91638 Addressing Inequality in South Asia Addressing Inequality in South Asia Martín Rama, Tara Béteille, Yue Li, Pradeep K. Mitra, and John Lincoln Newman © 2015 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 17 16 15 14 This work is a product of the staff of The World Bank with external contributions. The fi ndings, interpre- tations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: Rama, Martín, Tara Béteille, Yue Li, Pradeep K. Mitra, and John Lincoln Newman. 2015. Addressing Inequality in South Asia. South Asia Development Matters. Washington, DC: World Bank. doi: 10.1596/978-1-4648-0022-1. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations —If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party- owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-0022-1 ISBN (electronic): 978-1-4648-0023-8 DOI: 10.1596/978-1-4648-0022-1 Cover photo: © Chris Stowers / Panos Pictures. Used with the permission of Chris Stowers / Panos Pictures. Further permission required for reuse. Cover design: Critical Stages Library of Congress Cataloging-in-Publication Data has been requested. Contents Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Why inequality matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 The extent of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Drivers of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Limited opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Substantial mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Inadequate support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 1. Why Inequality Matters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Inequality of what? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Opportunities versus outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Monetary measures of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Multidimensional indicators of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Subjective well-being . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 The costs (and benefits) of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Intrinsic value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Incentives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Aspirations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Access to finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 Public goods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 Rent seeking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Conflict . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 v vi CONTENTS References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Why inequality matters: Main messages and policy implications . . . . . . . . . . . . . . . . . . . . 60 2. The Extent of Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 Monetary indicators of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 Nonmonetary dimensions of inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 Monetary inequality is increasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 Mixed trends in nonmonetary inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 What lies behind inequality? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 The extent of inequality: Main messages and policy implications . . . . . . . . . . . . . . . . . . . 93 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 3. Limited Opportunity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 Inequality in access to basic services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 Coverage is improving, equity less so . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 Who is covered? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 The role of inherited circumstances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 Limited resources and low progressivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Limited opportunity: Main messages and policy implications . . . . . . . . . . . . . . . . . . . . . 117 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 4. Substantial Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 Mobility across generations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 Mobility within the same generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 Nonfarm jobs drive mobility in villages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Migration is a major source of mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 Urban mobility is shaped by city characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 Substantial mobility: Main messages and policy implications . . . . . . . . . . . . . . . . . . . . . 146 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 5. Inadequate Support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Shocks and how households cope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 A scorecard for social protection programs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 The distributional impact of taxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 Wasteful and often regressive subsidies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 The promise of intergovernmental transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 Inadequate support: Main messages and policy implications . . . . . . . . . . . . . . . . . . . . . . 173 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 Boxes 1.1 Discrimination by teachers pushes children out of school . . . . . . . . . . . . . . . . . . . . . . 41 1.2 Standard statistical measures of monetary inequality . . . . . . . . . . . . . . . . . . . . . . . . . 44 1.3 Some monetary indicators may underestimate the true extent of inequality . . . . . . . . 46 1.4 Bhutan uses a happiness index to measure well-being . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.1 South Asian household surveys used in this report . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 3.1 In demographic transitions, inequality of opportunity increases inequality of outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 3.2 The Human Opportunity Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 4.1 For a given inequality of opportunity, mobility reduces the inequality of outcomes . . . 120 4.2 Measuring mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 4.3 How synthetic panels are constructed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 CONTENTS vii 4.4 International migration supports upward mobility in Bangladesh and Nepal . . . . . . 138 5.1 Support aims at offsetting inequality of outcomes stemming from bad luck . . . . . . . 152 5.2 Bangladesh has a rich and complex social protection architecture . . . . . . . . . . . . . . . 160 5.3 The adequacy of social assistance programs has declined in Bangladesh . . . . . . . . . . 164 Figures 1 Based on standard monetary indicators, South Asia has moderate levels of inequality . . . 2 2 Billionaire wealth in India is exceptionally large . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 3 The health outcomes of the poor are among the worst worldwide . . . . . . . . . . . . . . . . . 5 4 Returns to education create incentives to study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 5 Greater inequality reduces the quality of public services when the rich can opt out . . . 7 6 Poverty is higher in Indian districts suffering from Naxalite violence . . . . . . . . . . . . . . 8 7 The least wealthy are alarmingly vulnerable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 8 Inequality in health outcomes is wide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 9 Schooling among young adults is highly unequal in some countries in South Asia . . . 11 10 Richer countries tend to be more unequal in both South Asia and East Asia . . . . . . . . 12 11 Monetary inequality is increasing across most of South Asia. . . . . . . . . . . . . . . . . . . . 13 12 South Asians do not see an environment conducive to lower inequality. . . . . . . . . . . . 14 13 Multiple factors affect household outcomes relative to others in society . . . . . . . . . . . 15 14 Opportunities in education are better than in health or sanitation, as measured by the HOI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 15 Better opportunity is driven by greater coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 16 Parent’s education and location are critically important circumstances . . . . . . . . . . . . 19 17 Considerable occupational mobility exists across generations in India . . . . . . . . . . . . 21 18 Occupational mobility is higher for younger generations . . . . . . . . . . . . . . . . . . . . . . . 21 19 Upward mobility in South Asian countries is similar to that in the United States and Vietnam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 20 Upward mobility is much stronger in cities than in rural areas . . . . . . . . . . . . . . . . . . 24 21 The composition of urban employment varies with city size and governance in India. . . 27 22 In Pakistan, poorer and richer households cope with shocks in different ways . . . . . . 28 23 Social assistance is less adequate than social insurance but has greater coverage. . . . . 30 24 Electricity subsidies favor the better-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 25 Development spending per person is lower in poorer states and districts . . . . . . . . . . . 33 1.1 Estimates of expenditures differ between household surveys and national accounts . . . 45 1.2 Monetary and nonmonetary indicators can lead to opposite conclusions . . . . . . . . . . 48 1.3 Returns to education create incentives to study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 1.4 Greater inequality in landholdings is associated with lower asset accumulation among the poor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 1.5 Greater inequality reduces the quality of public services when the rich can opt out . . . 55 1.6 Inequality was highest under rent-seeking colonial rule . . . . . . . . . . . . . . . . . . . . . . . . 56 1.7 Poverty is higher in Indian districts suffering from Naxalite violence . . . . . . . . . . . . . 58 2.1 Based on standard monetary indicators, South Asia has moderate levels of inequality . . .69 2.2 Top incomes have been rising in India since the 1980s . . . . . . . . . . . . . . . . . . . . . . . . 72 2.3 The distribution of wealth is more concentrated than that of consumption in India . . .73 2.4 The least wealthy are alarmingly vulnerable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 2.5 Billionaire wealth in India is exceptionally large . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 2.6 Gaps in health outcomes are wide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 2.7 The health outcomes of the poor are among the worst worldwide . . . . . . . . . . . . . . . . 78 2.8 Schooling among young adults is highly unequal in some countries in South Asia . . . 79 viii CONTENTS 2.9 Gaps in educational attainment are much narrower among children than among adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 2.10 Among children, similarities in schooling hide disparities in learning in India . . . . . . 80 2.11 Growth is reducing poverty in South Asia, as it did in East Asia . . . . . . . . . . . . . . . . . 81 2.12 Richer countries tend to be more unequal in both South Asia and East Asia . . . . . . . . 82 2.13 Prosperity has been shared less widely in South and East Asia . . . . . . . . . . . . . . . . . . 83 2.14 Monetary inequality is increasing across most of South Asia. . . . . . . . . . . . . . . . . . . . 83 2.15 Inequality in health outcomes has remained stable or increased . . . . . . . . . . . . . . . . . 85 2.16 Inequality in educational attainment is generally decreasing . . . . . . . . . . . . . . . . . . . . 86 2.17 Learning outcomes have deteriorated in rural India and rural Pakistan . . . . . . . . . . . 87 2.18 The gap in learning outcomes between ethnic groups is declining in Sri Lanka. . . . . . 88 2.19 Education explains a growing share of overall inequality . . . . . . . . . . . . . . . . . . . . . . 88 2.20 The rural-urban divide is becoming a more important source of inequality . . . . . . . . 89 2.21 Caste is an important correlate of inequality in some Indian states . . . . . . . . . . . . . . . 90 2.22 South Asians do not see an environment conducive to lower inequality. . . . . . . . . . . . 91 2.23 Multiple factors affect household outcomes relative to others in society . . . . . . . . . . . 92 3.1 The Human Opportunity Index for basic health services is low in most of South Asia . . .99 3.2 The Human Opportunity Index for education is low in Afghanistan and Pakistan . . . 101 3.3 The Human Opportunity Index for sanitation is especially low . . . . . . . . . . . . . . . . 102 3.4 Opportunities have improved faster in primary education than in health services. . . 103 3.5 Better opportunities in health are driven by greater coverage of basic services . . . . . 103 3.6 Better opportunities in education reflect greater coverage and higher equity in some countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 3.7 The coverage of institutional births is lower in rural areas . . . . . . . . . . . . . . . . . . . . 105 3.8 The urban-rural gap in coverage remains large for secondary education . . . . . . . . . 106 3.9 Access to electricity is lower in rural areas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 3.10 In Bangladesh, children of slum dwellers have less access to education . . . . . . . . . . . 107 3.11 The coverage of health services is almost the same for boys and girls in South Asia . . . 108 3.12 Gender gaps in coverage are small for primary education but large for secondary education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.13 The coverage of health services differs widely by mother’s education . . . . . . . . . . . . 110 3.14 Parents’ education is highly correlated with children’s secondary school attainment . . . 111 3.15 Location and mother’s education are critically important circumstances in access to health services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 3.16 Gender and religion matter for access to education . . . . . . . . . . . . . . . . . . . . . . . . . . 113 3.17 Location is a critical circumstance for access to infrastructure services. . . . . . . . . . . 114 3.18 Limited public resources are spent on health services in South Asia . . . . . . . . . . . . . 114 3.19 Spending is progressive only for some health services in Bangladesh . . . . . . . . . . . . . 115 3.20 Limited public resources are spent on education services . . . . . . . . . . . . . . . . . . . . . . 116 3.21 In Bangladesh, education spending is progressive only at the primary level . . . . . . . . 116 4.1 Considerable occupational mobility exists across generations in India . . . . . . . . . . . 124 4.2 Occupational mobility is higher for younger generations . . . . . . . . . . . . . . . . . . . . . . 124 4.3 Consumption grows faster among the poor than among the better-off . . . . . . . . . . 126 4.4 Upward mobility in South Asian countries is similar to that of the United States and Vietnam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 4.5 Upward mobility in India is strong for the Scheduled Castes and Scheduled Tribes. . . 128 4.6 Upward mobility is substantial in rural Bangladesh and rural India . . . . . . . . . . . . . 129 4.7 Rural India and rural Pakistan have seen a consistent expansion of nonfarm employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.8 Casual rural jobs provide increasingly higher earnings in India . . . . . . . . . . . . . . . . 131 CONTENTS ix 4.9 Among men, permanent migration is driven by job aspirations . . . . . . . . . . . . . . . . . 133 4.10 Permanent migrants have higher economic status in India . . . . . . . . . . . . . . . . . . . . . 135 4.11 Seasonal migration is more common among poor and socially disadvantaged groups in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 4.12 Migration provides opportunities for occupational mobility to women in India . . . . 137 4.13 Upward mobility is much stronger in cities than in rural areas . . . . . . . . . . . . . . . . . 139 4.14 Even self-employment and casual work support upward mobility in urban areas . . . 141 4.15 Many informal sector workers are wage earners in urban India . . . . . . . . . . . . . . . . 141 4.16 South Asian countries are less urban than their peers . . . . . . . . . . . . . . . . . . . . . . . . 142 4.17 The composition of urban employment varies with city size in India . . . . . . . . . . . . . 144 4.18 The composition of urban employment also varies with city governance in India . . . 145 5.1 Health-related events and disasters are the most common shocks in Pakistan . . . . . . 155 5.2 Disasters in Pakistan affect rural populations much more than urban populations . . 156 5.3 In Pakistan, poorer and richer households cope with shocks in different ways . . . . . 158 5.4 Spending on health is mainly out of households’ pockets. . . . . . . . . . . . . . . . . . . . . . 158 5.5 Spending on social protection in South Asia is lower than in other developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 5.6 Absolute spending on social protection is progressive in Bangladesh and Sri Lanka . . . 161 5.7 Social assistance is less adequate than social insurance but has greater coverage. . . . 163 5.8 Tax revenue is lower than in other countries at a similar development level . . . . . . . 165 5.9 South Asian countries rely less on income taxes and more on trade taxes . . . . . . . . . 166 5.10 In Pakistan, even registered taxpayers fail to file tax returns . . . . . . . . . . . . . . . . . . . 167 5.11 Relative to their means, the poor in Pakistan pay almost as much tax as the middle class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 5.12 Much public spending goes into energy subsidies. . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 5.13 Electricity subsidies favor the better-off . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 5.14 Intergovernmental transfers benefit poorer states and provinces . . . . . . . . . . . . . . . . 171 5.15 Development spending per person is lower in poorer states and districts . . . . . . . . . . 172 Maps 1 Government revenue in South Asia is low compared with the rest of the world. . . . . . 31 5.1 Government revenue in South Asia is low compared with the rest of the world. . . . . 165 Tables 1 Changes in employment status reveal substantial mobility among migrant men in India. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2 Rural jobs allow people to escape poverty; urban jobs are a ticket to the middle class . . . 25 4.1 Occupational mobility has increased more for the most disadvantaged population groups in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 4.2 Upward mobility is considerable among the poor and the vulnerable . . . . . . . . . . . . 127 4.3 Changes in employment status reveal substantial mobility among migrant men in India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 4.4 Rural jobs allow people to escape poverty; urban jobs are a ticket to the middle class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .140 5.1 Bangladesh suffered 2 of the 10 most deadly natural disasters of recent times . . . . . 154 Foreword T he extent of inequality, and what to inequality which tend to hide as much as do about it, are among the most hotly they reveal. Indeed, high inequality should debated issues in economics. Every be less of a concern in a society with high faith and ideology has normative views on mobility, where people from disadvantaged how much inequality is tolerable, or desir- backgrounds and their offspring can pros- able. And to complicate matters, equality per. Conversely, even apparently low levels of along a dimension that matters for some inequality could justify corrective interven- members of society often entails inequal- tions if higher earnings result from connec- ity in some respect that others care about. tions and rent seeking, rather than from hard Debates on inequality seem to become more work and entrepreneurship. intense in periods of rapid structural trans- This report takes a fresh look at inequal- formation, both in advanced economies and ity in South Asia, one of the most dynamic in developing countries. As jobs migrate regions in the world. It does so with a abroad, or people move to cities, or a mod- focus on well-being, exploring both mon- ern sector emerges, the entire distribution etary and non-monetary dimensions of of well-being is shaken. Age-old rankings inequality. The former include income, that seemed cast in stone become compro- consu mpt ion a nd wea lt h. T he lat ter mised, new fortunes are made—sometimes comprise access to basic services in health, quickly—and the entire social fabric comes education, and infrastructure, as well as under stress. subjective assessments. Those are times when calls for action This combination of perspectives is reveal- abound. But they are also times of confusion: ing. In particular, countries in South Asia do on the extent of inequality, on its drivers, and not appear to be particularly unequal when on its implications. Debates are often framed focusing on consumption per person, but the in either-or terms which are not particularly region is host to both extravagant wealth at enlightening. Rivers of ink are devoted to the top and appalling human development discussing whether the government should outcomes at the bottom. This in itself says tackle inequality or boost growth “first,” something about what is working well—the for instance. There is also a tendency to ability of people to make a living—and what focus on simple, aggregate indicators of is not working so well—the delivery of basic xi xii FOREWORD social services and the channeling of support the social sectors can be derived from these to those in need. analyses. The report also makes an analytical con- There are also shortcomings in the support tribution, by providing a clear framework people receive throughout life. The region is to think about inequality and what to do home to some remarkable social protection about it. This framework considers not only programs, which have attracted worldwide opportunity in childhood, but also mobil- attention. But beyond these targeted interven- ity in adulthood, and support throughout tions, widespread tax avoidance and evasion, life. The focus on opportunity, dominant in combined with regressive subsidies for elec- recent applied work on equity, brings war- tricity and fuel, imply that the poor receive ranted attention to the social sectors: access relatively little support from the rich. Inter- to education, quality of schooling, coverage governmental transfers, channeling resources of health programs… Not surprisingly, this from rich to poor areas, are progressive; but focus has been associated with an emphasis for now they are not large enough to make on entitlements and rights-based approaches. a big difference. There is obviously a fi scal But limiting our attention to opportunity agenda out of these findings. also runs the risk of downplaying other very Perhaps the most important new insight important areas of public policy. The most from the report concerns mobility in adult- dramatic structural transformation South hood. Despite South Asia’s being a region Asia is going through is urbanization. The characterized by high informality and occupational and geographic transitions haphazard urbanization, jobs and migration associated with it are shaping mobility across have performed much better than could be and within generations. South Asia is also a expected. Occupational mobility is increas- region characterized by low tax revenue and ing steadily from one generation to the next. massive subsidies, affecting the capacity of Within the same generation, mobility in governments to provide adequate support earnings—measured by the ability to move throughout life. out of poverty and into the middle class—is The combination of a broad range of comparable to that of the United States or indicators with a focus on individuals’ well- Vietnam. Importantly, jobs seem to trump being throughout the lifecycle yields impor- caste as the extent of mobility is similar for tant insights. The report confirms that the groups across the entire social spectrum. region still has a way to go if it wants to Nurturing this dynamism should be the ensure opportunity in childhood. There has highest priority for those who care about been enormous progress in the coverage of addressing inequality. And that involves an primary school, but the quality of education agenda for urbanization and private sector received by the poor, and access to secondary development. school for girls, remain important challenges. The same applies to access to health ser- Philippe Le Houérou vices and sanitation, with rural areas being Vice-President for South Asia at a serious disadvantage. A clear agenda for The World Bank Acknowledgments T his report was prepared by a core Va r u n G au r i , K a rla Hof f, Sha h idu r team led by Martín Rama and includ- Khandker, Peter Lanjouw, Ghazala Mansuri, ing Tara Béteille, Yue Li, Pradeep K. Mario Picon, and Vyjayanthi Sankar. Mitra, and John Lincoln Newman. The Ravi Kanbur served as an external adviser. other members of the team were Gladys Pranab Bardhan, Peter Lanjouw, Julián López-Acevedo, Cem Mete, and Albertus Messina, and Branko Milanovic were the Voetberg. Important contributions were peer reviewers. They all contributed rich made by Mehtabul Azam, Olivier Dupriez, feedback throughout the preparation process. Virgilio Galdo, M inh Cong Nguyen, The production of the report and the María Florencia Pinto, and Kritika Saxena. logistics supporting it were assured by Additional research support was provided Neelam Chowdhry, Maya S. Krishnan, and by Claudia Berg, Bilgehan Gokcen, Rehan Muhammad Shafiq. Eva Franzuela Cardenas Rafay Jamil, Ayesha Raheem, and Amir was in charge of resource management. Sadegh Sadeghi. T he World B a n k’s P ublish i ng a nd Former Regional Vice President for South Knowledge Unit was in charge of the design, Asia Isabel Guerrero and Regional Vice typesetting, printing, and dissemination of President Philippe H. Le Houérou provided both the hard and soft version of the report. overall guidance and invaluable insights dur- Martha Gottron and Laura Glassman ing the preparation process. edited the report. Special thanks go to Aziz A set of background papers was commis- Gökdemir and Patricia Katayama for coordi- sioned for the report. They were prepared nating the entire process. by Luis Alberto Andrés, Yamini Aiyar, The team also thanks Alex Anthony Dan Biller, Anirvan Chowdhury, Hai-Anh H. Ferguson and Gabriela Aguilar Martínez for Dang, Ambrish Dongre, Agustin Echenique, their guidance on communication. xiii Abbreviations BBS Bangladesh Bureau of Statistics BISP Benazir Income Support Programme BLSS Bhutan Living Standards Survey CIT corporate income tax DHS Demographic and Health Survey DRM day reconstruction method FFE Food for Education GDP gross domestic product GE Generalized Entropy (Index) GNH Gross National Happiness HIES Household Income and Expenditure Survey HOI Human Opportunity Index IHDS India Human Development Survey LFS Labor Force Survey LPG liquefied petroleum gas MGNREG Mahatma Gandhi National Rural Employment Guarantee (Act) MLD mean log deviation NEREC National Education Research and Evaluation Centre [Sri Lanka] NLSS Nepal Living Standards Survey NRVA National Risk and Vulnerability Assessment NSS National Sample Survey PDS Public Distribution System [India] PPP purchasing power parity PSLM Pakistan Social and Living Standards Measurement OECD Organisation for Economic Co-operation and Development UN United Nations VAT value added tax VPA Vulnerability and Poverty Assessment WDI World Development Indicators WHO World Health Organization xv Overview F ifty years have passed since the Nobel Disentangling what lies behind the laureate, poet-turned-plenipotentiary “extremes” is even more challenging. What Octavio Paz, saw India, where he shapes opportunity at birth may be differ- was Mexico’s ambassador, as “a land of ent from what fosters mobility in adulthood. extremes.” The poet’s muse was his encoun- Moreover, inequality can be both good and ter with the profusion of sights and sounds, bad. When differences in fortunes reward colors and smells, people and animals that hard work and entrepreneurship, they pro- greeted him during his travels on the subcon- vide incentives for individuals to try their tinent. He described it as “the incredible opu- best and in doing so to contribute to every- lence” of the maharajahs surrounded by what body’s well-being. But when the extremes he saw as “equally unbelievable” poverty (Paz reflect suppressed aspirations at one end and 1997). Is that still the case today? And if so, rent seeking at the other, inequality leads what should be done (and not done) about it? to wasted talent and consolidates institu- Despite the enormous progress made tional arrangements with long-term negative on statistical data and analytical tools over impact on growth and development. Without these 50 years, assessing whether South Asia a clear understanding of the drivers of remains “a land of extremes” is a significant inequality, policy recommendations on what undertaking. The extent of inequality var- to do about it could be flawed. ies depending on the indicator of individual If standard monetary indicators are to be well-being considered. taken at face value, South Asia has modest The assessment also varies depending levels of inequality. Gini coefficients for con- on whether attention goes to inequality of sumption per capita range between 0.28 and opportunity in childhood or to inequality of 0.40 depending on the country, much lower outcomes in adulthood. Moreover, a static than in China, Mexico, or South Africa picture that describes the distribution of for- (figure 1). The share of the poorest 40 percent tunes at any point in time may miss important of households in total consumption also sug- insights compared to a more dynamic one that gests that inequality in South Asian countries considers mobility throughout the life cycle. is moderate by international standards. 1 2 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 1 Based on standard monetary indicators, South Asia has moderate levels of inequality a. Gini coefficients 70 60 50 40 30 20 10 0 do m ca a, p. t, A Fr ny M l hi ia Ire ali Ja ia ite Sp n d I ain Fe Tu ria M ysia C co So B hile fri l an en k n yp rmaan Ko rab nce lad da Et tral d In gdo ia et ia T am M Lan ia Un Ta ldiv a Th uta s ai n Ni land ra y Chion Ar hana M tin a nm ay Be lan k gh ed a p. Ne sh ite nz es St a ala a lg d Au ium Ba a taly pa h A zi Bh ate de rke a k ge n Fin ar Af Sw stri pa d ani Pa ista Au lan op Ki nd Vi nes Sr unis re Re Re ut ra i De orw e ge e ist i ng na ex n t a I n s G n C N i G ian Un Eg ss Ru b. Share of the poorest 40 percent 30 25 20 15 10 5 0 . d ay en tan tan ep ny ria m sh l ia li a d a in m ly pa op Ma nad lan ank Spa tna Ita sia nia an es nd ion tes ey na ina ia ia le il ys er hi raz a ge co a ca a lanorw wed nis kis ab R rma ust lgiu lade ne nza hut ldiv aila rat Sta urk Gha Ch di in si fri Ne Ethi Ar exi ala Nig C B ni nt In n a e L e o hA F N S gha P Ar Ge A Be ng i a r i i d a B Ma Th de ed T Tu C I Sr M V In T e it M t, ut Af yp Ba n F Un So Eg s sia Ru Sources: Based on World Bank’s World Development Indicators database, http://data.worldbank.org/data-catalog/world-development-indicators, and the Organisation for Economic Co-operation and Development’s (OECD) Income Distribution and Poverty data series, http://stats.oecd.org/Index.aspx?DataSetCode=IDD. Note: Orange and light brown bars indicate countries where inequality is estimated based on consumption per capita. Light blue bars indicate countries with estimates based on income per capita. OVERVIEW 3 Arguably, the comparison is tainted by the the assessment is inequality of outcomes in nature of the monetary indicators considered relation to both monetary and nonmonetary in different countries. In advanced economies dimensions of well-being. Inherited circum- as well as in many Latin American countries, stances and a variety of shocks affect inequal- inequality is measured on the basis of income ity of outcomes; public policies can offset the per capita. In South Asian countries, in con- contributions of these factors but may instead trast, most surveys convey information about amplify them. Within a person’s life cycle, consumption per capita. Within the same disparities in opportunity, mobility, and sup- country, income inequality is generally higher port drive gaps in outcomes. To understand than consumption inequality. However, the whether something should be done to address conclusion that monetary inequality in South inequality in South Asia—and what that Asia is moderate holds even when comparing should be—this report analyzes the region’s only countries for which data on consump- performance on these three aspects. tion per capita are available. Information on the assets held by the wealthiest offers a complementary perspec- Why inequality matters tive on monetary inequality, one in which Equality carries an intrinsic value for South Asians at the top are disproportion- most of the world’s faiths and ideologies— ately rich. In the only two countries in the religious or secular. Every normative theory region with publicly known billionaires, the of social arrangements that has stood the concentration of billionaire wealth appears test of time also seems to demand equal- to be unusually large in India. According ity of something. However, this report also to Forbes magazine (2014), total billion- cares about the ways in which inequality aire wealth amounts to 12 percent of gross affects social organization and economic domestic product (GDP) in 2012. As such, performance. In other words, it takes a posi- India is an outlier in the ratio of billionaire tive and not just a normative perspective. wealth to GDP among economies at a similar Seen this way, inequality is neither good nor development level (figure 2). bad. Some forms of inequality generate costs Nonmonetary indicators of well-being to society whereas others entail benefits. provide a more striking picture than mone- The issue is to identify the turning point at tary indicators. Despite not being the poorest which the costs of inequality start exceed- region in the world, South Asia has some ing its benefits. Doing so in a precise way is of the worst human development outcomes clearly out of reach. But economic analysis worldwide, and the comparison is even more helps identify the main costs and benefits dramatic when focusing on the outcomes of and aids in getting a sense of their order of the poorest quintile. The share of children magnitude. under five who are stunted among the poor- Inequality in outcomes profoundly affects est quintile is above 50 percent in Bangladesh how individuals and households behave. and Nepal and reaches 60 percent in India At the risk of oversimplifying, some degree (figure 3). India and Pakistan also have some of monetary inequality is needed to cre- of the highest infant mortality rates and ate incentives for people to study and accu- under-five child mortality rates among the mulate human capital, to work instead of poor across all comparators. Of 1,000 chil- taking leisure, to save for the future, and to dren born in India’s poorest population quin- invest in risky businesses. Returns to educa- tile, 82 will die within 12 months and 117 tion are a clear example of a differentiation within five years. The figures for Pakistan are in labor earnings that spurs the accumula- 94 and 120, respectively. tion of human capital and economic growth Joining others in this endeavor, this report but at the same time results in inequality in takes a positive perspective to assess the outcomes. In South Asia, returns are larger extent of inequality. The primary focus of the higher the educational attainment of the 4 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2 Billionaire wealth in India is exceptionally large 20 Philippines Aggregate net wealth (percent of GDP, 2012) 18 16 Ukraine 14 India Mexico 12 Thailand 10 8 6 Nepal 4 2 Vietnam Romania 0 0 5,000 10,000 15,000 20,000 GDP per capita (PPP, constant 2011 international dollars, 2012) Sources: Based on Forbes magazine’s Billionaires database, http://www.forbes.com/billionaires/, and World Bank’s WDI database, http://data.worldbank.org /data-catalog/world-development-indicators. Note: PPP = purchasing power parity. person (figure 4). Schooling is of course not relationship with others. (Hoff and Pandey the only determinant of labor earnings, and it 2006, 2012). accounts for only a fraction of earnings varia- The difference can be positive as well. In tion across individuals. But the relationship India, some villages have reserved the posi- between schooling and earnings is robust. tion of chief councillor (pradhan) for women. However, incentives may fail to change After about seven years of exposure to a behavior when economic mobility is lack- female pradhan, the gender gap in aspirations ing. Entrenched inequality of outcomes was sharply reduced for teenagers in these can significantly undermine individuals’ villages. Girls were less likely to want to be a aspirations in youth, affecting their subse- housewife, less likely to want their in-laws to quent educational and occupational choices. determine their occupation, and more likely Confi rmatory bias leads people to conform to want a job that requires more education. to the stereotype—for example, by discount- The gender gap in educational outcomes was ing the abilities of those who belong to a erased in these villages. Because little else marginalized group. And entrenched poverty changed in terms of actual policy or career may lead to depression and behavior akin to opportunities, seeing a woman achieve the “learned helplessness” (Hoff 2012). position of local head likely provided a role Effects of this sort have been found to model and affected aspirations, efforts, and make a substantial difference. In a controlled educational choices (Beaman and others experiment in India, boys from high and low 2012; Duflo 2012). caste displayed the same ability to solve mazes Although inequality of outcomes may under monetary incentives, but low-caste create incentives to accumulate human and boys performed worse if the name and caste physical capital, it may also affect the capac- of the boys were announced at the beginning ity of households to borrow for that purpose. of the session. Making caste salient may have If accumulation needs to build on individual evoked in the children memories that changed or household savings, those at the bottom of how they think about themselves and their the distribution may be unable to increase OVERVIEW 5 FIGURE 3 The health outcomes of the poor are among the worst worldwide a. Infant mortality 140 Mortality rate (deaths per 1,000 births) 120 among the poorest quintile 100 80 60 40 20 0 ep s a at a a Za a m Ph am h ria da In gua Gh a a Ug ia Ni n Pa ia Co ives Vi u Ni blic n R pine ng l Bo il bi Gu nesi an ny al di es a ta r b liv az Pe Ba Nep an ge em m In m n kis lad Ke ra u ald Br et ilip lo do ca M ica in Do b. Stunting 70 60 among the poorest quintile Share of children (percent) 50 40 30 20 10 0 ca n a M via ng ia ria h a a ru a da ar m a Co lic s l di bi ny ive a an gu pa es b nm na Pe ub ut ge an li In m m lad Ke Gh Ne ra Bo ald Bh et Ni ep Za Ug lo ya Vi M nR Ni Ba ica in m Do Source: Based on World Bank Health, Nutrition and Population Statistics database, http://datatopics.worldbank.org/hnp/WealthQuintiles. Note: Infant mortality rate is the number of deaths to children younger than 12 months per 1,000 live births. Stunting is the percentage of children younger than five years of age whose z-scores are two standard deviations or more below World Health Organization (WHO) Child Growth Standards. 6 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4 Returns to education create incentives to study 60 relative to no education (percent) 50 Earnings premium 40 30 20 10 0 Sri Lanka, 2008 India, 2010 Nepal, 2008 Pakistan, 2008 Incomplete primary Complete primary Complete lower secondary Complete higher secondary Source: World Bank 2011. their capital significantly, and that in turn most unequal villages. Notably, this nega- may perpetuate inequality. In rural Pakistan, tive impact of land concentration on teaching concentration of landholdings is the norm, quality does not extend to private schools. but the extent of concentration varies from These results imply that service quality and one village to another. Controlling for the access tend to decline at very high levels of initial poverty status and livestock holding of inequality but mainly for services for which households, the poor are more likely to accu- the wealthy can move to private providers— mulate livestock in villages with a lower level such as schools. of land inequality (Mansuri 2013). Inequality of outcomes does not gener- The influence of inequality goes beyond ate the right incentives when it rests on rents individual or household behavior. Inequality (Stiglitz 2013). In that case, rather than being also affects the ability of people to act col- encouraged to study or to accumulate, indi- lectively, the institutions they set up, and the viduals and households divert their efforts ways in which resources are allocated for the toward securing favoritism and protection. benefit of the group. A salient case is the pro- Rewarding such behaviors with high returns vision of public goods, where inequality can is costly to economic growth and social devel- have effects of opposite signs. On the one opment. It leads to a suboptimal allocation of hand, in a very unequal society, the better- resources in the short term and consolidates off typically have more power and are more institutional arrangements with negative effective at pulling in resources for the public long-term impacts on growth. goods they value. On the other hand, a high For instance, there is no doubt that India degree of inequality makes it more tempting has world-class entrepreneurs, commanding for the better-off to opt out of public services admiration for their innovation and man- altogether. agement capacity, and many of them oper- In the end, which of the two effects pre- ate successfully in highly competitive global vails is likely to depend on whether opting markets. At the same time, over a quarter of out is an option. In Pakistan, land inequal- India’s billionaire wealth is estimated to be ity is unambiguously associated with greater inherited, 40 percent is based on inheritance, access to services in the cases of electricity, and 60 percent originates from “rent-thick drainage, and access to public transport sectors,” such as real estate, infrastruc- (figure 5). In contrast, the teaching quality ture, construction, mining, telecommuni- in public primary schools is far poorer in the cations, cement, and media. This does not OVERVIEW 7 FIGURE 5 Greater inequality reduces the quality of public services when the rich can opt out Change in the availability of public goods when the share of land owned by the top quintile increases by 1 percent in Pakistani villages 100 80 60 The rich can opt out 40 Estimated change (percent) 20 0 –20 –40 –60 –80 The rich cannot opt out –100 –120 n t m ) ity d ol ol ic) nk te ni tio a ho ho te bl Ba ro ric iva hu rta ys pu sc sc ct d pr alt es ve po Ele y( te ic y( he bl Pa ag iva lit ns lit Pu ain ua sic tra Pr ua rq Ba Dr rq ic he bl he Pu ac ac Te Te Source: Based on Mansuri 2013 for this report. imply that wealth was acquired through violence for economic gain, such as the con- the exercise of influence, but highlights trol of resources, property, occupations, and that the potential for rent extraction business activities (Blattman and Miguel exists (Gandhi and Walton 2012). 2010; Collier and Hoeffler 2004). In others, Some connections also exist between economic factors lurk in the background of a inequality and confl ict, though a straight- conflict that erupts along social and political forward relationship would be hard to cleavages (Bardhan 2005; Horowitz 2000). establish between the two. Inequality may Inequality, especially deprivation, may inten- damage trust—the foundation for social sify the grievances felt by certain groups or cohesion— and thus weaken collective can reduce the opportunity costs of initiating decision making. The problem is particu- and joining a violent conflict. larly salient in management of common Confl ict may take many extreme forms. resources. Across irrigation communities In South Asia, it is more common in areas in south India and in Nepal, inequality is characterized by massive deprivation (Iyer found to make resolving disputes in water 2009). In the case of India, the probabil- allocation more difficult (Bardhan 2005; ity of a district being affected by Naxalites Lam 1998). (Maoist rebels) can be linked to the charac- More broadly, inequality affects the eco- teristics of the district. With the exception nomics of confl ict (Lichbach 1989). In some of Jharkhand, poverty incidence of rural cases, conflict reflects a systematic use of areas is higher in districts where Naxalites 8 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 6 Poverty is higher in Indian districts suffering from Naxalite violence Rural areas 60 50 Poverty headcount ratio (percent) 40 30 20 10 0 India Andhra Bihar Chhattisgarh Jharkhand Madhya Maharashtra Odisha Uttar West Bengal Pradesh Pradesh Pradesh LWEA districts Other districts in affected state Other districts in country Source: Based on India National Sample Survey (NSS) 2011–12. Note: LWEA = left-wing-extremism-affected districts. As defined by the Planning Commission in http://pcserver.nic.in/iapmis/state_district_list.aspx, India has 88 such districts. The headcount ratio is based on the national poverty line. are better implanted (figure 6). In Pakistan, spend their money—and to remind respon- the probability of violent attacks by insur- dents about them. Richer households also tend gents, sectarians, and terrorists is found to shun surveys of this sort. One indication of to increase with food insecurity and land underreporting is the size of the discrepancies inequality (Malik 2009, 2011). between levels and growth rates of consump- tion, as measured by household surveys and by national accounts. Disconnect between The extent of inequality the two major data sources is large in several Both monetary and nonmonetary indicators South Asian countries, especially in India. of well-being capture important dimensions Individual tax returns can be used to exam- of inequality in outcomes. Traditionally, the ine the extent of undercounting of the rich assessment of inequality is dominated by sin- in household surveys (Banerjee and Piketty gle-dimensional monetary indicators, repre- 2005). According to this data source, the sented by either income or consumption, both income share of India’s top 0.01 percent was in cash and in kind. These monetary indica- in the 1.5 percent to 2 percent range, whereas tors are generally computed based on indi- the share of the top 0.1 percent was in the vidual records from representative household 3 percent to 4.5 percent range. Assuming that surveys. However, household surveys may not the top 1 percent is not captured by house- capture well the income or the consumption hold surveys is not enough to account for the of the richest members of society. The survey full gap but explains 20 percent to 40 percent questionnaires usually focus on the relatively of it. This fraction is large enough to give basic basket of goods and services purchased credence to the hypothesis that traditional by those who live around the poverty line. In so income- or consumption-based monetary doing, they fail to capture the more diverse indicators are biased downward, probably by and sophisticated ways in which the better-off a substantial margin. OVERVIEW 9 The distribution of wealth provides a Monetary and nonmonetary indicators tend complementary perspective on monetary to be correlated, as shown by the fact that inequality in which considerable disparities health status or educational attainment is exist in asset holdings, net worth, and finan- typically higher among people who are better cial vulnerability. In India, at the household off in monetary terms. But the correlation is level, the Gini coefficient is 0.668 for asset not perfect because monetary and nonmon- holdings and 0.680 for net worth. As in other etary indicators capture different concepts countries, the wealth distribution is more and can vary independently. Therefore, non- concentrated than the distribution of income monetary indicators provide additional infor- and especially more concentrated than that of mation on distribution of well-being, beyond expenditures. But perhaps more striking than what is provided from monetary indicators of the extent of inequality is the vulnerability inequality. of the least wealthy. Wealth provides means Nonmonetary outcomes are also very to smooth consumption in the short run unevenly distributed in South Asia. A com- and to raise it in the long run. Wealth also parison of health status across population gives a sense of security. For a typical Indian quintiles, defined by a wealth index, is reveal- household among the top 10 percent, the net ing in this respect. Gaps in neonatal mortality worth could support consumption for more (death within the first 28 days of life) and in than 23 years. For a typical Indian household under-five child mortality (death within the in the bottom 10 percent, however, the net first five years of life) between the top and the worth was sufficient to support consumption bottom quintiles are large, especially in India for less than three months (figure 7). and Pakistan (figure 8). For children who live, Nonmonetary dimensions of inequality the main challenge is to be well nourished. capture the dispersion in human capabilities, Children belonging to the poorest quintile are as reflected, for example, in health and educa- more likely to be stunted in every country in tion outcomes. Differences in these outcomes the region, although the gap is relatively less can affect individuals’ abilities to do what glaring in Maldives and Sri Lanka. they would value doing and to convert differ- Inequality in educational attainment is ent means into well-being (Sen 1980, 1992). large as well, although varying widely across FIGURE 7 The least wealthy are alarmingly vulnerable 1,800,000 300 1,600,000 250 1,400,000 Months of consumption 1,200,000 200 Indian rupees 1,000,000 150 800,000 600,000 100 400,000 50 200,000 0 0 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th Asset-holding deciles, lowest to highest Average asset holding Average net worth Average net worth (Indian rupees) (Indian rupees) (months of consumption) Source: Based on NSS 2002–03. 10 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 8 Inequality in health outcomes is wide Under-five mortality Bangladesh 1st (poorest) 2nd 3rd 4th 5th (richest) India 1st (poorest) 2nd 3rd 4th 5th (richest) Maldives 1st (poorest) 2nd 3rd 4th 5th (richest) Nepal 1st (poorest) 2nd 3rd 4th 5th (richest) Pakistan 1st (poorest) 2nd 3rd 4th 5th (richest) Sri Lanka 1st (poorest) 2nd 3rd 4th 5th (richest) 0 20 40 60 80 100 120 140 160 180 200 Mortality rate (deaths per 1,000 births) Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka. Note: Under-five mortality rate is the number of deaths to children younger than five years per 1,000 live births. the region. From an international perspective, learning evaluations is generally low, consid- countries in South Asia seem to lie at both erable dispersion occurs in test scores across ends of the developing-country range. Among students from different backgrounds. For the countries for which comparable data are India, the inequality in learning outcomes available, Maldives and Sri Lanka exhibit can be seen by comparing test scores of chil- the lowest gaps in educational attainment dren whose households have both a radio and among young adults between the popula- a TV to those who have neither. The mean tion quintiles with the highest and the low- test scores for students in the fi rst group are est expenditures per capita (figure 9). At the higher across the entire distribution than for other end of the spectrum, gaps in Nepal, those from the second group. Gaps in learn- Afghanistan, and especially Bhutan are larger ing outcomes are large in other countries of than in all comparators. However, the inter- the region as well (Dundar and others 2014). national comparison is highly sensitive to the Trends in inequality also vary depend- age group considered because educational ing on the indicator considered. In South attainment is expanding rapidly throughout Asia, inequality in consumption per capita is the region. Gaps between children are being generally higher in the richer countries, and eliminated at the primary level. it has increased over time as countries grew For the younger South Asian cohorts, richer. This is consistent with growth being inequality in relation to education is increas- effective at reducing poverty in the region, ingly driven by quality rather than by access. as it was earlier in East Asia. Moreover, Although average performance in systematic South Asian countries have lower levels of OVERVIEW 11 FIGURE 9 Schooling among young adults is highly unequal in some countries in South Asia 6 ratio of richest to poorest quintile 5 Education attainment, 4 3 2 1 0 Ug nes Vi da Rw a ia In ria ka Za ia ng a Co blic ep u Ni h ilip a da Sr ives ca zil m a G n an a Bh n gh pal Pa a di Ba han Ph bi bi n R er es gu ta ny liv a s an na a an ge an In ne ut pi ist m m P Br kis e lad u Bo ald Ke ra et N iL an do lo M Ni Af ica in m Do Sources: Based on National Risk and Vulnerability Assessment (NRVA) 2007 for Afghanistan; Household Income and Expenditure Survey (HIES) 2010 for Bangladesh; Bhutan Living Standards Survey (BLSS) 2007 for Bhutan; NSS 2009–10 for India; HIES 2009–10 for Maldives; Nepal Living Standards Survey (NLSS) 2010 for Nepal; Household Integrated Economic Survey (HIES) 2010–11 for Pakistan; HIES 2009–10 for Sri Lanka; and World Bank Education Equality Country Profiles database, http://datatopics.worldbank.org/Education/wDHS/HProfiles.aspx. Note: Educational attainment is measured in years of schooling. The population considered are 20 to 29 years of age. inequality than East Asian countries had of the poorest 20 percent of the popula- when they were at similar levels of income per tion increases by 1.057 percent. The corre- capita. However, in both regions, countries sponding figure for the poorest 40 percent with a higher income per capita are charac- of the population is 1.004 percent. None terized by greater inequality (figure 10). of these estimates is significantly differ- Of course, no mechanical relationship ent from 1.0, meaning that grow th is exists between growth and inequality. A overall neutral with respect to distribu- well-known hypothesis in development tion. Breaking the results down by region economics—known as the Kuznets curve— uncovers some interesting variation, how- is that inequality initially increases as coun- ever. For the combined East and South tries grow into middle-income levels and Asia regions, the estimate is substantially then decreases as they become richer. But lower than 1.0 in both the 1990s and the the empirical evidence on this relationship 2000s (and significantly so in the 1990s). is mixed (Milanovic 2011). In addition, Consistent with the Kuznets curve hypoth- the extent of inequality depends on policy esis, monetary indicators of inequality have choices and not just on some economic fate. increased in the poorest countries in the That said, a rigorous statistical analy- South Asia region in recent years, whereas sis of the available microeconomic data they have decreased in the richest ones across countries suggests that grow th (figure 11). However, the only two countries has been more propitious to increasing for which a decrease is observable have a inequality in South Asia than in other combined population of less than 1 million regions (Dollar, Kleineberg, and Kraay people in a region accounting for a fourth of 2013). For the world as a whole, when mankind. The vast majority of South Asians average consumption per capita increases have experienced an increase in inequality, by 1 percent, the consumption per capita sometimes at a fast pace. 12 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 10 Richer countries tend to be more unequal in both South Asia and East Asia a. South Asian countries b. Selected East Asian countries 0.45 0.45 0.40 0.40 0.35 0.35 Bhutan China 0.30 0.30 Philippines Nepal MLD index MLD index 0.25 Maldives 0.25 Cambodia Thailand Vietnam 0.20 Sri Lanka 0.20 Indonesia India 0.15 Pakistan 0.15 Lao PDR Bangladesh 0.10 Afghanistan 0.10 0.05 0.05 0.00 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 GDP per capita (PPP, constant 2005 GDP per capita (PPP, constant 2005 international dollars, thousands) international dollars, thousands) Sources: For South Asia, based on NRVA 2005 and 2007 for Afghanistan; HIES 2000, 2005, and 2010 for Bangladesh; BLSS 2003 and 2007 for Bhutan; NSS 1993–94, 2004–05, and 2009– 10 for India; HIES 2002–03 and 2009–10, and Vulnerability and Poverty Assessment (VPA) 2004 for Maldives; NLSS 1995 and 2010 for Nepal; HIES 2001–02, 2004–05, 2007–08, and 2010–11 for Pakistan (Punjab); and HIES 1995–95, 2002–03, 2006–07, 2009–10 for Sri Lanka. For East Asia, based on the World Bank’s World Development Indicators database, http:// data.worldbank.org/data-catalog/world-development-indicators and PovcalNet tool, http://iresearch.worldbank.org/PovcalNet/index.htm: 1994–2009 for Cambodia; 1990–2009 for China; 1984–2010 for Indonesia; 1992–2008 for Lao PDR; 1985–2009 for Philippines; 1981–2010 for Thailand; and 1993–2008 for Vietnam. Note: MLD = mean log deviation; PPP = purchasing power parity. Trends in nonmonetary inequality are education, or have decreased. The decline in more mixed. Health outcomes show signs inequality is remarkable for the population at of widening inequality. The ratios between large (adults 15 to 65 years of age) as well, the neonatal and under-five child mortal- with the exception of Afghanistan. ity rates of the population quintiles with the highest and the lowest expenditures per cap- ita have either stayed constant or increased Drivers of inequality in most countries. Bangladesh had a reduc- The diversity of the observed trends, depend- tion in inequality in neonatal mortality. All ing on the indicator considered, makes other countries for which data are available relying on a single metric to assess changes at two points in time, on all indicators, show in inequality or progress toward shared pros- an increase in inequality in health outcomes. perity difficult. This difficulty comes on top The increase is especially marked in the case of the measurement issues associated with of stunting. some of the most common indicators. For In contrast, inequality in education out- instance, lower survey response rates among comes has unambiguously been narrow- better-off households and greater underesti- ing. Gaps in educational attainment among mation of their expenditures by the available children 6 to 11 years of age either have been survey instruments may bias standard mon- stable, mainly in countries that have achieved etary measures such as the Gini index or the close to universal coverage of primary income share of the bottom 40 percent. OVERVIEW 13 Even in the absence of bias, the story on FIGURE 11 Monetary inequality is increasing across most of inequality in South Asia would be different South Asia if it were told based on consumption indi- cators, on wealth indicators, or on health 0.80 indicators. At the same time, the diversity of assess- 0.60 Annual change in MLD index (percentage points per year) ments is welcome because it provides useful 0.40 insights on the drivers of inequality in South Asia. High inequality in health outcomes 0.20 is suggestive of limited access to basic care 0 among the poor, especially during pregnancy and childhood. High inequality of wealth –0.20 is to be expected in the presence of sizable –0.40 rents for the few and not much redistribu- tion toward the many. Moderate inequality –0.60 s an h ka a n l an pa ive di in consumption indicators, in contrast, could es ta an ut ist In Ne kis lad ald Bh iL an Pa ng reflect substantial mobility, allowing house- M Sr gh Ba Af holds from disadvantaged backgrounds and their offspring to prosper. Sources: Based on NRVA 2005 and 2007 for Afghanistan; HIES 2000 and 2010 for Bangladesh; BLSS A cursory analysis also reveals some 2003 and 2007 for Bhutan; NSS 1993–94 and 2009–10 for India; NLSS 1995 and 2010 for Nepal; HIES st rong — and somewhat pred ic table — 2002–03 and 2009–10 for Maldives; HIES 2001–02 and 2010–11 for Pakistan (Punjab); and HIES 1995–95 and 2009–10 for Sri Lanka. patterns of the impact of individual charac- Note: MLD = mean log deviation. teristics on inequality. Simple decompositions of consumption inequality between popula- tion groups suggest that characteristics such In Brazil, according to the Gallup World as educational attainment, location, or eth- Poll, satisfaction with access to public services nicity matter considerably. Both education is higher among poorer population groups, gaps and the rural-urban divide account for and a negative correlation exists between a growing share of consumption inequality. income per capita and expectations of a bet- The share is smaller in the case of ethnicity, ter life. Poorer population groups are also but caste remains relevant in northern and more satisfied with government efforts to help eastern Indian states. them. Taken together, these responses reveal The impact of these individual charac- a positive view of equality of opportunity, teristics on inequality can be mediated by upward mobility, and targeted support. All economic structures and public policies. of this is at odds with responses to the same However, opinion polls show that in South questions in South Asian countries (figure 12). Asia, the contribution public policies make The role played by individual character- is often seen under a negative light. The istics such as education, location, and caste Gallup World Poll asks respondents about provides some clues about what lies behind their satisfaction with basic services, their inequality. Opinion polls, in turn, are infor- assessment of future well-being, and their mative on how public policies may offset views on government efforts to help the or amplify the contribution from inherited poor. The six South Asian countries covered circumstances. by the Gallup World Poll can be compared A simple conceptualization of how these to Brazil, a country that has experienced a different variables come into play involves a substantial reduction in inequality in recent person’s life cycle (figure 13). Circumstances years. Moreover, this reduction is generally at birth, such as gender and caste, shape the attributed to policy changes (Barros and oth- options available to individuals, especially in ers 2010; de Souza 2012; Ferreira, Leite, and relation to the accumulation of human capi- Litchfield 2008). tal. As people age and enter the labor force, 14 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 12 South Asians do not see an environment conducive to lower inequality a. Access to public services 100 Health care 50 Satisfaction with access to public services (percent) 0 100 Education 50 0 100 Water 50 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil b. Future well-being 2.5 Views on future well-being relative to present 2 1.5 1 0.5 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil Bottom 40% Middle 40% Top 20% (continues next page) OVERVIEW 15 FIGURE 12 South Asians do not see an environment conducive to lower inequality (continued) c. Efforts to help the poor 60 Satisfaction with efforts to help the poor (percent) 50 40 30 20 10 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil Bottom 40% Middle 40% Top 20% Source: Based on Gallup World Poll. Note: Population groups are defined based on income or consumption per capita. Views on present (future) well-being are assessed on a scale from 1 (dis- satisfied) to 10 (satisfied). FIGURE 13 Multiple factors affect household outcomes relative to others in society Drivers of inequality MOBILITY Jobs Cities OPPORTUNITY Health Education SUPPORT CIRCUMSTANCES Social protection Gender Taxes and transfers Caste At birth During youth Throughout life Life cycle 16 ADDRESSING INEQUALITY IN SOUTH ASIA their job opportunities and the possibility of include clean water, improved sanitation, and reaping benefits from migration affect earn- electricity. ings prospects. Throughout life, differences The coverage of basic services is a first and in innate abilities, inherited wealth, and telling indicator of access. When the cover- shocks affect—positively or negatively— age of a specific service is universal, every- multiple aspects of well-being. In each of body has access to it; lower coverage rates these phases, public policies— explicitly or mean that some are necessarily excluded. implicitly—reduce or amplify the extent of Exclusion is typically not random, however. inequality. People from disadvantaged backgrounds are In sum, opportunity in childhood can be less likely to get access to services. The extent shaped by access to basic services, including to which access varies across clearly distinct health, education, and infrastructure; mobil- groups—defi ned along gender, ethnicity, or ity during adult life can be enhanced by eco- caste lines—is significant in this respect. The nomic growth and rapid urbanization; and greater the dispersion of access across groups, throughout life, support as government poli- the greater is the degree of inequality for the cies transferring resources to people or tax- same level of coverage. This simple intuition ing them can help mitigate shocks and offset is captured by a synthetic indicator, called disadvantages. the Human Opportunity Index (HOI). The HOI is computed by multiplying the coverage rate by a measure of the dispersion of access Limited opportunity across the relevant groups (Paes de Barros Equality of opportunity is considered a key and others 2009). The HOI varies from zero, condition for a society to ensure distribu- when nobody has access to services or the tional justice. Important outcomes—such dispersion is extremely high, to 100, when as income or health status—are seen as everybody has access, and it increases when determined by two main factors: efforts coverage expands or becomes more equitable and circumstances. Equality of opportunity across groups. requires compensating people for disadvan- Following the literature, this report uses tages related to circumstances so the distribu- characteristics that are more likely to be pre- tions of outcomes can be entirely attributed determined, hence unaffected by a child’s to efforts (Roemer 1998). Both conceptually own actions, to defi ne groups. These char- and empirically, completely distinguishing acteristics are the place of residency, either efforts from circumstances is difficult, hence urban or rural; a child’s gender, religion, and impeding disentanglement of opportunities caste; and the education level of the house- from outcomes (Kanbur 2009). However, hold head or the child’s mother, depending making access to basic services universal is at on data availability. Religion is used only the core of equality of opportunity (Paes de when data are available, and caste is used Barros and others 2009; World Bank 2005b). only in India’s case. General agreement exists that the set of In South Asian countries, despite the goods and services that every individual widespread commitment to rights, access to under 16 years of age should have access to services related to health and nutrition tends includes nutrition, health care, basic edu- to be limited (figure 14, panel a). The value of cation, and some forms of infrastructure. the HOI is worryingly low in the case of full Within health and nutrition, the focus is often immunization against vaccine-preventable on institutional births and full immunization. diseases among children two years of age and In basic education, opportunity is associ- younger. The HOI of most countries in the ated with primary school attendance and region does not cross the 50 percent mark. completion and, in some cases, with second- India and Pakistan perform poorly, with an ary school attendance. The forms of infra- HOI below 20 percent in the most recent year structure deemed essential for opportunity for which data are available. OVERVIEW 17 Access to primary education is far better. FIGURE 14 Opportunities in education are better than in health or Countries in the region have generally done sanitation, as measured by the HOI well in primary school attendance and even on completion (figure 14, panel b). In this, a. Full immunization South Asia resembles other regions, reflecting Bangladesh the global drive toward universal enrollment in primary education. The picture is less India encouraging for access to secondary school, Maldives especially in comparison with countries at a Nepal similar level of development. The HOI asso- ciated with secondary school completion Pakistan is below 50 percent across the region, with Sri Lanka the exception of Bhutan. Even in the best- 0 10 20 30 40 50 60 70 80 90 100 performing countries of the region—Bhutan, HOI (percent) Maldives, and Sri Lanka—the HOI is smaller than that in traditionally inequitable coun- b. Primary school completion (14–18 years of age) tries, such as Brazil and South Africa. International comparisons are less reliable Afghanistan with respect to infrastructure services. What Bangladesh it means to have access to improved water or to improved sanitation varies from one coun- Bhutan try to another. Moreover, access is often mea- India sured at the community level in South Asian Maldives countries but at the household level in others. If a power line arrives to a village, but only Nepal half the population in the village has electric- Pakistan ity, coverage is twice as high when measured Sri Lanka at the community level rather than the house- hold level. 0 10 20 30 40 50 60 70 80 90 100 Even when considering the more generous HOI (percent) access measure, at the community level, cov- erage of infrastructure services is low in most c. Improved sanitation South Asian countries (figure 14, panel c). Bangladesh Access to sanitation is generally dismal. With the exception of Maldives and Sri Lanka, Bhutan the HOI for improved sanitation services India does not exceed 40 percent in South Asian Maldives countries. Access to electricity fares bet- ter in Maldives and Sri Lanka. In contrast, Nepal Afghanistan still lags far behind, with an Pakistan HOI of about 10 percent. Access is also lim- Sri Lanka ited in Bangladesh and India, where the HOI hovers between 40 and 60 percent. 0 10 20 30 40 50 60 70 80 90 100 Opportunities in access to health and HOI (percent) education services have been improving in the region over the past decade. Whereas Sources: Based on DHS 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 most cou nt r ie s have reg istered HOI for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka for health; NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, BLSS 2007 for Bhutan, NSS 2009–10 for India, HIES 2009–10 for Maldives, increases in access to health services, prog- NLSS 2010 for Nepal, HIES 2010–11 for Pakistan, and HIES 2009–10 for Sri Lanka for education; and ress has been slower than for other basic ser- data from Andres and others 2013 for this report for infrastructure services. Note: The HOI increases with equality in access across population groups. The HOI for full vices. When considering full immunization, immunization is computed based on formal records instead of patient recall. 18 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 15 Better opportunity is driven by greater coverage a. Full immunization b. Primary school completion (14–18 years of age) Bangladesh Bangladesh Bhutan India India Maldives Nepal Nepal Pakistan Pakistan Sri Lanka –2 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 18 20 Annual change in HOI (percent) Annual change in HOI (percent) Sources: Based on DHS 1993 and 2011 for Bangladesh, DHS 1992 and 2005 for India, DHS 1996 and 2011 for Nepal, and DHS 1990 and 2007 for Pakistan for health; based on HIES 2005 and 2010 for Bangladesh, BLSS 2003 and 2007 for Bhutan, NSS 1993–94 and 2009–10 for India, HIES 2002–03 and 2009–10 for Maldives, NLSS 2003 and 2010 for Nepal, HIES 2001–02 and 2010–11 for Pakistan (Punjab), and HIES 2006–07 and 2009–10 for Sri Lanka for education. Note: Horizontal bars indicate the total change in the HOI; vertical lines indicate the change caused by increased coverage; the difference between the two reflects changes in equity. oppor t u nities increased over time in and reductions in the dispersion of cover- Bangladesh, Nepal, and India. By contrast, age rates across groups (equity effect). The they declined slightly in Pakistan. As for decomposition shows that greater coverage institutional births, Nepal registered the clearly drives the improvements of opportu- fastest improvement. But overall progress nities in health (figure 15). In the case of full has been slow, and inequality has declined immunization, almost all of the change in only slightly in recent times. The annual the HOI can be attributed to changes in cov- change in HOI stands between 0.5 and 1 erage, with inequality across groups remain- percentage point in most cases. ing stable. The decomposition of changes in Countries in South Asia have also made the HOI yields a more diverse picture in the significant strides in improving access to case of primary education. In Afghanistan, primary education. Maldives and Sri Lanka which showed rapid increases in its HOI led the region in achieving almost univer- for primary school attendance, the change sal primary education coverage. Progress has come mainly from greater coverage. In in Bangladesh, India, Bhutan, Nepal, and contrast, changes in equity in access play an Pakistan has also been significant, although important role in Nepal, a country that has at varying speeds. In Bhutan and Nepal, recorded notable growth in primary school improvements in primary school atten- attendance and completion. dance and completion were remarkable. In Although the measured access to health Bangladesh, the pace of change is slower, and education services in the region has but the starting point was higher (with HOIs generally improved, the magnitude of the around 85 percent in 2005). In Afghanistan, improvement may overestimate the improve- school attendance grew between 2005 and ment in equity. This is because the quality 2007, but the HOI remains low. of health and education services is likely to The drivers of inequality in access in vary considerably across population groups, South Asia can be better understood by and the quality of services has important decomposing the change in the HOI between implications for later-life opportunities. increases in coverage rates (scale effect) An improvement in HOI is unlikely to capture OVERVIEW 19 the fact that qualitative differences in services and completion across countries in the may not have diminished substantively. region. Location turns out to be a critical S evera l popu lat ion g roups receive circumstance for access to infrastructure systematically lower coverage of basic ser- services. vices because of their circumstances. Not sur- Several factors underlie South Asia’s lack- prisingly, location matters. Children residing luster performance in ensuring equality in in rural areas fare worse than those in urban access to basic services. Importantly, public areas with regard to basic health services, spending on education and health is relatively especially for institutional birth. The gap is often quite striking—in Nepal, for example, 32 percent of rural births are in a health FIGURE 16 Parent’s education and location are critically facility, compared with 71 percent of urban important circumstances births. Similar differences can be found in Bangladesh, India, and Pakistan. a. Full immunization The urban-rural gap is also evident in the Bangladesh provision of infrastructure services in the region, especially electricity. Rural areas fare India much worse in terms of access to electricity, particularly in Afghanistan and Bangladesh. Maldives Gender is another dimension along which important differences in coverage may exist. Nepal But this does not appear to be the case in South Asia, where boys’ advantage is gener- Pakistan ally very small for basic health services and primary education. Gender gaps are much Sri Lanka more pronounced at the secondary education 0 10 20 30 40 50 60 70 80 90 100 level, however. In Sri Lanka and Bangladesh, Contribution to measured inequality (percent) girls attend secondary school at a marginally higher rate than boys, but in Bangladesh, the Caste Gender Mother’s education female advantage in participation does not Religion Urban/rural continue through school completion. In gen- b. Primary school completion (14–18 years of age) eral, secondary school completion rates are 5 to 10 percentage points higher for boys than Afghanistan for girls. Bangladesh These analyses compare the coverage of basic services along a single dimension: Bhutan location, gender, or mother’s education. India But disadvantaged children are often disad- vantaged along several of those dimensions Maldives simultaneously. Overall, parent’s education Nepal and location are among the most important Pakistan circumstances behind inequality in access to health, education, and infrastructure Sri Lanka services (fi gure 16). In Bangladesh, India, 0 10 20 30 40 50 60 70 80 90 100 Nepal, and Sri Lanka, religion also explains Contribution to measured inequality (percent) some part of the inequality in access to pri- mary education. In India, caste explains Caste Gender Household head’s Religion Urban/rural education more than religion. In the case of secondary education, gender plays a significant role in explaining secondary school attendance (continues next page) 20 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 16 Parent’s education and location are critically important curative care tends to be regressive (World circumstances (continued) Bank 2003). In education, public spending tends to be c. Access to improved sanitation progressive at lower grade levels but regressive Bangladesh at secondary and especially tertiary levels. For instance, the poorest 40 percent of the Bhutan population in Bangladesh received 50 percent of public primary spending in 2010, up from India 32 percent in 2005. However, the richest Maldives 40 percent received about 80 percent of pub- lic spending directed to tertiary education Nepal (World Bank 2003, 2013a). A similar pattern Pakistan is found in India and Pakistan. Sri Lanka Substantial mobility 0 10 20 30 40 50 60 70 80 90 100 Economic mobility is an avenue to long- Contribution to measured inequality (percent) term equality and a source of efficiency Caste Urban/rural Household head’s education (Schumpeter 1955). At any point in time, differences exist in household well-being. Sources: Based on DHS 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 But they can be partially offset if the choices for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka for immunization; NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, BLSS 2007 for Bhutan, NSS 2009–10 for India, HIES 2009–10 households make on where to live and what for Maldives, NLSS 2010 for Nepal, HIES 2010–11 for Pakistan, and HIES 2009–10 for Sri Lanka for kind of work to do are more rewarding, in rel- education; and Andres and others 2013 for this report for infrastructure. ative terms, for the most disadvantaged. On the efficiency front, mobility leads to a better low in comparison with other countries at a use of talent and strengthens incentives. If the similar level of development. A meaningful distribution of creativity or resourcefulness assessment involves comparing public spend- across the population is less unequal than ing on health or education as a fraction of the distribution of income or consumption, GDP, rather than as a share of the budget, societies with greater mobility may be able to and controlling for the level of economic mobilize the talent of all population groups. development. This assessment shows clearly In a society where the poor and the rich that except for Bhutan and Maldives, coun- are equally likely to succeed or fail, people tries in the region spend much fewer public belonging to all groups have a higher moti- resources on health services than could be vation to work hard. Mobility fosters aspira- expected, given their income per capita. tion, efforts, innovation, and self-fulfillment. South Asian countries also spend less on Although economists tend to think of education than other countries at a similar mobility in terms of income and consumption, level of development. economic and social mobility are intertwined, Equity requires committing additional especially through jobs. In most societies, resources to disadvantaged groups to offset jobs are fundamental sources of self-respect their otherwise more limited access to basic and social identity. The distribution of jobs services. However, the opposite is often within society and perceptions about who observed in South Asia. In health, public has economic mobility can shape individu- spending is directed more toward the better- als’ perceptions of fairness and aspirations for off than to the poorest population groups the future (Akerlof and Kranton 2010; World (O’Donnell and others 2007). The extent of Bank 2012b). In South Asia, marginalized progressivity also varies across types of ser- population groups traditionally suffer from vices. In Bangladesh, for example, spending both material poverty and social indignity. on child care is clearly progressive, whereas This situation is most obvious in the case of OVERVIEW 21 the caste system, in which occupations are FIGURE 17 Considerable occupational mobility exists across essentially set for individuals at birth. generations in India In a perfectly mobile society, by contrast, 80 occupational choices should be independent Occupations of sons (percent) 70 across generations. For instance, the children 60 of manual and nonmanual workers would 50 have similar prospects of securing nonman- 40 ual jobs. If so, growth should allow the peo- 30 ple from marginalized groups to move into 20 the mainstream of society. 10 This intuition of occupational indepen- 0 dence can be followed to assess the extent Unskilled Farmers Skilled/ White collar of mobility across generations. Because of semiskilled limitations of data availability, a measure of Occupations of fathers occupational independence across genera- Downward mobility Persistence Upward mobility tions in South Asia can be computed only for India. There, a unique father-son matched Source: Based on India Human Development Survey (IHDS) 2004–05. data set exists, and it is based on a nationally representative sample. Data on self-reported occupation in the preceding year are avail- FIGURE 18 Occupational mobility is higher for younger able in the case of sons; in the case of fathers, generations data refer to occupation for most of life as mobile society (Altham statistic) reported by sons. For both generations, occu- 22.5 “Distance” from a perfectly pations are mapped into four categories in a 22.0 broadly ascending order of economic returns 21.5 and levels of social prestige: unskilled work- ers, farmers, skilled or semiskilled workers, 21.0 and white-collar workers. 20.5 The analysis reveals considerable occupa- tional mobility across generations (figure 17). 20.0 1945–54 1955–64 1965–74 1975–84 The sons of unskilled fathers and those of farming fathers both saw significant pros- Birth cohorts of sons pects of moving to higher-ranked jobs in Source. Based on IHDS 2004–05. terms of economic returns and social prestige. Note: “Distance” is measured as the Altham statistic comparing the actual transition matrix to that of Over 40 percent of the children of unskilled a hypothetical, perfectly mobile society. workers were holding other occupations. About 36 percent of the children of farmers worked as skilled or semiskilled workers or children of people in basic occupations have as white-collar workers. seen rising prospects of taking higher-ranked Occupational mobility across genera- occupations relative to the children of people tions has also increased over time (figure 18). in higher-ranked occupations. The conclu- The occupational transition matrix between sion appears to be robust to the classification fathers and sons of each birth cohort can of occupations and the methodology used for be compared with the hypothetical transi- the comparison across cohorts (Singh and tion matrix of a perfectly mobile society. Motiram 2012). The smaller the “distance” between the two The most notable improvements in mobil- matrices, the greater the mobility (Long and ity are found for Scheduled Castes, Scheduled Ferrie 2013). Between the first (1945–54) Tribes, and Other Backward Castes. The and the last (1974–84) cohorts, this distance transition matrices of socially marginal- has declined significantly. On average, the ized groups are compared in this case with 22 ADDRESSING INEQUALITY IN SOUTH ASIA those of higher-caste Hindus for each cohort. percent of the poor and vulnerable—moved The comparison shows that mobility among into the middle class. Households from Muslims has been similar to that of higher- Scheduled Castes and Scheduled Tribes expe- caste Hindus, whereas mobility among rienced upward mobility comparable to that Scheduled Castes and Scheduled Tribes and of the rest of the population. In Bangladesh, among Other Backward Castes has become between 2005 and 2010, about 19 percent of higher than that of higher-caste Hindus the total population, or half the poor, moved over time. Again, the conclusion appears above the poverty line, and about 13 percent to be robust to the choice of methodology of the total population, or over 15 percent (Hnatkovska, Lahiri, and Paul 2013). of the poor and vulnerable, moved into the A truly mobile society is arguably one middle class. Downward mobility was also in which poorer households can manage to considerable in both countries, however. climb up the income or consumption ladder By these measures, upward mobility through their own efforts within a single gen- within a generation in Bangladesh and India eration. Mobility of this sort requires that the was comparable to that of dynamic societ- growth in the income or the consumption ies such as the United States and Vietnam of the poor be faster than the growth of the (figure 19). Per capita consumption is higher average person. When splitting the total pop- and poverty incidence is lower in the United ulation by income status, a fraction should States and Vietnam than in South Asian be able to move above the poverty line while countries, but over a comparable period, some should be able to make solid progress the four countries saw similar fractions of into the middle class. the poor moving above the poverty line Mobility within a generation can be and a considerable fraction of the poor and assessed by comparing the income or con- vulnerable moving into the middle class. sumption of the same households between Downward mobility was much bigger in the two periods (Fields 2010; Fields and Ok two South Asian countries, however, reveal- 1996). But doing so requires nationally repre- ing the greater risks faced by the vulnerable sentative information on the same households and even the middle class. at two points in time—so-called panel data— At the level of villages, increasing mobil- which is generally not available for South ity is largely associated with occupational Asian countries. To overcome this difficulty, change. The timing and the pace have var- three synthetic panel data sets were built ied across countries, but the shift has con- especially for this report, linking different sistently involved an expansion of nonfarm rounds of nationally representative surveys. employment. While the new jobs are mainly Two of the synthetic panels are for India; casual, they have supported considerable they cover the periods between 1993–94 mobility. Wages of casual nonfarm workers and 2004–05, and between 2004–05 and were 30 percent to 50 percent higher than 2009–10. The third panel is for Bangladesh agricultural wages in rural India, Nepal, between 2005 and 2010. and Pakistan in the 2000s; they were 10 When splitting the population into three percent higher in rural Bangladesh dur- groups—poor, vulnerable, and middle ing the fi rst half of the 2000s (World Bank class—upward mobility was considerable for 2011). Although regular jobs tend to pay both the poor and the vulnerable. In India, better, the earnings gap between regular although some households fell into poverty and casual nonfarm jobs has narrowed over between 2004–05 and 2009–10, more of time in rural India, whereas the earnings them, about 15 percent of the total popula- gap between casual nonfarm jobs and agri- tion or 40 percent of the poor, moved above cultural jobs has increased (Himanshu and the poverty line. Meanwhile, a sizable pro- others 2013). portion of the poor and the vulnerable—over Internal migration has also helped South 9 percent of the total population or about 11 Asians find better jobs and investment OVERVIEW 23 FIGURE 19 Upward mobility in South Asian countries is similar to that in the United States and Vietnam 60 Share of consumption group (percent) 50 40 30 20 10 0 Bangladesh India United States Vietnam Bangladesh India United States Vietnam Bangladesh India United States Vietnam Bangladesh India United States Vietnam Moving out of poverty Moving up to middle class Falling back to poverty Falling out of middle class Sources: Based on Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The groups considered are the poor for moving out of poverty, the poor and the vulnerable for moving up to middle class, the nonpoor for falling back to poverty, and the middle class for falling out of middle class. opportunities and achieve economic mobil- City dynamism is an important driver of ity. Among men, internal migration is the increasing mobility observed in South primarily a mechanism to fulfi ll aspirations Asia. Mobility both across and within gen- for employment and investment opportu- erations is greater in urban than in rural nities (table 1). For women, it is said to be areas. In India, the sons of unskilled fathers associated to a greater extent with marriage in urban areas face a lower probability of and social reasons. However, special micro- staying in the same occupational category surveys focusing on female migrant workers than their rural counterparts; they also face in 20 Indian states found that a significant a much higher probability of taking on bet- propor tion of unemployed or house- ter jobs. Sons of farmers also face better bound women enter into paid employment prospects in urban areas than in rural areas through migration (Mazumdar, Neetha, and (figure 20). Agnihotri 2011). In Bangladesh, the scale In both Bangladesh and India, within the of female rural migration is such that it has same generation a larger fraction of the pop- changed social norms. About 90 percent of ulation manages to move above the poverty the female workers in urban garment sectors line in rural than in urban areas. Conversely, in the 1990s and early 2000s were estimated a larger fraction of the population makes it to be migrants from rural areas. Although to the middle class in urban than in rural a socially negative image of the garment areas (table 2). Downward mobility in the workers as “fallen women” prevailed in the form of falling below the poverty line is also early 1990s, these migrants effectively defied considerably smaller in urban than in rural and redefined their place in society over time areas. Urban households whose members are (Afsar 2003; Deshingkar and Grimm 2005; employed as regular wage or salaried work- Hossain, Sen, and Sawada 2014). ers experience the highest upward mobility 24 ADDRESSING INEQUALITY IN SOUTH ASIA TABLE 1 Changes in employment status reveal substantial mobility among migrant men in India After permanent migration (percent) Regular Unemployed or not Before permanent migration Self-employed employee Casual labor in labor force Total Self-employed 70.6 18.1 8.7 2.6 100 Regular employee 9.9 83.8 2.6 3.7 100 Casual labor 20.3 28.5 49.6 1.6 100 Unemployed or not in labor force 17.3 34.2 14.4 34.1 100 Total 24.1 39.7 19.2 17.0 100 Source: Based on NSS 2007–08. Note: The migrants considered are adult males 15 years of age and older. The circled cells show the shares of migrants who became regular wageworkers after migration. FIGURE 20 Upward mobility is much stronger in cities than in rural areas a. Urban India b. Rural India 70 Occupations of sons (percent) 90 Occupations of sons (percent) 80 60 70 50 60 50 40 40 30 30 20 20 10 10 0 0 Unskilled Farmers Skilled/ White collar Unskilled Farmers Skilled/ White collar semiskilled jobs semiskilled jobs Occupations of fathers Occupations of fathers Downward mobility Persistence Upward mobility Source: Based on IHDS 2004–05. and the lowest downward mobility. But self- driver of upward mobility. For instance, employment and casual employment also about 18 percent of total urban employ- support substantial improvements in living ment in India is accounted for by men who standards. In both Bangladesh and India, are regular wageworkers in the informal urban households whose members are self- sector and another 16 percent by men who employed or who work as casual labor expe- are casual wageworkers in the informal sec- rience stronger upward mobility and smaller tor. If both men and women are considered, downward mobility than rural households. about 57 percent of the informal workforce in This dynamism, including substantive urban areas earns wages. transitions into the middle class, is taking In Bangladesh, the urban formal sector has place despite the prevalence of informality in been expanding because of the rapid growth South Asia’s urban areas. The urban infor- of labor-intensive manufacturing, particu- mal sector includes a considerable number larly in garments and textiles. Women gained of wage jobs, and they could be the main more than men from the expansion of these OVERVIEW 25 TABLE 2 Rural jobs allow people to escape poverty; urban jobs are a ticket to the middle class Five years later (percent) Poor Vulnerable Middle class Total Bangladesh Rural households 2005 Poor 23.0 17.8 2.8 43.6 Vulnerable 12.4 22.3 8.2 42.9 Middle class 1.5 6.2 5.9 13.5 Total 36.9 46.3 16.9 100.0 Urban households 2005 Poor 9.9 11.9 2.9 24.7 Vulnerable 7.9 22.5 14.1 44.5 Middle class 1.4 10.2 19.1 30.8 Total 19.2 44.6 36.1 100.0 India Rural households 2004–05 Poor 26.3 15.5 0.8 42.7 Vulnerable 9.6 28.7 7.5 45.8 Middle class 0.3 4.7 6.5 11.5 Total 36.2 48.9 14.8 100.0 Urban households 2004–05 Poor 11.2 9.4 0.7 21.3 Vulnerable 7.1 29.3 11.3 47.7 Middle class 0.4 8.8 21.8 31.0 Total 18.7 47.5 33.8 100.0 Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The household head’s age is restricted to between 25 and 55 years on the first survey and adjusted accordingly for the second survey. The circled cells show the shares of the total population who experienced upward mobility. Percentages may not total to 100 because of rounding. industries. About 60 percent of garment defi nitions (UN 2012). However, countries workers were female in 2009; the share might in South Asia are still less urban than other have climbed to about 80 percent by 2012. countries at a similar level of development. Because the demand for female workers in From a mobility perspective, this amounts to the manufacturing sector has increased faster a missed opportunity. than that of male workers, women’s wages Urbanization also appears to be more have increased more rapidly than men’s in organic in South Asia. Whereas people come recent years (Ahmed, Bakht, and Yunus 2011; to cities in the form of migration, cities also Hossain, Sen, and Sawada 2014; Lopez- “come” to people through the densification Acevedo and Robertson 2012; Zhang and of population and the transformation of eco- others 2013). nomic activity in rural areas. These diverse Although urban areas present better pros- urbanization processes have led to a range pects of economic mobility than rural areas, of cities with different characteristics, not both the pace and the pattern of urbaniza- just in terms of their size but also in terms of tion in South Asia are reasons for concern. In their governance structure. These differences absolute terms, the urban population is mas- in size and governance matter for mobility, sive and the rate of urbanization is impres- because they shape the type of jobs available sive. Overall, 563 million people could be across different types of cities. considered urban residents in South Asia In India, districts can be classified depend- in 2011, according to the countries’ official ing on the size of their biggest city. This 26 ADDRESSING INEQUALITY IN SOUTH ASIA information, in turn, can be used to analyze Inadequate support the structure of urban employment at the district level. Districts with larger cities have Households in every country suffer from a higher proportion of urban regular wage- shocks. Most are minor and can be cushioned workers than do districts with smaller cities relatively easily, but some can have long- (figure 21, panel a). Between districts with lasting impacts, adversely affecting nutri- cities of more than 1 million people and the tion, human capital, and asset accumulation. other districts, the difference is large and sig- The vulnerability of households to shocks is nificant. Districts with larger cities also have determined partly by the types of risks they a higher proportion of urban jobs in manu- face. South Asian households are periodically facturing and services relative to districts exposed both to individual shocks and to with smaller cities. economy-wide shocks, such as natural disas- The governance of cities, in addition to their ters, food price spikes, and armed conflict. size, matters for urban economic mobility. In Natural disasters affected more than India, districts can be classified into six cat- 750 million people in the region between egories based on the administrative arrange- 1990 and 2008, resulting in approximately ments of their cities: state capitals, other cities 230,000 deaths and US$45 billion in dam- with municipal corporations, municipalities, ages (World Bank 2009). Bangladesh stands notified areas, nagar panchayat (including out as one of the countries that have expe- census towns), and industrial townships. The rienced the largest losses of human lives first five are in a broadly descending order worldwide. Moreover, the frequency and in terms of administrative autonomy, capac- magnitude of natural disasters are increas- ity, and financial resources of city authorities. ing, mainly because of climate change and The last category, industrial townships, covers accelerated snow melting in the Himalayas areas designated for industrial development (Memon 2012). Food price inflation has that have some of the characteristics of special been the main driver of headline inflation economic zones in other countries. Again, this throughout most of the region, for exam- information can be used to assess how urban ple, the widespread food price inflation of employment varies across city characteristics 2007–08. South Asia is also prone to conflict at the district level. and violence. The region accounted for at The share of regular wage jobs in urban least 40 percent of the world’s terrorist inci- employment broadly declines with the auton- dents in 2009 and 2010 (Global Terrorism omy, capacity, and financial resources of city Database 2009–14). authorities (figure 21, panel b). By contrast, The vulnerability of households also the share of self-employment increases. The depends on their own ability to manage those composition of urban employment in dis- risks. Typically, the poor have a relatively tricts with industrial townships resembles weak capacity to self-insure or to pool risks that of districts with municipal corpora- beyond extended families. Informal mecha- tions. Overall, districts with state capitals, nisms tend to be costly and inefficient, often municipal corporations, or industrial town- breaking down when shocks affect entire ships are associated with significantly greater communities. shares of urban employment in regular wage In Pakistan, poorer and richer households jobs and in all wage jobs. These districts also rely on different coping strategies (figure 22). have a higher share of urban employment in When faced with a shock, a large majority manufacturing and services. One could argue of households in the poorest quintile bor- that urban governance improves as city size row money, reduced expenditures, switched increases, so that the observed relationships to lower-quality food, or reduced the quan- would be misleading. However, the relation- tity of food they consume. Equally impor- ships hold even after controlling for the size tant, 11.5 percent of the poorest households category of the biggest city in the district. reported selling agricultural assets to cope OVERVIEW 27 FIGURE 21 The composition of urban employment varies with city size and governance in India a. City size Less than 50,000 Size of the biggest city in the district 50,000–100,000 100,000–500,000 500,000–1 million 1 million–4 million More than 4 million 0 20 40 60 80 100 Share of urban employment (percent) Regular wage Casual labor Self-employment b. City governance Nagar panchayat Governance category of cities in the district Notified area Municipality Industrial township Municipal corporation State capital 0 20 40 60 80 100 Share of urban employment (percent) Regular wage Casual labor Self-employed Sources: Based on NSS 2009–10 and RGCC 2012. Note: Adult workers considered are 15 years of age and older. 28 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 22 In Pakistan, poorer and richer households cope with shocks in different ways 80 70 Share of the population (percent) 60 50 40 30 20 10 0 e n s s s ey d d d d re av et ce tio oo oo an lan on itu ss ts an ac ll l yf ff la m no nd itt nt yo no Se lit ra w Re em pe Do ua ltu rro tit ke ex er -q an icu Ta Bo er ce or qu gr ow m du ll a ce ive Re ol du Se ht ce Re Re itc Sw Consumption quintiles 1st (poorest) 2nd 3rd 4th 5th (richest) Sources: Based on Pakistan Social and Living Standards Measurement (PSLM) surveys for 2007–08 to 2009–10. with the shock—a strategy that compromises luck is tolerable—or even desirable—is a their long-term ability to earn an income. In question to which different societies give contrast, the wealthiest groups are much less different answers. But most have developed likely to use these mechanisms. mechanisms for redress. Bad shocks can hit even before birth. Government-sponsored mechanisms to Individuals differ in their inherited wealth, cope with shocks are known as social pro- in their talent, in the value they attach to tection; they typically comprise social assis- future well-being relative to present well- tance and social insurance programs. They being, and in their willingness to work hard. are designed to prevent households that are Even children of the same parents can dif- affected by an adverse shock from expe- fer in these respects. In an ideal society with riencing too dramatic a fall in their living equality of opportunity and perfect mobil- standards. Mechanisms to redress deeper ity, these inherited differences would result differences in the fate of individuals and in inequality of outcomes. Given that no households fall under the broader heading society fully meets the ideal, the inequality of redistribution; the instruments used in of outcomes caused by inherited differences this case are taxes, subsidies, and transfers. is often amplified by the legacy of inequities This type of support aims at bringing dis- accumulating throughout life. Private trans- advantaged households to a higher level of fers are unlikely to be sufficient to offset the welfare than they could attain on their own. consequences of these forms of random- The reach, generosity, and efficiency of social ness. To what extent inequality of outcomes protection and redistribution mechanisms resulting from these different forms of bad vary from one society to another and, as a OVERVIEW 29 result, the extent of inequality of outcomes In most other countries, the coverage rate varies as well. of social assistance spending is between 20 South Asian countries have a mixed record percent and 30 percent, though coverage has on both fronts. Public spending on social pro- improved over time. Social assistance pro- tection programs has expanded over time. grams are not very well targeted. Across all Measured as a fraction of GDP, it is in line countries in the region, 60 percent to 80 per- with that of other countries at roughly the cent of the beneficiaries are not poor, and they same development level. On average, devel- receive between 50 percent and 80 percent of oping countries spend 1.53 percent of their the funding. Although some of the nonpoor GDP on safety nets, and with the exception may stand just above the poverty line, a dis- of Sri Lanka, most South Asian countries fall proportionately large fraction of resources is within the 0.25 percent to 2 percent range captured by the better-off. The benefit ade- (Gentilini, Honorati, and Yemtsov 2014). The quacy of social assistance programs remains design of these programs also shows impor- low in most countries considered. tant strengths. The Female Secondary Schools The record is arguably more questionable Stipends Program of Bangladesh is known in the case of redistribution. One of South worldwide as the precursor of modern con- Asia’s salient characteristics is the low level of ditional cash transfers. The Benazir Income its tax revenue relative to GDP, which implies Support Programme of Pakistan is remark- that the potential for the government to make ably well targeted to poorer households and a dent in inequality of outcomes is more lim- has helped them become more resilient to ited than elsewhere. In developing countries, natural disasters. The Mahatma Gandhi the overall tax collection as a percentage of National Rural Employment Guarantee Act GDP tends to be small (Chu, Davoodi, and of India—the largest public works program Gupta 2000). It is even lower than the aver- in the world—has dramatically reduced dis- age in South Asian countries, where govern- tress sales of land in years of drought. ment revenue averages between 10 percent Social protection spending is largely and 15 percent (map 1). The collection rates progressive across the entire region. When fall behind comparable developing countries, assessed on the basis of absolute spending such as Brazil, China, and Mexico. Even per person, social protection programs are after controlling for main structural factors, progressive in Bangladesh, Pakistan, and however, revenue mobilization in most South Sri Lanka. Bangladesh is the most effective Asian countries is still below the average for in directing social protection resources to countries at similar income levels (World the poor. In 2010, the poorest 40 percent Bank 2012a). of households received a little over half the Tax revenue is low not because South social assistance and over 75 percent of the Asian countries rely on unusual tax instru- social insurance. Because poorer households ments but rather because the revenue “pro- consume less, the same level of public spend- ductivity” of those instruments is unusually ing makes a greater difference in their case. low. A standard indicator of such produc- When assessed relative to the consumption tivity is the share of GDP in revenue collec- level of the beneficiaries, social protection tion for every percentage point of the basic spending turns out to be largely progressive tax rate. In countries such as Thailand or in all South Asian countries. Vietnam, the productivity of value added Overall, however, the coverage of social tax (VAT) exceeds 50 percent and that of protection programs in South Asia is par- corporate income tax (CIT) reaches 30 per- tial, their targeting is generally poor, and the cent. In contrast, the productivity of VAT in amount of resources they make available to South Asian countries varies between 20 and those who need them the most is often too 40 percent, and that of CIT hovers around modest (figure 23). In Nepal and Sri Lanka, a a meager 10 percent (World Bank 2010, little more than half the poor receive support. 2012a). Tax avoidance, tax evasion, and the 30 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 23 Social assistance is less adequate than social insurance but has greater coverage a. Afghanistan b. Bangladesh Targeting (percent of total transfers) Targeting (percent of total transfers) 50 116 40 96 30 76 20 56 10 36 0 –10 26 –20 –4 –10 0 10 20 30 40 50 60 –10 0 10 20 30 40 50 60 70 80 90 100 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) c. Maldives d. Nepal 30 100 Targeting (percent of total transfers) Targeting (percent of total transfers) 90 25 80 70 20 60 15 50 40 10 30 20 5 10 0 0 –10 –5 –20 –10 0 10 20 30 40 50 –20 –10 0 10 20 30 40 50 60 70 80 90 100 110 120 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) e. Pakistan f. Sri Lanka 80 70 Targeting (percent of total transfers) Targeting (percent of total transfers) 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 –10 –10 –20 –20 –20 –10 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) Social assistance programs Social insurance programs Remittances Sources: NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, VPA 2004 for Maldives, NLSS 2010 for Nepal, PSLM 2010–11 for Pakistan, and HIES 2006–07 for Sri Lanka. Note: Coverage is the percentage of the population in the bottom two deciles that receives the benefits. Targeting is the sum of transfers received by the bottom two deciles as a percentage of total transfers. Adequacy, measured by the size of the bubble, is the mean value of the transfer amount received by beneficiaries in the bottom two quintiles in per- cent of mean expenditure per capita in that group. Deciles are defined based on expenditure per capita net of social protection transfers. Remittances are used as a benchmark for government-funded social protection programs. OVERVIEW 31 MAP 1 Government revenue in South Asia is low compared with the rest of the world IBRD 41165 SEPTEMBER 2014 This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries. Government revenue, percent of GDP <10 10–19.99 20–29.99 30–39.99 40–49.99 ≥50 No data Source: Based on International Monetary Fund Data Mapper, http://www.imf.org/external/Datamapper/index.php. underreporting of taxable amounts are wide- In India, the Public Distribution System (PDS) spread, as are exemptions and special regimes is responsible for the provision of subsidized favoring the businesses with more clout. All food. It takes the largest share of resources of this results in a tax system of relatively low among all social protection programs: 43 bil- progressivity. lion Indian rupees, compared to around 30 Taxes are only one transfer mechanism billion Indian rupees devoted to Mahatma through which the distribution of income G a nd h i Nat iona l Ru ral E mploy ment or consumption can be modified. From an Guarantee funding (Union Budget 2013–14, individual’s point of view, taxes and public http://indiabudget.nic.in /budget2013-2014 spending can have similar consequences on /budget.asp). In Pakistan, subsidies absorbed income or consumption (although they typi- more than 230 billion Pakistani rupees, or cally affect incentives differently). 1.3 percent of GDP, in fiscal year 2010–11; In South Asia, a substantial share of public in comparison, the Benazir Income Support spending goes into subsidies. Some of them Programme consumed about 34 billion are regressive and can crowd out the provi- Pakistani rupees (Government of Pakistan sion of essential public goods. An unusually 2010; World Bank 2013b). large fraction of the typically low govern- The delivery mechanisms for these food ment revenue is often devoted to reducing the subsidies lend themselves to inefficiencies and final price of food, fertilizer, gas, and electric- leakages. In India, PDS was found to have ity. Equity concerns are among the justifica- strong poverty reduction impacts, account- tions for this type of spending. And in some ing for a significant fraction of the poverty cases, as for food, the justification has merit. decline between 2004– 05 and 2009–10. But in countries where the poor generally Several states have made substantial improve- lack access to the grid and often cook using ments in infrastructure and delivery sys- biomass, subsidizing electricity is bound to tems to plug leakage. However, the coverage be regressive. rates were around 53 percent in rural areas The bias toward food and price subsidies and 33 percent in urban areas in 2011–12. is especially marked in India and Pakistan. Take-up rates were progressive across 32 ADDRESSING INEQUALITY IN SOUTH ASIA quintiles, but coverage rates of richest 20 per- for an urban household in the richest quin- cent in rural areas remained high. The illegal tile is almost 100 percent (Goutam, Lahoti, diversion and leakages were about 44 percent and Suchitra 2012). In the case of Pakistan, by the end of 2007–08 and around 35 per- the poorest 40 percent of households used cent in 2011–12 (Himanshu 2013; Jha and to receive less than 30 percent of total elec- Ramaswami 2010; Khera 2011). In Pakistan, tricity subsidies, while the richest 20 percent both federal and provincial governments received close to 40 percent of total subsi- intervene in markets for food products dies. The distribution of benefits improved through mechanisms such as the “Utility after the October 2013 tariff increase, Store” arrangement. The complexity of these but electricity subsidies remain regressive mechanisms makes it difficult to quantify the (figure 24). Electricity subsidies are regressive full extent of the inefficiencies, but it is clear in Maldives as well. that part of the subsidy is appropriated by A positive note comes from spatial trans- wheat flour millers and traders (World Bank fers to lower-level administrative units, such 2013b). as provinces and districts. In countries with In South Asia, fertilizer subsidies are high levels of informality and important spa- crowding out investments in essential public tial disparities, intergovernmental transfers goods. Agricultural growth critically depends are a more effective tool to reduce inequal- on investments in research and development, ity than either taxes or transfers to individu- extension services, and water and irrigation als. When avoiding or evading rules is easy, infrastructure. Since the assets created by taxes encourage individuals and firms to such investments have the characteristics of remain informal. Both taxes and transfers public goods and thus tend to be underpro- to individuals affect incentives to work and vided by the market, public expenditures accumulate in ways that tend to be detrimen- are pivotal in their provision. In Bangladesh, tal to efficiency. By contrast, intergovern- investment in these sectors has fallen from mental transfers make resources available for 5.2 percent of total public agricultural expen- the provision of public goods in places that ditures to 2.7 percent over less than a decade, would not be able to mobilize the resources mainly because of increased spending on the to pay for them. The relevance of intergov- fertilizer subsidy (World Bank 2010). In Sri ernmental transfers is even greater in coun- Lanka, total public expenditure in agricul- tries that are large and diverse, as is the case ture in 2011 increased by 64 percent over in Bangladesh, India, and Pakistan. 2010 but mainly to finance growing spending Intergovernmental transfers are defined on the fertilizer subsidy (MoFP 2012). by country-specific institutions. In India, the Energy subsidies are arguably even more largest component of fiscal transfers comes regressive than food and fertilizer subsidies. from tax-sharing schemes, but discretion- In Bangladesh and Pakistan, energy subsi- ary transfers and the subsidies together are dies amount to more than 4 percent of GDP, almost as large as the tax shares. The over- a level in line with global energy exporters all system of intergovernmental transfers such as the Arab Republic of Egypt, Qatar, in India is generally progressive and leads and the República Bolivariana de Venezuela. to a more equitable distribution of fiscal In Sri Lanka and India, energy subsidies are resources across constituencies (Ghani, Iyer, about 2 percent, doubling the world average and Misra 2013). The same can be said (IMF 2013). of Pakistan and Sri Lanka, where poorer Energy subsidies disproportionately benefit regions receive higher per capita fiscal trans- the better-off. In the case of the subsidies for fers (Ghani 2010). liquefied petroleum gas (LPG) in India, the The relatively progressive nature of inter- average household in the poorest quintile has governmental transfers does not imply, how- less than a 20 percent probability of using ever, that public development spending per LPG; in contrast, the average probability person is progressive (figure 25). Subnational OVERVIEW 33 FIGURE 24 Electricity subsidies favor the better-off a. Maldives b. Pakistan 35 40 subsidies (percent) subsidies (percent) Share of electricity Share of electricity 30 25 30 20 20 15 10 10 5 0 0 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th (poorest) (richest) (poorest) (richest) Consumption quintiles Consumption quintiles FY2013 After October 2013 tariff increase Sources: Redaelli 2013; Trimble, Yoshida, and Saqib 2011. Note: FY2013 = Fiscal Year 2013. FIGURE 25 Development spending per person is lower in poorer states and districts a. Districts in Bangladesh b. States in India 4 2,000 Spending per capita (thousand takas) Spending in social services per capita 3.5 Maharashtra Kerala 3 Punjab 1,500 (Indian rupees) 2.5 Gujarat RajasthanKarnataka Haryana 2 Tamil Nadu 1.5 Jharkhand Andhra Pradesh 1,000 Odisha West Bengal 1 Bihar Chhattisgarh 0.5 Madhya Pradesh 500 Uttar Pradesh 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 10,000 20,000 30,000 40,000 Poverty headcount ratio (percent) GDP per capita (Indian rupees) Sources: Iyer, Ghani, and Mishra 2010; World Bank 2010. governments in poorer areas tend to have often leading to leakages or unspent funds much less locally generated revenue than (Murgai and Zaidi 2005). those in more affluent parts of the country. In addition, their capacity to spend their References resources is more limited. For instance, a Afsar, Rita. 2003. “Internal Migration and the study covering 533 blocks in Bihar—India’s Development Nexus: The Case of Bangladesh.” poorest state—found that one-third of them Paper presented at the Regional Conference on did not have any block development officers. Migration, Development and Pro-Poor Policy As a result, 20 percent of the funds allo- Choices in Asia, Dhaka, June 22–24. cated to the state had not been spent (World Ahmed, Nazneen, Zaid Bakht, and Md. Yunus. Bank 2005a). Weak capacity also under- 2011. “Size Structure of Manufacturing Indus- mines local monitoring of public spending, try and Implications for Growth and Poverty.” 34 ADDRESSING INEQUALITY IN SOUTH ASIA Bangladesh Country Paper, Bangladesh Insti- Working Paper 87, International Policy Centre tute of Development Studies, Dhaka. for Inclusive Growth, United Nations Develop- Akerlof, George A., and Rachel E. Kranton. ment Programme, Brasilia, Brazil. 2010. Identity Economics: How Our Identi- Deshingkar, Priya, and Sven Grimm. 2005. ties Shape Our Work, Wages, and Well-Being. Internal Migration and Development: A Princeton, NJ: Princeton University Press. Global Perspective. IOM Migration Research Andres, Luis, Dan Biller, Mario Picon, and Series No. 19. Geneva: International Organiza- Agustin Echenique. 2013. “Increasing Oppor- tion for Migration. tunities in South Asia— Infrastructure.” Dollar, David, Tatjana Kleineberg, and Aart Background paper for this report, World Bank, Kraay. 2013. “Growth Still Is Good for the Washington, DC. Poor.” Policy Research Working Paper 6568, Banerjee, Abhijit, and Thomas Piketty. 2005. World Bank, Washington, DC. “Top Indian Incomes, 1922–2000.” World Duflo, Esther. 2012. “Women Empowerment Bank Economic Review 19 (1): 1–20. and Economic Development.” Journal of Bardhan, Pranab K. 2005. Scarcity, Conflicts, Economic Literature 50 (4): 1051–79. and Cooperation: Essays in the Political and Dundar, Halil, Tara Béteille, Michelle Riboud, Institutional Economics of Development. and Anil Deolalika. 2014. Student Learning Cambridge, MA: MIT Press. in South Asia: Challenges, Opportunities, and Barros, Ricardo, Mirela de Carvalho, Samuel Policy Priorities. Washington, DC: World Bank. Franco, and Rosane Mendonça. 2010. “Mar- Educational Initiatives. 2010. Student Learning kets, the State, and the Dynamics of Inequal- Study. Status of Learning across 18 Sites of ity in Brazil.” In Declining Inequality in Latin India in Urban and Rural Schools. Ahmedabad, America: A Decade of Progress, edited by Luis India: Educational Initiatives. http://www F. López-Calva and Nora C. Lustig, 134–74. .ei-india.com/wp- content/uploads/2012/01 Baltimore: Brookings Institution Press. / Main_Report_ StudentLearningStudy_2010 Beaman, Lori, Esther Duflo, Rohini Pande, and _by_Educational_Initiatives1.pdf. Petia Topalova. 2012. “Female Leadership Ferreira Francisco, Phillippe Leite, and Julie Litch- Raises Aspirations and Educational Attain- field. 2008. “The Rise and Fall of Brazilian ment for Girls: A Policy Experiment in India.” Inequality: 1981–2004.” Macroeconomic Science 335 (6068): 582–86. Dynamics 12 (Suppl. S2): 199–230. Blattman, Christopher, and Edward Miguel. Fields, Gary S. 2010. “Does Income Mobil- 2010. “Civil War.” Journal of Economic Lit- ity Equalize Longer-Term Incomes? New erature 48 (1): 3–57. Measures of an Old Concept.” Journal of Chu, Ke-young, Hamid Reza Davoodi, and Economic Inequality 8 (4): 409–27. Sanjeev Gupta. 2000. “Income Distribution Fields, Gary S., and Efe A. Ok. 1996. “The and Tax and Government Spending Policies Meaning and Measurement of Income Mobil- in Developing Countries.” IMF Working Paper ity.” Journal of Economic Theory 71 (2): WP/00/62, International Monetary Fund, 349–77. Fiscal Affairs Department, Washington, DC. Forbes. 2014. “The World’s Billionaires.” http:// Collier, Paul, and Anke Hoeffler. 2004. “Greed www.forbes.com/billionaires/list. Accessed and Grievance in Civil War.” Oxford Eco- May 29, 2014. nomic Papers 56 (4): 563–95. Gandhi, Aditi, and Michael Walton. 2012. Dang, Hai-Anh, and Peter Lanjouw. 2014. “Where Do India’s Billionaires Get Their “Welfare Dynamics Measurement: Two Defi- Wealth?” Economic & Political Weekly, nitions of a Vulnerability Line and Their Appli- October 6, 10–14. cations.” Background paper for this report and Gentilini, Ugo, Maddalena Honorati, and Ruslan Policy Research Working Paper 6944, World Yemtsov. 2014. The State of Social Safety Nets Bank, Washington, DC. 2014. Washington, DC: World Bank. Dang, Hai-Anh, Peter Lanjouw, and Shahi- Ghani, Ejaz, ed. 2010. The Poor Half Billion in dur Khandker. 2014. “Poverty Dynamics in South Asia: What Is Holding Back Lagging Bangladesh: Recent Trends and Insights from Regions? New Delhi: Oxford University Press Synthetic Panel Data.” Background paper for and World Bank. this report, World Bank, Washington, DC. Ghani, Ejaz, Lakshmi Iyer, and Saurabh Misra. de Souza, Pedro H. G. 2012. “Poverty, Inequal- 2013. “Promoting Shared Prosperity in South ity and Social Policies in Brazil, 1995–2009.” Asia.” Economic Premise 110 (March): 1–8. OVERVIEW 35 Global Terrorism Database. 2009–14. National Iyer, Lakshmi, Ejaz Ghani, and Saurabh Mishra. Consortium for the Study of Terrorism and 2010. “Is Decentralization Helping the Lag- Responses to Terrorism, START: A Center of ging Regions?” In The Poor Half Billion in Excellence of the U.S. Department of Home- South Asia: What Is Holding Back Lagging land Security, University of Maryland, College Regions? ed. Ejaz Ghani, New Delhi: Oxford Park, MD. http://www.start.umd.edu/gtd/. University Press and World Bank. Goutam, Prodyumna, Rahul Lahoti, and J. Y. Jha, Shikha, and Bharat Ramaswami. 2010. Suchitra. 2012. “Subsidies for Whom? The “How Can Food Subsidies Work Better? Case of LPG in India.” Economic and Political Answers from India and the Philippines.” Weekly 47 (44). A DB E conomics Working Paper S eries Government of Pakistan. 2010. Poverty Reduc- no. 221, Asian Development Bank, Manila, tion Strategy Paper. Islamabad: Finance Philippines. Division, Government of Pakistan. Kanbur, Ravi. 2009. “Intergenerationalities: Some Himanshu, Abhijit Sen. 2013. “In-Kind Food Educational Questions on Quality, Quantity Transfers–I: Impact on Poverty.” Economic and Opportunity.” Working Paper No. 48922, and Political Weekly. Department of Applied Economics and Man- Himanshu, Abhijit Sen, Peter Lanjouw, Rinku agement, Cornell University, NY. Murgai, and Nicholas Stern. 2013. “Non-farm Khera, Reetika. 2011. “Trends in Diversion of Diversification, Poverty, Economic Mobility PDS Grain.” Working Paper No. 198, Centre and Income Inequality: A Case Study in for Development Economics, Delhi School of Village India.” Policy Research Working Paper Economics, Delhi, India. 6451, World Bank, Washington DC. Lam, Wai Fung. 1998. Governing Irrigation Sys- Hnatkovska, Viktoria, Amartya Lahiri, and tems in Nepal: Institutions, Infrastructure, Sourabh B. Paul. 2013. “Breaking the Caste and Collective Action. Oakland, CA: ICS Barrier: Intergenerational Mobility in India.” Press. Journal of Human Resources 48 (2): 435–73. Lichbach, Mark Irving. 1989. “An Evaluation Hoff, Karla. 2012. “The Effect of Inequality of ‘Does Economic Inequality Breed Politi- on Aspirations.” Background paper for this cal Confl ict?’ Studies.” World Politics 41 (4): report, World Bank, Washington, DC. 431–70. Hoff, Karla, and Priyanka Pandey. 2006. Long, Jason, and Joseph Ferrie. 2013. “Inter- “Discrimination, Social Identity, and Durable generational Occupational Mobility in Great Inequalities.” American Economic Review Britain and the United States since 1850.” 96 (2): 206–11. American Economic Review 103 (4): 1109–37. ———. 2012. “Making Up People—the Effect Lopez-Acevedo, Gladys, and Raymond Robert- of Identit y on Preferences and Perfor- son, eds. 2012. Sewing Success? Employment, mance in a Modernizing Society.” Policy Wages, and Poverty Following the End of the Research Working Paper 6223, World Bank, Multi-fibre Arrangement. Washington, DC: Washington, DC. World Bank. Horowitz, Donald L. 2000. Ethnic Groups in Malik, Sadia. 2009. “Horizontal Inequalities and Conflict. 2nd ed. Berkeley and Los Angeles: Violent Conflict in Pakistan: Is There a Link?” University of California Press. Economic and Political Weekly. H o s s a i n , M a h a b u b , B i n ay a k S e n , a n d ———. 2011. “An Empirical Investigation of the Yasuyuki Sawada. 2014. “Jobs, Growth Relationship between Food Insecurity, Land- and Development: Making of the ‘Other’ lessness, and Violent Conflict in Pakistan.” Bangladesh.” Background Paper for the World PIDE Working Paper, Pakistan Institute of Development Report 2013 , World Bank, Development Economics, Islamabad. Washington, DC. Mansuri, Ghazala. 2013. “Inequality and Devel- IMF (International Monetary Fund). 2013. opment.” Background paper for this report, “Energy Subsidy Reform: Lessons and Impli- World Bank, Washington, DC. cations.” IMF, Washington, DC, January 2. Mazumdar, Indrani, N. Neetha, and Indu Iyer, Lakshmi. 2009. “The Bloody Millennium: Agnihotri. 2011. “Gender and Migration Internal Conflict in South Asia.” Business, in India.” National Workshop on Internal Government, and the International Economy Migration and Human Development in India: Unit, Working Paper 09-086, Harvard Busi- Workshop Compendium, Vol. II. New Delhi: ness School, Cambridge, MA. UNESCO and UNICEF. 36 ADDRESSING INEQUALITY IN SOUTH ASIA Memon, Naseer. 2012. “Disasters in South Asia: Trimble, Chris, Nobuo Yoshida, and Mohammad A Regional Perspective.” Karachi: Pakistan Saqib. 2011. “Rethinking Electricity Tariffs Institute of Labour Education and Research. and Subsidies in Pakistan.” Policy Note, Milanovic, Branko. 2011. The Haves and the World Bank Report 62971-PK, World Bank, Have-Nots: A Brief Idiosyncratic History of Washington, DC. Global Inequality. New York: Basic Books. UN (United Nations Department of Economic MoFP (Ministry of Finance and Planning, and Social Affairs/Population Division). 2012. Sri Lanka). 2012. Annual Report 2011. World Urbanization Prospects: The 2011 Colombo: Treasury of Sri Lanka. Revision. New York: United Nations. Murgai, Rinku, and Salman Zaidi. 2005. “Effec- World B a n k . 2 0 03. B a ngl a d e s h: Pu bl i c tiveness of Food Assistance Programs in Expenditure Review. Report No. 24370-BD. Bangladesh.” Journal of Developing Societies Washington, DC: World Bank. 21 (1–2): 121–42. ———. 2005a. Bihar: Towards a Development Paes de Barros, Ricardo, Francisco H. G. Ferreira, Strategy. Washington, DC: World Bank. Jose R. Molinas Vega, and Jaime Saavedra ———. 2005b. World Development Report 2006: Chanduvi. 2009. Measuring Inequality of Equity and Development. Washington, DC: Opportunities in Latin America and the World Bank; New York: Oxford University Caribbean. Washington, DC: World Bank; Press. New York: Palgrave Macmillan. ———. 2009. South Asia: Shared Views on Devel- Paz, Octavio. 1997. In Light of India. Harcourt opment and Climate Change. Washington, Brace. DC: World Bank. Redaelli, Silvia. 2013. “Sustainability, Effi- ———. 2010. Bangladesh: Public Expenditure ciency and Equity of Electricity Subsidies in and Institutional Review : Towards a Better Maldives: Where Do We Stand and Options Qualit y of Public Expenditure . Report Moving Forward.” Processed, World Bank, No. 47767-BD. Dhaka and Washington, DC: Washington, DC. World Bank. RGCC (Registrar General and Census Com- ———. 2011. More and Better Jobs in South Asia. missioner). 2012. Census of India, 2011. Washington, DC: World Bank. New Delhi: Registrar General and Census ———. 2012a. Creating Fiscal Space through Commissioner, India, under Ministry of Home Revenue Mobilization. South Asia Economic Affairs, Government of India. Focus. Washington, DC: World Bank. Roemer, John E. 1998. Equality of Opportunity. ———. 2012b. World Development Report 2013: Cambridge, MA: Harvard University Press. Jobs. Washington, DC: World Bank. Schumpeter, Joseph Alois. 1955. Imperialism ———. 2013a. Bangladesh Education Sector [and] Social Classes: Two Essays. New York: Review: Seeding Fertile Ground—Education Meridian. That Works for Bangladesh. Washington, DC: Sen, Amartya. 1980. “Equality of What?” In The World Bank Group. Tanner Lectures on Human Values, Vol. 1, ———. 2013b. Pakistan—Towards an Integrated edited by S. McMurrin, 353–69. Cambridge: National Safety Net System: Assisting Poor Cambridge University Press. and Vulnerable Households, an Analysis of ———. 1992. Inequality Reexamined. New York: Pakistan’s Main Cash Transfer Program. Oxford University Press. Washington, DC: World Bank. Singh, Ashish, and Sripad Motiram. 2012. “How Zhang, Xiaobo, Shahidur Rashid, Kaikaus Close Does the Apple Fall to the Tree?” Ahmad, Valerie Mueller, Hak Lim Lee, Solo- Economic and Political Weekly 47 (40): 56–65. mon Lemma, Saika Belal, and Akhter Ahmed. Stiglitz, Joseph E. 2013. “Inequality Is Hold- 2013. “Rising Wages in Bangladesh.” IFPRI ing Back the Recovery.” New York Times, Discussion Paper 1249, International Food January 19. Policy Research Institute, Washington, DC. Why Inequality Matters 1 E quality carries an intrinsic value for economics toward equality of opportunities— most of the world’s great religions, often defined in terms of access to basic including Buddhism, Christianity, services. Hinduism, Islam, and Judaism, as it does for In South Asia, in particular, many children most other faith traditions and ideologies— suffer from discrimination in access to basic religious or secular. Every normative theory services because of their socioeconomic cir- of social arrangements that has stood the cumstances, such as their caste or their gen- test of time also seems to demand equality der. But between these circumstances and of something. their actual well-being as adults lies a range of What is meant by equality, however, var- additional factors affecting how individuals ies from one faith to another and from one function. The extent of mobility determines theory to another (Sen 1992). Reviewing how opportunity is converted into well-being what inequality means to different people at later stages in life. A range of shocks can and summarizing the long-standing intellec- pull individuals who had similar opportuni- tual pursuits around the concept would be an ties as children in completely different direc- exceedingly ambitious undertaking. Instead, tions over time. Measures of inequality in this report makes the deliberate choice of outcomes can capture the impact of these focusing on inequality of well-being among other factors, apart from opportunity, thus households. But justifying this choice requires allowing a better understanding of the distri- revisiting a major debate in modern develop- bution of well-being across the population. ment economics, namely inequality of oppor- A focus on the inequality of outcomes tunities versus inequality of outcomes. requires a clear metric to assess well-being. In A Theory of Justice , Rawls (1971) Traditionally, that metric has involved a single argued for both equality of opportunities monetary dimension represented by income and equality of outcomes. Roemer (1998a, or consumption per person, both in cash and 1998b), in contrast, made a case for com- in kind. However, the accuracy of this tra- pensating people for disadvantages related to ditional metric is limited by data quality. In circumstances but tolerating the remaining particular, concerns exist about how well sur- inequality of outcomes. Roemer’s conceptual- veys measure household income or consump- ization shifted the attention of development tion, especially among the better-off. Data 37 38 ADDRESSING INEQUALITY IN SOUTH ASIA constraints are even stronger when wealth, that spurs the accumulation of human capital rather than income or consumption, is used and economic growth but at the same time as the monetary indicator to assess inequality. results in inequality of outcomes. However, Over the past decade, the interest in entrenched inequality of outcomes can inequality analysis based on multidimensional undermine individuals’ aspirations in youth indicators has been growing. Nonmonetary and their subsequent educational and occu- indicators shed light on aspects of well-being pational choices. Incentives may also fail to for which markets do not exist or are under- change behavior when economic mobility developed. That is the case with access to is lacking. For example, entrenched poverty some basic services, such as health and edu- may lead to depression and behavior akin cation, on which data are often available. It to “learned helplessness.” Inequality of out- is also the case with exposure to pollution, comes may affect the capacity of households voice on public matters, or quality of social to borrow to accumulate human and physi- life, although these other aspects are harder cal capital. If accumulation needs to build on to measure. individual or household savings, those at the Using both monetary and nonmonetary bottom of the distribution may be unable to indicators of well-being is particularly infor- increase their capital significantly, and that in mative when the conclusions vary depending turn may perpetuate inequality. on the indicator used. Different conclusions Inequality also affects the ability of people can indeed shed additional light on what lies to act collectively, the institutions that they set behind inequality. up, and the ways in which resources are allo- Another metric that has spurred growing cated for the benefit of the group. A salient interest is subjective well-being. There is a case is the provision of public goods. On the long tradition reckoning that happiness is the one hand, in a very unequal society, the better- ultimate goal of existence and claiming that off typically have more power and are more people are to be trusted as the best judges of effective at pulling in resources for the public the overall quality of their lives. Subjective goods they value. On the other hand, a high well-being matters also because many human degree of inequality makes it more tempting thoughts and actions are based on perceptions for the better-off to opt out of public services and judgments, rather than data and reason- altogether. In the end, which of the two effects ing. Understanding the subjective assessments prevails is likely to depend on whether opting of well-being by different population groups out is an option. can complement inequality analyses based on Moreover, inequality of outcomes does not objective measures. generate the right incentives when it rests on In addition to the intrinsic value every rents. In that case, individuals and households great faith attaches to some form of equal- divert their efforts toward securing favoritism ity, this report is concerned with the ways in and protection instead of creating new wealth which inequality affects social organization or innovating. Inequality based on rents leads and economic performance. In other words, to a suboptimal allocation of resources and it takes a positive and not just a normative consolidates institutional arrangements with perspective. Seen this way, inequality is nei- negative long-term impacts on growth. ther good nor bad. Some forms of inequal- Last but not least, the existence of extreme ity generate costs to society, whereas others inequality exacerbates social fragmentations entail benefits. and undermines social cohesion—the capac- Inequality in outcomes has profound ity of a society to manage collective decision effects on how individuals and households making peacefully. Those suffering from behave. Some income inequality is necessary extreme deprivation may turn to violence to encourage people to study, work, save, and conflicts as the last resort to address their and invest. Returns to education are a clear concerns. This concern is not hypothetical in example of a differentiation in labor earnings South Asia’s often volatile societies. WHY INEQUALIT Y MAT TERS 39 This positive perspective is necessary to made worse off—avoids distributional understand why inequality should matter judgments altogether. In sum, the “social even to those who do not attach intrinsic welfare function” does not involve inter- value to equality. The positive perspective personal comparisons of well-being and, as also provides a framework to focus the atten- such, does not provide a useful framework tion of policy makers on the truly bad forms for distributional discussions (Milanovic of inequality and to pay attention to the costs 2009, 2011a, 2011b; Sen 1973, 1980). and benefits of measures aimed at reducing Moving beyond the social welfare func- inequality in outcomes. tion, a number of scholars uphold the intrin- sic value of economic justice and provide different conceptual frameworks for think- Inequality of what? ing about social arrangements. Most nota- Any defense or criticism of equality on nor- ble among them are John Rawls, Amartya mative grounds needs to begin by defin- Sen, Ronald Dworkin, and John Roemer. ing equality of what feature or equality in Their separate and important contributions what space—for instance, income, wealth, offer some guidance on how to reconcile opportunity, rights, or well-being. Among inequality and justice. Their differences those thinkers perceived as favoring equal- also set the foundation of the debate on ity, John Rawls (1971) focused on primary inequality of outcomes versus inequality of goods, Ronald Dworkin (1981a, 1981b) on opportunities. resources, Thomas Nagel (1986) on eco- Rawls (1971) argues that distributional nomic equality, and so on. The same is true justice requires two principles to be met. of those perceived as being against equality. The first is “an equal right to the most Robert Nozick (1974), for example, did not extensive basic liberty.” The second is social demand equality of well-being but rather arrangements “to everyone’s advantage” equality of libertarian rights. Because the and “attached to positions and offices open differences are substantive, demands by one to all.” “Open to all” can be interpreted as theory for equality along a particular dimen- relating to opportunities, whereas “everyone’s sion amount to a justification of inequality advantage” refers to outcomes. Clearly, in along some other dimension (Sen 1992). Rawls’s eyes, both are important, and equal- ity of opportunities alone is not sufficient. These two principles are established under Opportunities versus outcomes an important premise, “the veil of ignorance.” In modern economics, the thinking on social Rawls assumes that individuals do not know arrangements was initially shaped by wel- their position in the overall distribution of farism, with utilitarianism being its most assets, abilities, and preferences and therefore influential school of thought. According to ignore how alternative social arrangements utilitarianism, the social goal is to achieve will affect them. As a result, individuals can “the greatest happiness for the greatest assess principles of distributional justice only numbers” (Bentham [1781] 2000). In prac- on the basis of general considerations, with- tice, utilitarianism amounts to taking the out reference to the implication these prin- sum of the individual utilities as the metric ciples would have for them. of social welfare and assessing alternative “Open to all” could be either a formal or a social arrangements based on it. The lack substantive equality of opportunity. A formal of concern for the distribution of this sum equality of opportunity, such as careers open among individuals or households makes to all, means that all people have the same the approach unsuitable for judging equal- legal right to access. A substantive equality ity, despite its reputation as an egalitarian of opportunity, in contrast, recognizes that criterion. Similarly, the concept of Pareto given the distribution of natural endowments optimality—whereby nobody should be and social circumstances, not everybody has 40 ADDRESSING INEQUALITY IN SOUTH ASIA a fair chance in reality. Thus, a substantive Although not relying on exactly the same dif- equality of opportunity demands offsetting ference principle, Sen (1973, 1980) argues adverse circumstances over which individuals against the use of the principle of efficiency to have no control or for which they cannot be assess the distribution of outcomes. He asks held responsible. Rawls states that those who for the compensation of those who have less are at the same level of talent and ability and ability to convert income into welfare. He also have the same willingness to use them should demands that income distribution be assessed have the same prospects of success regardless based on need rather than talent or effort. of their initial circumstances. Dworkin (1981a, 1981b) and Roemer Similarly, “everyone’s advantage” can be (1998a, 1998b), in contrast, emphasize the assessed based on the principle of efficiency sufficiency of equality of opportunity, argu- or on the difference principle. The principle ing that individuals should take responsibil- of efficiency relates to Pareto optimality: ity for their well-being once differences of an arrangement is said to be efficient if no circumstances are eliminated. They believe rearrangement can make someone better off this approach can better answer conserva- without at the same time making anyone else tive critics of egalitarianism, which they say worse off. From this point of view, inequal- fails to hold individuals responsible for their ity of outcomes is justifiable as long as the fate. Roemer’s conceptualization, in particu- arrangement is efficient. The difference prin- lar, shifts the attention of many students of ciple is also called the maximin principle. It social arrangement and economic justice to states that inequality is justifiable only if it the equality of opportunity. is needed to raise the outcomes of the least Rawls’s case for social arrangements that advantaged in society. are both “open to all” and “to everyone’s Many of the thinkers who uphold the prin- advantage” speaks to the importance of ciple of justice have been greatly influenced equality of outcomes. The difference between by Rawls. Sen (1973, 1980, 1992) argues for equality of opportunity and equality of out- focusing on the differences in external factors comes may seem subtle, but it is nonethe- and internal characteristics of human beings less very important. Assessing inequality of and for aiming for the equality of functioning opportunities is necessary but not sufficient. across all. Dworkin (1981a, 1981b) demands Both measures are informative on their a distribution of resources that compensate own to understand the nature and extent of people for innate differences that they could inequality in the South Asia region. not have controlled, including differences in A substantive equality of opportunity talent. Roemer (1998a, 1998b) further devel- remains an elusive goal in South Asia. In terms ops this idea and suggests dividing factors of access to basic services, many children are that contribute to outcomes as either circum- still suffering from discrimination because of stances or efforts, with the former mainly their socioeconomic background. For exam- linked to the socioeconomic status of parents. ple, in Afghanistan, Bhutan, India, Nepal, and His idea of equal opportunity involves com- Pakistan, girls have fewer chances than boys pensating people for disadvantages related to to study because of cultural and social reasons. circumstances so that the distribution of out- Girls 12 to 18 years of age have both lower comes can be entirely attributed to the distri- secondary school attendance rates and lower bution of efforts. completion rates than boys of the same age Views on the equality of outcomes differ. group. In Afghanistan and Pakistan, the dif- For Rawls, a substantive equality of opportu- ference in attendance rates between boys and nity, together with an equal liberty, is insuffi- girls is as high as 25 percentage points and 14 cient to ensure justice in a social arrangement. percentage points, respectively. In India, the Judgment of the outcomes cannot be left to the difference between girls’ and boys’ completion principle of efficiency. The difference principle rates reaches 11 percentage points. Similarly, is required to protect the least advantaged. in parts of South Asia, some children receive WHY INEQUALIT Y MAT TERS 41 services of poorer quality because of their endowments matter, of course. But between caste, ethnicity, tribe, religion, or commu- them and actual well-being lies a range of nity. Although children who belong to certain internal and external factors affecting how groups are no longer systematically denied individuals function. Even children having schooling today, they remain vulnerable to the same access to basic services, building the bias and prejudice by teachers (box 1.1). same skills, and sharing the same aspirations Measuring and analyzing inequality of may not be able to attain the same level of opportunities can shed light on the extent of well-being. discrimination in access to services. Various The extent of economic mobility is one of methods have been proposed for doing so the factors determining how opportunity is (Roemer and Trannoy 2013). Among them converted into well-being in adult life. Work is the Human Opportunity Index developed opportunities, job transitions, and migra- by Paes de Barros and others (2009), which tion are among the mechanisms through is used extensively in this report. Assessing which economic mobility materializes. inequality of opportunities in the space of Similarly, actual well-being may be affected access to basic services is not sufficient, by shocks. From severe health conditions to however. Social circumstances and initial natural disasters to economic crises, a range BOX 1.1 Discrimination by teachers pushes children out of school Discriminatory treatment of lower-caste children Atrocities Act by which practicing untouchabil- in schools has persisted over decades, although ity is punishable by law. The kids were excited, in milder forms in recent years. Many children delighted as any kid would be, at the prospect of from lower castes remain scared to talk about not cleaning latrines. the unequal treatment against them, including “A few months later, all of them had dropped verbal abuse, physical punishment, or avoid- out. The teacher beat them. Called them stupid. ing touching them, by some of the upper-caste And rarely taught them anything. None of them teachers. Lower-caste children can be made to could read or write after six months in school. feel inferior in some schools. For example, they They were now out every morning collecting can be asked to sit separately and to perform plastic from garbage dumps. Earning Rs 20–30 unpleasant jobs such as cleaning toilets. They a day, they went to movies sometimes. School are seldom considered for leadership roles such was a distant dream, with not very pleasant as class monitor. These discriminatory practices memories.” and the entrenched social beliefs associated with Children can also be discriminated against or them can change the attitude of many lower- neglected in schools for their tribe or religion. caste children toward education and eventually In Uttar Pradesh state’s Sonbhadra district, for push them out of schools. example, students belonging to the Ghasiya The following story is a case in point tribal community reported that teachers often (Thekaekara 2004). “Near Lucknow I met pay them little attention, and the children suf- another group of bright, laughing kids. They’d fer from discrimination, such as being called all been admitted to school by a well-meaning as “dirty.” Many of these children, facing such social worker who had used a blend of threats obstacles, attend school only sporadically. Some and cajoling to convince the headmaster that stop going to school altogether. the kids had to be admitted. He reminded the principal about the existence of the Prevention of Sources: Adapted from Sedwal and Kamat 2011 and Human Rights Watch 2014. 42 ADDRESSING INEQUALITY IN SOUTH ASIA of unanticipated events can take two children means to smooth consumption in the short with similar opportunities in different direc- run and to raise it in the long run. Wealth tions. Measures of inequality in outcomes can gives a sense of security. It can also be used capture the impact of these internal and exter- to finance entrepreneurial activities, either nal factors and support a better understand- directly or as collateral (Davies and Shorrocks ing of the distribution of well-being across the 2000; Keister 2000). population. Motivated by these observations, However, flow and stock monetary indi- this report focuses on inequality of outcomes cators are not interchangeable. At the house- using a life cycle perspective. hold level, the correlation between wealth and income is relatively low. Many households are “asset rich and income poor,” meaning that Monetary measures of inequality they could be expected to have a higher stan- A focus on the inequality of outcomes dard of living than their income suggests. More requires a clear metric to assess well-being. generally, considerable dispersion of wealth Traditionally, that metric has involved a single exists within income categories, with gaps in dimension, represented by either individual wealth being much larger than gaps in income. income or individual consumption, both in Data availability has been the main hin- cash and in kind. Analyses have thus relied on drance to the analysis of wealth inequality, indicators of income inequality and poverty but the situation has rapidly improved for rates. Within this class of indicators, critical more advanced economies. For these coun- and sometimes difficult technical choices must tries, wealth indicators are estimated based be made, such as whether to use income or on household surveys, surveys of financial expenditure data and whether to use adult institutions, specialized private databases equivalent scales—taking into account that covering certain assets, and national balance the needs of a child are less than those of an sheets (Cagetti and De Nardi 2008; Davies adult. Overall, the methodology has been and others 2008; OECD 2013; Wolff 1996). refined with respect to various data and tech- Personal income tax filings have become an nical challenges and has stood the test of time. important source of information for this kind Most notably, monetary indicators have the of analysis (Piketty 2011; Piketty, Postel- attractive feature of using prices to aggre- Vinay, and Rosenthal 2006). For developing gate the various components of consumption countries, data remain a concern with some expenditure. Under relatively weak assump- exceptions. Thus, China and India have con- tions, relative prices are equal to the rate at ducted comprehensive household wealth which consumers themselves are willing to surveys, whereas Indonesia has household trade one such component for another. This surveys with a detailed wealth component (Li regularity is not affected by the income levels and Zhao 2008; Subramanian and Jayaraj or the potentially different utility functions of 2008). More and more household surveys consumers (Atkinson and Bourguignon 2000). also cover information on landholding and In addition to income and consumption, livestock. Personal tax income filings have wealth is an important but less frequently also been used in India’s case. used monetary indicator. Wealth is gener- Once the monetary indicator is chosen, ally measured as net worth, defined as total inequality is analyzed by using some sum- financial and nonfinancial assets net of total mary measure of the dispersion in the cho- liabilities (OECD 2013). Whereas income sen indicator within a population. Given the and consumption capture a flow of resources proliferation of such measures, some desir- over a period of time, wealth refers to a stock able properties may be spelled out to guide of resources at a point in time. Income can the choice between them (Atkinson and be saved to create wealth, and wealth can Bourguignon 2000). An example is the Pigou- be converted into income and used to sat- Dalton (or transfer) principle: for the chosen isfy consumption needs. Wealth provides the measure to be meaningful, a transfer from a WHY INEQUALIT Y MAT TERS 43 poorer person to a richer person, all things ( figure 1.1, panel b). Since the 1970s, a equal, must result in an increase of inequality. growing disconnect exists between the mea- Another example is the anonymity principle: sures. The consumption aggregate from the if two people switch positions in the overall NSS was about half the household consump- distribution, the measure of inequality should tion component of the national account in not change. Scale neutrality is yet another 2004–05. The gap rose further in the late example: if all incomes are multiplied by a 2000s and stabilized subsequently. constant, the measure should not change. National accounts, which estimate private These principles are selected based on consumption as a residual, do not necessar- value judgments, however. Most notably, the ily provide a benchmark to which household principle of scale neutrality reflects research- expenditure surveys should aspire. Differences ers’ preference for “relative inequality” over exist between what is being measured by the “absolute inequality.” Yet laboratory experi- two data sources. Private consumption data ments suggest that the general public is more from national accounts include sizable and ambivalent between these two concepts rapidly growing components of household (Amiel and Cowell 1999; Ravallion 2014). expenditures that are typically missing from The most frequently used measures of surveys (Deaton 2005; Deaton and Kozel monetary inequality—including the Gini 2005). Aside from these differences, household coefficient, the mean log deviation, and the surveys are vulnerable to low response rates Theil entropy index—all satisfy these axi- among the rich and underreporting of expen- oms (box 1.2). However, these summary ditures among those who respond. A large and measures may not necessarily generate the growing disconnect between two major data same ranking in levels of income inequal- sources is cause for concern of underestima- ity across population groups or between tion of expenditures of the better-off. Such two points in time for the same population underestimation may lead to biased indicators group. The main reason for these potential of inequality (box 1.3), although it is not a discrepancies in ranking is that each of the major concern when the focus is on poverty. measures embodies some assumption about Undersampling of the rich, low response the relative importance of inequality at dif- rates, and underreporting are even more ferent points in the income scale. Therefore, pronounced for wealth indicators. A prac- to reach robust conclusions, it is often neces- tical approach is to compare survey-based sary to rely on multiple measures. estimates with the balance sheet total for the Data availability and quality call for caution household sector and make adjustments. It is when one draws conclusions from monetary also informative to complement survey-based inequality. The representativeness of house- estimates with information from adminis- hold surveys illustrates this point. In South trative records and personal income tax fil- Asia, as in other regions, consumption expen- ings (Atkinson and Piketty 2010; Atkinson, ditures by different households are estimated Piketty, and Saez 2011; Banerjee and Piketty based on household-level surveys. Concerns 2005; Piketty and Saez 2003). For the ultra- exist about how well surveys measure income rich, a useful check is the lists compiled by or consumption (Datt and Ravallion 2009; Fortune and Forbes magazines (Davies and Ravallion 2003). Survey-based consumption Shorrocks 2000; Juster and Kuester 1991; and income aggregates for nationally repre- OECD 2013; Wolff 1987). sentative samples typically do not match the aggregates on private consumption obtained Multidimensional indicators of from national accounts (figure 1.1, panel a). inequality The discrepancies between levels and growth rates of consumption as measured by India’s Over the past decade there has been a grow- National Sample Surveys (NSS) and national ing interest in multidimensional indicators of accounts have been of particular concern well-being. In celebrating its 20th anniversary, 44 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 1.2 Standard statistical measures of monetary inequality The Gini coefficient is the most widely used mea- the top, the cumulative share would be the sure of inequality. In the figure, the horizontal one marked “line of equality,” signifying that axis stands for the poorest x percent of house- income is equally distributed. If, going to the holds. For that reason, the axis is labeled “cumu- other extreme, all income were concentrated lative share of households from lowest to highest in the richest household, the graph of the dis- incomes.” The vertical axis measures the income tribution would be the line from O to X to Y. share of the poorest x percent of households. The curve drawn in the figure, called the Lorenz Because shares are added while moving up the curve, shows the actual income distribution and ladder from the poorest 1 percent to the poor- lies between the lines of complete equality and est 2 percent to the poorest x percent, the verti- complete inequality. Thus, the more the Lorenz cal axis is labeled “cumulative share of income curve is bowed outward, the higher the inequal- earned.” ity in the distribution of income. What the Gini If the bottom x percent of households earned coefficient measures is the area between the line x percent of income and so on all the way to of complete equality and the Lorenz curve (A), relative to the maximum such area could attain (A + B). 100 Another class of indicators widely used by Y economists is the Generalized Entropy (GE) Index. These indicators are derived from the Cumulative share of income earned (%) A notion of entropy in information theory. Entropy is an expected information content calculated ) es as a weighted average of the information con- re eg tent of each observation. What varies across 5d (4 this class of indicators is the weight attached to ty ali qu each observation. For instance, the GE(0), or fe ve eo mean log deviation, attaches the same weight ur zc Lin n to all observations. The GE(1), or Theil index, re Lo attaches greater weight to observations that are further from the mean. One appealing feature B of the GE Index is that changes in inequality can be broken down into changes in inequal- 0 100 ity between subgroups—say, between rural and X urban areas—and within subgroups. Cumulative share of people from lowest to highest incomes (%) Source: Based on Atkinson and Bourguignon 2000. the authors of the Human Development they have the opportunity to be educated and Report of the United Nations state that whether they are free to use their knowledge “an individual’s well-being cannot be evalu- and talents to shape their own destinies” ated by money alone. Income is of course (UNDP 2010). The growing agreement on the crucial: without resources, any progress is dif- multidimensional nature of well-being, how- ficult. Yet we must also gauge whether peo- ever, is confronted by the ongoing disagree- ple can lead long and healthy lives, whether ment about how best to measure it. WHY INEQUALIT Y MAT TERS 45 FIGURE 1.1 Estimates of expenditures differ between household surveys and national accounts a. South Asian countries b. India 1960–2009 90 Sri Lanka 2009 80 India 2009 70 Indian rupees (trillions) Pakistan 2010 60 50 Afghanistan 2005 40 Bangladesh 2010 30 Bhutan 2007 20 10 Nepal 2010 0 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1960 1962 1965 1967 1969 1971 1974 1984 1988 1990 1992 1994 1996 2000 2002 2004 2005 2009 Household expenditure relative to private consumption Private consumption in national accounts Household expenditure in surveys Sources: Based on data from the World Bank’s World Development Indicators database, http://data.worldbank.org/data-catalog/world-development-indicators, and PovcalNet tool, http://iresearch.worldbank.org/PovcalNet/index.htm; Datt and Ravallion 2009; India’s National Sample Surveys (NSS). Choosing a monetary indicator as the contribution to well-being but also because measure of well-being implicitly assumes that they help capture the heterogeneity of indi- individuals or households can freely reallo- viduals, households, and countries. Access to cate their resources among, say, consumption, some basic services such as health and educa- health, and education. If so, measuring the tion are obvious examples, and measures are amount of resources available to individuals often available for these aspects of well-being. or households is enough to assess the maxi- For example, literacy and years of school- mum well-being they can attain. But in many ing are often used as proxies for the levels of cases a market for basic services and ameni- knowledge and skills. The life expectancy rate ties may not exist or may be heavily distorted. at birth, child mortality rate, and nutritional For instance, individuals or households may status are often used as proxies for the quality want to get more health services, but that of health. Other potentially relevant attributes simply may not be an option if no health care are exposure to pollution, voice on public facility is within a reasonable distance from matters, and quality of social life, although where they live. This issue has been exten- these other aspects are harder to measure. sively discussed in the economic literature The main challenge is to move from indi- (Arrow 1971; Atkinson and Bourguignon cators for individual dimensions of well-being 1982; Kolm 1977; Maasoumi 1986; Sen to indicators of overall well-being. In the 1973, 1992). It calls for extending the analy- case of consumption bundles, information sis beyond income or consumption and for on prices can be used to tell how individu- considering a broader array of determinants als or households see the trade-offs between of well-being. different consumption bundles. But no simi- Nonmonetary indicators of different lar, readily available weights are available to aspects of well-being are the starting point value nonmonetary dimensions of well-being. for a multidimensional analysis of inequality Thus, consensus remains elusive regarding or poverty. The focus is on attributes that are the functional forms and criteria to evalu- meaningful not only because of their direct ate well-being (Bourguignon 1999; Decancq 46 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 1.3 Some monetary indicators may underestimate the true extent of inequality The monetary indicators of inequality in this 100 Y report are based on household survey data. Deriving the extent to which they are biased would require information on nonresponse Cumulative share of income earned (%) rates and the underestimation of expenditures for those who respond. In the absence of such ) es re information, some assumptions must be made. eg 5d A good place to start is to assume that the gap (4 ty ali between measured expenditures and actual qu fe expenditures is greater for richer households. ve eo ur zc Lin With this assumption, it is intuitive that average en or consumption is underestimated, and the poverty eL u Tr rate overestimated. Assessing the implications of Empirical curve this assumption for the measurement of inequal- ity is less straightforward. One way to do so is to focus on the Lorenz curve, as in the figure. If the gap between measured and actual expen- 0 100 ditures increases with X, the true Lorenz curve is X Cumulative share of people from lowest higher than the observed Lorenz curve for richer to highest incomes (%) households and lower for poorer households. Given that both curves vary between zero and 100, it follows that they must intersect each poor but much larger for the rich. Consequently, other for some value of X. With neither Lorenz the Theil index is underestimated. Similarly, as curve lying completely above or below the other, long as the value of X for which the two curves no general statement about the Gini index can be intersect is between 10 percent and 90 percent, made in the general case. the ratio between the consumption of the top The bias from the underestimation of expen- and the bottom deciles is underestimated. ditures is clearer in the case of indicators of An assessment of the biases in measure- inequality that are more sensitive to the extremes ment is no substitute for better data, however. of the distribution. For instance, the Theil index Information on the nonresponse rates among gives more weight to larger gaps between actual different types of households and on the under- expenditure per capita and the level that would estimation of expenditures among those who do prevail if there were no inequality. The assump- respond would allow researchers to construct tion that the underestimation of expenditures more precise indicators of monetary inequality. increases with X implies that the true gaps are somewhat larger than the observed ones for the Source: Based on Korinek, Mistiaen, and Ravallion 2006. and Lugo 2013; Ferreira and Lugo 2013; these approaches can be classified into two Maasoumi 1999). groups (Lugo 2007; Maasoumi 1999; Tsui Since the pioneering work of Kolm (1977) 1999). The most common one involves the and Atkinson and Bourguignon (1982), a use of two-step indexes. The first step is number of approaches have been proposed to aggregate attributes for each individual to measure or analyze inequality in multiple or household, and the second step is to dimensions. At the risk of oversimplification, apply a standard measure of inequality to WHY INEQUALIT Y MAT TERS 47 the aggregate index. The second group of with an initial low level. But the conclusion approaches seeks to identify a set of appro- was the opposite when considering standard priate properties when multiple attributes poverty estimates. States that had higher exist and then looks for indexes that may standard poverty rates in 1993–94 showed satisfy these axioms. a faster reduction in poverty by 2004–05 Regardless of the approach chosen, the than those with low initial poverty rates. In marginal contribution of each dimension of other words, poverty rate declined faster in well-being to overall inequality needs to be states where poverty rates were lower to start identified. And handling the effects of correla- with when using a multidimensional poverty tions between attributes remains a challenge. index. But poverty rate declined slower in Intuitively, the various components of human states where poverty rates were lower when development are synergistically related to one using consumption-based poverty estimates. another. When all dimensions are strongly Presumably, a similar contrast in conclusions correlated, then higher achievement along one could arise if the focus had been on inequality dimension strongly reinforces higher achieve- rather than poverty. ment along other dimensions. In this case, Given that each perspective is informative focusing on just one dimension may be suf- on its own, this report complements analyses ficient for measuring well-being. Conversely, of inequality based on monetary consumption when the correlation among dimensions is with analyses based on nonmonetary indica- lower, multidimensional analyses become tors of well-being, including education and more informative (Seth 2010). health. In the absence of a consensus on the Using both monetary and nonmonetary functional form and criteria to construct mul- indicators of well-being often leads to dif- tidimensional inequality indexes, the report ferent conclusions regarding the extent of simply explores the level of inequality along inequality or its trend over time. These dif- each dimension as well as the correlation ferent conclusions, in turn, may shed addi- between monetary and nonmonetary indica- tional light on what lies behind inequality. tors. This combination provides the evidence For instance, a recent study compared needed for readers to make their own judg- consumption-based poverty estimates with ment about the “true” extent of inequality a multidimensional poverty index across 25 of well-being within countries in South Asia Indian states and over time (Alkire and Seth without imposing too much of the authors’ 2013). The multidimensional poverty index own judgments on the relationship between included education, health, and access to these different dimensions. basic services as the key nonmonetary dimen- sions of well-being (Alkire and Foster 2011). Subjective well-being For each of these dimensions, a threshold level of achievement (or access) was set, thus Subjective well-being is another area in allowing a determination of whether the per- which interest has been growing rapidly. son was deprived or not. An aggregate depri- While pushing the boundaries of traditional vation score was then computed for each approaches, multidimensional inequality person as the weighted sum of the depriva- analyses remain anchored in objective mea- tion scores along individual dimensions, and sures of well-being. The objective nature of a threshold level for the aggregated score was these measures is reflected in the substance defined to decide whether the person was matter considered (for example, years of poor or not. education) and in the assessment process (for A comparison of the poverty rate estimated instance, explicit criteria evaluated by exter- this way and the standard estimates of the nal observers). Instead, subjective measures poverty rate is striking (figure 1.2). States that of well-being seek to directly capture cogni- had a high multidimensional poverty index in tive life evaluation and mainly rely on self- 1999 saw slower progress by 2006 than those reporting by individuals. 48 ADDRESSING INEQUALITY IN SOUTH ASIA A subjective approach to well-being reck- indicators of well-being, subjective measures ons that happiness is the ultimate goal of exis- are more inclusive—not limited to the dimen- tence while admitting that everybody has his sions with available data. They also avoid or her own views on what a good life looks the issue of developing a joint distribution like. For this approach, observable indicators of different aspects of well-being (Veenhoven provide an incomplete measure of individual 2004). well-being. People are trusted to be the best An example of subjective evaluation of judge of the overall quality of their lives. well-being is the World Values Survey, which Individual scores of happiness or life satis- asks, “All things considered, how satisfied are faction are evaluated based on self-reported you with your life as a whole these days?” assessments of short-term experiences and Subjective well-being is assessed on a scale longer-term fulfillment (Frey and Stutzer from 1 (dissatisfied) to 10 (satisfied) (Inglehart 2002). In comparison with multidimensional 2000). The Eurobarometer surveys, covering FIGURE 1.2 Monetary and nonmonetary indicators can lead to opposite conclusions a. Multidimensional poverty in Indian states 1.0 0.5 0 Annual change, 1999 to 2006 (percentage points) –0.5 Bihar Rajasthan Madhya Pradesh Sikkim Uttar Pradesh –1.0 Nagaland Punjab West Bengal Haryana Tripura –1.5 Mizoram Assam Goa Manipur Gujarat Meghalaya Orissa Himachal Pradesh Karnataka –2.0 Maharashtra Andhra Pradesh –2.5 Tamil Nadu –3.0 –3.5 Kerala 10 20 30 40 50 60 70 80 Headcount ratio, 1999 (percent) (continues next page) WHY INEQUALIT Y MAT TERS 49 FIGURE 1.2 Monetary and nonmonetary indicators can lead to opposite conclusions (continued) b. Income poverty in Indian states 1.0 0.5 Tripura Goa Madhya Pradesh Mizoram Annual change, 1993–94 to 2004–05 (percentage points) 0 Punjab Orissa Sikkim Rajasthan –0.5 Gujarat West Bengal Uttar Pradesh Bihar Himachal Pradesh Maharashtra –1.0 Kerala Nagaland Haryana Andhra Pradesh Tamil Nadu Karnataka –1.5 Meghalaya Assam –2.0 Manipur –2.5 –3.0 –3.5 10 20 30 40 50 60 70 80 Headcount ratio, 1993–94 (percent) Source: Alkire and Seth 2013. all members of the European Union, and the method (DRM) has been developed more Gallup World Poll, surveying more than 160 recently (Kahneman and others 2004). In countries, ask respondents a similar question. this approach, people are asked to reflect on The emotional aspects of individuals’ how they felt during all the life episodes they well-being can be captured by questions went through the previous day. Compared on moment-to-moment affect (Kahneman to the life satisfaction evaluation, the DRM and Krueger 2006; Stutzer and Frey 2012). relies less on processes of retrospection. In Psychologists have been applying the experi- combination with the time use information, ence sampling method for years. It collects the DRM aims to establish a cardinal indica- information on individuals’ actual experi- tor of well-being by considering the time spent ences in real time in their natural environ- in a predominantly negative affective state. ments (Stone and Shiffman 1994). A related Because subjective survey data are based new approach called the day reconstruction on individual’s judgments, they are subject 50 ADDRESSING INEQUALITY IN SOUTH ASIA to both systematic and nonsystematic biases. with one another (Diener and Suh 1999; Frey The reported life satisfaction may depend in and Stutzer 2002). New approaches such as part on the order of questions, the wording DRM stand to reduce measurement error. of questions, the scales applied, the actual These measures have begun to inform mood and memory of the respondent, and the official statistics and have started to be dis- immediate context. Not surprisingly, reported cussed in policy debates. In his remarks to life satisfaction is found to fluctuate in natu- the General Conference of International ral settings over short time periods (Lucas, Association for Research in Income and Diener, and Suh 1996). Wealth, Federal Reserve chairman Ben S. However, the validity of data on reported Bernanke emphasized subjective measures of life satisfaction has been supported by empiri- well-being as a new direction of economic cal evidence. Correlations have been docu- measurement (Bernanke 2012). The Kingdom mented between measures of life satisfaction of Bhutan is another unique case in point. and various objective physiological and medi- Bhutan adopted Gross National Happiness as cal criteria (Cohen and others 2003; Kiecolt- the national indicator of well-being in 1972 Glaser and others 2002). Findings from (box 1.4). The index is multidimensional neuroscience research also lend some support and aims to include objective dimensions for the view that life satisfaction measures of well-being as well as subjective elements are related to individuals’ emotional states such as psychological wellness, balance with (Urry and others 2004). Some visible signs of the environment, community vitality, and cheerfulness, such as smiling, are positively strength of family and social ties. associated with self-reported happiness, while In addition to people’s intrinsic interest different measures of happiness correlate well in happiness, subjective well-being matters BOX 1.4 Bhutan uses a happiness index to measure well-being The Gross National Happiness (GNH) concept diversity, ecological resilience, living standards, was designed to measure the quality of life and health, education, and good governance. assess social progress in more holistic terms Each of the domains covers a range of specific than the gross domestic product (GDP) allows. indicators—33 in all. For instance, psychological The term gross national happiness was coined well-being includes life satisfaction, emotional in 1972 by Bhutan’s fourth Dragon King, Jigme balance, and spirituality. All three indicators are Singye Wangchuck, who opened the country to assessed based on self-reported cognitive assess- the age of modernization. He declared GNH ments. Community vitality covers social support, to be more important than GDP, and from that which depicts the civic contributions made; time onward the country has oriented its devel- community relationship, which refers to social opment plans and policy choices toward raising bonding and a sense of community; family rela- GNH. The concept was taken seriously, as the tionships; and perceived safety. Cultural diversity Centre for Bhutan Studies, under the leader- is measured in areas of language, artisan skills, ship of Karma Ura, developed a survey instru- cultural participation, and the way of harmony. ment to measure the population’s general level of Both community vitality and cultural diversity well-being. are constructed as a mixture of self-reported The current GNH Index is an aggregation of cognitive assessments and quantitative measures. performance across nine domains: psychological well-being, time use, community vitality, cultural Source: Based on Ura, Alkire, and Zangmo 2012. WHY INEQUALIT Y MAT TERS 51 because perceptions and intuitions influ- A more recent strand of literature, based ence individual decision making. The work on laboratory experiments, provides support by Daniel Kahneman and coauthors has to the idea of a shared human preference for profoundly challenged the validity of the equality and fairness (World Bank 2005). rational-agent model, the foundation of In these experiments, individuals interact modern economics (e.g., Kahneman 2003a, through behavioral games and play with real 2003b; Kahneman and Tversky 1984; money under tightly controlled conditions. Tversky and Kahneman 1974). Most criti- Results reject the standard economic hypoth- cally, most of the time people are found to esis that individuals are exclusively concerned make judgments and choices based on intu- with their own material gain. Instead, some ition rather than through rational reason- people behave in ways clearly inconsistent ing, and the rules governing intuition are with the rational self-interest hypothesis. generally similar to the rules of perception Additionally, people are heterogeneous. A siz- (Kahneman 2003b, 2011). In other words, able fraction of people in most experiments perception and intuition shape behaviors. engage in altruistic behavior while others Perceptions may or may not match reality, behave selfishly. Interestingly, fair-minded and intuitive judgments may or may not be people can behave selfishly, and self-interested accurate, but correcting them is not easy. people can behave altruistically. Taken Therefore, taking into account subjective together, these experimental results lend sup- assessments of well-being by different popu- port to the notion that equality and fairness lation groups is important when assessing matter intrinsically to people. inequality. The flip side is that people also value inequality. Endorsing equality along some particular feature easily leads to a theory The costs (and benefits) of justifying inequality along a different feature inequality (Sen 1992). More generally, ethical theories From a positive point of view, any assess- of social arrangements typically take a stance ment of the extent of inequality in a country on the acceptable levels of inequality. For requires some notion of optimality. Taking instance, by demanding equality of libertar- for granted that inequality is neither good ian rights, Nozick (1974) defends potentially nor bad, the issue is to determine the level large levels of inequality in well-being. At the that is preferable from a societal point of other end of the spectrum, Roemer (1998a, view. That level is the turning point for 1998b) narrows down the acceptable level of which the costs of inequality start exceed- inequality of outcomes by requiring equal- ing its benefits. Determining such a turning ity of opportunities, especially for children. point in a precise way is clearly out of reach. Even Rawls (1971) agrees that some level But economic analysis helps identify the of inequality in outcomes is acceptable if it main costs and benefits and aids in getting a is in the interest of the most disadvantaged sense of their order of magnitude. (that is, if the difference principle is satisfied). Inequality, it seems, also has some value for human beings. Intrinsic value The first and most straightforward reason Incentives why inequality can be costly is because peo- ple may not like it. Cultures and religions At the risk of oversimplifying, some degree around the world may differ in important of monetary inequality is needed to create respects, but they all seem to share a concern incentives for people to study and accumu- with fairness. The preference for equality, in late human capital, to work instead of tak- one form or another, may be attributed to ing leisure, to save for the future, and to entrenched social norms and cultural beliefs. invest in risky businesses. Without a certain 52 ADDRESSING INEQUALITY IN SOUTH ASIA difference in returns, people may not be youth, affecting their subsequent educa- motivated to undertake these activities. tional and occupational choices. A series Because of its implications, inequality was of experiments suggests that inequality of taken to be unambiguously good for growth outcomes can have “framing effects” (Hoff by some economists and philosophers, such 2012). An example of a frame, or stereo- as Max Weber and John Maynard Keynes type, is the idea that marginalized groups (Milanovic 2011a). Economic transition in are intellectually inferior. Confirmatory bias formerly socialist countries was all about leads people to conform to the stereotype— reintroducing incentives that had been for example, by discounting the abilities of undermined by excessive egalitarianism. those who belong to a marginalized group. Returns to education are a clear example Framing effects of this sort have been found of a differentiation in labor earnings that cre- to be salient in two experiments conducted ates incentives to study but at the same time in India. results in inequality in outcomes. In South In a laboratory experiment with high- and Asia, as in other regions of the world, returns low-caste boys, cues to one’s place in the are larger the higher the educational attain- traditional caste order are found to influ- ment of the person (figure 1.3). Schooling is ence the ability of the low-caste boys to learn of course not the only determinant of labor and the willingness of the high-caste boys to earnings, and it accounts for only a fraction of expend effort (Hoff and Pandey 2006, 2012). earnings variation across individuals. But the Students were assigned to sessions of six boys, relationship is robust, and it holds in South taught how to solve mazes, and then asked to Asia even though the quality of education is do it under monetary incentives. When names not always high. Positive returns to education and castes were not disclosed to others in the are consistent with the view that differences session, there were no differences in the abil- in earnings serve as rewards to efforts by chil- ity of high- and low-caste boys to learn this dren and their parents. skill. But the simple contrivance of calling out the name and caste of the boys at the begin- ning of a session produced a marked caste gap Aspirations in the ability to learn how to solve a maze. Entrenched inequality can significantly Compared with the anonymous sessions, undermine individuals’ aspirations in these sessions reduced the failure rate from FIGURE 1.3 Returns to education create incentives to study 60 relative to no education (percent) 50 Earnings premium 40 30 20 10 0 Sri Lanka, 2008 India, 2010 Nepal, 2008 Pakistan, 2008 Incomplete primary Complete primary Complete lower secondary Complete higher secondary Source: World Bank 2011. WHY INEQUALIT Y MAT TERS 53 8 percent to 2 percent among the high-caste Inequality of outcomes, especially in the boys and increased the failure rate among absence of avenues to prosper, can greatly the low-caste boys from 1 percent to 11 per- erode motivation for those at the bottom cent. The cue to caste was irrelevant to the of the distribution (Hoff 2012). The poor problem. But making caste salient may have and the disadvantaged who live in highly evoked in the children memories that changed unequal societies may come to think of how they think about themselves and about their places in the social order as fixed and the world around them. unchangeable. They have no opportunity to In another experiment, exposure to a play different roles or to be exposed to new more equal distribution of political power role models, so they have little chance to raised aspirations for girls. In 1993, change their worldviews and the way they India adopted gender quotas for village perceive themselves. Entrenched poverty governments. The position of chief coun- may lead to depression and behavior akin cilor ( pradhan ) was reserved for women to “learned helplessness.” The feeling that in one-third of the village governments, one has little power and few resources is which were randomly chosen. As a result, likely to diminish goals and sap the capac- the number of women holding this post ity for hope. increased dramatically. This program was A recent experiment in Bangladesh finds found to affect the aspirations of girls and evidence linking poverty to hopelessness and of their parents (Beaman and others 2012). inaction (Bryan, Chowdhury, and Mobarak After about seven years of exposure to a 2012). The experiment was conducted in female pradhan, the gender gap in aspira- a region where preharvest famine is recur- tions was sharply reduced for teenagers and rent, and yet out-migration is not common. their parents. In particular, girls were less In the experiment, a US$8 incentive was likely to want to be housewives, less likely provided to individuals to migrate during to want their in-laws to determine their the preharvest season. A very large effect on occupation, more likely to want to marry migration propensity was found. More than after 18, and more likely to want a job that 40 percent of those receiving an incentive requires more education. chose to migrate, whereas only 13 percent of These preferences were mirrored in some control households did. The gains for those tangible behavior. Whereas boys were more who migrated were large: about US$100 in likely than girls to be enrolled in school in extra consumption over the season. Other areas with no female pradhans and to be evidence suggests that this large effect can- able to read and write, the gender gap in not be explained by imperfect information on educational outcomes was completely erased employment opportunities at destination, nor in areas with seven years of exposure to a can it be plausibly explained by risk aversion. female pradhan. Because little else changed The reason that these poor people do not in terms of actual policy or career opportu- save the US$8 that they need to buy a ticket nities, seeing a woman achieving the position to migrate is likely that they have failed to of local head likely provided a role model think clearly about their opportunities. The and changed how the girls think about them- stimulus of the gift of the US$8 ticket leads selves and about the world around them. It them to make a major decision, with a very affected aspirations, which in turn affected large return, that most individuals were not educational choice (Banerjee and Duflo 2011; able to make otherwise (Banerjee and Duflo Duflo 2012). 2011; Duflo 2012). Evidence also indicates that poverty is a causal factor in depression and that in Behaviors poor regions depression is widespread (Case Incentives may fail to change behav- and Deaton 2009). In Bangladesh and in ior when economic mobility is lacking. West Bengal, India, aid to the very poor 54 ADDRESSING INEQUALITY IN SOUTH ASIA significantly affects their incomes, consump- 16 percent of all landholdings. Nor have any tion, and financial savings (Banerjee and substantive shifts in ownership taken place Duflo 2011; Banerjee and others 2011; Duflo because of land reforms or other policy or 2012). This experiment shows that a small market changes. In the absence of an active and temporary financial assistance can moti- market for land, farm plot sizes can be con- vate the very poor to action and lead them to sidered exogenous. Indeed, over the past make choices that substantially improve their 30 years, there has been little change in the well-being. Arguably, the outcome suggests distribution of land, so inequality in landown- that the very poor had become depressed and ership is unlikely to reflect differences in abil- were failing to perceive the choices that they ity or effort (Mansuri 2013). Concentration really faced. is the norm, with the median Gini coefficient for landholdings across villages being 0.74. But the extent of concentration varies from Access to finance one village to another, and this variation can Although inequality of outcomes may create be used to analyze whether inequality affects incentives to accumulate human and physi- the accumulation of physical capital by cal capital, it may also affect the capacity households. of households to borrow for that purpose. Accumulation of assets by households dif- If accumulation needs to build on individual fers significantly between the most equal vil- or household savings, those at the bottom of lages and the most unequal ones. The measure the distribution may be unable to increase of accumulation is the annual change of live- their capital significantly, and that in turn stock, which is perhaps the most divisible and may perpetuate inequality. liquid productive asset for rural households. Rural Pakistan provides ideal conditions Controlling for the initial poverty status and to explore this hypothesis. There, land is livestock holding of households, the poor are almost wholly acquired through inheritance, more likely to accumulate in villages with a with purchased land accounting for only lower level of land inequality (figure 1.4). FIGURE 1.4 Greater inequality in landholdings is associated with Public goods lower asset accumulation among the poor The influence of inequality goes beyond indi- vidual or household behavior. Inequality 120 also affects the ability of people to act col- lectively, the institutions they set up, and the Penalty on livestock accumulation relative 100 ways in which resources are allocated for 80 the benefit of the group. A salient case is the to the nonpoor (percent) provision of public goods, where inequality 60 can have effects of opposite signs. On the 40 one hand, in a very unequal society, the rich typically have more power and clout, which 20 makes them more effective at pulling in 0.0 development resources for the public goods they value. From that perspective, inequal- –20 ity may be conducive to a greater provision of public goods, thus benefiting the poor as –40 well. On the other hand, a high degree of Least equal 10 Middle 80 Most equal 10 percent percent percent inequality makes it more tempting for the Pakistani villages, by share of land owned by the top 1 percent rich and the upper-middle classes to opt out of public service because they can afford pri- Source: Based on data from Mansuri 2013 for this report. vate alternatives. In this case, the quality of WHY INEQUALIT Y MAT TERS 55 public services is bound to deteriorate, fur- is unambiguously associated with greater ther harming the poor. In the end, which of access to services (figure 1.5). The most the two effects prevails is likely to depend on unequal villages certainly do no worse than whether opting out is an option. the most equal villages. For education ser- Differences in land concentration among vices, however, things are somewhat differ- villages in Pakistan provide fertile ground to ent. Although the extent of land inequality test the effect of inequality when opting out has no correlation with the availability of of some services is an option (Mansuri 2013). public primary schools or with their physi- Land concentration at the village level has cal condition, the teaching quality in those been stable for decades in Pakistan, partly schools is far poorer in the most unequal because of distortions in land markets. This villages. Notably, this negative impact of long-term stability implies that land concen- land concentration on teaching quality does tration predates the current quality of public not extend to private schools. These results goods at the village level. If a causal relation- imply that service quality and access tend to ship exists, it arguably goes from land con- decline at very high levels of inequality for centration to the quality of public goods, and services for which the wealthy can move to not the other way around. private providers—such as schools. For services such as electricity, drainage, Experiences of decentralization to local access to public transport, and the near- governments shed some light on the relation- est bank branch, greater land inequality ship between inequality and elite capture. FIGURE 1.5 Greater inequality reduces the quality of public services when the rich can opt out Change in the availability of public goods when the share of land owned by the top quintile increases by 1 percent in Pakistani villages 100 80 60 The rich can opt out 40 Estimated change (percent) 20 0 –20 –40 –60 –80 The rich cannot opt out –100 –120 n t m ity ad ol ) ol ic) nk te ni tio ho ho te bl Ba ro ric iva hu rta ys pu sc sc ct d pr alt es ve po Ele y( te ic y( he bl Pa ag iva lit ns lit Pu ain ua sic tra Pr ua rq Ba Dr rq ic he bl he Pu ac ac Te Te Source: Based on from Mansuri 2013 for this report. 56 ADDRESSING INEQUALITY IN SOUTH ASIA In principle, the decentralization of service was more likely to select local projects that delivery brings government decisions closer generated less employment for the poor to the people, thereby improving responsive- when landlessness and land inequality were ness to local needs. When communities are greater (Bardhan and Mookherjee 2006). highly uneven and accountability is low, how- ever, decisions tend to be overly influenced Rent seeking by local elites at the expense of the rest of the population (Araujo and others 2008; Bardhan Inequality of outcomes does not generate 2005; Bardhan and Mookherjee 2005; the right incentives when it rests on rents Drèze and Sen 1991; Madison [1787] 1961). (Stiglitz 2013). In that case, rather than Comparing local outcomes before and after being encouraged to study or to accumu- decentralization episodes is a way to assess late, individuals and households divert their whether the extent of local inequality makes efforts toward securing favoritism and pro- a difference. tection. Rent seeking does not create new In Bangladesh, the central government wealth or lead to innovations. Instead, it delegated the identification of beneficiary aims to manipulate the social and political households under the Food-for-Education environment in the pursuit of privilege—for program to school-management commit- example, by “lobbying” for a lower tax rate tees at the community level. The targeting or “buying” a favorable regulatory environ- of the beneficiaries was worse in villages ment. Rewarding such behaviors with high with larger land inequality (Galasso and returns is costly to economic growth and Ravallion 2005). In West Bengal, India, the social development. It leads to a subopti- implementation of a set of development pro- mal allocation of resources in the short term grams had been decentralized to the local and consolidates institutional arrangements governments for over a quarter of a century. with negative long-term impacts on growth. In this case, the allocation of private goods Lucrative returns from rent seeking also fos- such as farm inputs and credits was not par- ter corruption and nepotism. ticularly worse in villages with greater land Colonialism was an extreme case of rent inequality. However, the local government seeking, associated with high inequality and a detrimental impact on local develop- FIGURE 1.6 Inequality was highest under rent-seeking ment. The extent of inequality in several colonial rule preindustrial economies, including colonial settlements, has been compared with that of 0.030 modern economies (Milanovic 2009, 2011a; Milanovic, Lindert, and Williamson 2011). 0.025 To ensure comparability across very different data sets, a consistent measure of inequality 0.020 is introduced. This is the inequality extrac- Frequency tion ratio, defined as the quotient between the 0.015 actual Gini coefficient and the maximum Gini coefficient. The maximum Gini coefficient 0.010 is computed assuming that all people other than the elite have incomes at subsistence 0.005 level whereas the elite earn what is left from aggregate income. Colonial settlements are 20 40 60 80 100 120 associated with the highest inequality extrac- Extraction ratio tion ratios, typically around 100 percent Modern Preindustrial noncolonies Preindustrial colonies (figure 1.6). The inequality extraction ratio Source: Personal communication from Branko Milanovic based on Milanovic, Lindert, and was as high as 113 percent in Mogul India in Williamson (2011). 1750 and 97 percent in British India in 1947. WHY INEQUALIT Y MAT TERS 57 These extreme levels of inequality were cases, conflict reflects a systematic use of vio- associated with crude rent-seeking behav- lence for economic gain, such as the control ior. Indigenous interests were suppressed, of resources, property, occupations, and busi- local resources and talents were exploited, ness activities (Blattman and Miguel 2010; and social and political institutions were Collier and Hoeffler 1998, 2004). In others, manipulated to maximize the returns to colo- economic factors lurk in the background of nial powers at the expense of the indigenous a conflict that erupts along social and politi- populations. The detrimental impact to local cal cleavages (Bardhan 2005; Horowitz 1998, economic growth and social development was 2000). Inequality, especially deprivation, may substantive and long-lasting. Evidence sug- intensify the grievances felt by certain groups gests that rent-seeking activities by colonial or can reduce the opportunity costs of initiat- powers may have left local areas with a worse ing and joining a violent conflict. institutional legacy—higher concentration of Conflict may take many extreme forms. landholdings, worse governance practices, In South Asia, violence is more common in and less access to justice—and lower long- areas characterized by massive deprivation. term economic growth (Naritomi, Soares, and Data from multiple sources, including the Assunção 2007). Native elites succeeding the Global Terrorism Database, the Rand-MIPT colonizers often relied on similar institutions, Terrorism Incident Database, the Indian however, and continued with similar policies Ministry of Home Affairs, the Informal of maximum extraction (Milanovic 2011a; Sector Services Center (a nongovernmental Milanovic, Lindert, and Williamson 2011). organization in Nepal), and the South Asia Intelligence Review, confirm that internal vio- lent conflicts—including communal violence, Conflict separatist movements, the Maoist insurgency, Although a straightforward relationship and terrorist attacks by religion extremists— between inequality and conflict would be are geographically correlated with the inci- hard to establish, some connections exist dence of poverty across countries in South between the two. Inequality may damage Asia (Iyer 2009). trust—the foundation for social cohesion— In the case of India, the probability of a dis- and thus weaken collective decision mak- trict being affected by Naxalites (Maoist reb- ing. The problem is particularly salient in els) can be linked to the characteristics of the management of common resources. Across district. Using data from the Indian Planning irrigation communities in south India and in Commission and the South Asian Intelligence Nepal, inequality is found to make resolving Review, the probability has been shown to rise disputes in water allocation more difficult. with the district’s poverty rate and to fall with Among 48 villages over six districts in India’s its literacy rate (Borooah 2008). Indeed, with Tamil Nadu state, the maintenance quality of the exception of Jharkhand and Maharashtra, distributaries and field channels was lower poverty incidence is higher in districts where and water-allocation rules were more likely Naxalites are better implanted (figure 1.7). to be violated in villages with higher land The difference is particularly large in the case inequality. When the water-allocation rules of rural areas. In Pakistan, the probability of were perceived as being crafted by the vil- violent attacks by insurgents, sectarians, and lage elite, dispute resolution was also more terrorists is also related to the socioeconomic difficult (Bardhan 2005). In Nepal, income characteristics of the district. Based on data inequality undermines the management of collected by the Pakistan Institute of Policy irrigation systems, as found in a comparison Studies, the probability is found to increase of 150 agency- and farmer-managed systems with food insecurity and land inequality (Lam 1998). (Malik 2009, 2011). More broadly, inequality affects the eco- Although these findings reflect only a cor- nomics of conflict (Lichbach 1989). In some relation, other analyses suggest causality. 58 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 1.7 Poverty is higher in Indian districts suffering from Naxalite violence a. Rural areas 60 50 Poverty headcount ratio (percent) 40 30 20 10 0 India Andhra Bihar Chhattisgarh Jharkhand Madhya Maharashtra Odisha Uttar West Bengal Pradesh Pradesh Pradesh b. Urban areas 40 35 Poverty headcount ratio (percent) 30 25 20 15 10 5 0 India Andhra Bihar Chhattisgarh Jharkhand Madhya Maharashtra Orissa Uttar West Bengal Pradesh Pradesh Pradesh LWEA districts Other districts in affected state Other districts in country Source: Based on NSS 2011–12 data for India. Note: LWEA = left-wing-extremism-affected districts. As defined by the Planning Commission in http://pcserver.nic.in/iapmis/state_district_list.aspx, India has 88 such districts. The headcount ratio is based on the national poverty line. Based on a more comprehensive database, (Eynde 2013; Kapur, Gawande, and Satyanath land inequality is found to strongly affect 2012). In the case of Nepal, an analysis track- Naxalite violence (Gomes 2011). A signifi- ing civil war incidence and casualties across cant relationship also exists between adverse space and over time found that poorer dis- shocks on agricultural production and renew- tricts were likely to be drawn into conflicts able resources such as forests and the inten- earlier and were associated with higher sity of conflict in the Maoist belt of India conflict-related deaths (Do and Iyer 2010). WHY INEQUALIT Y MAT TERS 59 Using data on the recruitment by the Maoist Araujo, M. Caridad, Francisco H. G. Ferreira, insurgency in Nepal, another study finds that Peter Lanjouw, and Berk Özler. 2008. “Local recruiting through abduction of young people Inequality and Project Choice: Theory and was more intensive in districts where inequal- Evidence from Ecuador.” Journal of Public Economics 92 (5): 1022–46. ity between those with land and the landless A rrow, Ken net h J. 1971. “A Ut i l it a r ia n had previously increased (Macours 2011). Approach to the Concept of Equality in The relationship between inequality and Public Expenditures.” Quarterly Journal of conflict is not mechanical, however. The Economics 85 (3): 409–15. impact of conflicts differs across income Atkinson, A nthony B ., and François groups. The poor may have less sympathy to Bourguignon. 1982. “The Comparison of violent conflicts as they tend to be the ones Multi-Dimensioned Distributions of Economic affected the most. In Pakistan, a 6,000-per- Status.” Review of Economic Studies 49 (2): son nationally representative survey was 183–201. conducted to measure attitudes toward four ———. eds. 20 0 0. H a n dbook of In c o m e militant organizations. Contrary to expec- Distribution. Vol. 1. Oxford: North-Holland, Elsevier. tations, poor Pakistanis dislike militants Atkinson, Anthony B., and Thomas Piketty, eds. more than middle-class citizens. The dislike 2010. Top Incomes: A Global Perspective. is strongest among the urban poor, particu- Oxford and New York: Oxford University larly those in violent districts, suggesting that Press. exposure to terrorists attacks reduces sup- Atkinson, Anthony B., Thomas Piketty, and port for militants (Blair and others 2012). Emmanuel Saez. 2011. “Top Incomes in the Groups divided by social and political cleav- Long Run of History.” Journal of Economic ages also tend to differ in their incomes, and Literature 49 (1): 3–71. as a result, changes in relative incomes across Banerjee, Abhijit, and Esther Duflo. 2011. Poor groups may exacerbate (or reduce) conflict. In Economics: A Radical Rethinking of the India, where Muslims tend to be poorer than Way to Fight Global Poverty. New York: PublicAffairs. Hindus, one study found that an increase in Banerjee, Abhijit, Esther Duflo, Raghabendra Muslim expenditure per capita at the com- Chattopadhyay, and Jeremy Shapiro. 2011. munity level generates a large and significant “Targeting the Hard-Core Poor: An Impact increase in local religious conflict. In contrast, Assessment.” J-PAL, Massachusetts Institute an increase in Hindu per capita expenditure of Technology, Boston. does not lead to higher conflict (Mitra and Banerjee, Abhijit, and Thomas Piketty. 2005. Ray 2013). In this case, lower inequality at “Top Indian Incomes, 1922–2000.” World the local level is associated with greater vio- Bank Economic Review 19 (1): 1–20. lence, whereas greater inequality is not. Bardhan, Pranab K. 2005. Scarcity, Conflicts, and Cooperation: Essays in the Political and Institutional Economics of Development. References Cambridge, MA: MIT Press. A l k i re, Sabi na, a nd Ja mes Foster. 2011. Bardhan, Pranab, and Dilip Mookherjee. 2005. “Counting and Multidimensional Poverty “Decentralizing Antipoverty Program Delivery Measurement.” Journal of Public Economics in Developing Countries.” Journal of Public 95 (7): 476–87. Economics 89 (4): 675–704. A l k i re , S abi n a , a nd Su m a n S e t h. 2013. ———. 2 0 0 6 . “ P ro - p o or Ta r g e t i n g a nd “Multidimensional Poverty Reduction in India Accountability of Local Governments in West between 1999 and 2006: Where and How?” Bengal.” Journal of Development Economics Oxford Poverty and Human Development 79 (2): 303–27. Initiative, University of Oxford, U.K. Beaman, Lori, Esther Duflo, Rohini Pande, and Amiel, Yoram, and Frank Cowell. 1999. Thinking Petia Topalova. 2012. “Female Leadership about Inequality: Personal Judgment and R a i s e s A s p i r a t i o n s a n d E du c a t i o n a l Income Distributions. Cambridge: Cambridge Attainment for Girls: A Policy Experiment in University Press. India.” Science 335 (6068): 582–86. 60 ADDRESSING INEQUALITY IN SOUTH ASIA Why inequality matters: Main messages and policy implications In recent years, debates on inequality in devel- undermine aspirations and lead to a psychologi- opment economics have been dominated by the cal state of helplessness, so that incentives do not concept of equality of opportunity. Implicit in translate into behaviors. Excessive inequality can this concept is a focus on access to basic ser- also get in the way of access to finance for the vices such as health, education, and basic most deprived and result in missed investment infrastructure in childhood, and especially on opportunities. Deprivation can undermine social eliminating discrimination in access because of cohesion, nurturing violence and fostering con- inherited circumstances, such as gender, ethnic- flict. Its effects on the provision of public goods ity, or location. However, well-being in adult life are more mixed and crucially depend on whether clearly may be affected by other forces beyond the rich and the better-off can opt out and rely on the control of individuals. Not everybody faces private alternatives. the same employment and migration choices, so Because inequality has both costs and bene- that children with equal access to immuniza- fits, it cannot be labeled as unambiguously good tion, schools, or sanitation may end up in very or bad. For instance, a high level of inequality different places. Similarly, shocks such as dis- of outcomes resulting from rent seeking, privi- ease or natural disasters may result in very dif- lege, and corruption is bound to be very bad for ferent levels of well-being in adult life. Although economic performance. But the same level of equality of opportunity is essential, mobility inequality in a society characterized by equal- and support matter as well. For this reason ity of opportunity in access to basic services and any comprehensive analysis needs to consider strong mobility through jobs and migration may inequality of outcomes and not only inequality provide all the right incentives and support rapid of opportunity. economic growth. From a policy perspective, Although most great faiths and political then, the issue is not just the extent of inequality movements have normative views on inequality, but also its sources. The goal is not necessarily a positive approach is warranted. Equality may to reduce (or increase) inequality but to focus have an intrinsic value that varies from one soci- efforts on reducing bad forms of inequality. And ety to another, but it also has instrumental value. even in this respect, the cost-benefit logic inher- Because of its effects on social organization and ent to the positive approach needs to be kept in economic performance, inequality actually has mind. Reducing bad forms of inequality, such tangible costs and benefits. On the benefits side, as rent seeking and discrimination, should be a some degree of inequality in outcomes is needed government priority. But the overall cost of the to encourage people to study, work, and save. instruments used to reduce inequality should not On the costs side, entrenched inequality can be higher than the expected benefits. Bentham, Jeremy. (1781) 2000. An Introduction Support for Militant Politics: Evidence from to the Principles of Morals and Legislation. Pakistan.” American Journal of Political Kitchener, ON: Batoche Books. Science. Bernanke, Ben S. 2012. “Economic Measurement.” Blattman, Christopher, and Edward Miguel. Remarks to the 32nd General Conference of 2010. “Civil War.” Journal of Economic the International Association for Research Literature 48 (1): 3–57. in Income and Wealth, Cambridge, MA (via Borooah, Vani K. 2008. “Deprivation, Violence, prerecorded video), August 6. and Conflict: An Analysis of Naxalite Activity Blair, Graeme, C. Christine Fair, Neil Malhotra, in the Districts of India.” International Journal and Jacob N. Shapiro. 2012. “Poverty and of Conflict and Violence 2 (2): 317–33. WHY INEQUALIT Y MAT TERS 61 Bourguignon, François. 1999. “Comment on Decancq, Koen, and María Ana Lugo. 2013. ‘Multidimensioned Approaches to Welfare “Weights in Multidimensional Indices of Analysis’ by E. Maasoumi.” In Handbook of Wellbeing: An Overview.” Econometric Income Inequality Measurement, edited by Reviews 32 (1): 7–34. Jacques Silber, 477–84. Norwell, MA: Kluwer Diener, Edward, and Eunkook M. Suh. 1999. Academic. “National Differences in Subjective Well- Bryan, Gharad, Shyamal Chowdhury, and Ahmed Being.” In Well-Being: The Foundations Mobarak. 2012. “Seasonal Migration and of Hedonic Psychology, edited by Daniel Risk Aversion.” Discussion paper 8739, Centre Kahneman, Edward Diener, and Norbert for Economic Policy Research, London. Schwarz, 434–50. New York: Russell-Sage. Cagetti, Marco, and Mariacristina De Nardi. Do, Quy-Toan, and L aksh m i Iyer. 2010. 2008. “Wealth Inequality: Data and Models.” “Geography, Poverty and Confl ict in Nepal.” Macroeconomic Dynamics 12 (suppl. S2): Journal of Peace Research 47 (6): 735–48. 285–313. Drèze, Jean, and Amartya K. Sen. 1991. Hunger Case, Anne, and Angus Deaton. 2009. “Health and Public Action. Oxford and New York: and Well-Being in Udaipur and South Africa.” Oxford University Press. In Developments in the Economics of Aging, Duflo, Esther. 2012. “Women Empowerment edited by David A. Wise, 317–49. Chicago: and Economic Development.” Journal of University of Chicago Press. Economic Literature 50 (4): 1051–79. Cohen, Sheldon, William J. Doyle, Ronald B. Dworkin, Ronald. 1981a. “What Is Equality? Turner, Cuneyt M. Alper, and David P. Skoner. Part 1: Equality of Welfare.” Philosophy and 2003. “Emotional Style and Susceptibility to Public Affairs 10 (3): 185–246. the Common Cold.” Psychosomatic Medicine ———. 1981b. “What Is Equality? Part 2: Equality 65 (4): 652–57. of Resources.” Philosophy and Public Affairs Collier, Paul, and Anke Hoeffler. 1998. “On 10 (4): 283–345. Economic Causes of Civil War.” Oxford Eynde, Oliver Vanden. 2013. “Targets of Violence: Economic Papers 50 (4): 563–73. Evidence from India’s Naxalite Conflict.” ———. 2004. “Greed and Grievance in Civil Working Paper. Paris School of Economics. War.” Oxford Economic Papers 56 (4): Ferreira, Francisco, and Maria Ana Lugo. 2013. 563–95. “Multidimensional Poverty Analysis: Looking Datt, Gaurav, and Martin Ravallion. 2009. for a Middle Ground.” World Bank Research “Has India’s Economic Growth Become More Observer 28 (2): 220–35. Pro-Poor in the Wake of Economic Reforms?” Frey, Bruno S., and Alois Stutzer. 2002. “What Policy Research Working Paper 5103, World Can Economists Learn from Happiness Bank, Washington, DC. Research?” Journal of Economic Literature Davies, James B., and Anthony F. Shorrocks. 40 (2): 402–35. 2000. “The Distribution of Wealth.” In Galasso, Emanuela, and Martin Ravallion. 2005. Handbook of Income Distribution, Vol. 1, “Decentralized Targeting of an Antipoverty edited by Anthony B. Atkinson and François Program.” Journal of Public Economics 89 Bourguignon, 605 –75. Oxford: Nor th- (4): 705–27. Holland, Elsevier. Gomes, Joseph Flavian. 2011. “The Political Davies, James B., Anthony Shorrocks, Susanna Economy of the Maoist Conflict in India: Sandström, and Edward N. Wolff. 2008. “The An Empirical Analysis.” Preliminary draft. World Distribution of Household Wealth.” In http://www.uclouvain.be/cps/ucl/doc/core Personal Wealth from a Global Perspective, /documents/Gomes.pdf. ed ited by Ja me s B . Dav ie s , 395 – 418. Hoff, Karla. 2012. “The Effect of Inequality New York: Oxford University Press. on Aspirations.” Background paper for this Deaton, Angus. 2005. “Measuring Poverty in report, World Bank, Washington, DC. a Growing World (or Measuring Growth in Hoff, Karla, and Priyanka Pandey. 2006. a Poor World).” Review of Economics and “Discrimination, Social Identity, and Durable Statistics 87 (1): 1–19. Inequalities.” American Economic Review Deaton, Angus, and Valerie Kozel. 2005. “Data 96 (2): 206–11. and Dogma: The Great Indian Poverty Debate.” ———. 2012. “Making Up People: The Effect of World Bank Research Observer 20 (2): 177–99. Identity on Preferences and Performance in a 62 ADDRESSING INEQUALITY IN SOUTH ASIA Modernizing Society.” Policy Research Working Global Development Working Paper No. 302, Paper 6223, World Bank, Washington, DC. Washington, DC. Horowitz, Donald L. 1998. “Structure and Keister, Lisa A. 2000. Wealth in America: Trends Strategy in Ethnic Confl ict.” Paper prepared in Wealth Inequality. New York: Cambridge for the Annual World Bank Conference on University Press. Development Economics, Washington, DC, Kiecolt-Glaser, Janice K., Lynanne McGuire, April 20–21. Theodore F. Robles, and Ronald Glaser. 2002. ———. 2000. Ethnic Groups in Conflict. 2nd ed. “Psychoneuroimmunology: Psychological Berkeley and Los Angeles: University of Influences on Immune Function and Health.” California Press. Journal of Consulting and Clinical Psychology Human Rights Watch. 2014 “They Say We’re 70 (3): 537–47. Dirty”: Denying an Education to India’s Kolm, Serge-Christophe. 1977. “Multidimensional Marginalized. New York: Human Rights Egalitarianisms.” Quarterly Journal of Watch. Economics 91 (1): 1–13. Inglehart, Ronald. 2000. World Values Surveys Korinek, Anton, Johan A. Mistiaen, and Martin and European Values Surveys, 1981–1984, Ravallion. 2006. “Survey Nonresponse 1990–1993 and 1995–1997. Ann Arbor, MI: and the Distribution of Income.” Journal of Inter-university Consortium for Political and Economic Inequality 4 (1): 33–55. Social Research. Lam, Wai Fung. 1998. Governing Irrigation Iyer, Lakshmi. 2009. “The Bloody Millennium: Systems in Nepal: Institutions, Infrastructure, Internal Conflict in South Asia.” Business, and Collective Action. Oakland, CA: ICS Press. Government, and the International Economy Li, Shi, and Renwei Zhao. 2008. “Changes in the Unit, Working Paper 09- 086, Harvard Distribution of Wealth in China, 1995–2002.” Business School, Cambridge, MA. In Personal Wealth from a Global Perspective, Juster, F. Thomas, and Kathleen A. Kuester. 1991. edited by James B. Davies, 93–111. New York: “Differences in the Measurement of Wealth, Oxford University Press. Wealth Inequality and Wealth Composition Lichbach, Mark Irving. 1989. “An Evaluation Obtained from Alternative U.S. Wealth of ‘Does Economic Inequality Breed Political Surveys.” Review of Income and Wealth 37 Conflict?’ Studies.” World Politics 41 (4): (1): 33–62. 431–70. Kahneman, Daniel. 2003a. “A Psychological Lucas, Richard, Edward Diener, and Eunkook Perspec tive on E conom ics.” A m e ric an M. Suh. 1996. “Discriminant Validity of Well- Economic Review 93 (2): 162–68. Being Measures.” Journal of Personality and ———. 2003b. “Maps of Bounded Rationality: Social Psychology 71 (3): 616–28. Psychology for Behavioral E conomics.” L u g o , M a r í a A n a . 2 0 0 7. “ C o m p a r i n g American Economic Review 93 (5): 1449–75. Multidimensional Indices of Inequality: ———. 2011. Thinking, Fast and Slow. New York: Methods and Application.” Research on Farrar, Straus and Giroux. Economic Inequality 14: 213–36. Kahneman, Daniel, and Alan B. Krueger. 2006. Maasoumi, Esfandiar. 1986. “The Measurement “Developments in the Measurement of and Decomposition of Multi-dimensional Subjective Well-Being.” Journal of Economic Inequality.” Econometrica: Journal of the Perspectives 20 (1): 3–24. Econometric Society 54 (4): 991–97. Kahneman, Daniel, Alan B. Krueger, David ———. 1999. “Multidimensioned Approaches to A. Schkade, Norbert Schwarz, and Arthur Welfare Analysis.” In Handbook of Income A. Stone. 2004. “A Survey Method for Inequality Measurement, edited by Jacques Characterizing Daily Life Experience: The Day Silber, 437– 84. Nor well, M A: K luwer Reconstruction Method.” Science 306 (5702): Academic. 1776–80. Macours, Karen. 2011. “Increasing Inequality and Kahneman, Daniel, and Amos Tversky. 1984. Civil Conflict in Nepal.” Oxford Economic “Choices, Values, and Frames.” American Papers 63 (2011). Cambridge, U.K.: Oxford Psychologist 39 (4): 341–50. University Press. Kapur, Devesh, Kishore Gawande, and Shanker Madison, James. (1787) 1961. “Federalist 10.” Satyanath. 2012. “Renewable Resource Shocks In Alexander Hamilton, James Madison, and Conflict in India’s Maoist Belt.” Center for and John Jay, The Federalist, edited by B. F. WHY INEQUALIT Y MAT TERS 63 Wright, 77–84. Cambridge, MA: Belknap Piketty, Thomas. 2011. “On the Long-Run Press of Harvard University. Evolution of Inheritance: France 1820–2050.” Malik, Sadia. 2009. “Horizontal Inequalities and Quarterly Journal of Economics 126 (3): Violent Conflict in Pakistan: Is There a Link?” 1071–131. Economic and Political Weekly. Piketty, Thomas, Gilles Postel-Vinay, and ———. 2011. “An Empirical Investigation of Jean-Laurent Rosenthal. 2006. “Wealth the Relationship between Food Insecurity, Concentration in a Developing Economy: Paris Landlessness, and Violent Conflict in Pakistan.” and France, 1807–1994.” American Economic PIDE Working Paper, Pakistan Institute of Review 96 (1): 236–56. Development Economics, Islamabad. Piketty, Thomas, and Emmanuel Saez. 2003. Mansuri, Ghazala. 2013. “Inequality and “Income Inequality in the United States, Development.” Background paper for this 1913–1998.” Quarterly Journal of Economics report, World Bank, Washington, DC. 118 (1): 1–41. Milanovic, Branko. 2009. “Global Inequality Ravallion, Martin. 2003. “Measuring Aggregate and the Global Inequality Extraction Ratio: Welfare in Developing Countries: How Well The Story of the Past Two Centuries.” Policy Do National Accounts and Surveys Agree?” Research Working Paper 5044, World Bank, Review of Economics and Statistics 85 (3): Washington, DC. 645–52. ———. 2011a. The Haves and the Have-Nots: ———. 2014. “I ncome I nequa l it y i n t he A Brief Idiosyncratic History of Global Developing World.” Science 344 (6168): Inequality. New York: Basic Books. 851–55. ———. 2011b. World s Apar t: Me a suring Rawls, John. 1971. A Theor y of Justice . In t e r n a tio n al a n d Global In equ alit y . Cambridge, MA: Harvard University Press. Princeton, NJ: Princeton University Press. Roemer, John E. 1998a. Equality of Opportunity. Milanovic, Branko, Peter H. Lindert, and Jeffrey G. Cambridge, MA: Harvard University Press. Williamson. 2011. “Pre-Industrial Inequality.” ———. 1998b. Theories of Distributive Justice. Economic Journal 121 (551): 255–72. Cambridge, MA: Harvard University Press. M it ra , A n i rba n , a nd D ebraj R ay. 2013. Roemer, John E., and Alain Trannoy. 2013. “Implications of an Economic Theory of “ E q u a l i t y o f O p p o r t u n i t y.” C o w l e s Conflict: Hindu-Muslim Violence in India.” Foundation Discussion Paper No. 1921, NBER Working Paper No. 19090, National Cowles Foundation for Research in Economics, Bureau of Economic Research, Cambridge, MA. Yale University, New Haven, CT. Nagel, Thomas. 1986. The View from Nowhere. Sedwal, Mona, and Sangeeta Kamat. 2011. New York: Oxford University Press. “Education and Social Equity in Elementary Naritomi, Joana, Rodrigo R. Soares, and Education.” In Who Goes to School, edited Juliano J. Assunção. 2007. Rent Seeking by R. Govinda, 87–123. Cary, NC: Oxford and the Unveiling of ‘De Facto’ Institutions: University Press. Development and Colonial Heritage within Sen, Amartya, ed. 1973. On Economic Inequality. Brazil.” NBER Working Paper No. 13545, New York: Oxford University Press. National Bureau of Economic Research, ———. 1980. “Equality of What?” In The Tanner Cambridge, MA. Lectures on Human Values, Vol. 1, edited by Nozick, Robert. 1974. Anarchy, State, and S. McMurrin, 353–69. Cambridge: Cambridge Utopia. New York: Basic Books. University Press. OECD (Organisation for Economic Co-operation ———. 1992. Inequality Reexamined. New York: and Development). 2013. OECD Framework Oxford University Press. fo r S t a t i s t i c s o n t he D i s t r i b u t i o n of Seth, Suman. 2010. “A Class of Distribution Hou sehold Income, C on sumption an d and Association Sensitive Multidimensional Wealth. Paris: OECD Publishing. http://dx.doi Welfare Indices.” Journal of Economic .org/10.1787/9789264194830-en. Inequality 11 (2): 1–30. Paes de Barros, Ricardo, Francisco H. G. Ferreira, Stiglitz, Joseph E. 2013. “Inequality Is Holding José R. Monlinas Vega, and Jaime Saavedra Back the Recovery.” New York Times , Chanduvi. 2009. Measuring Inequality of January 19. Opportunities in Latin America and the Stone, A rthur, and Saul Shiffman. 1994. Caribbean. Washington, DC: World Bank. “Ecological Momentary Assessment (EMA) in 64 ADDRESSING INEQUALITY IN SOUTH ASIA Behavioral Medicine.” Annals of Behavioral Pathways to Human Development. New York: Medicine 16 (3): 199–202. Palgrave Macmillan. Stutzer, Alois, and Bruno S. Frey. 2012. “Recent Ura, Karma, Sabina Alkire, and Tshoki Zangmo. Developments in the Economics of Happiness: 2012. GNH and GNH Index . Thimpu, A Selective Overview.” Discussion Paper No. Bhutan: Centre for Bhutan Studies. 7078, Forschungsinstitut zur Zukunft der Urry, Heather, Jack Nitschke, Isa Dolski, Daren Arbeit (Institute for the Study of Labor), Bonn, Jackson, Kim Dalton, Corrina Mueler, Melissa Germany. Rosenkranz, Carol Ryff, Burton Singer, and Subramanian, S., and D. Jayaraj. 2008. “The Richard Davidson. 2004. “Making a Life Worth Distribution of Household Wealth in India.” Living.” Psychological Science 15 (6): 367–72. In Personal Wealth from a Global Perspective, Veenhoven, Ruut. 2004. “Subjective Measures of edited by James B. Davies, 112–33. New York: Well-Being.” WIDER Discussion Paper No. Oxford University Press. 2004/07, United Nations University World Thekaekara, Mari Marcel. 2004. “No Hope, Institute for Development Economics Research No Fut u re.” T he Hi n d u , Febr ua r y 8. (UNU-WIDER). http://w w w.hindu.com /mag /20 04/02 /08 Wolff, Edward N. 1987. “Estimates of Household /stories/2004020800010400.htm. Wealth Inequality in the U.S., 1962–1983.” Tsui, Kai-yuen. 1999. “Multidimensional Review of Income and Wealth 33 (3): 231–56. Inequality and Multidimensional Generalized ———. 1996. “International Comparisons of Entropy Measures: An Axiomatic Derivation.” Wealth Inequality.” Review of Income and Social Choice and Welfare 16 (1): 145–57. Wealth 42 (4): 433–51. Tversky, Amos, and Daniel Kahneman. 1974. World Bank. 2005. World Development Report “Judgment under Uncertainty: Heuristics and 2006: Equity and Development. Washington, Biases.” Science 185 (4157): 1124–31. DC, and New York: World Bank and Oxford U N DP (United Nations Development University Press. Programme). 2010. Human Development ———. 2011. More and Better Jobs in South Asia. Report 2010: The Real Wealth of Nations: Washington, DC: World Bank. The Extent of Inequality 2 A ssessing the extent of inequality in a the bottom are among the most destitute particular country or region can be people worldwide. These additional pieces of sensitive to the indicators used. This information cast doubts on the picture that is particularly so in South Asia, where mon- emerges from standard monetary indicators etary and nonmonetary indicators of well- computed from household surveys. being yield sometimes radically different The assessment of trends in inequality pictures. Monetary indicators are based on also depends on the indicator considered. cash and in-kind income or consumption; Monetary inequality is increasing in most they may also refer to wealth. Nonmonetary of South Asia. Growth has been effective indicators capture aspects of well-being, at reducing poverty in the region, as it was such as access to basic services, environmen- earlier in East Asia. Yet growth has led to tal quality, or voice in public matters. increasing inequality in both regions, despite Standard indicators based on consumption being neutral overall with respect to distribu- per capita suggest that South Asia has modest tion at the global level. Monetary indicators of levels of inequality compared to other regions inequality have increased in the poorest coun- of the world. If Gini indexes are to be taken tries in the South Asia region in recent years at face value, the extent of inequality in South and fallen only among the richest ones. A Asia is much lower than in China, Mexico, common hypothesis in economic development or South Africa. Inequality in human devel- (the so-called Kuznets curve) is that inequality opment outcomes is higher than inequality increases until countries attain a certain level in expenditures per capita, but it is not high of income and only declines as they become enough to make South Asia an outlier among richer. This hypothesis is controversial, and developing regions. no real consensus exists about its validity. But Yet the combination of monetary and monetary indicators from South Asia are con- nonmonetary indicators yields a somewhat sistent with it. If so, a vast majority of South different perspective. Figures on tax fil- Asians could be confronted with an increase ings and asset holdings suggest that South in monetary inequality in the years to come. Asians at the top are disproportionately rich Trends in nonmonetar y inequalit y on a global scale. At the same time, figures are more mixed. On some dimensions, on malnutrition reveal that South Asians at South Asia is becoming more equal, whereas 65 66 ADDRESSING INEQUALITY IN SOUTH ASIA on others, gaps keep widening. For example, complementary evidence for policy makers to the region has made substantial progress in act upon. educational attainment at the primary level; The diversity of the economic forces however, inequality is increasing in other underlying the level of inequality and its dimensions, notably in health outcomes. changes over time is useful to articulate a The diversity of the observed trends, conceptual framework to be used in the rest depending on the indicator considered, of this report. Some circumstances at birth, makes it difficult to rely on a single metric such as gender and caste, shape the options to assess changes in inequality or progress available to individuals. As children age, the toward shared prosperity. This difficulty opportunities to accumulate human capital comes on top of the measurement issues become critically important; once they enter associated with some of the most com- the labor force, job opportunities and the mon indicators. For instance, lower survey possibility of reaping benefits from migration response rates among better-off households affect earnings prospects. Throughout life, and greater underestimation of their expendi- people experience shocks and are affected— tures by the available survey instruments may positively or negatively—by government poli- bias standard inequality measures such as the cies transferring resources to them or taxing Gini index or the income share of the bottom them, explicitly or implicitly. Through each 40 percent. But even in the absence of bias, of these phases, public policies may be able to the story on inequality in South Asia would affect the extent of inequality. be different if it were told based on monetary indicators, on education indicators, or on health indicators. Monetary indicators of This diversity of assessments is actually inequality welcome when interpreting what lies behind Different indicators are used in practice to the level of inequality and its changes over measure inequality in the distribution of time. Some of the indicators can be decom- income or consumption. Each one captures posed between population groups defined by different dimensions of inequality. Some, characteristics such as gender, location, and such as the Gini coefficient or the mean log caste, providing some clues. This exercise pro- deviation (MLD), focus on the entire dis- vides guidance on how to design policies in tribution and can be seen as measuring the relation to basic services to ensure equality of gap between actual consumption or income opportunity. Other indicators can be traced and the consumption or income that would over time for different population groups. prevail if there were no inequality. A Gini For instance, changes in occupational status coefficient of zero expresses perfect equal- between fathers and sons or in expenditures ity, whereas a Gini coefficient of one corre- per capita within the same generation reveal sponds to the extreme where all consumption the extent of mobility for each of the groups. or income is in the hands of one person Findings help disentangle the contribution or household. As for the MLD, an intui- jobs and migration make to overall inequal- tive explanation is that, multiplied by 100, ity. The same logic can be applied to shocks it shows the percentage difference between and public support: not all population groups the consumption of a randomly selected are equally affected by disease, economic individual and the average per capita con- crises, or natural disasters, and the magni- sumption in the country. The more unequal tude of the net transfers they receive from the distribution, the bigger the percent- government, or make to it, varies as well. age difference. Other indicators emphasize Results can be used to inform how social the extremes of the distribution. The Theil protection, taxation, and transfers should be index, for instance, increases more than organized. Patterns of subjective assessments linearly as the share of the rich increases. of inequality across population groups offer Finally, indicators such as the ratio between THE EXTENT OF INEQUALITY 67 the 90th and the 10th percentile focus on matter of values or ideology. The same ana- particular points in the distribution. lyst may choose one particular inequality None of the indicators is intrinsically bet- indicator to address an issue and another one ter than others. A focus on different points or to deal with a different issue. portions of the distribution or on the distri- Monetary indicators of inequality are gen- bution in its entirety reflects a judgment on erally computed based on individual records what aspect of inequality matters most. For from representative household surveys instance, some may prefer to put the spot- (box 2.1). Overall, these indicators suggest light on the top 1 percent and others on the that inequality in South Asia is moderate by bottom 40 percent. This is not necessarily a international standards (figure 2.1). BOX 2.1 South Asian household surveys used in this report Acronym Name Countries Years Description BLSS Bhutan Living Bhutan 2003, 2007 BLSS is a nationally representative survey. It gathers data on Standards Survey consumption expenditure and other household information, including demographic characteristics, education, health and employment of household members, and household remittances, housing, and access to public facilities and services. The BLSS sample size is 4,007 in 2003 and 9,798 in 2007. DHS Demographic and Bangladesh 1993, 2011 DHSs are nationally representative household surveys that Health Survey India 1992, 2005 provide data for a wide range of monitoring and impact Maldives 2009 evaluation indicators in the areas of population, health, and Nepal 1996, 2011 nutrition. Pakistan 1990, 2007 Sri Lanka 2007 HIES Household Income Bangladesh 2000, 2005, 2010 HIES of Bangladesh is the core nationally representative survey and Expenditure to provide important data such as income, expenditure, Survey consumption, and poverty situation. In particular, it contains information on demographic characteristics, education, health and employment of household members, and household housing, economic activities, income, and consumption. The first HIES round was conducted in 1973–74, and 15 rounds have been completed since then. The sample size of the 2010 survey is 12,240 households; 7,840 were from rural areas and 4,400 from urban areas. HIES Household Income Maldives 2002–03, HIES of Maldives collects detailed information on the and Expenditure 2009–10 expenditure, income, demographic, and socioeconomic Survey characteristics from local households living in the administrative islands of the country. The main objective is to produce reliable statistics on different components of income and expenditure of households in the capital, Malé, and the Atolls to assess the economic well-being of the population. Specifically, the results are used to bring about improvements in the country’s national accounts, consumer price index, and vulnerability and poverty statistics. The survey covers 834 households from Malé and 40 islands in the 2002–03 round, and 2,060 households from Malé and 39 islands in the 2009–10 round. (continues next page) 68 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 2.1 South Asian household surveys used in this report (continued) Acronym Name Countries Years Description HIES Household Income Sri Lanka 1995–96, HIES of Sri Lanka provides information on people’s household and Expenditure 2002–03, income and expenditure to measure their levels and changes Survey 2006–07, in living conditions. Data collected from this survey are used 2009–10 to observe the consumption patterns to compute various other socioeconomic indicators, such as poverty price indexes. Generally, the survey is conducted over a 12-month period to capture seasonal variations. The general sample size is 2,500 housing units. HIES Household Integrated Pakistan 2001–02, HIES of Pakistan was started in 1963 and has continued to Economic Survey 2004–05, be carried out with breaks. It was merged with the Pakistan 2007–08, Integrated Household Survey in 1998–99 and 2001–02. As 2010–11 provincial level surveys, they are also part of the Pakistan Social and Living Standards Measurement surveys. HIES provides important information on household income, savings, liabilities, consumption expenditure, and consumption patterns at national and provincial levels with urban/rural breakdown. It provides the requisite data on consumption for estimation of poverty. IHDS India Human India 2004–05 The IHDS is a nationally representative, multi-topic survey Development Survey of 41,554 households in 1,503 villages and 971 urban neighborhoods across India. The first round of interviews was completed in 2004–05. NLSS Nepal Living Nepal 1995, 2003, The main objectives of NLSS are to update data on living Standards Survey 2010 standards and assess the impact of various government policies and programs. Further, the survey tracks changes experienced by previously enumerated households during the past 15 and 7 years. It enumerated 7,020 households, of which 5,988 were from the cross-section sample and the remaining 1,032 were from the panel sample. NRVA National Risk Afghanistan 2005, 2007 NRVA is a nationally representative multipurpose survey, and Vulnerability covering a variety of development sectors and cross-cutting Assessment themes, such as poverty and the position of women. NSS National Sample India 1993–94, NSS is a large, nationally representative household sample Survey 2002–03, survey program launched for collection of data on the 2004–05, various aspects of the national economy required by different 2007–08, agencies of the government, both central and state. 2009–10, 2011–12 PSLM Pakistan Social and Pakistan Annually, PSLM is designed to provide social and economic indicators in Living Standards 2004–05 to the alternate years at provincial and district levels. It provides Measurement 2011–12 a set of representative, population-based estimates of social indicators and their progress under the Poverty Reduction Strategy Paper/Millennium Development Goals. The PSLM surveys are conducted at district and provincial levels, respectively, in alternate years. The sample size of PSLM surveys is approximately 80,000 households at district level and approximately 18,000 at provincial level. (continues next page) THE EXTENT OF INEQUALITY 69 BOX 2.1 South Asian household surveys used in this report (continued) Acronym Name Countries Years Description VPA Vulnerability and Maldives 2004 VPA covers households on all 200 inhabited islands. The scope Poverty Assessment of the survey includes demographic characteristics, education, health, and employment of household members, household housing, economic activities, consumption, and access to public facilities and services. It also reports information at the community level. The 2004 survey is the second round and the sample size is 2,728. Sources: Based on World Bank’s PovcalNet tool, http://iresearch.worldbank.org/PovcalNet/index.htm?4, and International Household Survey Network Survey Catalog, http://catalog.ihsn.org/index.php/. FIGURE 2.1 Based on standard monetary indicators, South Asia has moderate levels of inequality a. Gini coefficients 70 60 50 40 30 20 10 0 do m ca a, p. t, A Fr ny M l hi ia Ire ali Ja ia ite Sp n d I ain Fe Tu ria M ysia C co So B hile fri l an en k n yp rmaan Ko rab nce lad da Et tral d In gdo ia et ia Tu am M Lan ia Un Ta ldiv a Th uta s ai n Ni land ra y Chion Ar hana M tin a nm ay Be lan k gh ed a p. Ne sh ite nz es St a ala a lg d Au ium Ba a taly pa h A zi Bh ate de rke Fin ar a k ge n Af Sw stri pa d ani Pa ista Au lan op Ki nd Vi nes Sr nis re Re Re ut ra i De orw e ge e ist i ng na ex n t a I n s G n C N i G ian Un Eg ss Ru (continues next page) 70 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.1 Based on standard monetary indicators, South Asia has moderate levels of inequality (continued) b. Share of the poorest 40 percent 30 25 20 15 10 5 0 . d ay en an an ep y ria m sh al ia ali da d ka in m ly sia ia an es d on es ey na na ia ia le il ys er hi az a sia ge ico a ca an rw ed nist kist b R man ust lgiu ade ep hiop M ana elan Lan Spa etna Ita one zan hut ldiv ilan rati Stat urk Gha Chi di in ala Nig C Br fri l ni Ar ex nt In n o w a r A e l N i d an B Ma Tha ede ed T hA i a a t r i Tu F N S gh P Ar Ge B ng C I Sr M E V In T M t, it ut Af a p B n F Un So y Eg s sia Ru c. Ratio of richest to poorest decile 60 50 40 30 20 10 0 Ge ade . Eg Nor an Ba rab en Ne ria lg a i L ia Ire am a Fe hail ia n M ina ala o ge ria ut Ch a Br a l p kis n rm sh Et ium ald ia nz n Tu nia Ch na Ni ysia h A ile et a Sp s Bh ain t, A e y Au any ite urk y Gh tes Be Indii d ey an nd M l In ana d ra d il ive al do d in c M exic tio pa Vi ank Pa ista ng Re Ta uta yp S wa Un T Ital Sr iop ian T nis M nes az C n de an fri t d st Ar ge a nt a n a gh la a St l h w Af Fin So ss Ru Sources: Based on World Bank’s World Development Indicators (WDI ) database, http://data.worldbank.org/data-catalog/world-development-indicators, and Organisation for Economic Co-operation and Development’s Income Distribution and Poverty data series, http://stats.oecd.org/Index.aspx?DataSetCode=IDD. Note: Orange and light brown bars indicate countries where inequality is estimated based on consumption per capita. Light blue bars indicate countries with estimates based on income per capita. THE EXTENT OF INEQUALITY 71 Arguably, the comparison is tainted by the consumption of the richest members of soci- nature of the monetary indicators considered ety. The survey questionnaires usually focus in different countries. In advanced economies, on the relatively basic basket of goods and as well as in many Latin American countries, services purchased by those who live around inequality is measured on the basis of income the poverty line. In so doing, they fail to per capita. In South Asian countries, in con- capture the more diverse and sophisticated trast, most surveys convey information about ways in which the better-off spend their household expenditure, allowing the estima- money—and to remind respondents about tion of consumption per capita but not of them. Richer households also tend to shun household income. Within a same country, surveys of this sort. The monetary compen- income inequality is generally higher than sation offered to respondents may provide a consumption inequality. However, the con- sufficient incentive to participate for the poor clusion that monetary inequality in South and near-poor but may be seen as a pittance Asia is moderate holds even when comparing by wealthier households. The latter may also only countries for which data on expenditure be concerned about possible tax implications per capita are available. of their responses. One indication of under- Gini coefficients of the most recent avail- reporting is the size of the discrepancies able years ranged between 0.28 and 0.40 in between levels and growth rates of consump- South Asian countries. Based on these esti- tion as measured by sample surveys and by mates, inequality in South Asian countries national accounts. Disconnect between the was clearly lower than in China, Mexico, two major data sources is large in several or South Africa. The share of the poorest South Asian countries, especially in India. 40 percent of households in total consump- Individual tax returns can be used to tion also suggests that inequality in South examine the extent of undercounting of the Asian countries is not very high by interna- rich in sample surveys. According to this tional standards. Within-country inequal- data source, the income share of India’s top ity looks even less severe based on the gap 0.01 percent had more than doubled during between the observed extremes: the ratios the 1980s—from less than 0.4 percent to between top and bottom deciles were smaller more than 0.8 percent (Banerjee and Piketty than 10 in South Asian countries, whereas 2005). The trend was similar for the top the ratio was 19 in Mexico, 22 in Nigeria, 1 percent in the 1980s. But in the 1990s, and a staggering 44 in South Africa. a clear divergence arose between what was The conclusion based on international happening in the top 0.01 percent, and in benchmarking needs to be assessed with sev- the rest of the top 1 percent. From the mid- eral caveats in mind, however. The first is 1990s to the end of the decade, the share of the already mentioned distinction between the top 0.01 percent was in the 1.5 percent inequality in consumption per capita and to 2 percent range, whereas the share of inequality in income per capita. Not many the top 0.1 percent was in the 3 percent to household surveys in South Asia collect accu- 4.5 percent range (figure 2.2). The rapid rate data on both, which would allow a direct increase in the share of the former sug- comparison of measures of inequality based gests that economic growth in recent years on the two variables. An exception is the was more strongly biased in favor of the 2004–05 IHDS. Based on this survey, the Gini ultra-rich. index for consumption was estimated at 0.34, The rapid growth of income at the top while the Gini index for income was estimated of the distribution goes some way toward to be 20 points higher at 0.54. This is less explaining the gap observed during the than the income inequality observed in South 1990s between average consumption esti- Africa, but it is still more than in Mexico. mates based on sample surveys and those A second caveat is that household sur- based on national accounts. Assuming veys may not capture well the income or the that the top 1 percent is not captured by 72 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.2 Top incomes have been rising in India since the 1980s inequality measures based on consumption are biased downward, probably by a sub- a. Income share of the top 0.1 percent, 1922–99 stantial margin. 9 The distribution of wealth provides a complementary perspective on monetary 8 inequality. Every decade, a special round of the Indian NSS provides comprehensive 7 information on asset holdings and debt for 6 the population at large. The 59th round of the NSS, for example, presented detailed 5 wealth information for both rural and urban Percent households in 2002–03. Eight broad types 4 of assets were covered, including land, building, agricultural machinery, nonfarm 3 business equipment, transportation equip- 2 ment, consumer durables, and financial assets. Debts included cash loans, in-kind 1 loans, and other payables by households. Net worth is calculated as total financial 0 and nonfinancial assets net of total debts. This information indicates consider- 19 2 26 19 0 19 4 19 8 19 2 19 6 19 0 54 19 8 19 2 19 6 19 0 74 19 8 19 2 19 6 19 0 19 4 98 2 3 3 3 4 4 5 5 6 6 7 7 8 8 9 9 19 19 19 19 able inequality in asset holdings and in net b. Income share of the top 0.01 percent, 1922–99 worth in India. At the household level, the 3.5 Gini coefficient is 0.668 for asset holdings and 0.680 for net worth; the MLD reaches 1.049 for asset holdings and 1.045 for net 3.0 worth. As in other countries, the wealth distribution is more concentrated than the 2.5 distribution of income per capita and espe- cially more concentrated than that of expen- 2.0 ditures per capita (figure 2.3). Percent Perhaps more striking than the extent of 1.5 inequality is the vulnerability of the least wealthy. Net worth provides a measure of the ability of households to support their con- 1.0 sumption in the event of an adverse shock. Households at the bottom are obviously more 0.5 vulnerable than those at the top. But what is remarkable is how much more vulnerable 0 they are (figure 2.4). According to this data source, the aver- 19 2 19 6 30 19 4 19 8 19 2 19 6 19 0 19 4 19 8 62 19 6 70 19 4 78 19 2 19 6 19 0 19 4 98 2 2 3 3 4 4 5 5 5 6 7 8 8 9 9 19 19 19 19 19 age net worth of the top 10 percent of the population was more than 380 times that of Source: Based on Banerjee and Piketty 2005. the bottom 10 percent. For a typical house- hold among the top 10 percent, this net household surveys is not enough to account worth could support consumption for more for the full gap but explains 20 percent than 23 years. In contrast, for a typical to 40 percent of it, depending on the household in the bottom 10 percent, this net assumptions. This fraction is large enough worth was sufficient to support consump- to give credence to the hypothesis that tion for less than three months. THE EXTENT OF INEQUALITY 73 Measuring the net worth of the better- over a quarter of India’s billionaire wealth off is even more challenging than measur- is estimated to be inherited, 40 percent is ing their income or expenditure. Lack of based on inheritance, and 60 percent origi- response and misrepresentation are bound to nates from “rent-thick sectors” such as real be even more serious problems, resulting in a estate, infrastructure, construction, mining, distorted picture of the wealth status of the very rich. To address this shortcoming of the FIGURE 2.3 The distribution of wealth is more data, direct information on billionaire wealth concentrated than that of consumption in India can be useful. The concentration of billionaire wealth appears to be unusually large in India but 100 much less so in Nepal. These are the only two countries in the region with publicly known Cumulative monetary proportion billionaires. According to Forbes magazine (2014), total billionaire wealth represented about 10 percent of gross domestic product (GDP) in 2012. As such, India is an outlier in the ratio of billionaire wealth to GDP among economies at a similar development level (figure 2.5). One concern is whether extraordinary wealth at the top of the distribution is the result of exceptional entrepreneurship or 0 100 substantial rent seeking. There is no doubt Cumulative household proportion that India has world-class entrepreneurs, Line of perfect equality commanding admiration for their innova- Lorenz curve of monthly consumption tion and management capacity, and many Lorenz curve of net worth of them operate successfully in highly com- petitive global markets. At the same time, Source: Based on NSS 2002–03. FIGURE 2.4 The least wealthy are alarmingly vulnerable 1,800,000 300 1,600,000 250 1,400,000 Months of consumption 1,200,000 200 Indian rupees 1,000,000 150 800,000 600,000 100 400,000 50 200,000 0 0 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th Asset-holding deciles, lowest to highest Average asset holding Average net worth Average net worth (Indian rupees) (Indian rupees) (months of consumption) Source: Based on NSS 2002–03. 74 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.5 Billionaire wealth in India is exceptionally large 20 Philippines Aggregate net wealth (percent of GDP, 2012) 18 16 Ukraine 14 India Mexico 12 Thailand 10 8 6 Nepal 4 2 Vietnam Romania 0 0 5,000 10,000 15,000 20,000 GDP per capita (PPP, constant 2011 international dollars, 2012) Sources: Based on Forbes magazine’s billionaires database, http://www.forbes.com/billionaires/, and World Bank’s WDI database, http://data.worldbank.org /data-catalog/world-development-indicators. Note: PPP = purchasing power parity. telecommunications, cement, and media. as reflected, for example, in health and This does not necessarily mean that wealth education outcomes. Differences in these was acquired through the legal or illegal outcomes can affect individuals’ abilities exercise of influence, but highlights that the to do what they would value doing and to potential for rent extraction exists (Gandhi convert different means into well-being. and Walton 2012). Sen (1980, 1992) often takes the case of a person with physical disabilities to illustrate the point: such a person cannot Nonmonetary dimensions of function in the way an able-bodied person inequality can. It follows that the extent of depriva- Both monetary and nonmonetary indicators tion of a person with physical disabili- capture important dimensions of inequality ties in comparison with others cannot be in outcomes. And they tend to be correlated, adequately judged by looking at incomes, as shown by the fact that health status or edu- because the person may be greatly disad- cational attainment is typically higher among vantaged in converting income into the people who are better off in monetary terms. desirable achievements. The same logic can But the correlation is not perfect because be applied to a person who is malnour- monetary and nonmonetary indicators cap- ished, ill, or illiterate. ture different concepts and can vary indepen- Nonmonetary indicators of well-being are dently. Therefore, nonmonetary indicators very unevenly distributed among the South provide additional information on distribu- Asian population. A comparison of health tion of well-being, beyond what is provided outcomes across population quintiles, defined from monetary indicators of inequality. by a wealth index, is revealing in this respect. Nonmonetary dimensions of inequality Gaps in neonatal mortality (death within the include the dispersion in human capabilities first 28 days of life) and in under-five child THE EXTENT OF INEQUALITY 75 mortality (death within the first five years of diversity, with Maldives, Sri Lanka, and life) between the top and the bottom quintiles Bangladesh being the countries with the are large, especially in India and Pakistan. lowest levels of inequality. But the disper- For children who live, the main challenge is sion is much narrower among younger to be well nourished. Children (under two population cohorts. years old) belonging to the poorest quintile For the younger South Asian cohorts, are more likely to be stunted in every country inequalit y in relation to education is in the region, although the gap is relatively increasingly driven by quality, rather than less glaring in Sri Lanka and the Maldives by access. Evidence on learning outcomes (figure 2.6). is sparser than data on educational attain- By international standards, the health out- ment, but it is quite consistent (World comes of the poor in South Asia are among Bank 2013). An important source of infor- the worst worldwide. The share of children mation on the extent of cognitive skills (under five years old) who are stunted among acquired in school is a study that covered the poorest quintile is above 50 percent 2,399 government-run schools in India in in Bangladesh and Nepal and reaches 2009 (Educational Initiatives 2010). The 60 percent in India (figure 2.7). India and coverage included samples from approxi- Pakistan also have some of the highest infant mately 74 percent of the population from mortality rates and under-five child mortality both urban and rural India. The actual rates in the poor across all comparators. Of number of test takers was 101,643, repre- 1,000 children born in India’s poorest popu- senting a presence rate of 66 percent—more lation quintile, 82 will die within 12 months or less in line with the regular presence and 117 within five years. The figures for rate found in government schools—and Pakistan are 94 and 120, respectively. about 5,600 teachers took the background Inequality in educational attainment questionnaires. is large as well, although showing wide Average test performance was low: stu- differences across the region. From an dents scored barely 50 percent in grade international perspective, countries in 4, between 46 and 48 percent in grade 6, South Asia seem to lie at both ends of the and between 46 and 47 percent in grade 8. developing-country range. A mong the But considerable dispersion existed in test countries for which comparable data are scores across students from different back- available, Maldives and Sri Lanka exhibit grounds. The inequality in learning out- the lowest gaps in educational attainment comes can be seen by comparing test scores between the population quintiles with the of children whose households have both a highest and the lowest expenditures per radio and a television to those who have nei- capita (figure 2.8). At the other end, gaps in ther (figure 2.10). The mean test scores for Nepal, Afghanistan, and especially Bhutan students in the first group are higher across are larger than elsewhere. the entire distribution than for those from However, the international comparison the second group, especially in the lower is highly sensitive to the age group con- grades. sidered, because educational attainment is Gaps in learning outcomes are large expanding rapidly throughout the region. in other countries of the region as well. The ratio of the years of education between Bangladesh and Sri Lanka have achieved a the richest and the poorest quintiles in the high level of equality in years of schooling, distribution is generally high among people but learning outcomes differ significantly 20 to 29 years of age (figure 2.9), but gaps across income groups. In Bangladesh, stu- are being eliminated at the primary level, dent achievement in English was 250 percent as reflected in the moderate dispersion in higher among the richest quintiles than among years of education among children 6 to 11 the poorest. In written math competency, years of age. Again, there is considerable the difference was more than 100 percent 76 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.6 Gaps in health outcomes are wide a. Neonatal mortality Bangladesh 1st (poorest) 2nd 3rd 4th 5th (richest) India 1st (poorest) 2nd 3rd 4th 5th (richest) Maldives 1st (poorest) 2nd 3rd 4th 5th (richest) Nepal 1st (poorest) 2nd 3rd 4th 5th (richest) Pakistan 1st (poorest) 2nd 3rd 4th 5th (richest) Sri Lanka 1st (poorest) 2nd 3rd 4th 5th (richest) 0 5 10 15 20 25 30 35 40 45 50 55 60 65 Mortality rate (deaths per 1,000 births) b. Under-five mortality Bangladesh 1st (poorest) 2nd 3rd 4th 5th (richest) India 1st (poorest) 2nd 3rd 4th 5th (richest) Maldives 1st (poorest) 2nd 3rd 4th 5th (richest) Nepal 1st (poorest) 2nd 3rd 4th 5th (richest) Pakistan 1st (poorest) 2nd 3rd 4th 5th (richest) Sri Lanka 1st (poorest) 2nd 3rd 4th 5th (richest) 0 20 40 60 80 100 120 140 160 180 200 Mortality rate (deaths per 1,000 births) (continues next page) THE EXTENT OF INEQUALITY 77 FIGURE 2.6 Gaps in health outcomes are wide (continued) c. Stunting Bangladesh 1st (poorest) 2nd 3rd 4th 5th (richest) India 1st (poorest) 2nd 3rd 4th 5th (richest) Maldives 1st (poorest) 2nd 3rd 4th 5th (richest) Nepal 1st (poorest) 2nd 3rd 4th 5th (richest) Pakistan 1st (poorest) 2nd 3rd 4th 5th (richest) Sri Lanka 1st (poorest) 2nd 3rd 4th 5th (richest) 0 10 20 30 40 50 60 Share of children (percent) Sources: Based on DHS 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 1990 for Pakistan, and DHS 2007 for Sri Lanka. Note: Neonatal mortality rate is the number of deaths within the first 28 days of life per 1000 live births. Under-five mortality rate is the number of deaths to children younger than five years per 1,000 live births. Stunting is the percentage of children younger than two years whose height-for-age ratio is two standard deviations or more below the World Health Organization (WHO) Child Growth Standards. (Asadullah and others 2009). In Sri Lanka, on less than $1.25 a day in purchasing the test scores were a full standard deviation power parity (PPP). The pattern is clearly higher among students from the richest quin- discernable when plotting comparable tiles, relative to those from the poorest stu- estimates of the poverty rate for countries dents (Aturupane, Glewwe, and Wisniewski in the two regions against the income per 2013). Learning outcomes are also closely capita (also in PPP terms) those countries related to ethnicity. In Nepal, Madhesis and had in the years to which the poverty rates Dalits score much lower than other ethnic refer (figure 2.11). East Asian countries are groups. In Sri Lanka, Burgher children score on average richer than their South Asian highest on standardized tests and Tamil chil- counterparts, but both regions demonstrate dren lowest (World Bank 2013). enough diversity in incomes per capita to find countries at almost any point in the range from low- to middle-income levels of Monetary inequality is increasing development. Growth has been effective at reducing The steepness of the relationship between poverty in South Asia, as it was earlier in the poverty rate and income per capita indi- East Asia. In both regions, higher levels cates how good growth has been for the of income per capita have been associated poor. For instance, the lower the initial with a lower share of the population living poverty rate, the more difficult is achieving 78 Mortality rate (deaths per 1,000 births) Mortality rate (deaths per 1,000 births) among the poorest quintile among the poorest quintile M M 100 150 200 250 50 100 120 140 20 40 60 80 0 0 ald ald iv ive Co es Co s lo lo m m bi bi a a Do Pe Pe m Vi ru ru in e Vi et ica tna Do n nR m m Ph am ep in ica ilipp Ph ubl n R in ilip ic ep es p u ADDRESSING INEQUALITY IN SOUTH ASIA Ni ines Ni blic ca ca ra ra g g In ua In ua do do ne ne sia Gu si at a Ne em Gu pa ala at l e Gh Ba mal an ng a a lad Ne es Ba p h ng l a Ke lad ny es a. Infant mortality a h b. Under-five mortality Br Ke az ny il Za a Gh m an bi a Ug a Bo an FIGURE 2.7 The health outcomes of the poor are among the worst worldwide liv ia da In In di di Pa a a kis Br ta az Ug n i an Bo l d liv Za a Pa ia m kis bi a ta n Ni Ni ge ge ria ria (continues next page) THE EXTENT OF INEQUALITY 79 FIGURE 2.7 The health outcomes of the poor are among the worst worldwide (continued) c. Stunting 70 60 among the poorest quintile Share of children (percent) 50 40 30 20 10 0 an a ia a ria h a a ru a da ar m a lic s l di bi bi ny ive an gu pa es liv nm na Pe ub ut ge an In m m lad Ke Gh Ne ra Bo ald Bh et Ni ep Za Ug lo ya ca ng Vi Co M nR M Ni Ba ica in m Do Source: Based on World Bank Health Nutrition and Population Statistics database, http://datatopics.worldbank.org/hnp/WealthQuintiles. Note: Infant mortality rate is the number of deaths to children younger than 12 months per 1,000 live births. Under-five mortality rate is the number of deaths to children younger than five years per 1,000 live births. Stunting is the percentage of children younger than five years whose z-scores are two standard deviations or more below WHO Child Growth Standards. FIGURE 2.8 Schooling among young adults is highly unequal in some countries in South Asia 6 Education attainment, ratio of richest 5 to poorest quintile 4 3 2 1 0 Ug nes Vi da Rw a ia In ria ka Za ia ng na Co blic ep u Ni h ilip a da Sr ives ca l a m Gh n an a an an l Pa a Ni razi di a Ph bi bi n R Per es gu ta ny liv s an ep na an a ge an In ne ut pi ist m m kis lad u Bo ald Ke ra B et N iL Bh do lo M gh Ba Af ica in m Do Sources: Based on NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, BLSS 2007 for Bhutan, NSS 2009–10 for India, HIES 2009–10 for Maldives, NLHS 2010 for Nepal, HIES 2010–11 for Pakistan, HIES 2009–10 for Sri Lanka, and World Bank Education Equality Country Profiles database, http://datatopics.worldbank .org/Education/wDHS/HProfiles.aspx. Note: Educational attainment is measured in years of schooling. The population considered are 20 to 29 years of age. 80 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.9 Gaps in educational attainment are much narrower among children than among adults 6 Education attainment, ratio of richest 5 to poorest quintile 4 3 2 1 0 an ka n a h an s l ive di pa ta es ist an ut In kis lad Ne ald an Bh iL Pa ng gh Sr M Ba Af Ages 6 to 11 Ages 20 to 29 Sources: Based on NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, BLSS 2007 for Bhutan, NSS 2009–10 for India, HIES 2009–10 for Maldives, NLHS 2010 for Nepal, HIES 2010–11 for Pakistan, and HIES 2009–10 for Sri Lanka. Note: Educational attainment is measured in years of schooling. FIGURE 2.10 Among children, similarities in schooling hide disparities in learning in India a. Test scores for Class 6 b. Test scores for Class 8 80 80 Average scores (in points) Average scores (in points) 60 60 40 40 20 20 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Percentiles Percentiles No radio or TV Both radio and TV Source: Based on Educational Initiatives 2010. THE EXTENT OF INEQUALITY 81 FIGURE 2.11 Growth is reducing poverty in South Asia, as it did in East Asia a. South Asian countries b. Selected East Asian countries 75 75 70 Bangladesh 70 Lao PDR 65 65 Poverty headcount ratio (percent) Poverty headcount ratio (percent) 60 60 55 55 50 50 45 45 40 40 35 India 35 30 Pakistan 30 25 25 Indonesia 20 20 15 Nepal 15 Cambodia 10 10 China Vietnam 5 Sri Lanka Bhutan 5 Philippines Thailand 0 0 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 GDP per capita (PPP, constant 2005 international GDP per capita (PPP, constant 2005 international dollars, thousands) dollars, thousands) Sources: For South Asia, based on HIES 2000, 2005, and 2010 for Bangladesh; BLSS 2003 and 2007 for Bhutan; NSS 1993–94, 2004–05, and 2009–10 for India; NLSS 1995 and 2010 for Nepal; HIES 2001–02, 2004–05, 2007–08, and 2010–11 for Pakistan (Punjab); and HIES 1995, 2002, 2006, and 2009 for Sri Lanka. For East Asia, based on the World Bank’s WDI data- base, http://data.worldbank.org/data-catalog/world-development-indicators and PovcalNet tool, http://iresearch.worldbank.org/PovcalNet/index.htm: 1994–2009 for Cambodia, 1990–2009 for China, 1984–2010 for Indonesia, 1992–2008 for Lao PDR, 1985–2009 for Philippines, 1981–2010 for Thailand, and 1993–2008 for Vietnam. Note: The headcount ratio is based on $1.25 a day (PPP, constant 2005 international dollars). PPP = purchasing power parity. a steep reduction in poverty for the same countries with a higher income per capita are growth in income per capita. Thus, the characterized by higher inequality. decline in poverty is gentler in Thailand, Of course, no mechanical relationship where the initial poverty rate (in 1980) was exists between growth and inequality. A less than 25 percent, than it is in Vietnam, well-known hypothesis in development where the initial poverty rate was close to economics—known as the Kuznets curve— 65 percent (in 1992). Up to 2010, India had is that inequality initially increases as coun- the gentlest rate of poverty decline among tries grow into middle-income levels and the South Asian countries, but preliminary then decreases as they become richer. But analyses based on data for 2011–12 sug- the empirical evidence on this relationship gest that the speed of poverty reduction has is mixed (Milanovic 2011). In addition, accelerated substantially in recent years. the extent of inequality depends on pol- Growth has also led to increasing inequal- icy choices and not just on some economic ity in both regions. The relationship can be fate. In East Asia, for instance, substan- visualized by replacing the poverty rate con- tial heterogeneity exists, with increases in sidered in the previous analysis by the MLD inequality in China and Indonesia that are of per capita consumption. In South Asia, higher than those observed in South Asia. the consumption of a randomly selected The stable inequality displayed by Vietnam individual is between 10 percent and 30 per- and the decline in inequality observed in cent lower than the average consumption Thailand are additional indications that per capita (figure 2.12). That is less than the inequality does not inevitably increase as average inequality observed in East Asian GDP per capita grows. countries when they were at similar levels of Overall, inequality increases about as income per capita. However, in both regions, often as it falls during spells of growth 82 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.12 Richer countries tend to be more unequal in both South Asia and East Asia a. South Asian countries b. Selected East Asian countries 0.45 0.45 0.40 0.40 0.35 0.35 Bhutan China 0.30 0.30 Philippines Nepal MLD index MLD index 0.25 Maldives 0.25 Cambodia Thailand Vietnam 0.20 Sri Lanka 0.20 Indonesia India 0.15 Pakistan 0.15 Lao PDR Bangladesh 0.10 Afghanistan 0.10 0.05 0.05 0.00 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 GDP per capita (PPP, constant 2005 GDP per capita (PPP, constant 2005 international dollars, thousands) international dollars, thousands) Sources: For South Asia, based on NRVA 2005 and 2007 for Afghanistan; HIES 2000, 2005, and 2010 for Bangladesh; BLSS 2003 and 2007 for Bhutan; NSS 1993–94, 2004–05, and 2009–10 for India; HIES 2002–03 and 2009–10, and Vulnerability and Poverty Assessment (VPA) 2004 for Maldives; NLSS 1995 and 2010 for Nepal; HIES 2001–02, 2004–05, 2007–08, and 2010–11 for Pakistan (Punjab); and HIES 1995–95, 2002–03, 2006–07, 2009–10 for Sri Lanka. For East Asia, based on the World Bank’s WDI database, http://data.worldbank.org /data-catalog/world-development-indicators and PovcalNet tool, http://iresearch.worldbank.org/PovcalNet /index.htm: 1994–2009 for Cambodia, 1990–2009 for China, 1984–2010 for Indonesia, 1992–2008 for Lao PDR, 1985–2009 for Philippines, 1981–2010 for Thailand, and 1993–2008 for Vietnam. Note: MLD = mean log deviation; PPP = purchasing power parity. (Ferreira and Ravallion 2008). New data neutral overall with respect to distribution actually reveal less income inequality in the (Dollar, Kleineberg, and Kraay 2013). developing world than 30 years ago because That said, a rigorous statistical analy- of falling inequality between countries. sis of the available microeconomic data Average inequality within developing coun- across countries suggests that grow th tries has been slowly rising, though staying has been more propitious to inequality in fairly flat since 2000. As a rule, higher rates South Asia—and especially in East Asia— of growth in average incomes have not put than elsewhere. Looking at the preceding upward pressure on inequality within coun- analysis by geographical region, one can tries (Ravallion 2014). see that for the combined East and South For the world as a whole, when aver- Asia regions the estimated elasticity of the age consumption per capita increases by consumption per capita of the poorest with 1 percent, the consumption per capita of respect to average consumption per capita the poorest 20 percent of the population is substantially lower than 1.0 percent in increases by 1.057 percent (figure 2.13). The both the 1990s and the 2000s (and sig- corresponding figure for the poorest 40 per- nificantly so in the 1990s) (figure 2.13). cent of the population is 1.004 percent. None The fastest-growing countries, notably of these estimates is significantly different China, have increases in income inequality, from 1.0 percent, meaning that growth is making the income growth of the bottom THE EXTENT OF INEQUALITY 83 FIGURE 2.13 Prosperity has been shared less widely in South and East Asia Increase in the expenditure of the population group for a 1 percent increase in average expenditure a. Bottom 20 percent b. Bottom 40 percent 1.4 1.4 1.2 1.2 1.0 1.0 Percent Percent 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 sia ca ia So ica ar ia t a Cari ia ha frica ha frica ic ic es No ean nt es rth n Am ope e c rld co ld As a s cif cif s fri tin Euro ome Wor ica nd C ntri hA i fr lA e E the ral A e ica nd untr ur com Wo nd ibb bb nA Pa Pa nA A b- rth A As uth ra ut u d d nt o ra ra o an an as e C S No e Ce ia ia Sa Sa nd h As t b- c a a nd ta nd -in -in st st Su Su Am pe Ea Ea as a a gh gh eE Hi Hi er er E dl dl id id M M tin La La Source: Dollar, Kleineberg, and Kraay 2013. Note: Data are from 118 countries for which household surveys are available for at least two years since the 1970s. 40 percent lag behind average income FIGURE 2.14 Monetary inequality is increasing across most of growth. Latin America shows the opposite South Asia tendency in the 2000s, with an estimate greater than 1.0 percent for the bottom 0.80 20 percent. This means that in faster- growing Latin American countries, the 0.60 Annual change in MLD index (percentage points per year) income shares of the bottom quintiles also 0.40 increased more. Consistent with the Kuznets cur ve 0.20 hypothesis, monetary indicators of inequal- 0 ity have increased in the poorest countries in the South Asia region in recent years, –0.20 whereas they have decreased in the richest –0.40 ones (figure 2.14). However, the only two countries for which a decrease is observable –0.60 s an h ka a n l n have a combined population of less than pa ive di es ta sta an ut In Ne kis lad ald ni Bh iL 1 million people, in a region accounting for Pa ha ng M Sr g Ba Af a fourth of humankind. The vast majority of South Asians have experienced an increase in inequality, sometimes at a fast pace. Sources: Based on NRVA 2005 and 2007 for Afghanistan, HIES 2000 and 2010 for Bangladesh, BLSS 2003 and 2007 for Bhutan, NSS 1993–94 and 2009–10 for India, NLSS 1995 and 2010 for Nepal, HIES Although wealth data are available for 2002–03 and 2009–10 for Maldives, HIES 2001–02 and 2010–11 for Pakistan (Punjab), and HIES only India, they suggest that inequality has 1995–95 and 2009–10 for Sri Lanka. Note: MLD = mean log deviation. persisted or slightly widened over the last few decades. The Gini coefficients for asset holdings and net wealth remained roughly 84 ADDRESSING INEQUALITY IN SOUTH ASIA constant between 1991–92 and 2002–03. the Taliban years led to fewer educational An analysis going back to 1961–62 confirms opportunities for children, especially girls, this rough stability (Subramanian and Jayaraj all other countries in the region have expe- 2008). Meanwhile, the ratio of billionaire rienced a decrease in the ratio between the wealth to GDP has increased substantially years of education of the top and bottom in India, suggesting an increasing concen- quintiles for people 15–65 years of age. In tration of wealth at the top. The ratio rose some cases, the decline has been substantial. from around 1 percent in the mid-1990s to Learning assessments have been con- 22 percent at the peak of the economic boom ducted among rural households in India in 2008; the ratio was still about 10 percent and Pakistan since 2005. The findings from of GDP in 2012 (Gandhi and Walton 2012). them suggest a general decline in average learning achievement. For instance, between 2008 and 2011 both the proportion of stu- Mixed trends in nonmonetary dents in grade 3 who could read at grade 1 inequality level and the proportion of students in grade Although monetary indicators of well-being 5 who could read at grade 2 level in rural show that inequality has been on the rise in India fell significantly. In arithmetic, student most of South Asia, the picture that emerges achievement appears to have plunged even from nonmonetary indicators is less clear- further: the percent of third-graders able to cut. Health outcomes show signs of wid- perform subtraction fell from 40 percent to ening inequality (figure 2.15). The ratios 30 percent and of fifth-graders able to per- between the neonatal and under-five child form division from 36 to 28 percent. In rural mortality rates of the population quintiles Pakistan, between 2008 and 2011 at every with the highest and the lowest expendi- grade a small decline occurred in the per- tures per capita have either stayed constant centage of students who can read a story or or increased in most countries. Bangladesh perform division (figure 2.17). Because rural had a reduction in inequality in neonatal households tend to be more disadvantaged, mortality. All other countries for which data the decline in the learning outcomes of their are available at two points in time, on all children can be interpreted as a sign of grow- indicators, show an increase in inequality in ing inequality (Das and Zajonc 2008). health outcomes. The increase is especially In Sri Lanka, however, inequality in learn- marked in the case of stunting. ing outcomes has either remained stable or This increase in inequality is not inconsis- fallen over time. Although academic perfor- tent with absolute progress for all population mance remains lower among children for groups, reflecting cumulative improvements whom Tamil is the language of instruction, in coverage of a range of services in health the gap with the academic performance of as well as improvements in living standards, children learning in Sinhala has narrowed in rising parental education, and decreasing bar- recent years, at least in the case of the first riers to access basic services. But progress gen- language and mathematics. It has remained erally has been faster among the better-off. roughly unchanged in the case of English In contrast, inequality in education out- (figure 2.18). Tamil households are generally comes has unambiguously been on a down- poorer than Sinhala households, so this trend ward trend (figure 2.16). Gaps in educational can be interpreted as evidence of declining attainment among children 6 to 11 years inequality along an important nonmonetary of age either have been stable, mainly in dimension. countries that have achieved close to uni- One explanation of the difference between versal coverage of primary education, or India and Pakistan, on the one hand, and have decreased. The decline in inequality Sri Lanka, on the other, is the rapid expan- is remarkable for the population at large. sion in school coverage that took place in With the exception of Afghanistan, where the former two countries in recent years. THE EXTENT OF INEQUALITY 85 FIGURE 2.15 Inequality in health outcomes has remained stable or increased a. Neonatal mortality 3.5 3.0 Ratio of poorest to richest quintile 2.5 2.0 1.5 1.0 0.5 0 1993 2011 1992 2005 2009 1996 2011 1990 2007 2007 Bangladesh India Maldives Nepal Pakistan Sri Lanka b. Under-five mortality 3.5 3.0 Ratio of poorest to richest quintile 2.5 2.0 1.5 1.0 0.5 0 1993 2011 1992 2005 2009 1996 2011 1990 2007 2007 Bangladesh India Maldives Nepal Pakistan Sri Lanka c. Stunting 4.5 4.0 3.5 Ratio of poorest to richest quintile 3.0 2.5 2.0 1.5 1.0 0.5 0 1993 2011 1992 2005 2009 1996 2011 1990 2007 2007 Bangladesh India Maldives Nepal Pakistan Sri Lanka Initial year Final year Sources: Based on DHSs. Note: Neonatal mortality rate is defined as number of deaths to children younger than 28 weeks per 1,000 live births. Under-five mortality rate is defined as number of deaths to children younger than five years per 1,000 live births. Stunting is defined as percentage of children younger than two years whose z-scores are below minus two standard deviations from the median of the WHO Child Growth Standards. Much of this expansion occurred through corresponding figure in the case of children increased access to education for some of the from scheduled castes and scheduled tribes most disadvantaged groups. For instance, in was 6.6 percentage points. This composition India school enrollment rates for children effect is absent in Sri Lanka’s case, where 6 to 10 years of age rose by 4.7 percent- school enrollment rates have been consis- age points between 2005 and 2010, but the tently high for decades. 86 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.16 Inequality in educational attainment is generally decreasing a. Educational attainment (6 to 11 years of age) b. Educational attainment (15 to 65 years of age) 2.0 2.0 1.8 1.8 Final ratio of richest to poorest quintile Final ratio of richest to poorest quintile Bhutan 1.6 1.6 Pakistan 1.4 1.4 Afghanistan Bangladesh 1.2 Bhutan 1.2 Afghanistan Sri Lanka/Maldives Nepal Nepal India 1.0 1.0 0.8 0.8 India Bangladesh 0.6 0.6 Sri Lanka Pakistan 0.4 0.4 Maldives 0.2 0.2 0 45° 0 45° 0 0.5 1.0 1.5 2.0 2.5 0 0.5 1.0 1.5 2.0 2.5 Initial ratio of richest to poorest quintile Initial ratio of richest to poorest quintile Sources: Based on NRVA 2005 and 2007 for Afghanistan, HIES 2000 and 2010 for Bangladesh, BLSS 2003 and 2007 for Bhutan, NSS 1993–94 and 2009–10 for India, NLSS 1995 and 2010 for Nepal, HIES 2002–03 and 2009–10 for Maldives, HIES 2001–02 and 2010–11 for Pakistan (Punjab), and HIES 1995–96 and 2009–10 for Sri Lanka. Note: Educational attainment is measured in years of schooling. What lies behind inequality? of total inequality in India in 2009–10, up from less than 13 percent in 1993–94 A cursory analysis of the data available (these figures are computed based on the on monetary indicators of inequality in MLD). A similar upward trend is found in South Asian countries reveals some strong— Bangladesh, Bhutan, Nepal, and Pakistan. and somewhat predictable—patterns. Some Only in Sri Lanka did the contribution of monetary indicators of well-being, includ- education to total inequality somewhat ing the MLD and the Theil index, allow decline between 1995–96 and 2009–10 decomposing total inequality as the sum of (figure 2.19). Precise estimates depend on the inequality between population groups and assumptions made to compute the decompo- inequality within population groups. The sition, but the conclusion that educational former is computed as the extent of inequal- attainment accounts for a growing share of ity that would prevail if everyone in each of inequality in South Asia seems robust to the the groups had the average income or con- methodology used. sumption per capita of the group as a whole. Another obvious correlate on inequal- Population groups are typically defined ity is location. Except in Sri Lanka, where along individual characteristics that are pre- differences between urban and rural areas determined or diffi cult to modify, such as are becoming less relevant over time, else- gender, ethnicity, or location. The estimated where in the region the urban-rural divide inequality between groups provides insights accounts for a substantially larger share of on the contribution the selected individual overall inequality now than it did barely 10 characteristics make to total inequality. or 15 years ago. In India, when households A standard partition of the population are partitioned depending on whether they is on the basis of educational attainment. live in cities or not, differences between For instance, households can be classi- groups account for nearly 20 percent of total fied depending on whether the household inequality in 2009–10, more than three times head has at least some secondary educa- the ratio observed in 1993–94. The ratio tion or not. With this partition, inequality reached 17 percent in Nepal in 2010, from between groups accounted for 26 percent less than 5 percent in 1995 (figure 2.20). THE EXTENT OF INEQUALITY 87 FIGURE 2.17 Learning outcomes have deteriorated in rural India and rural Pakistan a. Academic competency, rural India 60 50 Share of students (percent) 40 30 20 10 0 Can read Can subtract Can read Can divide grade 1 text grade 2 text Third graders Fifth graders b. Academic competency, rural Pakistan 90 80 Share of students (percent) 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 Can read a story Can divide Grade1 to grade 8 students Grade 1 to grade 8 students 2008 2011 Sources: Annual Status of Education Report 2009, 2012; South Asian Forum for Education Development 2009, 2012. Whereas education and place of residence in 2009–10. This finding is somewhat mis- are easily observable correlates of inequality, leading, however. Caste accounts for a sub- other less visible but not less powerful forms stantially larger share of overall inequality in of social exclusion may also be at play. In some parts of India than in others. Indeed, India’s case, a relevant partition of the pop- caste turns out to be an important contribu- ulation is between households whose head tor to total inequality in most northern states belongs to a Scheduled Caste or a Scheduled (figure 2.21). On a more positive note, the Tribe and other households. Contrary to contribution of caste to overall inequality expectations, caste explains a relatively small declined in Andhra Pradesh and in Bihar, part of between-group inequality in con- two major states that together account for 15 sumption per capita—just around 7 percent percent of India’s population. 88 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.18 The gap in learning outcomes between ethnic groups is declining in Sri Lanka 80 Average test scores (in points) 70 60 50 40 30 Sinhala Tamil Sinhala Tamil Sinhala Tamil medium of medium of medium of medium of medium of medium of instruction instruction instruction instruction instruction instruction First language English Mathematics 2003 2007 2009 Source: National Research and Evaluation Centre (NEREC) 2009. FIGURE 2.19 Education explains a growing share of overall inequality 30 Contribution of inequality between groups to 25 total inequality (percent) 20 15 10 5 0 2000 2010 2003 2007 1993 2009 1995 2010 2001 2010 1995 2010 Bangladesh Bhutan India Nepal Pakistan Sri Lanka Initial year Final year Sources: Based on HIES 2000 and 2010 for Bangladesh, BLSS 2003 and 2007 for Bhutan, NSS 1993–94 and 2009–10 for India, NLSS 1995 and 2010 for Nepal, HIES 2001–02 and 2010–11 for Pakistan (Punjab), and HIES 1995–96 and 2009–10 for Sri Lanka. THE EXTENT OF INEQUALITY 89 FIGURE 2.20 The rural-urban divide is becoming a more important source of inequality 20 Contribution of inequality between groups to total 15 inequality (percent) 10 5 0 2000 2010 1993 2009 1995 2010 2001 2010 1995 2010 Bangladesh India Nepal Pakistan Sri Lanka Initial year Final year Sources: Based on HIES 2000 and 2010 for Bangladesh, NSS 1993–94 and 2009–10 for India, NLSS 1995 and 2010 for Nepal, HIES 2001–02 and 2010–11 for Pakistan (Punjab), and HIES 1995–96 and 2009–10 for Sri Lanka. The preceding decompositions describe Gallup World Poll asks respondents about the contribution different individual charac- their satisfaction with basic services, their teristics make to total inequality. But they do assessment of future well-being, and their not explain why those characteristics matter, views on government efforts to help the poor. and in particular they are totally silent on The six South Asian countries covered by the the interaction between those characteristics Gallup World Poll are compared to Brazil, and a social and economic environment that which is chosen as the benchmark because is shaped by public policies. Based on the it has experienced a substantial reduction decompositions, households from lower castes in inequality in recent years. Moreover, this are clearly at a disadvantage, but that could reduction is generally attributed to policy well be because they have fewer opportuni- changes leading to declining returns to edu- ties to access education and health services cation, pronounced rural-urban convergence, when young. Similarly, education emerges as increases in social assistance transfers tar- a growing force behind inequality, but that geted to the poor, and a possible decline in could be the result of very different jobs being racial inequality (Paes de Barros and others available to those with and without skills. Job 2010; Ferreira, Leite, and Litchfield 2008; opportunities may also differ substantially Lustig, Lopez-Calva, and Ortiz-Juarez 2012; between rural and urban areas, and exposure and de Souza 2012). to major shocks such as droughts and natu- I n Brazil, according to the Gallup ral disasters might be different as well. Public World Poll, satisfaction with access to pub- policies can, in turn, result in better access to lic services is higher among poorer popula- basic services, greater mobility through jobs, tion groups, and a negative correlation exists or improved social protection. between income per capita and expectations In South Asia, the contribution public poli- of a better life. Poorer population groups cies make is often seen in a negative light. The are also more satisfied with government 90 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.21 Caste is an important correlate of inequality in some Indian states Orissa Kerala Karnataka Goa Pondicherry Tamil Nadu Southern Dadra and Nagar Haveli states Gujarat Daman & Diu Chhattisgarh Andhra Pradesh Maharastra Andaman and Nicobar Chandigarh Haryana Punjab West Bengal Delhi Uttarakhand Rajasthan Madhya Pradesh Northern states Jharkhand Uttar Pradesh Himachal Pradesh Bihar Sikkim Tripura Assam Manipur 0 5 10 15 20 25 30 35 Contribution of inequality between groups to total inequality (percent) 1993 2009 Sources: Based on India NSS 1993–94 and 2009–10. efforts to help them. Taken altogether, these or amplify the contribution from inherited responses reveal a positive view of equality of circumstances. opportunity, upward mobility, and targeted A simple conceptualization of how support. All of this is at odds with responses these different variables come into play to the same questions in South Asian coun- involves a person’s life cycle (figure 2.23). tries (figure 2.22). Circumstances at birth, such as gender and The role played by household character- caste, shape the options available to individu- istics such as education, location, and caste als, especially in relation to the accumulation provides some clues about what lies behind of human capital. As people age and enter the inequality. Opinion polls, in turn, are infor- labor force, their job opportunities and the mative on how public policies may offset possibility of reaping benefits from migration THE EXTENT OF INEQUALITY 91 FIGURE 2.22 South Asians do not see an environment conducive to lower inequality a. Access to public services 100 Health care 50 Satisfaction with access to public services (percent) 0 100 Education 50 0 100 Water 50 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil b. Future well-being 2.5 Views on future well-being relative to present 2 1.5 1 0.5 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil Bottom 40% Middle 40% Top 20% (continues next page) 92 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 2.22 South Asians do not see an environment conducive to lower inequality (continued) c. Efforts to help the poor 60 Satisfaction with efforts to help the poor (percent) 50 40 30 20 10 0 Afghanistan Bangladesh India Nepal Pakistan Sri Lanka Brazil Bottom 40% Middle 40% Top 20% Source: Based on Gallup World Poll. Note: Population groups are defined based on income or consumption per capita. Views on present (future) well-being are assessed on a scale from 1 (dissatisfied) to 10 (satisfied). FIGURE 2.23 Multiple factors affect household outcomes relative to others in society Drivers of inequality MOBILITY Jobs Cities OPPORTUNITY Health Education SUPPORT CIRCUMSTANCES Social protection Gender Taxes and transfers Caste At birth During youth Throughout life Life cycle THE EXTENT OF INEQUALITY 93 affect earnings prospects. Throughout life, including health, education, and infrastruc- people also experience shocks and are influ- ture; mobility throughout adult life can be enced by innate differences. Through each of enhanced by economic growth and rapid these phases, public policies may affect the urbanization; and support as government extent of inequality. transferring resources to people or tax- In sum, opportunity in childhood can ing them can mitigate shocks and offset be shaped by access to basic services, disadvantages. The extent of inequality: Main messages and policy implications Standard monetary indicators underestimate the Trends in inequality are also different true extent of inequality in South Asia. Taken at depending on the indicator considered. Monetary face value, the Gini index, the ratio between the inequality is unambiguously increasing, except expenditures of the top and bottom 10 percent in the richest countries in the region. Overall, of the population, and the share of the bottom the bottom 20 and 40 percent of the population 40 percent, all point out in the direction of mod- have seen their expenditures grow more slowly erate inequality. The numbers suggest that coun- than the average in South Asia. Although consid- tries in South Asia are much less unequal than erable debate exists about the validity of the so- China, Mexico, or South Africa. called Kuznets curve, trends in the region seem Nonmonetary indicators yield a more nuanced to conform to it, suggesting that pressures for picture. Inequalities in human development out- monetary inequality will increase further in the comes such as infant mortality, under-five child coming years. Trends in nonmonetary inequal- mortality, or educational attainment are wide ity are less clear. South Asia is becoming even and make the extent of inequality in most coun- more unequal in relation to health outcomes. tries of South Asia look similar to that of other In contrast, the expansion of school coverage is developing countries. Several countries in the reducing inequalities in educational attainment, region are among the most unequal for which although it may at the same time be widening comparable data are available. disparities in learning outcomes. But the combination of monetary and non- Important insights are possible into the drivers monetary indicators is most telling. Based on of inequality in the region. Simple decompositions data on asset holdings, the wealth in the hands of total inequality between population groups sug- of India’s billionaires is disproportionately high gest that characteristics such as educational attain- compared with that of other countries at similar ment or location matter a lot. Both education gaps income levels. Meanwhile, on human develop- and the rural-urban divide account for a growing ment indicators such as malnutrition, the poorest share of total inequality. The share is smaller in South Asians fare worse than most anyone else the case of caste, but it remains relevant in north- in the world. Seen this way, South Asia remains ern and eastern Indian states. However, the impact a land of extremes. of these individual characteristics on inequality is The diversity of assessments, depending on mediated by economic structures and public poli- which indicator is used, is somewhat disturbing. cies. Opinion polls suggest that South Asians have It implies that a single metric to assess progress negative views on access to basic services, pros- toward shared prosperity, such as the share of pects for improvement through life, and help for the bottom 40 percent, could be seriously mis- the poor. Beyond subjective assessments, any thor- leading. But at another level this diversity should ough assessment of the dynamics of inequality in be welcome because it provides useful insights on the region, and the policies to address it, needs to the drivers of inequality in South Asia. focus on opportunity, mobility, and support. 94 ADDRESSING INEQUALITY IN SOUTH ASIA References of B ra z i l ia n I nequ a l it y: 1981–20 0 4.” Macroeconomic Dynamics 12 (Suppl. S2): Asadullah, M., N. Chaudhury, D. Prajuli, L. R. 199–230. Sarr, Y. Savchenko, and S. R. Al-Zayed. 2009. Forbes. 2014. “The World’s Billionaires.” Secondary Education Quality and Access http://w w w.forbes.com / billionaires/ list. Enhancement Project (SEQAEP): Baseline Accessed May 29, 2014. Report. Washington, DC: World Bank. Gandhi, Aditi, and Michael Walton. 2012. Aturupane, Harsha, Paul Glewwe, and Suzanne “Where Do India’s Billionaires Get Their Wisniewski. 2013. “The Impact of School Wealth?” Economic & Political Weekly, Quality, Socio-Economic Factors and Child October 6, 10–14. Health on Students’ Academic Performance: Lustig, Nora, Luis F. Lopez-Calva, and Eduardo Evidence from Sri Lankan Primary Schools.” Ortiz-Juarez. 2012. “Declining Inequality Education Economics 21 (1): 2–37. in Latin America in the 2000s: The Cases of Banerjee, Abhijit, and Thomas Piketty. 2005. Argentina, Brazil, and Mexico.” Department “Top Indian Incomes, 1922–2000.” World of Economics Working Paper 1218, Tulane Bank Economic Review 19 (1): 1–20. University, New Orleans, LA. Das, Jishnu, and Tristan Zajonc. 2010. “India Milanovic, Branko. 2011. The Haves and Shining and Bharat Drowning: Comparing two the H ave - Not s: A Br i ef Idios y n cratic Indian States to the Worldwide Distribution History of Global Inequality. New York: Basic in Mathematics Achievement.” Journal of Books. Development Economics. Ravallion, Martin. 2014. “Income Inequality in de Souza, Pedro H. G. 2012. “Poverty, Inequality the Developing World.” Science 344 (6168): and Social Policies in Brazil, 1995–2009.” 851–55. Working Paper 87, International Policy Ricardo, Paes de Barros, Mirela de Carvalho, Centre for Inclusive Growth, United Nations Samuel Franco, and Rosane Mendonça. 2010. Development Programme, Brasilia, Brazil. “Markets, the State, and the Dynamics of Dollar, David, Tatjana Kleineberg, and Aart Inequality in Brazil.” In Declining Inequality Kraay. 2013. “Growth Still Is Good for the in Latin America: A Decade of Progress , Poor.” Policy Research Working Paper 6568, edited by Luis F. López-Calva and Nora World Bank, Washington, DC. C. Lustig, 134 –74. Baltimore: Brookings E du c a t i o n a l I n i t i a t i v e s . 2 010 . S t u d e n t Institution Press. Learning Study: Status of Learning across Sen, Amartya. 1980. “Equality of What?” In 18 S ites of In di a in Urban an d R ural The Tanner Lectures on Human Values , Schools . Ahmedabad, India: Educational Vol. 1, edited by S. McMurrin, 353 – 69. I n it i at ive s . ht t p : // w w w . e i - i n d i a . c o m / Cambridge: Cambridge University Press. w p - content /uploads / 2012 / 01 / M a i n _ ———. 1992. Inequality Reexamined. New York: Report_ StudentLearningStudy_ 2010_by_ Oxford University Press. Educational_Initiatives1.pdf. Subramanian, S., and D. Jayaraj. 2008. “The Ferreira, Francisco H. G., and Martin Ravallion. Distribution of Household Wealth in India.” 2008. “Global Poverty and Inequality: A In Personal Wealth from a Global Perspective, Review of the Evidence.” Policy Research edited by James B. Davies, 112–33. New York: Wo r k i n g P a p e r 4 6 2 3 , Wo r l d B a n k , Oxford University Press. Washington, DC. World Bank. 2013. Student Learning in South Ferrei ra, Francisco, Ph illippe L eite, and Asia: Challenges, Opportunities, and Policy Julie Litchfield. 2008. “The Rise and Fall Priorities. Washington, DC: World Bank. Limited Opportunity 3 E quality of opportunity is considered a Declaration of Human Rights, the United key condition for a society to ensure Nations Convention on the Rights of the distributional justice. Those who Child, and the Millennium Development have the same talent and ability and have the Goals (MDGs) all reflect this consensus. The same willingness to use them should have idea of opportunities as equal access to basic the same prospects of success regardless of services by children is also consistent with the their initial circumstances (Rawls 1971). “rights approach” upheld by several countries Important outcomes—such as income or in South Asia. health status—are seen as determined by General agreement exists that the set of two main factors: efforts and circumstances. goods and services that every individual under The latter includes inherited characteristics 16 years of age should have access to includes such as gender or ethnicity as well as luck nutrition, health care, basic education, and (Arneson 1989; Cohen 1989; Dworkin some forms of infrastructure. Within health 1981a, 1981b; Roemer 1998). Equality of and nutrition, the focus is often on institu- opportunity requires compensating people tional births and full immunization. In basic for disadvantages related to circumstances education, opportunity is associated with pri- so the distributions of outcomes can be mary school attendance and completion and entirely attributed to efforts (Roemer 1998). in some cases with secondary school atten- Both conceptually and empirically, dance. The forms of infrastructure deemed completely distinguishing efforts from essential for opportunity include clean water, circumstances is difficult, hence imped- improved sanitation, and electricity. ing disentanglement of opportunities from Reaching universal coverage of basic outcomes (Kanbur 2009). However, broad services is associated with equality of oppor- consensus exists that making access to basic tunity, but increases in coverage along the way services universal is at the core of equality may amplify inequalities (box 3.1). In devel- of opportunity (Paes de Barros and others oping countries, and especially in the poorest 2009; World Bank 2005). Access to basic ones, coverage is far from universal, and services improves the likelihood of a child households and individuals from more privi- maximizing her human potential and adding leged backgrounds usually gain access first. value to the world around her. The Universal Whereas MDG targets have given impetus 95 96 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 3.1 In demographic transitions, inequality of opportunity increases inequality of outcomes A simple diagrammatical representation can fi nal well-being in the vertical axis. Inequality illustrate how inequality of opportunity shapes in outcomes is captured by the area between the dynamics of inequality of outcomes. Consider households along each of these two axes. the well-being of individuals at two points in life: If inequality in outcomes were persistent, in childhood (initial well-being) and in adult- households would be aligned on a 45-degree line: hood (final well-being). In the figure, initial well- differences in well-being would be as large in being is represented in the horizontal axis and adulthood as they were in childhood. However, if richer households have better access to services than poorer households do, they are also in a Final better position to live in good health and accu- well-being mulate skills. Throughout life, a higher human capital amplifies the differences stemming from ET Impact on initial outcomes, and households end up on a inequality convex curve located above the 45-degree line. The figure refers to the same households at two points in their lives, but inequality in out- Health, comes is measured for the entire population. It skills is thus some weighted average of the inequal- eT ity captured in the horizontal axis and in the Initial vertical axis, with weights given by the size of capabilities the corresponding population groups. In coun- 45° tries undergoing a demographic transition, the e0 E0 Initial weight of the adult group is increasing steadily. Poorer Richer well-being In their case, therefore, inequality of opportu- household household nity leads to a growing inequality in outcomes. to campaigns to increase coverage, equity in electricity is also limited in most countries, access has received less attention. When cov- while countries with a relatively high cover- erage is low, some households are excluded, age face the challenge of improving the qual- particularly those lacking the abilities or ity of access. resources to utilize these services fully. Several population groups receive sys- In South Asia, access to health and edu- tematically lower coverage of basic services cation services has improved in recent times. because of their circumstances. Not sur- However, the lack of access for children prisingly, rural areas fare worse than urban remains acute in relation to health and nutri- areas. Gaps in coverage are also pronounced tion. Access is better in the case of primary when comparing children whose mothers or education but remains low in secondary edu- parents have different levels of educational cation. The picture becomes more diversified attainment. Differences in coverage between in the case of infrastructure services: access girls and boys are small, however. Overall, to improved sanitation is dismally low in measurable circumstances of children— most South Asian countries, with the excep- including their place of residence, their gen- tion of Maldives and Sri Lanka; access to der, their religion (or ethnicity), or their LIMITED OPPORTUNITY 97 parents’ education—explain a large fraction extremely high, to 100 percent, when cover- of the observed inequality in access to health, age is universal (box 3.2). This index can thus education, and infrastructure services. be used to assess inequality of opportunities Several factors explain South Asia’s lack- in access to basic services in South Asia. luster performance in ensuring equality in An important question concerns the fac- access to basic services. Importantly, public tors that define children’s background or spending on education and health is relatively circumstances in a particular country or set low in comparison with other countries at a of countries. Following the literature, this similar level of development. Equity requires report uses factors that are more likely to be committing additional resources to disadvan- predetermined, hence unaffected by a child’s taged groups to offset their otherwise more own actions. These factors are the place of limited access to basic services. However, residency, either urban or rural; a child’s the opposite is often observed in South Asia. gender, religion, and caste; and the educa- In health, public spending is directed more tion level of the household head or the child’s toward the better-off than to the poor- mother, depending on data availability. est population groups. In education, public Religion is used only when data are available, spending tends to be progressive at lower and caste is used only in India’s case. grade levels but regressive at secondary and The idea of opportunities as equal access especially tertiary levels. to basic services by children is consistent with rights-based approaches to service delivery. Most countries in South Asia have Inequality in access to basic explicitly recognized the right of children services to education. Sri Lanka was a precursor, The coverage of basic services, measured as when it did so in 1945—three years before a percentage of the relevant population, is independence. Bangladesh passed the a fi rst and telling indicator of access. When Compulsory Primary Education Act, which the coverage of a specific service is univer- made primary education both free and com- sal, everybody has access to it; lower cov- pulsory, in 1993. Nepal enshrined in the erage rates mean that some are necessarily Constitution of 1990, and further in the excluded. Exclusion is typically not random, Seventh Amendment of the Education Act however. People from disadvantaged back- of 2001, the right of every child to free pri- grounds are less likely to get access to mary education. India’s Right to Education services. The extent to which access var- Act, passed in 2009, recognized access to ies across clearly distinct groups—such primary education as a constitutional right. as women or ethnic minorities or lower- The following year, Pakistan did likewise caste groups—is significant in this respect. with the passage of the 18th amendment to The greater the dispersion of access across the constitution, under article 25A. groups, the greater the degree of inequality More recently, four South Asian countries for the same level of coverage. joined the Scaling Up Nutrition movement, This simple intuition is captured by which had been launched in 2010 and renewed a synthetic indicator, called the Human in 2012 with updated goals. These four South Opportunity Index (HOI). The HOI is com- Asian countries are Bangladesh, Nepal, puted by multiplying the coverage rate by a Pakistan, and Sri Lanka. Complementing the measure of the dispersion of access across MDGs, Scaling Up Nutrition aims at saving the relevant groups. When people from all and improving lives through greater availabil- groups have equal access on average, the HOI ity of nutritious food. is equal to the coverage rate. As the disper- Despite this widespread commitment to sion of access increases, the HOI declines. By rights, access to services related to health construction, the HOI varies between zero, and nutrition tends to be limited in South when coverage is nil or when the dispersion is Asia (figure 3.1). The value of the HOI is 98 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 3.2 The Human Opportunity Index The HOI measures the availability of services 50 percent of all children go to school. From the that are necessary to progress in life, adjusted perspective of overall coverage, both countries by how unequally the services are distributed look alike. Now suppose that in country A, no among different groups in the population. Two girl attends school, but in country B, 50 per- countries that have identical coverage of nutri- cent of both girls and boys attend school. The tion services for infants, for instance, may have a HOI discounts the coverage rate of 50 percent different HOI if the infants that lack this service in country A because access is more unequal. systematically share a personal circumstance For country B, there is no discounting because beyond their control, such as gender, caste, there is no gender inequality, making the HOI parental income, or place of birth. Put simply, 50 percent, or equal to the coverage. Because the HOI is coverage corrected for equity. country B has a higher HOI, it is more equal than The calculation of the HOI focuses on the dis- country A, even though the enrollment rate is the similarity index (D), originally a demographic same in both countries. measure of evenness widely used in the analy- The HOI has practical appeal because it sis of social mobility and typically applied to allows summarizing the evolution of the inequal- dichotomous outcomes. The D-index is defined ity of opportunity over time without having to as the weighted average of absolute differences track the different circumstances one by one. In of group-specific access rates (pi) from the over- addition, it has technical properties that make all average access rate (p): it an attractive measure of inequality. First, it is sensitive to scale—if access improves for 1 n all groups by, say, a factor of K (additively or D= ∑ i =1 βi pi − p . 2p multiplicatively), then the HOI changes by the same factor, K. Second, the HOI rewards Pareto By construction the D-index varies between improvement: if the coverage rate improves for zero and one. A value of zero indicates that one circumstance group, without decreasing cov- access rates for all groups considered are the erage rates for the remaining groups, the HOI same, while positive values indicate that certain rises. Third, the HOI always improves if access groups of individuals have a lower probability changes in a way that favors groups with cover- of access to the service considered. In practical age rates below the average. terms, the D-index reflects the percentage of the However, the HOI provides a lower bound coverage rate of a particular opportunity that on the inequality prevalent in a given place. Its has to be discounted to obtain the HOI, that is, calculation can include only those circumstances that are measurable and for which data exist. If HOI = p (1 − D). more circumstances were added to the set con- sidered in the analysis, the HOI would increase. The HOI can be increased by providing more Having a lower bound measure complicates services to all (“scale effect”) or by distributing comparison between countries or geographical services more fairly (“equalization effect”). regions, especially if the purpose is to investi- Consider two countries, A and B, and con- gate which country or region is more inequitable sider a basic opportunity such as access to pri- overall and not which one has a lower minimum mary education. Suppose that in both countries, level of inequality. Sources: Paes de Barros and others 2009; Molinas and others 2010. LIMITED OPPORTUNITY 99 FIGURE 3.1 The Human Opportunity Index for basic health services is low in most of South Asia a. Institutional births 100 90 80 70 HOI (percent) 60 50 40 30 20 10 0 h n l a s ka pa ive di es ta an In Ne kis lad ald iL Pa ng M Sr Ba b. Full immunization 100 90 80 70 HOI (percent) 60 50 40 30 20 10 0 n ria a l da h ru ia m a ya s a s ka pa ne ive di bi an ta es s na Pe n an an ge ne In m Ne kis ad pi Ke Gh ald et iL Ug Ni lo do ilip Pa gl Vi M Co Sr In n Ph Ba Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, DHS 2007 for Sri Lanka, and data from World Bank Visualize Inequality dashboard (http://www1.worldbank.org/poverty/visualizeinequality/) for other countries. Note: The HOI for full immunization is computed based on formal records instead of patient recall. HOI estimates differ from HOI reported by the World Bank Visualize Inequality dashboard because a different set of circumstances is used. In this report, circumstances include only predetermined factors, namely the gender of the child, parental education, geographic location (urban or rural), caste, and religion. 100 ADDRESSING INEQUALITY IN SOUTH ASIA worryingly low in the case of full immuni- the household level. Hence, the analysis of zation against vaccine-preventable diseases inequality of opportunity in access to infra- among children two years of age and structure services can be conducted with younger. The HOI of most countries in the some level of confidence at the country level region does not cross the 50 percent mark. but not across countries. In Bangladesh, it stands at approximately Still worth noting is that by construction 50 percent; in Nepal, it is 30 percent. India the HOI cannot exceed the coverage rate. and Pakistan perform poorly with an HOI And even when considering the more gener- below 20 percent on the most recent year ous access measure, at the community level, for which data are available. The extent of coverage is low in most South Asian countries inequality in access in these two countries (figure 3.3). has consistently remained the highest in The smaller countries in the region per- the region and has only decreased slightly form well in the case of access to electricity. over time. Maldives has virtually universal access, and Access to primary education is far better. in Sri Lanka the HOI exceeds 80 percent. In Countries in the region have generally done contrast, Afghanistan still lags far behind, well in primary school attendance and even with an HOI of about 10 percent. Access on completion (figure 3.2). Bhutan and India to electricity is also limited in Bangladesh report HOIs for primary school completion and India, where the HOI hovers between between 80 percent and 90 percent, while 40 percent and 60 percent. Access to sani- Bangladesh, Maldives, and Sri Lanka are tation is generally dismal. With the excep- above 90 percent. Nepal and Pakistan have tion of Maldives and Sri Lanka, the HOI for also done well for primary school atten- improved sanitation services does not exceed dance though less so for completion. In this, 40 percent of the population in South Asian South Asia resembles other regions, reflecting countries. the global drive toward universal enrollment in primary education. The picture is less encouraging for access Coverage is improving, equity to secondary school, especially in compari- less so son with countries at a similar level of devel- Opportunities in access to health and educa- opment. The HOI associated with secondary tion services have been improving in most school completion is below 50 percent across countries over the past decade. Whereas the region, with the exception of Bhutan. most countries have registered HOI increases In Afghanistan, it is less than 5 percent. in access to health services, progress has Even in the best-performing countries of the been slower than for other basic services region—Bhutan, Maldives, and Sri Lanka— (figure 3.4). When considering full immuni- the HOI is smaller than that in traditionally zation, opportunities increased most rapidly inequitable countries, such as Brazil and in Bangladesh, as measured by the annual South Africa. change in the HOI. By contrast, they declined International comparisons are less reli- slightly in Pakistan. As for institutional able with respect to infrastructure services. births, Nepal registered the fastest improve- What it means to have access to improved ment. But overall progress has been slow, water or to improved sanitation varies from and inequality has declined only slightly one country to another. Moreover, access is in recent times. The annual change in HOI measured at the community level in South stands between 0.5 and 1 percentage point in Asian countries but at the household level most cases. in others. If a power line arrives to a village, Countries in South Asia have made signifi- but only half the population in the village has cant strides in improving access to primary electricity, coverage is twice as high when education. Maldives and Sri Lanka led the measured at the community level rather than region in achieving almost universal primary LIMITED OPPORTUNITY 101 FIGURE 3.2 The Human Opportunity Index for education is low in Afghanistan and Pakistan a. Primary school attendance (6–11 years of age) 100 90 80 70 HOI (percent) 60 50 40 30 20 10 0 an ria a da n an a a h sia l m ru s a ka pa ive an di ny bi ta es na Pe an ist ut an ge ne In m Ne kis lad Ke Gh ald et Bh iL an Ug Ni lo do Pa Vi ng M Co Sr gh In Ba Af b. Primary school completion 100 90 80 70 HOI (percent) 60 50 40 30 20 10 0 ha a n a ria n a Ph pal s Vi a m ru a ng n h sia ka s ne ive nd an ny bi di sta ta a es na Pe an ut ge ne In m Ne kis pi lad Ke Gh a ald ni et Bh iL Ug Ni lo do ilip Pa M Co Sr In g Ba Af Sources: Based on National Risk and Vulnerability Assessment 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2010 for Bangladesh, Bhutan Living Standards Survey 2007 for Bhutan, National Sample Survey 2009 for India, HIES 2009 for Maldives, Nepal Living Standards Survey 2010 for Nepal, HIES 2010 for Pakistan, HIES 2009 for Sri Lanka, and data from World Bank Visualize Inequality dashboard (http://www1.worldbank.org/poverty /visualizeinequality/) for other countries. Note: The HOI is computed based on gross completion rates and not on students completing primary school at the appropriate age. HOI estimates differ from HOI reported by the World Bank Visualize Inequality dashboard because a different set of circumstances is used. In this report, circumstances include only predetermined factors, namely the gender of the child, parental education, geographic location (urban or rural), caste, and religion. 102 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.3 The Human Opportunity Index for sanitation is especially low a. Electricity b. Improved sanitation Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Sri Lanka Sri Lanka 0 20 40 60 80 100 0 20 40 60 80 100 HOI (percent) HOI (percent) Source: Based on Andres and others 2013 for this report. Note: Definition of what it means to have access to improved sanitation varies across countries. education coverage. The HOIs associated HOI can be attributed to changes in cover- with both attending and completing primary age, with inequality across groups remaining school were greater than 95 percent in these stable. The bolstering of vaccination pro- two countries already in the mid-2000s, grams through initiatives such as National and they are by now only 2 percentage Immunization Days has contributed to this points short of the 100 percent mark, which increase in coverage. For institutional births, explains their slower rate of change. Progress the main driver of the increase in the HOI has in Bangladesh, India, Bhutan, Nepal, and also been greater coverage, although there Pakistan has also been significant, although has also been an increase in equity across all at varying speeds. In Bhutan and Nepal, the countries covered in the analysis. improvements in primary school atten- The decomposition of changes in the HOI dance and completion were remarkable. yields a more diverse picture in the case of pri- In Pakistan, school attendance saw more mary education (figure 3.6). For instance, in rapid progress than school completion. In Afghanistan, which showed rapid increases Bangladesh, the pace of change is slower, in its HOI for primary school attendance, the but the starting point was higher (with HOIs change has come mainly from greater cover- around 85 percent in 2005). In Afghanistan, age. In contrast, changes in equity in access school attendance grew between 2005 and play an important role in Bhutan, India, 2007, but the HOI remains low. Nepal, Maldives, and Pakistan. As a country The drivers of inequality in access in that has recorded notable growth in primary South Asia can be better understood by school attendance and completion, Nepal is decomposing the change in the HOI between a case in point. The overall increase in HOI increases in coverage rates (scale effect) and in Nepal is much more pronounced than reductions in the dispersion of coverage rates the scale effect would imply, because of an across groups (equity effect). The decompo- important reduction in inequality in access. sition shows that greater coverage clearly This equity effect is large enough to account drives the improvements of opportunities in for approximately one-third of the annual health (figure 3.5). In the case of full immu- increase in the HOI for both attendance and nization, almost all of the change in the completion. LIMITED OPPORTUNITY 103 FIGURE 3.4 Opportunities have improved faster in primary education than in health services a. Health services b. Primary education 1.6 6 5 1.2 Annual change in HOI (percentage points) Annual change in HOI (percentage points) 4 0.8 3 0.4 2 0.0 1 –0.4 0 an h an a s l n ka pa n ive di es ta a h l ta an ist ut pa di In Ne es kis lad ald kis Bh iL an In Ne lad Pa ng Pa M Sr gh ng Ba Af Ba Institutional births Full immunization Attendance Completion Sources: For health, based on Demographic and Health Survey (DHS) 1993 and 2011 for Bangladesh, DHS 1992 and 2005 for India, DHS 1996 and 2011 for Nepal, and DHS 1990 and 2007 for Pakistan; for education, based on National Risk and Vulnerability Assessment 2005 and 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2005 and 2010 for Bangladesh, Bhutan Living Standards Survey 2003 and 2007 for Bhutan, National Sample Survey 1993 and 2009 for India, HIES 2003 and 2009 for Maldives, Nepal Living Standards Survey 2003 and 2010 for Nepal, HIES 2001 and 2010 for Pakistan (Punjab), and HIES 2006 and 2009 for Sri Lanka. FIGURE 3.5 Better opportunities in health are driven by greater coverage of basic services a. Institutional births b. Full immunization Bangladesh Bangladesh India India Nepal Nepal Pakistan Pakistan –2 0 2 4 6 8 10 12 14 16 –2 0 2 4 6 8 10 12 14 16 Annual change in HOI (percent) Annual change in HOI (percent) Sources: Based on Demographic and Health Survey (DHS) 1993 and 2011 for Bangladesh, DHS 1992 and 2005 for India, DHS 1996 and 2011 for Nepal, and DHS 1990 and 2007 for Pakistan. Note: Horizontal bars indicate the total change in the HOI; vertical lines indicate the change caused by increased coverage; the difference between the two reflects changes in equity. Although the measured access to health education services is likely to vary consider- and education services in the region has gen- ably across population groups, and the qual- erally improved, the magnitude of the change ity of services has important implications for may overestimate the improvement in equity. later-life opportunities. For instance, whereas This is because the quality of health and the improvement in the HOI for primary 104 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.6 Better opportunities in education reflect greater coverage and higher equity in some countries a. Primary school attendance b. Primary school completion (ages 6–11 years) (14–18 years of age) Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Annual change in HOI (percent) Annual change in HOI (percent) c. Secondary school attendance d. Secondary school completion (ages 12–18 years) (ages 22–26 years) Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka –20 –10 0 10 20 30 40 50 60 70 80 90 –20 –10 0 10 20 30 40 50 60 70 80 90 Annual change in HOI (percent) Annual change in HOI (percent) Sources: Based on National Risk and Vulnerability Assessment 2005 and 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2005 and 2010 for Bangladesh, Bhutan Living Standards Survey 2003 and 2007 for Bhutan, National Sample Survey 1993 and 2009 for India, HIES 2003 and 2009 for Maldives, Nepal Living Standards Survey 2003 and 2010 for Nepal, HIES 2001 and 2010 for Pakistan (Punjab), and HIES 2006 and 2009 for Sri Lanka. Note: Horizontal bars indicate the total change in the HOI; vertical lines indicate the change caused by increased coverage; the difference between the two reflects changes in equity. school attendance in India may reflect a HOI is unlikely to capture the fact that quali- narrowing of enrollment gaps between dif- tative differences in services may not have ferent castes, in reality the type of schools diminished substantively. high-caste and low-caste students are likely Another important caveat refers to differ- to attend could be quite different in terms of ences within countries. These differences can resources and learning. An improvement in be wide in the case of basic health services. LIMITED OPPORTUNITY 105 For example, the HOI associated with insti- including location, gender, education level tutional births in Kerala stands at almost of the mother, ethnicity, or caste. 100 percent, much higher than for India as Children residing in rural areas fare worse a whole. In Nepal, coverage in the mountain- than those in urban areas with regard to basic ous region stands at 19 percent, compared health services, especially when it comes to with 41 percent in the Tarai. Significant dif- institutional birth (figure 3.7). The gap is ferences also exist in coverage across religious often quite striking—in Nepal, for example, groups in all countries. In India, for instance, 32 percent of rural births are in a health coverage has always been higher among facility, compared with 71 percent of urban Christians and Buddhists than among other births. Similar differences can be found in religious groups. The same holds true for Bangladesh, India, and Pakistan. Hindus in Bangladesh. In the case of primary education, the gap Despite the substantial increase in cover- between urban and rural areas is generally age, differences between states or provinces low, with the exception of Afghanistan and can also be sizable in the case of primary Pakistan (figure 3.8). This holds true not education opportunities. Bangladesh and just for primary school attendance but also Nepal show much less within-country varia- for primary school completion. The picture tion than India and Pakistan. In these last is different in the case of secondary schools. two countries, federalism implies that states Overall coverage rates are much lower, and or provinces take a leading role in designing the gap between rural and urban areas is service-delivery policies in their jurisdiction. large, especially for school completion. The urban-rural gap is also evident in the provision of infrastructure services in Who is covered? the region, especially electricity (figure 3.9). Several population groups receive system- Rural areas fare much worse in terms of atically lower coverage of access to basic access to electricity, especially in Afghanistan services because of their circumstances, and Bangladesh. FIGURE 3.7 The coverage of institutional births is lower in rural areas a. Institutional births b. Full immunization Bangladesh Bangladesh India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of children (percent) Rural Urban Estate Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka. 106 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.8 The urban-rural gap in coverage remains large for secondary education a. Primary school attendance b. Primary school completion (ages 6–11 years) (ages 14–18 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of children (percent) c. Secondary school attendance d. Secondary school completion (ages 12–18 years) (ages 22–26 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of population group (percent) Rural Urban Sources: Based on National Risk and Vulnerability Assessment 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2010 for Bangladesh, Bhutan Living Standards Survey 2007 for Bhutan, National Sample Survey 2009 for India, HIES 2009 for Maldives, Nepal Living Standards Survey 2010 for Nepal, HIES 2010 for Pakistan, and HIES 2009 for Sri Lanka. Within cities, the disparity in access to dwellers is only part of the story. Evidence services between slum and nonslum urban suggests a shortage of schools exists for slum areas adds more nuance to the general pic- areas and this shortage is most acute at the ture of an urban-rural divide. For instance, secondary level (World Bank 2013). in Bangladesh, children of slum dwellers tend The disparity in access to basic services to have lower rates of school participation between slum and nonslum urban areas is than children living in nonslum urban areas also evident in the case of health care. In the (figure 3.10). The relatively higher opportu- urban slums in Dhaka, slum dwellers use nity costs of schooling faced by poor slum pharmacies as their primary source of health LIMITED OPPORTUNITY 107 FIGURE 3.9 Access to electricity is lower in rural FIGURE 3.10 In Bangladesh, children of slum dwellers have less areas access to education 120 Afghanistan Bangladesh 100 Bhutan 80 Percent India 60 Maldives 40 Nepal 20 Sri Lanka 0 Primary Secondary Primary Secondary 0 10 20 30 40 50 60 70 80 90 100 Gross enrollment rate Net enrollment rate Share of communities (percent) Rural Urban Slum extreme poor Slum poor Slum nonpoor Urban average Source: Based on Andres and others 2013 for this report. Source: World Bank 2013. care, just as residents of rural areas do, despite at a marginally higher rate than boys, but the fact that cheaper government health care in Bangladesh the female advantage in par- services are generally available. But the ser- ticipation does not continue through school vices provided by pharmacies are often of completion. In general, secondary school poor quality. For instance, only 8 percent of completion rates are 5 to 10 percentage the drug dispensers working in pharmacies in points higher for boys than for girls. Dhaka correctly treated dysentery patients. Although access to infrastructure facili- Similar situations have been observed in ties such as sanitation and electricity concerns many of the urban slums of South Asia locations more than individuals, the implica- (Khan, Grübner, and Krämer 2012). tions can be different for men and women. For Gender is another dimension along which instance, lack of toilets at school is well known important differences in health coverage to create a more severe obstacle for attendance may exist. But this does not appear to be the among girls than among boys. Similarly, lack case in South Asia, where boys’ advantage is of electricity at home reduces the efficiency generally very small for basic health services of domestic chores such as cooking, thereby (figure 3.11). penalizing labor force participation by women Gender differences in coverage are more than participation by men. Evidence also low in the case of primary education also indicates that electrification reduces fer- (figure 3.12). Attendance rates are similar tility rates in rural areas. One probable impact for boys and girls, except in Afghanistan of electrification on fertility is through televi- and Pakistan, where girls are at a disadvan- sion. A small but important body of evidence tage. Completion rates are slightly lower for on the impacts of television on gender roles girls than for boys, but the gap is not too and women’s empowerment supports this different from that observed for attendance. conclusion. For example, access to cable tele- In Bangladesh and Bhutan, completion rates vision is found to result in lower acceptance are actually higher for girls. of spousal abuse, lower son preference, more Gender gaps are much more pronounced female autonomy, and a greater likelihood of at the secondary level, however. In Sri Lanka sending young girls to school in rural India and Bangladesh, girls attend secondary school (Jensen and Oster 2009). 108 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.11 The coverage of health services is almost the same for boys and girls in South Asia a. Institutional births b. Full immunization Bangladesh Bangladesh India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of children (percent) Female Male Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka. Gaps in coverage are also associated case of first-generation learners. Illiterate par- with the educational attainment of par- ents or guardians often fail to fully under- ents. In Bangladesh, India, Nepal, and stand the schooling process. As a result, Pakistan the coverage of institutional births first-generation learners receive limited guid- is dramatically higher among mothers with ance and support for schooling at home; higher education than among those with no they are more prone to nonenrollment, low education. In Bangladesh, India, only 13 attendance, and dropout and face more percent of children with mothers who have difficulty in making the transition from pri- no education are fully immunized, compared mary to secondary education (Govinda and with 52 percent of children with mothers Bandyopadhyay 2011). who have higher education (figure 3.13). These analyses compare the coverage of Parents’ level of education generally does basic services along a single dimension, be not make a big difference in a child’s access it location, gender, or mother’s education. to or completion of primary education. But But disadvantaged children are often disad- parents’ level of education plays an important vantaged along several of those dimensions role in whether the child attends and com- simultaneously. An individual typically does pletes secondary school (figure 3.14). Children not belong to just one disadvantaged circum- whose parents’ have only 1 to 6 years of edu- stance group but to many such groups. These cation have much lower secondary participa- multiple disadvantages reinforce one other, tion and completion rates than those whose resulting in much lower coverage. mothers have 13 years of education or more. Even in Sri Lanka, where coverage is high, the gap between the two groups reaches 23 per- The role of inherited centage points for school participation and a circumstances staggering 67 percent for school completion. The insight that children’s characteristics, The pattern is similar throughout the region. such as their place of residence, their gen- The disadvantages faced by children with der, or their mother’s education, affect their low parental education are illustrated by the access to basic services can be used to assess LIMITED OPPORTUNITY 109 FIGURE 3.12 Gender gaps in coverage are small for primary education but large for secondary education a. Primary school b. Primary school attendance (ages 6–11 years) completion (ages 14–18 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 20 40 60 80 100 0 20 40 60 80 100 Share of children (percent) Share of children (percent) c. Secondary school d. Secondary school attendance (ages 12–18 years) completion (ages 22–26 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 20 40 60 80 100 0 20 40 60 80 100 Share of children (percent) Share of population group (percent) Female Male Sources: Based on National Risk and Vulnerability Assessment 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2010 for Bangladesh, Bhutan Living Standards Survey 2007 for Bhutan, National Sample Survey 2009 for India, HIES 2009 for Maldives, Nepal Living Standards Survey 2010 for Nepal, HIES 2010 for Pakistan, and HIES 2009 for Sri Lanka. 110 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.13 The coverage of health services differs widely by mother’s education a. Institutional births b. Full immunization Bangladesh Bangladesh India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 010 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of children (percent) No education Primary Secondary Higher Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka. the overall contribution of inherited circum- adding one more circumstance. Because cir- stances to inequality of opportunity. Several cumstances are correlated with each other, approaches have been proposed to do this the change in inequality when a circumstance (e.g., Ferreira and Gignoux 2011). Among is added depends on the initial set of circum- them, the HOI framework is particularly stances to which it is added. To calculate the appealing because it also allows decomposing impact of each circumstance, the average of inequality of opportunity between specific all possible changes to different combinations inherited circumstances and assessing how of other circumstances is computed. the contribution of each varies over time. Besides the three circumstances considered However, the HOI provides a lower bound above—location, gender, and parents’ educa- on the inequalities prevalent in a given place. tion—the decomposition exercise for South Its calculation can include only those circum- Asian countries takes into account religion stances that are measurable and for which and, in India’s case, caste. With the exception data exist. If more circumstances were added of Sri Lanka, location and mother’s educa- to the set considered in the analysis, the HOI tion explain a large fraction of the observed would increase (Paes de Barros and others inequality in access to health services (fig- 2009; Molinas and others 2010). The contri- ure 3.15). Taken together, they contribute at bution of each circumstance thus corresponds least 70 percent of the inequality in full immu- to the measurable lower bound of inequali- nization, rising to 94 percent in Pakistan and ties, not to the actual inequalities in access. 96 percent in Maldives. The explanatory In practice, the contribution of inher- power ranges between 43 percent (Sri Lanka) ited circumstances to inequality of oppor- and 98 percent (Pakistan) in the case of insti- tunity can be assessed by applying Shapley tutional births. Only in Sri Lanka is religion decompositions on the dissimilarity index. the most important correlate of inequality in Shapley decompositions, originally pro- access to health services. posed by Shorrocks (1999), show by how Location and the education level of much inequality changes as a consequence of household head also explain a large part of LIMITED OPPORTUNITY 111 FIGURE 3.14 Parents’ education is highly correlated with children’s secondary school attainment a. Primary school attendance b. Primary school completion (ages 6–11 years) (ages 14–18 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of children (percent) c. Secondary school attendance d. Secondary school completion (ages 12–18 years) (ages 22–26 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Share of children (percent) Share of population group (percent) No education Primary Secondary Higher Sources: Based on National Risk and Vulnerability Assessment 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2010 for Bangladesh, Bhutan Living Standards Survey 2007 for Bhutan, National Sample Survey 2009 for India, HIES 2009 for Maldives, Nepal Living Standards Survey 2010 for Nepal, HIES 2010 for Pakistan, and HIES 2009 for Sri Lanka. the inequality in access to primary school- important, gender plays a significant role in ing and in its completion (figure 3.16). In explaining secondary school attendance and several countries in the region, importantly completion across countries in the region. Bangladesh, India, Nepal, and Sri Lanka, reli- For infrastructure, both the location of the gion also explains some part of the inequality household and the education level of house- in access to primary education. In India, caste hold head are critically important circum- explains more than religion. stances. Their relative contribution varies The relevant circumstances are more by the types of services and across countries diverse in the case of secondary education. (figure 3.17). The location of the household Although the education of the household is the most important circumstance in access head and location of residence continue to be to electricity in Afghanistan, Bangladesh, 112 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.15 Location and mother’s education are critically important circumstances in access to health services a. Full immunization b. Institutional births Bangladesh Bangladesh India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Contribution to measured inequality (percent) Contribution to measured inequality (percent) Caste Religion Gender Urban/rural Mother’s education Sources: Based on Demographic and Health Survey (DHS) 2011 for Bangladesh, DHS 2005 for India, DHS 2009 for Maldives, DHS 2011 for Nepal, DHS 2007 for Pakistan, and DHS 2007 for Sri Lanka. Bhutan, and India. It is also the dominant Low levels of public spending are not circumstance for access to sanitation in immediately apparent when considering the Maldives and Pakistan. In Nepal and Sri share of government expenditures devoted to Lanka, the education level of the household health care and education. For most countries head is more important in explaining the in the region, that share is indeed close to the inequality of opportunity in access to infra- average for advanced countries. But advanced structure services. In India, caste also plays countries have much larger government an important role. expenditures relative to their gross domestic product (GDP) than their South Asian coun- terparts, so this comparison is only partially Limited resources and low relevant. progressivity A more meaningful comparison involves Several factors explain South Asia’s lackluster comparing public spending on health care performance in ensuring equality in access as a fraction of GDP, rather than as a share to basic services. Importantly, public spend- of the budget, and controlling for the level ing on education and health is relatively low. of economic development. This comparison No doubt, higher public spending by itself shows clearly that except for Bhutan and may not be conducive to better outcomes, Maldives, all countries in the region spend because the efficiency of spending mat- much less public resources on health services ters as well. However, bigger public outlays than could be expected, given their income are a prerequisite for improving coverage. per capita (figure 3.18). Moreover, equity requires committing addi- Public spending on health is also low in tional resources to children of disadvantaged comparison with out-of-pocket spending by groups to offset their more unfavorable cir- households. Whereas public expenditures cumstances. Unfortunately, the opposite is on health care are below the global aver- often observed in the region, with the better- age, out-of-pocket expenditures are well off benefiting from higher levels of public above, with the exception of Bhutan. Out- spending than those with greater needs. of-pocket expenditures accounted for about LIMITED OPPORTUNITY 113 FIGURE 3.16 Gender and religion matter for access to education a. Primary school attendance (ages 6–11 years) b. Primary school completion (14-18 years of age) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Contribution to measured inequality (percent) Contribution to measured inequality (percent) Caste Religion Gender Urban/rural Household head’s education c. Secondary school attendance (ages 12–18 years) d. Secondary school completion (ages 22–26 years) Afghanistan Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Contribution to measured inequality (percent) Contribution to measured inequality (percent) Caste Religion Gender Urban/rural Household head’s education Sources: Based on National Risk and Vulnerability Assessment 2007 for Afghanistan, Household Income and Expenditure Survey (HIES) 2010 for Bangladesh, Bhutan Living Standards Survey 2007 for Bhutan, National Sample Survey 2009 for India, HIES 2009 for Maldives, Nepal Living Standards Survey 2010 for Nepal, HIES 2010 for Pakistan, and HIES 2009 for Sri Lanka. 58 percent of total health expenditures in with a higher vulnerability of the popula- the South Asia region in 2012. By contrast, tion to fall into poverty or remain poor the average out-of-pocket spending in the because of catastrophic expenditures. world is about 18 percent. The combina- Regional trends over the past 10 years do tion of low public outlays and high out-of- not indicate major changes regarding this pocket expenditures is typically associated vulnerability. 114 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.17 Location is a critical circumstance for access to infrastructure services a. Access to electricity b. Access to improved sanitation Afghanistan Bangladesh Bangladesh Bhutan Bhutan India India Maldives Maldives Nepal Nepal Pakistan Sri Lanka Sri Lanka 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Contribution to measured inequality (percent) Contribution to measured inequality (percent) Caste Urban/rural Household head’s education Source: Based on Andres and others 2013 for this report. FIGURE 3.18 Limited public resources are spent on health services in South Asia 18 16 Public expenditure on health (percent of GDP, 2012) 14 12 10 8 6 4 Maldives Nepal Bhutan 2 Afghanistan India Bangladesh Pakistan Sri Lanka 0 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Log of GDP per capita (PPP, constant international 2011 dollars, 2012) Source: Based on data from the World Bank’s World Development Indicators database, http://data.worldbank.org/data-catalog/world-development -indicators. Not only are public resources allocated to in a relatively even manner in Sri Lanka. But health care low in most South Asian coun- it is directed more toward the better-off than tries, but in most of the region they are also to the poor in Bangladesh, India, and Nepal. allocated with low progressivity. Overall, Furthermore, public spending still works to public spending on health care is distributed reduce the gap between the rich and the poor LIMITED OPPORTUNITY 115 in Bangladesh but does not appear to do so FIGURE 3.19 Spending is progressive only for some health services in India and Nepal (O’Donnell and others in Bangladesh 2007). This is in clear contrast to public health spending patterns in East Asian econ- omies, which are much more pro-poor. 30 The progressivity of public spending varies Share of spending (percent) 25 across health services. Public spending on hospital care, which tends to support more 20 sophisticated services, is biased toward the 15 better-off. By contrast, spending on nonhos- pital care is pro-poor. Because a large share 10 of public spending on health care is generally 5 allocated to hospital care, the overall spend- ing pattern is regressive. 0 Nepal is a case in point. Spending on hos- Family planning Limited Maternal Child All public pitals and mobile clinics is found to benefit the and control of curative health health health communicable care expenditure richer population more while gross subsidies diseases for local health facilitates are significantly pro- Consumption quintiles gressive. Even after taking fee payments into 1st (poorest) 2nd 3rd 4th 5th (richest) account, public spending on hospitals—net spending—is still biased against the bottom Source: World Bank 2003. population quintile, because of the prohibitive access costs (Silva-Leander 2012). The extent received 50 percent of public primary spend- of progressivity also varies across types of ser- ing in 2010, up from 32 percent in 2005. vices. In Bangladesh, for example, spending on However, the richest 40 percent received child care is clearly progressive, whereas cura- about 80 percent of public spending directed tive care tends to be regressive (figure 3.19). to tertiary education (figure 3.21). A similar South Asian countries also spend less on pattern is found in India and Pakistan. education than other countries at a similar The different extent of progressivity level of development (figure 3.20). In par- across education levels results partly from the ticular, Bangladesh, India, Pakistan, and very different shares of the relevant popula- Sri Lanka all sit far below the international tion covered by each level. Access to primary trend line between public spending on educa- education has been broadened greatly in the tion and income level. India’s public spend- region, but access to secondary and tertiary ing per student on primary education puts it education remains low. However, public among the lowest quintile of all countries with spending per student is generally higher for available information. Its public spending per the latter. In India, for instance, public spend- student on secondary education is higher but ing per student is above the world average for still falls below the average of comparators. tertiary education but below it for primary Trends have not been encouraging either. and secondary education. Because access to Overall, rapid economic growth in South Asia higher education is highly correlated with did not lead to significant increases in per stu- household income, geographical location, dent expenditure, which generally remained and social background, the result is a more constant (World Bank 2014). regressive pattern of spending when mov- The extent of progressivity in public spend- ing from lower to higher levels of education ing varies across education levels. It tends to (World Bank 2014). be progressive in the case of primary educa- Public spending on education is also tion but regressive at secondary and especially found not to reach its intended beneficiaries. at tertiary levels. For instance, the poorest Because the more disadvantaged suffer dis- 40 percent of the population in Bangladesh proportionately from resource waste of this 116 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 3.20 Limited public resources are spent on education services 14 Public expenditure on education (percent of GDP, 2009) 12 10 8 Maldives 6 Nepal Bhutan 4 India Bangladesh 2 Pakistan Sri Lanka 0 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 Log of GDP per capita (PPP, constant international 2011 dollars, 2012) Source: Based on data from the World Bank’s World Development Indicators database, http://data.worldbank.org/data-catalog/world-development -indicators. FIGURE 3.21 In Bangladesh, education spending is progressive sort, public spending can be less progressive only at the primary level than figures suggest. An example is the use of funds under Sarva Shiksha Abhiyan—India’s 70 flagship program for universalizing elemen- tary education. Schools frequently do not get 60 their money in whole or on time (Dongre, Share of spending (percent) 50 Chowdhury, and Aiyar 2012). Among nine districts of seven Indian states, no district 40 gets all the funds allocated to it. Considerable variation occurs between districts and across 30 schools, however. Some districts get a much 20 higher percentage of the funds allocated to them; in some districts, only 11 percent 10 to 12 percent of schools report receiving the grants annually. Many districts receive 0 Primary Secondary Higher secondary Tertiary funds late in the school year, which prevents Consumption quintiles them from procuring important teaching and learning materials or from undertaking 1st (poorest) 2nd 3rd 4th 5th (richest) school maintenance until they are five or six Source: World Bank 2013. months into the school year. LIMITED OPPORTUNITY 117 Limited opportunity: Main messages and policy implications Equality of opportunity is associated with Circumstances at birth—including location notions of fairness. It is widely believed that along the rural-urban divide, gender, religion or people who have the same talent and the same caste, and parents’ education—remain crucial willingness to work hard should have the same determinants of access. The coverage of basic prospects of succeeding, regardless of their cir- services is consistently lower in rural areas than cumstances at birth. Access to basic services such in urban areas. Gender is not an important cor- as health, education, and infrastructure, espe- relate of access in the case of health services and cially at the beginning of life, is seen as essential primary education, but girls have lower access to ensure they do. Equality of opportunity also to secondary education than boys. Mother’s matters for positive reasons. Typically, children education makes a substantial difference in from better-off households have greater access to access to health services and in school attain- basic services in their childhood, which allows ment. Overall, location and mother’s education them to live a healthier life and to accumulate account for most of the inequality in access to more human capital than children from more health, whereas gender and religion also mat- disadvantaged backgrounds. As a result, the ter for the inequality in access to education. gap in well-being is bigger in adulthood than in Location is a critical factor in relation to access childhood. In societies undergoing demographic to basic infrastructure services. transitions, the share of adults increases steadily, Low coverage of basic services and limited implying that a stable inequality of opportunity equity in accessing them can be traced to insuf- translates into a growing inequality of outcomes. ficient and often regressive public spending. In South Asia, access to basic health services, Compared with spending in other countries at including immunization and institutional birth, a similar level of development, public spend- tends to be limited. Access to primary education ing on health and education represents a tends to be better, because most countries in the lower share of GDP in South Asian countries. region have reached universal school coverage or Depending on the service considered, public are on the way to attaining it. In between these spending can be progressive—more tilted two extremes, the picture on access to secondary toward the poor—or regressive. For instance, education and to basic infrastructure services such spending on child health and on primary edu- as electricity and sanitation is more mixed. The cation tends to be progressive, whereas outlays coverage of all these services is generally expand- related to hospital care or tertiary education are ing, but progress is less in equity, because disad- regressive. Although the number of beneficia- vantaged population groups remain consistently ries of the regressive programs is smaller, spend- excluded. Even the progress in coverage is some- ing on them is larger. This makes the overall what misleading, because the expansion in access public spending on the provision of basic ser- to basic services often occurs at levels of quality vices regressive, thus undermining equality of below those available for better-off groups. opportunity. References Arneson, Richard J. 1989. “Equality and Equal Opportunity for Welfare.” Philosophical Andres, Luis, Dan Biller, Mario Picon, and Studies 56 (1): 77–93. Ag usti n E chen ique. 2013. “I ncreasi ng Cohen, Gerald A. 1989. “On the Currency of Opportunities in South Asia—Infrastructure.” Egalitarian Justice.” Ethics 99 (4): 906–44. Background paper for this report, World Bank, Dongre, Ambrish, Anirvan Chowdhury, and Washington, DC. Yamini Aiyar. 2012. “Unpacking School 118 ADDRESSING INEQUALITY IN SOUTH ASIA Finances.” Background paper for this report, Sh iva R aj Ad h i k a r i , D en i H a rbia nto, World Bank, Washington, DC. Charu C. Garg, Piiya Hanvoravongchai, Dworkin, Ronald. 1981a. “What Is Equality? Mohammed N. Huq, Anup Karan, Gabriel Part 1: Equality of Welfare.” Philosophy and M. Leung, Chiu Wan Ng, Badri Raj Pande, Public Affairs 10 (3): 185–246. Keith Tin, Kanjana Tisayaticom, Laksono ———. 1981b. “What Is Equality? Part 2: Trisnantoro, Yuhui Zhang, and Yuxin Zhao. Equality of Resources.” Philosophy and Public 2007. “The Incidence of Public Spending Affairs 10 (4): 283–345. on Healthcare: Comparative Evidence from Ferreira, Francisco H. G., and Jérémie Gignoux. Asia.” World Bank Economic Review 21 (1): 2011. “The Measurement of Inequality of 93–123. Opportunity: Theory and an Application Paes de Barros, Ricardo, Francisco H. G. Ferreira, to Latin America.” Review of Income and Jose R. Molinas Vega, and Jaime Saavedra Wealth 57 (4): 622–57. Chanduvi. 2009. Measuring Inequality of Govinda, R., and Madhumita Bandyopadhyay. Opportunities in Latin America and the 2011. “Access to Elementary Education in Caribbean. Washington, DC: World Bank; India: Analytical Overview.” In Who Goes New York: Palgrave Macmillan. to School? Exploring Exclusion in Indian Rawls, John. 1971. A Theor y of Justice . Education, edited by R. Govinda. New Delhi: Cambridge, MA: Harvard University Press. Oxford University Press. Roemer, John E. 1998. Equality of Opportunity. Jensen, Robert, and Emily Oster. 2009. “The Cambridge, MA: Harvard University Press. Power of TV: Cable Television and Women’s Shorrocks, Anthony F. 1999. “Decomposition Status in India.” Quarterly Journ al of Procedures for Distributional A nalysis: Economics 124 (3): 1057–94. A Unified Framework Based on the Shapley Kanbur, Ravi. 2009. “Intergenerationalities: Value.” Department of Economics, University Some Educational Questions on Quality, of Essex, Colchester, U.K. Qu a nt it y a nd Oppor t u n it y.” Work i ng Silva-Leander, Sebastian. 2012. “Benefit Incidence Paper No. 48922, Department of Applied Analysis in Health.” Nepal Health Sector E conom ic s a nd M a n agement , C or nel l Support Programme (NHSSP), Government University, NY. of Nepal. Khan, M. M. H., Oliver Grübner, and Alexander World B a n k . 2 0 03. B a ngl a d e s h: Pu bl i c Krämer. 2012. “Frequently Used Healthcare Expenditure Review. Report No. 24370-BD. Services in Urban Slums of Dhaka and Adjacent Washington, DC: World Bank. Rural Areas and Their Determinants.” Journal ———. 2005. World Development Report 2006: of Public Health 34 (2): 261–71. Equity and Development. Washington, DC: Molinas, José R., Ricardo Paes de Barros, Jaime World Bank. Saavedra, and Marcelo Giugale. 2010. Do ———. 2013. Bangladesh Education Sector Our Children Have a Chance?: The 2010 Review: Seeding Fertile Ground—Education Hum an Oppor tunit y Re por t for L atin That Works for Bangladesh. Washington, DC: America and the Caribbean . Conference World Bank Group. Edition. Washington, DC: World Bank. ———. 2014. Student Learning in South Asia: O’Donnell, Owen, Eddy van Doorslaer, Ravi C halle nges, Oppor tunities, and Polic y P. Rannan-Eliya, Aparnaa Somanathan, Priorities. Washington, DC: World Bank. Substantial Mobility 4 E conomic mobility has often been seen incentives. In a society where the poor and as an avenue to long-term equality the rich (or at least their children) are equally (box 4.1). The celebrated Austrian likely to succeed or fail, people belonging to economist Joseph Schumpeter (1955) is said all groups have a higher motivation to work to have likened income distribution to a hard. Mobility has been seen to foster aspira- hotel. The rooms at the top are luxurious, tion, efforts, innovation, and self-fulfillment. those on the middle levels are ordinary, and Last but not least, mobility reduces waste. those in the basement are downright shabby. Highly mobile societies are less prone to At any given time, the occupants of the hotel social conflict and less tempted by redistribu- experience very unequal accommodations. tive policies, both of which have negative At a later time, if one reexamines who is liv- implications for economic growth. ing where, one fi nds that some have moved Although economists tend to think of to higher floors, some have moved to lower mobility in terms of income and consump- floors, and some have stayed where they tion, economic and social mobility are inter- were. The difference in the quality of hotel twined, especially through jobs. In most rooms provides a static measure of inequal- societies, jobs are fundamental sources of ity. The movement of hotel guests among self-respect and social identity. The distribu- rooms of different quality is mobility. The tion of jobs within society and perceptions more movement of guests that occurs among about who has economic mobility can shape floors, the greater is the likelihood of long- individuals’ perceptions of fairness and aspi- term equality in accommodations. rations for the future. In South Asia, mar- Economic mobility is also a foundation ginalized population groups traditionally for efficiency. Mobility leads to a better use suffer from both material poverty and social of talent. If the distribution of creativity or indignity. This situation is most obvious in resourcefulness across the population is less the case of a caste system, where occupations unequal than the distribution of income or are essentially set for individuals at birth. In a consumption, societies with greater mobil- perfectly mobile society, by contrast, occupa- ity may be able to mobilize the talent of all tional choices should be independent across population groups and not only those of generations. For instance, the children of the better-off. Mobility also strengthens manual and nonmanual workers would have 119 120 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 4.1 For a given inequality of opportunity, mobility reduces the inequality of outcomes Even if equality of opportunity were assured, at This simple analysis has important implica- any point in time differences in well-being would tions. Because much of the debate on inequality exist across households. Even when they have has focused on opportunities, policy emphases the same access to health, education, and basic have been on access to basic services, includ- infrastructure, people may differ in their talent ing health, education, and infrastructure as or in their natural inclination to work hard. As seen from the household perspective. But the a result, some inequality of outcomes can be role of mobility as a driver of reduced inequal- expected even in youth, when people are just ity of outcomes calls for a focus on jobs and the leaving school and making their fi rst choices on rural-urban transformation. Even if this case also where to live, whether to work, and what kind of has implications for infrastructure, the kind of work to do. investments needed could be quite different. This initial inequality of outcomes would of course be even greater if no real equality of oppor- tunity exists, as is generally the case in South Asia. Final However, it does not follow that countries lacking well-being equality of opportunity are condemned to experi- ence a large and potentially growing inequality of ET Impact on outcomes. Whether this happens depends in large inequality part on how the workings of the economy reward the choices people make about where to live and eT what kind of work to do. In rapidly urbanizing Location, jobs societies, switching out of agriculture, moving to cities, and working for a wage may offer avenues for mobility to some of the poorest population groups. Thus, despite being at a disadvantage in Initial earnings terms of health or skills, poorer households could 45° end up experiencing bigger gains in well-being e0 E0 Initial than the better-off, and inequality of outcomes well-being Poorer Richer in adulthood could be smaller than in childhood. household household The figure illustrates that possibility. similar prospects of securing nonmanual and greater social prestige. Moreover, the jobs. If so, growth should allow the people association between the occupations of sons from marginalized groups to move into the and fathers has declined steadily over time. mainstream of society. That is, occupational mobility has become These intuitions can be followed to assess higher for younger generations. The improve- the extent of mobility across and within gen- ment is more pronounced for the more erations in South Asian countries. Contrary socially marginalized populations, namely, to expectations, economic mobility has the Scheduled Castes, the Scheduled Tribes, become substantial in the region in recent and Other Backward Castes. decades, especially for socially marginal- In the case of Bangladesh and India, when ized groups. In India, the sons of unskilled splitting the population into three groups— or farming fathers see significant prospects of poor, vulnerable, and middle class—a siz- working in occupations with higher income able proportion of the poor moves up to the SUBSTANTIAL MOBILIT Y 121 vulnerable group within the same generation, status in urban areas in Bangladesh and India while a large proportion of the vulnerable despite the existence of a large urban infor- moves into the middle class. In India, the mal sector. Urban wage employment, regard- upward mobility of households belonging less of whether it is formal or informal, offers to Scheduled Castes and Scheduled Tribes is better prospects of economic mobility for a similar to that of the rest of the population. massive number of households in South Asia. In a way, the dynamism of market forces Urbanization is taking place through diversi- appears to have trumped the rigidity of social fied processes, however. People move to cities institutions. through migration, but cities also “move” At the level of villages, increasing mobil- to people through the transformative effect ity is largely associated with occupational of economic activity in formerly rural areas. change. Rural areas in South Asian countries These diverse processes have created cities with have seen nonfarm activities expanding and different characteristics. The availability and generating many of the new jobs. Although nature of jobs—hence mobility prospects— nonfarm employment is mainly casual, it vary across cities of different types. has supported mobility. The earnings gap between regular and casual nonfarm jobs has narrowed over time, whereas the earn- Mobility across generations ings gap between casual nonfarm jobs and Whereas economists tend to assess mobility agricultural jobs has increased. Growing on the basis of monetary indicators such as integration with urban economies, expand- income or consumption, mobility has social ing markets, and migration opportunities are dimensions as well. Sociologists generally also shaking up traditional social structures see mobility through the lens of classes and, and offering a chance for the most disadvan- often, occupational groups. Some of them taged in these villages to prosper. believe that occupation is the most critical Urbanization and the associated geo- factor in an individual’s social standing, life graphical differences in economic dynamism chances, and level of material comfort and are important forces underpinning mobility. that occupation defi nes the extent to which Those geographical differences lie behind an individual is advantaged or disadvan- internal migration. In South Asia, permanent taged (Giddens 2009). internal migration accounts for a significant Increasingly, economists have started to fraction of the population across countries. adopt occupational change as a metric for Many move to fulfill their aspiration for measuring mobility, partly because of data jobs and investment opportunities. Indeed, availability (Long and Ferrie 2007, 2013). permanent migration is an important strat- More important, economists have started to egy for upward mobility for both men and recognize the broader implications of occupa- women. Seasonal migration has been con- tions and jobs. Jobs contribute to individuals’ sistently identified as more likely to originate view on who they are and their relationships in groups that are poorer and more socially with others. In most societies, jobs are fun- marginalized and in places that are more damental sources of self-respect and social prone to natural disasters and confl ict. For identity. Some jobs are empowering and oth- people from these groups and regions, tem- ers are less so. In the extreme case, the lack porary migration is a viable accumulation of jobs can contribute to violence. The distri- strategy, an effective seasonal distress reduc- bution of jobs within society and perceptions tion mechanism, and even an escape route about who has access to job opportunities from social discrimination. influence individuals’ perceptions of fairness Cities themselves support greater mobil- and aspirations for the future (Akerlof and ity than rural areas, both within and across Kranton 2010; World Bank 2012b). generations. A larger fraction of poor and In South Asia, economic and social vulnerable households achieves middle-class mobilit y are historically intert wined. 122 ADDRESSING INEQUALITY IN SOUTH ASIA The caste system, an inflexible and lineage- are available in the case of sons; in the case of based social ordering of Hinduism, is a clear fathers, data refer to occupation for most of manifestation of the links between the two life as reported by sons. For both generations, dimensions. The ancient Hindu society strat- occupations are mapped into four categories ified the population into mutually exclusive, in a broadly ascending order of economic hereditary, and occupational-specific com- returns and levels of social prestige: unskilled partments. The occupations reserved for workers, farmers, skilled or semiskilled the top castes were generally nonmanual workers, and white-collar workers. and not linked to the production of objects The analysis reveals considerable occu- or the provision of services. In contrast, pational mobilit y across generations the occupations mandated for out-castes (figure 4.1). The sons of unskilled fathers and (Scheduled Castes) were menial in nature: those of farming fathers both saw signifi- for example, skinning dead cows and sweep- cant prospects of moving to higher-ranked ing roads. Occupations were essentially jobs in terms of economic returns and social set for individuals at birth, leaving little prestige. Over 40 percent of the children of space for aspiration, efforts, innovation, unskilled workers were holding other occu- and fulfillment (Deshpande 2000). Other pations. About 36 percent of the children socially marginalized populations, including of farmers worked as skilled or semiskilled Scheduled Tribes, and other religious and workers or as white-collar workers. The chil- ethnic minorities traditionally suffer from dren of white-collar fathers experienced the social indignity and material poverty. highest downward mobility rate among all In contrast, a perfectly mobile society is four groups—over 55 percent. one where occupation in one generation is not Occupational mobility across genera- particularly informative to predict occupa- tions is not only substantial, but it has also tion in the next one. That would imply that increased over time (figure 4.2). Based on the the sons of fathers in occupations with higher notion of occupational independence, the economic returns and social prestige have no analysis compares the occupational transi- advantage in securing such an occupation tion matrices of sons from each of the four relative to sons of fathers in occupations with birth cohorts (1945–54, 1955–64, 1965–74, lower economic returns and social prestige. and 1975–84) with the hypothetical transi- This extreme situation, described as “occu- tion matrix of a perfectly mobile society. pational independence,” captures the notion The comparison shows that the differences of perfect mobility across generations. Under between two successive 10-year cohorts are complete occupational independence, the chil- not always significant. However, the mobility dren of farmers and those of engineers would experienced by sons relative to the perfectly have similar prospects of securing an engi- mobile society has increased significantly neering job, while the children of Scheduled between the first and the last cohorts. On Castes and those of other population groups average, the children of people in basic occu- would face similar probabilities of doing non- pations have seen rising prospects of taking manual jobs. higher-ranked occupations relative to the Because of limitations of data availabil- children of people in higher-ranked occupa- ity, a measure of occupational independence tions. The conclusion appears to be robust across generations in South Asia can be to the classification of occupations and the computed only for India (box 4.2). There, methodology used for the comparison across matched father-son information is reported cohorts (Singh and Motiram 2012). by the IHDS and it is based on a nation- The increases in occupational mobility ally representative sample. The sons in this across generations over time may be more data set include birth cohorts from as early pronounced than what is being observed. as 1945 and as late as 1984. Data on self- Two important caveats relate to the data. reported occupation in the preceding year Fathers and sons potentially are of different SUBSTANTIAL MOBILIT Y 123 BOX 4.2 Measuring mobility Economists have developed multiple measures seen in a positive light. In the second case, only to assess the extent of mobility. These mea- greater movement from low- to high-prestige sures reflect different notions of what economic attainment—greater upward mobility—is seen mobility actually is (Björklund and Jäntti as socially desirable (Erikson and Goldthorpe 2000; Fields 2008; Fields and Ok 1996, 2000; 1992; Ganzeboom and Treiman 1996; and Salverda, Nolan, and Smeeding 2011). Broadly Björklund and Jäntti 2000 for a survey). speaking, three types of measures exist: move- Assessing occupational mobility requires ment, origin independence, and long-term computing how people’s occupations change equalization of outcomes such as earnings and over time, under the form of a transition matrix. incomes (see Ferreira and others 2011 for sur- The question is whether having a father in a veys; Fields 2008). Movement measures focus higher-ranked occupation improves the odds on changes in the distribution of outcomes over of a son being in a higher-ranked occupation time. Origin-independence measures focus on and by how much. When there are only two the correlation between positions in the distri- occupations, that question can be answered by bution of outcomes at two points in time. Long- looking at the cross-product ratio, defined as term equalization measures focus on inequality the ratio of (a) the odds that sons of fathers in in permanent outcomes. These measures can the basic occupation will be in the basic rather be analyzed for the same individuals over time than the higher occupations to (b) the odds that (intragenerational mobility) or for related indi- sons of fathers in the higher occupation will be viduals across generations (intergenerational in the basic rather than the higher occupation. mobility). Because the analysis requires infor- If the knowledge of a father’s occupation yields mation on the same individuals or households no information about the odds of a son being at different points in time, data availability in the basic or higher occupation the cross- heavily influences the selection of indicators of product ratio is equal to one. mobility. When there are more than two occupa- Comparing occupations over time is in a way tions, the number of cross-products increases. more complex than comparing income or con- Altham statistics can be used to combine all of sumption. Empirically, classes and occupational them to compare the actual transition matrix groups are taken as either intrinsically discrete to the hypothetical matrix of a perfectly mobile and unordered or as associated with differ- society, in which all cross-product ratios are ent levels of prestige. In the first case, greater equal to one (Ferrie 2005; Long and Ferrie fluidity among the groups—in any direction—is 2007). ages and thus may be at different points age differences. Without accounting for of their life cycle. Individuals of the four age differences, the analysis may underes- cohorts are of different ages and are likely timate the increases in occupational mobil- to be at different points of their life cycle as ity over time. well. When individuals are younger, they T he most notable improvements in are more likely to start with the profession mobility are found for Scheduled Castes, of their parents or to land jobs with lower Scheduled Tribes, and Other Backward economic and social prestige. The data do Castes. The analysis is conducted sepa- not allow more refi nement to control for rately for sons from two broader birth 124 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4.1 Considerable occupational mobility exists across mobility among Muslims is similar to that generations in India of higher-caste Hindus, whereas mobil- ity among Scheduled Castes and Scheduled 80 Tribes and among Other Backward Castes Occupations of sons (percent) 70 has become higher (table 4.1). Again, the 60 conclusion appears to be robust to the choice 50 of methodology (Hnatkovska, Lahiri, and 40 Paul 2013). 30 20 10 Mobility within the same 0 generation Unskilled Farmers Skilled/ White collar semiskilled A truly mobile society is arguably one in Occupations of fathers which poorer households can manage to Downward mobility Persistence Upward mobility climb up the income or consumption ladder Source: Based on India Human Development Survey (IHDS) 2004–05. through their own efforts within a single generation. Mobility of this sort requires that the growth in the income or the con- FIGURE 4.2 Occupational mobility is higher for younger generations sumption of the poor be faster than the growth of the average person. When split- ting the population into three groups—poor, mobile society (Altham statistic) 22.5 “Distance” from a perfectly vulnerable, and middle class—a fraction of 22.0 the poor should be able to move above the 21.5 poverty line while some of the vulnerable 21.0 should be able to make solid progress into the middle class. A truly mobile society is 20.5 one in which this prospect also exists for 20.0 socially marginalized groups. 1945–54 1955–64 1965–74 1975–84 Mobility within the same generation can Birth cohorts of sons be assessed through the average growth rate of household income or consumption Source: Based on IHDS 2004–05. between two periods (Ferreira and others Note: “Distance” is measured as the Altham statistic comparing the actual transition matrix to that of a hypothetical, perfectly mobile society. 2012; Fields 2010; Fields and Ok 1996). But doing so requires nationally represen- cohorts, 1945– 64 and 1965–84, to have tative information on the same households sufficiently large samples. The transition at two points in time—so-called panel matrices of socially marginalized groups are data—which is generally not available for compared in this case with those of higher- South Asian countries. One exception is caste Hindus. The comparison shows that the Pakistan Social and Living Standards TABLE 4.1 Occupational mobility has increased more for the most disadvantaged population groups in India Occupational mobility, cohort born in Occupational mobility, cohort born in Disadvantaged group 1945–64 1965–84 Muslims No difference from higher castes No difference from higher castes Scheduled Castes and Scheduled Less than among higher castes More than among higher castes Tribes Other Backward Castes No difference from higher castes Slightly more than among higher castes Source: Based on IHDS 2004–05. SUBSTANTIAL MOBILIT Y 125 Measurement survey, but the period it National Sample Survey (NSS) 1993–94, covers—between 2007 and 2009—is too NSS 2004 – 05, and NSS 2009–10. The short to meaningfully track economic and third panel is for Bangladesh between 2005 social mobility. To overcome this difficulty, and 2010 and is based on the Household three synthetic panel data sets were espe- Income and Expenditure Survey (HIES) for cially built for this report, linking different 2005 and 2010. rounds of nationally representative surveys For India, results show that consump- (box 4.3). Two of the synthetic panels are tion growth was faster among the poor than for India; they cover the periods between among the better-off (figure 4.3). The trend 1993–94 and 2004–05 and between 2004– is not apparent when comparing the con- 05 and 2009 –10 and are based on the sumption per capita of any one decile of BOX 4.3 How synthetic panels are constructed Panel data conventionally underpin the analysis consumption per capita. For each survey, of economic mobility over time. However, such household consumption per capita is predicted data are not readily available for most develop- based on observable household characteristics ing countries. In South Asia none of the pub- and is broken down into two components. The licly available household expenditure surveys is first component is associated with observable a real panel over a sufficient period of time— characteristics of the households—such as size, more than two years—to allow a meaningful age composition, education, or assets—and assessment of economic mobility. The method- the corresponding estimated coefficients. The ology to construct synthetic panels was intro- second component is associated with nonob- duced to address this limitation in data. The servable household characteristics—such as robustness of results derived from these panels, entrepreneurship or work ethic—and is cap- relative to real panel data, is still being assessed. tured through the error term. But consensus exists that synthetic panels pro- The first component of predicted consump- vide a promising way forward in the absence of tion is constructed by applying the coeffi- real panel data. cients estimated for the 2009–10 survey to To illustrate the way the synthetic panel the corresponding observable characteristics methodology works, consider the Indian NSS in the 2004–05 survey. The second compo- of 2004–05 and that of 2009–10. The approach nent is estimated based on the estimated cor- consists of using data on households appearing relation coefficient of the error terms between in the 2004–05 survey as the base year and pre- both surveys. If a point estimate on this cor- dicting their consumption in 2009–10 through relation coefficient is used, a point estimate of an imputation methodology. Probabilities are this component can be constructed; otherwise, predicted for households falling into different if a range of estimates is used, an upper and combinations of living standards in the two a lower bound estimate result. The predicted periods (e.g., being poor in both periods, or consumption in 2009–10 of each household being poor in the first period but nonpoor in appearing in the 2004–05 survey is estimated the second period, and so on). Underlying the as the sum of the predicted values of the two predicted status in each period is household components. Source: Based on Dang and Lanjouw 2013. 126 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4.3 Consumption grows faster among the poor than among the better-off a. India, 1993–94 to 2004–05 b. India, 2004–05 to 2009–10 4 4 Annual change in consumption Annual change in consumption 3 3 per capita (percent) per capita (percent) 2 2 1 1 0 0 –1 –1 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th (poorest) (richest) (poorest) (richest) Consumption deciles Consumption deciles Anonymous Nonanonymous Source: Based on data from Dang and Lanjouw 2014 for this report. the distribution in two years—the so-called upward and downward—in a particular anonymous distribution—simply because the society. households in a particular decile are unlikely To distinguish between the vulnera- to be the same in the two years. A more ble and the middle class, the concept of a meaningful assessment involves compar- vulnerability line needs to be introduced. ing the consumption per capita of the same For each nonpoor household in the initial households in the initial year and in the final year, one can compute a vulnerability index, year. Based on this other approach—the so- defi ned as the probability that the house- called nonanonymous distribution—mobil- hold will have fallen into poverty by the ity has become particularly strong in more fi nal year. For any level of the vulnerability recent years. index, a vulnerability line can be identified Another way to assess mobility is to (Dang and Lanjouw 2014). The synthetic focus on the fraction of households that panel data sets constructed for South transit from one well-defined population Asia allow studying household transitions group to another. For instance, the popu- between 2005 and 2010 in Bangladesh and lation can be classified into three mutually between 2004– 05 and 2009–10 in India exclusive categories: the poor, the vulner- using this approach. If the vulnerability able, and the middle class. The poor are index is set at 20 percent, the vulnerability those whose consumption falls below the line for Bangladesh is 1,582 taka per person official poverty line. The vulnerable are per month at 2005 prices, and the corre- households that are not poor in the initial sponding line for India is 998 rupees per year but that face a significant risk of fall- person per month at 2004 prices. ing into poverty in the fi nal year. And the The analysis of household transitions middle class is made up of the remaining using these vulnerability lines shows that households: those who are not poor in the upward mobility was considerable for both initial year and for whom the risk of fall- the poor and the vulnerable (table 4.2). ing into poverty in the final year is low. In India, the share of the population liv- Household transitions between these three ing below the pover t y line (the poor groups reveal the extent of mobility—both group) declined from almost 37 percent in SUBSTANTIAL MOBILIT Y 127 TABLE 4.2 Upward mobility is considerable among the poor and the vulnerable Five years later (percent) Poor Vulnerable Middle class Total Bangladesh 2005 Poor 19.5 16.2 2.8 38.5 Vulnerable 11.2 22.4 9.7 43.3 Middle class 1.5 7.3 9.4 18.1 Total 32.1 45.9 22.0 100.0 India 2004–05 Poor 22.2 13.9 0.8 36.8 Vulnerable 8.9 28.8 8.5 46.3 Middle class 0.3 5.8 10.7 16.8 Total 31.5 48.5 20.0 100.0 Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The household head’s age is restricted to between 25 and 55 years of age for the first survey and adjusted accordingly for the second survey. All cell numbers are significantly different from zero. The circled cells show the shares of the total population who experienced upward mobility. Percentages may not total to 100 because of rounding. 2004–05 to about 32 percent in 2009–10. moving above the poverty line and a consid- Although some households fell into poverty erable fraction of the poor and vulnerable between the two periods, more of them— moving into the middle class. Downward about 15 percent of the total population, mobility was much bigger in the two South or about 40 percent of the poor—moved Asian countries, however, revealing the above the pover t y line. Meanwhile a greater risks faced by the vulnerable and sizable proportion of the poor and the even the middle class. vulnerable—over 9 percent of the total I n I nd i a , t he ex p e c t at ion i s t h at population, or about 11 percent of the poor households from Scheduled Castes and and vulnerable—moved into the middle Scheduled Tribes would experience less class. Downward mobility was also consid- upward mobility than the rest of the pop- erable, however. ulation. However, this is not true when The pattern is similar in Bangladesh, considering the fraction of these popula- where both upward and downward mobil- tion groups moving up in the distribu- ity were strong over the period considered. tion (figure 4.5). Between 2004 – 05 and Between 2005 and 2010, about 19 percent of 20 09 –10, their poverty rate fell from the total population, or half the poor, moved 51 percent to 44 percent, following the above the poverty line, and about 13 percent same trend as the rest of the popula- of the total population, or over 15 percent tion. Taking downward mobilit y into of the poor and vulnerable, moved into the account, about 35 percent of the poor from middle class. Scheduled Castes and Scheduled Tribes By these measures, upward mobility moved out of poverty in five years. And in Bangladesh and India was comparable 7 percent of poor and vulnerable house- to that of dynamic societies such as the holds from these disadvantaged popula- United States and Vietnam (figure 4.4). Per tion groups moved into the middle class. capita consumption is higher and poverty This figure is lower than for the rest of the incidence is lower in the United States and population because Scheduled Castes and Vietnam than in South Asian countries, Scheduled Tribes have more poor and vul- but over a comparable period, the four nerable members, thus the denominator countries saw similar fractions of the poor of the ratio is bigger. But it suggests that 128 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4.4 Upward mobility in South Asian countries is similar to that of the United States and Vietnam 60 Share of consumption group (percent) 50 40 30 20 10 0 Bangladesh India United States Vietnam Bangladesh India United States Vietnam Bangladesh India United States Vietnam Bangladesh India United States Vietnam Moving out of poverty Moving up to middle class Falling back to poverty Falling out of middle class Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The groups considered are the poor for moving out of poverty, the poor and the vulnerable for moving up to middle class, the nonpoor for falling back to poverty, and the middle class for falling out of middle class. FIGURE 4.5 Upward mobility in India is strong for the Scheduled Castes and Scheduled Tribes 50 Share of consumption group (percent) 45 40 35 30 25 20 15 10 5 0 Scheduled Other groups Scheduled Other groups Scheduled Other groups Scheduled Other groups Castes and Castes and Castes and Castes and Scheduled Scheduled Scheduled Scheduled Tribes Tribes Tribes Tribes Moving out of poverty Moving up to middle class Falling back to poverty Falling out of middle class Source: Based on data from Dang and Lanjouw 2014 for this report. Note: The groups considered are the poor for moving out of poverty, the poor and the vulnerable for moving up to middle class, the nonpoor for falling back to poverty, and the middle class for falling out of middle class. SUBSTANTIAL MOBILIT Y 129 upward mobility for these socially disad- Rural households in Bangladesh and India vantaged groups is not taking place only at have experienced strong upward and down- the bottom of the distribution. ward mobility, consistent with the national trend. But upward mobility was slightly higher and downward mobility slightly lower Nonfarm jobs drive mobility in among households primarily engaging in villages nonfarm activities (figure 4.6). Mobility is taking place even at the level In South Asia’s rural areas, many of the of villages, where most South Asians still new jobs come from the expansion of non- live, and it is largely associated with an farm activities. This change in the struc- occupational shift from farm to nonfarm ture of rural employment was underpinned activities. Casual employment in services by the modernization of agricultural pro- and construction has driven this transition duction. In Bangladesh, for instance, the in most rural areas, in parallel with higher modernization process started with the agricultural productivity. Employment adoption of new high-yielding rice variet- in manufacturing, most often casual, has ies. Land productivity increased as a result; risen rapidly in the vicinity of major urban together with higher food prices, this led centers. Modernization of agricultural pro- to an upswing in real agricultural wages, duction, growing integration with urban especially in the second half of the 2000s. economies, expanding markets, and migra- Agricultural wages grew faster in villages tion opportunities are shaking up centuries- experiencing larger increases in land pro- old social structures and offering a chance ductivity. Higher real agricultural wages, for the most disadvantaged to prosper. in turn, contributed to income growth and FIGURE 4.6 Upward mobility is substantial in rural Bangladesh and rural India 70 Share of consumption group (percent) 60 50 40 30 20 10 0 Moving Moving Falling Falling Moving Moving Falling Falling Moving Moving Falling Falling Moving Moving Falling Falling out of up to back to out of out of up to back to out of out of up to back to out of out of up to back to out of poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle class class class class class class class class Bangladesh Bangladesh India India Rural farm households Rural nonfarm households Rural farm households Rural nonfarm households Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The groups considered are the poor for moving out of poverty, the poor and the vulnerable for moving up to middle class, the nonpoor for falling back to poverty, and the middle class for falling out of middle class. 130 ADDRESSING INEQUALITY IN SOUTH ASIA growing demand for nonfood products (figure 4.7, panel b). Bhutan, Maldives, and among the rural poor. They also encouraged Sri Lanka also experienced strong expansion the adoption of labor-saving techniques. of nonfarm employment in rural areas in the For instance, machines currently do over past decade. 80 percent of tillage operations (Hossain The timing and the pace of the shift and Bayes 2009; Hossain, Sen, and Sawada from farm to nonfarm activities have varied 2014; Narayan and Zaman 2008; World across countries, however. In Bangladesh, Bank 2013a). the expansion of nonfarm employment domi- The occupational shift from farm to non- nated from the late 1980s to the first half of farm activities has been common across the the 2000s. Among rural households, the aver- region. In rural India, between 1993–94 age number of adults working in nonagricul- and 2004–05, nonfarm employment grew tural activities increased from 0.65 in 1988 more rapidly than agricultural employ- to 0.92 in 2004, whereas the number work- ment. The trend accelerated after 2004–05, ing in agricultural activities decreased from as agricultural employment started declin- 1.17 to 0.97 (Nargis and Hossain 2006). ing in absolute levels. In net terms, nonfarm The expansion of nonfarm employment has activities accounted for all the expansion of stopped since 2005, however. Returns to employment in rural areas between 2004–05 farm activities have risen ever since, largely and 2009–10 (figure 4.7, panel a). In rural because food price increases have affected Pakistan, the shift of employment from farm the dynamics between the two sectors. As to nonfarm activities has taken a strong hold for Nepal, it has seen only the nonfarm sec- since 2001. The drought in 2000–02, one tor share of rural employment increase by of the most severe weather-related shocks in about 1 percentage point over the last decade decades, partly triggered the shift, but more (World Bank 2011c, 2013a). fundamental trends consolidated it. Nonfarm Rural nonfarm employment is almost exclu- employment increased more rapidly than sively casual. In India, nearly 40 percent of farm employment in rural areas between the rural nonfarm employment was in casual 1999 and 2001, and the sectoral distribution work in 2009–10, and another 40 percent of employment remained stable afterward was in self-employment. In Bangladesh, the FIGURE 4.7 Rural India and rural Pakistan have seen a consistent expansion of nonfarm employment a. India b. Pakistan 90 60 Share of employment (percent) 80 50 Share of paid employment 70 60 40 (percent) 50 30 40 30 20 20 10 10 0 0 83 4 0 5 0 3 00 02 06 08 –9 00 –0 –1 –9 19 20 – – – –2 93 04 09 92 01 05 07 – 19 99 20 20 19 99 20 20 20 19 19 Farm Nonfarm Sources: Himanshu and others 2013 and World Bank 2010. SUBSTANTIAL MOBILIT Y 131 share of casual labor within the rural non- FIGURE 4.8 Casual rural jobs provide increasingly higher earnings farm sectors increased considerably in the first in India half of the 2000s. In rural Pakistan, casual 3.0 employment in manufacturing and construc- tion increased significantly between 2000 and 2.5 2008. The increase accounted for almost all net wage employment creation for unskilled Ratio rural workers during that period (Himanshu 2.0 and others 2013; World Bank 2010, 2011c). However, casual employment is not 1.5 inconsistent with upward economic mobil- ity. Wages of casual nonfarm workers were 1.0 30 percent to 50 percent higher than agri- 1983 1993–94 1999–2000 2004–05 2009–10 cultural wages in rural India, Nepal, and Regular over casual nonfarm wages (mean) Pakistan in the 2000s; they were 10 per- Casual nonfarm over agricultural wages (mean) cent higher in rural Bangladesh during the Source: Himanshu and others 2013. fi rst half of the 2000s (World Bank 2011c). Although regular jobs tend to pay better, the earnings gap between regular and casual India, assessed the extent of the change in nonfarm jobs has narrowed over time in rural their economic activities between 1990 and India, whereas the earnings gap between 2007. In one of the blocks, 30 percent of casual nonfarm jobs and agricultural jobs the households had shifted their occupation has increased (figure 4.8). toward nonfarm employment; in the other, The shift in the structure of employment the figure was 60 percent (Kapur and others created new earning opportunities for rural 2010). In parallel, agricultural production households. Even after controlling for other has seen changes altering traditional eco- characteristics such as education or asset nomic and social relations. By 2007, almost ownership, this shift in occupations has been no household from Scheduled Castes in identified as a major contributor to income these two rural blocks was subject to forms growth among the poor (Himanshu and of bonded labor, such as halwaha , and others 2013; Nargis and Hossain 2006). very few worked on land owned by house- Unskilled workers benefited from these trans- holds from the upper castes. Conversely, formations as well, even if they did so to a sharecropping has risen considerably by lesser extent than their skilled counterparts. households from Scheduled Castes, which Nonfarm jobs are better paid partly because are also more likely to rent in plowing ser- they are higher-skilled jobs, and nonfarm vices, a manual task usually performed by workers have higher average levels of edu- upper-caste men. cation than agricultural workers. However, The experience of Palanpur, a village the majority of industry and service jobs pay in western Uttar Pradesh that has been the more than agricultural casual labor even after focus of sustained research attention for accounting for levels of education and other several decades, could well mirror broader individual characteristics. Thus, unskilled trends across northern India. In Palanpur, workers achieved a reallocation gain even the expansion of nonfarm employment has without additional investment in education translated into upward mobility for a sig- (World Bank 2010, 2011c). nificant number of households that had pre- S ocial ly d isadva nt aged popu lat ion viously appeared mired in absolute poverty. groups have been among the beneficiaries of Economic mobility has increased among the the transformation of rural areas. A survey poor, and especially for households from of households from Scheduled Castes in two Scheduled Castes. The flip side of this greater rural blocks in the state of Uttar Pradesh, occupational change has been a clear trend 132 ADDRESSING INEQUALITY IN SOUTH ASIA toward rising income inequality in the village intention of returning to the area of origin. (Himanshu and others 2013). Because of the time spells involved, perma- The transformation of rural areas has also nent and semipermanent migrants share led to a change in the structure of female similar characteristics; importantly, they are employment. Perhaps the best-known and also relatively well captured by official cen- most-often-told story is that of Bangladesh, sus or nationwide surveys. By contrast, most where the rapid growth of the garment indus- analyses of seasonal or circular migration try created wage employment opportunities need to rely on smaller-scale surveys and for young women from the villages. But in field studies. Pakistan, too, women benefited more than In India, permanent migration accounted men from the expansion of nonfarm activi- for about 30 percent of the population in ties. The number of unskilled women work- 2001, a share that has remained roughly ing as agricultural laborers fell by a third stable since then (NSSO 2010; RGCC 2012). between 2000 and 2008, whereas the num- But rural-to-urban and urban-to-urban ber of those working in manufacturing and migration flows have increased, especially construction tripled. This shift is more sig- among men. The share of rural-to-urban nificant than that among rural unskilled migrants in total male migrants increased male workers, for whom employment in from 34 percent in 1999–2000 to 39 per- manufacturing and construction increased cent in 2007–08; over the same period, the by about 60 percent. The new employment share of urban-to-urban migrants increased opportunities have resulted in labor force from 23 percent to 25 percent (Bhagat 2010; participation by women who were less likely Kundu and Saraswati 2012; NSSO 2010; to work before. The manufacturing sector Srivastava 2011, 2012). does not require a high level of education, In Pakistan, permanent migrants account but it offers significantly higher wages com- for 12 percent to 15 percent of the popula- pared to agricultural work. Its expansion in tion 10 years of age and older. The shares rural areas particularly attracted semiedu- of rural-to-urban and urban-to-urban cated women (with class 1 to 8 education), migration within this aggregate are similar who usually participated in the labor market (Hamid 2010; Khan, Shehnaz, and Ahmed far less than those with no formal education 2000; Narayan and Zaman 2008). But these and those with class 9 or higher education figures are underestimates because they do (World Bank 2010). not take into account migration within the same district. In Bangladesh, the number of permanent migrants was estimated to Migration is a major source of be 12.7 million in 2004—about 9 percent mobility of the total population—increasing from Internal migration has helped South Asian 6.6 million in 1982 (BBS 2011). By the early households find better jobs and invest- 2000s, rural-to-urban transitions were the ment opportunities and achieve signifi- dominant form of migration, especially cant and increasing economic mobility. among men. Since then, the share of urban- Internal migration can be classified into to-urban migrants in total male migrants has two major types, depending on the length increased more rapidly (BBS 2011). In Nepal, of migrants’ stay in the area of destination: permanent migration stands at about (a) permanent and semipermanent migra- 19 percent of the population 15 years of age tion and (b) seasonal or circular migration. and older (CBS 2012). At 18 percent, the Permanent migration involves an open- figure is similar for Sri Lanka (DCS 2012). ended change in location. Semipermanent Among men, permanent migration is pri- migration—also called long-term circu- marily a strategy to fulfill aspirations for lar migration—involves fi nding a foothold employment and investment opportunities in the area of destination but with the (figure 4.9). The economic motivation is SUBSTANTIAL MOBILIT Y 133 FIGURE 4.9 Among men, permanent migration is driven by job aspirations 70 60 Share of migrants (percent) 50 40 30 20 10 0 Rural to Urban to Rural to Rural to Urban to Rural to Urban to Rural to urban urban rural urban urban urban urban rural India Nepal Pakistan Economic-related reasons Studies Other reasons Sources: Based on NSS 2007–08 for India, Labour Force Survey 2009–10 for Pakistan, and CBS 2012 for Nepal. Note: Economic-related reasons refer to employment and business. The migrants considered are adult males 15 years of age and older. Urban-to-rural migration is omitted because the sample sizes are too small. stronger among those who migrated to urban of those who were in casual employment areas. In India, economic-related reasons before migrating had become regular wage- account for about 70 percent of adult male workers. A similar pattern was observed migrants who moved from rural to urban among those who were unemployed or not in areas and 55 percent of those who moved the labor force before migration. between urban areas. The pattern is similar Evidence also suggests greater economic in Nepal and Pakistan. For women, migra- mobility among migrants than among tion is generally associated to a greater extent nonmigrants. In both India and Pakistan, with marriage and social reasons. In coun- migrants to urban areas are more likely tries such as Bangladesh, migration has also to work as wage employees (54 percent been an important response to natural haz- to 58 percent) than nonmigrants (about ards and conflicts (Afsar 2003; Marshall and 46 percent); in rural areas, they are more Rahman 2013). likely to work in self-employment than non- For India, a comparison of the occupa- migrants. In the case of Dhaka, Bangladesh, tions held before and after migration shows which attracted the lion’s share of migrants considerable upward occupational mobility. to urban areas, the unemployment rate was Regular wageworkers account for almost estimated to be 4 percent among working- 40 percent of employment among permanent age members of migrant households, half the adult male migrants, compared with about rate found among nonmigrant households 20 percent before migration. This increase (Afsar 2000, 2003). in regular wage employment is particularly Substantial upward mobility is also pronounced among rural-to-urban migrants found among migrants who ended up in (from 10 percent to 45 percent). The occu- urban slums. According to a slum survey pational transition matrix elucidates the in Delhi, many of the migrants living there composition of regular wage workers after had moved from casual jobs to regular wage migration (table 4.3). About 18 percent of jobs (Mitra 2010). Another survey for the those who were self-employed and 28 percent Indian cities of Jaipur, Ludhiana, Mathura, 134 ADDRESSING INEQUALITY IN SOUTH ASIA TABLE 4.3 Changes in employment status reveal substantial mobility among migrant men in India After permanent migration (percent) Regular Unemployed or not Before permanent migration Self-employed employee Casual labor in labor force Total Self-employed 70.6 18.1 8.7 2.6 100 Regular employee 9.9 83.8 2.6 3.7 100 Casual labor 20.3 28.5 49.6 1.6 100 Unemployed or not in labor force 17.3 34.2 14.4 34.1 100 Total 24.1 39.7 19.2 17.0 100 Source: Based on NSS 2007–08. Note: The migrants considered are adult males 15 years of age and older. The circled cells show the shares of migrants who became regular wageworkers after migration. and Ujjain reached a similar conclusion. about 30 percent of urban nonmigrants. For example, among the Jaipur workers In Nepal, 45 percent of rural-to-urban engaged as semiprofessionals at the time of adult permanent migrants and 64 percent the survey, about 19 percent had worked of urban-to-urban permanent migrants earlier in sales and trade (Gupta and Mitra have more than secondary education (CBS 2002). This pattern is also evident in other 2012). In Bangladesh, data covering 62 cities. villages between 2000 and 2008 show Permanent migrants tend to have higher that education correlates positively with economic status than nonmigrants. In permanent internal migration (Hossain, India, when the total population is clas- Sen, and Sawada 2014). sifi ed based on monthly per capita expen- In contrast, seasonal migration is more ditu re, permanent mig rants tu rn out frequent among poorer and more socially to be overrepresented at the top of the marginalized groups. In India, seasonal distribution—and underrepresented at the migrants are characterized by lower eco- bottom—relative to nonmigrants (Bhagat nomic and educational attainment than 2010; Kundu and Sarangi 2007; Srivastava the neighbors; they also tend to come from 2012). The contrast is sharper in the case households with smaller landholdings of permanent migrants who moved for (Keshri and Bhagat 2012). The rate of sea- economic-related reasons, and even more sonal migration is much higher among the so in the case of urban-to-urban permanent worse-off and among the socially margin- migrants (figure 4.10). alized populations, especially in rural areas Permanent migrants—especially those (figure 4.11). Among the poorest population migrating to urban areas—are relatively in rural areas, of 1,000 people, 52 migrated better educated than nonmigrants. In to urban areas and 22 migrated to other India, almost half the adult rural-to-urban rural areas, whereas the figures are 12 and 6, migrants have at least secondary educa- respectively, among the richest population. tion, and the fi gure is even higher among Similarly, rural residents from the Scheduled urban-to-urban migrants. In contrast, only Tribes have a much higher tendency to 30 percent of urban nonmigrants have at engage in seasonal migration to both urban least secondary education. In Pakistan, and other rural areas than other groups. 40 percent to 50 percent of adult perma- For rural-to-rural migration, village-level nent migrants to urban areas have at least surveys in Andhra Pradesh and Madhya secondary education in comparison with Pradesh confirm the general patterns. SUBSTANTIAL MOBILIT Y 135 FIGURE 4.10 Permanent migrants have higher economic status in India a. Rural-to-urban migration b. Urban-to-urban migration 60 60 50 50 Share of population Share of population group (percent) group (percent) 40 40 30 30 20 20 10 10 0 0 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th (poorest) (richest) (poorest) (richest) Consumption quintiles at place of destination Consumption quintiles at place of destination Permanent migrants Economic-related permanent migrants Urban nonmigrants Source: Based on NSS 2007–08. Note: Economic-related reasons refer to employment and business. Only adult males 15 years of age and older are included. FIGURE 4.11 Seasonal migration is more common among poor and socially disadvantaged groups in India a. By expenditure quintile b. By social group 5th (richest) Consumption quintiles at place of origin Others 4th Other Backward Class 3rd Scheduled Caste 2nd Scheduled 1st (poorest) Tribe 0 10 20 30 40 50 60 0 10 20 30 40 50 Seasonal migrants per 1,000 persons Seasonal migrants per 1,000 persons Urban to urban Urban to rural Rural to rural Rural to urban Source: Based on NSS 2007–08. Note: The population considered are adult males 15 years of age and older. These villages have hardly any permanent Seasonal migration is much more dif- migrants among Scheduled Tribes and only ficult to quantify than permanent migra- a few among Scheduled Castes, but seasonal tion, and numbers are often subject to migration is common among both groups heated debates. In India, official surveys (Deshingkar and Akter 2009; Deshingkar suggested about 13.6 million people were and others 2008). seasonal migrants; of these, 13.1 million 136 ADDRESSING INEQUALITY IN SOUTH ASIA were of working age (Keshri and Bhagat uncertainty of finding work, urban wages 2012; NSSO 2010). But a number of having increased, and dependence on con- scholars argue that these fi gures are gross tractors having declined over time. Seasonal underestimates (Deshingkar and Akter migration has reduced borrowing for con- 2009; Deshingkar and Farrington 2009; sumption, improved debt repayment capac- Deshingkar and others 2008). ity, and given migrants greater confidence Village-level surveys provide comple- and bargaining power (Deshingkar and mentary information. A village-level survey Akter 2009; Deshingkar and others 2008; in Madhya Pradesh, India, suggests that Mosse and others 2002). 52 percent of households were involved in As such, seasonal migration offers the seasonal migration, mainly to work in the Scheduled Castes and Scheduled Tribes an construction sector (Deshingkar and others escape route from social discrimination. For 2008). In 42 villages in central- western instance, in Jharkhand, migration to work India, about 65 percent of households in brick kilns has given youth the opportu- and 48 percent of the adult population nity to pursue romantic relationships away were involved in seasonal mig ration, from the social restrictions they face in the overwhelmingly for urban construction villages. Similarly, migrant workers belong- work (Mosse, Gupta, and Shah 2005). ing to Scheduled Castes and Scheduled Tribes But a large survey of 1,460 villages in 31 in Bihar reported that working outside the states fi nds that only about 58 percent of village had given them dignity and freedom the villages reported having any seasonal (Deshingkar and Akter 2009; Shah 2005, migrants (Shah 2005, 2010). 2010). This finding is consistent with the For Bangladesh, regional household findings from village surveys of Scheduled surveys offer some insights on the scale Caste households in Uttar Pradesh (Kapur of seasonal migration. A survey of 1,600 and others 2010). rural households in the northwest region of I n Bang ladesh, seasonal m ig ration Bangladesh suggests that 19 percent of total stands out as an effective strategy to reduce households and 25 percent of chronically seasonal distress. The northwest region, poor households were involved in seasonal also known as the greater Rangpur region, migration (Afsar 2003; Hossain, Khan, and is characterized by an acute lean season Seeley 2003). A more recent survey of more of agricultural activities (monga). Over a than 480,000 poor households in the same third of the poor rural households in the region suggests that about 36 percent of region use seasonal migration as a mech- these households engaged in seasonal migra- anism to cope with seasonal deprivation tion in 2006–07 (Khandker, Khalily, and caused by monga. Indeed, seasonal migra- Samad 2012). tion is associated with significant reduc- Seasonal migration is not only a cop- tions in the starvation rate and the general ing strategy for the poorest; arguably, it is food deprivation rate (Khandker, Khalily, also an accumulation strategy. According and Samad 2012). Another case in point to a village-level survey in Madhya Pradesh, is the workers of the brick-making indus- India, earnings from seasonal migration try in areas adjacent to Dhaka. Most of account for over 30 percent or the largest the workers come from the disaster-prone share of total income among households with areas of both the southern (coastal belt) at least one migrant (Deshingkar and oth- and northern (river erosion belt) parts of ers 2008). The share is even higher among the country. Most of these migrant workers socially marginalized groups. Resurveys were engaged in agriculture and nonagri- of the same villages suggest that seasonal cultural wage labor as their main occupa- migration has become more prevalent over tion before their migration. Seasonal work time. This may be the result of new oppor- significantly increases the average monthly tunities in urban areas having reduced the income of migrant workers and reduces SUBSTANTIAL MOBILIT Y 137 the extreme food deprivation rate among FIGURE 4.12 Migration provides opportunities for occupational the migrant households (Hossain, Sen, and mobility to women in India Sawada 2014). Migration is an especially important a. Rural areas avenue for upward mobility in the case of women. Special microsurveys conducted Agriculture by the Centre for Women’s Development Construction Studies in 20 Indian states shed some light on the links between female migra- Manufacturing tion and female occupational mobility Paid domestic work (Mazumdar, Neetha, and Agnihotri 2011, 2013). Results point to a higher incidence Unemployed of seasonal migration among women from Scheduled Castes and Scheduled Tribes, Unpaid family work as well as a higher incidence of permanent Others migration among women from upper castes and more affluent families. In addition, 0 10 20 30 40 50 a significant proportion of unemployed Share of female migrants (percent) or housebound women are found to enter b. Urban areas into paid employment through migration (figure 4.12). Among female migrants to urban areas, about 13 percent reported as Agriculture engaging only in unpaid family work before Construction migration while the share for this category is zero after migration. The fraction of Manufacturing female migrants to urban areas working as Paid domestic work low-skilled manufacturing workers—such as tailors or textile spinning and weaving Retail services workers—almost tripled after migration Education and health (12 percent). The fraction of them work- ing in low- and medium-skilled occupations Other services in services—such as beauticians, nurses, teachers, call center employees, and techni- Unemployed cal employees—increased from 3 percent to Unpaid family work 17 percent after migration. In Bangladesh, the scale of female rural Others migration is such that it has changed social norms. Whereas initially more men than 0 5 10 15 20 25 30 women migrated to cities, over time the Share of female migrants (percent) gender ratio became more balanced, largely Before migration After migration because of the boom in the garment indus- try since the mid-1980s. This rebalancing Source: Mazumdar, Neetha, and Agnihotri 2011. took place despite a socio-religious seclu- Note: The migrants considered are female adults 15 years of age and older. sion of women that could have reduced their freedom of movement. About 90 percent of the female workers in urban garment Deshingkar and Grimm 2005). Although sectors in the 1990s and early 2000s were a socially negative image of the garment estimated to be migrants from rural areas; workers as “fallen women” prevailed in three-quarters of them came from landless the early 1990s, these migrants effectively or very poor households (Afsar 2000, 2003; defied and redefi ned their place in society. 138 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 4.4 International migration supports upward mobility in Bangladesh and Nepal Every year Bangladesh and Nepal send abroad By now, remittances account for 12 percent and scores of migrant workers and receive substan- 25 percent of GDP, respectively, for Bangladesh tial flows of remittances in exchange. Between and Nepal. The picture is similar when seen from 2000 and 2012, 5.5 million Bangladeshis and a household perspective. In 2010, remittances 2.4 million Nepalese migrated abroad. The Gulf represented 14 percent of household income in region and other Southeast Asian countries are Bangladesh and 20 percent in Nepal. their most important destinations. These migrants International migration is behind massive typically have limited education. In Bangladesh, transitions out of poverty in both countries. the share of unskilled or semiskilled labor migrants One-fifth of the poverty reduction observed in increased from 50 percent in 2000 to 76 percent Nepal between 1996 and 2004 can be attributed in 2010. In Nepal, only 5 percent of international to remittances, and the share is presumably much migrants can be considered skilled (World Bank higher in recent years. Similarly, 18 percent of 2011b). International migrants from these two the poverty reduction experienced by Bangladesh countries also tend to be overwhelmingly male. between 2000 and 2005, or 11 percent of the Remittances have become a major source of decline in poverty observed between 2000 and household income for the two countries over the 2010, is attributable to remittances. International last decade. Flows to Bangladesh increased from migration also makes households less vulnerable US$1.9 billion in 2000 to US$14.1 billion in to downward mobility. In Bangladesh, house- 2012, while flows to Nepal surged from US$0.1 holds receiving remittances are 6 percent less billion to US$4.7 billion over the same period. likely to fall into poverty. Sources: Based on BBS 2010; BMET 2014; DOFE 2013; Inchauste and others 2012; Lokshin, Bontch-Osmolovski, and Glinskaya 2010; Raihan and others 2009; and Migra- tion and Remittances Data (database), Development Prospects Group, World Bank, Washington, DC, http://go.worldbank.org/092X1CHHD0. Although they often wear hijab in public places, they are also increasingly assertive Urban mobility is shaped by city on gender rights (Hossain, Sen, and Sawada characteristics 2014). City dynamism is another driver of the International migration is another impor- increasing mobility observed in South Asia. tant avenue for South Asian households to Mobility both across and within generations fulfi ll their aspiration for jobs and upward is greater in urban areas than in rural areas, economic mobility (box 4.4). Between 2000 despite the existence of a large urban infor- and 2012, 5.5 million Bangladesh workers mal sector. Urban wage employment, regard- migrated overseas and 2.4 million Nepalese less of whether it is formal or informal, workers migrated (BMET 2014; DOFE offers better prospects of economic mobility 2013). Migration of low-skilled workers for a massive number of households in South has become a significant feature for both Asia. Urbanization is taking place through countries. Over the last decade, interna- diversified processes, however. People move tional remittance flows to Bangladesh and to cities through migration, but cities also Nepal have been on the rise. In 2012, remit- “move” to people through the transforma- tances accounted for 12 percent of the gross tion of economic activity in formerly rural domestic product (GDP) in Bangladesh and areas. These diverse processes have cre- 25 percent in Nepal. ated cities with different characteristics. SUBSTANTIAL MOBILIT Y 139 The availability and nature of jobs vary standards (figure 4.14). In both Bangladesh across cities of different types. and India, urban households whose mem- Solid evidence indicates that cities are bers are self-employed or who work as associated with greater mobility than rural casual labor experience stronger upward areas. In India, for instance, the sons of mobility and smaller downward mobility unskilled fathers in urban areas face a lower than rural households. probability of staying in the same occupa- This dynamism, including substantive tional category than their rural counterparts; transitions into the middle class, is taking they also face a much higher probability of place despite the prevalence of informality taking on better jobs. Sons of farmers also in South Asia’s urban areas. This is because face better prospects in urban areas than in the urban informal sector includes a con- rural areas (figure 4.13). siderable number of wage jobs, and they In both Bangladesh and India, within could be the main driver of upward mobil- the same generation a larger fraction of ity (Mukhopadhyay 2011). For instance, the population manages to move above the about 18 percent of total urban employ- poverty line in rural areas than in urban ment in India is accounted for by men who areas. Conversely, a larger fraction of the are regular wageworkers in the informal population achieves middle-class status in sector and another 16 percent by men who urban areas than in rural areas (table 4.4). are casual workers in the informal sec- And downward mobility in the form of tor (fi gure 4.15). If both men and women falling below the poverty line is consider- are considered, about 57 percent of the ably smaller in urban areas than in rural informal workforce in urban areas earns areas. wages. Upward mobility is higher, and down- The sectoral distribution of informal ward mobility lower, among households employment further confirms that the infor- whose members are employed as regu- mal sector is an integral part of the urban lar wage or salaried workers. But self- economy. In India, about half the informal employment and casual employment also urban male workers can be found in man- support substantial improvements in living ufacturing, construction, and transport, FIGURE 4.13 Upward mobility is much stronger in cities than in rural areas a. Urban India b. Rural India 70 Occupations of sons (percent) 90 Occupations of sons (percent) 80 60 70 50 60 50 40 40 30 30 20 20 10 10 0 0 Unskilled Farmers Skilled/ White collar Unskilled Farmers Skilled/ White collar semiskilled jobs semiskilled jobs Occupations of fathers Occupations of fathers Downward mobility Persistence Upward mobility Source: Based on IHDS 2004–05. 140 ADDRESSING INEQUALITY IN SOUTH ASIA TABLE 4.4 Rural jobs allow people to escape poverty; urban jobs are a ticket to the middle class Five years later (percent) Poor Vulnerable Middle class Total Bangladesh Rural households 2005 Poor 23.0 17.8 2.8 43.6 Vulnerable 12.4 22.3 8.2 42.9 Middle class 1.5 6.2 5.9 13.5 Total 36.9 46.3 16.9 100.0 Urban households 2005 Poor 9.9 11.9 2.9 24.7 Vulnerable 7.9 22.5 14.1 44.5 Middle class 1.4 10.2 19.1 30.8 Total 19.2 44.6 36.1 100.0 India Rural households 2004–05 Poor 26.3 15.5 0.8 42.7 Vulnerable 9.6 28.7 7.5 45.8 Middle class 0.3 4.7 6.5 11.5 Total 36.2 48.9 14.8 100.0 Urban households 2004–05 Poor 11.2 9.4 0.7 21.3 Vulnerable 7.1 29.3 11.3 47.7 Middle class 0.4 8.8 21.8 31.0 Total 18.7 47.5 33.8 100.0 Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The household head’s age is restricted to between 25 and 55 years on the first survey and adjusted accordingly for the second survey. The circled cells show the shares of the total population who experienced upward mobility. Percentages may not total to 100 because of rounding. whereas 56 percent of the informal urban A round 60 percent of garment work- female workers are in manufacturing, con- ers were female in 2009; the share might struction, education, and health. Thus, have climbed to about 80 percent by 2012. rather than being segregated into low-pro- Because the demand for female workers ductivity activities, households in the infor- in the manufacturing sector has increased mal sector can be somewhat integrated faster than that of male workers, female with the more modern parts of the urban wages have increased more rapidly than economy. male wages in recent years (Ahmed, Bakht, In Bangladesh, the urban formal sector and Yunus 2011; Hossain, Sen, and Sawada has been expanding because of the rapid 2014; Lopez-Acevedo and Robertson 2012; growth of labor-intensive manufacturing— Narayan and Zaman 2008; World Bank in par ticular, garments and tex tiles. 2012a; Zhang and others 2013). Together, these activities account for nearly While urban areas present better pros- 75 percent of total manufacturing employ- pects of economic mobility than rural areas, ment. Although their expansion has been both the pace and the pattern of urbaniza- sustained on low labor costs, real wages tion in South Asia are reasons for concern. have grown more rapidly in manufacturing In absolute terms, the urban population industries than in agriculture. Women, and is massive, and the rate of urbanization is especially migrants from rural areas, gained impressive. India had 388 million people from the expansion of these industries. residing in urban areas at the end of the SUBSTANTIAL MOBILIT Y 141 FIGURE 4.14 Even self-employment and casual work support upward mobility in urban areas 70 60 Share of consumption group (percent) 50 40 30 20 10 0 Moving Moving Falling Falling Moving Moving Falling Falling Moving Moving Falling Falling Moving Moving Falling Falling out of up to back to out of out of up to back to out of out of up to back to out of out of up to back to out of poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle poverty middle class class class class class class class class Bangladesh Bangladesh India India Urban self-employed Urban regular wage Urban self-employed Urban regular wage and casual labor and salaried and casual labor and salaried Sources: Based on data from Dang and Lanjouw 2014 for this report and Dang, Lanjouw, and Khandker 2014 for this report. Note: The groups considered are the poor for moving out of poverty, the poor and vulnerable for moving up to middle class, the nonpoor for falling back to poverty, and the middle class for falling out of middle class. FIGURE 4.15 Many informal sector workers are wage earners in urban India a. Male employment b. Female employment 70 12 employment (percent) employment (percent) 60 10 Share of urban Share of urban 50 8 40 6 30 4 20 10 2 0 Self-employed Regular wage Casual labor Total Self-employed Regular wage Casual labor Total Self-employed Regular wage Casual labor Total Self-employed Regular wage Casual labor Total Informal sector Formal sector Informal sector Formal sector Source: Based on NSS 2009–10. Note: The workers considered are adults 15 years of age and older. 142 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4.16 South Asian countries are less urban than their peers 100 Urban share of total population (percent in 2011) 80 60 40 Pakistan Bhutan Maldives India Bangladesh Afghanistan 20 Nepal Sri Lanka 0 6 8 10 12 Log of GDP per capita (PPP, constant 2011 international dollars, 2012) Sources: Based on UN 2012 and World Bank’s World Development Indicators database, http://data.worldbank.org/data-catalog/world-development-indicators. Note: PPP = purchasing power parity. century’s fi rst decade, Pakistan 64 million, the transformation of economic activity in and Bangladesh 43 million. Overall, 563 rural areas. In India, for example, areas that million people could be considered urban are administratively rural but economically residents in South Asia in 2011, according urban are called census towns. Between 2000 to the countries’ official definitions (UN and 2011, a total of 2,553 administratively 2012). India saw its urban population grow rural areas were reclassified as census towns, at about 3 percent a year in the 2000s, dou- accounting for an estimated 30 percent of ble the rural population growth rate. And the urbanization recorded during that period Bangladesh has one of the world’s fastest- (Pradhan 2013; RGCC 2012). Satellite imag- growing urbanization rates. However, ery of contiguously built areas suggests that countries in South Asia are still less urban urban extent has expanded beyond officially than other countries at a similar level of defined urban boundaries, even when tak- development (figure 4.16). From a mobil- ing census towns into account (Denis and ity perspective, this amounts to a missed Marius-Gnanou 2011). opportunity. These diverse urbanization mechanisms Urbanization is not only slower in South have led to a range of cities with different Asia, but it also appears to be more organic. characteristics, not just in terms of their size Whereas people come to cities in the form but also in terms of their governance struc- of migration, cities also “come” to people ture. With more than 7,900 cities and towns, through the densification of population and India spans the entire urban spectrum. SUBSTANTIAL MOBILIT Y 143 The eight largest metropolitan areas have is significantly smaller. Districts with more than 4 million people. Another 38 larger cities also have a higher propor- cities have a population between 1 million tion of urban jobs in manufacturing and and 4 million, 46 cities have a population services, relative to districts with smaller between 500,000 and 1 million, and so on cities. Construction, manufacturing, min- (RGCC 2012; UN 2012). Large cities are ing, and service activities account for over governed by municipal corporations, smaller 80 percent of total employment in districts ones are governed by municipalities and vari- with cities of more than 4 million people, ous types of town councils, and some are still whereas the share is below 66 percent in administratively rural. districts whose biggest city hosts fewer than In other countries in the region, one or two 100,000 people. metropolitan areas dominate the urban hier- More disaggregated city-level analysis archy. In Bangladesh, Dhaka and Chittagong finds that the largest metropolitan areas offer account for over 48 percent of the urban more job opportunities in business services population. In contrast, over 60 percent and sophisticated manufacturing industries of other cities have populations of 50,000 than other cities do (World Bank 2013b). or less. Similarly, in Pakistan, Karachi and In particular, sophisticated manufacturing Lahore account for about 32 percent of the activities show a noticeable concentration urban population, whereas 40 percent of in the largest metropolitan areas, compared other towns have populations of 50,000 or with other urban areas. For example, only less. About 20 percent of Nepal’s urban 17 percent of total employment in these met- population resides in Kathmandu, but areas ropolitan areas is in low-tech manufacturing around Birgunj and Biratnagar, which industries, but the proportion reaches almost are close to the Indian border, are also 36 percent for high-tech industries. The dynamic. In Sri Lanka, about 28 percent of pattern is similar for business services. the total population lives in the Colombo In addition to size, the governance of cit- Metropolitan Region, whereas many small ies matters for urban economic mobility. cities cluster along the coast (Lall and Astrup In India, population censuses provide 2009; Muzzini and Aparicio 2013a, 2013b; detailed information on the administrative Narayan and Zaman 2008; UN 2012; World arrangements applying to each city or town. Bank 2012a, 2013a). Using this information, one can classify These differences in size and governance urban areas into six categories: state capi- matter for mobility because they shape the tals, other cities with municipal corporations, types of jobs available across different types municipalities, notified areas, nagar pan- of cities. In India, districts can be classified chayat (including census towns), and indus- depending on the size of their biggest city. trial townships. The first five are in a broadly This information, in turn, can be used to descending order in terms of administrative analyze the structure of urban employment autonomy, capacity, and financial resources. at the district level. Districts with larger The last category, industrial townships, cov- cities have a higher proportion of urban ers areas designated for industrial develop- regular wageworkers than do districts with ment that have some of the characteristics smaller cities (figure 4.17). For districts with of special economic zones in other countries. cities of more than 4 million people, regular Again, this information can be used to assess wage jobs account for 54 percent of urban how the composition of urban employment employment, and the share still exceeds varies across city characteristics at the dis- 40 percent for those with cities having a trict level. population between 1 million and 4 million. The share of regular wage jobs in urban In contrast, for districts whose biggest city employment broadly declines with the auton- has fewer than 100,000 people, the share of omy, capacity, and financial resources of regular wageworkers in urban employment city authorities (figure 4.18). By contrast, the 144 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 4.17 The composition of urban employment varies with city size in India a. By type of job Less than 50,000 Size of the biggest city in the district 50,000–100,000 100,000–500,000 500,000–1 million 1 million–4 million More than 4 million 0 20 40 60 80 100 Share of urban employment (percent) Regular wage Casual labor Self-employment b. By sector of activity Less than 50,000 Size of the biggest city in the district 50,000–100,000 100,000–500,000 500,000–1 million 1 million–4 million More than 4 million 0 20 40 60 80 100 Share of urban employment (percent) Agriculture Mining, manufacturing, and construction Services Other Sources: Based on NSS 2009–10 and RGCC 2012. Note: The workers considered are adults 15 years of age and older. share of self-employment increases. The com- have a smaller share of urban employment in position of employment in districts with agriculture; the same holds true for districts industrial townships resembles that of dis- in which an industrial township exists. One tricts with municipal corporations. Overall, could argue that urban governance improves districts with state capitals, municipal cor- as city size increases, thus making the porations, or industrial townships are asso- observed relationships misleading. However, ciated with significantly greater shares of the relationships hold even after controlling regular wage jobs. Similarly, districts with for the size category of the biggest city in the a state capital or a municipal corporation district. SUBSTANTIAL MOBILIT Y 145 FIGURE 4.18 The composition of urban employment also varies with city governance in India a. By type of job Governance category of cities in the district Nagar panchayat Notified area Municipality Industrial township Municipal corporation State capital 0 20 40 60 80 100 Share of urban employment (percent) Regular wage Casual labor Self-employed b. By sector of activity Governance category of cities in the district Nagar panchayat Notified area Municipality Industrial township Municipal corporation State capital 0 20 40 60 80 100 Share of urban employment (percent) Agriculture Mining, manufacturing, and construction Services Other Sources: Based on NSS 2009–10 and RGCC 2012. Note: The workers considered are adults 15 years of age and older. 146 ADDRESSING INEQUALITY IN SOUTH ASIA Substantial mobility: Main messages and policy implications Mobility is an avenue to long-term equality and classified into three groups—the poor, the vul- a source of economic efficiency. At any point in nerable, and the middle class—upward mobil- time, differences exist in household well-being, ity within a generation is considerable for both and these differences are amplified in the absence the poor and the vulnerable. A large fraction of equality of opportunity. But they can be par- of the poor moves above the poverty line while tially offset if the choices households make on some of the vulnerable make solid progress into where to live and what kind of work to do are the middle class. The upward mobility of house- more rewarding, in relative terms, for the most holds belonging to the Scheduled Castes and disadvantaged. Through migration and changes Scheduled Tribes is similar to that of the rest of in occupation, a rapidly urbanizing society offers the population. to rural households—typically the poorest in A spatial perspective sheds light on the drivers society—a chance to improve their condition. of mobility in South Asia. At the village level, Assessing the actual extent of mobility is mobility is associated with the development of demanding, because it requires data for the same nonfarm employment in rural areas and with the individuals or households at different—and ide- migration of household members to urban areas, ally distant—points in time. Few data sources of whether permanent or temporary. Migration, not this sort, including a matched father-son data set just from rural to urban areas but also between are available in South Asia. This chapter supple- urban areas, emerges from the analysis as a pow- mented those sources with the construction of erful source of mobility. But overall, mobility is synthetic panels for Bangladesh and India, allow- higher in urban than in rural areas, and this is ing the comparison of household expenditures so even though a vast majority of urban jobs in per capita with an interval of several years. These South Asia are informal. Regular wage and sala- data sources are admittedly partial, but they ried jobs, formal or informal, appear to be the yield a consistent picture. Moreover, they pro- households’ ticket to the middle class. vide information on both expenditures and occu- Recognizing the importance of mobility in pations, which is valuable because occupational adulthood—and not just of equality of opportu- change is arguably the most powerful driver of nity in childhood—has important policy impli- mobility. cations. The focus on human opportunity that Contrary to expectations, the extent of mobil- has characterized much recent work on inequal- ity in South Asia turns out to be substantial. The ity focuses policy attention on the provision of occupations held by sons are increasingly inde- basic health, education, and infrastructure ser- pendent from those their parents used to have, vices. A focus on mobility calls for attention to and the movement is in the direction of leaving issues such as jobs and urbanization. Even the unskilled jobs and farming. In India, mobil- kind of infrastructure development to emphasize ity across generations is greater for households is different in both cases. And from the mobil- belonging to the Scheduled Castes and Scheduled ity perspective, the structure of cities, hence their Tribes and to Other Backward Castes than it is capacity to support the creation of regular wage for higher-caste Hindus. When population is and salaried jobs, becomes crucial. SUBSTANTIAL MOBILIT Y 147 References ———. 2014. “Welfare Dynamics Measurement: Two Defi nitions of a Vulnerability Line and Afsar, Rita. 2000. Rural-Urban Migration in Their Applications.” Background paper for Bangladesh: Causes, Consequences, and this report and Policy Research Working Paper Challenges. Dhaka: University Press. 6944, World Bank, Washington, DC. ———. 2003. “Internal Migration and the Dang, Hai-Anh, Peter Lanjouw, and Shahidur Development Nexus: The Case of Bangladesh.” Khandker. 2014. “Poverty Dynamics in Paper presented at the Regional Conference on Bangladesh: Recent Trends and Insights from Migration, Development and Pro-Poor Policy Synthetic Panel Data.” Background paper for Choices in Asia, Dhaka, Bangladesh, June this report, World Bank, Washington, DC. 22–24. DCS (Department of Census and Statistics). Ahmed, Nazneen, Zaid Bakht, and Md. Yunus. 2012. Census of Population and Housing, 2011. “Size Structure of Manufacturing 2011. Colombo: Department of Census and I ndustr y and I mplications for Grow th Statistics, Government of Sri Lanka. and Poverty.” Bangladesh Country Paper, Denis, Eric, and Kamala Marius-Gnanou. 2011. Bangladesh Institute of Development Studies, “Toward a Better Appraisal of Urbanization Dhaka. in India: A Fresh Look at the Landscape of Akerlof, George A., and Rachel E. Kranton. 2010. Morphological Agglomerates.” Cybergeo: Identity Economics: How Our Identities European Journal of Geography [Online], Shape Our Work, Wages, and Well-Being. Systems, Modelling, Geostatistics, Document Princeton, NJ: Princeton University Press. 569. ht t p: //c ybergeo.rev ues.org / 24798. BBS (Bangladesh Bureau of Statistics). 2010. doi: 10.4000/cybergeo.24798. Re por t of the Hou sehold Income an d Deshingkar, Priya, and Shaheen Akter. 2009. Expenditure Survey 2010. Dhaka: Bangladesh “Migration and Human Development in Bureau of Statistics, Statistics Division, India.” Human Development Research Paper Ministry of Planning. HDRP 2009-13, United Nations Development ———. 2011. Report on Sample Vital Registration Programme, New York. System-2010 . Dhaka: Bangladesh Bureau Deshingkar, Priya, and John Farrington. 2009. of Statistics, Statistics Division, Ministry of Circular Migration and Multilocational Planning. Livelihood Strategies in Rural India . New Bhagat, Ram B. 2010. “Internal Migration in York and Oxford: Oxford University Press. India: Are the Underprivileged Migrating More?” Asia-Pacific Population Journal Deshingkar, Priya, and Sven Grimm. 2005. 25 (1): 27–45. Inte r n al M ig ration an d De velopm e nt: Björklund, Anders, and Markus Jäntti. 2000. A Global Pe rspective. IOM M ig ration “I ntergenerat iona l Mobi l it y of S o cio - Research Series No. 19. Geneva: International Economic Status in Comparative Perspective.” Organization for Migration. Nordic Journal of Political Economy 26 (1): Deshingkar, Priya, Pramod Sharma, Sushil 3–32. Kumar, Shaheen Akter, and John Farrington. BMET (Bureau of Manpower, Employment and 20 08. “Circular M igration in Madhya Training). 2014. Statistical Reports. Dhaka: Pradesh: Changing Patterns and Social Bureau of Manpower, Employment and Protection Needs.” European Journal of Training, Ministry of Expatriate Welfare Development Research 20 (4): 612–28. and Overseas Employment, Government of Deshpande, Ashwini. 2000. “Does Caste Still Bangladesh. Define Disparity? A Look at Inequality in CBS (Central Bureau of Statistics). 2012. National Kerala, India.” American Economic Review Population and Housing Census 2011. 90 (2): 322–25. Kathmandu: Central Bureau of Statistics, DOFE (Department of Foreign Employment). National Planning Commission Secretariat, 2013. Annual Report. Kathmandu: Ministry Government of Nepal. of Labour Employment, Department of Dang, Hai-Anh, and Peter Lanjouw. 2013. Foreign Employment, Government of Nepal. “Measuring Poverty Dynamics with Synthetic Erikson, Robert, and John H. Goldthorpe. 1992. Panels Based on Cross-Sections.” Policy The Constant Flux: A Study of Class Mobility Research Working Paper 6504, World Bank, in Industrial Societies. Oxford, U.K.: Oxford Washington, DC. University Press. 148 ADDRESSING INEQUALITY IN SOUTH ASIA Ferreira, Francisco H. G., Julian Messina, Jamele India.” Journal of Human Resources 48 (2): Rigolini, Luis-Felipe López-Calva, Maria 435–73. Ana Lugo, and Renos Vakis. 2012. Economic Hossain, Mahabub, and Abdul Bayes. 2009. Mobility and the Rise of the Latin American Rural Economy and Livelihoods: Insights Middle Class. Washington, DC: World Bank. from Bangladesh. Dhaka: A H Development Ferrie, Joseph P. 2005. “History Lessons: The Publishing House. End of American Exceptionalism? Mobility H o s s a i n , M a h a b u b , B i n ay a k S e n , a n d in the United States since 1850.” Journal of Yasuyuki Sawada. 2014. “Jobs, Growth Economic Perspectives 19 (3): 199–215. and Development: Making of the ‘Other’ Fields, Gary S. 2008. “A Brief Review of the Bangladesh.” Background Paper for the World Literature on Earnings Mobility in Developing Development Report 2013 , World Bank, C ou nt r ie s .” Work i ng Pap er s , C or nel l Washington, DC. University, ILR School site. http://digitalcom- Hossain, Munshi I., Iqbal A. Khan, and Janet mons.ilr.cornell.edu/workingpapers/101. Seeley. 2003. “Surviving on Their Feet: ———. 2010. “Does Income Mobility Equalize Charting the Mobile Livelihoods of the Poor Longer-Term Incomes? New Measures of an in Rural Bangladesh.” Paper presented at the Old Concept.” Journal of Economic Inequality conference Staying Poor: Chronic Poverty and 8 (4): 409–27. Development Policy, Institute for Development Fields, Gary S., and Efe A. Ok. 1996. “The Policy and Management, Universit y of Meaning and Measu rement of I ncome Manchester, U.K., April 7–9. Mobility.” Journal of Economic Theory 71 (2): Inchauste, Gabriela, Sergio Olivieri, Jaime 349–77. Saavedra, and Hernan Winkler. 2012 . ———. 2000. “Income Mobility: Concepts “What Is Behind the Decline in Poverty since and Measures.” In New M arkets, New 2000? Evidence from Bangladesh, Peru and Opportunities? Economic and Social Mobility Thailand.” Policy Research Working Paper in a Changing World, edited by Nancy Birdsall 6199, World Bank, Washington, DC. and Carol Graham, 101–33. Washington, Kapur, Devesh, Chandra Bhan Prasad, Lant DC: Brookings Institution and Carnegie Pr itchet t , a nd D. Shya m B abu. 2010. Endowment Press. “Rethinking Inequality: Dalits in Uttar Ganzeboom, Harry B. G., and Donald J. Pradesh in the Market Reform Era.” Economic Treiman. 1996. “Internationally Comparable and Political Weekly 45 (35): 39–49. Measures of Occupational Status for the Keshri, Kunal, and Ram B. Bhagat. 2012. 1988 International Standard Classification of “Temporary and Seasonal Migration: Regional Occupations.” Social Science Research 25 (3): Pattern, Characteristics and Associated 201–39. Factors.” Economic and Political Weekly 47 Giddens, Anthony. 2009. Sociology. 6th ed. (4): 81–88. Cambridge, U.K.: Polity Press. Khan, Aliya H., Lubna Shehnaz, and Ather Gupta, Indrani, and Arup Mitra. 2002. “Rural Maqsood Ahmed. 2000. “Determinants of Migrants and Labour Segmentation: Micro- Internal Migration in Pakistan: Evidence Level Evidence from Delhi Slums.” Economic from the Labour Force Survey, 1996–97.” and Political Weekly 37 (2): 163–68. Pakistan Development Review 39 (4) Part II: Hamid, Shahnaz. 2010. “Rural to Urban 695–712. M i g r a t i o n i n Pa k i s t a n : T h e G e n d e r Khandker, Shahidur R., M. A. Baqui Khalily, and Perspective.” Working Papers and Research Hussain A. Samad. 2012. “Seasonal Migration R e p o r t s 2 010 , Pa k i s t a n I n s t i t u t e of to Mitigate Income Seasonality: Evidence from Development Economics, Islamabad. Bangladesh.” Journal of Development Studies Himanshu, Peter Lanjouw, Rinku Murgai, 48 (8): 1063–83. and Nicholas Stern. 2013. “Non-farm Kundu, Amitabh, and Niranjan Sarangi. 2007. Diversification, Poverty, Economic Mobility “Migration, Employment Status and Poverty: and Income Inequality: A Case Study in An Analysis across Urban Centres.” Economic Village India.” Policy Research Working Paper and Political Weekly 47 (26): 219–27. 6451, World Bank, Washington DC. Kundu, Amitabh, and Lopamudra Ray Saraswati. Hnatkovska, Viktoria, Amartya Lahiri, and 2 01 2 . “M i g r a t i o n a n d E x c lu s i o n a r y S ou rabh B . Pau l. 2013. “Brea k i ng t he Urbanisation in I ndia.” Economic and Caste Barrier Intergenerational Mobility in Political Weekly 42 (4): 299–306. SUBSTANTIAL MOBILIT Y 149 Lall, Somik V., and Claus Astrup. 20 09. S t a t u s f r o m a n U r b a n Pe r s p e c t i v e . Sri Lanka: Reshaping Economic Geography: Washington, DC: World Bank. Connecting People to Prosperity. Washington, ———. 2013b. Urban Growth and Spatial DC: World Bank. Transition in Nepal: An Initial Assessment. Lokshin, Michael, Mikhail Bontch-Osmolovski, Washington, DC: World Bank. and Elena Glinskaya. 2010. “Work-Related Narayan, Ambar, and Hassan Zaman. 2008. Migration and Poverty Reduction in Nepal.” Poverty Assessment for Bangladesh: Creating Review of Development Economics 14 (2): Opportunities and Bridging the East-West 323–32. Divide. Bangladesh Development Series Paper Long, Jason, and Joseph Ferrie. 2007. “The Path to no. 26. Washington, DC: World Bank. Convergence: Intergenerational Occupational Nargis, Nigar, and Mahabub Hossain. 2006. Mobility in Britain and the US in Three Eras.” Income Dynamics and Pathways out of Economic Journal 117 (519): C61–71. Rural Poverty in Bangladesh, 1988–2004.” ———. 2013. “Intergenerational Occupational Agricultural Economics 35 (Suppl. s3): Mobility in Great Britain and the United 425–35. States since 1850.” American Economic NSSO (National Sample Survey Office). 2010. Review 103 (4): 1109–37. Migration in India 2007–2008: NSS 64th L o p e z - A c e ve d o , G l a d y s , a n d R ay m o n d Round (July 2007–June 2008). NSS Report Robertson, eds. 2012. Sewing Success? No. 533 (64/10.2/2). New Delhi: National Employment, Wages, and Poverty Following Sample Survey Office, Ministry of Statistics the End of the Multi-fibre Arrangement . and Programme Implementation, Government Washington, DC: World Bank. of India. Marshall, Richard, and Shibaab Rahman. 2013. Pradhan, Kanhu Charan. 2013. “Unacknowledged “Internal Migration in Bangladesh: Character, Urbanisation: The New Census Towns of Drivers and Policy Issues.” Paper by the India.” Economic and Political Weekly United Nations Development Programme, 47 (36): 43–51. Bangladesh. Raihan, Selim, Bazlul H. Khondker, Guntur Mazumdar, Indrani, N. Neetha, and Indu Sugiyarto, and Shikha Jha. 2009. “Remittances Agnihotri. 2011. “Gender and Migration and Household Welfare: A Case Study of in India.” National Workshop on Internal Bangladesh.” Working Paper Series No. 189, Migration and Human Development in Asian Development Bank, Manila, Philippines. India: Workshop Compendium, Vol. II. New RG C C ( R e g i s t r a r G e n e r a l a n d C e n s u s Delhi: UNESCO and UNICEF. Commissioner). 2012. Census of India, 2011. ———. 2013. “Migration and Gender in India.” New Delhi: Registrar General and Census Economic and Political Weekly 48 (10): Commissioner, India, under Ministry of Home 54–64. Affairs, Government of India. Mitra, Arup. 2010. “Migration, Livelihood and Salverda, Wiemer, Brian Nolan, and Timothy M. Well-Being: Evidence from Indian City Slums.” Smeeding, eds. 2011. The Oxford Handbook Urban Studies 47 (7): 1371–90. of Economic Inequality. Oxford, U.K.: Oxford Mosse, David, Sanjeev Gupta, Mona Mehta, University Press. Vidya Shah, Julia Rees, and the KRIBP Project Schumpeter, Joseph Alois. 1955. Imperialism Team. 2002. “Brokered Livelihoods: Debt, [and] Social Classes: Two Essays. New York: Labour Migration and Development in Tribal Meridian. Western India.” Journal of Development Shah, Amita. 2005. “Land Degradation and Studies 38 (5): 59–88. Migration in a Dry Land Region in India.” Mosse, David, Sanjeev Gupta, and Vidya Shah. SANDEE Working Paper No. 10, South Asian 2005. “On the Margins in the City: Adivasi Network for Development and Environmental Seasonal Labour Migration in Western India.” Economics, Kathmandu, Nepal. Economic and Political Weekly 40 (28): ———. 2010. “Migration as an Exit Route: How 3025–38. Does It Work for the Chronic Poor in India?” Mukhopadhyay, Partha. 2011. “Formality and Paper presented at International Conference on Functionality in Indian Cities, Op-Ed.” Ten Years of War against Poverty: What Have Seminar Web edition (January), New Delhi. We Learned since 2000; What Should We Do Muzzini, Elisa, and Gabriela Aparicio. 2013a. 2010 –2020, Manchester, U.K., September Bangladesh: The Path to Middle-Income 8–10. 150 ADDRESSING INEQUALITY IN SOUTH ASIA Singh, Ashish, and Sripad Motiram. 2012. “How ———. 2011c. More and Better Jobs in South Close Does the Apple Fall to the Tree?” Asia. Washington, DC: World Bank. Economic and Political Weekly 47 (40): 56–65. ———. 2012a. Bangladesh: Towards Accelerated, Srivastava, Ravi. 2011. “Labour Migration in Inclusive and S ust ain able Growth: India: Recent Trends, Patterns and Policy Opportunities and Challenges. Washington, Issues.” Indian Journal of Labour Economics DC: World Bank. 54 (3): 411–40. ———. 2012b. World Development Report 2013: ———. 2012. “Internal Migration in India: An Jobs. Washington, DC: World Bank. Overview of its Features, Trends, and Policy ———. 2013a. Bangladesh Poverty Assessment: Challenges.” National Workshop on Internal Assessing a Decade of Progress in Reducing Migration and Human Development in India: Poverty, 2000–2010. Washington, DC: World Workshop Compendium, Vol. II. New Delhi: Bank. UNESCO and UNICEF. ———. 2013b. Urbanization beyond Municipal UN (United Nations Department of Economic B o u n d a r i e s : N u r t u r i ng M e t ro po l i t a n and Social Affairs/Population Division). 2012. Economies and Connecting Peri-Urban Areas World Urbanization Prospects: The 2011 in India. Washington, DC: World Bank. Revision. New York: United Nations. Zhang, Xiaobo, Shahidur Rashid, Kaikaus Wo rl d B a n k . 2 010 . “ Pa k i s t a n Pov e r t y Ahmad, Valerie Mueller, Hak Lim Lee, Assessment.” World Bank, Washington, DC. Solomon Lemma, Saika Belal, and Akhter Unpublished. Ahmed. 2013. “Rising Wages in Bangladesh.” ———. 2011b. L arge- Scale Migration and IFPRI Discussion Paper 1249, International Remittance in Nepal: Issues, Challenges, Food Policy Research Institute, Washington, and Opportunities. Report No. 55390-NP. DC. Washington, DC: World Bank. Inadequate Support 5 H ouseholds in every country suffer innate characteristics is often amplified by from shocks. Most are minor and the legacy of inequities accumulating through can be cushioned relatively easily, but life. Parents’ support for their less fortunate some can have long-lasting impacts, adversely offspring can go some way toward redressing affecting nutrition, human capital, and asset disparities within a household, whereas char- accumulation. The vulnerability of house- ity inspired by the generosity of those who holds to shocks is determined partly by the do well can reduce disparities more broadly. types of risks they face. For instance, urban But private transfers are unlikely to be suffi- and rural populations are exposed in dif- cient to offset the consequences of these other ferent degrees to the same natural disasters. forms of randomness. The vulnerability of households also depends To what extent inequality of outcomes on their own ability to manage those risks. resulting from these different forms of bad Typically, the poor have a relatively weak luck is tolerable—or even desirable—is a capacity to self-insure or to pool risks beyond question to which different societies give extended families. Informal mechanisms tend different answers. But most—if not all— to be costly and inefficient, often breaking have developed mechanisms for redress down when shocks affect entire communities. (box 5.1). Bad luck can hit even before birth. Government-sponsored mechanisms to Individuals differ in their inherited wealth, in cope with shocks are known as social pro- their talent, in the value they attach to future tection; they typically comprise social assis- well-being relative to present well-being, and tance and social insurance programs. They in their willingness to work hard. Even chil- are designed to prevent households that are dren of the same parents can differ in these affected by an adverse shock from expe- respects. In an ideal society with equality riencing too dramatic a fall in their living of opportunity and perfect mobility, these standards. Mechanisms to redress deeper inherited differences would result in inequal- differences in the fate of individuals and ity of outcomes. Given that no society fully households fall under the broader heading meets the ideal, the inequality of outcomes of redistribution; the instruments used in caused by inherited wealth, talent, and other this case are taxes, subsidies, and transfers. 151 152 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 5.1 Support aims at offsetting inequality of outcomes stemming from bad luck Consider a society with equality of opportunity Final and perfect mobility. In such society, inequality well-being of outcomes is still possible because individuals differ in their inherited wealth, their talent, their foresightedness, or their aversion to effort. ET Impact on inequality In the absence of any government interven- tion, differences in well-being are unlikely to be resorbed as people age. Imagine now that all Social eT protection, individuals experience an adverse shock at some subsidies point in their adult life, but the magnitude of the shock varies. In the figure, for example, poorer Shocks, taxes individuals are affected to a greater extent—in Initial relative terms—than richer ones. 45° earnings The level of final well-being crucially depends e0 E0 Initial on the support received from the government. If Poorer Richer well-being social protection programs are indeed targeted household household to the poor, and if taxes and subsidies are pro- gressive, then the actual inequality of outcomes could be even lower than the initial one. In most can be taken in different directions, depending countries, after-tax inequality of outcomes is on their talent and the shocks they experience. indeed lower than pretax inequality. And government support, although partially mit- However, as the figure shows, some inequality igating the effects of this randomness, is unlikely of outcomes would still occur. People who had to fully offset it. The policy challenge is not the same opportunity in childhood and the same how to ensure equality of outcomes but rather chances to prosper through mobility in adult life how to provide support in an efficient manner. This type of support aims at bringing dis- well targeted to poorer households and has advantaged households to a higher level of helped them become more resilient to natu- welfare than they could attain on their own. ral disasters such as floods and earthquakes. The reach, generosity, and efficiency of social Similarly, the Mahatma Gandhi National protection and redistribution mechanisms Rural Employment Guarantee (MGNREG) vary from one society to another and, as a Act of India—the largest public works result, the extent of inequality of outcomes program in the world—has dramatically varies as well. reduced distress sales of land in years of South Asian countries have a mixed drought. record on both fronts. Public spending on But the social protection programs of social protection programs has expanded the region also have distinct features that over time. Measured as a fraction of GDP, make them less than ideal. Compared with it is in line with other countries at roughly other regions, a much higher proportion of the same development level. The design resources goes into direct subsidies for house- of these programs also shows important holds to purchase items such as food, fuel, strengths. Thus, the Benazir Income Support and electricity. The justification for some of Programme (BISP) of Pakistan is remarkably these subsidies may be conceptually strong, as INADEQUATE SUPPORT 153 in the case of food, where the aim is to reduce has a high population density; 23.8 per- the risk of hunger, but weak implementa- cent of world’s population lives on 3.6 tion results in high leakage and substantial percent of the world’s land—and approxi- waste. Subsidies for fuel and electricity tend mately 68 percent of the region’s popula- to be regressive in that they benefit better- tion lives in poverty. South Asia’s economy off households. Because of these features, is strongly oriented toward agriculture, social protection programs in the region are which makes countries more vulnerable less well geared to reducing inequality than to natural disasters and weather-related they could be. Their coverage is partial, their shocks. Indeed, roughly 70 percent of targeting is generally poor, and the amount the South Asian population lives in rural of resources they make available to those areas and about half relies on agriculture who need them the most is often too modest. for its livelihood. In addition, the region The record is arguably more questionable experiences frequent conflicts, including in the case of redistribution. One of South guerrilla warfare, riots, and antigovern- Asia’s salient characteristics is the low level ment demonstrations. of its tax revenue relative to GDP, which Health shocks affect households and com- implies that the potential for the government munities through births, deaths, disability, to make a dent in inequality of outcomes is illness, accidents, and other adverse indi- more limited than elsewhere. Tax revenue vidual health events. Shocks emanating from is low not because South Asian countries noncommunicable diseases such as cardio- rely on unusual tax instruments but rather vascular problems, cancer, diabetes, and the because the revenue “productivity” of those like affect individual households, whereas instruments is unusually low. Tax avoid- shocks from communicable diseases such as ance, tax evasion, and the underreporting tuberculosis, HIV and AIDS, hepatitis, and of taxable amounts are widespread, as are the like threaten entire communities. In addi- exemptions and special regimes favoring tion to the impact on health, shocks of this the businesses with more clout. All of this sort often result in catastrophic household results in a tax system of relatively low expenditures, thus undermining the well- progressivity. being of all household members. A positive note comes from spatial transfers Many natural disasters, such as tidal to poorer administrative units, such as prov- waves and earthquakes, occur with little inces and districts. Several countries in the warning and have widespread, devastat- region are large enough for these transfers to ing effects. Other weather-related events be a very important component of local bud- are recurrent annual or periodic prob- gets. Transfers designed on a spatial basis also lems. More than 900 disasters have been have the advantage of not directly distorting reported in South Asia since 1970, with the incentives faced by households and firms. floods constituting almost half the events. Transfers of this sort are clearly progressive in Floods and changes in traditional agricul- South Asia, but they are still modest in rela- tural patterns can severely affect house- tion to the needs of poorer areas. holds. An example is the Rangpur region of Bangladesh, where hunger is prevalent during the monga season (Khandker 2012; Shocks and how households cope World Bank 2013a). South Asian households are periodically Natural disasters affected more than exposed both to individual shocks and 750 million people in the region between to economy-wide shocks, such as natural 1990 and 2008, resulting in approximately disasters, food price spikes, and armed 230,000 deaths and US$45 billion in dam- conflict. The region’s frequent exposure ages (World Bank 2009b). Bangladesh stands to shocks relates to geographic, histori- out as one of the countries that have expe- cal, and socioeconomic factors. South Asia rienced the largest losses of human lives 154 ADDRESSING INEQUALITY IN SOUTH ASIA TABLE 5.1 Bangladesh suffered 2 of the 10 most deadly natural disasters of recent times Victims Date Type of disaster Area 300,000 1970 Drought Ethiopia 300,000 1970 Storm and flood catastrophe Bangladesh 255,000 1976 Earthquake China 220,000 2004 Earthquake, tsunami in Indian Ocean Indonesia, Thailand 150,000 1983 Drought Sudan 138,000 1991 Tropical cyclone Gorky Bangladesh 133,655 2008 Cyclone Nargis Myanmar 87,476 2008 Earthquake China, India, Pakistan 73,300 2005 Earthquake Afghanistan 66,000 1970 Earthquake Peru Source: Based on EM-DAT, the International Disaster Database, Centre for Research on the Epidemiology of Disasters, Brussels, Belgium, http://www.emdat.be/. worldwide (table 5.1). Moreover, the fre- and Bangladesh (approximately 15 percent), quency and magnitude of natural disasters to very high in Pakistan (approximately are increasing, mainly because of climate 20 percent). Sri Lanka and Afghanistan expe- change and accelerated snow melting in the rienced food price inflation of more than Himalayas (Memon 2012). 30 percent. Economic crises that start in industrial South Asia is also prone to conflict and countries can quickly affect developing violence. Over the past decade, the number countries in today’s interconnected global of terrorist incidents increased significantly economy. Compared with the rest of the in South Asia (Global Terrorism Database world, the effect of the most recent global 2009–14). The region accounted for more financial crisis on South Asia was relatively than 30 percent of the world’s terrorist inci- modest. But Bangladesh was hit especially dents between 2004 and 2008 and at least hard by the reduction in global demand 40 percent in 2009 and 2010. for apparel, which is the country’s leading A sizable number of households in parts industry. Another important transmission of South Asia are directly affected by conflict. channel for economic shocks is remittances. For example, in a World Food Programme In relative terms, remittances are particularly survey carried out in parts of Khyber important for Nepal, Bangladesh, and Sri Pakhtunkhwa and federally administered Lanka; in 2009 they accounted for 23, 12, tribal areas in Pakistan during 2010, 16 and 8 percent of GDP, respectively. The percent of households reported high levels global crisis reduced the average growth rate of insecurity from strict rules imposed by of remittances to South Asia from 33 percent militants, curfews or bans on freedom of in 2008 to less than 6 percent in 2009 movement, military operation against mili- (Migration and Remittances Data, World tants, displacement, and property damage Bank n.d.). (Pop 2010). Food price inflation has been the main Households are affected differently by driver of headline inflation throughout most these various kinds of shocks, however of South Asia. The effects of the widespread (Glewwe and Hall 1998; Heltberg and Lund food price inflation of 2007–08 varied signifi- 2009). In Pakistan, almost two-thirds of cantly among countries in the region. Inflation respondents had experienced one or more ranged from relatively moderate in India shocks over the three-year study period. (approximately 7 percent) to high in Nepal Slightly more than half the shocks were INADEQUATE SUPPORT 155 related to health events, including disease, FIGURE 5.1 Health-related events and disasters are the most death, accident, and disability (figure 5.1). common shocks in Pakistan Interestingly, the wealthiest quintile was Share of population experiencing the most likely to face a large decrease in 30 each type of shock (percent) consumption. Studies focusing on specific shocks shed 20 additional light on who is most affected. For instance, natural disasters significantly affect poor children who cannot con- 10 tinue their education because of lost fam- ily income and housing as well as delays in 0 matriculation (UN/ISDR 2008). In India, 1st 2nd 3rd 4th 5th natural disasters are strongly correlated (poorest) (richest) with more working hours devoted to non- Consumption quintile farm activities at lower wages, especially Health Others Disaster Economic when child labor is involved (Kochar 1995, Sources: Based on PSLM 2007–08 to 2009–10. 1999; Rosenzweig and Stark 1989). As a consequence of the destructive 1988 floods in Bangladesh, the nutrition of children from (Arimond and Ruel 2004). Micronutrient landless households in affected areas signifi- deficiencies and early childhood malnutri- cantly worsened (Foster 1995). Within the tion, in particular chronic malnutrition, have flood-exposed households in Bangladesh, been linked to impaired cognitive develop- the percentage of stunted children in house- ment, physical and mental disabilities, child holds at the bottom 40th percentile of the and maternal deaths, and lower productivity consumption distribution remained much (Alderman, Hoddinott, and Kinsey 2006; higher than that for households in the Glewwe and Jacoby 1995; Glewwe and King top 20th percentile a year after the flood 2001; Micronutrient Initiative and others (Del Ninno and Lundberg 2005). 2009). Countries adjust to economic shocks It seems natural to expect conflict and in different ways, with potentially differ- violence to negatively influence all house- ent implications for workers. One study holds within the affected regions, but even found that adjustments for 41 developing then, wealth can make a difference. In two countries occurred through the quality of conflict-affected regions of Pakistan, poor jobs rather than through the number of households were more likely to incur debts, jobs (Khanna, Newhouse, and Paci 2010). to sustain house damage during fighting, to Growth in earnings slowed much more be displaced, to become victims of crime and than employment growth, unemployment theft, and to be affected by drought than changed little, and hourly wage growth wealthier households (Pop 2010). Because of did not collapse. Adjustments came mainly attacks by militants, many households did through large reductions in the number of not send children to school, especially their hours worked. girls, suggesting a gender effect. In Nepal, the Because poorer households in the region conflict adversely affected female education, spend a considerable amount of their house- but overall it did not significantly reduce the hold budget on food (half of it, on average, number of years of education for either sex in 2011), even minor changes in food prices (Valente 2011). can profoundly affect their well-being. Households can mitigate the effects of Short-term declines in food security can shocks by reducing consumption, altering have serious implications. In particular, low expenditure patterns, and changing the allo- levels of dietary diversity have been statisti- cation of their time (Frankenberg, Smith, cally linked to inadequate nutrient intake and Thomas 2003). Their strategies include 156 ADDRESSING INEQUALITY IN SOUTH ASIA depleting savings and selling physical assets Employment diversification within house- (Deaton 1990), increasing labor force par- holds is a usual way to mitigate risk, akin in ticipation, finding new jobs or producing spirit to portfolio diversification in fi nance. goods for home consumption, accessing Employment diversification includes having interhousehold transfers (Cox 1987), bor- household members who take other jobs rowing, and altering consumption patterns in the same location or who migrate for (World Bank 1990). But coping strategies jobs elsewhere (Banerjee and Duflo 2007). can lead to persistent negative effects on In India, rural areas with higher rainfall human capital, such as removing children variability or those affected by the severe from school and putting them to work flood of 1998 exhibited a significantly (Jacoby and Skoufi as 1997; Kochar 1995; higher level of intrahousehold employ- Moser 1998). ment diversification. But this excessive In Pakistan, shocks negatively affected diversification negatively affected house- household consumption regardless of hold welfare and increased the probability the indicator considered and regardless of of falling into poverty (Bandyopadhyay and the number of shocks experienced. But the Skoufias 2012). impact was not always statistically signifi- South Asian countries exhibit high rates cant. Among households that reported natu- of informal sector employment (World ral disasters, rural households were more Bank 2012c). That could seem an advan- likely to experience a drop in consump- tage in dealing with economic shocks, tion than those in urban areas (figure 5.2). because flexibility is higher in the informal Conversely, among households that reported sector. Indeed, the formal sector tends to economic shocks, those in urban areas were adjust more through the number of jobs, likely to experience a drop in consumption, whereas in the informal sector the bur- which was not the case in rural areas. Health den of the adjustment falls on earnings, shocks are also significantly related to a drop with fewer people being left totally out in household consumption in urban areas. of employment. In the case of aggregate FIGURE 5.2 Disasters in Pakistan affect rural populations much more than urban populations a. Urban b. Rural 30 30 Share of population experiencing Share of population experiencing each type of shock (percent) each type of shock (percent) 20 20 10 10 0 0 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th (poorest) (richest) (poorest) (richest) Consumption quintile Consumption quintile Health Others Disaster Economic Source: Based on PSLM 2007–08 to 2009–10. INADEQUATE SUPPORT 157 shocks, however, self-employment does not purchasing power and relative price changes always provide a safety net, and the num- (D’Souza and Jolliffe 2010). ber of hours of work may fall short of the Recent data on the whole set of cop- desired level (McKenzie 2004). ing mechanisms are available for Pakistan Without formal social safety nets and (figure 5.3). When faced with a shock, a with limited or no access to formal credit large majority of households in the poor- and insurance markets, households in est quintile borrow money, reduced expen- the developing world often resort to self- ditures, switched to lower-quality food, or help and mutual insurance mechanisms. reduced the quantity of food they consumed. Self-insurance strategies include accumu- Equally important, 11.5 percent of the poor- lating assets—stored grain or small and est households reported selling agricultural large livestock in rural areas or jewelry assets to cope with the shock—a strategy and durables in urban areas—that can be that compromises their long-term ability to sold during hard times. Informal insur- earn an income. In contrast, the wealthi- ance mechanisms comprise local borrow- est groups are much less likely to use these ing schemes with friends and neighbors mechanisms. as well as private transfers originating T he f ull welfare impact of shocks from relatives. The most common infor- depends on whether they are individual or mal mechanisms of lending in the rural aggregate. Food price fluctuations are a economy are informal group loans, micro- case in point. For an individual household finance programs, and rotating savings that is a net buyer of food, the individual and credit associations. Borrowing can shock is clearly negative. But higher prices be costly in imperfect markets, especially also encourage food production and tend to for poor households (Banerjee and Dufl o result in a higher demand for rural labor. 2007; Behrman 1988). Mitigating risks Some of the poorest households work as through production choices can be costly, rural laborers, so they benefit from this too, because expected profits must be sac- indirect effect, and the net outcome may be rificed to achieve a lower risk (Morduch positive in their case. 1995). The significant price fluctuations that Poor households in some South Asian took place during 2000–10 make measure- countries also rely significantly on remit- ment of the impact of food price infl ation tances. For example, more than 50 percent on the poor possible. In Bangladesh, the of the poorest two deciles of the population poor bore the brunt of higher food prices in Bangladesh report receiving remittances, in the short term. Once wages adjusted to for an average amount corresponding to the price shock, however, the impact was 142 percent of their pretransfer consump- largely equalized across population groups. tion. Similarly, about 50 percent of Pakistani In the longer term, as the price shock perme- households in the poorest two deciles report ated other sectors, the negative short-term receiving remittances, the amount received impact on the poor was reversed. Because corresponding to 76 percent of their average Bangladesh is still to a large extent an consumption. agrarian country, in the end all groups ben- Households have developed i n for- efited from higher food prices (Jacoby and mal strategies to cope with food price Dasgupta 2012). infl ation. In Afghanistan, poor households Regardless of the origin of the shock, the demonstrate large negative price elastici- impact on household well-being depends ties of food consumption but much smaller on the nature of the social protection price elasticities of caloric intake, suggesting programs in place. Health shocks illus- that households trade quality of food for trate the trade-offs (figure 5.4). On one quantity of calories in response to declining hand, in countries without well-developed 158 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 5.3 In Pakistan, poorer and richer households cope with shocks in different ways 80 70 Share of the population (percent) 60 50 40 30 20 10 0 e n s s s ey d d d d re av et ce tio oo oo an lan on itu ss ts an ac ll l yf ff la m no nd itt nt yo no Se lit ra w Re em pe Do ua ltu rro tit ke ex er -q an icu Ta Bo er ce or qu gr ow m du ll a ce ive Re ol du Se ht ce Re Re itc Sw Consumption quintiles 1st (poorest) 2nd 3rd 4th 5th (richest) Sources: Based on PSLM 2007–08 to 2009–10. FIGURE 5.4 Spending on health is mainly out of households’ pockets 90 80 Out-of-pocket expenditure (percent of 70 total health expenditure) 60 50 40 30 20 10 0 an n a ia h ka an s l ld ive di a pa es As an ist ist ut or In ad Ne ald h an Bh k W iL Pa ut l ng Sr M gh So Ba Af 2000 2005 2012 Source: Based on World Bank Health Nutrition and Population Statistics database, http://datatopics.worldbank.org/hnp/. INADEQUATE SUPPORT 159 social protection architecture, such as FIGURE 5.5 Spending on social protection in South Asia is lower Afghanistan and India, catastrophic health than in other developing countries shocks confront households with massive out-of-pocket expenditures. On the other hand, in Maldives and especially Bhutan, Afghanistan the public sector bears most of the cost through the corresponding social security India agencies. Bangladesh Bhutan A scorecard for social protection programs Pakistan S o c i a l prot e c t ion help s i nd iv idu a l s , Nepal households, and communities create assets, cope with risk and volatility, and cushion Maldives the impact of crisis, structural transforma- Sri Lanka tions, and other shocks. By targeting the poor and the most vulnerable, spending 0 1 2 3 on social protection programs should be Spending on social protection (percent of GDP) progressive and thus reduce inequality. By South Asian countries Developing World (mean) reducing the variability of earnings, expen- countries (mean) ditures, and well-being more generally, social protection programs also cushion Source: Based on World Bank 2014. individual household transitions for a given level of aggregate inequality. However, the impact of social protection on inequality BISP, introduced in 2008, helped increase can be undermined by low coverage, poor spending on social safety net programs targeting, administrative leakage, and from 0.3 percent of GDP in 2003– 04 to inadequacy of benefits. over 0.7 percent of GDP in 2010. On average, developing countries spend When assessed on the basis of abso- 1.53 percent of their GDP on safety nets, lute spending per person, social protection and with the exception of Sri Lanka, programs are progressive in Bangladesh, most South Asian countries fall within Pakistan, and Sri Lanka (figure 5.6). t he 0. 25 percent to 2 percent ra nge Bangladesh is the most effective in directing (figure 5.5). This, in itself, reflects the social protection resources to the poor. In recent strides countries in the region have 2010, the poorest 40 percent of households made. Bangladesh provides a clear exam- received a little over half the social assistance ple (box 5.2). Among its public safety net and over 75 percent of the social insurance, programs, the Female Secondary Schools whereas the wealthiest 20 percent of house- Stipends Program—the precursor of mod- holds received about 12 percent of both types ern conditional cash transfers—is known of assistance. In contrast, resources are more worldwide. But Bangladesh is not alone. tilted toward the better-off in Afghanistan, In India, the MGNREG Act represents Maldives, and Nepal. a significant milestone in the design and However, because poorer households execution of public works, supported consume less, the same level of public spend- by massive government resources. The ing makes a greater difference in their case. Rashtriya Swasthya Bima Yojana health When assessed relative to the consumption insurance program for the poor is also level of the beneficiaries, social protection pathbreaking in its design though still in an spending turns out to be largely progressive early stage of development. In Pakistan the across the entire region. 160 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 5.2 Bangladesh has a rich and complex social protection architecture The social protection architecture of Bangladesh increased emphasis on cash transfers reflects rec- builds on a long history of efforts emanat- ognition of their greater cost-effectiveness and ing from the need to roll out emergency relief lower risk of misappropriation. Nevertheless, measures in response to a cyclone or a fam- food transfer programs remain an important ine. These efforts instilled a culture of experi- pillar of Bangladesh’s food security strategy and mentation and innovation, depending on the serve a secondary role in turning over the coun- nature of the crisis, the instruments available, try’s emergency grain supplies. and the actors involved. Large numbers of non- Under the government-financed caption of governmental organizations, which provide “social protection and social empowerment,” an extensive safety net system for the poor in 99 social protection programs together account Bangladesh, were part of these relief efforts. for 14 percent of the budget and 2.4 percent of The many programs in place and the relatively GDP. These programs are of varying sizes and poor coordination and harmonization between can be grouped under a number of broad cat- them are largely the result of this long and rich egories: food-based emergency or seasonal relief history of efforts. and public works programs, pension programs, Until 2005, the bulk of public transfers in transfers linked to health and education, and Bangladesh were allocated to expensive and cash allowance programs for special groups. “leaky” food transfer programs. However, since Agricultural programs aimed at supporting then, the government boosted its cash transfer farmers, microcredit and rural employment pro- programs and thereby their share of total program grams, and programs addressing climate change spending. The Food for Education program was are also included in the social protection budget. transformed into a cash-based stipend program, Despite the varied mix of the programs, the larg- and Cash for Work is gradually being incor- est 10 absorb over 70 percent of the total social porated into the Food for Work program. The assistance budget. Source: World Bank 2013a. The progressivity of social safety net (19.4 and 18.8 percent). These figures rep- programs is partially explained by the loca- resent a significant improvement over time: tion of the recipients. In Bangladesh, the in 2005, the coverage of safety nets was vast majority of them reside in the country- negatively correlated with regional level side—7.2 million, about 10 times the number poverty rates. of recipient households in urban areas. Overall, the potential of a safety net pro- Although the coverage of safety net pro- gram to reduce inequality depends on its grams varies significantly by region within coverage of the intended beneficiaries, the Bangladesh, it is closely correlated with local share of total resources spent on intended poverty rates. For example, Barisal, with beneficiaries, the spillover to unintended the highest poverty rate (39 percent), has beneficiaries, and the generosity of the ben- the second-highest coverage of safety nets efits provided. In more technical terms, the among all divisions (34 percent). In contrast, four dimensions to assess are coverage, Chittagong and Dhaka, which have the low- targeting, leakage, and adequacy. Household est poverty rates in the country (respectively, survey data from 2005 to 2010 can be used 26 and 31 percent), have the lowest coverage to conduct such an assessment for most of INADEQUATE SUPPORT 161 FIGURE 5.6 Absolute spending on social protection is progressive in Bangladesh and Sri Lanka a. Afghanistan b. Bangladesh 100 100 90 90 Share of spending (percent) Share of spending (percent) 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 1st (poorest) 2nd 3rd 4th 5th (richest) 1st (poorest) 2nd 3rd 4th 5th (richest) Consumption quintiles Consumption quintiles c. Maldives d. Nepal 100 100 90 90 Share of spending (percent) Share of spending (percent) 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 1st (poorest) 2nd 3rd 4th 5th (richest) 1st (poorest) 2nd 3rd 4th 5th (richest) Consumption quintiles Consumption quintiles e. Pakistan f. Sri Lanka 100 100 90 90 Share of spending (percent) Share of spending (percent) 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 1st (poorest) 2nd 3rd 4th 5th (richest) 1st (poorest) 2nd 3rd 4th 5th (richest) Consumption quintiles Consumption quintiles All social protection All social insurance All social assistance Sources: Based on NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, VPA 2004 for Maldives, NLSS 2010 for Nepal, PSLM 2010–11 for Pakistan, and HIES 2006–07 for Sri Lanka. 162 ADDRESSING INEQUALITY IN SOUTH ASIA the countries in the region (data on benefit poor households, the real value of transfers adequacy are unfortunately unavailable for has halved between 2005 and 2010, falling India). Given the variation in program objec- from 22 percent to 11 percent (box 5.3). tives and design, distinguishing between Adequacy is higher in the case of social social assistance and social insurance is insurance programs. important (figure 5.7). Although countries in South Asia are Social protection spending in South Asia spending more on social protection, a large is characterized by low to moderate coverage. proportion of that spending is in direct In Nepal and Sri Lanka, a little more than subsidies for households to purchase food, half the poor receive support. In most other fertilizer, or electricity. These subsidies are countries, the coverage rate of social assis- justified on poverty reduction grounds, tance spending is between 20 and 30 percent. but they are not always progressive, and in However, coverage has improved over time. some cases they end up being unambigu- In Bangladesh, for instance, 33 percent of ously regressive. Maldives and Sri Lanka the poor benefitted from at least one social rely more on contributory social insur- assistance program in 2010, compared with ance than on social safety nets (DNP 2013; 21 percent in 2005. MoFP 2012). With the exception of these Generally, social assistance programs two countries, spending on social insur- are not very well targeted. Across all coun- ance tends to be small because a majority of tries in the region, 60 percent to 80 percent workers in South Asia are informal sector of the beneficiaries are not poor, and they employees who are not eligible to participate receive between 50 percent and 80 percent of in programs such as old-age pensions and the funding. Although some of the nonpoor health insurance. may stand just above the poverty line, a dis- The bias toward food and price sub- proportionately large fraction of resources sidies is especially marked in India and is captured by the better-off. This high rate Pakistan. In India, the Public Distribution of leakage significantly weakens the impact System is responsible for the provision of of social protection spending in South Asia. subsidized food. In fi scal year 2003/04, it In Bangladesh, for example, the propor- absorbed about 3 percent of GDP, almost tion of program recipients who are not poor triple the average spending on food secu- increased from 44 percent in 2005 to almost rity in advanced economies. Since then 60 percent in 2010; in parallel, the share of the share has declined, but in fiscal year total program spending accruing to the poor 2008/09 it still absorbed about 1 per- dropped from 52.6 percent to 35.3 percent. cent of GDP. This is the largest share of Pakistan is an exception, however, and the resources among all social protection pro- 2010 data even underestimate its actual per- grams: 43 billion Indian rupees, compared formance, because the data only partially to around 30 billion Indian rupees devoted reflect recent improvements in targeting from to MGNREG funding (Union Budget of the implementation of a proxy means test to India 2013–14, http:// indiabudget .nic.in identify the poor. /budget2013-2014/budget.asp). In Pakistan, Finally, the benefit adequacy of social part of the increase in social protection assistance programs remains low in most spending resulted from higher spending on countries considered. In Nepal, the trans- fuel and food subsidies in response to the fer amount represents less than 10 percent global fi nancial crisis. Subsidies absorbed of the consumption of the poorest two more than 230 billion Pakistani rupees, or deciles of the population. Average trans- 1.3 percent of GDP, in fi scal year 2010/11; fer adequacy is also low in Bangladesh, in comparison, BISP consumed about 34 and it has actually declined over the years: billion Pakistani rupees (Government of as a share of the expenditure per capita of Pakistan 2010; World Bank 2013b). INADEQUATE SUPPORT 163 FIGURE 5.7 Social assistance is less adequate than social insurance but has greater coverage a. Afghanistan b. Bangladesh Targeting (percent of total transfers) Targeting (percent of total transfers) 50 116 40 96 30 76 20 56 10 36 0 –10 26 –20 –4 –10 0 10 20 30 40 50 60 –10 0 10 20 30 40 50 60 70 80 90 100 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) c. Maldives d. Nepal 30 100 Targeting (percent of total transfers) Targeting (percent of total transfers) 90 25 80 70 20 60 15 50 40 10 30 20 5 10 0 0 –10 –5 –20 –10 0 10 20 30 40 50 –20 –10 0 10 20 30 40 50 60 70 80 90 100 110 120 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) e. Pakistan f. Sri Lanka 80 70 Targeting (percent of total transfers) Targeting (percent of total transfers) 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 –10 –10 –20 –20 –20 –10 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 Coverage (percent of the poorest deciles) Coverage (percent of the poorest deciles) Social assistance programs Social insurance programs Remittances Sources: Based on NRVA 2007 for Afghanistan, HIES 2010 for Bangladesh, VPA 2004 for Maldives, NLSS 2010 for Nepal, PSLM 2010–11 for Pakistan, and HIES 2006–07 for Sri Lanka. Note: Coverage is the percentage of the population in the bottom two deciles that receives the benefits. Targeting is the sum of transfers received by the bottom two deciles in percent of total transfers. Adequacy, measured by the size of the bubble, is the mean value of the transfer amount received by beneficiaries in the bottom two quintiles in percent of mean expenditure per capita in that group. Deciles are defined based on expenditure per capita net of social protection transfers. The color blue is for social assistance programs and the color orange for social insurance programs. The color brown refers to remittances, used here as a benchmark for government-funded social protection programs. 164 ADDRESSING INEQUALITY IN SOUTH ASIA BOX 5.3 The adequacy of social assistance programs has declined in Bangladesh The success of Bangladesh’s commendable social frequently debated. In Bangladesh, strong protection architecture is partially negated evidence from the mid-1990s suggests that by implementation problems. Outcomes are the FFE program succeeds in attracting poor encouraging in the case of the Vulnerable Group children to school. Specifi cally, estimates Feeding program: in 2010 beneficiaries received show that participation in the FFE pro- on average 88 percent of their expected cash gram increases the probability of attending transfer and 90 percent of the in-kind transfer school by 20 percent on average. Evidence they were entitled to. But substantial variation shows that these gains have been sustained. occurred across beneficiaries. Often program Furthermore, the presumption that recipients officials retain part of the benefits to recover the would misuse the benefits is not supported cost of bagging and transporting the food. In by the available evidence (Devereux 2002; the case of public works, part of the allocation Hanlon, Barrientos, and Hulme 2010). is frequently used to fund nonwage costs of small Likewise, the size of the benefits received infrastructure projects. is small and has increased only marginally More disturbing is the gap between actual in recent years. The most generous transfer and expected transfers in the Food for Education remains the old-age allowance, with an average (FFE) program. Estimates based on the 2000 transfer of Tk 501 a month, which represents HIES suggest that an overwhelming 75 percent only 28 percent to 44 percent of the poverty of program allocations did not reach any house- line. The average amount of rice given out by hold, whether an intended or unintended ben- the Gratuitous Relief program is 15.7 kilograms eficiary. Such disconnect between program per beneficiary, which represents 20 percent to allocations and survey estimates indicates sub- 33 percent of the poverty line. Similarly, the stantial administrative leakage. Similar calcula- benefit from the Vulnerable Group Feeding tions using the earlier 1995–96 HIES showed a program is 17 percent to 26 percent of the substantially lower discrepancy, suggesting that poverty line. Thus, despite increased spend- problems of leakage worsened over time. ing on safety net programs between 2005 and Whether these support mechanisms make 2010, the improvement on the ground has been a difference to their intended beneficiaries is limited. Source: World Bank 2013a. The distributional impact of taxes The government’s capacity to redistrib- ute income through taxes is more limited Whether deliberate policy or unintended in developing countries, because the overall effect, taxes alter the distribution of income tax collection as a percentage of GDP tends or consumption. The distinction between to be signifi cantly smaller (Chu, Davoodi, pretax and after-tax inequality is common and Gupta 2000). It is even lower than the in advanced economies, where tax collec- average in South Asian countries, where tion is more effective than elsewhere. Recent government revenue averages between 10 series of studies suggest that income tax is a percent and 15 percent (map 5.1). The col- major factor in explaining the evolution of lection rates fall behind comparable devel- income distribution in advanced economies oping countries or emerging economies, in the post–World War II period (Atkinson, including Brazil, China, Mexico, and the Piketty, and Saez 2011). INADEQUATE SUPPORT 165 Russian Federation, where revenue collec- agricultural shares of GDP, and lower tions average about 20 percent. They are international trade than most other regions much lower than the collection rates in experience (World Bank 2012a). Even after advanced economies. controlling for the main structural factors, Low rates of revenue collection in South however, revenue mobilization in most South Asia result in part from structural fac- Asian countries is still below the average for tors, notably lower income levels, higher countries at similar income levels (figure 5.8). MAP 5.1 Government revenue in South Asia is low compared with the rest of the world IBRD 41165 SEPTEMBER 2014 This map was produced by the Map Design Unit of The World Bank. The boundaries, colors, denominations and any other information shown on this map do not imply, on the part of The World Bank Group, any judgment on the legal status of any territory, or any endorsement or acceptance of such boundaries. Government revenue, percent of GDP <10 10–19.99 20–29.99 30–39.99 40–49.99 ≥50 No data Source: Based on International Monetary Fund Data Mapper, http://www.imf.org/external/Datamapper/index.php. FIGURE 5.8 Tax revenue is lower than in other countries at a similar development level 40 Tax revenue (percent of GDP, 2010) 30 20 Nepal Pakistan Bhutan 10 Maldives Afghanistan India 0 6 7 8 9 10 11 Log of GDP per capita (PPP, constant 2005 international dollars, 2010) Source: Based on World Bank World Development Indicators database, http://databank .worldbank .org/data/views/reports/tableview.aspx. Note: PPP = purchasing power parity. 166 ADDRESSING INEQUALITY IN SOUTH ASIA Furthermore, the gaps have persisted in declined in the past decades, South Asian the 2000s. countries still see relatively high revenue These revenue gaps cannot be accounted collection from trade. In contrast, trade tax for by the tax structures of South Asian collection has fallen in many other regions countries, which resemble those of other because of trade liberalization. Moreover, low-income countries. The tax instruments some of the countries in the region depend used, and even the tax rates, are standard. heavily on nontax revenue to fi nance their Most countries in the region have introduced public expenditures. That is the case of personal income tax, corporate income tax A fghanistan, whose budget is mostly (CIT), some form of general services tax or funded by donors, and of Bhutan, which value added tax (VAT), excise taxes, and relies on revenue from electricity exports trade taxes. to India. One important difference with other Exemptions, avoidance, and evasion developing countries lies in the relative account for low tax revenue in South Asian weight of each of these tax types in total countries to a much greater extent than the revenue. Compared to their peers at a simi- formal tax structure does. In India, a meager lar development level, South Asian coun- 2.8 percent of the population pays personal tries rely less on income taxes and more on income tax. Stepped-up efforts to increase trade taxes (figure 5.9). Income tax collec- tax collection by the ministry of finance tion, especially for personal income tax, is include a unique online system for monitor- low, except in India. This situation is not ing suspicious transactions through real-time unusual: revenues for personal income coordination among revenue intelligence tax represent about 2 percent of GDP agencies. Yet these efforts concern a few in developing countries compared with thousand cases and less than 0.2 percent of 11 percent of GDP in advanced economies GDP in lost tax revenue, showing that there (Zolt 2008). Although import tariffs have is still some way to go (World Bank 2012a). FIGURE 5.9 South Asian countries rely less on income taxes and more on trade taxes 20 15 Tax revenue (percent of GDP) 10 5 0 Bangladesh Pakistan Nepal Sri Lanka India Indonesia Korea, Rep. Thailand China Income tax VAT or sales tax Other indirect tax Customs duties Other Source: World Bank 2012a. INADEQUATE SUPPORT 167 In Pakistan, actual tax receipts were FIGURE 5.10 In Pakistan, even registered taxpayers fail to file tax estimated at only 21 percent of potential returns receipts in fiscal year 2007/08. This number is likely to be an underestimate because it 90 filing tax returns (percent) does not include the workers welfare fund, 80 Share of taxpayers not 70 the workers profit participation fund, the 60 capital value tax, or the wealth tax (which 50 together represent less than 2 percent of 40 total federal tax receipts). Data limita- 30 tions also rule out an analysis of provincial 20 taxes, which account for about 4 percent 10 of Pakistan’s total tax collection (World 0 Association of Taxpayers registered Taxpayers registered Bank 2009a). persons with RTOs with LTUs Another important difference with other Income tax Sales tax developing countries concerns the “pro- ductivity” of taxes. A standard indicator Source: World Bank 2012b. of such productivity is the share of GDP Note: RTOs = regional tax offices; LTUs = large tax units. in revenue collection for every percentage point of the basic tax rate. In countries be broadly or mildly progressive (Martinez- such as Thailand or Vietnam, the produc- Vazquez 2008; Shah and Whalley 1991) even tivity of VAT exceeds 50 percent and that though many transactions take place in the of CIT reaches 30 percent. This means informal sector. A weak tax administration that a VAT rate of 10 percent, for example, combined with the ability of residents to brings in more than 5 percent of GDP to shift assets outside the country also limits government coffers. In contrast, the pro- the collection of income tax. These limita- ductivity of VAT in South Asian countries tions have forced developing countries to rely varies between 20 and 40 percent, and that more on indirect taxes on consumption and of CIT hovers around a meager 10 percent less on direct taxation. However, consump- (World Bank 2012a). tion taxes, including VAT, are also found Despite some improvements in tax to be progressive in some developing coun- ad m i n ist ration, ta x en forcement has tries (Sahn and Younger 2003; Younger and remained weak, creating incentives for eva- others 1999). sion. Currently, only 17 percent of taxpay- Evidence of the redistributive impact of ers registered as an “association of persons” taxes is scant for South Asian countries, as in Pakistan are systematically filing their it is for most other developing countries. income tax returns (figure 5.10). A large Several incidence analyses have been con- proportion of the taxpayers formally reg- ducted for Pakistan, however, including for istered with regional tax offi ces, and even general services taxes and for federal taxes. with large taxpayer units, fail to file tax Most recently, the incidence of direct and returns. This situation clearly points to indirect taxes at both federal and provincial weak tax enforcement by the Federal Board levels was assessed, making extensive use of of Revenue. Similarly, tax audits, discon- microeconomic data. Although incidence tinued because of weak Federal Board of assumptions and data sources differ, fi nd- Revenue capacity, have remained suspended ings from these studies largely corroborate for years, despite signifi cant improvement each other. Overall, the system is mildly in capacity. progressive, with the progressivity com- The redistributive impact of taxes depends ing mainly from direct taxes (Martinez- on tax scale and structure as well as on Vazquez 2006; Wahid and Wallace 2008). the effectiveness of tax administration. In Households in the lowest deciles pay about developing countries, tax systems tend to 2.4 percent of total taxes, while they 168 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 5.11 Relative to their means, the poor in Pakistan pay Wasteful and often regressive almost as much tax as the middle class subsidies 40 Taxes are only one transfer mechanism through which the distribution of income or consumption can be modifi ed. From an 30 individual’s point of view, taxes and pub- lic spending can have similar consequences Percent 20 on income or consumption (although they typically affect incentives differently). In 10 developing countries, public expenditure policies are found to be more effective than tax policies in redistributing income or in 0 modifying the distribution of consumption. 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th However, expenditures that have intended (poorest) (richest) equity goals may not be progressive in Income deciles practice. Share of taxes paid Effective tax rate In South Asia, a substantial share of Source: Wahid and Wallace 2008. public spending goes into subsidies. Some of them are regressive and can crowd out the provision of essential public goods. An account for 3 percent of total consumption; unusually large fraction of the typically households in the highest decile pay a little low government revenue is often devoted over 40 percent while accounting for about to reducing the final price of food, fertil- 32 percent of total consumption. Effective izer, gas, and electricity (figure 5.12). Equity tax rates are basically flat across most of the concerns are among the justifications for population until the richest group is reached this type of spending. And in some cases, (figure 5.11). as for food, the justification has merit. But Substantial differences exist between in countries where the poor generally lack direct and indirect taxes, however. On access to the grid and often cook using bio- one hand, indirect taxes are neutral or mass, subsidizing electricity is bound to be slightly progressive because the general regressive. Moreover, the delivery mecha- services tax and customs duties are neu- nisms for these subsidies can lend them- tral to slightly progressive, whereas the selves to inefficiencies and leakages. excises are actually regressive (Refaqat I n I nd i a , t he gove r n m e nt sp e nd s 2003; Wahid and Wallace 2008). Direct more on subsidies (relative to GDP) than taxes, on the other hand, are more pro- much richer countries do. For example, gressive (Wahid and Wallace 2008). The only 68 percent of fertilizer subsidies top income group bears over 70 percent of are estimated to reach farmers, with the the burden. The personal income tax paid bulk being paid to large farmers in irri- by salaried workers is progressive, and that gated areas (Herd and Leibfritz 2008). paid by the self-employed slightly so. The The Public Distribution System is another burden of the CIT also falls disproportion- example. It covers about 20 percent of the ately on the top income group, because the population, much more than any other effective tax of the top decile is more than social protection program. It was found triple that of the bottom decile. Because to have strong poverty reduction impacts, CIT accounts for the lion’s share of direct accounting for a significant fraction of taxes, it greatly contributes to the overall the poverty decline between 2004 – 05 progressivity of direct taxes. and 2009–10. Several states have made INADEQUATE SUPPORT 169 substantial improvements in infrastruc- FIGURE 5.12 Much public spending goes into energy subsidies ture and delivery systems to plug leakage. However, the coverage rates were around World 53 percent in rural areas and 33 percent in Sri Lanka urban areas in 2011–12. Take-up rates were progressive across quintiles, but coverage India rates of the richest 20 percent in rural areas Argentina remained high. Because of the price differ- Thailand ence between subsidized grain and grain Indonesia sold through regular marketing channels, Qatar powerful incentives exist to arbitrage and Pakistan make illegal profits. In fiscal year 2004/05, Bangladesh the level of leakage of Public Distribution Venezuela, RB System grains countrywide was estimated to reach above 50 percent. The situation Egypt, Arab Rep. improved later: the illegal diversion and 0 2 4 6 8 10 12 leakages declined to about 44 percent by the Energy subsidies (percent of GDP) end of 2007/08 and to around 35 percent in 2011/12 (Himanshu 2013; Jha and Source: Based on IMF 2013. Ramaswami 2010; Khera 2011). The story is similar in Pakistan. Both federal and provincial governments inter- amounted to more than 250 billion Indian vene in markets for food products through rupees. According to one estimate, the aver- mechanisms such as the “Utility Store” age household in the poorest quintile has arrangement. The complexity of these less than a 20 percent probability of using mechanisms makes quantifying the full LPG; in contrast, the average probability for extent of the inefficiencies difficult, but an urban household in the richest quintile part of the subsidy is clearly appropriated is almost 100 percent (Goutam Lahoti, and by wheat flour millers and traders. The Suchitra 2012). interventions have also resulted in a signifi- Energy subsidies disproportionately ben- cant excess capacity—about 300 percent in efit the better-off in Maldives and Pakistan the wheat milling industry—while crowd- as well. In the case of Pakistan, the poorest ing out the private sector. In addition, fer- 40 percent of households used to receive less tilizer subsidies create a large fi scal burden than 30 percent of total electricity subsidies, because the domestic price of fertilizer is while the richest 20 percent received close to almost 60 percent higher than it was before 40 percent of total subsidies. The distribu- the crisis (World Bank 2013b). tion of benefits improved after the October Energy subsidies are arguably more 2013 tariff increase, but electricity subsidies regressive than food and fertilizer subsidies, remain regressive (figure 5.13). suffering as they do from looser targeting In South Asia where the energy sec- and being more prone to capture by the rich tor suffers from severe capacity or delivery than the subsidized grain programs. India’s shortfalls, subsidies exacerbate the effects subsidies for liquefied petroleum gas (LPG) of the shortages. Subsidized energy use are a case in point. More than half the total stimulates demand by widening the sup- cost of an LPG cylinder is paid for by the ply and demand gaps. Cross-subsidies as a government through a direct price subsidy popular policy in South Asia raise another as well as through transfers to the oil mar- concern. Under cross-subsidies, households keting companies. In fiscal year 2010/11, and especially agricultural users of electric- t h is combi ned gover n ment sp end i ng ity tend to pay well below the incremental 170 ADDRESSING INEQUALITY IN SOUTH ASIA FIGURE 5.13 Electricity subsidies favor the better-off a. Maldives b. Pakistan 35 40 subsidies (percent) subsidies (percent) Share of electricity Share of electricity 30 25 30 20 20 15 10 10 5 0 0 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th (poorest) (richest) (poorest) (richest) Consumption quintiles Consumption quintiles FY2013 After October 2013 tariff increase Sources: Redaelli 2013; Trimble, Yoshida, and Saqib 2011. Note: FY2013 = Fiscal Year 2013. cost of supply, whereas rates are higher for Fertilizer subsidies are crowding out these industrial customers. These subsidies moti- investments. vate industrial customers to turn to alter- Similarly, in Bangladesh ever-growing natives that make good economic sense for subsidies have crowded out investments them but are societally inefficient. Endemic in core public goods such as research and inefficiencies in the supply of energy services development, extension services, and water also increase the cost of energy production, and irrigation infrastructure. Investment in addition to constraining its overall supply in these sectors has fallen from 5.2 percent (Chattopadhyay 2004; Toman 2014). of total public agricultural expenditures to In Sri Lanka, total public expenditure in 2.7 percent over less than a decade. In par- agriculture in 2011 increased by 64 percent ticular, expenditures on research and devel- over 2010, mainly because of increased opment have contracted from 9.9 percent of spending on the fertilizer subsidy. The sub- total agricultural expenditures in 2000 to sidy, which had been available for paddy since 2.8 percent in 2007. This is the equivalent of 2005, was extended to other crops in 2011. only 0.3 percent of agricultural GDP, com- By providing a fertilizer subsidy to farmers, pared with 0.62 percent for other develop- the government absorbed about 90 percent ing countries and 2.80 percent for developed of the total fertilizer cost for paddy in 2011, countries as a group (World Bank 2010). while absorbing 65 percent of the fertilizer Expenditures on core public goods in the costs for other eligible crops (MoFP 2012). agricultural sector have been declining as Ye t ag r ic u lt u ra l g row t h cr it ic a l ly a proportion of total budgetary resources depends on investments in rural infra- devoted to t he sec tor. Discret iona r y structure, including irrigation, fl ood pro- resources, which fund research projects, tection, rural roads, and power, as well have taken the biggest hit, while a large as on investments in agricultural research share of research and development expen- and extension services. These investments ditures is spent on overhead costs of the have generally been found to provide high 10 research institutes under the umbrella economic returns. Because the assets cre- of Bang ladesh Ag ricu lt u ral Resea rch ated by such investments have the charac- Council. Although public expenditures teristics of public goods and thus tend to on agricultural research as a percent- be underprovided by the market, public age of agricultural GDP have remained expenditures are pivotal in their provision. at roughly the same levels (mainly as INADEQUATE SUPPORT 171 result of agricultural growth being slower (Chakraborty, Mukherjee, and Nath 2010). than grow th in overall public expen- The other major component is the direct diture levels), they are low by interna- transfer of resources to districts and other tional standards (World Bank 2010). implementing agencies, thereby bypassing the state budget. Most of this flow is through the Ministry of Rural Development (55 per- The promise of cent), the Ministry of Human Resource intergovernmental transfers Development (29 percent), and the Ministry One can argue that in countries with high of Health and Family Welfare (11 percent). levels of informality and important spatial Transfers to fund Sarva Shiksha Abhiyan— disparities, intergovernmental transfers are India’s fl agship program for universalizing a more effective tool to reduce inequality elementary education—and the MGNREG than either taxes or transfers to individu- together constituted almost half the total als. When avoiding or evading rules is associated with these three ministries. easy, taxes encourage individuals and fi rms to remain informal. And both taxes and FIGURE 5.14 Intergovernmental transfers benefit poorer states transfers to individuals affect incentives to and provinces work and accumulate in ways that tend to be detrimental to efficiency. By contrast, a. India intergovernmental transfers make resources 2,000 Orisha Fiscal transfer per capita available for the provision of public goods in Chhattisgarh places that would not be able to mobilize the 1,500 Bihar Madhya Pradesh West Bengal KarnatakaKerala resources to pay for them. The relevance of Jharkhand Andhra Pradesh Punjab Uttar Pradesh Rajasthan Tamil Nadu intergovernmental transfers is even greater 1,000 Gujarat Haryana in countries that are large and diverse, as is Maharashtra the case in Bangladesh, India, and Pakistan. 500 Intergovernmental transfers are defi ned 10,000 20,000 30,000 40,000 by country-specifi c institutions. In federal GDP per capita (constant prices) nations such as India and Pakistan, intergov- b. Pakistan ernmental transfers are typically enshrined Balochistan Fiscal transfer per capita in the constitution. The magnitude of these 3,500 transfers and the rules and procedures that 3,000 Sindh govern them are often controversial politi- 2,500 cal questions. The controversies involve the magnitude of the transfers, their relationship 2,000 Khyber Pakhtunkhwa to indicators of needs or performance, and 1,500 Punjab the ability of local authorities to raise addi- 4,000 5,000 6,000 7,000 tional revenue. Recent debates around the GDP per capita (constant prices) 18th Constitutional Amendment in Pakistan c. Sri Lanka is a case in point. But every country has an 6,000 Fiscal transfer per capita array of earmarked transfers around specific North East budget lines or government programs. 4,000 North Central The specifics vary from country to coun- Uva Central North West try. In India, subnational governments Southern 2,000 Sabaragamuwa receive transfers from the central govern- Western ment through tax shares and grants from 0 the Finance Commission, through plan 0.06 0.08 0.1 0.12 0.14 0.16 grants from the Planning Commission, GDP per capita (constant prices) and through discretionary grant transfers in the form of centrally sponsored schemes Source: Iyer, Ghani, and Mishra 2010. 172 ADDRESSING INEQUALITY IN SOUTH ASIA In Pakistan, transfers occur through across constituencies. The same can be said tax shares from the National Finance of Pakistan and Sri Lanka, where poorer Commission and through development and regions receive higher per capita fiscal trans- nondevelopment grants allocated by the fers (figure 5.14). federal government (Cheema, Khwaja, and The relatively progressive nature of Qadir 2006). In Sri Lanka, fiscal transfers intergovernmental transfers does not imply, from the central government are based on the however, that public development spend- recommendation of the Finance Commission ing per person is progressive (figure 5.15). and come in four major ways: block grants, Subnational governments in poorer areas Provincial Service Development Grants, tend to have much less locally generated rev- criteria-based grants, and matching grants. enue than those in more affluent parts of the And in Bangladesh, a highly centralized country. In addition, their capacity to spend country, fiscal transfers occur through ear- their resources is more limited. For instance, marked sectoral grants, through grants for a study covering 533 blocks in Bihar— specific transfer programs, through block India’s poorest state—found that one-third development grants, and through recurrent of them did not have any block development expenditure grants (Ghani 2010). officers. As a result, 20 percent of the funds The largest component of fiscal transfers allocated to the state had not been spent in India comes from tax-sharing schemes, but (World Bank 2005). Because of weak capac- discretionary transfers and the subsidies ity, districts with lower revenues and poorer together are almost as large as the tax shares. educational outcomes are found to be less The subsidy component benefits richer states. capable of using Sarva Shiksha Abhiyan For both food and fertilizer subsidies, the funds; in particular, they are found to be value of the transfers is indeed strongly cor- less able to implement programs that target related with the states’ per capita income disadvantaged groups (Jhingran and Sankar (Ghani, Iyer, and Misra 2013). However, the 2006, 2009). Weak capacity also under- overall system of intergovernmental transfers mines local monitoring of public spending, in India is generally progressive and leads to a often leading to leakages or unspent funds more equitable distribution of fiscal resources (Murgai and Zaidi 2005). FIGURE 5.15 Development spending per person is lower in poorer states and districts a. Districts in Bangladesh b. States in India 4 2,000 Spending per capita (thousand takas) Spending in social services per capita 3.5 Maharashtra Kerala 3 Punjab 1,500 (Indian rupees) 2.5 Gujarat RajasthanKarnataka Haryana 2 Tamil Nadu 1.5 Jharkhand Andhra Pradesh 1,000 Odisha West Bengal 1 Bihar Chhattisgarh 0.5 Madhya Pradesh 500 Uttar Pradesh 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 10,000 20,000 30,000 40,000 Poverty headcount ratio (percent) GDP per capita (Indian rupees) Sources: Iyer, Ghani, and Mishra 2010; World Bank 2010. INADEQUATE SUPPORT 173 Inadequate support: Main messages and policy implications Individuals may be hit by shocks, some of which offer modest support. The benefits provided by can have long-lasting impacts on well-being, at social insurance programs are more adequate, any time. Although self-insurance and reliance but the coverage of these programs is thin and on extended families and communities can help favors the better-off. One of the striking fea- them cope, informal mechanisms of this sort tures of South Asian countries is how much tend to be costly and inefficient and to break of the social protection spending goes into down when shocks are widespread. Randomness subsidies—especially for food, gas, and electric- hits even before birth, because individuals differ ity. Subsidies under the form of food distribu- in their inherited wealth, their talent, their tion are prone to leakage; subsidies for gas and preference for the present, and their willingness electricity are clearly regressive. to work hard, all of which have lifelong impli- Outright redistribution is the other typi- cations for well-being. Care by parents for their cal instrument to address bad luck. But across less fortunate offspring and charity by more most countries in the region the fiscal space for fortunate individuals can go some way toward redistribution is constrained by low government redressing these gaps, but they are unlikely to revenue relative to GDP. This is not because of offset them. Thus, even a society with equality the nature of the tax instruments used, although of opportunity and perfect mobility is bound to South Asia relies more on trade taxes than other experience inequality of outcomes. Social protec- regions. Tax rates are not unusual either, but tion and redistribution are the usual instruments the “productivity” of taxes is abnormally low. to address this type of inequality. This is caused by numerous exemptions and spe- Seen from a household perspective, health- cial regimes. It is also the result of massive tax related events and disasters are the most com- avoidance and tax evasion. When considering mon types of shocks in South Asia, with disasters the overall incidence of the tax system, including affecting the rural populations most. Confronted indirect taxation, the poor pay almost as much with shocks, the poor more frequently reduce in taxes as the middle class. expenditures, including quantity and quality of In a region characterized by wide spatial dis- food consumed; they are also more prone to bor- parities, transfers between levels of government row and sell agricultural assets. bear much promise as a mechanism to redress Social protection programs are the standard inequality of outcomes. Intergovernment trans- instrument to address these shocks. By now fers can bridge the gap between assignment of South Asian countries’ spending on social pro- responsibilities—including for social protection tection is roughly the same share of their GDP and disaster management—and assignment of as other countries at a similar level of develop- revenue. Because they do not reach individuals ment. Absolute spending on social protection and households directly, transfers are less likely per person is progressive in some countries to distort incentives to work, save, and invest. and generally not regressive in the rest; spend- With poorer regions generally facing greater ing relative to expenditure per capita is clearly needs and fewer resources, intergovernment progressive. But social protection programs transfers are generally progressive. This is the differ considerably in their strengths. Social case in South Asia, too, but they are often too assistance programs have broad coverage but small to make a difference. 174 ADDRESSING INEQUALITY IN SOUTH ASIA References Development Economics 1989, vol. 1, 61–96. Washington, DC: World Bank. Alderman, Harold, John Hoddinott, and Bill Del Ninno, Carlo, and Mattias Lundberg. 2005. Kinsey, 2006. “Long Term Consequences of “Treading Water: The Long-Term Impact of Early Childhood Malnutrition.” Oxford Eco- the 1998 Flood on Nutrition in Bangladesh.” nomic Papers 58 (3): 450–74. Economics & Human Biology 3 (1): 67–96. Arimond, Mary, and Marie T. Ruel. 2004. Devereux, Stephen. 2002. “Can Social Safety “Dietary Diversity Is Associated with Child Nets Reduce Chronic Poverty?” Development Nutritional Status: Evidence from 11 Demo- Policy Review 20 (5): 657–75. graphic and Health Surveys.” Journal of DNP (Department of National Planning), Minis- Nutrition 134 (10): 2579–85. try of Finance and Treasury, Republic of Mal- Atkinson, Anthony B., Thomas Piketty, and dives. 2013. Statistical Yearbook of Maldives Emmanuel Saez. 2011. “Top Incomes in the 2013. http://planning.gov.mv/yearbook2013 Long Run of History.” Journal of Economic /yearbook.html. Literature 49 (1): 3–71. D’Souza, Anna, and Dean Jolliffe. 2010. “Rising Bandyopadhyay, Sushenjit, and Emmanuel Skoufias. Food Prices and Coping Strategies: Household- 2012. “Rainfall Variability, Occupational Level Evidence from Afghanistan.” Policy Choice, and Welfare in Rural Bangladesh.” Research Working Paper 5466, World Bank, Review of Economics of the Household: 1–46. Washington, DC. Banerjee, Abhijit V., and Esther Duflo. 2007. Foster, A nd rew D. 1995. “Prices, Cred it “The Economic Lives of the Poor.” Journal of Markets and Child Growth in Low-Income Economic Perspectives 21 (1): 141–68. Rural Areas.” Economic Journal 105 (430): Behrman, Jere R. 1988. “Nutrition, Health, Birth 551–70. Order and Seasonality: Intrahousehold Alloca- Frankenberg, Elizabeth, James P. Smith, and tion among Children in Rural India.” Journal Duncan Thomas. 2003. “Economic Shocks, of Development Economics 28 (1): 43–62. Wealth, and Welfare.” Journal of Human Chakraborty, Pinaki, Anit N. Mukherjee, and Resources 38 (2): 280–321. H. K. Amar Nath. 2010. “Interstate Distribu- Ghani, Ejaz, ed. 2010. The Poor Half Billion in tion of Central Expenditure and Subsidies.” South Asia: What Is Holding Back Lagging Working Paper No. 2010-66, National Institute Regions? New Delhi: Oxford University Press of Public Finance and Policy, New Delhi, India. and World Bank. Chattopadhyay, Pradip. 2004. “Cross-Subsidy Ghani, Ejaz, Lakshmi Iyer, and Saurabh Misra. in Electricity Tariffs: Evidence from India.” 2013. “Promoting Shared Prosperity in South Energy Policy 32 (5): 673–84. Asia.” Economic Premise 110 (March): 1–8. Cheema, Ali, Asim I. Khwaja, and Adnan Glewwe, Paul, and Gillette Hall. 1998. “Are Some Qadir. 2006. “Local Government Reforms in Groups More Vulnerable to Macroeconomic Pakistan: Context, Content and Causes.” In Shocks Than Others? Hypothesis Tests Based Decentralization and Local Governance on Panel Data from Peru.” Journal of Develop- in Developing Countries: A Comparative ment Economics 56 (1): 181–206. Perspective, edited by Pranab Bardhan and Glewwe, Paul, and Hanan G. Jacoby. 1995. “An Dilip Mookherjee, 257–84. Cambridge, MA: Economic Analysis of Delayed Primary School MIT Press. Enrollment in a Low Income Country: The Chu, Ke-young, Hamid Reza Davoodi, and Role of Early Childhood Nutrition.” Review Sanjeev Gupta. 2000. “Income Distribu- of Economics and Statistics 77 (1): 156–69. tion and Tax and Government Spending Glewwe, Paul, and Elizabeth M. King. 2001. Policies in Developing Countries.” I M F “The Impact of Early Childhood Nutritional Working Paper W P/00/62 , International Status on Cognitive Development: Does the Monetary Fund, Fiscal Affairs Department, Timing of Malnutrition Matter?” World Bank Washington, DC. Economic Review 15 (1): 81–113. Cox, Donald. 1987. “Motives for Private Income Global Terrorism Database. 2009–14. National Transfers.” Journal of Political Economy Consortium for the Study of Terrorism and 95 (5): 957–76. Responses to Terrorism, START: A Center of Deaton, Angus. 1990. “Saving in Developing Excellence of the U.S. Department of Home- Countries: Theory and Review.” In Proceed- land Security, University of Maryland, College ings of the World Bank Annual Conference on Park, MD. http://www.start.umd.edu/gtd/. INADEQUATE SUPPORT 175 Goutam, Prodyumna, Rahul Lahoti, and J. Y. Khandker, Shahidur R. 2012. “Seasonality of Suchitra. 2012. “Subsidies for Whom? The Income and Poverty in Bangladesh.” Journal Case of LPG in India.” Economic and Political of Development Economics 97 (2): 244–56. Weekly 47 (44). Khanna, Gaurav, David Newhouse, and Pierella Government of Pakistan. 2010. Poverty Reduc- Paci. 2010. “Fewer Jobs or Smaller Paychecks? tion Strategy Paper. Islamabad: Finance Divi- Labor Market Impacts of the Recent Crisis sion, Government of Pakistan. in Middle-Income Countries.” Economic Hanlon, Joseph, Armando Barrientos, and David Premise 11 (April): 1–4. Hulme. 2010. Just Give Money to the Poor: Khera, Reetika. 2011. “Trends in Diversion of The Development Revolution from the Global PDS Grain.” Working Paper No. 198, Centre South. Sterling, VA: Kumarian Press. for Development Economics, Delhi School of Heltberg, Rasmus, and Niels Lund. 2009. Economics, Delhi, India. “Shocks, Coping, and Outcomes for Pakistan’s Kochar, Anjini. 1995. “Explaining Household Poor: Health Risks Predominate.” Journal of Vulnerability to Idiosyncratic Income Shocks.” Development Studies 45 (6): 889–910. American Economic Review 85 (2): 159–64. Herd, Richard, and Willi Leibfritz. 2008. “Fis- ———. 1999. “Smoothing Consumption by cal Policy in India: Past Reforms and Future Smoothing Income: Hours-of-Work Responses Challenges.” Economics Department Work- to Idiosyncratic Agricultural Shocks in Rural ing Paper No. 595, Organisation for Economic India.” Review of Economics and Statistics Co-operation and Development, Paris. 81 (1): 50–61. Himanshu, Abhijit Sen. 2013. “In-Kind Food Martinez-Vazquez, Jorge. 2006. “Pakistan: A Transfers–I: Impact on Poverty.” Economic Preliminary Assessment of the Federal Tax Sys- and Political Weekly. tem.” International Studies Program Working IMF (International Monetary Fund). 2013. Paper 06-24, Andrew Young School of Policy “Energy Subsidy Reform: Lessons and Impli- Studies, Georgia State University, Atlanta, GA. cations.” IMF, Washington, DC, January 2. ———. 2008. “The Impact of Budgets on the Iyer, Lakshmi, Ejaz Ghani, and Saurabh Mishra. Poor: Tax and Expenditure Benefit Incidence 2010. “Is Decentralization Helping the Lag- Analysis.” In Public Finance for Poverty ging Regions?” In The Poor Half Billion in Reduction: Concepts and Case Studies from South Asia: What Is Holding Back Lagging Africa and Latin America, edited by Blanca Regions? ed. Ejaz Ghani, New Delhi: Oxford Moreno - Dodson a nd Quenti n Wodon, University Press and World Bank. 113–62. Washington, DC: World Bank. Jacoby, Hanan G ., and Basab Dasgupta. McKenzie, David J. 2004. “Aggregate Shocks and 2012. “Household Exposure to Food Price Urban Labor Market Responses: Evidence from Increase in Rural Bangladesh.” World Bank, Argentina’s Financial Crisis.” Economic Devel- Washington, DC. opment and Cultural Change 52 (4): 719–58. Jacoby, Hanan G., and Emmanuel Skoufias. Memon, Naseer. 2012. “Disasters in South Asia: 1997. “Risk, Financial Markets, and Human A Regional Perspective.” Karachi: Pakistan Capital in a Developing Country.” Review of Institute of Labour Education and Research. Economic Studies 64 (3): 311–35. Micronutrient Initiative, Flour Fortification Ini- Jha, Shikha, and Bharat Ramaswami. 2010. tiative, USAID, GAIN, WHO, World Bank, “How Can Food Subsidies Work Better? and UNICEF. 2009. Investing in the Future: A Answers from India and the Philippines.” ADB United Call to Action on Vitamin and Mineral Economics Working Paper 221, Asian Devel- Deficiencies; Global Report 2009. Ottawa, opment Bank, Manila, Philippines. Canada: Micronutrient Initiative. Jhingran, Dhir, and Deepa Sankar. 2006. “Ori- MoFP (Ministry of Finance and Planning, Sri enting Outlays toward Needs: An Evidence- Lanka). 2012. Annual Report 2011. Colombo: Based, Equity-Focused Approach for Sarva Treasury of Sri Lanka. Shiksha Abhiyan.” Working paper, World Morduch, Jonathan. 1995. “Income Smoothing Bank, Washington, DC. and Consumption Smoothing.” Journal of ———. 2009. “Addressing Educational Disparity: Economic Perspectives 9 (3): 103–14. Using District Level Education Development Moser, Caroline. 1998. “The Asset Vulnerabil- Indices for Equitable Resource Allocations in ity Framework: Reassessing Urban Poverty India.” Policy Research Working Paper 4955, Reduction Strategies.” World Development 26 World Bank, Washington, DC. (1): 1–19. 176 ADDRESSING INEQUALITY IN SOUTH ASIA Murgai, Rinku, and Salman Zaidi. 2005. “Effec- Wahid, Umar, and Sally Wallace. 2008. “Incidence tiveness of Food Assistance Programs in of Taxes in Pakistan: Primer and Estimates.” Bangladesh.” Journal of Developing Societies International Studies Program Working Paper 21 (1–2): 121–42. 08-13, Andrew Young School of Policy Studies, Pop, Luana. 2010. “Covariate Shocks and Vulner- Georgia State University, Atlanta, GA. abilities in KP and FATA.” Processed, World World Bank. n.d. Migration and Remittances Bank, Washington, DC. Data [database]. World Bank, Washington, Redaelli, Silvia. 2013. “Sustainability, Effi- DC. http://go.worldbank.org/092X1CHHD0). ciency and Equity of Electricity Subsidies in ———. 1990. World Development Report 1990: Maldives: Where Do We Stand and Options Poverty. New York: Oxford University Press. Moving Forward.” Processed, World Bank, ———. 2005. Bihar: Towards a Development Washington, DC. Strategy. Washington, DC: World Bank. Refaqat, Saadia. 2003. “Social Incidence of Gen- ———. 2009a. Pakistan: Tax Policy Report: eral Sales Tax in Pakistan.” IMF Working Tapping Tax Bases for Development. Vol 2. Paper WP/03/216, International Monetary Washington, DC: World Bank. Fund, Washington, DC. ———. 2009b. South Asia: Shared Views on Rosenzweig, Mark R., and Oded Stark. 1989. Development and Climate Change. Washing- “Consumption Smoothing, Migration, and ton, DC: World Bank. Marriage: Evidence from Rural India.” Jour- ———. 2010. Bangladesh: Public Expenditure and nal of Political Economy 93 (4): 905–26. Institutional Review: Towards a Better Qual- Sahn, David E., and Stephen D. Younger. 2003. ity of Public Expenditure. Report No. 47767- “Estimating the Incidence of Indirect Taxes BD. Dhaka and Washington, DC: World Bank. in Developing Countries.” In The Impact of ———. 2012a. Creating Fiscal Space through Economic Policies on Poverty and Income Revenue Mobilization. South Asia Economic Distribution: Evaluation Techniques and Focus. Washington, DC: World Bank. Tools, edited by François Bourguignon and ———. 2012b. “Pakistan: From Raising Spending Luiz A. Pereira da Silva, 27–40. New York: to Spending for Results: A Review of Public World Bank and Oxford University Press. Expenditure, Procurement and Financial Man- Shah, Anwar, and John Whalley. 1991. “Tax agement Practices.” Processed, World Bank, Incidence Analysis of Developing Countries: Washington, DC. An Alternative View.” World Bank Economic ———. 2012c. More and Better Jobs in South Review 5 (3): 535–52. Asia. Washington, DC: World Bank. Toman, Michael. 2014. “Energy and Sustainable ———. 2013a. Bangladesh Poverty Assessment: Economic Development in South Asia: Assess- Assessing a Decade of Progress in Reducing ment of Knowledge and Identification of Key Poverty, 2000–2010. Dhaka and Washington, Issues for Further Investigation.” World Bank, DC: World Bank. Washington, DC. ———. 2013b. Pakistan—Towards an Integrated Trimble, Chris, Nobuo Yoshida, and Moham- National Safety Net System: Assisting Poor mad Saqib. 2011. “Rethinking Electricity Tar- and Vulnerable Households, an Analysis of iffs and Subsidies in Pakistan.” Policy Note, Pakistan’s Main Cash Transfer Program. World Bank Report 62971-PK, World Bank, Washington, DC: World Bank. Washington, DC. ———. 2014. The State of Social Safety Nets UN/ISDR (United Nations Office for Disaster 2014. Washington, DC: World Bank. Risk Reduction), 2008. Linking Disaster Risk Younger, Stephen D., David E. Sahn, Steven Reduction and Poverty Reduction: Good Haggblade, and Paul A. Dorosh. 1999. “Tax Practices and Lessons Learned. Geneva: Incidence in Madagascar: An Analysis Using United Nations. Household Data.” World Bank Economic Valente, Christine. 2011. “What Did the Maoists Review 13 (2): 303–31. Ever Do for Us? Education and Marriage of Zolt, Eric M. 2008. “Inequality, Collective Women Exposed to Civil Confl ict in Nepal.” Action, and Taxing and Spending Patterns Policy Research Working Paper 5741, World of State and Local Governments.” Tax Law Bank, Washington, DC. Review 62 (4): 445. ECO-AUDIT Environmental Benefits Statement The World Bank is committed to preserving Saved: endangered forests and natural resources. • 25 trees Addressing Inequality in South Asia was • 11 million BTU of total printed on recycled paper with 30 percent energy postconsumer fiber in accordance with the rec- • 2,164 pounds of net ommended standards for paper usage set by greenhouse gases (CO2 the Green Press Initiative, a nonprofit program equivalent) supporting publishers in using fiber that is not • 11,735 gallons of waste sourced from endangered forests. For more water information, visit www.greenpressinitiative • 785 pounds of solid .org. waste