the poverty and income distribution implications of aquaculture growth in Bangladesh.4 However, empirical studies have only recently begun to emerge. Even though the literature provides important insights, there remain significant gaps in terms of both conceptual understanding and empirical methodologies.5 Conceptually, the aquaculture-poverty linkage literature closely resembles the agricultural- growth linkage literature in that almost all available studies hypothesize aquaculture to have both direct and indirect benefits on poverty. However, there are ambiguities regarding the definition and quantification of these direct and indirect benefits. Existing literature has important methodologi- cal weaknesses, as almost all available studies are based on nonrepresentative samples. As a result, although these studies provide important insights into selected issues, no generalizable conclusions can be drawn from them. Perhaps the most important gap is that none of the available studies answer the basic question: To what extent has aquaculture growth contributed to poverty reduction? This chapter attempts to fill these gaps in the literature. In particular, it uses microsimulation methods to analyze the impacts of aquaculture growth

2 According to official data, production of aquaculture fish grew by about 15 percent in the 1990s, as compared with a growth in rice production of about 3 percent during the Green Revolution. 3 The nutritional implications are particularly important for Bangladesh because fish accounts for 63 percent of animal-sourced protein (FAO 2005) in Bengali diets; it is the most important source of high-quality protein, essential fatty acids, and micronutrients (Roos et al. 2007); and it is the most frequently consumed, nutrient-rich, animal-sourced food among all income groups in Bangladesh (Toufique and Belton 2014). 4 Another reason for increased interests on the aquaculture-poverty link is that donors have taken an active interest in promoting aquaculture. 5 Ahmed and Lorica (2002) presented a conceptual framework, and Toufique and Belton (2014) presented a good summary of available studies of the aquaculture-poverty links in Bangladesh. Welfare and Poverty Impacts of Aquaculture Growth 79

on income distribution and poverty in Bangladesh using an expanded version of Deaton’s (1989) model and several rounds of nationally representative household survey data. The chapter is organized as follows: The next section presents key trends in aquaculture production, exports, consumption, and prices in Bangladesh. An overview of the concepts and previous research is followed by a description of the data and methods used in the chapter. The results of the welfare impact of real price decline and productivity growth are then presented. The chapter concludes with a summary and implications.

Trends in Aquaculture in Bangladesh

Production Bangladesh has made remarkable progress in promoting aquaculture. From 1983/1984 to 1992/1993 the average annual production of aquaculture fish from inland fisheries was only about 178,000 metric tons; this number jumped to about 1.3 million metric tons in the most recent decade (Table 5.1). Production of capture and marine fisheries also increased in absolute terms since 1983/1984, but their average shares in total production declined from 53 to 36 percent and from 26 to 19 percent, respectively. In 2015/2016, the latest year for which data are available, total aquaculture production was 2.1 million metric tons, equivalent to 56 percent of the total fish production of 3.7 million metric tons (Shamsuzzaman et al. 2017). Of the four subcategories of culture fisheries, pond culture has experienced the fastest growth. Average annual pond fish production increased from about 150,000 metric tons in the 1980s to more than 1.0 million metric tons since 2008, equivalent to about 80 percent of the total culture production of 1.3 million metric tons. The growth has been faster in recent years, with production from pond culture jumping from 750,000 metric tons in 2007/2008 to 1.45 million metric tons in 2012/2013. Shrimp production, which receives much more attention in policy discussion, increased only by about 65,000 metric tons during the same period. Another new development in aquaculture in Bangladesh is the seasonal production of cultured fish in haors and baors (the depressions that get flooded during the monsoon season). The Department of Fisheries (DoF) began collecting data on seasonal culture fish production in 2009/2010, when it recorded a total production of about 46,000 metric tons; this number increased to 201,000 metric tons by 2012/2013. Thus this appears to be another growth area for Bangladesh aquaculture. 80 chapter 5

Table 5.1 Trends in fish production in Bangladesh, 1983/1984–2012/2013

Annual average production (metric tons) and production shares 1983/1984– 1993/1994– 2003/2004– Sources 1992/1993 Shares (%) 2002/2003 Shares (%) 2012/2013 Shares (%) Inland fisheries Capture River 175,285 20.6 151,561 10.0 140,552 5.1 Sunder ban 6,852 0.8 9,865 0.6 18,220 0.7 Beel (lake) 47,397 5.6 67,403 4.4 79,136 2.9 Kaptai lake 3,714 0.4 6,490 0.4 8,096 0.3 Flood lands 220,170 25.8 404,291 26.5 728,193 26.4 Total 453,418 53.2 639,610 42.0 974,197 35.3 Culture Ponds 153,258 18.0 479,867 31.5 1,010,184 36.6 Semiclose 1,293 0.2 3,277 0.2 5,261 0.2 Baors/culture — — — — 107,782 3.9 Shrimp 23,378 2.7 79,611 5.2 151,610 5.5 Total 177,929 20.9 562,755 37.0 1,274,837 46.1 Marine fisheries Industrial 11,393 1.3 17,411 1.1 42,805 1.5 Artisanal 210,092 24.6 303,123 19.9 471,245 17.1 Total 221,485 26.0 320,534 21.0 514,050 18.6 National total 852,832 100.0 1,522,899 100.0 2,763,084 100.0

Source: Based on data compiled by the authors from Fisheries Statistical Year Book of Bangladesh (1983–2013), published by the DoF’s Ministry of Fisheries and Livestock. Note: — = data not available.

Exports Bangladesh exports seven categories of fish products—frozen shrimp, frozen fish, dry fish, salted fish, shark fin, turtle, and others; shrimp is by far the largest export in terms of value and volume. Of the total export of 85,000 metric tons in 2012/2013, more than 50,000 metric tons (or about 60 percent) was shrimp produced by large commercial farmers with the primary objective of exporting. Shrimp is mainly exported to the United States of America (26%), Belgium (21%), the United Kingdom (13%), Germany (8%), and the Netherlands (5%) (Kabir 2015, percentages based on data from fiscal year 2010/2011). The share of shrimp in the total export value was even larger, at 81 percent of the total export of $533 million. By contrast, the total volume of frozen fish was only about 11,000 metric tons, which included not Welfare and Poverty Impacts of Aquaculture Growth 81

only farm-raised fish but also marine and other capture fisheries. Therefore the actual volume of aquaculture fish export was much smaller than 11,000 metric tons, which is less than 1 percent of the 1.3 million metric tons of cultured fish. These numbers imply that the growth in the production of common aquaculture fish varieties is overwhelmingly driven by domestic demand. This is a key point because multiplier effects are much larger, both in terms of nutrition and income generation, for aquaculture than for other fish subsectors not considered in this study.

Consumption The small volume of nonshrimp aquaculture exports indicates that the growth in production has translated into increased consumption at the aggregate level. However, it tells little about the patterns of consumption by location, gender, and income groups, which are central to drawing welfare implications. To this end, this chapter analyzes three rounds of Household Income and Expenditure Survey (HIES) data. The only problem in carrying out this analysis is that HIES does not clearly disaggregate nonshrimp aquaculture from other types of fish consumption, especially from inland capture fisheries. The authors follow Toufique and Belton’s (2014) strategy of cat- egorizing fish into four major groups: (1) primarily aquaculture, (2) inland capture and culture, (3) primarily inland capture, and (4) primarily marine. Table 5.2 presents the results of the analysis. The aggregate estimates of rural and urban consumption are fairly similar to those in Toufique and Belton (2014), although two additional dimensions are added to this analysis: disag- gregation by gender and by income quintile. Three main points can be drawn from the results. First, fish consumption has increased for every category of household shown—rural, urban, male, female—as well as for each income quintile. At the national level, per capita annual fish consumption increased from 13.4 kilograms in 2000 to 18.1 kilograms in 2010, implying 35 percent growth over the decade or 3 percent per year. This achievement is remarkable because, not too long ago, achieving 18.0 kilograms per capita annual consumption was a national target and seen as a potential for aquaculture growth. This goal is reflected in an FAO (2005) report: “The present per capita annual fish consumption in Bangladesh stands at about 14 kg/year against a recommended minimum requirement of 18 kg/ year; hence there is still need to improve fish consumption in the country.” Now that the country has achieved the target, perhaps a different set of questions needs to be asked. In particular, policies can now focus on improving distribution across income groups and perhaps accessing international markets. 82 chapter 5

Table 5.2 Changes in annual per capita fish consumption (kilograms per person per year) in Bangladesh, 2000–2010

Sex of the household Location head* Income quintile Year Fish sources Rural Urban Male Female 1st 2nd 3rd 4th Richest All 2000 Primarily aquaculture 3.4 3.9 3.3 3.5 1.6 2.3 3.0 3.9 6 3.5 Inland capture and 4.5 3.8 5.1 4.3 3.2 4.0 4.5 4.8 5.1 4.4 culture Primarily inland 3.6 3.5 3.6 3.6 2.8 3.6 3.5 4.0 3.9 3.6 capture Primarily marine 2.0 3.6 3.0 2.3 0.8 1.3 2.3 3.0 3.6 2.3 Total 13.6 14.9 15.0 13.7 8.4 11.2 13.3 15.7 18.5 13.4 2005 Primarily aquaculture 5.5 6.2 6.5 5.6 3.2 4.0 4.8 6.3 8.7 5.7 Inland capture and 5.1 4.6 5.7 4.9 4.6 4.5 4.7 5.1 5.5 4.9 culture Primarily inland 2.7 3.4 2.8 2.9 2.3 2.4 2.7 3.1 3.4 2.9 capture Primarily marine 1.5 4.1 2.5 2.1 1.3 1.6 2.0 2.2 3.1 2.1 Total 14.7 18.3 17.5 15.5 11.4 12.4 14.2 16.8 20.8 15.6 2010 Primarily aquaculture 7.5 7.8 8.2 7.5 5.3 6.2 6.7 8.5 10.0 7.5 Inland capture and 5.3 6.4 6.4 5.5 4.7 4.9 5.3 5.9 6.7 5.6 culture Primarily inland 2.4 3.2 2.9 2.6 2.0 2.2 2.5 2.8 3.3 2.6 capture Primarily marine 1.6 4.5 2.8 2.3 1.3 1.4 1.9 2.7 4.0 2.4 Total 16.7 21.9 20.2 17.8 13.2 14.6 16.4 19.8 23.9 18.1 Change since 2000 (%) 23.8 46.9 34.7 29.9 57.1 30.4 23.3 26.1 29.2 35.1

Source: Authors’ calculations from BBS HIES rounds 2000, 2005, 2010. Note: * The differences between mean consumptions are statistically significant.

Second, in relative terms, households in the bottom two income quintiles experienced the fastest growth in fish consumption between 2000 and 2010. The poorest households increased their consumption by more than 57 percent—larger than any other income groups in Table 5.2. This finding is consistent with studies suggesting that both price and income elasticity of demand for fish are higher (in absolute value) for the poor than for richer households (Dey, Alam, and Bose 2010). Finally, the results show that levels of fish consumption are diverging between urban and rural households. In 2000 per capita consumption in urban households was 14.9 kilograms—only about 10 percent higher than consumption by rural households. In 2010 per Welfare and Poverty Impacts of Aquaculture Growth 83

capita annual consumption by urban households jumped to almost 22.0 kilograms—about 31 percent higher than for rural households.

Price Trends Since the main way that the boom in Bangladeshi aquaculture helps nonproducers is through the price effect, it is important to explore trends in fish prices over this period. Estimates from the HIES data suggest that fish prices declined between 2000 and 2005 but increased again in 2010. The HIES data refer to unit values of farm-level sales, not prices per se, and may be influenced by seasonality and changes in quality. More reliable data are collected by the DoF on a regular basis in larger wholesale markets. Figure 5.1 presents trends in market prices of a few popular varieties, along with three point estimates from the HIES data. The DoF market prices are higher than the HIES producer prices because of the transaction costs of collecting and transporting fish to wholesale markets. Except for ilish (primarily a marine fish), all prices have been declining in real terms since 2002, which is consistent with growth of aquaculture.6 For instance, commercial production of non-native pangas began in Mymensingh in 1993 (Belton, Haque, and Little 2012), and the DoF began collecting high-frequency price data for pangas and other cultured fish only in the early 2000s. The prices of both pangas and smaller carps (two main varieties of culture fish) have been declining ever since the DoF began recording prices in the early 2000s. Finally, real prices of all fish varieties, both inland and marine, increased rapidly in the 1980s, which was when the Green Revolution took root, rice prices fell, and the economy enjoyed overall growth. This last point was the subject of much debate in the 1990s. One set of studies argued that income growth was not sufficient to improve nutrition (see, for example, Behrman and Deolalikar 1987; Bouis and Haddad 1992); another strand refuted it (see, for instance, Subramanian and Deaton 1996). A recurrent observation at the time was that although the Green Revolution resulted in declining rice prices, prices of other food items—such as pulses, vegetables, fish, and animal products—were increasing. For instance, Bouis (2000) reported that although rice prices in selected Asian countries declined by 40 percent, the real prices of pulses, vegetables, and animal products had increased 25–50 percent since the onset of the Green Revolution. Figure 5.1

6 Ilish is from the shad family of marine fish. In monsoon season they swim toward upstream rivers, like the Padma. The fish are caught at both the Bay of Bengal and the large rivers. This popular fish in Bengal has a rapidly growing market in the countries of Europe, the Middle East, and North America, where there are Bangladeshi immigrants or migrants. 84 chapter 5

Figure 5.1 Real prices (2010 = 100) of selected fish varieties in Bangladesh, 1986–2014

HIES prices (aquaculture average per kilogram) Big katla (above 5 kilograms)

350 Ilish 450 Big rui (above 5 kilograms)

Small pangas 400 300 Small katla (1.5–2.0 kilograms) 350 250 300 t er or 200 250 200 150 150 100 t

100 S rce t er or

ru 50 o 50

rce 0 0

86 88 90 92 94 96 98 10 12 14 19 19 19 19 19 19 19 2000 2002 2004 2006 2008 20 20 20

Source: Authors’ construction based on the Bangladesh DoF price series (1994, 1997, 2006, and 2015) and BBS HIES data (2010). demonstrates that this is no longer a concern in Bangladesh, at least in the case of fish; real prices have been declining for all fish varieties, except marine fish, which is largely consumed by the rich and middle-income groups.7 The empirical evidence presented suggests that demand has increased with population growth and higher incomes, but supply has increased even more rapidly, resulting in a decline in real price. Because Bangladesh has experienced sustained economic growth, and because aquaculture is a small part of the country’s overall economy, it is safe to assume that the shift in demand curve has resulted mainly from overall growth, not aquaculture per se. Given that aquaculture’s share in gross domestic product is small, its contribution to the shift in demand should also be relatively small. This suggests that the shift in the supply curve has been the main channel through which aquaculture has contributed to welfare. This is the point of departure of our analysis.

7 An emerging body of literature suggests that the prices of small indigenous species (SIS), which used to be consumed by the poor, are increasing. Studies suggest that because of the way SIS are eaten, their contribution to Recommended Nutritional Intake (RNI) is higher. For a detailed analysis on Bangladesh, see Bogard et al. (2015). Welfare and Poverty Impacts of Aquaculture Growth 85

Concepts and Previous Research The conceptual framework for analyzing the welfare implications of aquaculture growth is no different from that for any other economic sector. Following Hirschman’s (1957) work in Latin America, many studies have examined linkages between economic sectors. The agricultural growth linkage literature made a distinction between direct and indirect effects in terms of improving rural employment and household well-being (Johnston and Mellor 1961; Mellor and Lele 1972; Adelman and Morris 1973). This literature was a strong force in changing the then-prevailing views—mainly emerging from Latin America—that public investment should be directed toward the industrial sector; this view was based on the presumption that public investment had greater linkages to the overall economy. The underlying ideas of this strand of literature continue to be relevant, as reflected in the recommendations of the World Bank’s (2008) World Development Report and de Janvry and Sadoulet’s (2010) studies of China. The central premise of the agricultural growth linkage literature is that the growth in agriculture resulting from the alleviation of supply constraints— such as technological innovation or infrastructural improvement—generates higher multiplier effects (or growth links) through increased demand of nontradable goods outside of agriculture.8 The existing aquaculture poverty linkage literature diverges significantly from this conceptual framework, as is evident in several studies. This chapter builds on two studies that discuss conceptual issues of aquaculture development and their implications for empirical methods and interpretation of results. Ahmed and Lorica (2002) argued that adoption of new aquaculture technology could affect food security through three pathways: income links, employment links, and consumption links. In the income link, the incremental income from aquaculture production improves the ability of households to diversify their diet and consume higher-quality foods. In the employment link, better diets (including more fish) raise the labor productivity of households involved in aquaculture. In the

8 To illustrate the point, suppose that (1) a rural economy is divided into two sectors, agriculture and nonagriculture; (2) marginal propensity to consume (MPC) is low, at 0.5, for agricultural commodities; and (3) MPC is high, at 0.9, for nonagricultural commodities. The second and third assumptions are particularly realistic for poor economies in that a large share of incremental income is likely to be spent on nonagricultural goods and services (for example, housing, clothing, education, and other nontradables) than on agricultural goods (for example, rice and wheat). Given these assumptions, 1 multiplier effects for nonagriculture, given by Mna = (1–0.9) = 10, will be five times larger than the 1 multiplier effects of agriculture, given by Ma = (1–0.5) = 2. This was the key insight of Johnston and Mellor’s (1961) seminal study, as well as of many subsequent studies, on the role of agriculture in economic development. 86 chapter 5

consumption link, aquaculture technology adoption has both direct effects (through consumption from own production) and indirect effects (through lower fish prices, leading to increased consumption among nonproducers). Although Ahmed and Lorica (2002) provide a comprehensive discussion of the links between aquaculture and food security, they do not propose a strategy to measure these links. If income rises among aquaculture producers, how much can be attributed to aquaculture per se (income channel) and how much can be attributed to improved labor productivity (employment channel)? Similarly, if fish consumption rises among producers, how much is attributed to income growth (income channel) and how much to the lower (implicit) price of fish (consumption link)? A recent study by Toufique and Belton (2014) extend the Ahmed-Lorica framework in two ways: (1) they provide more nuance to direct and indirect links, and (2) they adapt a framework for assessing the degree to which aquaculture growth is pro-poor. The study defines the following four links: direct consumption links (increased consumption from own production), indirect consumption links (increased availability and accessibility of fish), direct income links (increased income for aquaculture producers), and indirect income links (employment in the fish value chain and consumption links). The other contribution of Toufique and Belton (2014) is that they present a conceptual framework for assessing whether aquaculture growth is pro-poor. Drawing from the pro-poor growth literature proposed in Ravallion (2004, 2009) and Kakwani, Khandker, and Son (2004), they classify four types of pro-poorness: benchmarked, weakly pro-poor, strong relative pro-poor, and strong absolute pro-poor. Growth in aquaculture is defined as benchmarked pro-poor if it is associated with an increase in fish consumption by households below the poverty line; weakly pro-poor if fish consumption by the poor increases at a faster rate than in the past; strong relative if fish consumption by the poor increases at a faster rate than by the nonpoor; and strong absolute if the amount of fish consumed by the poor is higher than by the nonpoor.9 Although this characterization helps illustrate the pro-poorness of aquaculture, it has some limitations. The original pro-poor economic growth classification applied to the entire economy, not to any particular sector within an economy. More specifically, pro-poor growth literature compares two mea- surable outcomes—growth and poverty—that are identifiable. This is not the case for a given sector, as both growth and overall poverty reduction can be influenced by the economy’s other sectors. The key variable in Toufique

9 Poor/nonpoor distinction is based on the national poverty line generated by the BBS. Welfare and Poverty Impacts of Aquaculture Growth 87

and Belton’s framework—that is, fish consumption—can also be affected by economic factors outside of aquaculture. For example, Toufique and Belton (2014) state that “aquaculture has proven unequivocally pro-poor in terms of the ‘indirect consumption’ pathway.” An implicit assumption of this conclusion is that the increase in fish consumption resulted from the growth in aquaculture alone. However, an increase in fish consumption can be caused by a host of other factors, including growth in other sectors and subsectors. In this chapter’s assessment, all of the available evidence confirms that there has been an increase in fish consumption by all income groups and that the real prices have declined due to technological innovation and a rightward shift in the supply curve. But neither of these can be attributed to the growth in aquaculture alone, as there are other contributing factors, including years of overall economic growth. In this chapter we propose a strategy for estimating the contribution of aquaculture development to overall income growth and poverty reduction. As described in the next section, we extend methods originally proposed by Deaton (1989) to estimate the distributional effect of price and income changes attributable to aquaculture development.

Data and Methods

Data This study uses three rounds of Household Income and Expenditure Survey (HIES), conducted by the Bangladesh Bureau of Statistics (BBS) in collabo- ration with the World Bank. The sampling methods and the key results from various rounds of the HIES are presented in HIES reports produced after the completion of each round.10 Although this chapter uses three rounds of surveys (2000, 2005, 2010) to explore trends in aquaculture, the microsimulation analysis is based on the latest round (2010). The sampling from that round was based on the Integrated Multipurpose Sample, which consisted of 1,000 primary sampling units. This unit is defined as two or more contiguous enumeration areas used in the 2001 Bangladesh census. Following a two-stage stratified random sampling method, a total of 12,240 households were sampled, of which 7,840 were from rural areas and 4,400 were from urban areas. Sampling weights from the HIES are used in all calculations to compensate for oversampling and undersampling of households in different locations.

10 The report on the 2010 round of the HIES is available from the BBS. 88 chapter 5

In addition, secondary data from other sources were used. Production data, disaggregated by location and sources, were compiled from the Department of Fisheries (DoF) Fisheries Yearbook published by the Ministry of Fisheries and Livestock. This yearbook is based on the Fisheries Resources Survey System administered by the DoF. Historical data on fish prices, disaggregated by location and varieties, and on export-import data were also obtained from the DoF. These data were particularly helpful in analyzing the trends and triangu- lating the estimates generated from the HIES data.

Methodology This study employs an expanded version of Deaton’s (1989) model for analyzing the welfare implications of food price changes. Deaton’s model has been widely used to analyze the welfare effects of price shocks, most notably following the 2007/2008 global food crisis (Minot and Goletti 1998; Ivanic and Martin 2008). Deaton (1989) represents the proportional impact on real income of a commodity price change as:

dY PQ PC dP Y = ( Y − Y ) P where Y is household income (or expenditure), P is the price of the commodity, Q is the quantity produced, and C is the quantity consumed. Thus, PQ/Y is the value of production of the commodity as a proportion of income, PC/Y is the value of consumption of the commodity as a proportion of income, and dP/P is the proportional change in the price of the commodity. Deaton calls the expression in parentheses the net benefit ratio (NBR). It can be considered the short-term elasticity of welfare with respect to the commodity price. We extend this welfare measure in three ways. First, we add a term to reflect changes in income, since we plan to simulate the impact of technology improvements that shift the supply of aquaculture products. Second, we add second-order terms to reflect the welfare-enhancing responses of producers and consumers to price changes. And third, we drop the assumption that producer prices and consumer prices change by the same proportion. The result, shown below, is the expression we used to estimate the proportional impact on the real income of a household resulting from technological change and the resulting price changes:

P Q dQ P (Q+dQ) dP P Q dP dY = p + p p + 1 p ( p)2 − PcC dPc − 1 PcC ( dPc )2 Y Y ϑ Q Y Pp 2 εs Y Pp Y Pc 2 εD Y Pc where Pp is the producer price of aquaculture products, Q is the household production of aquaculture products, ϑ is the ratio of producer surplus (profit) to gross revenue in aquaculture, εS is the elasticity of supply, Pc is the retail Welfare and Poverty Impacts of Aquaculture Growth 89

price of aquaculture products, C is the quantity of aquaculture products consumed by the household, and εD is the price elasticity of demand for aquaculture products. The first three terms describe the effect of technological change and changes in the producer prices on aquaculture producers. More specifically, the first term on the right side describes the producer surplus from the technological shift in aquaculture supply as a share of household income. This is calculated as the product of (a) the share of gross aquaculture income in total income, (b) the ratio of net to gross income in aquaculture, and (c) the proportional shift in supply. The second term describes the first-order proportional effect on producer income of the change in aquaculture producer prices (before any price response). The third term is the second-order proportional effect on producer income of the change in producer prices. The last two terms describe the effect of changes in aquaculture retail prices on consumers of aquaculture production. The fourth term is the first-order proportional effect on real income of changes in retail prices of aquaculture products (before any price response). The fifth term is the second- order proportional effect on real income of changes in retail prices of aquaculture products. Since we are simulating an increase in aquaculture supply resulting in a reduction of aquaculture prices, dQ/Q > 0, dPP/PP < 0, and dPc/Pc < 0. This implies that the first term is positive, reflecting the benefits from technological improvement. The second term is negative, measuring the loss in aquaculture income due to lower prices. The third term is positive, as producer response to the price reduction partially offsets the impact of the decline in prices. The fourth term (including the negative sign) is positive, reflecting the gains to consumers from lower prices for aquaculture products. And the last term is also positive, as the consumer response to lower prices further enhances welfare. To estimate the welfare impact for each household in the Bangladesh HIES, we need to combine data from the household survey with information on price and production trends, and some assumptions about the market for aquaculture products. PpQ/Y represents the value of aquaculture production as a share of income, also called the production share, while PcC/Y is the value of aquaculture consumption as a share of income, also called the consumption share. These are calculated for each household in the 2000 Bangladesh HIES. The value of household consumption expenditure, Y, a proxy for household income, is also calculated from the HIES. The proportional change in producer prices, dPp/Pp, is based on the weighted average of the inflation-adjusted 90 chapter 5

producer price of three fish species widely raised in aquaculture, big rui (more than 5 kilograms), large katla (more than 5 kilograms), and ilish in Kuhlna (the division with the largest aquaculture production). Between 2000 and 2010 the weighted average of the real producer prices fell 36 percent. The proportional change in consumer prices, dPc/Pc, is calculated as the weighted average of the inflation-adjusted retail prices of the same three species in Dhaka (the main consumption center). This weighted price decreased 45 percent over the same period. We adopt a price elasticity of demand for aquaculture products from Dey, Alam, and Paraguas (2011), a comprehensive study of fish demand in Bangladesh. They provide compensated and uncompensated elasticity estimates for eight types of fish and seafood products as well as information on consumer spending on each. Based on information provided by Toufique and Belton (2014) on the main source of each type of fish in Bangladesh (marine capture, inland capture, and aquaculture), we used the elasticity estimates of five of the eight types: Indian carp, exotic carp, live fish, tilapia, and high-value fish. We calculated the value-weighted average of the compensated price elasticities of demand for these five types of fish. The compensated (or Hicksian) elasticities are appropriate for calculating welfare impact, although the difference between compensated and uncompensated demand elasticities is modest in this case. The weighted-average Hicksian demand elasticity for aquaculture products is -0.47. Estimated supply elasticities for aquaculture are more difficult to find. In the absence of recent estimates for Bangladesh, we adopt estimates by Kumar, Dey, and Paraguas (2006) of the supply elasticities for three aquaculture products in India: tilapia, milkfish, and shrimp/prawns. We use the simple average of these three elasticities: 1.33. Given the uncertainty about the elasticity estimates, we carried out an analysis of the sensitivity of the results to alternative elasticity estimates. The results are quite similar with no major qualitative differences. These results are available from the authors. Between 2000 and 2010 the quantity of aquaculture products increased by 104 percent. Taking into account population growth over this period, this represents a per capita increase of 76 percent. We use this figure to represent the supply shift, although this is a conservative estimate because it does not take into account the fact that the increase in output occurred in spite of lower fish prices. To calculate producer surplus, we need an estimate of ϑ, the ratio of producer surplus to the gross value of production. The simplest assumption is that the supply function has a constant elasticity, implying a double-log func- tional form. Given our adoption of a supply elasticity of 1.33, the area under Welfare and Poverty Impacts of Aquaculture Growth 91

the supply curve (costs) is 42.9 percent of the area representing gross revenue. This implies that θ, ratio of net revenue to gross revenue, is approximately 0.571. In other words, the producer surplus can be calculated as 57.1 percent of the gross value of aquaculture production. We now have the data and parameter assumptions needed to estimate the proportional change in income for each household in the survey, taking into consideration the importance of aquaculture production and consumption in their income. With the welfare effect of the price and technology changes for each household, it is possible to aggregate the results to the national level using the sampling weights. The results are also aggregated to groups of households defined by location (urban or rural), sex of head of household, region, occupation (fish farmer or not), and income quintile. For each group of households we focus on two measures of improvement: the average proportional change in real income (dY/Y) and the change in the incidence of headcount poverty.

Welfare Implications of Aquaculture Growth

Market Positions and Net Benefit Ratios Based on data from the 2000 HIES survey, Table 5.3 shows descriptive statistics for the NBR of aquaculture for various types of households in Bangladesh. The first column indicates the proportion of all households in that category. The next three columns show the average production ratio (PpQ/Y), the average consumption ratio (PcC/Y), and the average NBR (PpQ/Y - PcC/Y), respectively. The last three columns present the share of households in that category that are net sellers, autarkic (neither buying nor selling), or net buyers. The results are disaggregated by household location (rural or urban), administrative division, the sex of the household head, fishing practices, and income quintile. Starting with the national estimates (the last row of the table), aquaculture production accounts for 2.07 percent of income, while aquaculture consumption represents 3.77 percent of income, giving an NBR of –1.7 percent.11 This means that a 10 percent increase in all aquaculture prices would result in a 0.17 percent decrease in income for Bangladeshi households. The last three columns of the last row indicate that about 80.0 percent of households are net buyers, 12.7 percent are autarkic, and about 7.5 percent are net sellers.

11 Throughout this analysis total household consumption expenditure is used as a proxy for income because income is subject to more measurement errors. 92 chapter 5

Table 5.3 Net positions of households in aquaculture fish in Bangladesh

Percentage of total Percentage of households in income each category Percentage Household of all Production Consumption Net benefit Net Net category households ratio ratio ratio seller Autarkic buyer Location Rural 80 2.51 3.90 −1.38 9.1 13.2 77.6 Urban 20 0.32 3.27 −2.96 1.2 10.4 88.4 Sex of household head Male 91 2.21 3.80 −1.59 8.0 11.9 80.1 Female 9 0.61 3.48 −2.86 2.5 20.1 77.5 Regional Barisal 7 1.99 2.09 −0.10 9.7 34.5 55.8 Chittagong 23 1.77 4.37 −2.60 5.4 12.8 81.8 Dhaka 33 1.78 3.87 −2.09 6.9 8.6 84.6 Khulna 12 5.38 3.74 1.64 14.0 11.6 74.5 Rajshahi 25 1.18 3.54 −2.36 6.7 12.5 80.8 Occupation Fish farmer 23 9.06 4.94 4.12 32.9 9.2 57.9 Other 77 — 3.42 −3.42 — 13.7 86.3 Income quintile Poorest 20 1.52 3.15 −1.63 4.7 21.4 73.9 2nd 20 2.21 3.67 −1.45 8.1 15.1 76.8 3rd 20 2.35 4.16 −1.81 8.6 11.8 79.7 4th 20 2.61 4.10 −1.49 9.5 10.4 80.1 Richest 20 1.65 3.78 −2.13 6.7 4.6 88.6 National 100 2.07 3.77 −1.70 7.5 12.7 79.8

Source: Authors’ calculations based on the BBS HIES 2000. Note: — = data not available.

A couple of other results are worth highlighting. First, estimated NBRs are positive only for fish farmers and for households in the Khulna division, implying that these two groups will gain from an increase in aquaculture prices and lose from a decline in prices, at least in the short run. For fish farmers a positive NBR is obvious. For Khulna, the largest fish-producing region in the country, a positive NBR means that the gains to producers from a price increase will outweigh the losses to consumers. Second, among the household categories with negative NBRs, the estimates vary, from as low as –3.42 for Welfare and Poverty Impacts of Aquaculture Growth 93

nonfarmers to as high as –0.10 percent in Barisal. Similarly, the NBR magnitudes vary by income quintile as well. This implies that the impacts of an increase (decrease) in aquaculture fish prices will be felt differently depending on the household categories. For instance, urban households would benefit more, on average, from a price decline than would rural households. However, the magnitude of welfare gain (loss) due to a decrease (increase) in fish prices is small. A doubling of fish prices would lead to only about a 3 percent decline in the welfare of urban households. The same argument goes for male-headed households.

Impacts of Aquaculture Growth on Income Distribution This section discusses the distributional impact of improved aquaculture technology leading to an expansion of supply and the lower prices resulting from the shift. As discussed earlier, we estimate the welfare impact of these changes for each household in the HIES survey, then aggregate to different categories of households. Table 5.4 shows the percentage change in real income. Similar to Table 5.3, the results are disaggregated by location, gender of the household head, region, occupation (fish farmer or not), and income quintile. Both short- and long-run impacts of aquaculture growth have been analyzed. The first column of figures in Table 5.4 shows the baseline level of per capita consumption expenditure. Not surprisingly, urban households are better off than rural ones. Fish farmers are neither poorer nor richer than other households, on average. The second column gives the short-term price effect, excluding the effect of technological improvement and household response to price changes. Starting at the national level, the short-run price effect results in an increase in real income by 0.88 percent. However, non–fish farmers, representing 77 percent of the Bangladeshi population, gain 1.54 percent from the decline in aquaculture fish prices. Fish farmers lose because a decline in prices reduces their revenues. Although small in magnitudes, Khulna as a whole loses as well. This suggests that the reduction in revenue due to a fall in prices outweighs the gain by consumers in the division. This finding is consistent with the fact that Khulna is the largest producer of aquaculture fish in Bangladesh. The third column of Table 5.4 shows the long-run price effect in terms of the percentage increase in real income. On average, households gain 1.25 percent in real income due to the decline in aquaculture prices. The gains are larger for every group (or the losses are smaller) as households adapt to the price changes. The fourth column presents the percentage of income changes for aquaculture producers. The average gain is 3.33 percent. Every group gains, 94 chapter 5

Table 5.4 Impacts of aquaculture growth on household income

Short-run Long-run effect (Percentage-point change in poverty) Baseline price effect expenditure (Percentage- Price and Price and quantity Household (BDT/capita/ point change in Only price quantity effect effect on all category year) poverty) effect on fishers households (……….………....…….% change in income……...…………..……) Location Rural 11,394 0.77 1.19 3.34 2.23 Urban 17,003 1.32 1.50 3.08 1.63 Sex of household head Male 12,475 0.84 1.23 3.34 2.12 Female 13,186 1.40 1.63 2.88 1.88 Regional Barisal 11,380 0.05 0.37 2.23 1.31 Chittagong 12,685 1.23 1.61 3.97 2.38 Dhaka 13,886 1.02 1.37 3.59 2.10 Khulna 12,563 −0.28 0.38 4.03 2.45 Rajshahi 10,822 1.11 1.39 2.53 1.87 Occupation Fish farmer 12,418 −1.13 −0.10 3.33 3.33 Other 12,553 1.54 1.70 — 1.70 Income quintile Poorest 5,774 0.84 1.12 3.04 1.70 2nd 8,002 0.96 1.31 3.23 2.05 3rd 10,107 0.98 1.38 3.46 2.27 4th 13,382 0.76 1.22 3.51 2.36 Richest 25,339 0.85 1.24 3.30 2.15 National 12,519 0.88 1.25 3.33 2.11

Source: Authors’ analysis based on the BBS HIES 2000. Note: The data are based on the following assumptions: (1) Consumer and producer prices decline by 45 and 36 percent, respectively. (2) Growth in production is 104 percent. (3) Demand and supply elasticity estimates are –0.47 and 1.33, respectively. — = data not available. BDT = Bangladeshi taka. as the gains from technology improvement exceed the losses associated with lower prices. The largest gain is for fish farmers in Khulna, where aquaculture production is concentrated. The last column shows the percentage of change in real income for all households taking into account both price and technology changes. On average, real household income rises by 2.11 percentage points. Every group benefits, but the size of the gain varies. Rural households (including many fish Welfare and Poverty Impacts of Aquaculture Growth 95

farmers) gain more than urban ones. Male-headed households (who are more likely to be net sellers) gain more than female-headed households. Households in Khulna (many of whom are net sellers) gain more than those in other regions. The poorest gain somewhat less than other income groups. This is because they are less likely to be net sellers of aquaculture products and less likely to benefit from falling prices because fish expenditure is a relatively small portion of their budgets.

Impacts of Aquaculture Growth on Poverty Between 2000 and 2010, production of aquaculture fish (excluding shrimp) has more than doubled, and the real prices of major varieties of aquaculture at the retail and farm-gate level declined by 45 percent and 36 percent, respectively. This section presents an analysis of how these changes translate into poverty reduction. We begin by estimating the baseline poverty rate by household categories using the 2000 HIES. The national poverty rate in 2000 was 48.8 percent; rural and urban poverty rates were 52.0 and 35.0 percent, respectively. Therefore these estimates are close to the ones presented in World Bank (2013) and the government’s figures based on the HIES data (BBS 2010). The first column of Table 5.5 shows the poverty rate for each category of household. The poverty rate is similar for male- and female-headed households. Across regions it ranges from 45 percent to 57 percent. It is worth noting that fish farmers have a lower poverty rate (44 percent) than other households. The second column assesses how the poverty rates decline in the short run following a reduction in the real price of aquaculture products. Two points should be highlighted from these results. First, the magnitude of overall reduction in poverty due to price change is small: a 45 percent decline in the retail price leads to a 1.07 percentage point decline in poverty. This finding is not surprising in that, on average, aquaculture products account for less than 4 percent of the household budget. Second, a decline in price does not reduce poverty rates among the bottom two quintiles of households. This does not mean that those households do not gain from price decline; rather, it means that the gains are not large enough to pull those households out of poverty. Finally, poverty among fish-farming households increases by 0.56 percentage points due to short-run price effects, which implies that fish-farming households just above the poverty line could slip into poverty when only price decline is considered. However, this simulation does not account for productivity gains and the second-round effects specified in the second equation on page 88. The long-run estimates are presented in the last three columns of Table 5.5. The first of these three columns reports the long-run effects of a 96 chapter 5

Table 5.5 Impacts of aquaculture growth on poverty reduction

Long-run effect Price and Price and quantity Household Baseline Short-run Price quantity effect effect on all category poverty rate (%) price effect effects on fishers households (………………. percentage point change in poverty ……………….) Location Rural 52 −0.84 −1.14 −2.63 −1.91 Urban 35 −0.50 −0.78 −2.35 −0.85 Sex of household head Male 49 −0.71 −1.03 −2.72 −1.70 Female 47 −1.71 −1.71 0.00 −1.71 Regional Barisal 53 −0.98 −1.15 −3.54 −2.15 Chittagong 45 −0.74 −1.14 −3.63 −1.79 Dhaka 46 −0.70 −1.03 −2.84 −1.56 Khulna 45 −0.47 −0.47 −1.05 −1.87 Rajshahi 57 −1.00 −1.31 −2.18 −1.56 Occupation Fish farmers 44 0.56 −0.10 −2.63 −2.63 Others 50 −1.21 −1.39 — −1.39 Income quintile Poorest 100 0.00 0.00 0.00 0.00 2nd 100 0.00 0.00 0.00 0.00 3rd 44 −4.33 −5.82 −12.2 −8.49 4th 0 0.27 0.27 0.00 0.00 Richest 0 0.21 0.21 0.00 0.00 National 49 −0.77 −1.07 −2.63 −1.70

Source: Authors’ analysis based on BBS HIES 2000. Note: This set of simulations is based on the following assumptions: (1) Consumer and producer prices decline by 45 and 36 percent, respectively. (2) Growth in production is 104 percent. (3) Demand and supply elasticity estimates are –0.47 and 1.33, respectively. — = data not available. decline in prices, taking into account household response; the second column shows the joint impact of price decline and productivity growth on fish-farming households only; and the final column presents the long-run impacts on all households. Four general conclusions can be drawn from these results. First, the long-run poverty impacts are much larger than the short-run impacts. The impact of price decline on poverty in the long run is 1.07 percentage points, which is almost 40 percent larger than the short-run Welfare and Poverty Impacts of Aquaculture Growth 97

impacts. Second, households in the third income quintile benefit the most. The magnitudes of poverty reduction are as high as 12.2 percentage points if only fish-farming households are considered; they are 8.49 percentage points if all households are considered. Since the poverty rate is 49 percent across the population, many households in the third quintile are close to the poverty line. Third, among the other household categories, overall poverty reduction is greatest among fish-farming households (2.63 percentage points), followed by households in Barisal (2.15 percentage points). However, the impacts are also substantial for other divisions. The final points to highlight are that there is no change in the poverty rate for the poorest two quintiles and the richest two quintiles. These households are too far from the poverty line for these changes in income to affect their poverty status. To summarize, the all-inclusive long-run results suggest that, at the aggregate level, the growth of aquaculture in Bangladesh between 2000 and 2010 was responsible for a 2.11 percent increase in real incomes and a 1.70 percentage point reduction in poverty. Official statistics show that overall poverty in the country declined from 48.9 percent to 31.5 percent during the same period, for a decrease of 17.4 percentage points. As a first approximation, this implies that about 10 percent (2.06/17.4) of the overall reduction in poverty can be attributed to the growth in aquaculture during 2000–2010.12

Conclusion Defying many earlier predictions, global fish production has grown faster than the world’s population in the past couple of decades. Aquaculture has been the biggest contributor to this growth, with its share in total fish production rising from about 16 percent in 1990 to more than 50 percent in 2012. This remarkable transformation has important implications for food security, poverty, and environment, especially for developing countries of Asia. This chapter contributes to that literature by clarifying some conceptual issues and presenting the quantitative evidence on the income and poverty impacts of aquaculture growth in Bangladesh. Conceptually, the aquaculture–poverty linkage literature closely resembles agricultural growth linkage literature. Evolved in the 1960s, the agricultural growth linkage literature demonstrated that in addition to its direct

12 A more in-depth analysis would decompose the sources of poverty reduction by growth in different sectors, taking into account interactions and general equilibrium effects. However, there is no reason to believe that our analysis systematically overestimates the share of poverty reduction that can be attributed to growth in aquaculture. 98 chapter 5

benefits, agriculture generates multiplier effects through increased demand of nontradable goods outside of agriculture. The existing studies on aquaculture’s link to poverty and income distribution appear to miss this impact channel, however, as they focus only on aquaculture and ignore links to other sectors. The existing empirical literature has two important gaps—problems of generalizability of results and the quantification of income and poverty impacts. Because most of the studies are not based on representative samples, their conclusions cannot be generalized, even though they offer useful insights into the process. Perhaps the most important gap in the literature is that previous studies did not quantify the degree to which aquaculture has contributed to income growth and poverty reduction. This chapter addresses both of these gaps by estimating the income and poverty impacts using nationally representative household survey data from Bangladesh, a country that has been successful in promoting aquaculture and reducing poverty since the early 2000s. The results of this study suggest that the impacts of aquaculture growth on income distribution and poverty reduction in Bangladesh have been substantial, even though the impacts on households in the bottom income quintile have been modest. We estimate that aquaculture’s contribution to income growth between 2000 and 2010 was 2.11 percent, including both price and quantity effects. This income growth has translated into an estimated poverty reduction of 1.7 percentage points nationwide. Although these estimates seem small, they represent a substantial share of overall poverty reduction between 2000 and 2010. For instance, national headcount poverty rates declined from 48.9 percent in 2000 to 31.5 percent in 2010 (World Bank 2013). This implies that the growth in aquaculture has been responsible for almost 10 percent of the overall poverty reduction in Bangladesh during the first decade of the 21st century. Put differently, of the 18 million Bangladeshis who escaped poverty during 2000–2010, more than 2 million of them managed to do so because of aquaculture. Although these are impressive numbers, this is only part of the story. More analysis is needed to better understand the true benefits and costs of promoting a Blue Revolution strategy. First, a general equilibrium analysis could incorporate additional impact channels, such as through wage rates, which would likely generate larger benefits for the daily wage worker who belongs to the bottom income quintile. Second, it would be useful to explore the nutritional impact of the expansion of aquaculture in Bangladesh through both changes in income and the reduced price of fish and other aquaculture products. Finally, the environmental impact of the Blue Revolution needs to be Welfare and Poverty Impacts of Aquaculture Growth 99

incorporated into the analysis, raising difficult questions regarding the quantification of costs.

References Adelman, I., and C. Morris. 1973. Economic Growth and Social Equity in Developing Countries. Stanford, CA, US: Stanford University Press. Ahmed, M., and M. H. Lorica. 2002. “Improving Developing Country Food Security through Aquaculture Development—Lessons from Asia.” Food Policy 27 (2): 125–141. Bangladesh, DoF (Department of Fisheries). 1994. Fisheries Statistical Year Book of Bangladesh 1992– 1993. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock. —. 1997. Fisheries Statistical Year Book of Bangladesh 1995–1996. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

—. 2006. Fisheries Statistical Year Book of Bangladesh 2004–2005. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

—. 2015. Fisheries Statistical Year Book of Bangladesh 2013–2014. Dhaka: Fisheries Resource Survey System, DoF, Ministry of Fisheries and Livestock.

BBS (Bangladesh Bureau of Statistics). 2000. Household Income and Expenditure Survey 2000. Dhaka. —. 2005. Household Income and Expenditure Survey 2005. Dhaka. —. 2010. Report of the Household Income and Expenditure Survey 2010. Dhaka: Ministry of Planning. Behrman, J. R., and A. B. Deolalikar. 1987. “Will Developing Country Nutrition Improve with Income? A Case Study for Rural South India.” Journal of Political Economy 95 (3): 492–507. Belton, B., M. M. Haque, and D. C. Little. 2012. “Does Size Matter? Reassessing the Relationship between Aquaculture and Poverty in Bangladesh.” Journal of Development Studies 48 (7): 904–922. Bogard, J. R., S. H. Thilsted, G. C. Marks, M. A. Wahab, M.A.R. Hossain, J. Jakobsen, and J. Stangoulis. 2015. “Nutrient Composition of Important Fish Species in Bangladesh and Potential Contribution to Recommended Nutrient Intakes.” Journal of Food Composition and Analysis 42: 120–133. Bouis, H. E. 2000. “Commercial Vegetable and Polyculture Fish Production in Bangladesh: Their Impacts on Household Income and Dietary Quality.” Food and Nutrition Bulletin 21 (4): 482–487. Bouis, H. E., and L. J. Haddad. 1992. “Are Estimates of Calorie-Income Elasticities Too High? A Recalibration of the Plausible Range.” Journal of Development Economics 39: 333–364. 100 chapter 5

Deaton, A. 1989. “Rice Prices and Income Distribution in Thailand: A Non-parametric Analysis.” Economic Journal 99 (395): 1–37. De Janvry, A., and E. Sadoulet. 2010. “Agricultural Growth and Poverty Reduction: Additional Evidence.” World Bank Research Observer 25 (1): 1–20. Dey, M. M., M. F. Alam, and M. L. Bose. 2010. “Demand for Aquaculture Development: Perspectives from Bangladesh for Improved Planning.” Reviews in Aquaculture 2 (1): 16–32. Dey, M. M., M. F. Alam, and F. J. Paraguas. 2011. “A Multistage Budgeting Approach to the Analysis of Demand for Fish: An Application to Inland Areas of Bangladesh.” Marine Resource Economics 26 (1): 35–58. Economist. 2003. “Fish Farming: The Promise of a Blue Revolution.” Accessed December 1, 2017. www.economist.com/node/1974103. FAO (Food and Agriculture Organization of the United Nations). 2005. “Bangladesh National Aquaculture Sector Overview.” Accessed December 11, 2017. www.fao.org/fishery/ countrysector/naso_bangladesh/en.

—. 2014. The State of World Fisheries and Aquaculture: Opportunities and Challenges. Rome: FAO Fisheries and Aquaculture Department. Hirschman, A. O. 1957. “Investment Policies and ‘Dualism’ in Underdeveloped Countries.” American Economic Review 47 (5): 550–570. Ivanic, M., and W. Martin. 2008. “Implications of Higher Global Food Prices for Poverty in Low-Income Countries.” Agricultural Economics 39 (s1): 405–416. Johnston, B. F., and J. W. Mellor. 1961. “The Role of Agriculture in Economic Development.” American Economic Review 51 (4): 566–593. Kabir, S. H. 2015. Sea Food Export from Bangladesh and Current Status of Traceability. Accessed August 21, 2018. www.unescap.org/sites/default/files/6-%20%20Sea%20Food%20 Export%20from%20Bangladesh-Kabir.pdf.

Kakwani, N., S. Khandker, and H. H. Son. 2004. Pro-Poor Growth: Concepts and Measurement with Country Case Studies. International Poverty Centre Working Paper 1. Brasilia, Brazil: United Nations Development Programme, International Poverty Centre. Kumar, P., M. M. Dey, and F. J. Paraguas. 2006. “Fish Supply Projections by Production Environments and Species Types in India.” Agricultural Economics Research Review 19 (2): 327–351.

Mellor, J. W., and U. Lele. 1972. “Growth Linkages of New Agricultural Technology.” Indian Journal of Agricultural Economics 18: 10–15. Minot, N., and F. Goletti. 1998. “Export Liberalization and Household Welfare: The Case of Rice in Vietnam.” American Journal of Agricultural Economics 80 (4): 738–749. Welfare and Poverty Impacts of Aquaculture Growth 101

Naylor, R. L., R. J. Goldburg, J. H. Primavera, N. Kautsky, M. C. Beveridge, J. Clay, C. Folke, J. Lubchenco, H. Mooney, and M. Troell. 2000. “Effect of Aquaculture on World Fish Supplies.” Nature 405 (6790): 1017–1024. Pingali, P. 2001. “Environmental Consequences of Agricultural Commercialization in Asia.” Environment and Development Economics 6 (4): 483–502. Ravallion, M. 2004. Pro-Poor Growth: A Primer. Washington, DC: Development Research Group, World Bank.

—. 2009. “Defining Pro-Poor Growth: A Response to Kakwani.” IPC-IG Collection of One Pagers. Brasilia, Brazil: United Nations Development Programme International Poverty Centre. Roos, N., M. A. Wahab, C. Chamnan, and S. H. Thilsted. 2007. “The Role of Fish in Food-Based Strategies to Combat Vitamin A and Mineral Deficiencies in Developing Countries.” Journal of Nutrition 137 (4): 1106–1109. Sachs, J. D. 2007. “The Promise of the Blue Revolution: Aquaculture Can Maintain Living Standards while Averting the Ruin of the Oceans.” Scientific American 297 (1): 37–38. Shamsuzzaman, M. M., M. M. Islam, N. J. Tania, M. A. Al-Mamun, P. P. Barnum, and X. Xu. 2017. “Fisheries Resources of Bangladesh: Present Status and Future Direction.” Aquaculture and Fisheries 2: 145–156. Subramanian, S., and A. Deaton. 1996. “The Demand for Food and Calories.” Journal of Political Economy 104 (1): 133–162. Toufique, K. A., and B. Belton. 2014. “Is Aquaculture Pro-Poor? Empirical Evidence of Impacts on Fish Consumption.” World Development 64: 600–620. World Bank. 2008. The World Development Report 2008: Agriculture for Development. Washington, DC.

—. 2013. Bangladesh Poverty Assessment: Assessing a Decade of Progress in Reducing Poverty 2000–2010. Accessed December 1, 2017. https://openknowledge.worldbank.org/bitstream/ handle/10986/16622/785590NWP0Bang00Box0377348B0PUBLIC0.pdf?sequence=1.

Chapter 6

FUTURE SCENARIOS (PROJECTIONS TO 2050)

Paul Dorosh and Andrew Comstock

Introduction The Bangladesh fish sector has experienced both rapid growth and rapid change over the past several decades. With plentiful waterways, access to the sea, and a subtropical climate, prospects for future production growth are equally bright. Domestic demand for fish products is also increasing as rising incomes and more efficient value chains make fish products more accessible and affordable for both rural and urban households. How fast supply grows relative to demand will determine not only future movements in real prices of fish products but also the feasibility of significant levels of exports. Our analysis of future production trends involves modeling different growth rates of the various fish production systems: inland capture, aquaculture, and marine fisheries. Aquaculture development has recently been stressed by both private and public sector organizations (FAO 2014). Indeed, aquaculture productivity has increased, although this has not been the case for other fishery systems such as inland capture and marine fisheries. All these systems are beset with their own constraints, but each have unique opportunities to contribute to future growth. The analysis of future demand is based on econometric estimates of own price and expenditure elasticities for four different fish groups using a modified Quadratic Almost Ideal Demand System (QUAIDS). Elasticities are estimated within rural and urban household groups, and within these subsets, poor and nonpoor. These parameters are then used, along with other exogenous parameters, in a multimarket model. Using this model, we can provide a baseline estimation of future fish production, demand, and prices. We also model the effects of hypothetical shocks to fish production to investigate their impacts on the fish economy. The next section of the chapter provides a brief overview of fish production systems in Bangladesh along with their historical trends. The following

103 104 chapter 6

section discusses the estimation of demand parameters, including the equations and variables used as well as the empirical results. Thereafter, we provide a description of the multimarket model and the other parameters involved, along with the results of the baseline estimations and the policy simulations. The final section concludes.

Fish Production Systems: Present and Past As discussed in Chapter 1, there are three primary systems of fish production in Bangladesh: aquaculture, inland capture, and marine capture. Aquaculture is primarily made up of pond culture, which accounts for nearly 86 percent of this fish production system. Although this type of fish production is more easily accessible at the household level, unless a household invests with the intent of making aquaculture its primary income source, production levels tend to remain low. Inland capture involves the capture of wild fish from streams, rivers, and lakes, and the types of various freshwater fish produced from inland capture have historically been staples in Bangladeshi diets. Finally, marine fishing (including shrimp), the third fish production system, is mostly dominated by small-scale fisheries and faces the significant challenge of overfishing. All three production systems have notable costs and benefits, and all play considerable roles in the Bangladeshi economy and Bengali diets. The prominence of aquaculture has increased significantly in Bangladesh since 2004, both as a share of fish production and as kilograms produced per capita. As discussed in Chapter 1, of the three systems, aquaculture production has grown the most over the past 15 years. Marine and inland capture also grew but not to the extent of aquaculture. The historical consumption data mirrors these patterns, with the consumption of aquaculture fish growing significantly and that of inland capture decreasing in importance (consumption of marine fish decreased in rural areas but increased in urban areas, likely because it is considered a luxury good). In terms of value shares of fish in total expenditures, urban areas previously consumed the larger share of aquaculture relative to total consumption compared with rural areas; however, this flipped by 2010, likely due to expanding production and lower prices. Finally, calculating the real prices over time shows that the real prices of fish from mixed, inland capture, and marine systems rose by 21, 29, and 32 percent respectively, but prices of aquaculture fish dropped by 12 percent. Future Scenarios (Projections to 2050) 105

Determinants of Fish Demand

Economic Specification To obtain expenditure and price elasticities for the simulations, we estimate a QUAIDS model using data from the Bangladesh Bureau of Statistics (BBS) Household Income and Expenditure Survey (HIES) in the years 2000, 2005, and 2010. The main variables of interest were those of quantity of items consumed, total expenditures on items, and price (which was imputed from the total expenditure and quantity consumed of each item).1 A three-stage methodology was used for the analysis, adapted from Ecker and Qaim (2010). The first stage estimates the elasticity for food versus nonfood. This eliminates the need for predicting missing nonfood prices but relies on the assumption that consumption decisions are first made by the individual at a food versus nonfood level, rather than nonfood versus any particular food category. The second stage estimates the elasticities within the six major food groups of the study, and the third stage estimates the elasticities within the four different types of fish. Again, the assumption is that consumption decisions are more likely to be made within these smaller groups than across them. Unconditional elasticities were then calculated using the estimates from the three stages. The households in the sample were divided into rural and urban subgroups, and the methodology was applied separately within these groups. This was done with the idea that the rural and urban samples would most likely differ in their responses to price and expenditure changes. Full sample estimates were potentially troublesome due to excessive heterogeneity within the sample. As such, full sample elasticities were calculated by adding up the quantity weighted elasticities of the two smaller samples. Elasticities were also calculated, within the rural and urban samples, at the means for the bottom three and top two quintiles of expenditure (in later tables these are referred to as “poor” and “nonpoor”). Standard errors for the elasticities estimation were calculated via bootstrapping at the sample mean for the poor and nonpoor populations. The number of repetitions was calibrated using Poi’s bssize command in Stata statistical software.2

1 Outliers for quantity and expenditure observations were defined as those outside of five standard deviations from the mean, within each category, and were removed from the final sample. 2 The bssize (Poi 2004) command was calibrated to identify the number of repetitions needed to return standard errors that did not deviate by more than 5 percent from the ideal bootstrapped values with 99 percent probability. 106 chapter 6

Possible endogeneity of total expenditures was a concern in estimation. To address this, in all stages of the estimation, an instrumental regression was used to obtain a predicted value of total expenditures. The instrumental regression used is given in equation 1:

Eh = β0 + β1(lnYh) + εh (1)

where Eh is total household expenditures, Yh is the real income of household h,

and εh is an error term. A description of the variables in this and the following regressions is in Table 6.A1. In the first stage of the estimation, a Working–Leser (Working 1943; Leser 1963) model was used to estimate the price and expenditure elasticities

of food versus nonfood consumption. The expenditure share of food, wif , was estimated using equation 2:

wif = β1 log(êl) + β2Pf + Σβ3(Demi) + Σβ4(Districti) + εij (2)

where log(êl) is the log of the estimated total expenditure (on food and non-

food) from the instrumental regression, Pf is the average price of food, Demi is a vector of household-level demographics that includes household size, age and sex of the head of household, square footage of the household, and the log of total household assets. Districti is a set of district dummies for all 94 districts, and εij is an error term. The conditional expenditure and Marshallian price elasticities were calculated using Leser’s (1963) formulae shown in equations 3 and 4: = 1 + β1 (3) ei wi = − + β2 (4) eif δij wi where ei is the expenditure elasticity, and eif is the price elasticity. The Kronecker delta is represented by δ. Within the second and third stages a modified QUAIDS model is estimated. The model (as well as the Stata statistical software code) was again adapted from Ecker and Qaim (2010). Adjustments were made to the standard QUAIDS model to account for censoring, explained below, and potential endogeneity, explained above. The same demographic control variables used in the Working–Leser model were also accounted for. In the food consumption data there were a number of zero-consumption observations, particularly among the observations for the separated four types of fish. To account for these, first a multivariate probit model was estimated for household consumption of the six categories of food, in the second stage, Future Scenarios (Projections to 2050) 107

and the four categories of fish, in the third stage. The probit model was estimated using equation 5:

dij = β1 log(êt) + Σβ2(Demi) + eij (5) where i and j index households and commodity subgroups respectively, dij indicates whether a household consumed a certain commodity (with dij = 1 if household i consumed good j), Demi is the same vector of household characteristics used before, and eij is an error term. From these probit models the respective probability density and cumulative distribution functions were estimated and used to create the inverse mills ratio for each of the zero-observation commodities. These were incorporated into the final QUAIDS models as demographic shifters to act as instruments correcting for zero observations. With these adjustments the final QUAIDS model is shown in equation 6:

n  λj  wij = αi + Σ j=1 yij ln Pi + βiln[a(p)] + b(p) {ln[a(p)]}2 (6) where wij is the consumption share of good j in household i, and ln Pi and p are the trans-log price aggregator functions.3 Demographic shifters are incorporated into the equation through the α’s. Finally, unconditional price elasticities were calculated using the estimates from all three stages. The expenditure elasticities are calculated relatively simply, by multiplying through the elasticities for all stages by good. The unconditional Marshallian were calculated using Edgerton’s (1997) formula in equation 7:

u c * c * u eij = δrse(r)ij + E(r)iw(s)j(ers + Erwsef) (7) where i and j represent the subgroups at the third stage, r and s are the larger food subgroups in the second stage, and f represents the first stage of aggregate food. The Kronecker delta is calculated at the second stage and is again represented by δ.4

3 Table 6.A2 in Annex III contains the descriptive statistics of all consumption shares and prices for the various food groups. 4 Alternative elasticity estimates were also calculated via an iterative linear least squares approach (using Lecoq and Robin’s aidsills command [2015]) and a QUAIDS model which did not account for censoring (using Poi’s QUAIDS command [2012]). Standard errors in the first of these alternative models are calculated from the command itself using the delta methodology, while standard errors for the second of these were calculated in the same method as our primary specification. Results can be found in Tables 6.A6a and 6.A6b. Generally, the expenditure elasticity estimates produced by the alternate methods fell within the range of the model sensitivity testing described in Annex II. Those that did not were nearly all larger in magnitude and in the marine fish category—the category with the most censoring. The alternative price elasticity estimates also mostly fell within the sensitivity testing range. All of those outside the range were on the higher end. Since neither of the 108 chapter 6

Demand Parameter Results Table 6.1 presents the results of the demand parameter estimation, including estimated expenditure elasticities for all three years of the HIES used in this study. Consistent with the historical increase in aquaculture fish consumption, aquaculture expenditure elasticities increased for all household groups from 2000 to 2010. Inland capture, which saw much slower growth than aquaculture and increasing prices, showed mixed results across rural and urban households. Inland capture expenditure elasticities decreased in rural areas and increased in urban areas, with the biggest shift being a decrease for the rural poor. As this is the largest group in Bangladesh, the overall effect was a decrease in expenditure elasticities. Note that expenditure elasticities for the rural poor decreased for all types of fish except for aquaculture. For the other household groups, however, expenditure elasticities increased for most of the categories of fish. As shown in Table 6.2, estimates of the own-price elasticities of aquaculture decrease (that is, become more negative) for all household groups from 2000 to 2010. Thus demand for aquaculture fish has become less responsive to price changes over time. In contrast, the price responsiveness of most other fish has increased (that is, become more elastic) over time. As we saw before, expenditure elasticities for aquaculture increased across all groups, so this move toward a more inelastic price response is surprising. Further data and analysis, disaggregated by type of fish, is needed. Also surprising is that the cross-price elasticity between aquaculture and inland capture decreases by relatively large margins within each group, in some cases with changes in sign from positive to negative across the three surveys. This could indicate that the two are becoming stronger complements rather than substitutes. However, further detailed analysis of consumption by fish type is needed to untangle the reasons for these changes. Note also that the cross price of elasticities of demand of inland capture fish with respect to the price of aquaculture increase for nearly all household groups, with positive cross-price elasticities of demand for all household groups in 2010 (indicating the two types of fish are substitutes). Again, further analysis is needed.5

alternative estimates account for censoring, it makes sense that the estimates produced by these methods are larger, as a household’s zero-consumption observation is always treated as a direct response to increases in price or income. As such, we believe our primary estimation technique, and estimates, to be more valid. 5 Annex III contains alternative estimates of own-price and cross-price elasticities used to validate the parameters estimated for the model. Future Scenarios (Projections to 2050) 109

Table 6.1 Bangladesh: Expenditure elasticities by fish production system (QUAIDS model estimates)

Rural Standard Rural Standard Urban Standard Urban Standard poor error nonpoor error poor error nonpoor error 2000 expenditure elasticities Primarily aquaculture 0.99 (0.0001) 0.66 (0.002) 0.91 (0.0001) 0.57 (0.004) Mixed 0.95 (0.0001) 0.64 (0.002) 1.16 (0.0001) 0.80 (0.006) Primarily inland 1.04 (0.0001) 0.68 (0.002) 0.97 (0.0001) 0.68 (0.005) capture Primarily marine 0.98 (0.0001) 0.64 (0.002) 0.70 (0.0001) 0.32 (0.002) 2005 expenditure elasticities Primarily aquaculture 1.29 (0.002) 1.07 (0.004) 1.44 (0.004) 1.15 (0.009) Mixed 0.84 (0.001) 0.57 (0.002) 0.94 (0.003) 0.61 (0.005) Primarily inland 0.91 (0.001) 0.67 (0.002) 1.02 (0.003) 0.72 (0.005) capture Primarily marine 1.23 (0.002) 1.11 (0.004) 1.38 (0.004) 1.19 (0.009) 2010 expenditure elasticities Primarily aquaculture 1.03 (0.001) 0.93 (0.002) 1.19 (0.001) 1.19 (0.003) Mixed 0.83 (0.0001) 0.77 (0.001) 0.62 (0.001) 0.53 (0.001) Primarily inland 0.72 (0.0001) 0.57 (0.001) 1.13 (0.001) 0.99 (0.003) capture Primarily marine 0.86 (0.001) 0.74 (0.001) 0.93 (0.001) 0.78 (0.002)

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: QUAIDS = Quadratic Almost Ideal Demand System.

Model Simulations

Model Specification To project future fish supply, demand, and prices and to analyze the effects of various shocks, we use a simple partial equilibrium multimarket model. 6 We specify a set of demand and supply equations for the fish market in Bangladesh and input parameters across four types of fish and four household groups. As explained above, some of the demand parameters were estimated using the various HIES results in Bangladesh. The HIES was also used to calculate household incomes and consumption and population. The other

6 Multimarket models are particularly useful for looking at sectoral-level analysis rather than full economies and can allow for disaggregated policy impact evaluation (Braverman and Hammer 1986; Sadoulet and de Janvry 1995; Croppenstedt et al. 2007). 110 chapter 6

Table 6.2 Bangladesh: Econometric estimates of price elasticities of demand for fish, 2000, 2005, and 2010

2000 price elasticities Rural Urban Primarily Primarily Primarily Standard Standard inland Standard Primarily Standard Primarily Standard Standard inland Standard Primarily Standard aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −0.71 (0.0001) 0.10 (0.0001) −0.01 (0.0001) −0.09 (0.0001) −0.69 (0.0001) 0.11 (0.0001) 0.13 (0.0001) 0.03 (0.0001) (households aquaculture in bottom Mixed −0.10 (0.0001) −0.86 (0.0001) −0.06 (0.0001) −0.19 (0.0001) 0.18 (0.0001) −0.83 (0.0001) 0.19 (0.0001) −0.02 (0.0001) 40% of expenditures) Primarily −0.06 (0.0001) 0.06 (0.0001) −0.91 (0.0001) −0.16 (0.0001) 0.09 (0.0001) 0.12 (0.0001) −0.81 (0.0001) 0.08 (0.0001) inland capture Primarily 0.04 (0.0001) 0.21 (0.0001) 0.12 (0.0001) −0.88 (0.0001) 0.02 (0.0001) 0.01 (0.0001) 0.17 (0.0001) −0.69 (0.0001) marine Nonpoor Primarily −0.81 (0.0000) 0.01 (0.0001) −0.07 (0.0001) −0.14 (0.0001) −0.69 (0.0001) 0.03 (0.0001) 0.02 (0.0001) −0.02 (0.0001) (households aquaculture in top 60% of Mixed 0.03 (0.0001) −0.88 (0.0000) −0.02 (0.0001) −0.12 (0.0001) 0.34 (0.0001) −0.89 (0.0001) 0.08 (0.0001) 0.02 (0.0001) expenditures) Primarily 0.05 (0.0001) 0.07 (0.0001) −0.89 (0.0000) −0.06 (0.0001) 0.28 (0.0001) 0.13 (0.0001) −0.86 (0.0001) 0.09 (0.0001) inland capture Primarily 0.08 (0.0001) 0.13 (0.0001) 0.10 (0.0001) −0.88 (0.0000) −0.02 (0.0001) 0.05 (0.0001) 0.06 (0.0001) −0.66 (0.0001) marine 2005 price elasticities Rural Urban Primarily Primarily Primarily Standard Standard inland Standard Primarily Standard Primarily Standard Standard inland Standard Primarily Standard aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −0.83 (0.0001) 0.10 (0.0001) −0.05 (0.0001) −0.16 (0.0001) −0.53 (0.0001) 0.41 (0.0001) 0.25 (0.0001) 0.14 (0.0001) (households aquaculture in bottom Mixed 0.02 (0.0001) −0.88 (0.0001) −0.14 (0.0001) −0.13 (0.0001) 0.25 (0.0001) −0.66 (0.0001) 0.09 (0.0001) 0.09 (0.0001) 40% of expenditures) Primarily 0.24 (0.0001) −0.04 (0.0001) −0.86 (0.0001) 0.02 (0.0001) 0.39 (0.0001) 0.11 (0.0001) −0.71 (0.0001) 0.17 (0.0001) inland capture Primarily −0.23 (0.0001) 0.38 (0.0001) 0.14 (0.0001) −0.91 (0.0001) −0.06 (0.0001) 0.54 (0.0001) 0.31 (0.0001) −0.75 (0.0001) marine Nonpoor Primarily −0.92 (0.0001) 0.01 (0.0001) −0.14 (0.0001) −0.21 (0.0001) −0.56 (0.0001) 0.37 (0.0001) 0.22 (0.0001) 0.15 (0.0001) (households aquaculture in top 60% of Mixed 0.15 (0.0001) −0.94 (0.0001) −0.11 (0.0001) −0.11 (0.0001) 0.29 (0.0001) −0.80 (0.0001) 0.03 (0.0001) 0.03 (0.0001) expenditures) Primarily 0.44 (0.0001) −0.12 (0.0001) −0.92 (0.0001) 0.05 (0.0001) 0.55 (0.0001) −0.01 (0.0001) −0.81 (0.0001) 0.16 (0.0001) inland capture Primarily −0.22 (0.0001) 0.44 (0.0001) 0.15 (0.0001) −0.92 (0.0001) −0.06 (0.0001) 0.60 (0.0001) 0.31 (0.0001) −0.77 (0.0001) marine Future Scenarios (Projections to 2050) 111

Table 6.2 Bangladesh: Econometric estimates of price elasticities of demand for fish, 2000, 2005, and 2010

(continued) 112 chapter 6

Table 6.2 Continued

2010 price elasticities Rural Urban Primarily Primarily Primarily Standard Standard inland Standard Primarily Standard Primarily Standard Standard inland Standard Primarily Standard aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −0.81 (0.0001) −0.03 (0.0001) −0.20 (0.0001) −0.21 (0.0001) −0.92 (0.0001) 0.09 (0.0001) −0.27 (0.0001) −0.18 (0.0001) (households aquaculture in bottom Mixed 0.10 (0.0001) −0.79 (0.0001) −0.08 (0.0001) −0.17 (0.0001) 0.11 (0.0001) −0.78 (0.0001) −0.14 (0.0001) −0.09 (0.0001) 40% of expenditures) Primarily 0.12 (0.0001) 0.14 (0.0001) −0.80 (0.0001) 0.00 (0.0001) 0.15 (0.0001) 0.14 (0.0001) −0.83 (0.0001) 0.17 (0.0001) inland capture Primarily 0.25 (0.0001) 0.15 (0.0001) 0.04 (0.0001) −0.80 (0.0001) 0.16 (0.0001) 0.06 (0.0001) 0.12 (0.0001) −0.80 (0.0001) marine Nonpoor Primarily −0.87 (0.0002) −0.03 (0.0001) −0.25 (0.0001) −0.20 (0.0001) −0.98 (0.0001) 0.06 (0.0001) −0.25 (0.0001) 0.01 (0.0001) (households aquaculture in top 60% of Mixed 0.20 (0.0001) −0.81 (0.0002) −0.08 (0.0001) −0.14 (0.0001) 0.17 (0.0001) −0.76 (0.0001) −0.14 (0.0001) −0.01 (0.0001) expenditures) Primarily 0.15 (0.0001) 0.11 (0.0001) −0.85 (0.0002) 0.05 (0.0001) 0.12 (0.0001) 0.02 (0.0001) −0.80 (0.0001) 0.35 (0.0001) inland capture Primarily 0.30 (0.0001) 0.05 (0.0001) 0.01 (0.0001) −0.81 (0.0002) 0.04 (0.0001) −0.07 (0.0001) 0.04 (0.0001) −0.82 (0.0001) marine

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. exogenous variables and parameters derive from a variety of sources, including World Bank data (growth rates of population and GDP per capita) and BBS data (quantities of fish production and exports).7 The full list of model variables and parameters is given in Table 6.3 below. The model consists of five sets of equations. Household income per capita is estimated as the base level of per capita income (Y0h) multiplied by an exogenous rate of growth (1 + ygrh) (equation 8):

Yh = Y0h * (1 + ygrh) (8) Quantity supplied (production) of each fish type i is calculated as the base level of quantity QS0 supplied, the (exogenous) productivity growth rate

(qsgri) and current to base year prices (Pi / P0i):

esi = 0 * (1 + ) * Pi (9) QSi QS i qsgri P0i where esi is the elasticity of supply of fish type i.

7 A table documenting the historical trend of exports can be found in Table 6.A3 in Annex III. Future Scenarios (Projections to 2050) 113

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010.

Table 6.3 Model variables and parameters

QDh,i Per capita quantity demanded/consumed

Yh Per capita household income

ygrh Household income growth rate

Poph Population, by household group

popgrh Population growth rate

QSi Quantity produced, by fish type

qsgri Production growth rate E Exports

Pi Fish price

esi Elasticity of supply

edhhi,i,h Household price elasticity

edyi,h Household income elasticity

Source: Authors’ calculations. 114 chapter 6

Per capita household demand of each fish type i for household h is calculated as a function of the base level of demand, QD0hi, ratios of current to base year prices (Pi / P0i), and the ratio of current to base year household income (Yh / Y0h):

edhij Y edyhi = 0 Π Pi h QDhi QD hi * P0i * Y0h (10) where edij is the elasticity of demand of fish type i with respect to the price of commodity j. Aggregate demand is simply the sum of per capita household demands multiplied by the population of the household group (the base year population multiplied by one plus the exogenous population growth rate) (equation 11).

QDi = Σ(QDhi * (Poph * (1 + popgrh))) (11) Finally, equation 12 defines the market equilibrium condition: total quantity demanded by domestic households is equal to supply minus exports.8

QDi = QSi − Ei (12)

Simulation Results In the base scenario we set productivity growth for aquaculture and mixed production systems to an average of 3.7 and 1.5 percent per year, respectively— rates that are slightly lower than recent historical growth rates of production (Table 6.4). For inland capture and marine capture we assume no productivity growth, given possible limits on sustainable increases in production.9 Consistent with recent trends, urban population is projected to grow much faster than rural population (3.0 and 0.4 percent per year, respectively), with faster growth with overall incomes in urban areas as well (Table 6.5). Overall, we model five policy scenarios with various combinations of productivity growth and increased demand (arising from exogenous increases in household incomes). Table 6.6 presents the simulation results in terms of annual growth rates of production and prices from 2015 to 2030. In the base simulation total fish production from 2015 to 2030 increases by an average of 2.61 percent per

8 Exports in the model are assumed to be 20 percent of the baseline production for all categories. For the following years, aquaculture exports are set at 50 percent of the previous year’s production to reflect increased exports in this subcategory. 9 These productivity shocks are chosen to produce trends in production and real prices similar to those observed from 2010 to 2015. Similar assumptions were used in World Bank (2007). Future Scenarios (Projections to 2050) 115

Table 6.4 Model simulation assumptions for fish productivity growth

Inland Aquaculture capture Mixed Marine Average growth rate (%) Base 3.7 0.0 1.5 0.0 High productivity 5.3 0.8 2.1 0.0 Very high productivity 6.7 0.8 2.1 0.0 Cumulative growth (%) Base 71.5 0.0 25.0 0.0 High productivity 118.0 12.7 36.6 0.0 Very high productivity 163.1 12.7 36.6 0.0

Source: Authors’ calculations. Note: Productivity growth rates for aquaculture in the base scenario are set equal to 5 percent for 2016 to 2020 and 3 percent for 2021 to 2030. For the high-productivity scenarios, productivity growth rates for aquaculture are set equal to 5 and 3 percent for 2016–2020 and 2021–2030. Similarly, for the very high-productivity growth scenarios, aquaculture productivity growth is set equal to 8 and 6 percent in the two periods, respectively. year (47.2 percent in fifteen years), with very slow rates of growth for inland capture and marine fish production (1.56 and 1.37 percent per year, respectively), given the assumptions of zero productivity growth in these sectors. Production of inland capture fish in 2030 is projected to be only 26.1 percent higher than in 2015, and marine fish production is only 22.6 percent higher than in 2015. With increased demand as population and incomes rise, real prices of these types of fish increase over time, by 5.29 and 4.63 percent per year, respectively (with cumulative real price increases of 116.7 and 97.3 percent). By contrast, aquaculture and mixed system production increase relatively rapidly over time, by 3.65 and 1.76 percent per year, respectively, so that prices of aquaculture actually fall and mixed system production rises only slightly—by –0.02 and 0.84 percent per year (with cumulative price effects of –0.3 and 13.4 percent by 2030).10 Fish consumption per capita rises by 27 percent overall, between 2015 and 2030, with particularly large increases for urban households (Table 6.7).

10 Earlier fish projections by World Bank (2007) for 2001 to 2015 assumed a growth rate of aquaculture of 10 percent per year, as compared with our calculated growth rate for this period of 8.6 percent and our base line growth rate of 3.65 percent for 2016 to 2030. Both analyses assume minimal growth of marine fish production. The two projections differ more substantially for inland capture, however. World Bank (2007) projects a growth rate of -7.0 percent for 2001 to 2015. Our calculations for the period show a positive (though somewhat low) growth rate of 0.8 percent, with a projected growth rate of 1.56 percent for 2016 to 2030 in the base simulation. 116 chapter 6

Table 6.5 Model simulation assumptions for population and income growth

Urban Rural Poor Nonpoor Poor Nonpoor Total Base Population 2010 (millions) 20.50 18.75 52.50 57.50 149.25 Population growth (%) 3.0 3.0 0.4 0.4 1.1 Income growth (%) 6.0 4.0 3.0 2.0 4.2 Per capita income growth (%) 2.9 1.0 2.6 1.6 3.1 High demand Population growth (%) 3.0 3.0 0.4 0.4 1.1 Income growth (%) 8.0 6.0 5.0 4.0 6.5 Per capita income growth (%) 4.9 2.9 4.6 3.6 5.4

Source: Authors’ calculations.

Table 6.6 Bangladesh fish production and prices: Simulation results

Annual growth rates: 2015–2030 Production Aquaculture Inland capture Mixed Marine Total fish Base 2015 level 1,654.8 467.5 1,022.2 496.5 3,641.0 Growth rates (%) Base 3.65 1.56 1.76 1.37 2.61 Sim 1 4.89 2.07 2.17 1.32 3.43 Sim 2 4.17 2.08 2.19 1.81 3.10 Sim 3 5.41 2.59 2.61 1.75 3.92 Sim 4 5.82 2.03 2.14 1.28 3.94 Sim 5 6.35 2.55 2.58 1.72 4.44 Price Aquaculture Inland capture Mixed Marine Average fish Base 2015 level 103.99 127.61 109.01 220.20 140.2 Growth rates (%) Base −0.02 5.29 0.84 4.63 3.8 Sim 1 −1.06 4.26 0.22 4.45 3.2 Sim 2 1.24 7.10 2.30 6.16 5.3 Sim 3 0.19 6.06 1.6 5.97 4.8 Sim 4 −1.97 4.12 0.12 4.34 3.0 Sim 5 −0.73 5.91 1.56 5.84 4.6

Source: Model simulations. Note: Sim 1: High productivity, all systems; Sim 2: Increased household fish demand; Sim 3: Sim 1 with increased household demand; Sim 4: Sim 1 with extra aquaculture productivity gains; Sim 5: Sim 4 with increased household demand. Future Scenarios (Projections to 2050) 117

In Simulation 1 we model increased productivity and output of fish. In particular, we increase productivity of aquaculture by 5 percent per year for five years (2016–2020) and then slow the rate of productivity growth to 3 percent per year for the remaining ten years of the simulation (2021–2030). For inland capture and mixed systems we assume constant annual increases in productivity, increases of 0.8 and 2.1 percent per year over all fifteen years of the simulation. We assume no change in the productivity of marine fish (see Table 6.4). Under these assumptions aquaculture production rises by 4.89 percent per year (120.50 percent over the 15-year period). Prices of aquaculture fish fall steeply, by 1.06 percent per year (14.70 percent by 2030). With faster growth in inland capture fish (2.07 percent per year, compared with 1.56 percent per year in the base simulation), prices rise by only 4.26 percent per year (86.9 percent by 2030), compared with an increase of 5.29 percent per year (116.70 percent by 2030) in the base simulation. Likewise, production of the mixed system rises by 2.17 percent per year (compared with 1.76 percent per year in the base simulation) and prices of mixed system fish only increase by 0.22 percent per year (3.30 percent by 2030). Total fish consumption per capita in 2030 is 56 percent higher than in 2015; 29 percent higher than in 2030 in the base simulation. Simulation 2 models faster household income growth that results in increased demand for fish, along with higher prices relative to the base simulation. Increased demand leads to moderate increases in prices and a corresponding supply response. Overall, fish production increases by 3.10 percent per year, compared with 2.61 percent per year in the base simulation. By 2030, increased demand results in an overall gain of 58 percent in fish production relative to the base simulation. Simulation 3 is a combination of the first two simulations (higher productivity growth along with increased fish demand). Overall, the effects of increased productivity and supply outweigh the impacts of increased demand. Overall fish production in 2030 is 17 percent higher than in the base simulation and only 5 percent lower than in the high-productivity scenario with no demand increase (Simulation 1). Simulation 4 models an additional increase in productivity growth and production of aquaculture, relative to the high-productivity scenario (Simulation 1). As a result, aquaculture production increases by 5.82 percent per year, compared with 3.65 percent per year in the base simulation. Prices of aquaculture fall by 1.97 percent per year, as compared with a decline of 0.02 percent per year in the base simulation, resulting in a decline of 118 chapter 6

Table 6.7 Bangladesh fish consumption: Simulation results (per capita consumption)

Aquaculture Inland capture Mixed Marine All fish Level (kilogram per capita) Base 2015 Rural nonpoor 12.09 2.29 5.72 1.72 21.81 Rural poor 7.46 1.88 4.33 0.89 14.56 Urban nonpoor 13.72 3.78 6.88 5.78 30.15 Urban poor 11.23 3.03 5.94 2.75 22.96 All Bangladesh 10.49 2.46 5.41 2.16 20.53 Percent change, 2015–2030 (%) Base Rural nonpoor −5.1 −35.4 −2.4 −26.8 −9 Rural poor 16.7 −24.3 9.5 −10.7 8 Urban nonpoor 69.1 22.8 10.4 −7.3 35 Urban poor 105.5 61.7 31.4 44.1 73 Total 42.2 9.1 12.7 10.3 27 Sim 3 Rural nonpoor 47.2 −15.8 34.3 −1.9 33 Rural poor 65.8 −3.3 37.1 14.6 45 Urban nonpoor 125.9 60.5 15.6 −1.0 68 Urban poor 89.2 55.8 16.1 26.9 58 Total 83.0 29.8 29.8 19.0 56 Sim 5 Rural nonpoor 67.7 −16.4 32.7 −4.6 44 Rural poor 87.1 −3.5 37.4 12.0 56 Urban nonpoor 159.9 59.3 14.6 −0.1 83 Urban poor 116.8 54.4 16.2 25.4 72 Total 108.9 28.9 29.2 18.1 69

Source: Model simulations. Note: Sim 1: High productivity, all systems. Sim 2: Increased household fish demand. Sim 3: Sim 1 with increased household demand. Sim 4: Sim 1 with extra aquaculture productivity gains. Sim 5: Sim 4 with increased household demand.

26 percent and 0.30 percent, respectively, in 2030. Finally, Simulation 5 combines the higher income growth of Simulation 2 with the accelerated productivity gains of Simulation 4. Again, increased demand results in a smaller price decrease in aquaculture (only 0.73 percent per year, as compared with 1.97 percent per year in Simulation 4). Figure 6.1 summarizes the results of base scenario, high-productivity/high- demand scenario (Simulation 3), and the highest-productivity aquaculture/ high-demand scenario (Simulation 5). As indicated, total fish production is Future Scenarios (Projections to 2050) 119

Figure 6.1 Simulation results: Bangladesh fish production (thousand metric tons)

8 Aquaculture Inland capture Mixed 7 Marine

5 to

3 o etrc

0 Base Base Simulation Simulation Base Simulation Simulation 3 5 3 5

Source: Model simulations. projected to grow rapidly if aquaculture investment and productivity continue to increase, potentially reaching 6.48 million metric tons in 2030 in Simulation 3 and 6.99 million metric tons in Simulation 5 (20.9 and 30.3 percent increases relative to the baseline figure). Annex II (Tables 6.A8 and 6.A9) presents sensitivity analysis using an alternative set of (more inelastic) demand parameters equal to 0.6 times those used in the main simulations. As expected, more inelastic demand implies larger price declines due to production increases (for example, Simulation 1a), but greater rises in the simulations involving exogenous increases in household incomes (for example, Simulation 2a). Aquaculture fish prices fall by 1.34 percent per year in Simulation 3a and 2.62 percent per year in Simulation 5a, compared with small yearly increases of 0.19 percent and decline of 0.73 percent per year in the corresponding simulations using the base parameters. More inelastic demand and resulting larger price declines also dampen production increases, so that overall fish production rises by only 3.41 percent and 3.81 percent per year in Simulations 3a and 5a. 120 chapter 6

Conclusion Fish production, particularly production from aquaculture, has increased rapidly in Bangladesh in the past decade. As a result, despite rising household incomes and consumer demand, real prices of fish produced from aquaculture systems have fallen. Moreover, prospects for future growth of household incomes are good, given public, private, and foreign investment in infrastructure. Improved infrastructure and changes in information and telecommunication and technology could further raise productivity and household incomes. Thus household demand for fish products will likely increase substantially if per capita incomes continue to rise as expected. Our analysis of potential increases in fish production (most of which would likely come from aquaculture) suggests that if present rates of investment in aquaculture and productivity growth continue, fish production growth is likely to outpace these increases in demand. In our base (moderate productivity growth) scenario, aquaculture and total fish production increase by an average of 3.65 percent and 2.61 percent per year from 2015 through 2030, contributing to a decline in real prices of aquaculture by 0.02 percent per year through 2030. Increases in aquaculture investment and productivity could lead to greater overall increase in production of as much as 120 percent in 2030 relative to 2015. If demand also increases rapidly, real aquaculture prices may fall by only 0.73 percent through 2030. Even greater aquaculture investments and larger productivity gains (by another 2 percent) could raise production to 6.99 million metric tons by 2030, 152 percent greater than in 2015 and a 69 percent increase in per capita consumption. Poor households who currently consume only small quantities of fish (14.56 kilograms per capita in 2015 for the rural poor as compared with 30.15 kilograms per capita for the urban nonpoor) stand to gain significantly from greater production and lower prices, with potential increases of 6.5 and 20.5 kilograms per capita, respectively, provided that improvements in storage, transport, and basic processing can increase their access to these increased supplies. These trends also imply significant changes in relative prices of various types of fish. Investments in aquaculture have the potential to lead to significant increases in production and possibly reduce real prices of aquaculture fish. In contrast, ecosystem constraints on sustainable fish production in inland capture and marine fishing are likely to limit production gains in these systems. All the model simulations indicate major shifts in consumption patterns and relative prices arising from these significant differences in production potential. Future Scenarios (Projections to 2050) 121

Thus inadequate demand is unlikely to be a major constraint on the Bangladesh fish sector, particularly for aquaculture. Although other production systems (inland capture and marine fishing) face serious constraints and are unlikely to increase their output significantly over time, the simulations suggest that sustained increases in aquaculture production at rates approxi- mating those of recent years would be sufficient to meet rising fish demand in Bangladesh. Moreover, increased supply could lead to moderate declines in the real price of fish products and enable poor Bangladeshi households to increase their consumption of this nutrient-rich food.

References BBS (Bangladesh Bureau of Statistics). 2000. Household Income and Expenditure Survey 2000. Dhaka.

—. 2005. Household Income and Expenditure Survey 2005. Dhaka. —. 2010. Household Income and Expenditure Survey 2010. Dhaka. —. 2016. Statistical Yearbook of Bangladesh 2015. Ministry of Planning. Dhaka. Braverman, A., and J. Hammer. 1986. “Multimarket analysis of agricultural pricing policies in Senegal.” In Agricultural Household Models: Extensions, Applications, and Policy, 233– 254. Washington, DC: World Bank.

Croppenstedt, A., L. G. Bellú, F. Bresciani, and S. DiGiuseppe. 2007. Agricultural Policy Impact Analysis with Multimarket Models: A Primer. ESA Working Paper 07–26. Rome: Food and Agriculture Organization of the United Nations.

Dervis, K., J. de Melo, and S. Robinson. 1982. General Equilibrium Models for Development Policy. World Bank Research Publication. Washington, DC: World Bank.

Dorosh, P. A. 2006. Accelerating Income Growth in Rural Bangladesh. Background Report to World Bank (2007). Washington, DC: World Bank. Ecker, O., and M. Qaim. 2010. “Analyzing Nutritional Impacts of Policies: An Empirical Study for Malawi.” World Development 39 (3): 412–428. Edgerton, D. L. 1997. “Weak Separability and the Estimation of Elasticities in Multistage Demand Systems.” American Journal of Agricultural Economics 79 (1): 62–79. FAO (Food and Agriculture Organization of the United Nations). 2014. Fishery and Aquaculture Country Profiles: Bangladesh. Country Profile Fact Sheets. Rome. FishBase. 2018. “List of Freshwater Fishes Reported from Bangladesh.” Accessed June 2018. www.fishbase.org. 122 chapter 6

FSR (Fisheries Sector Review). 2003. Fisheries Sector Review and Future Development: Theme Study: Economic Performance. Dhaka: World Bank, Danida (Danish International Development Assistance), USAID (United States Agency for International Development), FAO (Food and Agriculture Organization of the United Nations), and UK, DFID (United Kingdom, Department for International Development).

Heckman, J. 1979. “Sample Selection Bias As a Specification Error.” Econometrica 47 (1): 153–161. Lecoq, S., and J. M. Robin. 2015. “Estimating Almost-Idea Demand Systems with Endogenous Regressors.” The Stata Journal 15 (2): 554–557. Leser, C. E. V. 1963. “Forms of Engel Functions.” Econometrica 31 (4): 694–703. Poi, B. 2004. “From the Help Desk: Some Bootstrapping Techniques.” The Stata Journal 4 (3): 312–328.

—. 2012. “East Demand-System Estimation with QUAIDS.” The Stata Journal 12 (3): 433–446.

Portley, N. 2016. Report on the Shrimp Sector: Asian Shrimp Trade and Sustainability. Honolulu, HI, US: Sustainable Fisheries Partnership.

Sadoulet, E., and A. de Janvry. 1995. Quantitative Development Policy Analysis. Baltimore; London: Johns Hopkins University Press. Wikipedia. 2018. “List of Fishes in Bangladesh.” Wikimedia Foundation. Accessed August 2, 2018. www.wikipedia.org/wiki/List_of_fishes_in_Bangladesh.

Working, H. 1943. “Statistical Laws of Family Expenditure.” Journal of the American Statistical Association 38 (221): 43–56. World Bank. 2007. Bangladesh: Strategy for Sustained Growth. Bangladesh Development Series 18. Washington, DC.

Annex I: Alternate Demand Elasticity Estimates Given the importance of the household demand parameters for the model results, we present alternative estimates of these parameters based on econometric estimates using a Linear Expenditure System specification (Dervis, de Melo, and Robinson 1982). We first estimate expenditure elasticities from simple Engel functions using a Heckman two-step methodology (Heckman 1979; Leser 1963). Following this methodology, we correct for the possibility of zero consumption for each fish type by running a multivariate probit model (shown below). = 1 ( ) + 2( ) + dij β log Eij Σβ Demi eij Future Scenarios (Projections to 2050) 123

where i and j index households and commodity subgroups respectively, dij indicates whether a household consumed a certain commodity (with dij = 1 if household i consumed good j), Eij is household i’s expenditure on good j, and

Demi is a vector of household-level demographics that includes household size, age and sex of the head of household, square footage of the household, and district. From these probit models, the respective probability density and cumulative distribution functions were estimated and used to create the inverse mills ratio for each of the zero-observation commodities. We then correct for endogeneity of household expenditures using an instrumental regression to obtain a predicted value of total expenditures. The instrumental regression used was:

TEij = β1(lnYi) + eij where TEij and Yi are the total expenditure and real incomes of household h. Next, the instrumented total expenditures and inverse mills ratio were used in the following regression:

= 1 ( ) + 2 * ( )2 + 3 + 4( ) + whfood β ln Têh β ln Têh β lnPfood ∑β Demi 5( ) + + ∑β Districti λi eij where ln(Têh) is the log of the estimated total expenditure (on food and nonfood) from the instrumental regression, lnPfood is the natural log of the average price of food, Demi is the vector of household level (now excluding the district variable), Districti is a set of district dummies for all 94 districts, and λi is the inverse mills ratio. Finally, expenditure elasticities are calculated from these Engel functions using the following equation: = 1 + β1 ei wi where ei is the expenditure elasticity. And then own-price and cross-price elasticities are calculated using the Linear Expenditure System (LES) equations as such:

AvgP 1 = − ( i * TEi − ) eii ei wi φ P = − ( j * TEj) eij ei wi where eii is the own-price elasticity, eij is the cross-price elasticity, and φ is the Frisch parameter (Dervis, de Melo, and Robinson 1982). 124 chapter 6

The resulting elasticities are presented in Table 6.A5. We have done this exercise across the four fish categories used in the main analysis, but we have also done it between freshwater and marine fish categories for further exploration. The freshwater category was calculated by simply combining the three categories that are not marine. Comparing these results to the ones used in the model, we can see similarities in the direction of the signs and in the general magnitudes. For example, the own-price elasticities are all mostly negative and less than 1. Note that the magnitude of the cross-price elasticities are smaller than those estimated using the QUAIDS model (Table 6.A6). The small magnitudes of the cross-price elasticities may be due to aggregating various types of fish by production system rather than by characteristics reflecting consumer preferences. Further work on fish demand by fish type could shed light on these issues.

Annex II: Model Calibration and Validation After the model was constructed, calibration was employed to match the simulation productivity and price growth rates with the historical price and growth trends. Calibration targets were calculated using available BBS data on price and production for different fish types. The exogenous parameter we used in calibration was supply elasticity. Table 6.A7 shows the final results of the calibration, with the most reasonable price and production errors we arrived at given plausible supply elasticities. The highest overall error rates are those for price in aquaculture and production in mixed. It is interesting that we see too high of a price decrease in our model even though we also see too low of a production rate increase. More could be done to investigate this dichotomy. Validation was also undertaken to investigate how well the trends from the model matched the historical trends from the BBS data. Figure 6.A1 has the results of this. We graphed the BBS data from 2000 to 2015 and then over- played the graphs of our model results from 2010 to 2015. The model matched the results relatively well. Although the initial levels differ due to the model being based on the HIES data and the historical trends being from the BBS, the slopes tend to match well. For aquaculture we provide additional trend lines. The solid line represents production results using the demand parameters estimated from the QUAIDS model. The lines above and below this represent the results when we double the price elasticities for aquaculture and halve them, respectively. Both of these have a visible effect on the trend line, with halving the elasticities causing a more dramatic shift in production. In the end, we decided on the original parameters for the final estimates. Future Scenarios (Projections to 2050) 125

Annex III: Figures and Tables

Figure 6.A1 Model validation: Production by fish system (million metric tons)

1.8 Aquaculture Inland capture model 1.6 Inland capture Mixed model Mixed 1.4 Marine model Marine 1.2 Aquaculture model - High Aquaculture model 1.0 Aquaculture model - Low

.6 o etrc to .4

0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

Source: Model simulations. 126 chapter 6

Table 6.A1 Demographic variables used in the econometric estimation of household demand parameters, 2000, 2005, and 2010

2000 2005 2010

Variable Description Mean Std. dev. Min Max Mean Std. dev. Min Max Mean Std. dev. Min Max log real income Natural log of real income 11.40 0.90 4.94 15.68 10.97 0.88 5.99 16.25 10.84 0.81 5.39 16.38 sqft Square footage of residence 402 420 0 20,000 391 308 30 4,207 368 1,435 0 120,000 hhsize Continuous household size 4.5 1.9 1.0 17.0 4.9 2.1 1.0 20.0 5.2 2.2 1.0 25.0 sex Binary variable denoting sex of 0.86 0.35 0.00 1.00 0.90 0.30 0.00 1.00 0.09 0.29 0.00 1.00 head of household lnage Natural log of age of head of 3.78 0.30 2.40 4.80 3.77 0.30 2.48 4.60 3.75 0.30 2.48 4.60 household tasset Total household assets, in 43 280 −2,240 13,100 18 85 −2,800 4,094 25 257 −45 16,300 thousands

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: Std. dev. = standard deviation. Future Scenarios (Projections to 2050) 127

Table 6.A1 Demographic variables used in the econometric estimation of household demand parameters, 2000, 2005, and 2010

2000 2005 2010

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: Std. dev. = standard deviation. 128 chapter 6

Table 6.A2 Share and price variables: Descriptive statistics, 2000, 2005, and 2010

2000 2005 2010

Variable Description Mean Std. dev. Min Max Mean Std. dev. Min Max Mean Std. dev. Min Max w1 Consumption share of primarily 3 4 0 38 4 4 0 30 5 5 0 36 aquaculture (%) w2 Consumption share of mixed (%) 4 4 0 48 4 4 0 40 4 4 0 42 w3 Consumption share of primarily 4 4 0 30 3 3 0 32 2 3 0 34 inland capture (%) w4 Consumption share of primarily 2 4 0 39 2 4 0 39 3 4 0 53 marine (%) w5 Consumption share of food grains 45 15 0 87 44 15 0 100 40 15 0 86 and pulses (%) w6 Consumption share of dairy, eggs, 6 5 0 52 6 4 0 39 7 5 0 48 and meat (%) w7 Consumption share of vegetables 11 10 0 64 11 10 0 62 12 11 0 80 and fruit (%) w8 Consumption share of oils, fats, 14 5 0 44 13 5 0 48 13 5 0 61 and sugars (%) w9 Consumption share of misc. (%) 12 7 0 100 13 8 0 100 13 7 0 100 p1 Price of primarily aquaculture 55.9 19.9 6.6 300.0 58.1 16.6 8.0 200.0 103.4 25.1 40.0 300.0 p2 Price of mixed 47.7 19.0 10.0 400.0 56.3 20.9 12.0 200.0 109.9 33.0 40.0 500.0 p3 Price of primarily inland capture 48.3 20.3 7.1 328.0 62.5 22.7 4.0 750.0 126.8 42.9 48.0 800.0 p4 Price of primarily marine 80.9 41.0 10.9 2,000.0 103.1 34.2 10.0 800.0 220.4 84.8 40.0 1,000.0 p5 Price of food grains and pulses 12.5 1.8 5.5 36.0 17.4 1.9 3.6 29.0 33.3 4.5 20.4 76.0 p6 Price of dairy, eggs, and meat 26.9 10.0 2.4 103.6 38.0 11.9 2.5 512.5 66.9 27.6 11.1 283.1 p7 Price of vegetables and fruits 448.0 913.7 9.0 4,000.0 537.9 1,066.1 11.8 5,000.0 1,084.8 2,091.5 14.0 8,000.0 p8 Price of oils, fats, and sugars 8.6 3.0 2.5 38.5 9.7 3.7 3.0 81.1 18.0 7.9 4.2 88.7 p9 Price of misc. 86.6 78.6 7.9 5,457.5 87.0 40.0 18.3 500.0 132.3 66.2 32.8 1,108.8

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: Prices are directly from HIES and have not been adjusted for inflation. Future Scenarios (Projections to 2050) 129

Table 6.A2 Share and price variables: Descriptive statistics, 2000, 2005, and 2010

2000 2005 2010

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: Prices are directly from HIES and have not been adjusted for inflation. 130 chapter 6

Table 6.A3 Bangladesh fish exports, 2000–2014 (thousand metric tons)

Frozen Frozen Dry Salted Turtles/ Shark Year shrimp fish fish fish crab fish+others Total 2000 28.51 9.48 0.22 0.81 0.11 0.26 39.39 2001 29.71 7.97 0.14 0.84 0.15 0.18 38.99 2002 30.21 9.86 0.52 0.29 0.34 0.26 41.48 2003 36.86 8.85 0.33 0.53 0.63 0.17 47.37 2004 42.94 10.23 0.47 0.38 0.12 0.18 54.14 2005 46.53 15.76 0.27 0.77 0.04 0.17 63.38 2006 49.32 17.43 0.15 0.59 1.11 0.08 68.83 2007 53.36 18.38 0.08 0.44 1.12 0.24 73.70 2008 49.91 23.52 0.21 0.66 0.44 0.27 75.30 2009 50.37 19.29 0.34 0.08 1.22 0.28 72.89 2010 51.60 21.46 0.62 0.19 0.69 0.96 77.64 2011 54.89 16.74 0.62 0.58 4.49 1.78 96.47 2012 48.01 15.51 1.00 0.41 5.77 1.76 92.48 2013 50.33 11.44 1.28 0.54 7.43 1.60 84.91 2014 47.64 11.68 2.63 0.26 7.71 2.39 77.33

Source: BBS Statistical Yearbook (2016). Note: To compute production values for use in the model, processing loss was assumed to be 50 percent (Port- ley 2016). As such, the base levels of exports in the model are double what is presented in the table.

Table 6.A4 Scientific names for all fish by category

Category Fish name Scientific names Aquaculture Rui Labeo rohita Katla Catla catla Mrigel Cirrhinus cirrhosus Kalibaus Cyprinus calbasu Silver carp Hypophthalmichthys molitrix Grass carp Ctenopharyngodon idella Mirror carp Cyprinus carpio carpio Pangas Pangasius pangasius Boaal Wallago attu Air Bagarius bagarius Mixed Kai Xenentodon cancila Magur Clarias batrachus Shingi Heteropneustes fossilis Khalisha Tirchogaster chuna Future Scenarios (Projections to 2050) 131

Category Fish name Scientific names Mixed (cont.) Koi Anabas testudineus Mala–kachi Corica soborna Chala–chapila Gudusia chapra Puti Puntius sophore Big puti Barbonymus gonionotus Tilapia Oreochromis mossambicus Nilotica Oreochromis niloticus niloticus Inland capture Shoal Channa striata Gajar Channa marulius Taki Channa punctata Tangra Devario anomalus Eelfish Gong magor Balia Glossogobius giuris Tapashi Polynemus paradiseus Linnaeus Shrimp Macrobrachium rosenbergii Marine Ilish Tenualosa ilisha

Source: Wikipedia 2018; FishBase 2018.

Table 6.A5 Own-price elasticity estimates using the Linear Expenditure System (LES)

Own-price elasticity Category Rural poor Rural nonpoor Urban poor Urban nonpoor Primarily aquaculture −0.24 −0.42 −0.52 −0.08 Inland capture and culture −0.22 0.14 −0.38 −0.73 Primarily inland capture −0.06 −0.53 −0.03 −0.04 Primarily marine 0.08 −0.18 0.21 −1.60 Freshwater −0.29 −0.26 −0.22 −0.26 Marine 0.08 −0.18 0.21 −1.60

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. 132 chapter 6

Table 6.A6 Cross-price elasticity estimates using the Linear Expenditure System (LES)

Cross-price elasticities Primarily Primarily inland Primarily Wealth group aquaculture Mixed capture marine Primarily aquaculture Rural poor — −0.0023 −0.0013 −0.0021 Rural nonpoor — −0.0039 −0.0022 −0.0048 Urban poor — −0.0076 −0.0045 −0.0105 Urban nonpoor — −0.0011 −0.0007 −0.0025 Mixed Rural poor −0.0024 — −0.0012 −0.0019 Rural nonpoor 0.0018 — 0.0007 0.0015 Urban poor −0.0053 — −0.0033 −0.0076 Urban nonpoor −0.0106 — −0.0069 −0.0240 Primarily inland capture Rural poor −0.0006 −0.0005 — −0.0005 Rural nonpoor −0.0069 −0.0049 — −0.0060 Urban poor −0.0004 −0.0004 — −0.0006 Urban nonpoor −0.0006 −0.0005 — −0.0013 Primarily marine Rural poor 0.0009 0.0008 0.0004 — Rural nonpoor −0.0023 −0.0017 −0.0009 — Urban poor 0.0029 0.0030 0.0018 — Urban nonpoor −0.0229 −0.0218 −0.0148 — Freshwater Marine Freshwater Rural poor — −0.0025 Rural nonpoor — −0.0030 Urban poor — −0.0044 Urban nonpoor — −0.0083 Marine Rural poor 0.0020 — Rural nonpoor −0.0049 — Urban poor 0.0076 — Urban nonpoor −0.0605 —

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Note: — = not applicable. Future Scenarios (Projections to 2050) 133

Table 6.A6a Alternative specifications of expenditure elasticity parameters, 2000, 2005, and 2010

Iterated linear least-squares approach Rural Standard Rural Standard Urban Standard Urban Standard Poor error nonpoor error Poor error nonpoor error 2000 expenditure elasticities Primarily aquaculture 1.46 (0.109) 1.46 (0.080) 1.76 (0.296) 1.47 (0.187) Mixed 0.87 (0.059) 0.81 (0.072) 0.74 (0.198) 0.76 (0.244) Primarily inland capture 0.79 0.068) 0.06 (0.081) 0.45 (0.160) 0.29 (0.284) Primarily marine 0.97 (0.143) 1.23 (0.098) 1.00 (0.261) 1.16 (0.219) 2005 expenditure elasticities Primarily aquaculture 1.18 (0.055) 1.10 (0.043) 1.20 (0.074) 1.12 (0.063) Mixed 0.97 (0.045) 0.94 (0.052) 0.79 (0.067) 0.62 (0.088) Primarily inland capture 0.83 (0.064) 0.71 (0.079) 0.85 (0.097) 0.87 (0.104) Primarily marine 0.89 (0.105) 1.27 (0.099) 1.19 (0.133) 1.45 (0.105) 2010 expenditure elasticities Primarily aquaculture 1.07 (0.048) 1.02 (0.044) 1.07 (0.063) 0.99 (0.068) Mixed 0.89 (0.048) 0.83 (0.056) 0.87 (0.060) 0.77 (0.075) Primarily inland capture 0.94 (0.077) 0.90 (0.086) 1.15 (0.108) 0.99 (0.100) Primarily marine 1.13 (0.112) 1.37 (0.078) 0.96 (0.125) 1.31 (0.084) Alternative QUAIDS approach 2000 expenditure elasticities Primarily aquaculture 1.32 (0.031) 1.49 (0.039) 1.05 (0.056) −0.45 (1.005) Mixed 0.67 (0.029) 0.50 (0.033) 1.06 (0.04) 2.34 (0.68) Primarily inland capture 0.62 (0.065) 0.57 (0.039) 0.51 (0.11) −3.83 (3.661) Primarily marine 1.23 (0.025) 1.40 (0.039) 1.46 (0.15) 2.01 (0.412) 2005 expenditure elasticities Primarily aquaculture 1.20 (0.008) 1.15 (0.005) 1.00 (0.003) 0.98 (0.002) Mixed 0.79 (0.02) 0.70 (0.016) 0.70 (0.03) 0.54 (0.034) Primarily inland capture 0.91 (0.005) 0.94 (0.003) 1.04 (0.011) 1.10 (0.012) Primarily marine 1.26 (0.031) 1.35 (0.025) 1.53 (0.048) 1.52 (0.036) 2010 expenditure elasticities Primarily aquaculture 0.91 (0.007) 0.84 (0.008) 0.58 (0.023) 0.29 (0.034) Mixed 0.74 (0.019) 0.60 (0.019) 0.55 (0.033) 0.41 (0.03) Primarily inland capture 0.99 (0.003) 0.98 (0.004) 1.03 (0.01) 0.91 (0.009) Primarily marine 1.72 (0.048) 2.19 (0.086) 2.73 (0.134) 2.89 (0.166)

Source: Authors’ calculations from the BBS HIES 2000, 2005, and 2010. 134 chapter 6

Table 6.A6b Alternative specifications of price elasticity parameters, 2000, 2005, and 2010

Iterated linear least-squares approach Iterated linear least-squares approach 2000 price elasticities 2000 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −2.25 (0.15) −0.31 (0.12) −0.20 (0.11) 0.28 (0.08) −2.26 (0.37) −0.07 (0.27) −0.37 (0.31) 0.78 (0.23) (households in aquaculture bottom 40% of Mixed 0.02 (0.08) −0.95 (0.07) −0.24 (0.07) 0.20 (0.05) 0.13 (0.26) −0.76 (0.23) 0.09 (0.25) 0.10 (0.17) expenditures) Primarily 0.55 (0.09) 0.25 (0.08) −0.59 (0.08) −0.21 (0.06) 0.29 (0.22) 0.07 (0.19) −0.47 (0.20) −0.07 (0.14) inland capture Primarily 1.24 (0.19) −0.16 (0.18) 0.13 (0.16) −1.69 (0.14) 1.32 (0.33) −0.27 (0.29) −0.36 (0.33) −2.19 (0.24) marine Nonpoor Primarily −2.03 (0.10) −0.28 (0.10) −0.23 (0.09) 0.23 (0.07) −1.93 (0.24) −0.03 (0.22) −0.27 (0.24) 0.60 (0.17) (households in aquaculture top 60% of Mixed 0.01 (0.09) −0.93 (0.08) −0.25 (0.08) 0.24 (0.06) 0.12 (0.29) −0.74 (0.27) 0.09 (0.29) 0.12 (0.11) expenditures) Primarily 0.55 (0.10) 0.31 (0.09) −0.42 (0.09) −0.22 (0.07) 0.33 (0.29) 0.09 (0.26) −0.28 (0.30) −0.07 (0.19) inland capture Primarily 1.01 (0.16) −0.19 (0.14) −0.11 (0.13) −1.55 (0.10) 1.07 (0.31) −0.28 (0.25) −0.38 (0.27) −2.02 (0.20) marine 2005 price elasticities 2005 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −1.71 (0.08) −0.26 (0.06) 0.02 (0.05) 0.12 0.04 −1.55 (0.08) −0.24 (0.07) 0.04 (0.06) 0.45 (0.05) (households in aquaculture bottom 40% of Mixed 0 (0.06) −1.09 (0.05) −0.14 (0.04) 0.31 0.04 0.23 (0.07) −1.06 (0.06) −0.03 (0.05) 0.25 (0.04) expenditures) Primarily 0.75 (0.09) 0.34 (0.07) −1.02 (0.06) −0.37 0.05 0.41 (0.10) 0.16 (0.09) −1.22 (0.08) −0.33 (0.06) inland capture Primarily 0.64 (0.15) 0.38 (0.12) 0.42 (0.11) −1.61 0.09 0.24 (0.14) 0.47 (0.12) 0.28 (0.10) −2.13 (0.09) marine Nonpoor Primarily −1.57 (0.06) −0.19 (0.05) 0.04 (0.04) 0.10 0.04 −1.47 (0.07) −0.19 (0.06) 0.05 (0.05) 0.40 (0.04) (households in aquaculture top 60% of Mixed 0 (0.07) −1.10 (0.06) −0.15 (0.05) 0.36 0.04 0.33 (0.09) −1.01 (0.08) −0.01 (0.07) 0.31 (0.05) expenditures) Primarily 0.89 (0.12) 0.42 (0.09) −0.98 (0.08) −0.43 0.06 0.43 (0.11) 0.16 (0.09) −1.23 (0.08) −0.36 (0.06) inland capture Primarily 0.57 (0.16) 0.24 (0.12) 0.23 (0.10) −1.61 0.08 0.10 (0.12) 0.26 (0.10) 0.18 (0.09) −1.95 (0.08) marine Future Scenarios (Projections to 2050) 135

Table 6.A6b Alternative specifications of price elasticity parameters, 2000, 2005, and 2010

Table 6.A6b Continued

Iterated linear least-squares approach Iterated linear least-squares approach 2010 price elasticities 2010 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −1.69 (0.08) −0.15 (0.06) 0.24 (0.05) 0.25 0.03 −1.17 (0.08) −0.31 (0.07) −0.05 (0.05) 0.10 (0.04) (households in aquaculture bottom 40% of Mixed 0.2 (0.08) −0.97 (0.06) −0.13 (0.05) 0.11 0.04 −0.12 (0.08) −0.80 (0.07) 0.04 (0.05) 0.13 (0.04) expenditures) Primarily 0.61 (0.12) 0.28 (0.09) −1.79 (0.08) −0.39 0.06 −0.14 (0.14) 0.02 (0.12) −1.50 (0.09) −0.23 (0.06) inland capture Primarily 0.92 (0.16) 0.01 (0.13) 0.71 (0.12) −1.60 0.09 0.77 (0.16) 0.29 (0.14) 0.51 (0.11) −1.27 (0.08) marine Nonpoor Primarily −1.64 (0.07) −0.13 (0.05) 0.23 (0.05) 0.25 0.03 −1.15 (0.09) −0.29 (0.07) −0.04 (0.06) 0.11 (0.04) (households in aquaculture top 60% of Mixed 0.24 (0.09) −0.94 (0.07) −0.14 (0.06) 0.13 0.04 −0.12 (0.09) −0.74 (0.08) 0.06 (0.06) 0.16 (0.04) expenditures) Primarily 0.68 (0.14) 0.32 (0.11) −1.86 (0.09) −0.43 0.06 −0.08 (0.13) 0.09 (0.11) −1.46 (0.09) −0.21 (0.06) inland capture Primarily 0.62 (0.14) −0.10 (0.10) 0.50 (0.09) −1.49 0.06 0.43 (0.12) 0.06 (0.10) 0.30 (0.08) −1.22 (0.05) marine Alternative QUAIDS approach Alternative QUAIDS approach 2000 price elasticities 2000 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −1.84 (0.06) −0.10 (0.03) 0.03 (0.03) 0.58 (0.04) −2.37 (0.23) −0.07 (0.01) −0.05 (0.01) 1.43 (0.25) (households in aquaculture bottom 40% of Mixed 0.1 (0.05) −0.89 (0.03) −0.10 (0.02) 0.22 (0.02) −0.08 (0.02) −0.78 (0.05) −0.23 (0.04) 0.02 (0.01) expenditures) Primarily 0.48 (0.14) −0.33 (0.08) −0.10 (0.07) −0.68 (0.05) 0.14 (0.03) −0.15 (0.04) −0.26 (0.17) −0.24 (0.06) inland capture Primarily 1.46 (0.17) 0.27 (0.02) −0.97 (0.12) −1.99 (0.09) 2.40 (0.51) −0.03 (0.03) −0.63 (0.14) −3.19 (0.55) marine Nonpoor Primarily −1.65 (0.04) −0.36 (0.04) −0.23 (0.03) 0.75 (0.05) −1.44 (0.46) 0.28 (0.23) 0.22 (0.17) 1.39 (0.25) (households in aquaculture top 60% of Mixed −0.1 (0.02) −0.69 (0.03) 0.08 (0.02) 0.21 (0.01) −0.58 (0.32) −1.02 (0.16) −0.49 (0.12) −0.25 (0.12) expenditures) Primarily 0.11 (0.06) −0.03 (0.04) 0.04 (0.07) −0.68 (0.04) 1.99 (1.73) 0.72 (0.82) 0.70 (0.68) 0.41 (0.59) inland capture Primarily 2.02 (0.18) 0.22 (0.02) −1.39 (0.12) −2.25 (0.10) 2.02 (0.70) −0.16 (0.08) −0.70 (0.23) −3.17 (0.71) marine Future Scenarios (Projections to 2050) 137

Table 6.A6b Continued

Alternative QUAIDS approach Alternative QUAIDS approach 2005 price elasticities 2005 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −1.73 (0.03) −0.17 (0.01) 0.34 (0.01) 0.36 (0.02) −1.73 (0.03) −0.08 (0.01) 0.30 (0.01) 0.50 (0.02) (households in aquaculture bottom 40% of Mixed −0.02 (0.01) −1.05 (0.03) 0.01 (0.01) 0.28 (0.02) 0.00 (0.01) −1.13 (0.05) 0.01 (0.02) 0.41 (0.04) expenditures) Primarily 0.62 (0.03) −0.01 (0.01) −1.14 (0.01) −0.37 (0.02) 0.52 (0.03) −0.08 (0.03) −1.33 (0.02) −0.15 (0.02) inland capture Primarily 1.03 (0.07) 0.62 (0.04) −0.75 (0.05) −2.16 (0.08) 1.19 (0.11) 0.57 (0.07) −0.33 (0.03) −2.96 (0.16) marine Nonpoor Primarily −1.58 (0.02) −0.22 (0.01) 0.30 (0.01) 0.35 (0.01) −1.69 (0.03) −0.18 (0.01) 0.35 (0.01) 0.55 (0.02) (households in aquaculture top 60% of Mixed −0.1 (0.01) −0.84 (0.02) 0.00 (0.00) 0.23 (0.01) −0.11 (0.01) −0.69 (0.05) −0.12 (0.02) 0.38 (0.04) expenditures) Primarily 0.75 (0.03) −0.07 (0.01) −1.18 (0.01) −0.45 (0.02) 0.61 (0.03) −0.28 (0.02) −1.31 (0.02) −0.12 (0.02) inland capture Primarily 1.1 (0.08) 0.42 (0.04) −0.75 (0.05) −2.11 (0.06) 0.92 (0.07) 0.34 (0.04) −0.25 (0.02) −2.53 (0.09) marine 2010 price elasticities 2010 price elasticities Rural Urban Primarily Primarily Primarily Std. Std. inland Std. Primarily Std. Primarily Std. Std. inland Std. Primarily Std. aquaculture error Mixed error capture error marine error aquaculture error Mixed error capture error marine error Poor Primarily −1.92 (0.04) 0.05 (0.00) 0.58 (0.03) 0.39 (0.02) −0.81 (0.01) −0.21 (0.01) 0.20 (0.01) 0.24 (0.02) (households in aquaculture bottom 40% of Mixed 0.12 (0.01) −1.03 (0.00) 0.10 (0.01) 0.06 (0.00) −0.24 (0.01) −0.63 (0.02) 0.16 (0.01) 0.15 (0.01) expenditures) Primarily 1.06 (0.05) 0.10 (0.01) −2.00 (0.04) −0.14 (0.01) 0.26 (0.02) 0.14 (0.01) −1.42 (0.02) 0.00 (0.01) inland capture Primarily 0.56 (0.03) −0.16 (0.01) −0.28 (0.01) −1.84 (0.04) −0.17 (0.05) −0.33 (0.03) −0.23 (0.02) −2.00 (0.11) marine Nonpoor Primarily −1.86 (0.03) 0.06 (0.00) 0.57 (0.02) 0.39 (0.01) −0.65 (0.02) 0.01 (0.29) 0.01 (0.25) 0.02 (0.02) (households in aquaculture top 60% of Mixed 0.19 (0.01) −0.98 (0.00) 0.14 (0.01) 0.06 (0.00) −0.24 (0.01) −0.52 (0.02) 0.22 (0.01) 0.14 (0.01) expenditures) Primarily 1.17 (0.05) 0.12 (0.01) −2.11 (0.05) −0.17 (0.01) 0.35 (0.02) 0.20 (0.01) −1.44 (0.02) −0.02 (0.00) inland capture Primarily 0.56 (0.04) −0.35 (0.03) −0.41 (0.03) −2.00 (0.07) −0.48 (0.05) −0.49 (0.05) −0.33 (0.03) −1.60 (0.07) marine

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. Future Scenarios (Projections to 2050) 139

Source: Authors’ calculations from BBS HIES 2000, 2005, and 2010. 140 chapter 6

Table 6.A7 Model calibration

Price Current Historical Current Sim target sim price production sim prod. Current Current price Prod. growth growth growth rate growth prod. elast. of change Price error rate (%) rate (%) (%) rate (%) shock supply (%) error (%) (%) Aquaculture −2.00 −5.24 8.61 6.72 0.08 0.4 −5.28 −3.31 −1.74 Inland capture −2.00 −1.49 2.13 1.97 0.021 0.3 −1.52 0.52 −0.15 Mixed −2.00 −3.19 −0.12 4.34 0.08 0.3 −3.19 −1.21 4.47 Marine 0.30 −0.30 3.01 1.10 0.03 0.3 −0.30 −0.60 −1.85

Source: Authors’ calculations from simulations. Note: All growth rates and changes are from 2010 to 2015.

Table 6.A8 Bangladesh fish production and prices: Simulation results (alternative parameters)

Annual growth rates (%): 2015–2030 Production Base 2015 level Aquaculture Inland capture Mixed Marine Total fish (million metric tons) 1.5 0.5 1.0 0.5 3.5 Growth rates (%) Base 3.23 1.39 1.63 1.31 2.31 Sim 1 4.27 1.80 1.96 1.25 2.96 Sim 2 3.72 1.83 2.01 1.68 2.75 Sim 3 4.76 2.24 2.34 1.62 3.41 Sim 4 5.04 1.75 1.93 1.21 3.36 Sim 5 5.53 2.19 2.30 1.58 3.81 Price Base 2015 level Aquaculture Inland capture Mixed Marine Average fish (Bangladeshi taka per kilogram) 103.99 127.61 109.01 220.20 140.2 Growth rates (%) Base −1.04 4.71 0.43 4.43 3.4 Sim 1 −2.51 3.33 −0.44 4.23 2.8 Sim 2 0.14 6.23 1.68 5.71 4.8 Sim 3 −1.34 4.84 0.79 5.50 4.1 Sim 4 −3.77 3.18 −0.56 4.10 2.6 Sim 5 −2.62 4.68 0.67 5.36 3.9

Table 6.A9 Bangladesh fish consumption: Simulation results (alternative parameters)

Per capita consumption Level (kilograms per capita) Aquaculture Inland capture Mixed Marine All fish Base 2015 Rural nonpoor 11.67 2.36 5.72 1.77 21.52 Rural poor 7.03 1.90 4.25 0.89 14.07 Urban nonpoor 12.30 3.61 6.77 5.72 28.40 Urban poor 9.77 2.78 5.69 2.59 20.83 All Bangladesh 9.78 2.43 5.33 2.15 19.69 Percent change, 2015–2030 (%) Base Rural nonpoor 6.7 −21.3 0.3 −18.4 0 Rural poor 19.9 −13.4 8.4 −8.1 10 Urban nonpoor 51.4 15.5 8.2 −3.4 25 Urban poor 69.9 36.0 21.0 23.4 46 Total 33.7 6.0 10.4 9.1 21 Sim 3 Rural nonpoor 47.5 −4.6 25.2 −2.6 32 Rural poor 56.4 3.6 28.6 6.6 38 Urban nonpoor 90.7 39.7 14.7 2.2 48 Urban poor 71.4 37.0 16.5 14.5 45 Total 66.8 22.4 24.4 15.9 44 Sim 5 Rural nonpoor 64.4 −5.4 23.7 −5.2 40 Rural poor 72.9 3.2 28.6 4.3 46 Urban nonpoor 114.4 38.6 13.7 2.7 58 Urban poor 92.0 35.6 16.3 13.1 54 Total 86.2 21.5 23.6 15.0 53

Table 6.A10 Bangladesh fish production projections to 2015

Growth rates (%) 1995/1996 2000/2001 2014/2015 1996–2001 2001–2015 Inland capture 1,366 950 344 −7.00 −7.00 Inland culture 317 651 2,471 15.50 10.00 Marine 628 597 519 −1.00 −1.00 Total 2,311 2,198 3,334 −1.00 3.00

Source: FSR (2003), Dorosh (2006), and authors’ calculations. Note: Projections for 2014/2015 assume no change in real prices of fish. Inland culture is Department of Fisheries data. Inland capture is residual of Household Expenditure Survey (HES) data of total fish less DoF inland culture. Marine is HES consumption figure for dried fish (10:1 fresh-to-dry conversion). 1995/1996 and 2000/2001 are historical data. 2014/2015 is a projection. Chapter 7

SUMMARY AND IMPLICATIONS

Shahidur Rashid and Xiaobo Zhang

Introduction Led by aquaculture, the fishery sector in Bangladesh has been remarkably successful in rapidly increasing production, reducing prices, and meeting rising domestic demand. The trend has defied many earlier predictions, and the success clearly deserves to be labeled a Blue Revolution. In the early 1990s, when the country was celebrating the success of the Green Revolution, per capita annual fish consumption was only 10 kilograms, with widespread concerns that consumption could decline even further because of rising prices (Bouis and Haddad 1992). The policy ambition was not high even in the early 2000s. In 2005 a Food and Agriculture Organization (FAO) report argued that reaching per capita consumption of 18 kilograms per year would be a big accomplishment. The country far exceeded that target by 2010; and according to the latest estimates, per capita fish consumption in Bangladesh reached 23 kilograms per year in 2016 (BBS 2017). This book has attempted to understand the enablers, impacts, and prospects of this unprecedented growth. There are three important motivations for undertaking the research presented in this book. First, there is a limited set of robust empirical analysis on aquaculture value chains through systematic data gathering. Most of the existing studies focus on exportable fish (that is, mainly shrimp and marine fish) and are generally based on small nonrepresentative surveys, which are inadequate for analyzing internal mechanics of value chain transformation. Furthermore, the previous research mainly stressed subsistence fish farming, paying little attention to the emergence of commercial value chains. Thus an important contribution of the book is the generation and analysis of a unique dataset, which we hope will help relevant government agencies gain better insights and design better survey instruments. Second, although several earlier studies looked at the negative poverty and income distributional impacts of promoting aquaculture (Minkin and Boyce 1994; Ahmed and Lorica 2002; Sachs 2007), to the best of our knowledge, there have been no robust analyses on the welfare implications of promoting aquaculture. This book has

143 144 chapter 7

presented such evidence with simulation exercises using several rounds of nationally representative household surveys. Third, with domestic consumption reaching its historic high, there are uncertainties about the prospects of future growth. This book has presented midterm demand-and-supply scenarios for the Bangladesh fisheries sector, based on a multimarket model.

Enablers of the Blue Revolution Understanding the enablers of the recent growth has relied on a specially designed survey, with samples involving all key actors in the fish value chain. A community-level survey was also conducted to gain perspective about the transformation in the fish value chain. The results from these surveys, presented in Chapter 3 and Chapter 4, offer some useful insights into the various drivers of the Blue Revolution. Clearly the aquaculture value chain has transformed. There were several enabling factors—such as the availability of technology, improved infrastructure, demonstration and extension, and youth training—that helped trigger the revolution. However, a central mes- sage from this set of analyses is that the demand in domestic markets, fueled by years of sustained economic growth, has been the key driver of the transformation in the fish value chain in Bangladesh. More than 90 percent of farmed fish (excluding shrimp) are consumed domestically. Based on the primary fish value chain survey and other secondary data sources, the book shows that the Blue Revolution has happened in Bangladesh, with the farmed fish market growing by a factor of 25 times in three decades. The enlarging market size due to rising demand and decreasing transportation costs has driven the specialization and formation of fish clusters at various geographic locations. As the chapters have argued, the transformation in the value chain is marked with three defining features: growth, intensification, and clustering. The growth in the sector is extensive across all the segments of the fish value chain. As shown in Chapter 3, the number of actors along the fish value chain have witnessed phenomenal growth. Based on the fish value chain survey, the number of hatcheries, feed mills, feed dealers, and fish traders more than doubled during 2004–2014, while the number of fish farmers grew by 63 percent over that period. At the same time, market efficiency has gone up as output dealt in per trader (or production per farmer) has increased in recent decades. The intensification of the sector has been rapid, as fish farmers are using more hatchery-produced seed, purchasing more floating formulated feed, applying more chemicals, hiring more labor, and investing more in quasi- fixed capital. Such intensification has led to the diversification of upstream Summary and Implications 145

value chain segments, specifically feed and seed, which are now specifying their products based on the different fish species and even the varying characteristics within species. Finally, fish production has become increasingly clustered, as shown in Chapter 4. In the fish production areas, fish pond areas, the number of fish traders, fish farmers, and feed dealers have simultaneously increased. The proximity of these actors to one another is a key feature of fish production clusters. Fish traders are acting more frequently as commission agents, broker- ing deals between buyers and sellers. Feed dealers bring in feed from large and medium-sized feed mills, which are not necessarily near the fish farming areas. The demand for feed has induced hatcheries to expand beyond only the North region. Clustering has created many positive externalities for the fish value chain actors. With easier access to markets, fish farmers in highly clustered regions use more modern inputs, face fewer difficulties in marketing, and cooperate more among the actors in the value chain. Clustering has also brought markets closer to the farmers, with numerous traders competing for market shares. Moreover, actors in the same segment of fish value chains are more likely to collaborate in highly clustered regions than less clustered areas. They often share market information, tools, and labor. The positive externalities in clusters have offset the disadvantages of intense competition.

Poverty and Welfare Implications Many earlier studies highlighted the income distribution implications of promoting aquaculture (Lewis 1997; Ahmed and Lorica 2002; ADB 2005). These concerns seemed valid as rural Bangladesh is characterized by marked inequalities regarding access to land, with more than half of households being functionally landless. Since aquaculture requires land and capital, a common concern was that the poor could neither produce nor afford to consume fish. Thus, the argument goes, the promotion of aquaculture would lead to increased inequality (Ahmed, Rab, and Bimbao 1995). Another reason was that until the early 1990s, aquaculture in Bangladesh largely referred to shrimp culture, which was dominated by wealthy landowners and generated little benefits for the poor. Our analysis clearly shows a large positive welfare effect of the growth in aquaculture for all income groups. The economic pathways for overall welfare improvement are easy to understand: fish supply has increased relatively more than the overall demand (the supply curve shifted relatively more toward the right), resulting in lower 146 chapter 7

prices. Improvement in the value chain as a result of reductions in transaction costs has also helped to lower prices. Reductions in prices helped to improve consumers’ budgets through both income effects and substitution effects. The quantitative assessment of these impacts—presented in Chapter 5— shows that increases in fish consumption have been more equitable than expected. Fish consumption grew across the lines of gender, region, and income quintile, with poorer households experiencing higher growth than other groups. The welfare analysis in Chapter 5 shows that in the short-run all household categories (except for fish-farming households) benefit from the decline in fish price, but the magnitude of losses for fish-farming households is small and is outweighed by the gains by most of the households (note that the short- run analysis only considers price changes on consumption and not on production). In the long-run, when change in production is also considered, fish farmers gain the most, although all households gain to some degree. Overall estimates of aquaculture’s contribution to income growth between 2000 and 2010 range from about 1 percentage point in the short run (only price effects) to more than 2 percentage points in the long run (price and quantity effects). The corresponding estimates for poverty reduction are 0.70 and 1.74 percentage points, respectively. Between 2000 and 2010 the poverty headcount in Bangladesh declined by 17.4 percentage points (from 48.9 to 31.5 percent). Therefore aquaculture’s contribution to overall poverty reduction during this period was about 10 percent, significantly higher than the sector’s contribution to GDP. In summary, the rapid growth in aquaculture has helped reduce poverty.

Demand and Supply Projection to 2030 The analysis of the enablers and impacts of a Blue Revolution in Bangladesh presented in Chapter 6 describes quite a positive story. However, the sector also faces challenges in future growth, as the per capita consumption has reached a historic high. A critical question is whether the country’s growth in aquaculture can continue to be fueled by the domestic demand, as has been the case over the past two decades. Chapter 6 addresses this issue by generating future scenarios with a multimarket model. The analysis of future increases in fish production, mostly coming from aquaculture, shows that fish production growth is likely to outpace the increases in demand between 2015 and 2030 with a moderate decline in prices. Increases in aquaculture investment and productivity could lead to greater overall increases in production Summary and Implications 147

of as much as 120 percent in 2030 relative to 2015. If demand also increases rapidly, real aquaculture prices may fall by only 0.73 percent through 2030. Even greater aquaculture investments and larger productivity gains (by another 2 percent) could raise production to 6,986 million metric tons by 2030, 152 percent greater than in 2015 and a 69 percent increase in per capita consumption. Chapter 6 also demonstrates that poor households, currently consum- ing small quantities, will gain significantly from greater production and lower prices. However, growth in productivity will require targeted investments, especially since inland capture and marine fishing face serious ecosystem constraints.

Conclusion and Future Research Despite rather limited interventions from the government, the aquaculture- led Blue Revolution has taken off in Bangladesh. One key reason is that the aquaculture sector fits well with the comparative advantage of Bangladesh— numerous ponds and abundant labor. However, further promotion of the sector will be complex, due to both known and unknown risks associated with human health, biodiversity, and environmental consequences. This book has not covered many of those issues. Yet the book points to several impor tant implications for policies and future research. These implications can be broadly grouped into productivity enhancement, market development, and institutional capacity. As is obvious from the value chain analysis, the transformation of the nonshrimp aquaculture value chain has been largely driven by the domestic markets in a broadly unregulated environment. The projection analysis suggests that at reasonable growth rates and adequate investment, the domestic market can continue to absorb the increased production. However, as Chapter 2 discussed, the potential for productivity growth is much higher. To be specific, per hectare productivity of shrimp in Bangladesh is less than a third of both Thailand and Vietnam. The productivity of striped catfish under an intensive cultivation system is even lower: 60–70 metric tons per hectare in Bangladesh relative to 300 or more metric tons per hectare in Vietnam (Phan et al. 2009). Overall, the country’s aquaculture productivity was only about 4.26 metric tons per hectare in 2014 (a total production of 1.61 million metric tons in 377,968 hectares of pond land). Even if half of the pond area can be converted into an intensive system, with a productivity of 100 metric tons per hectare, there will be more than a 12-fold increase in aquaculture production. 148 chapter 7

The large gap in productivity with other East Asian countries implies great growth potentials in the aquaculture sector in Bangladesh. Yet two issues follow from the discussion about productivity potentials. The first is whether Bangladesh can achieve the productivity level of Thailand and Vietnam. In demonstration ponds, WorldFish has shown that it is possible to achieve up to 200 metric tons per hectare of striped catfish in Bangladesh using existing technologies, implying that the technology is not a binding constraint. Therefore it is important to understand the other factors that constrain farmers achieving high productivity. According to a recent survey on problems and challenges faced by the aquaculture farmers (Bangladesh Shrimp and Fish Foundation 2017), among the top five answers, three are related to feed: specifically the high price of fish feed, difficulty in getting fish feed on time, and poor quality of fish feed. The problems with fish feed are likely to be a major factor behind the low productivity level in Bangladesh. More research is needed to establish the causal link. Even with current low aquaculture productivity, the domestic fish price is already low. In fact, 65 percent of aquaculture farmers report “not getting the right price of fish” as a key challenge for them (Bangladesh Shrimp and Fish Foundation 2017). Supposing the country can achieve the potential productivity targets in a cost-effective way, the next challenge will be to find markets outside its borders. Even though Bangladesh has been successful in exporting shrimp to industrialized countries, accessing international markets with other aquaculture fish might not be easy. Unlike shrimp, which is dominated by large entrepreneurs, the other aquaculture fish are farmed by smallholders. Therefore it will be difficult for them to meet the international safety and cer- tification, especially since nonshrimp pond culture has grown as a response to domestic markets, where food safety standards are not enforced. As a result, promoting export beyond niche markets will require setting up new institutional and regulatory frameworks. This constitutes an important area of future research. The issues related to governance and institutional capacity building are of great importance. Although the country had a wide range of ordinances, rules, and acts, an integrated National Fisheries Policy (NFP) was adopted only in 1998. A separate strategy document, called the National Fisheries Strategy (NFS), was released in 2006. Two points are worth highlighting. The first is that, although some of the provisions are relevant to the subsector, there is no separate section on aquaculture in the NFP or the NFS. Second, these policy documents highlight a long list of objectives, ranging from promoting economic growth to reversing environmental imbalance. The NFP extends Summary and Implications 149

to all government organizations involved in fisheries (and to all water bodies used for fisheries), but it is unclear how each of these entities would effectively coordinate. This is perhaps the reason why the Department of Fisheries (Bangladesh, DoF 2015) points out “poor institutional linkages” as one of the key challenges to the sector. The need of institutional capacity building within the relevant public agencies is evident from the review and experience of conducting the surveys. The weaknesses in regulatory capacity is clear by the fact that some of the key regulations related to seed, feed, and other inputs had not been formulated until 2010/2011. Both fish feed and hatchery regulations were passed in 2011—more than a decade after the aquaculture growth picked up. The need for data-management capacity development is also clear. For the researchers of this book project, reconciling even basic data, such as production and consumption, proved difficult. Without understanding the true picture of aquaculture development, it is hard for policymakers to make sound policy decisions. In fact, a 2016 DoF presentation at the FAO’s 26th session of the Asia Pacific Commission highlights some of the challenges that are consistent with our experiences. For example, the DoF used an old survey framework (1983/1984), recognized that it lacks human capacity, and acknowledged that it needs a comprehensive framework for inland fisheries resources. The main research chapters in this book have pointed out areas of critical future research. In addition to many institutional issues pointed out earlier, the book has not addressed the “utilization” aspect of food security. In other words, although the book has investigated the economics of the value chain and welfare, it has not gone into the nutritional issues, as conceptualized in a “food system” framework. Future studies on assessing the impacts and iden- tifying the bottlenecks in improving households’ nutritional outcomes will shed important light on broader food security implications. Another topic to study is the impact of clustering on productivity and quality. Chapter 4 discussed how the fish value chain actors in the clusters choose to work together rather than to work in stiff competition with one another; however, the chapter did not empirically show how the clusters have affected the final products on the fish markets, both in terms of quantity and quality. Finally, this body of literature would benefit from additional research on overcoming the barriers—for example, access to finance and risks of adopting technology—that the smallholders face in switching to a more intensive form of fish farming, which would subsequently increase production. And to complement that, studying the dynamics and impacts of extensive versus intensive fish farming in Bangladesh would also be beneficial. 150 chapter 7

While Bangladesh has not reached the productivity levels of East Asian countries, recent growth in aquaculture has defied many earlier predictions and deserves to be termed a Blue Revolution. Research presented throughout this book has provided important insights into this recent phenomenon, including its determinants, structure, and current and projected impacts. Bangladesh is a unique country with regards to its two endowments: labor and water. Just as the availability of cheap labor fueled the garment export boom (serving as an important push toward structural transformation), so can use of the country’s water resources be effective in triggering multiplier effects. Thus studying the growth in this sector provides a unique perspective on how all these dynamics (and more) interact. The hope is that this book will provide a basis on which to improve the aquaculture policymaking processes in Bangladesh so the two most important resources of the country—its people and water—can serve each other in a sustainable way.

References ADB (Asian Development Bank). 2005. An Evaluation of Small-Scale Freshwater Rural Aquaculture Development for Poverty Reduction. Manila: Operations Evaluation Department. Ahmed, M., and M. H. Lorica. 2002. “Improving Developing Country Food Security through Aquaculture Development—Lessons from Asia.” Food Policy 27 (2): 125–141. Ahmed, M., M. A. Rab, and M. A. P. Bimbao. 1995. Aquaculture Technology Adoption in Kapasia Thana, Bangladesh: Some Preliminary Results from Farm Record-Keeping Data. Manila: WorldFish.

Bangladesh, DoF (Department of Fisheries). 2015. Fisheries Statistical Year Book of Bangladesh 2013–2014. Fisheries Resource Survey System, DoF. Dhaka: Ministry of Fisheries and Livestock.

Bangladesh Shrimp and Fish Foundation. 2017. Comprehensive Study on the Aquaculture Sector in Bangladesh with Particular Focus on Existing Policy Support Received by the Sector and Scope for Improvements. Dhaka. BBS (Bangladesh Bureau of Statistics). 2017. Preliminary Report on the Household Income and Expenditure Survey of 2016. Dhaka: Ministry of Planning, Government of the People’s Republic of Bangladesh. Bouis, H. E., and L. J. Haddad. 1992. “Are Estimates of Calorie-Income Elasticities Too High? A Recalibration of the Plausible Range.” Journal of Development Economics 39: 333–364. Summary and Implications 151

FAO (Food and Agriculture Organization of the United Nations). 2005. Bangladesh National Aquaculture Sector Overview. Accessed December 11, 2017. www.fao.org/fishery/country sector/naso_bangladesh/en. Lewis, D. 1997. “Rethinking Aquaculture for Resource-Poor Farmers: Perspectives from Bangladesh.” Food Policy 22 (6): 533–546. Minkin, S. F., and J. K. Boyce. 1994. “Net Losses: ‘Development’ Drains the Fisheries of Bangladesh.” Amicus Journal 16 (3): 36–40. Phan, L. T., T. M. Bui, T. T. T. Nguyen, G. J. Gooley, B. A. Ingram, H. V. Nguyen, P. T. Nguyen, and S. S. De Silva. 2009. “Current Status of Farming Practices of Striped Catfish, Pangasianodon hypophthalmus in the Mekong Delta, Vietnam.” Aquaculture 296 (3): 227–236. Sachs, J. D. 2007. “The Promise of the Blue Revolution: Aquaculture Can Maintain Living Standards While Averting the Ruin of the Oceans.” Scientific American 297 (1): 37–38.

AUTHORS

Kaikaus Ahmad is the senior secretary of the Power Division, Ministry of Power, Energy, and Mineral Resources, Government of Bangladesh. At the time he contributed to this work, he was an associate research fellow in the Development Strategy and Governance Division of the International Food Policy Research Institute (IFPRI), Washington, DC. Akhter Ahmed is a senior research fellow in the Poverty, Health, and Nutrition Division, and country representative to Bangladesh at IFPRI, Dhaka. Ben Belton is an assistant professor in the Department of Agricultural, Food, and Resource Economics at Michigan State University, East Lansing, US. Qingqing Chen is an applied economics doctoral student at the University of Pennsylvania, Philadelphia. At the time she contributed to this work, she was a research assistant in the Development Strategy and Governance Division of IFPRI, Washington, DC. Andrew Comstock is a senior research analyst in the Development Strategy and Governance Division of IFPRI, Washington, DC. Paul Dorosh is the director of the Development Strategy and Governance Division of IFPRI, Washington, DC. Peixun Fang is a research analyst in the Development Strategy and Governance Division of IFPRI, Washington, DC.

153 154 Authors

Ricardo Hernandez is an agrifood economist for the Asia Region of CIAT, Viet Nam. Chaoran Hu is an assistant professor in the School of Economics at Sichuan University, Chengdu, China. At the time she contributed to this work, she was a doctoral student in the Department of Agricultural, Food, and Resource Economics at Michigan State University, East Lansing, US. Solomon Lemma is a data science/BI tool analyst at the Bank of America, Charlotte, NC, US. At the time he contributed to this work, he was a senior research assistant in the Markets, Trade, and Institutions Division of IFPRI, Washington, DC. Nicholas Minot is a senior research fellow in the Markets, Trade, and Institutions Division of IFPRI, Washington, DC. Shahidur Rashid is the director of the South Asia office of IFPRI, New Delhi. Thomas Reardon is a professor in the Department of Agricultural, Food, and Resource Economics at Michigan State University, East Lansing, US. Gracie Rosenbach is a senior research analyst in the Development Strategy and Governance Division of IFPRI, Washington, DC. Xiaobo Zhang is a senior research fellow in the Development Strategy and Governance Division of IFPRI, Washington, DC, and chair professor of economics at Peking University, China. INDEX

Page numbers for entries occurring in figures are followed by an f; those for entries in notes, by an n; and those for entries in tables, by a t.

Ahmed, M., 85–86 Cluster-based aquaculture growth, 57–58; Aquaculture in Bangladesh, 5, 12; growth surveys, 58 in, 86; trends in, 7, 8t, 79–84 Clustering, 73–75, 145; and accessibility Aquaculture production by fish category, to truck rental companies, 73, 75f; and 49, 50t adoption of modern inputs, 67, 69t; vs. cooperation among input dealers Aquaculture production trends, 79, 80t; and traders, 70–73, 71f, 72t, 74t (see historical, 7, 8t also Cooperative behavior in clusters); Aquaculture sector in Bangladesh: impor- degree of clustering in sample districts, tance, 19–20; opportunities and chal- 62, 64, 65f; fish farmers’ output lenges, 20–23; transformation in, 10–11 specialization and, 64–67, 66t, 68t; impact, 64–73; positive externalities for Belton, Ben, 32, 43, 86–87 fish value chain actors created by, 145 Blue Revolution, 77, 98, 143, 147, 150; Clusters: defined, 58; districts drivers and implications of the, in fish-clustering areas, 58, 59f; 10–12; enablers of the, 143–45; features of fish production clusters, environmental impact, 98–99; in the 145; measuring, 58–59, 61–62, food security debate, 1–4; origin of the 63–64t, 64, 67, 70–71; structural change in feed dealer clusters over term, 1n. See also specific topics 10 years, 38, 39t; structural change in Bouis, H. E., 2–3 feed mill clusters over 10 years, 36–37, 38t; structural change in fish farmer Capital (equipment), rapid increase/ clusters over 10 years, 38–40, 40t (see investment in quasi-fixed, 52–53 also Fish farm segment restructuring); structural change in hatchery clusters Capital intensification in farm sector, over 10 years, 36, 37t; structural 53–54 change in rural fish trader clusters over CGIAR, 3 10 years, 42t, 43 Chemicals, rapid increase in use of, 51 Commercial aquaculture, defined, 31

155 156 Index

Competition, 57, 67. See also Cooperative Feed mills, 33t, 44, 53–54; defined, 33t, behavior in clusters 37; distribution by district, 60f Cooperative behavior in clusters, traders’ Fish consumption, 115, 117, 118t, 140t; and feed dealers’, 57–58, 67–75, 71f, consumption of most consumed 72t; determinants of, 71–73, 72t, 74t; species in rural Bangladesh, 48–49, 49f horizontal cooperation, 68–71, 70t Fish consumption trends, 8–9, 9t, 81–83, Cross-price elasticities, 108, 123, 124, 132t 82t Fish demand, determinants of: demand Deaton, A., 79, 87, 88 parameter results, 108, 109t; economic specification, 105–7 Demand, fish, 117–21; aquaculture and, 77; in Bangladesh, 3, 90, 117, Fish farm capital-to-labor ratio by year and 121; demographic variables used in zone, 52, 53t estimation of household demand Fish farm commercialization, 45, 46t, 47t, parameters, 126t 50t Demand and supply projection to 2030, Fish farm harvest: disposal by final user 146–47 location, 45, 47t; disposal by final user Demand elasticity estimates, alternate, type, 45, 46t 122–24 Fish farm segment restructuring, 38–41, Demand elasticity for aquaculture 40–42t, 43 products. See Price elasticity: of Fish farmers: distribution by district, 60f. demand for aquaculture products See also specific topics Department of Fisheries (DoF), 24, 83, 88 Fish feed. See Feed Fish price trends, historical, 9–10, 10t Entrepreneurial ponds, 5 Fish prices, 114–15, 116t, 140t Equipment. See Capital Fish production, 114–15, 116t, 118–20, Expenditure elasticities, 103, 105–8, 109t, 119f, 140t; production projections to 122, 123 2015, 142t; in various geographic areas, 20, 21t. See also specific topics Expenditure elasticity parameters, alternative specifications of, 133t Fish production systems, 5; expenditure elasticities and, 108, 109t; historical Exports, 6, 19–20, 58–59, 80–81, 114, trends, 7–10; past and present, 104. 130t See also specific systems Fish productivity growth, model Farm capital. See Capital simulation assumptions for, 114, 115t Farmers, 33t Fish seed. See Seed Feed, fish: rapid increase of purchased, Fish species, names of, 17t, 130–31t 50–51. See also Input supply Fish supply chain, 59, 62; major actors in, Feed commercialization, 44 61; spatial linkage of, 61, 62t Feed dealer segment restructuring, 38, 39t Fish supply chain survey, 64 Feed mill clusters: structural change in, 37, Fish traders: distribution by district, 60f. 38t. See also Clusters See also specific topics Feed mill segment restructuring, 36–37, Fish wholesale segment restructuring, 42t, 38t 43 Index 157

Fisheries, Department of. See Department Lewis, D., 3 of Fisheries Lorica, M. H., 85–86 Fisheries legislation, 23 Fisheries subsector policy environment, Marine capture, 6 23–24 Market equilibrium condition, 114 Food security: aquaculture and, 12, 19, 25, 77; pathways through which aquaculture technology could affect, Nadvi, K., 57 85–86 National Fisheries Policy (NFP), 23–25, Food security debate, Blue Revolution in 148–49 the, 1–4. See also Blue Revolution Nutrition, 1, 2, 4, 78 Food supply chain, transformation and modernization of, 43. See also Fish Own-price elasticities of aquaculture, 103, supply chain 108, 123–24, 131t

Green Revolution, 1, 2, 35, 83 Poor/nonpoor distinction, 86, 105, 109– 12t, 116t, 118t, 131–38t, 141t Hatcheries, 33t; distribution by district, Population growth, model simulation 60f assumptions for, 114, 116t Hatchery segment restructuring, 36, 37t Poverty, impacts of aquaculture growth on, Hired labor, increase of use of, 51–52 95–98, 96t, 145–46 Homestead ponds, 5, 40, 50 Price elasticity: of demand for aqua- Household Income and Expenditure culture products, 88–90, 103, Survey (HIES), 26–28 105–7, 131t (see also Cross-price elasticities; Own-price elasticities of aquaculture); of demand for fish, 108, Ilish, 83 110–13t Immanent vs. interventionist development, Price elasticity parameters, alternative 31–32 specifications of, 134–39t Income (distribution), impacts of Prices of selected fish varieties, 83–84, 84f aquaculture growth on household, Price trends, 83–84 93–95, 94t Product cycle, 48–49, 49f; stages of, 48 Income growth, model simulation assumptions for, 114, 116t Pro-poorness: of aquaculture, 86–87; types of, 86 Inland capture, 5–6 Input dealers, 33t Quadratic Almost Ideal Demand System Input supply, 23, 26. See also Feed; Seed (QUAIDS), 103, 105–7, 109t, 124, Input supply segments, growth and 133t, 136–39t technological change in, 53–54 “Quiet revolution” in agrifood systems, 32 Interventionist development, 32 Schmitz, H., 57 Labor, hired, 51–52 Seed: rapid increase of purchased, 50–51. Leser, C. E. V., 106 See also Input supply 158 Index

Seed commercialization, 43–44 Value chain (VC), transformation Specialization, fish farmers’ output: of structure and conduct in clustering and, 64–67, 66t, 68t; the, 35, 144; commercialization commercialization and, 43 and spatial elongation, 43–47; growth and concentration, 35–41, Supply constraints, 85 43; technological cum product Supply elasticity, 88, 90 composition/product cycle change and patterns, 48–53 Toufique, K. A., 86–87 Value chain (VC) actors, 31, 32 Traders, 33t; distribution by district, 60f; Value chain (VC) survey, 25–26 rise of trader segment to urban areas, Value chains (VCs), 31 45–47. See also Cooperative behavior in clusters Welfare implications of aquaculture Truck rental companies, clustering and growth, 145–46; market positions and accessibility to, 73, 75f net benefit ratios, 91–93, 92t.See also Income; Poverty The Making of a in Bangladesh Enablers, Impacts, and the Rashid Blue Revolution Path Ahead for Aquaculture & Zhang

Maria Litwa/laif/Redux Cover photo:Cover is a senior is research fellow in the Development info Gareth Johnstone Gareth Lawrence Haddad Lawrence Director General, WorldFish General, Director Md. Raisul AlamMd. Raisul Mondal www.ifpri. & 2018 World Food& 2018 Prize Winner

| Xiaobo Zhang Xiaobo

+1-202-862-5606 www.ifpri.org F.

is director is the of South Asia office the of International Food Policy Research |

Washington, DC 20005 USA

, Executive Global Director, Alliance for Improved Nutrition (GAIN) Secretary, Ministry Fisheries of and Livestock, Government Bangladesh of [email protected] +1-202-862-5600 T. of protein, vitamins, and minerals for the human diet, but the world’s oceans but the world’s the human for and diet, minerals vitamins, protein, of makes an invaluable contribution to future research and policymaking and research future to contribution an invaluable makes policymakers and investors on how they can make these nutritious foods foods these nutritious on can how make they and investors policymakers relevant, and timely. Bangladesh is committed to developing the fisheries the fisheries developing to committed Bangladesh is and timely. relevant, the analytical frameworks that I believe objective. that achieve made to Institute (IFPRI), NewInstitute Delhi. (IFPRI), sector in a sustainable manner; and significant investments are being being are investments and significant manner; sector a sustainable in a commercially viable and environmentally sustainable roadmap for for roadmap sustainable and environmentally viable a commercially the most vulnerable.” all, especially to accessible are mostly overfished and so we need to look at environmentally sustainable sustainable and overfished mostly so need we look environmentally are at to book This sector how documents answers. has this grown for aquaculture and findings of the book will prove useful in policymaking.” in useful the book of prove and will findings in Bangladesh and what it needs to do to meet future demand. As demand. usual meet Bangladesh and future in needs do to what it to Shahidur Rashid at Pekingat China. University, Strategyand Governance Division IFPRI, of Washington, DC, and chair professor economics of to ensuring access to a sustainable supply of safe, affordable, and nutritious and nutritious affordable, safe, supply of a sustainable to access ensuring to fish and aquatic foods for all.” for foods andfish aquatic from IFPRI it combines pragmatism, policy, and first-rate evidence to create create to evidence and first-rate policy, pragmatism, combines it IFPRI from 1201 Eye1201 Street, NW Email: “A robust and insightful analysis of aquaculture in Bangladesh. This book Bangladesh. This in aquaculture of analysis and robust insightful “A “This book is very timely. Fish, shellfish, and mollusks are important sources and are mollusks shellfish, Fish, book“This very is timely. “A compelling set of analyses on aquaculture in Bangladesh that is rigorous, rigorous, is Bangladesh that in analyses on aquaculture set of compelling “A