Achieving the Millennium Development Goals (Studies in Development Economics and Policy)

Achieving the Millennium Development Goals Edited by Mark McGillivray Studies in Development Economics and Policy Gen...

Author: Mark McGillivray

27 downloads 655 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Achieving the Millennium Development Goals Edited by

Mark McGillivray

Studies in Development Economics and Policy General Editor: Anthony Shorrocks UNU WORLD INSTITUTE FOR DEVELOPMENT ECONOMICS RESEARCH (UNU-WIDER) was established by the United Nations University as its first research and training centre and started work in Helsinki, Finland, in 1985. The purpose of the Institute is to undertake applied research and policy analysis on structural changes affecting the developing and transitional economies, to provide a forum for the advocacy of policies leading to robust, equitable and environmentally sustainable growth, and to promote capacity strengthening and training in the field of economic and social policy-making. Its work is carried out by staff researchers and visiting scholars in Helsinki and through networks of collaborating scholars and institutions around the world. UNU World Institute for Development Economics Research (UNU-WIDER) Katajanokanlaituri 6B, FIN-00160 Helsinki, Finland

Titles include: Tony Addison, Henrik Hansen and Finn Tarp (editors) DEBT RELIEF FOR POOR COUNTRIES Tony Addison and George Mavrotas (editors) DEVELOPMENT FINANCE IN THE GLOBAL ECONOMY The Road Ahead Tony Addison and Alan Roe (editors) FISCAL POLICY FOR DEVELOPMENT Poverty, Reconstruction and Growth George G. Borjas and Jeff Crisp (editors) POVERTY, INTERNATIONAL MIGRATION AND ASYLUM Ricardo Ffrench-Davis and Stephany Griffith-Jones (editors) FROM CAPITAL SURGES TO DROUGHT Seeking Stability for Emerging Economies David Fielding (editor) MACROECONOMIC POLICY IN THE FRANC ZONE Basudeb Guha-Khasnobis (editor) THE WTO, DEVELOPING COUNTRIES AND THE DOHA DEVELOPMENT AGENDA Prospects and Challenges for Trade-led Growth Basudeb Guha-Khasnobis, Shabd S. Acharya and Benjamin Davis (editors) FOOD INSECURITY, VULNERABILITY AND HUMAN RIGHTS FAILURE Basudeb Guha-Khasnobis and Ravi Kanbur (editors) INFORMAL LABOUR MARKETS AND DEVELOPMENT Basudeb Guha-Khasnobis and George Mavrotas (editors) FINANCIAL DEVELOPMENT, INSTITUTIONS, GROWTH AND POVERTY REDUCTION Aiguo Lu and Manuel F. Montes (editors) POVERTY, INCOME DISTRIBUTION AND WELL-BEING IN ASIA DURING THE TRANSITION

George Mavrotas (editor) DOMESTIC RESOURCE MOBILIZATION AND FINANCIAL DEVELOPMENT George Mavrotas and Anthony Shorrocks (editors) ADVANCING DEVELOPMENT Core Themes in Global Economics Mark McGillivray (editor) ACHIEVING THE MILLENNIUM DEVELOPMENT GOALS Mark McGillivray (editor) HUMAN WELL-BEING Concept and Measurement Mark McGillivray (editor) INEQUALITY, POVERTY AND WELL-BEING Robert J. McIntyre and Bruno Dallago (editors) SMALL AND MEDIUM ENTERPRISES IN TRANSITIONAL ECONOMIES Vladimir Mikhalev (editor) INEQUALITY AND SOCIAL STRUCTURE DURING THE TRANSITION E. Wayne Nafziger and Raimo Väyrynen (editors) THE PREVENTION OF HUMANITARIAN EMERGENCIES Machiko Nissanke and Erik Thorbecke (editors) GLOBALIZATION AND THE POOR IN ASIA Can Shared Growth be Sustained? Machiko Nissanke and Erik Thorbecke (editors) THE IMPACT OF GLOBALIZATION ON THE WORLD’S POOR Transmission Mechanisms Matthew Odedokun (editor) EXTERNAL FINANCE FOR PRIVATE SECTOR DEVELOPMENT Appraisals and Issues Laixiang Sun (editor) OWNERSHIP AND GOVERNANCE OF ENTERPRISES Recent Innovative Developments Guanghua Wan (editor) UNDERSTANDING INEQUALITY AND POVERTY IN CHINA Methods and Applications Studies in Development Economics and Policy Series Standing Order ISBN: 978 0 333 96424 8 hardback Series Standing Order ISBN: 978 0 230 20041 8 paperback (outside North America only) You can receive future titles in this series as they are published by placing a standing order. Please contact your bookseller or, in case of difficulty, write to us at the address below with your name and address, the title of the series and one of the ISBNs quoted above. Customer Services Department, Macmillan Distribution Ltd, Houndmills, Basingstoke, Hampshire RG21 6XS, England

Achieving the Millennium Development Goals Edited by

Mark McGillivray

in association with the United Nations University – World Institute for Development Economics Research

© United Nations University 2008 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No portion of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, Saffron House, 6–10 Kirby Street, London EC1N 8TS. Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identiﬁed as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2008 by PALGRAVE MACMILLAN Palgrave Macmillan in the UK is an imprint of Macmillan Publishers Limited, registered in England, company number 785998, of Houndmills, Basingstoke, Hampshire RG21 6XS. Palgrave Macmillan in the US is a division of St Martin’s Press LLC, 175 Fifth Avenue, New York, NY 10010. Palgrave Macmillan is the global academic imprint of the above companies and has companies and representatives throughout the world. Palgrave® and Macmillan® are registered trademarks in the United States, the United Kingdom, Europe and other countries ISBN-13: 978−0−230−21723−2 ISBN-10: 0−230−21723−0 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. Logging, pulping and manufacturing processes are expected to conform to the environmental regulations of the country of origin. A catalogue record for this book is available from the British Library. A catalog record for this book is available from the Library of Congress. 10 17

9 16

8 15

7 14

6 13

5 12

4 11

3 10

2 09

Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne

1 08

Contents

List of Tables and Figures

viii

Acknowledgements

xiii

Notes on the Contributors

xiv

List of Abbreviations

xvi

Foreword by Anthony Shorrocks

1

2

3

xviii

The Millennium Development Goals: Overview, Progress and Prospects Mark McGillivray Introduction The MDGs and progress towards them Volume structure and contents Projecting Progress Towards the Millennium Development Goals Howard White and Nina Blöndal Introduction Approaches to making projections Income poverty Mortality Education Nutrition HIV/AIDS Conclusions Appendix: projection methods Achieving Health, Wealth and Wisdom: Links between Aid and the Millennium Development Goals David Fielding, Mark McGillivray and Sebastián Torres Introduction A brief literature review Data definition and measurement v

1 1 2 14

20 20 20 23 28 34 39 42 43 45

55 55 57 60

vi

Contents

Empirical analysis Conclusion Appendix 4

5

6

7

Achieving the Water and Sanitation Millennium Development Goal P. B. Anand Introduction Hypotheses, methodology, data and limitations Analysis Conclusions and further issues Appendix Measuring Pro-Poor Progress Towards the Non-Income Millennium Development Goals Melanie Grosse, Kenneth Harttgen and Stephan Klasen Introduction The concept of pro-poor progress Methodology Data Results Conclusion Links between Childhood Mortality and Economic Growth and Their Implications for Achieving the Millennium Development Goals in India Sonia Bhalotra Introduction Why growth? Related research Data Descriptive statistics The econometric model Results Conclusions Achieving the Millennium Development Goal for Primary Schooling in India Sonia Bhalotra and Bernarda Zamora Introduction Data and definitions Related literature and contributions

65 84 86

90 90 92 95 111 113

123 123 125 126 127 129 144

150 150 151 153 156 157 160 164 169

174 174 175 179

Contents

Analytical approach Empirical model Results Decomposition and simulation Conclusions Appendix 8

The Burden of Government Debt in the Indian States: Implications for the MDG Poverty Target Indranil Dutta Introduction Debt and MDGs Methodology Results and analysis Simulation Conclusion

Index

vii

182 182 184 194 199 200

208 208 210 212 214 223 226 229

List of Tables and Figures Tables 1.1 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14 2.15 2.A1 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8

The Millennium Development Goals Income poverty forecasts (% living below $1 a day) for 2015 from three studies Subregional income poverty estimates 2015 ($1 a day) Rural and urban poverty in 2015 headcount ratio and millions of people ($1 a day) Mortality estimates 2015 (rates per 1000) Subregional estimates of infant and under-5 mortality, 2015 Rural urban differentials in under-5 mortality, 2015 Access to water in 2015 Net primary enrolment rate, 2015 (proportion of age cohort enrolled) Subregional primary enrolment rates (NER) and numbers out of school, 2015 (millions) Numbers of children out of school, 2015 (millions) Gender equality in education, female/male ratio, 2015 Literacy in 2015 by region Alternative estimates of underweight Per capita food consumption (calories per person per day) Number of HIV/AIDS-related deaths Summary of main approaches to projecting MDG indicators Countries included in the analysis Descriptive statistics for the asset weights Summary statistics Variable definitions and model structure Fitted regression coefficients Main econometric results Equilibrium effects on each variable of 1 standard error shock to each equation Predicted percentage change in each variable for a 100 per cent increase in aid

viii

4 24 26 26 29 30 33 34 35 36 38 38 38 39 41 43 50 61 62 66 68 70 72 76 79


3.9

3.10

3.A1 4.1

4.2 4.3 4.4 4.5 4.6

4.7 4.8 4.9 4.10 4.11 4.12 4.13 4.A1 4.A2 5.1 5.2 5.3 6.1

Percentage increases in each wellbeing indicator with 10, 50 and 100 per cent increases in aid, disaggregated by quintiles Percentage increases in each wellbeing indicator with 10, 50 and 100 per cent increases in aid, disaggregated by initial wellbeing levels Data sources for the conditioning variables Mean and standard deviation of percentage of population having access to water and sanitation, 1990 and 2000 Distribution of number of countries as per access to water and sanitation, 2000 Access to water or sanitation: is it a function of per capita GDP? HDI and access to water and sanitation Synergy in providing access to water and access to sanitation Proportion of population with access to water and sanitation in 2000 is a function of the proportion of population having such access in 1990 Starting point effect and being in sub-Saharan Africa Top ten countries in terms of progress with regard to access to water, 1990–2000 (% of population) Top 10 movers with regard to per cent of population having access to sanitation, 1990 to 2000 Were high achievers different? Results of multiple regression analysis: dependent variable is access to water (% of population), 2000 Results of multiple regression analysis: dependent variable is access to sanitation (% of population), 2000 Whether child mortality rate (2000) is affected by access to water and sanitation Access to water: forecasts for year 2015 by country based on regression models Access to sanitation: forecasts by country based on regression models Selected MDGs in Bolivia (1989 and 1998) Non-income achievements by income decile (conditional on income, Bolivia, 1989 and 1998) Pro-poor growth and pro-poor progress in Bolivia (between 1989 and 1998) Level and change of under-5 mortality and income: all-India and states, 1970–98

ix

82

83 86

95 96 97 100 102

104 105 105 106 106 107 109 110 114 117 131 133 136 158

x


6.2 6.3 6.4 6.5 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.A1 7.A2 8.1 8.2 8.3 8.4 8.5 8.6 8.7

Changes in under-5 mortality and income: all-India and states, 1970–81 and 1982–94 The income elasticity of under-5 mortality: alternative sets of control variables Growth elasticities and fixed effects by state Was there a ‘structural break’ in the income elasticity? Selection of the samples for analysis Microdata sample statistics (weighted by all-India sample weight) Annual growth rates of state level variables (per cent p.a.), 1982–99 Probit estimates of school attendance among 6–11-year-old children Probit estimates of school completion among 12-year-old children Stage-2 results: regression of the state fixed effects on state-level variables Decomposition and prediction based on the linear probability model Linear probability model of school attendance for children aged 6–11 Linear probability model of school completion for children aged 12 years Random-effect models on log of the head count ratio Fixed-effect model on log of the headcount ratio 2SLS and random-effect models on log of the headcount ratio Predicted values of poverty in 2007 Predicted values of poverty in 2015 Simulated values of poverty in 2005, with varied levels of debt ratio Simulated values of poverty in 2015, with varied levels of health expenditure per capita

160 165 166 168 177 178 180 185 189 193 197 200 203 215 216 219 220 222 224 225

Figures 1.1 1.2 1.3 1.4 1.5 1.6 2.1

Progress towards MDG1, MDG2, MDG4 and MDG6 MDG progress in sub-Saharan Africa MDG progress in South Asia Progress towards MDG7 Progress towards MDG8 Total ODA from DAC member countries, 1990–2010 Share of global poverty (proportion of poor in each region using $US1 a day poverty line)

6 7 10 11 12 13 24


2.2 2.3 2.4 2.5 2.6 2.7 2.8 3.1 4.1

4.2

4.3 4.4 4.5 4.6 4.7 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10

Income poverty 2015 (growth-based projections, differential elasticities) Under-5 mortality 2015 (UN Population Projection, 2004 Revision) Mortality trends Net enrolment rates 2015 (naïve projections) Patterns and trends from FAO data Trends in agricultural output per person (2000 = 100) HIV/AIDS three scenarios for Africa Frequencies of values of the five wellbeing indicators GDP per capita (average for 1997–9 in $US, 2000 prices) and percentage of people having access to water in 2000 GDP per capita (average for 1997–9 in $US, 2000 prices) and percentage of people having access to sanitation in 2000 Change in HDI and change in percentage of population with access to water, 1990–2000 Change in HDI and change in percentage of population with access to sanitation, 1990–2000 Countries as percentages with access to water and sanitation in 2000 Access to water in 1990 and in 2000 Access to sanitation in 1990 and in 2000 Growth incidence curve and absolute change for income Conditional NIGIC and absolute change for stunting (z-score ∗ 100) Conditional NIGIC and absolute change for primary completion rate Conditional NIGIC and absolute change for literacy rate Conditional NIGIC and absolute change for ratio of education of women to men Conditional NIGIC and absolute change for share of women in wage employment Conditional NIGIC and absolute change for under-5 survival rate Conditional NIGIC and absolute change for under-1 survival rate Conditional NIGIC and absolute change for vaccinations Conditional NIGIC and absolute change for birth attendance rate

xi

27 31 32 37 40 41 42 63

98

99 100 101 102 103 104 135 137 138 138 139 139 140 141 142 142

xii


5.11 5.12 6.1 6.2 6.3 6.4

Conditional NIGIC and absolute change for access to water Conditional NIGIC and absolute change for access to sanitation Trends in under-5 mortality by state Trends in real log income per capita by state The relationship of under-5 mortality and state income: quadratic fit by state All-India trends in under-5 mortality and state income: population-weighted averages

143 143 159 161 161 162

Acknowledgements This volume originates from the UNU-WIDER research project entitled ‘The Millennium Development Goals: Assessing and Forecasting Progress’. The Board of UNU-WIDER provided valuable suggestions during the early stages of the project. Tony Shorrocks, Director of UNU-WIDER, provided considerable encouragement, advice and support throughout the life of the project. Tony Addison, former Deputy Director of UNU-WIDER, provided very useful advice, especially at the early stages of the project. Valuable comments from anonymous referees on a previous draft of the volume were most useful in shaping the final product. The Millennium Development Goals: Assessing and Forecasting Progress project was supported by many UNU-WIDER staff. Special thanks are due to Anne Ruohonen and Adam Swallow. Anne served as the project secretary, providing extremely efficient support and consistently good humour. The book could not have ever come to completion without Anne’s inputs. Adam provided incisive and timely publications advice, handling often complex matters with sound judgement and skill. UNU-WIDER gratefully acknowledges the financial contributions to the research programme by the governments of Denmark (Royal Ministry of Foreign Affairs), Finland (Ministry for Foreign Affairs), Norway (Royal Ministry of Foreign Affairs), Sweden (Swedish International Development Cooperation Agency – Sida) and the United Kingdom (Department for International Development).

xiii

Notes on the Contributors P. B. Anand is Reader in Environmental Economics and Public Policy at the Centre for International Development, University of Bradford, UK. His research interests include collective action, new institutional economics, human rights and capability approach. He is currently researching agency and governance issues with a focus on the interaction between civil society and local governance institutions. Sonia Bhalotra is Reader in Economics at the University of Bristol, UK. She obtained an MPhil and a DPhil in Economics from the University of Oxford, UK, and a BSc in Economics from Delhi University, India. She has previously worked at the Universities of Cambridge and Sussex, UK, and has consulted for UNICEF, the World Bank, UNU-WIDER, UNESCO and the ILO. Nina Blöndal obtained her MSc from the London School of Economics, UK, in 2002 and has subsequently worked within the field of international development, focusing on development effectiveness and impact evaluation. She has worked on several impact evaluations for the World Bank’s Independent Evaluation Group and as a technical adviser for Danida and is currently studying for a PhD in economics at the University of Copenhagen, Denmark. Indranil Dutta is Lecturer in Economics at the University of Manchester, UK. He was previously a research fellow with UNU-WIDER. His research interests are mainly in the area of development economics, particularly on issues of poverty and inequality. David Fielding is Professor of Economics at the University of Otago, New Zealand. His research interests are in development macroeconomics and quantitative political economy. He has previously held positions at Oxford, Nottingham and Leicester Universities in the UK. His most recent Palgrave Macmillan book is Macroeconomic Policy in the Franc Zone (2005). Melanie Grosse is a PhD student in development economics at the University of Göttingen, Germany. She holds a first-class degree in economics at the University of Göttingen. Her work focuses on measuring poverty, inequality and pro-poor growth in income and non-income dimensions, with applications to Latin American countries. She is also the research manager of an interdisciplinary research group focusing on sustainable resource use in Indonesia. xiv

Notes on the Contributors

xv

Kenneth Harttgen studied economics at the University of Göttingen, Germany, and recently completed his PhD there on empirical dynamics of determinants, distribution and dynamics of poverty, with case studies from sub-Saharan Africa and Latin America. He currently holds a post-doctoral position at the University of Göttingen and has worked as a consultant to several international development agencies including UNDP, UNESCO, GTZ and the OECD. Stephan Klasen is Professor of Development Economics at the University of Göttingen, Germany. He holds a PhD from Harvard University and has since held positions at the World Bank, the University of Cambridge, UK, and the University of Munich, Germany. His research focuses on measurement and analysis of poverty and inequality in developing countries. Mark McGillivray is Chief Economist of the Australian Agency for International Development. He was previously Deputy Director of UNUWIDER. Mark is also honorary Professor of Development Economics at the University of Glasgow, UK, an External Fellow of the Centre for Economic Development and International Trade at the University of Nottingham, UK, and an Inaugural Fellow of the Human Development and Capability Association. His research focuses on aid effectiveness and allocation and measuring achieved human well-being. Sebastián Torres studied economics at the University of the Republic of Uruguay, development economics at the Institute of Social Studies, The Hague, The Netherlands, and has completed his PhD thesis on simultaneous equation models of economic development and income inequality at the University of Leicester, UK. He holds a post-doctoral grant from the Economic and Social Research Council and has worked as a consultant to UNESCO, UNDP, UNIDO and UNU-WIDER. Howard White is Fellow of the Institute of Development Studies, University of Sussex, UK. He has worked for many different agencies in countries across Africa and Asia and has over fifty publications in refereed journals, focusing on aid effectiveness and poverty reduction. His books include Targeting Development: Critical Perspectives on the MDGs and Econometrics and Data Analysis for Developing Countries. Bernarda Zamora holds a PhD in Economics from University Carlos III of Madrid, Spain. She has worked as an economist at the International Monetary Fund. Since 2002 she has worked in Spain as a Visiting Professor at the University of Alicante and University Jaume I in Castellón, and is currently a Research Associate at the School for Policy Studies of the University of Bristol, UK.

List of Abbreviations CHIM DAC Danida DHS EKC FAO GEP GIC GLS GMM GNI GRIM HDI HNP IDS IFPRI IIPS INFHS IMR ISS LAC LDCs MAMS MDGs MPTFE NER NFHS NIGIC NGO NSS ODA ODI OECD PPCH

change in mean Development Assistance Committee, OECD Danish International Development Agency demographic and health survey, World Bank environmental Kuznets curve United Nations Food and Agriculture Organization Global Economic Prospects, World Bank growth incidence curve generalized least squares generalized method of moments gross national income growth rate in the mean Human Development Index health, nutrition, population (poverty data), World Bank Institute of Development Studies, UK International Food Policy Research Institute, USA International Institute for Population Sciences, India Indian National Family Health Survey infant mortality rate Institute of Social Studies, The Netherlands Latin America and the Caribbean least-developed countries Maquette for MDG Simulations, World Bank Millennium Development Goals Millennium Project Task Force on Education net enrolment rate National Family Health Survey of India non-income growth incidence curve non-governmental organization national sample survey official development assistance Overseas Development Institute, UK Organisation for Economic Cooperation and Development pro-poor change

xvi

List of Abbreviations

PPGR Sida SimSIP SSA TI UNAIDS UN-DESA UNDP UNICEF UNFPA UNU-WIDER UPE WHO

pro-poor growth rate Swedish International Development Cooperation Agency Simulations for Social Indicators and Poverty, World Bank sub-Saharan Africa Transparency International Joint United Nations Programme on HIV/AIDS United Nations Department of Economic and Social Affairs United Nations Development Programme United Nations Childrens Fund United Nations Population Fund United Nations University World Institute for Development Economics Research universal primary education World Health Organization

xvii

Foreword

The Millennium Development Goals (MDGs) agreed at the United Nations Millennium Summit in September 2000 constitute the most ambitious enterprise ever undertaken by the international development community. They aim to reduce extreme income poverty and hunger, achieve universal primary education, promote gender equality and empower women, reduce child mortality, improve maternal health, combat HIV/AIDS and other major diseases, provide access to water and sanitation, ensure environmental sustainability and establish a global partnership for development. Each of these goals is defined in terms of a specific target to be achieved by 2015. While there has been tremendous progress in many countries, overall global progress towards achieving the MDG targets has been mixed, with many countries well behind the schedule for 2015. This volume results from the UNU-WIDER research project on ‘The Millennium Development Goals: Assessing and Forecasting Progress’. It suggests how the MDGs can be achieved, by presenting empirical analyses of the core determinants of MDG target variables. A key insight is that most of the MDG targets are causally related in one way or another. Recognition of these interdependencies is crucial not only in analysing the MDGs but also in devising strategies for their achievement. The book provides both broad mapping and details of each of the MDGs and their corresponding targets. It also reports actual and projected progress towards their achievement, examines the role that aid can play in relation to the education and child health objectives, analyses the determinants of the water and sanitation targets, and examines how the measurement of pro-poor growth can be improved with respect to the non-income dimensions of poverty targeted by the MDGs. Special attention is given to achieving the income poverty, education, and child health targets in India, the country which is home to more than 20 per cent of the world’s poor. This book will be of value to a wide audience, including not only those involved in designing and implementing strategies for the MDGs, but also readers with a general interest in poverty and human wellbeing in developing countries. ANTHONY SHORROCKS Director, UNU-WIDER xviii

1 The Millennium Development Goals: Overview, Progress and Prospects Mark McGillivray

Introduction The international community has over the years embraced various campaigns that have aimed to achieve certain development goals, variously defined. These include the United Nations ‘Education for All’ and ‘Health for All’ campaigns adopted in 1978 and 1990 respectively. Education for All aimed to achieve universal worldwide access to primary education by the year 2000. Health for All aimed to provide universal primary education and universal access to healthcare by the same year. Unanimously adopted by all member states at the United Nations Millennium Summit in September 2000, the Millennium Development Goals (MDGs) are to date the most ambitious and comprehensive developmental undertaking ever embraced by the international community. The MDGs involve the eradication of extreme income poverty and hunger, achieving universal primary education, promoting gender equality and empowering women, reducing child mortality, improving maternal health, combating HIV/AIDS and other major diseases, ensuring environmental sustainability and developing a global partnership for development. In most cases the corresponding targets are to be achieved by 2015. A key component of the United Nations strategy to achieve these highly ambitious goals is the doubling of official development assistance (ODA) from its 2003 level, to approximately $US135 billion per year by 2006 and to further increase it to $US195 billion per year by 2015 (United Nations Millennium Project 2005). The official donor community certainly appears to be taking both these calls and the MDGs seriously, to the extent that it has responded both with enthusiastic words at the 2002 Monterrey Conference on Financing for Development and, more importantly, with substantial increases in ODA in subsequent 1

2

Mark McGillivray

years. Many core activities of a number of key international organizations (OECD, UNDP, the World Bank and others within the United Nations system) have also been geared towards the MDGs, and civil society worldwide has embraced the goals. There was also the Second United Nations Millennium Summit in September 2005, at which the international community reaffirmed its commitment to achieving the MDGs. The preceding factors should not be taken to imply that the MDGs have been enthusiastically embraced by all concerned or that they have not been the subject of criticisms, valid or otherwise. Among the criticisms of the MDGs is that they reflect a poor level of analysis, hide more than they reveal about global development challenges, carry the potential to distort meaningful intellectual and research agendas and could serve as a harmful vehicle for a realignment of the political economy of development at the global level (Saith 2006). These factors do, however, combine to suggest that one should not reject the MDGs as merely another symbolic or hollow gesture of the international community, the apparent failure of the Education for All and Health for All campaigns notwithstanding. Achieving the Millennium Development Goals aims to provide analytical insights into how the MDG targets might be achieved. It does this by presenting original and rigorous empirical analyses of key behavioural relationships and how they are likely to impact on progress towards the MDGs. A key recognition is that most of the MDG targets are casually related in one way or another. No one goal can be looked at in isolation from the others, nor from key macroeconomic outcomes not built directly into or recognized within the MDGs. Central to achieving the MDGs is a recognition of these interpendencies, and any robust and insightful analysis of them must take this into account. This first chapter has two remaining aims. The first is to outline briefly each MDG and the progress made towards the targets on which it is based. To this extent, it provides a broad informational context for the remaining chapters. The second is to provide an overview of the volume and briefly to outline the contents of those chapters.

The MDGs and progress towards them The MDGs The United Nations General Assembly, at the 2000 United Nations Millennium Summit, adopted unanimously what is known as the Millennium Declaration. The MDGs were a component of this Declaration.The Declaration itself is much broader than the MDGs, containing inter alia

Overview, Progress and Prospects

3

statements of principle relating to freedom, equality, solidarity, tolerance, respect for nature and shared responsibility (United Nations 2000). The MDGs actually emerge from the section of the Declaration addressing development and poverty eradication. They can be seen largely as a response by the international community to the intolerably low levels of wellbeing experienced by so many people living in developing countries, and the growing gaps in living standards achieved between the richest developed and poorest developing countries. It should come as no surprise, therefore, that seven of the eight MDGs focus on achieving living standards or wellbeing outcomes primarily in developing countries. These MDGs, MDG1 through to MDG7, can be interpreted as intrinsic, in that they involve outcomes worth achieving in their own right. Nor should it come as a surprise that the remaining goal, MDG8, mainly has instrumental roles vis-à-vis these outcomes, calling on certain actions from developed countries. The articulation of each MDG is a statement of broad principle or intent that at best is open to interpretation or at worst vague without clear and precise meaning. MDG1, for example, is to ‘eradicate extreme poverty and hunger’. For most goals this vagueness is removed by one or more targets with which each is associated. Each target and the MDG to which it corresponds are outlined in Table 1.1. There are 18 targets in total. Some of them address long-held priorities of the UN, for which strategies have been in place for some time, such as addressing the needs of the least-developed countries, landlocked countries and small island developing states (MDG8, Targets 8.2 and 8.3). The target for MDG2, ensuring that by 2015 all children complete a full course of primary schooling, builds on the above-stated aim of the UN Education for All campaign. Ten of the MDG targets are time-bound and defined in reasonably precise quantitative terms. Nine targets are intended to be achieved by 2015, and one, improving the lives of at least 100 million slumdwellers (Target 7.3), is to be achieved by 2020. The main goal, that which receives most attention and emphasis, is the income poverty reduction target. This target is to reduce by 2015 the proportion of people whose income is less than one dollar per day, measured in terms of international purchasing power dollars ($PPP), to half of what it was in 1990. It follows that the measure of extreme income poverty on which MDG1 is partly based is the World Bank poverty headcount threshold of $PPP1 per day. The remaining eight MDG targets are more qualitative in nature and are perhaps better described as statements of principle or intent. With the exception of Target 7.1, which corresponds to MDG7, all of these

4 Table 1.1 The Millennium Development Goals MDG1: Eradicate Extreme Poverty and Hunger Target 1.1: halve, between 1990 and 2015, the proportion of people living on less than a dollar a day. Target 1.2: halve, between 1990 and 2015, the proportion of people who suffer from hunger. MDG2: Achieve Universal Primary Education Target 2: ensure by 2015 that all boys and girls complete a full course of primary schooling. MDG3: Promote Gender Equality and Empower Women Target 3: eliminate gender disparity in primary and secondary education preferably by 2005 and in all levels of education by 2015. MDG4: Reduce Child Mortality Target 4: reduce by two thirds, between 1990 and 2015, the mortality rate among children under five. MDG5: Improve Maternal Health Target 5: reduce by three quarters, between 1990 and 2015, the maternal mortality ratio. MDG6: Combat HIV/AIDS, Malaria and Other Diseases Target 6.1: halt by 2015 and begin to reverse the spread of HIV/AIDS. Target 6.2: halt by 2015 and begin to reverse the incidence of malaria and other major diseases. MDG7: Ensure Environmental Sustainability Target 7.1: integrate the principles of sustainable development into country policies and programmes and reverse loss of environmental resources. Target 7.2: halve, between 1990 and 2015, the proportion of people without sustainable access to safe drinking water and basic sanitation. Target 7.3: to improve the lives of at least 100 million slum dwellers by 2020. MDG8: Develop a Global Partnership for Development Target 8.1: develop further an open, rule-based, predictable, non-discriminatory trading and financial system, including a commitment to good governance, development, and poverty reduction both nationally and internationally. Target 8.2: address the special needs of the least developed countries, including tariff and quota free access for least developed countries’ exports; enhanced programme of debt relief for HIPCs and cancellation of official bilateral debt; and more generous ODA for countries committed to poverty reduction. Target 8.3: address the special needs of landlocked countries and small island developing states. Target 8.4: deal comprehensively with the debt problems of developing countries through national and international measures in order to make debt sustainable in the long term. Target 8.5: in cooperation with developing countries, develop and implement strategies for decent and productive work for youth. Target 8.6: in cooperation with pharmaceutical companies, provide access to affordable essential drugs in developing countries. Target 8.7: in cooperation with the private sector, make available the benefits of new technologies, especially information and communications. Source: United Nations (2007).


5

qualitative targets correspond to MDG8, ‘Developing a Global Partnership for Development’. Target 8.2, for instance, involves ‘more generous ODA for countries committed to poverty reduction’. Putting aside the issue of recipient-country commitments to poverty reduction, there are various ways ‘more generous ODA’ can be interpreted. For example, it could involve donors giving larger proportions of their gross national incomes (GNIs) as ODA. Likewise, more generous ODA might simply be interpreting in absolute rather than relative volumes or in terms of more aid provided in the forms of grants rather than loans. Each of these interpretations points to obvious measures on the basis of which quantitative targets could be defined, such as the level of ODA as a percentage of donor gross national income. This measure is the basis of the well-known and longstanding 0.7 per cent target, to which there is no reference in the Millennium Declaration. The international community could have easily adopted precise, quantitative targets for each of the areas addressed in MDG8 had there been the commitment to do so. Such commitment is lacking, it seems. Progress towards the MDGs The MDGs, as mentioned, are a response to the intolerably low living standards of so many of the world’s population. The vast majority of these people – practically all, according to some indicators – live in developing countries. With the possible exception of tackling the spread of HIV/AIDS and tuberculosis, which are essentially global epidemics, achieving the first seven MDGs is essentially about progress in these countries. It therefore follows that tracking performance towards them, and establishing in which parts of the world the greatest challenges will be faced, requires us to focus on the developing world. Now we have made these points, it is clear that of effort required to meet the MDGs is very unevenly distributed across the regions of the developing world. Most of these regions will achieve most of the MDGs. But achieving most of the MDGs in sub-Saharan Africa (SSA) is unlikely in the extreme, it seems, based on a simple examination of the relevant data. Indeed, if the developing world as a whole does not achieve some MDGs, it will be due to a lack of progress in SSA. A lack of progress in South Asia will also play a part in this outcome. Figures 1.1 and 1.2 help illustrate the comments just made regarding overall developing-country and SSA progress towards the MDGs. We return to the progress in South Asia later. All data shown in these figures, and those appearing later in the chapter, have been either taken from or calculated using information in UNAIDS (2006), OECD (2007a; 2007c),

6

1990

1999

2005

2015 (Target)

100 80 60 40 20 0 MDG1: Income Poverty (% of population)

Figure 1.1

MDG1: Hunger (% of children)

MDG2: Primary Schooling (% of school age population)

MDG4: Child Mortality (deaths per 1000 live births)

MDG6: HIV Prevalence (millions of people)

MDG6: Tuberculosis (millions of people)

Progress towards MDG1, MDG2, MDG4 and MDG6

Notes: (i) Data on MDG1, MDG2 and MDG4 are for developing regions only, while data for MDG6 are for the world. (ii) 2005 data for income poverty and child mortality are actually for 2004. (iii) Hunger and child mortality data are not available for 1999.

1990

1999

200 180 160 140 120 100 80 60 40 20 0 MDG1: Income Poverty (% of population)

Figure 1.2




MDG6: HIV Prevalence (millions of people)

MDG progress in sub-Saharan Africa

Notes: (i) 2005 data for income poverty and child mortality are actually for 2004. (ii) Hunger and child mortality data are not available for 1999.

7

8

Mark McGillivray

UN (2007), World Bank (2007) and WHO (2007). Only progress towards those targets for which sufficient statistical data are available are shown. The targets shown in these figures have been calculated directly from published 1990 data or, in cases where the target is not precisely articulated, have been inferred. The income poverty target, for instance, has been calculated by taking 50 per cent of the number of people living in poverty worldwide according to the information reported in UN (2007). Likewise, the child mortality target has been obtained by taking one-third of the 1990 child mortality rate reported in this source. The official targets for MDG6 are a little vague in that they simply mention halting and beginning to reverse the spread of the diseases in question. The MDG6 targets shown in Figures 1.1 and 1.2 are inferred, therefore, being premissed on the possibly generous assumption that ‘halting and beginning to reverse the spread’ involves keeping the incidence of these diseases at their 1999 levels. Consider first the income poverty target. The income poverty data shown in Figures 1.1 and 1.2 are for the percentage of the world and SSA population living on less than $PPP1 per day. In developing regions as a whole, this percentage fell by 13 points, from 32 in 1990 to 19 in 2004. Such a fall corresponds to the number of people living on less than this income dropping from 1.25 billion in 1990 to 980 million in 2004 (UN 2007). The target is 16 per cent by 2015 and, if this progress continues, the MDG1 income poverty target will, in all probability, be met in the developing world as a whole. Africa’s march to the income poverty target is much less certain. In 1990, 47 per cent of the population in SSA lived below the $PPP1 per day poverty line. This percentage fell to 41 in 2004. But meeting the income poverty target requires a reduction of 18 percentage points, to 23 per cent, which on face value would appear unlikely given the trend since 1990. Similar scenarios exist with respect to the hunger target for MDG1, and the targets for MDG2, MDG4 and MDG6, as Figures 1.1 and 1.2 demonstrate. The hunger data relate to the percentage of children underweight, which is interpreted as an indicator of hunger and is used by international agencies to monitor progress with respect to the second MDG1 target. A 10 percentage point reduction between 2005 and 2015 is required if the developing regions as a whole are to reach the hunger target. A 12-point reduction over the same period is required in SSA. Developing regions collectively need to achieve a further 12 percentage point increase in primary school enrolments if MDG2 is to be achieved.1 A 20-point increase is required in SSA. A seemingly unlikely 48-unit reduction in the child mortality rate is required between 2004 and 2015


9

if MDG4 is to be achieved in developing countries as a whole. A seemingly impossible 104-unit reduction is required over the same period in sub-Saharan Africa. The most striking trends are those in HIV prevalence. While all other target variables shown in Figure 1.1 show progress, albeit not apparently sufficient in some cases to achieve the corresponding target, the number of people worldwide with HIV continues to rise. Between 1990 and 2005, the number of people worldwide with HIV increased from 8 million to 39 million. Keeping the spread of HIV at its 1999 level is clearly not happening, as its global prevalence rose by almost 10 million people between 1999 and 2005. A close inspection of Figure 1.2 shows that the majority of the world’s HIV-infected people are actually in SSA. While this region’s contribution to world infections has declined over time, 64 per cent of people in the world with HIV are in SSA. Little more needs to be said about these numbers. This is clearly a problem of enormous magnitude. Better news, not shown in Figure 1.1, is that the global rate of increase in HIV prevalence among 15-to-49-year-olds has begun to taper off in recent years. While rising from 0.3 per cent in 1990 to just under 1 per cent in 2002, since 2003 it has remained at 1 per cent. A similar trend is evident in SSA, where 6 per cent of those in the 15-to-49-year age group are infected with HIV (UNAIDS 2006). Similar news applies to tuberculosis, as Figure 1.1 suggests. While an estimated 8.8 million new cases were reported in 2005, the prevalence of tuberculosis fell from 16.6 million people in 1999 to 14.4 million in 2005 (UN 2007; WHO 2007). If the target for tuberculosis is interpreted as maintaining its incidence at the 1999 level, then the target will be achieved, provided its incidence can be kept at its current level or slightly higher. It is widely recognized that the MDGs will be hardest to achieve in SSA, consistent with the evidence just presented. This should not imply however that there will be little or no difficulty in reaching some MDGs in other parts of the world. There are, indeed, widespread concerns that a number of MDGs may not be met in South Asia, as was alluded to above. Figure 1.3 helps explain why. There are some concerns about whether the income poverty target will be achieved, and a recognition that progress in India is important in this regard (UN 2007). The most profound concerns are for the MDG1 hunger and MDG4 child mortality targets. There has been comparatively little progress towards the first of these targets. As Figure 1.3 shows, the proportion of children underweight fell by 7 percentage points between 1990 and 2005. A further drop of 25 percentage points, from 46 to 21 per cent, is required if the MDG1 hunger target is to be reached. Child mortality fell by 44 deaths per 1000 live

10 Mark McGillivray 1990

1999

2005

2015 (Target)

140 120 100 80 60 40 20 0 MDG1: Income Poverty (% of population)




Figure 1.3 MDG progress in South Asia Notes: (i) 2005 data for income poverty and child mortality are actually for 2004. (ii) Hunger and child mortality data are not available for 1999.

births between 1990 and 2005. This fall is substantial, but it comes from a very high base of 126 deaths per 1000 live births, and a seemingly improbable further decline of 42 deaths per 1000 live births is required by 2015 if MDG4 is to be achieved. Progress with respect to MDG7 and MDG8 is shown in Figures 1.4 and 1.5. Progress in the context of these goals is necessarily vague, given the absence of clearly specified targets. The exception is the MDG sanitation target, which is specified precisely. On face value, progress towards this target needs to be accelerated if it is to be reached. What can be said about the remaining MDG7 and MDG8 targets? While falling between 1990 and 1999 in overall volume terms, the level of ODA provided by countries that are members of the OECD Development Assistance Committee (DAC) is higher in 2005 than in 1990. This is shown in Figure 1.5. This applies to total ODA and to that allocated to leastdeveloped countries (LDCs). It is evident that developing-country access to developed-country markets increased between 1999 and 2005; the debt of developing countries fell between 1990 and 2005, and both youth unemployment and the use of new technology increased between 1999 and 2005 (see Figure 1.5). While the proportion of slum-dwellers has decreased, a less pleasing picture emerges for the remaining MDG7 target variables (see Figure 1.4). This is to the extent to which deforestation has remained virtually constant between 1990 and 2005.2

1990

1999

2005

2015 (Target)

80 70 60 50 40 30 20 10 0 MDG7: Basic Sanitation (% of population)

Figure 1.4

MDG7: Deforestation (% of land area)

MDG7: Greenhouse Gas Emissions (Developing Regions) (billions of tonnes)

MDG7: Greenhouse Gas Emissions (Developed Regions) (billions of tonnes)

MDG7: Slum Dwellers (% of urban population)

Progress towards MDG7

Notes: (i) 2005 data for basic sanitation and greenhouse gas emissions are actually for 2004. (ii) Sanitation and greenhouse gas emission data are not available for 1999.

11

12

1990

1999

2005

90 80 70 60 50 40 30 20 10 0 MDG8: ODA (Total) (billions of dollars)

Figure 1.5

MDG8: ODA (LDCs) (billions of dollars)

MDG8: Market Access (% of imports)

MDG8: Debt (% of exports)

MDG8: Youth (% unempolyed)

MDG8: New MDG8: New Technology Technology (Developed Regions) (Developing Regions) (users per 100 (users per 100 population) population)

Progress towards MDG8

Notes: (i) 2005 youth unemployment data are actually for 2006. (ii) Youth and new technology not available for 1990.


13

2000

2010

140000 Actual

Predicted

Net disbursements ($US millions, 2005 prices)

120000 100000 80000 60000 40000

2009

2008

2007

2006

2005

2004

2003

2002

2001

1999

1998

1997

1996

1995

1994

1993

1992

1991

0

1990

20000

Figure 1.6 Total ODA from DAC member countries, 1990–2010

It was mentioned at the outset of this chapter that a key component of the strategy to achieve the MDGs is the doubling of ODA from its 2003 level to approximately $US135 billion per year by 2006 and to further increase it to $US195 billion by 2015. It was also mentioned that the UN has called on donors to actually provide these levels of ODA. Further information on the level of ODA is provided in Figure 1.6, which shows the actual annual levels of global aid from 1990 to 2006 as well as anticipated or projected levels, based on donor pledges and forward commitments, to 2010. The ODA data shown in Figure 1.6 have been obtained from OECD sources (2007a; 2007b; 2007c). A scaling up of ODA is clearly evident and the level of these flows in 2005, which amounted to $US107 billion, is the highest ever provided by OECD DAC members. The scaling up did not, however, result in the hoped-for doubling of ODA by 2006. In fact the level of ODA in 2006 fell back from its 2005 level, albeit slightly, to $US104 billion. This in part reflects the nature of the very much increased ODA in 2005, which was driven largely by increases in debt forgiveness that do not wholly reflect a high real allocation of public funds to aid budgets. If the trend in anticipated levels is sustained and if anticipations are correct, with donors fully delivering on pledges and commitments, ODA will reach in 2005 prices somewhere between $US160 billion and $US170 billion by 2015. This is clearly well short of the call from the UN for ODA to reach $US195 billion by 2015. The implications of this shortfall for obtaining the MDGs remain to be seen.

14 Mark McGillivray

Finally, having examined ‘progress towards the MDGs’ it would be remiss, at this stage of the chapter, not to consider what this actually means. What does ‘progress towards the MDGs’ or for that matter ‘MDG achievement’ actually mean? The preceding discussion has, to some extent, been vague with respect to these questions. Answers to them logically follow from each other, so let us focus on the meaning of the second question. There are three possible interpretations for any given goal: (i) achievement in all countries and therefore worldwide; (ii) achievement in all regions of the world but not necessarily achievement in each country within each region; or (iii) achievement in the world as a whole but not necessarily in each region or country. The second and third scenarios could be interpreted as referring to average achievements, in that overachievement in some countries compensates for failure to achieve the goals in others. Some associated with the design and implementation of strategies to achieve the MDGs have recently sought to provide clarification on precisely what ‘MDG achievement’ actually means. One position is that assessing whether progress is ‘on track’ for meeting the 2015 targets can be done only at the global level and cannot, therefore, be done for any specific region or particular country. This corresponds to interpretation (iii) above. Vandemoortele (2007:6) adopts this position, specifically asserting that ‘it is erroneous . . . to lament that sub-Saharan Africa will not meet the MDGs’. The current chapter does not seek to resolve this issue, but one point is worth making. To be educated, to be healthy and to have an adequate material standard of living reflects universal human values. They are identified in the United Nations Charter on human rights: each is, in fact, a universal, unalienable human right. This is why most of the MDGs can be viewed as having intrinsic value. These recognitions, which are reflected in the Millennium Declaration, provide a case for defining the MDGs as targets that are to be met within each country. To claim, for instance, that the MDGs have succeeded in eradicating extreme poverty and hunger when at the same time these conditions persist across an entire region or in a number of countries would appear to be inconsistent with the spirit of the Millennium Declaration. Put differently, relying purely on global aggregates seems inconsistent with the principles on which the MDGs are founded and would appear to be reflecting somewhat shaky ethical grounds.

Volume structure and contents Achieving the Millennium Development Goals consists of seven more chapters, each of which examines or uses empirical research methods.


15

Chapters 2 to 5 look at generic issues that are notnecessarily specific to any single country or developing-country group. Among the issues considered are projections of progress towards the goals, the impact of aid and interrelations between goals, the attainment of the targets relating to water and sanitation, and pro-poor growth measurement in non-income dimensions of poverty. Chapters 6 and 7 examine the health and education outcomes on which the MDGs focus – in particular child mortality and school enrolment in particular – using household data for Indian states. Chapter 8 looks at links between efforts within Indian states to service foreign debt and progress toward the MDG income poverty reduction target. More detailed descriptions of each chapter follow below. These descriptions highlight the main finding or findings from each chapter, in particular those that are policy-relevant. Why do three of this volume’s eight chapters focus on India, especially given that the greatest challenges in meeting the MDGs are in sub-Saharan Africa? There are four reasons for this. First, in the early 2000s roughly a quarter of the world’s poor – those living on less than $PPP1 per day, lived in India. Approximately 850 million Indians – 80 per cent of the country’s population – lived on less than $PPP2 per day in the early 2000s (UNDP 2006). Second, despite India’s considerable economic progress in recent years, there are widespread concerns that India will not achieve some of the MDGs. This point was partly alluded to above, when reference was made to the slow progress towards the MDG1 hunger target and MDG4 in the South Asian region as a whole. The rate of progress towards these goals in India is a factor contributing to this, due in part to the size of the Indian population. Third, much of the country- or region-specific research on the MDGs has tended, rightly, to focus on SSA. Comparatively little research has been undertaken for other parts of the world. But achieving the MDGs in other regions of the world is important and here there is an apparent void in the literature. Fourth, data sets relating to living conditions in India are much better than those for most other developing countries. More rigorous investigation into achieving the MDGs is thus possible for India, and many questions can be addressed. Chapter 2 assesses progress towards the millennium development goals. Since the adoption of the International Development Targets, and their successors the MDGs, a growing number of publications have presented estimates of development outcomes in 2015 which, as mentioned, is the target year for most of the goals. What most of these projections show is that the developing world as a whole is ‘off track’ with respect to a number of targets and many countries will fall far short. Chapter 2 examines the soundness of these somewhat dire projections. On the

16 Mark McGillivray

basis of this examination, it concludes inter alia that many of the gains achieved in the outcomes targeted by the MDGs will continue into the new millennium, although not usually fast enough to achieve the ambitious targets set by goals. Neither the goal for income poverty reduction nor that for lower mortality will be met in the vast majority of countries. Attaining universal primary education is the one area where the goal looks achievable in many countries, though by no means all. The chapter warns, however, that these projections are based on various assumptions, including the assumption of business as usual and that various adverse shocks may result in far worse scenarios. Chapter 3 examines aid and the MDGs in terms of health, wealth and education. It uses a relatively new cross-country data set to estimate: (i) the strength of the links between a number of MDG target and related variables, including health, educational status and access to water and sanitation; and (ii) the extent to which aid impacts on these variables. The chapter differs from previous studies of links between wellbeing variables and investigations of aid effectiveness by analysing data for different population subgroups in each country, thus avoiding a number of drawbacks of using national-level data. Among the chapter’s findings is that child mortality is the central variable, where decreases lead to the largest beneficial changes in the other MDG or MDG-similar variables under consideration. It is also the variable on which aid has the largest quantitative impact. This implies that if aid flows are to achieve the maximum benefit, donors should prioritize primarily the MDG4 target for child mortality. The authors also find that while aid is effective overall, the poorest subgroups within each country are typically not the principal beneficiaries of these inflows. This suggests that if the wellbeing of these groups and inequality reduction are priorities, donors need to try harder to target these groups more effectively. Failure to do so will result in a more inequitable world, even if the MDGs are achieved. Chapter 4 uses cross-country regression analysis to develop models to forecast the projected proportion of population with access to water and sanitation in 2015, based on current variables. This study also revisits the issue of whether per capita GDP, levels of human development and governance impact on access to water and sanitation. Further, an attempt is made to explore whether the synergy effect is significant in a statistical sense. This involves examining whether the countries that have made significant progress with one target are more likely to make significant progress with other related targets, and whether and to what extent the achievement or lack of progress on these two targets can impinge on performance in relation to other MDGs or targets. The author finds that on


17

current trends the water target will be either just barely achieved or else narrowly missed in the majority of countries, while the sanitation target will be missed in the great majority of countries. The chapter also points to a strong relationship between access to water and sanitation and child mortality, which suggests that the international community needs to address more seriously the prospect that the MDG7 might not be met. Chapter 5 looks at the non-income-related MDGs. As noted above, all but one of the MDGs involve a target that is not defined in terms of income. While there are plenty of measures designed to track progress in incomes, there are no corresponding measures for tracking the distribution of progress in non-income dimensions of poverty, and thus the distribution of progress towards MDGs 2–7. Chapter 5 proposes to extend the pro-poor growth measurement to non-income dimensions of poverty, particularly health and education. It illustrates empirically the proposed approach for Bolivia and shows that it allows a much more detailed assessment of progress towards MDGs 2–7 by focusing on the distribution of progress. Furthermore, this extension also allows an explicit assessment of the linkage between progress in MDG1 and MDGs 2–7 as well as extending traditional incidence analysis by quantifying outcomes in non-income dimensions of poverty along the income distribution. Chapter 6 links childhood mortality and economic growth in India. As such, it addresses MDG4 and picks up on the concerns that South Asia might not achieve this goal. Using state-level data obtained from the Indian National Family Health Survey (INFHS), the chapter investigates the extent to which the decline in child mortality in India over the last three decades can be attributed to economic growth. In doing this, it exploits the considerable variation in growth over this period, across states and over time. Empirical estimates reported in Chapter 6 are then used to produce a crude estimate of the rate of economic growth that would be necessary to achieve the MDG of reducing the under-5 mortality, by the year 2015,to a third of its level in 1990. The main conclusion is that while growth does have a significant impact on mortality risk, growth alone cannot be relied upon to achieve the goal. Chapter 7 addresses achievement of the MDG for primary schooling in India, and uses two large repeated cross-sections, one for the early 1990s and one for the late 1990s, to describe growth in school enrolment and completion rates for boys and girls, and to explore the extent to which enrolment and completion rates have developed over time. The data are also taken from the INFHS. It decomposes this growth into one component due to changes in the characteristics that determine schooling, and a second associated with changes in the responsiveness of

18 Mark McGillivray

schooling to given characteristics. The chapter’s analysis performs illustrative simulations relevant to the question of whether India will be able to achieve the universal primary education target by the year 2015. The simulations suggest that India will achieve universal attendance, but that primary-school completion rates will not exhibit much progress. Chapter 8 explores what impact, if any, Indian state government debts have on achieving the income poverty target of the MDGs. To fulfil this and many of the other MDG targets, national governments, especially in the developing world, have to undertake major investments in the social sector; but how much they will really be able to do so will depend on the conditions of their finances. The chapter finds that government investment in the social sector is extremely important for the Indian states in reducing poverty, but the government’s debt burden is actually stopping several states from attaining the poverty target. Specifically the chapter finds that while the impact of the debt on poverty is not very harmful in the medium term, it has significant negative impact in the longer run. The chapter’s main conclusion is, therefore, that for policy purposes, reductions in debt should be given priority. The topics covered in this book address important issues relating to the achievement of the MDGs, both in India and elsewhere. They also attempt to give some insight into the state of MDG-relevant research. While the chapters are useful in their individual focus, it is also hoped that they will stimulate further discussion aimed at better, more effective progress towards worldwide achievement of the MDG targets and, more generally, towards a more equitable and stable world.

Acknowledgements The author is grateful to two anonymous referees for useful comments on this chapter. The usual disclaimer applies.

Notes 1. The primary school enrolment data shown in Figures 1.1 and 1.2 are the number of students of primary school age, enrolled in either primary or secondary school, as a percentage of the total population in that age group. See UN (2007) for further details. 2. The market access data shown in Figure 1.5 are the percentage of imports (excluding arms and oil) from developing countries admitted duty-free to developed countries. Debt data are external debt payments as a percentage of export revenue. New technology data are the number of internet users.


19

References OECD (Organisation for Economic Co-operation and Development) (2007a) OECD Journal on Development: OECD DAC Development Co-operation Report 2006. Paris. OECD (Organisation for Economic Co-operation and Development) (2007b) ‘Development Aid from OECD Countries Fell 5.1% in 2006’. Press release. Paris. OECD (Organisation for Economic Co-operation and Development) (2007c) International Development Statistics On-line. Paris. Saith, A. (2006) ‘From Universal Values to Millennium Development Goals: Lost in Translation’. Development and Change, 37(6): 1167–99. UNAIDS (Joint United Nations Programme on HIV/AIDS) (2006) 2006 Report of the Global AIDS Epidemic. Geneva. UNDP (United Nations Development Programme) (2006) Human Development Report, 2006. New York: Palgrave Macmillan. United Nations (2000) United Nations Millennium Declaration: Resolution Adopted by the General Assembly, Fifty-fifth Session, Agenda Item 60(b) New York. United Nations (2007) The Millennium Development Goals Report 2007. New York. United Nations Millennium Project (2005) Investing in Development: A Practical Plan for Achieving the Millennium Development Goals. New York: UNDP. Vandemoortele, J. (2007) ‘The MDGs: “M” for Misunderstood?’. WIDER Angle 2007/1: pp. 6–7, Helsinki: UNU-WIDER. WHO (World Health Organization) (2007) Global TB Database. Geneva. World Bank (2007) World Development Indicators, 2007. Washington, DC.

2 Projecting Progress Towards the Millennium Development Goals Howard White and Nina Blöndal

Introduction Since the adoption of the International Development Targets, and their successor Millennium Development Goals (MDGs), a growing number of publications have presented estimates of development outcomes in 2015, the target year for most of the goals. What the majority of these projections show is that the developing world as a whole is ‘off track’ with respect to most targets. They will not, in aggregate, be met and many countries will fall far short. The MDGs seem set to pass into history as another set of missed development targets. This chapter looks at the soundness of these projections. We discuss briefly the basis for making such projections and then examine the findings for the major MDG targets: income poverty, mortality, education, nutrition and HIV/AIDS.

Approaches to making projections1 Projections may vary because of differences in assumptions regarding (i) determinants of the outcome of interest, (ii) model parameters or (iii) future values of the determinants. The simplest models, and by far the most common approach, take time as the only determinant – that is, the future is forecast based on historical trends, an approach called here naïve projections. Naïve projections tell us if a country is on track or not to meet the relevant goal, which is certainly of policy interest, but are of less use in predicting actual expected values at some point in the future unless its determinants really are uncertain or unforecastable. The next most common approach is to use an outcome–income elasticity to base the forecast on projections of economic growth, the 20

Projecting Progress

21

latter usually being taken from some other source, such as the World Bank’s Global Economic Prospects. Since income is highly correlated with most of the outcomes of interest, this approach can be expected to give a reasonable first approximation. However the correlation with income is imperfect, so adding more variables to the right-hand side of the determinants equation will help, at least if the future values of these determinants can be predicted with any degree of confidence. More sophisticated versions utilize multi-equation models. There are undoubted advantages to such models, which allow a wider range of policy simulations, but they are quite resource-intensive, the assumptions more hidden than is the case for simpler projections. Results from the most recent multi-equation model, the World Bank Maquette for MDG Simulations (MAMS), were not available at the time of writing, and such approaches are not the basis of the projections presented here. However, models allow the flexibility to vary the underlying assumptions and so conduct policy experiments. Such an approach should be of interest to policy-makers, with the caveat that many are wary of CGEs because of their apparent complexity. This is no reason not to use them, but is a reason to be as explicit about the assumptions as possible. Some indicators have risen in recent years at a far greater rate than their historical precedent – primary enrolment being the most obvious case. In these cases, if naïve projections are based on very recent trends, then such forecasts may be superior to those which are model-based. But naïve projections will underestimate future growth if the period used to calculate expected growth includes many years prior to any recent expansion. The main point is that policy can make a difference to the achievement of goals – as shown by the experience of countries that do not follow the expected outcome–income trajectory (see World Bank 2005b for a discussion of Bangladesh, which has been remarkably successful in reducing fertility and mortality faster than income growth would suggest). Were the right policies to be implemented, then all future outcomes would be better than suggested here, though this of course begs the question as to what the ‘right policies’ are. A further source in variation of projections is the level of disaggregation at which they are made. Although most sources report results at the regional level, it is preferable that these regional aggregates be based on country-level projections using country-specific data and parameters. This is not always the case. All methods are of course dependent on the quality of data used. Data availability has improved greatly in the last two decades, though most outcomes are not available on an annual basis.2 There are also exceptions to the general increase in data availability, such

22 Howard White and Nina Blöndal

as maternal mortality and literacy, for which data are less reliable. Moreover, even where data are good, there are frequently groups missed from the data, notably those not in fixed households (nomads, street children, those in institutions, and so on). Basis of own projections Our own projections are made at the country level and are reported using either the UN or the World Bank regional classifications. The baseline data (that is, pre-projection) are from the UN Millennium Indicators Database, supplemented by the World Bank’s World Development Indicators and Demographic and Health Survey (DHS) data as necessary. Population data are from the UN Population Database, 2004 Revision. Projections are made using one of three methods: (i) naïve projections, (ii) growth-based estimates, and (iii) model-based estimates (that is, growth plus other factors). The latter two approaches use differential elasticities to capture the greater poverty-reducing effect of growth in low-inequality settings. Rural–urban rates are based on countryspecific estimates of the rural–urban differential. Specific details of the projections are as follows:

• Income poverty is modelled as a function of income growth alone,

•

• • •

with the elasticity varying according the level of initial inequality. No allowance is made for the fact that the elasticity also changes over time; while this would be a more realistic approach, too little information is available to make such estimates in a sufficiently informed manner. Under-5 mortality is modelled as a function of income poverty (elasticity = 0.4) and female enrolments (elasticity = −1.3) and an autonomous decline of 1.5 per cent per annum. These elasticities are taken from a review of papers on the cross-country determinants of under-5 mortality (for example Hanmer et al. 2003). Undernutrition is modelled as a function of income alone, with the elasticity again varying by the level of inequality. Net enrolments are based on naïve projections, made separately for male and female, capped at 98 and the female rate capped at the male rate.3 Literacy is calculated in two ways: (i) naïve projections, done separately for male and female, and capped at 99.5, and (ii) an accounting approach based on age structure and net enrolments, removing older, less-literate people from the population and adding those graduating

Projecting Progress

23

from school who are assumed to be literate. Again the data are capped at 99.5. • Separate rural and urban estimates are made by using the ratio of rural to urban incidence, based on observed values where available and modelled (as a function of poverty, urbanization, education and regional dummies) when not available.

Income poverty Early estimates of progress towards reducing income poverty come from studies conducted by the Institute of Social Studies for the Sida 2015 project (hereinafter the ISS study, Hanmer et al. 1997a, 1997b), and work undertaken by the Overseas Development Institute (ODI).4 These two studies made forecasts based on estimates of the elasticity of poverty with respect to income, allowing the elasticity to vary according to the degree of inequality.5 The World Bank has since begun publishing annual estimates of income poverty in 2015 in Global Economic Prospects. These estimates are the same as those to be found in the Bank’s MDG Global Monitoring Report. The methodology used in these studies is not public, but appears to be based on household data rather than cross-country elasticities. Elasticities based on household data are generally higher than those from cross-country regressions, so the estimates of future numbers of poor will be lower than if elasticities estimated using cross-country data are used. Our own estimates follow the ODI–Sida approach applied on a countryby-country basis. Projected growth rates are based on regional averages from Global Economic Prospects, with a few exceptions. The poverty elasticity is assumed to vary according to initial inequality.6 To calculate regional figures of absolute numbers of poor, the total for the region is scaled up to capture the population not covered by the data. Table 2.1 summarizes the various results. The most recent World Bank estimates present a more positive picture than do the earlier studies, except with respect to sub-Saharan Africa (SSA). Our own estimates give figures more similar to those of the earlier studies. In general this will be because we assume lower elasticities than those used by the World Bank, especially for high-inequality countries, so the rate of poverty reduction as a consequence of growth is less. However, in the case of Africa, our estimates are more positive. This discrepancy is most likely caused by the World Bank using recent country-specific poverty estimates not available to us. Discrepancies also arise from both differing assumed growth rates and data revisions to poverty estimates in the base year (1990). More

24 Howard White and Nina Blöndal Table 2.1 Income poverty forecasts (% living below $1 a day) for 2015 from three studies

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

Sida 2015

Hanmer and Naschold

World Bank (2005)

Own estimates

32.4 3.4 7.2 18.6 14.3 2.7

33.2 1.9 10.1 23.3 15.7 3.9

38.4 0.9 0.9 12.8 6.9 0.4

30.0 4.2 8.4 16.8 16.1 4.3

Source: See text.

Share of global poverty (%)

60 50 40 30 20 10 0

East Asia and Pacific

Europe and Central Asia

1990

Latin America and Caribbean

Middle East and North Africa

2001

2015

South Asia

Sub-Saharan Africa

Figure 2.1 Share of global poverty (proportion of poor in each region using $US1 a day poverty line) Source: Calculated from World Bank Global Economic Prospects data.

recent growth forecasts are generally more positive than the older ones7 and estimates of poverty in 1990 have been revised downward, especially for Latin America and the Caribbean (LAC). Income poverty will be concentrated increasingly in Africa (Figure 2.1), with absolute numbers of poor under the $US1 a day standard in Africa overtaking South Asia at around the current time. If a $US2 a day poverty line is used then South Asia will continue to have the bulk of the world’s income-poor in 2015 (46 per cent), followed by SSA (31 per cent).

Projecting Progress

25

Subregional estimates Subregional figures (Table 2.2) are based on our own income poverty estimates. The total number of poor in 2015 (736 million) sits well within the range of earlier World Bank estimates, though it is higher than their current projection. The headcount ratio is highest in West and East Africa. Central Africa appears better off than Central and South America but it should be recalled that most of the African countries do not have data, especially those countries affected by conflict, where poverty will be high. These patterns are also evident in the map, although the data limitations also show up there (Figure 2.2). As shown in Table 2.2, the reasonable progress in reducing income poverty is not usually sufficient to meet the MDG of halving poverty by 2015. Of the 71 countries for which there are data, less than a third (21) are expected to achieve the goal (using our own growth-based projections), with Asian countries accounting for a disproportionate share of success. India, China, Indonesia, Bangladesh and Vietnam are all predicted to meet the target. Together these five countries will account for close to half (46 per cent) of the population of the developing world in 2015. Hence expressed in terms of the world’s population the prospects appear better (about half will reach the goal, which of course means that half will not) than when expressed in terms of countries (over two-thirds will not meet the target). Rural versus urban poverty The poverty rate is generally higher in rural areas than in urban ones. Data were available on rural and urban poverty rates using the national poverty line for 56 countries for various years in the 1990s. The mean ratio of numbers of rural and urban poor was 1.86 (and the median 1.53). For only 7 of the 56 countries was urban poverty higher than rural, compared with 18 in which it was over twice as high in rural areas. In order to generate urban and rural poverty estimates for a larger number of countries, this ratio was imputed based on a regression of determinants.8 The estimates are made at the country level, using (i) mostly regional growth rates with some country-specific adjustments, and (ii) elasticities which vary according to the degree of initial inequality. The main finding is that in 2015 two-thirds of the world’s income-poor will be in rural areas. The majority of the populations in Africa and Asia will be rural until after 2020. Given the higher poverty rate in rural areas (16 per cent globally compared with 7 per cent in urban areas), it follows that these areas must account for a disproportionate number of the poor (Table 2.3).

26 Table 2.2 Subregional income poverty estimates 2015 ($1 a day) Progress in halving income poverty

Income poverty in 2015

No. of countries Headcount Population Contribution 2015 expected ratio Number share to poverty headcount to meet (% poor) (millions) (%) (%) as % 19901 goal2 East Africa Central Africa North Africa Southern Africa West Africa Eastern Asia South-Central Asia South-East Asia Western Asia Eastern Europe Northern Europe Southern Europe Caribbean Central America South America Total

33.8 7.5 1.4 11.3 37.5 6.1 14.5

122 11 3 6 124 98 271

4.6 3.3 3.0 1.0 1.0 1.7 9.7 8.5 10.2

29 9 9 1 2 1 16 36 736.3

5.0 2.0 3.1 0.8 4.6 22.2 25.8

16.6 1.5 0.4 0.8 16.8 13.3 36.8

84 – 64 89 78 21 37

0/7 0/0 0/3 1/5 1/9 1/1 6/9

8.6 3.6 3.9 1.4 2.1 0.6 2.3 5.9 100.0

3.9 1.2 1.2 0.1 0.2 0.1 2.2 4.9 100.0

29 141 74 25 52 55 51 97 –

4/6 0/2 2/9 1/2 – 1/3 2/5 2/10 21/71

Notes: 1. Calculated using only countries for which estimates available for both 1990 and 2015. 2. Second figure is the number of countries for which early 1990s income poverty data were available. (Data availability for this column is less than that used for poverty projections, as more recent poverty data were usually used for the projections.)

Table 2.3 Rural and urban poverty in 2015 headcount ratio and millions of people ($1 a day)

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

Headcount ratio

Millions of people

Total

Urban

Rural

Total

Urban

Rural

% Rural

30.0 2.8

21.6 1.6

37.4 4.2

282 10

86 4

196 6

69.6 63.6

3.2

2.5

4.3

15

8

7

47.9

5.7 14.8 8.7

3.1 10.6 7.0

8.4 16.8 16.1

120 255 54

33 59 35

87 196 19

72.5 76.8 35.3

11.8

7.4

16.3

737

225

512

69.5

Percent of population ⬍ 4.3 ⬎ 4.3 to 10.3 ⬎ 10.3 to 24.4 ⬎ 24.4 Not available

Figure 2.2

Income poverty 2015 (growth-based projections, differential elasticities)

27


Summary Recent estimates show that Africa is overtaking South Asia as the region with the largest number of those living on less than a dollar a day, and by 2015 half of those below this poverty line globally will be in that region. The prospects for most of Africa meeting the MDG of halving income poverty seem remote. These trends reflect both the region’s poorer growth performance and the fact that it has high and growing levels of inequality. By contrast the share of East Asia in world poverty is falling rapidly and will be only 3 per cent by 2015; this region will meet the MDG as will other parts of Asia. Latin American countries will be close to achieving the goal, but excluding several in South America. If a higher poverty line of $US2 a day is used then the global poverty profile shifts, with South Asia again having most of the world’s poor, followed by Africa. Income poverty is higher in rural areas than urban ones. In 2015, 70 per cent of the income-poor will live in rural areas.

Mortality Different estimates Hanmer and Naschold (2000) provide regression-based estimates of infant and under-5 mortality allowing for income, HIV/AIDS, education and health services.9 However, there have been substantial upward revisions to mortality estimates in the base year (1990) since they carried out their analysis, so it is to be expected that their results should have a downward bias. Indeed, the figures are lower than those available from other sources. Using data from the World Bank Global Monitoring Report, we made our own naïve projections at the regional level, and then adjusted these to allow for expected higher future growth. We also made country-specific estimates using a model based on poverty levels and female primary enrolment. The final estimates are those from the UN World Population Prospects, 2004 revision. These estimates are based on demographic trends, without reference to economic conditions, but appear nonetheless broadly consistent with the growth-adjusted estimates we made. However, the model-based estimates show a stronger reduction in mortality, notably in SSA and South Asia, this discrepancy being driven by the rapid growth of female enrolments in some countries and the fact that the model does not allow for additional AIDS-related deaths (Table 2.4).

Projecting Progress Table 2.4

29

Mortality estimates 2015 (rates per 1000) Own projections Hanmer and Naïve Growth Model UN population Naschold projection adjusted based projections

Under-5 mortality Sub-Saharan Africa Middle East and North Africa East Asia and Pacific South Asia Latin America and Caribbean Eastern Europe and Central Asia

118 34 8 55 15 6

168 42 30 78 25 29

138 30 32 75 10 30

97 31 27 50 24 25

130 28 30 69 22 30

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

67 22 7 43 14 10

96 53 24 54 19 25

89 28 27 56 29 31

89 28 27 56 29 31

80 25 25 51 17 6

Note: Own projections are at regional level for model-based. Source: See text.

Subregional and country data Table 2.5 shows the subregional figures using the UN Population projection estimates. Rates are highest in West and Central Africa. These are both areas in which tropical diseases remain important and so have unusually high ratios of child to infant deaths. But they are also areas in which conflict has adversely affected mortality. The maps show the scar of premature death running across Africa (Figure 2.3). Some trends All three major developing country regions have seen declining under-5 mortality, though it has been more rapid in Asia and LAC than Africa (Figure 2.4). Indeed, the rate of reduction in Africa has been insufficient to keep up with population growth, so the number of deaths is continuing to climb. As a result, Africa’s share of under-5 deaths is growing. In the case of infant deaths this trend represents a dramatic reversal. In the 1960s three-quarters of all infant deaths were in Asia, but by 2015 Africa will account for the majority, as it has done for child deaths for some years now. However, as with income poverty reduction, the rate of decline is rarely sufficient to meet the ambitious goal of a two-thirds reduction by 2015. Our own projections suggest that only five countries will

Table 2.5

Subregional estimates of infant and under-5 mortality, 2015 Progress toward reducing Under-5 mortality (U5M) in 2015 U5M by two-thirds

IMR Number U5M Number (per 000) (millions) Share (per 000) (millions) Share East Africa Central Africa North Africa Southern Africa West Africa Eastern Asia South-Central Asia South-Eastern Asia Western Asia Eastern Europe Northern Europe Southern Europe Western Europe Caribbean Central America South America Northern America Australia/New Zealand Melanesia Micronesia Polynesia

71 95 28 29 93 23 50 26 27 12 4 6 4 27 14 18 6 4 46 35 17

0.9 0.6 0.1 0.0 1.1 0.5 1.9 0.3 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0

15.8 10.2 2.3 0.5 18.8 7.7 32.7 4.6 2.6 0.5 0.1 0.1 0.1 0.6 0.7 2.0 0.5 0.0 0.2 0.0 0.0

113 162 37 47 149 27 67 32 32 15 6 7 5 39 18 22 5 5 n.a. 8 14

1.3 0.8 0.2 0.1 1.5 0.5 2.5 0.3 0.2 0.0 0.0 0.0 0.0 0.0 0.1 0.2 0.0 0.0 0.0 0.0 0.0

16.9 10.4 2.6 1.3 19.5 6.5 32.5 3.9 2.6 0.0 0.0 0.0 0.0 0.0 1.3 2.6 0.0 0.0 0.0 0.0 0.0

Total

44

6.0

100.0

60

7.8

100.0

Source: UN Population Projections 2004.

U5M in 2015 Number of as a per cent of countries expected rate in 1990 to meet goal 65 82 39 73 69 58 55 41 47 71 – – – 52 36 42 50 51 68 26 33

0/16 0/9 0/6 0/5 0/17 0/5 0/14 3/11 0/18 2/10 – – – 0/12 0/8 0/12 – – – – – 5/143

30

Infant mortality in 2015

Deaths per 1,000 live births ⬍ 27 ⬎ 27 to 64 ⬎ 64 to 118 ⬎ 118 Not available

Figure 2.3

Under-5 mortality 2015 (UN Population Projection, 2004 Revision)

31

Infant mortality rate

32 Howard White and Nina Blöndal 200 180 160 140 120 100 80 60 40 20 0 1950

1970

No. of deaths (millions)

2010 Asia

2030 LAC

3.0

12

2.5

10

2.0

8

Asia

2025

2035

2045

LAC

0.5

2

(e)

2015

1.0

4

90 80 70 60 50 40 30 20 10 0 1940

2005 Africa

1.5

6

0 1940

1995

(b)

14

1960

1980 Africa

(c)

Share of deaths

1990

Africa

(a)

90 80 70 60 50 40 30 20 10 0 1985

2000

2020

Asia

2040

2060

LAC

0.0 1985

1995

2005 Africa

(d)

2015 Asia

2025

2035

2045

2035

2045

LAC

80 70 60 50

1960

1980 Africa

2000 Asia

2020

2040 LAC

2060

40 30 20 10 0 1985

1995

(f)

2005 Africa

2015 Asia

2025 LAC

Figure 2.4 Mortality trends: (a) infant mortality rate; (b) child mortality rate; (c) number of infant deaths; (d) number of child deaths; (e) regional shares of infant deaths (per cent); (f) regional shares of child deaths (per cent) Source: Based on UN Population Projections 2004 revision.

do so: Indonesia and Vietnam are the only two of significance (the others are Montserrat and Vanuatu; outside the sample, the Czech Republic, Moldova and Portugal will also do so). The UN Population projections are more optimistic, showing 10 countries expected to reach the target, with Egypt, Syria and Tunisia being among those added to the list. Rural and urban differences Mortality is higher in rural areas than urban ones in virtually all countries. Data from the last 15 years from 126 Demographic and Health Surveys (DHS) were examined. In only eight countries was infant mortality lower in rural areas than urban ones: on average the infant mortality


33

Rural urban differentials in under-5 mortality, 2015

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

U5M rate (per 000)

Number (millions)

Urban

Rural

Total

Urban

Rural

Total

81 24 21 19 36 22 37

108 39 32 35 57 36 63

97 31 25 27 50 24 51

4.9 0.6 0.4 1.4 2.0 0.9 10.3

9.2 0.6 0.3 2.5 6.7 0.4 19.8

14.1 1.2 0.7 3.9 8.7 1.3 30.0

rate (IMR) was just over a third higher in rural areas, though the ratio was over 2 for Peru. The contrast is stronger for child mortality, with the average ratio being 1.7 (median 1.6); in only 5 cases was rural mortality lower than that in urban areas, and rural mortality was double that in urban areas in 30 cases, reaching a ratio of 4.9 in Armenia. There are no clear regional patterns in these differentials. For simulation purposes, actual values of the ratio for under-5 mortality are used for those countries covered by the survey, with remaining countries set at the sample median (1.4).10 The results are shown in Table 2.6 (applied to the modelbased estimates). Of the 30 million under-5 deaths in 2015, 20 million (that is, two-thirds) will take place in rural areas. Water supply Lack of access to water will be a largely rural phenomenon by 2015: three-quarters of the 761 million without access to water will be in rural areas. Lack of access will be concentrated in East Asia and SSA, which will together account for 80 per cent of those without access (Table 2.7). Summary All three major developing country regions have seen declining mortality, though it has been more rapid in Asia and LAC than in Africa. But only in rare cases has the rate of reduction been at the 1.6 per cent a year required to achieve the MDG of a two-thirds reduction by 2015. Indeed, the rate of reduction in Africa has been insufficient to keep up with population growth, so the number of deaths is continuing to climb. As a result, Africa’s share of under-5 deaths will continue to grow. In the case of infant deaths this trend represents a dramatic reversal. In the 1960s

34 Howard White and Nina Blöndal Table 2.7 Access to water in 2015

Access to water (% of population)

Share of total number Urban Rural Total Urban Rural Total globally Sub-Saharan Africa Middle East and N. Africa Europe and Central Asia East Asia and Pacific South Asia L. America and Caribbean Total

Number without access (millions)

Share of those without access in rural areas

82.2 96.0

57.6 79.4

69.3 89.7

72 9

227 30

288 37

38 5

79 80

99.4

83.2

94.1

2

28

29

4

98

86.3 95.7 98.0

78.2 94.9 85.4

85.4 95.6 96.5

147 24 10

225 59 18

308 77 22

41 10 3

73 77 81

91.4

81.5

87.8

264

586

761

100

77

three-quarters of all infant deaths were in Asia, but by 2015 Africa will account for the majority, as it has done for child deaths for some years now, with the highest rates in Western and Central Africa. Two-thirds of these deaths will be in the rural areas of Africa, which is a similar share of under-5 deaths in rural areas across the world. Similar patterns reveal themselves with respect to critical determinants of under-5 mortality: access to water and immunization. For example, West and Central Africa are also the two subregions with the lowest access to water expected in 2015. Over three-quarters of those without water in 2015 will be in rural areas. Immunization coverage fell in some countries in the 1990s and there remain a significant minority of countries with unacceptably low coverage rates. Falling under-5 mortality, other health improvements which increase longevity and the demographic transition are increasing the share of the elderly in the population of all regions. This trend will in time create new problems for developing countries.

Education The Millennium Project Task Force on Education (MPTFE) is the only study to estimate primary net enrolment rates.11 Projections are made based on predictive power of the s-curve (that is, that later enrolments are harder to achieve) and suggest that SSA will continue to lag behind


35

Net primary enrolment rate, 2015 (proportion of age cohort enrolled)

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

MPTFE

Own projections

79.6 86.1 88.0 95.6 96.3 97.0

73.9 92.4 89.3 98.0 91.3 90.8

Source: Millennium Project Task Force on Education (MPTFE) and own projections.

other regions with an estimated enrolment rate of just under 80 per cent. South Asia and the Middle East and North Africa are also estimated to lie below 90 per cent while other regions are expected to lie within 5 per cent of universal enrolment. This ‘last 5 per cent’ is increasingly recognized as being ‘problem groups’ requiring different policies. Our own countrylevel naïve projections do not allow for this effect, which may explain the higher figures in South Asia and LAC. Naïve projections seem appropriate since the policy push behind primary enrolments has enabled increases over and above those expected by economic performance or other possible determinants. It is the possibility of such a push in Africa which explains the higher rates there estimated by MPTFE. However, India began a very recent push, not yet shown in the data and so not picked up by naïve projections (Table 2.8). The subregional breakdown shows the problem in Africa to be least in Southern Africa (Table 2.9), which is confirmed partially by the map, but it is patchy owing to limited data (Figure 2.5). Defining universal primary enrolment as a net enrolment rate (NER) of 98 per cent shows that the majority of countries appear on track to achieve this goal. Breaking down our own projections we see that SSA will account for nearly half of all the 80 million children out of primary school by 2015 (Table 2.10), reflecting the lower enrolment rates across the subcontinent (Figure 2.5). Gender dimension Areas with low enrolments have a disproportionate number of girls out of school (see Table 2.10). The MPTFE study also provides estimates of gender equality in primary education measured by female-to-male net primary enrolment. Sub-Saharan Africa is estimated to have the lowest ratio, with 93 girls to 100 boys (Table 2.11).

36 Howard White and Nina Blöndal Table 2.9 Subregional primary enrolment rates (NER) and numbers out of school, 2015 (millions) No. of countries with UPE No. NER by 2015 East Africa 13.9 76.5 Central Africa 5.6 77.3 North Africa 4.2 84.8 Southern Africa 0.6 90.5 West Africa 15.4 70.8 Eastern Asia 9.3 91.8 South-Central 19.1 91.2 Asia South-Eastern Asia 6.3 90.2 Western Asia 2.0 93.6 Eastern Europe 0.8 95.2 Northern Europe 0.2 97.6

9/17 7/9 6/7 1/5 12/18 4/7 10/14 7/11 13/18 5/10 11/13

No. of countries with UPE No. NER by 2015 Southern Europe 0.3 Western Europe 0.2 Caribbean 0.1 Central America 0.4 South America 0.9 North America 0.6 Australia/ 0.0 New Zealand Melanesia 0.2 Micronesia 0.0 Polynesia 0.0 Total 80.0

96.3 98.0 98.0 98.0 98.0 98.0 98.0

5/10 – 11/13 24/24 8/8 – –

85.8 – 98.0 – 98.0 – 89.1 182/230

Notes: NER = net enrolment rates; UPE = universal primary education.

Literacy No estimates of literacy were located. Country-level projections were made for this report using two methods: (i) naïve projections, and (ii) an accounting approach based on the net enrolment rate.12 The two approaches are not entirely consistent, with naïve projections predicting lower literacy. This is because recent rapid increases in enrolments are not yet reflected in higher literacy, but the accounting approach picks up these increases. However, both methods find that the vast majority of illiterates in 2015 will be in South Asia, with the majority of the remainder in SSA (Table 2.12). Summary Sub-Saharan Africa will continue to lag behind other regions with respect to primary-school enrolments, with an estimated enrolment rate of just under 80 per cent, though the problem is less in Southern Africa than elsewhere. South Asia and the Middle East and North Africa are also estimated to lie below 90 per cent while other regions are expected to lie within 5 per cent of universal enrolment. However, this ‘last 5 per cent’ is increasingly recognized as consisting of ‘problem groups’ requiring different policies.

Percent ⬍ 65 ⬎ 65 to 85 ⬎ 85 to 95 ⬎ 95 Not available

Figure 2.5

Net enrolment rates 2015 (naïve projections)

37

38 Table 2.10 Numbers of children out of school, 2015 (millions) Proportion Share of total Boys Girls Total girls out of school Sub-Saharan Africa 17.9 21.4 Middle East and North Africa 2.4 2.4 Eastern Europe and Central Asia 1.3 1.8 East Asia and Pacific 9.5 5.9 South Asia 6.8 8.5 Upper middle income 0.7 0.7 High income 0.7 0.7 Total 39.2 41.4

39.3 4.7 3.1 15.4 15.3 1.3 1.4 80.6

54.5 49.8 59.0 38.4 55.7 50.0 50.0 51.4

48.8 5.9 3.9 19.1 19.0 1.7 1.7 100.0

Table 2.11 Gender equality in education, female/male ratio, 2015

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

MPTF on education (primary)

MPTF on gender (primary, gross)

Own estimates (primary)

95.3 93.2 99.1 99.3 98.8 96.0

111.0 94.6 96.2 98.3 98.4 100.0

99.6 98.8 100.0 99.2 101.1 100.0

Note: MPTF = Millennium Project Task Force. Source: UN Millennium Project (2005b), and own estimates.

Table 2.12 Literacy in 2015 by region Naïve projections

Accounting based

No. of Literacy No. of illiterates Literacy illiterates rate (millions) rate (millions) Sub-Saharan Africa Middle East and North Africa Europe and Central Asia East Asia and Pacific South Asia Latin America and Caribbean Developed countries Total

81.7 86.6 99.1 99.1 68.1 98.1 99.1 89.6

99 33 4 16 387 9 8 555

93.7 91.9 99.3 99.3 80.9 99.4 96.8 93.8

34 20 3 11 232 3 26 329

Projecting Progress

39

Table 2.13 Alternative estimates of underweight Children Adults Smith and FAO (2015) Haddad (2020) Onis et al. (2015) Prevalence South Asia Sub-Saharan Africa East Asia Near East and North Africa Latin America and Caribbean

12 23 6 7 6

37.4 28.8 12.8 5 1.9

26.2 29.2 3.0 7.4 3.4

Numbers South Asia Sub-Saharan Africa East Asia Near East and North Africa Latin America and Caribbean Developing countries

195 205 135 37 40 612

66 48.7 21.4 3.2 1.1 140.4

61.8 42.7 3.0 3.4 2.0 112.9

Shares South Asia Sub-Saharan Africa East Asia Near East and North Africa Latin America and Caribbean Developing countries

31.9 33.5 22.1 6.0 6.5 100.0

47.0 34.7 15.2 2.3 0.8 100.0

54.7 37.8 2.7 3.0 1.8 100.0

Source: See text.

Nutrition Different estimates Nutrition outcomes have been comprehensively modelled in three different sources: two IFPRI studies (Smith and Haddad 2000; von Braun et al. 2005) and FAO’s agricultural projections (FAO 2002).13 In addition Onis et al. (2004) present subregional-based naïve projections. The FAO study refers to adults who are undernourished (defined with reference to a calorie requirement of 1900 kcal per day), whereas the IFPRI studies and Onis et al. refer to children only (using anthropometric measurement). The IFPRI results shown forecast for 2020 rather than for 2015 (Table 2.13). The different estimates do not, in this case, give a similar picture. Smith and Haddad’s estimates are in general higher than those of Onis, and there is a marked discrepancy in the case of East Asia – which Smith and

350 Millions underweight

40 35 30 25

20 15 10 5 0 (c)

East Asia

(b)

South Asia

East Asia

South Asia


Near East/ North Africa

5


10

40 35 30 25 20 15 10 5 0

Sub-Saharan Africa

15

0 (a)

Per cent underweight

Share of undernutrition

20

Sub-Saharan Africa

Per cent underweight

25

Near East/ North Africa


300 250 200 150 100 50 0

1990/92

1997/92

Sub-Saharan Africa LAC East Asa

2015

2030

North-East North Africa South Asia

(d)

1990/92

1997/99

Sub-Saharan Africa LAC East Asa

2015

2030

North-East North Africa South Asia

Figure 2.6 Patterns and trends from FAO data: (a) prevalence of adult underweight, 2015; (b) share of adult underweight, 2015; (c) prevalence of adult underweight, 1990–2030; (d) number of adult underweight, 1990–2030

Haddad have as accounting for 15 per cent of children underweight by 2020, compared with Onis’s 3 per cent in 2015. While both have South Asia’s share as exceeding that of Africa, Smith and Haddad predict higher prevalence in South Asia than Africa, but Onis the reverse. FAO forecasts for adults give the two regions an equal share. The different studies do give broadly similar results for prevalence in SSA, as does another recent IFPRI study which suggests prevalence of child malnutrition to be 30.4 per cent in 2015 by their baseline ‘business as usual’ scenario (Rosegrant et al. 2005). However, the point of the study is to show how different policy interventions – such as rural infrastructure and investing in agricultural research – can make an enormous difference to nutritional outcomes. The different scenarios place the number of malnourished children in 2025 at between 9.4 million and 55.1 million.

Some trends Figure 2.6 shows trends from the FAO data. As mentioned, the shares of Africa and Asia are about the same in 2015, reflecting a growing number


41

Per capita food consumption (calories per person per day) 1964–6

1997–9

2015

2054 2058 2290 2393 2017 1957

2681 2195 3006 2824 2403 2921

2850 2360 3090 2980 2700 3060

Developing countries Sub-Saharan Africa Middle East and North Africa Latin America and the Caribbean South Asia East Asia

Index of agricultural production

Source: FAOSTAT; see http://faostat.fao.org/.

140 120 100 80 60 40 20 0 1960

1970

1980

1990

Sub-Saharan Africa

East Asia

LAC

South Asia

2000

Figure 2.7 Trends in agricultural output per person (2000 = 100) Source: Calculated from FAOSTAT; http://faostat.fao.org/.

of undernourished people in Africa, compared with a steady decline in South Asia. These trends reflect differing trends in food availability. The FAO report data on food consumption, based on calculations of food production and the net trade balance. Food consumption in Africa is set to rise slowly, falling further behind that of other regions (Table 2.14). This trend reflects the region’s dismal agricultural performance. In contrast to other regions which have experienced rising agricultural productivity since the 1960s, that in Africa has declined (Figure 2.7). Inequality The FAO (2002) report graphs differing levels of malnutrition for three different distributions against average food consumption. With high

42 Howard White and Nina Blöndal 40

6.0

35 30

4.0

Millions

Prevalence (%)

5.0

3.0 2.0

25 20 15 10

1.0

5

0.0 (a) 1980

1990

Baseline

2000 Optimistic

2010

2020 Pessimistic

0 1980

(b)

1990 Baseline

2000 Optimistic

2010

2020 Pessimistic

Figure 2.8 HIV/AIDS three scenarios for Africa: (a) adult HIV prevalence in Africa by three different scenarios; (b) number of adults living with HIV/AIDS in Africa Source: UNAIDS (2005).

inequality, which is experienced in SSA, the region’s average food availability yields undernutrition of 23 per cent. But if the region were to have low inequality then, with the same level of food availability, undernutrition would be just 10 per cent.

HIV/AIDS Projections related to diseases are plagued by additional uncertainty regarding both current prevalence rates, the appropriate models to be used for prediction purposes and the extent to which effective action will be taken to combat the disease. Based on current trends, UNFPA (no date) estimate that there will be 46 million new HIV infections by 2010, but only a third that number (17 million) with an appropriate policy response. The three scenarios in a recent UNAIDS (2005) report on AIDS in Africa show that the current figure of 25 million adults living with HIV/AIDS may either fall to 14 million by 2025 or rise to 38 million – nearly two-thirds more than the optimistic estimate (Figure 2.8). Even the most pessimistic estimates suggest nonetheless that prevalence will decline, reflecting the belief that within 15 years (by around 2020) there will be no new AIDS cases. As AIDS in Africa is brought somewhat under control, attention is shifting to Asia, where the epidemic is in its early stages: the number of AIDS-related deaths in Asia will quadruple in the next 15 to 20 years, compared with an increase of about a third in Africa. But, despite its larger population share, Asia will never account for even half of all deaths, a dubious honour that remains with Africa into the longer term

Projecting Progress

43

Table 2.15 Number of HIV/AIDS-related deaths 1990–5

2000–5

2010–15

Absolute number of excess deaths from HIV/AIDS (000s) Sub-Saharan Africa 3,216 14,807 18,933 Asia 1,314 3,461 10,872 LAC 201 697 774 More developed 427 789 1,692 Total 5,158 19,754 32,271 Share of excess deaths (per cent) Sub-Saharan Africa 62.3 Asia 25.5 LAC 3.9 More developed 8.3 Total 100.0

75.0 17.5 3.5 4.0 100.0

2020–5

2045–50

18,585 17,078 730 1,557 37,950

7,510 4,821 37 49 12,417

58.7 33.7 2.4 5.2 100.0

49.0 45.0 1.9 4.1 100.0

60.5 38.8 0.3 0.4 100.0

Note: More developed are Russia and USA. Source: UN Population Division.

(Table 2.15).14 Africa’s share is likely understated in this table given the efficacy with which Asian countries have been tackling the epidemic, and the continuing opportunity for some of these countries to head it off, compared with the more mixed picture in Africa. Those data that are available, though not systematic, suggest that the incidence of HIV/AIDS is higher in urban areas,15 and many affected rural residents go to urban areas for treatment. However, since most of the population in Africa and Asia are rural it is possible that the absolute number of HIV/AIDS cases is higher in rural areas than urban ones.

Conclusions There has been substantial progress in poverty reduction . . . The second half of the last century witnessed remarkable gains in the reduction of many forms of poverty. Mortality fell, and life expectancy rose, across the developing world at historically unprecedented rates – much faster than had been achieved in the now developed countries. With the exception of countries affected by HIV/AIDS, under-5 mortality has continued to decline in most countries even in times of economic stagnation or decline, thanks to increased immunization coverage, access to safe water and so on. The last decade has seen many countries adopt programmes to ensure universal primary education.


. . . and this progress will continue, though not fast enough to achieve the MDGs . . . All projections suggest that these gains will continue into the new millennium, although not usually fast enough to achieve the ambitious targets set by the MDGs. Neither the goal for income poverty reduction nor that for lower mortality will be met in the vast majority of countries. Attaining universal primary education is the one area where the goal looks achievable in many countries, though by no means all. . . . so there will still be substantial poverty . . . Even where these goals are achieved substantial poverty will remain – the aim is only to halve poverty, not eliminate it. Current projects suggest there will be around three-quarters of a billion people living on less than a $1 a day in 2015. In absolute terms the number of children dying prematurely in Africa will continue to rise for some years to come. . . . of which an increasing share will be in Africa Progress has been, and will continue to be, slowest in Africa, opening up a growing gap between the region and the rest of the developing world. South Asia, formerly the poorest region, continues to have substantial numbers of undernourished children, and the most children out of school (though recent efforts in India are not captured in the data). And while more poverty will be in urban areas, it will still be predominately rural Urban areas are growing rapidly, and with them slums. So poverty will become more and more an urban issue. But, other than HIV/AIDS, poverty indicators are worse in rural areas in virtually all countries. Since rural residents will remain the majority of the world’s population, the bulk of the poor (60–70 per cent, depending on the indicator) will still be in rural areas in 2015. The nature of problems changes . . . As poverty falls so does its character. The policies required when the bulk of the population are poor differ from those needed when only a minority are affected, especially as many of these remaining poor may share common characteristics (remote, ethnic minority, nomadic, and so on). As under-5 mortality falls it becomes more and more concentrated in the first months or even days, of life. General socioeconomic development and public health interventions can reduce high mortality rates, but once they are lower the required policies become more medically

Projecting Progress

45

intensive. ‘The last 5 per cent’ not in school – children of nomads, parents who think girls should not go to school, street children – are harder to reach, with simply providing schools not being enough. . . . new problems emerge . . . Development brings with it new problems. The demographic transition will increase the share of the elderly in the population. Obesity rises with urbanization. Smoking becomes a major health problem. . . . some remain hidden But some age-old problems – notably the plight of people with disabilities – remain with us and are largely ignored. Finally, we should try to expect the unexpected Predictions of progress assume ‘business as usual’. But shocks of various kinds should be expected, though their precise timing and location cannot be predicted with accuracy. Smaller shocks are frequent, but local in their effects and so with little impact on aggregate trends. Except for protracted conflict, smaller shocks also have short-lasting effects. But many low-income countries are vulnerable to very frequent small shocks, the cumulative impact of which seriously undermines poverty reduction efforts. More seriously still, there are strong grounds for believing there will be a major reversal, possibly related to either global conflict or environmental catastrophe, before the middle of this century.

Appendix: projection methods The value of an outcome indicator, Yi,t , for country or region at time t is given in general by Yi,t = β0,i + β1,i Xi,t

(2.A1)

where X is a vector of determinants comprising one or more variables. Hence the prediction of Y when t is at some point in the future (for example, 2015) involves three unknowns: (i) Model specification: which variables to include in X. (ii) Projections of independent variables: the projected value of X at time t.


(iii) Parameterization: The model parameters (estimation of which is usually based on historical data for X and Y). The general specification given in equation (2.A1) allows the parameter values to vary between countries/regions. Nearly all projections may be written in a form equivalent to equation (2.A1).16 The main exceptions are those based on demographic models, such as those produced for the UN and some analyses of HIV/AIDS, which rely on demographic accounting identities.17 However, even for these exceptions, the formulation in equation (2.A1) identifies the three potential sources of differences in projections. Different model specifications (sets of X-variables) are now discussed in turn. Constant time trend: naïve projections The simplest model is to assume that the future will be like the past, that is, to pass future trends on historical ones. The most common way of doing this is to use a version of equation (2.A1) in which the outcome (Y)-variable is logged and T is simply a time trend: ln Yi,t = β0,i + β1,i T

(2.A2)

so β1 = the annual rate of growth, and β0 is the constant.18 Estimates of β1 may be made in two ways: (i) The most usual method is to estimate the rate of growth based on past data, usually from 1990 to the most recently available year. Hence it is being assumed that the indicator will continue to change at the same rate as it has done since 1990. This approach is implicit in all discussions as to whether countries are ‘on track’ to meet the MDGs, and is the most commonly used with respect to social indicators (for example, the Global Monitoring Report 2005, World Bank 2005a, presentations for primary school completion and under-5 mortality). (ii) Demery and Walton (1998) assume a decline in under-5 mortality of 1.5 per cent a year, which they say has been identified historically as an autonomous element in mortality reduction. These estimates are then augmented to allow for the effects of female education and income. The first of these approaches is labelled as ‘naïve projections’,19 since it takes no account of possible changes in the determinant variables

Projecting Progress

47

between the previous period and the next.20 However, the method does have the virtue of simplicity: (i) there is no need to project the X-variable (since the value of the time trend is of course known in the future); and (ii) the underlying model is easy to grasp: policy-makers can readily understand a statement such as ‘at the current rate of progress, the target of halving the proportion of the people who suffered from hunger in 1990 will not be met by 2015’ (UN MDG Report, Goal 1; United Nations 2005:5). Naïve projections do give information of value. If a country is off track there have to be compelling reasons for believing that performance will change if it is thought the target might still be met. However, they are of dubious value in predicting expected values for future years, which is the main purpose of this report, for variables which are linked closely to underlying determinants. But for variables which can be autonomously driven by policy then naïve projections may prove superior predictors to the models discussed below. Income-based projections The next most common approach, and the dominant one for income poverty projections, is to model Y as depending solely on income (GDP per capita, preferably adjusted for PPP). The form of equation (2.A1) takes logs of both dependent and independent variables: ln Yi,t = β0,i + β1,i ln (INCi,t )

(2.A3)

where INCi,t is GDP per capita for country/region i at time t. Given equation (2.A3), β1 is the elasticity of Y with respect to income, that is the percentage change in Y given a 1 per cent change in income per capita. The elasticity can be obtained from cross-country regressions. This procedure means that a single value of β1 is used for all countries. However β1 may be allowed to vary across subsamples, either by running subsample regressions or else by allowing for a slope dummy for the required categories. For example, Hanmer and Naschold (2000) estimate different income elasticities for income poverty according to the level of inequality. While in principle time series data could be used to produce countryspecific estimates of β1 there are in practice few countries with sufficient time series of the required indicators.21 But in the case of income poverty (see pages 23–8) the elasticity is estimated by the slope of cumulative distribution function at the poverty line. Hence a single household income survey will allow a country-specific estimate of this elasticity.22


This model specification assumes distribution-neutral growth, that is the income of all income groups grows at the same rate (equal to the overall rate of growth). Such occurrences are a historical rarity; while it is difficult to detect any systematic statistical relationship between the rate of growth and inequality,23 this is not the same as saying that distribution does not change during growth episodes. It does, falling in roughly equal measure between growth episodes which are pro-poor (growth of income among the poor exceeds the average) and those which are antipoor (growth of income among the poor is less than the average). Some analyses allow for changes in distribution during growth;24 for example, Demery and Walton calculated the growth required to halve poverty with distribution-neutral growth and, for selected countries, assuming distribution worsens to a specified level (given by the current level in a comparator country). There remains some debate as to the relative role of income versus other factors in determining social indicators. However, the forecaster is not so concerned with channels. The correlation between income per capita and most social indicators is high, with the R2 from the simple regression being typically in the range 0.6–0.8. It is undoubtedly the case that the income term is picking up the effect of other determinants which are correlated with income, such as female education. But for forecasting purposes it is not necessary to separate these effects. (Of course if female education has an effect independent from that of income, and education is not perfectly correlated with income, then the fit and therefore the forecast would be improved by adding this variable.) Given this reasonably high R2 from the simple regression, income-based forecasts can be taken as good first approximations. However, as will emerge in the subsequent discussions, ignoring distributional issues is a major shortcoming.

Adding more explanatory variables Additional explanatory variables will improve the fit of the equation, and so also the reliability of the forecast, provided that the predictions of the X-variables are not too wildly inaccurate. Several studies have forecast various indicators using more explanatory variables. The main example is the World Bank project, Simulations for Social Indicators and Poverty (SimSIP; www.worldbank.org/simsip). This project, which grew out of projections made for the LAC region, provides spreadsheets which can be downloaded and used to make country-level projections of a range of MDG-type indicators. The projections may be

Projecting Progress

49

based on either historical trends (using the best-fitting of four possible ways of fitting a trend) or model-based elasticities, where the independent variables are economic growth, population growth, urbanization and a time trend. Income poverty measures are disaggregated by rural and urban. In addition to income, Hanmer and Naschold (2000) used HIV prevalence in the under-5 and maternal mortality equations, and the number of physicians in the former and literacy in the latter. Demery and Walton used female literacy in their under-5 mortality equation. Multi-equation models Any solvable multi-equation model can, in principle, be written as a reduced-form equation in the syle of equation (2.A1), that is, expressing the outcome as a function of the exogenous variables in the model. In practice, however, such models are run as computer simulations, showing the effect of different assumptions regarding the trajectories of the exogenous variables (and possibly under different parameter assumptions reflecting different policy scenarios). The World Bank has developed a model called MAMS, which is a CGE model incorporating social indicators. At present MAMS has been applied only to the case of Ethiopia. While the CGE approach allows far more detailed modelling, critics of CGE-based analyses are wary of the extent to which the results are driven by model assumptions, which may in the end be derived by fairly crude methods. Having said that, multi-equation approaches can pick up the complementarities between the different indicators, which single-equation estimates will not.25 Other multi-equation models have been used with respect to nutrition. FAO use country-level modelling of agricultural production and trade, from which they calculate nutrition outcomes. Two IFPRI studies have calculated child nutrition outcomes from different models (Smith and Haddad 2002; von Braun et al. 2005). The level of analysis Analysis is preferably carried out at country level, and the result aggregated to present regional and global forecasts. No study can do this in its entirety, since there are countries for which data do not exist. Thus, either implicitly or explicitly, values are assigned to countries for which there are no data, most usually assuming their performance will equal the average for countries in that region (this is unlikely to be a good assumption since countries without data are more likely poor performers).

50 Howard White and Nina Blöndal Table 2.A1

Summary of main approaches to projecting MDG indicators

Model specification Parameterization Naïve projections

Projection of independent variables

Calculation of historical growth rate (1) Elasticities taken from existing studies;

Not necessary (only independent variable is time trend, which is known) Income-based (1) Use of regional growth forecasts by recognized authority (usually World Bank Global Economic Prospects); (2) regression-based (2) regression-based growth estimates estimate based on initial values Income plus models As for income-based As for income-based Multi-equation (1) Regression-based; Assumed values for different models (2) CGE approach policy simulations Source: See text.

The level of analysis matters not only because each country may require different parameters and exogenous values, but because aggregation may conceal country-level limits. This is the case for many of the MDGs for which it is true that the world as a whole is on track to meet the goal simply by virtue of China’s performance. It would therefore be wrong to say that if current trends continue then the goal will be met – not only because it will be met in China but not many other places (though this is true), but because China cannot logically sustain its rate of progress (more than 100 per cent of children cannot go to school, and less than 0 per cent cannot live below the poverty line). Summary Projections may vary because of differences in assumptions regarding (i) determinants, (ii) model parameters or (iii) future values of the determinants. The simplest models – and by far the most common approach – take time as the only determinant. The result tells us if a country is on track or not to meet the relevant goal, which is certainly of policy interest, but of less use in predicting the actual expected value at some point in the future. Since income is highly correlated with most of the outcomes of interest, a first approximation can be given by applying the relevant elasticity to projected growth rates. This is the basis for many projections. However the correlation with income is imperfect, so adding more variables to the right-hand side will help, at least if their future values

Projecting Progress

51

can be predicted with any degree of confidence. It is argued below that distribution is a right-hand-side variable which should not be ignored. More sophisticated versions utilize multi-equation models. There are undoubted advantages to such models, which allow a wider range of policy simulations, but they are quite resource-intensive and assumption-dependent.

Acknowledgements This chapter is based on a report prepared for the Rockefeller Foundation. However, the views expressed here are those of the authors.

Notes 1. The Appendix presents a fuller discussion of the different approaches. 2. They may be reported on an annual basis in sources such as UNDP’s Human Development Report, but this is not the same thing. These figures are based on interpolation between available data points. 3. While the latter is generally a reasonable assumption, the emergence of stipend schemes in recent years has resulted in girls’ enrolment exceeding that of boys in some countries, for example Bangladesh. 4. The same person, Lucia Hanmer, was responsible for the design of the quantitative aspects of both these studies. 5. Forecasting was also carried out by Demery and Walton (1998) from the World Bank using an approach similar to that used in the ISS and ODI studies, though the results were not presented in a way allowing regional tabulations. 6. The elasticity is taken as −0.5, −0.8 and −1.2 for high, middle- and lowinequality countries respectively, where the former has a Gini coefficient of ≥0.54 and the latter one of ≤0.40. 7. These more positive estimates reflect the relative strength of recent growth performance. Making estimates in this way ignores the cyclical nature of growth and possible threats to growth in some regions, such as the growing dominance of Asian economies. 8. The model contains net enrolments, the degree of urbanization and subregional dummies. 9. Infant mortality is the probability of death in the first year of life and under-5 mortality that before the fifth birthday. These rates are expressed per thousand live births. Child mortality is the probability of death between first and fifth birthdays. 10. Given the importance of China, which is not covered by the DHS data, other sources were consulted. The 1997 China Human Development Report gives under-5 mortality in rural area as 71 compared with 21 in urban areas, a ratio of 3.5. This ratio is much larger than the sample maximum (2.2 for Peru), which might just reflect data differences, but may also be a real disparity on account of the strong differences in development in the western provinces


11. 12.

13. 14.

15.

16.

17. 18.

19.

20.

21.

22.

23.

compared with the eastern seaboard and the stronger enforcement of the one-child policy in urban areas (personal communication: Hilary Standing and Gerry Bloom, IDS). The ratio for China is set to 2 for the simulation. UN Millennium Project (2005a). Older people with lower literacy were removed from the population and younger people who have been to school added to it. It is being assumed that all those enrolled are literate on completion. This is certainly a rather heroic assumption, but it is one that already lays behind the literacy data themselves. Literacy is capped at 99.5 in all countries. The forecasts from the latter IFPRI study are not given here, as their results are available only in graphical form. The table shows HIV/AIDS deaths from the UN–DESA Population Division medium variant. No global projections of the number of people living with HIV/AIDS have been found. The HIV/AIDS Surveillance Database of the US Census Bureau (http:// www.census.gov/ipc/www/hivtable.html) reports data as ‘major city’ and ‘outside major city’, which is not the same as urban–rural. Nonetheless, these data show higher incidence in cities than outside in virtually all cases for which there are data. The most sophisticated models are multi-equation models, such as MAMS. However, even these approaches can be written as a single reduced-form equation such as equation (2.A1). There is also an exception when calculating income poverty, which is noted below. An alternative specification would be to leave Y unlogged, which would imply a constant absolute change in Y. This is not a very plausible specification, but is one of the four used by Ramadas et al. (2002) for the World Bank SimSIP project (see pages 48–9). De Onis et al. (2004) use a logistic formulation for their analysis of malnutrition, thus allowing for the fact that the dependent variable is bounded between zero and one. The terminology follows that used in economics; naïve expectations are the belief that the variable of interest will not change value from one period to the next. Since the dependent variable in equation (2.2) is logged, this means that the rate of change of determinants is being assumed constant in making naïve projections, not their level (which would be worse than naïve). Data sources such as World Development Indicators and the Human Development Report give a misleading impression of the availability annual estimates for most indicators. However, these series are based on less than annual data collection. The intervening years are obtained by interpolation. Estimates obtained in this way enable cross-country comparisons on an annual basis, but cannot be used meaningfully for time series analysis. A non-regression-based approach can thus be used for income poverty estimates using household survey data. Assuming distribution neutral growth, the income of all households is increased by the same amount as the overall rate of GDP growth and the poverty measures recalculated. It is the absence of such a relationship which is in fact being demonstrated in the influential Dollar and Kraay paper ‘Growth Is Good for the Poor’ (2004),

Projecting Progress

53

rather than the conclusion, widely drawn, that growth generally (or even necessarily) benefits the poor. See White and Anderson (2001) for an elaboration of this point, and for operationalization of the definitions of pro-poor growth given here. 24. This is different from the approach of Hanmer and Naschold mentioned above (and Hanmer et al. 2000), in which the elasticity varies according to the initial level of inequality. 25. This has elsewhere been called this the ‘basic needs multiplier’ (White 1997) and in more recent work on Ghana identified the importance of the ‘inter-generational multiplier’ whereby parents who have themselves been educated are more likely to send their own children to school (World Bank 2004).

References De Onis, M., M. Blossner, E. Borghi, E. A. Frongillo and R. Morris (2004) ‘Estimates of Global Prevalence of Childhood Underweight in 1990 and 2015’. Journal of the American Medical Association, 291(21):2600–06. Demery, L. and M. Walton (1998) Are Poverty Reduction and Other 21st Century Social Goals Attainable? Washington, DC: World Bank. Dollar, D. and A. Kraay (2004) ‘Growth Is Good for the Poor’. In A. Shorrocks and R. van der Hoeven (eds), Growth, Inequality and Poverty: Prospects for Pro-Poor Economic Development. Oxford University Press for UNU-WIDER. FAO (2002) World Agriculture: Towards 2015/2030. Rome. Hanmer, L. and F. Naschold (2000) ‘Attaining the International Development Targets: Will Growth Be Enough?’. Development Policy Review, 18(1):11–36. Hanmer, L., J. Healey and F. Naschold (2000) ‘Will Growth Halve Poverty by 2015?’ Overseas Development Institute Policy Briefing 8. London. Hanmer, L., N. de Jong, R. Kurian and J. Mooij (1997a) ‘Social Development: Past Trends and Future Scenarios’. Sida Project 2015. The Hague: Institute of Social Studies. Hanmer, L., N. de Jong, R. Kurian and J. Mooij (1997b) ‘Poverty and Human Development: What Does the Future Hold?’ Institute of Social Studies Working Paper 259. The Hague. Hanmer, L., R. Lensink and H. White (2003) ‘Infant and Child Mortality in Developing Countries: Analysing the Data for Robust Determinants’. Journal of Development Studies, 40(1):101–18. Ramadas, K., D. van der Mensbrugghe and Q. Wodon (2002) SimSIP Poverty: Poverty and Inequality Comparisons Using Group Data, Washington, DC: World Bank. Rosegrant, M. W., S. Cline, W. Li, T. Sulser and R. Valmonte-Santos (2005) ‘Looking Ahead: Long-Term Prospects for Africa’s Agricultural Development and Food Security’. International Food Policy Research Institute 2020 Discussion Paper 41. Washington, DC. Smith, L. C. and L. Haddad (2000) ‘Overcoming Child Malnutrition in Developing Countries: Past Achievements and Future Choices’. Food, Agriculture and Environment Discussion Paper. Washington, DC: International Food Policy Research Institute.

54 Howard White and Nina Blöndal UN Millennium Project (2005a) Toward Universal Primary Education: Investments Incentives and Institutions, Millennium Project Task Force on Education and Gender Inequality Report. New York: United Nations. UN Millennium Project (2005b) Taking Action: Achieving Gender Equality and Empowering Women. Millennium Project Task Force on Education and Gender Inequality Report. London: Earthscan. UNAIDS (Joint United Nations Programme on HIV/AIDS) (2005) AIDS in Africa: Three Scenarios to 2025. Geneva. UNFPA (United Nations Population Fund) (no date) ‘The Price of Inaction’. Available at: www.unfpa.org/icpd/inaction.htm United Nations (2005) The Millennium Development Goals Report 2005. New York: United Nations. Von Braun, J. et al. (2005) ‘New Risks and Opportunities for Food Security Scenario Analyses for 2015 and 2050’. International Food Policy Research Institute 2020 Discussion Paper 39. Washington, DC. White, H. (1997) ‘The Economic and Social Impact of Adjustment in Africa: Further Empirical Analysis’. Institute of Social Studies Working Paper 245. The Hague. White, H. and E. Anderson (2001) ‘Growth versus Distribution: Does the Pattern of Growth Matter?’ Development Policy Review, 19(3):267–89. World Bank (2004) Books, Buildings, and Learning Outcomes: An Impact Evaluation of World Bank Support to Basic Education in Ghana. Washington, DC: Operations Evaluation Department, World Bank. World Bank (2005a) Global Monitoring Report 2005. Washington, DC: World Bank. World Bank (2005b) Maintaining Momentum to 2015? An Impact Evaluation of Interventions to Support Maternal and Child Health and Nutrition in Bangladesh. Washington, DC: Operations Evaluation Department, World Bank. World Bank (annual) Global Economic Prospects. Washington, DC: World Bank. Available at: www.worldbank.org/prospects

3 Achieving Health, Wealth and Wisdom: Links between Aid and the Millennium Development Goals David Fielding, Mark McGillivray and Sebastián Torres

Introduction Achieving the MDGs is an ambitious undertaking and one that in planning relies heavily on the developmental impact of aid. The achievement of the MDGs by 2015, or later, requires inter alia a recognition and understanding of their interdependence and the efficient intra-country allocation of development aid. This chapter addresses both issues. Using a new cross-country data set, it empirically examines: (i) the strength of the links between a number of MDG target and related variables, including health, educational status and access to water and sanitation; and (ii) the extent to which aid impacts on these variables. Identification of the key links between the different target variables can help to inform prioritization of the MDGs, by suggesting the areas of endeavour that are likely to have the largest and widest impact. A particular interest of the chapter is whether there is a central or pivotal variable, which, if targeted, leads to the greatest beneficial impact on the remaining variables. Importantly, the chapter’s data set provides information on population subgroups within each country. This allows the chapter also to draw inferences regarding the impact of aid on the poorest groups in each country. Such information is clearly crucial for strategies to move towards achieving the MDGs, especially in countries that are a long way from or not tracking towards achieving the goals. Overall, the chapter seeks not only to provide policy-relevant information on achieving the MDGs, but also to contribute to the literatures on the determinants of cross-country wellbeing achievement and aid effectiveness. This analysis differs in three ways from previous work on the determinants of the cross-country variation in the level of wellbeing. First, we 55

56 David Fielding, Mark McGillivray and Sebastián Torres

model simultaneously five variables capturing different aspects of wellbeing, along with a measure of aid. They relate respectively to the level of material prosperity, the supply of water, educational attainment, fertility and health. One of these variables, the health indicator, is an MDG target variable. Three (water supply, sanitation and educational attainment) are very similar to three other target variables. The water supply variable, for example, is access to piped water. The corresponding MDG target is the proportion of the population with access to an improved water source. Each of these wellbeing variables has the clear potential to impact on each of the others. Modelling all simultaneously permits us to identify the linkages that are the most quantitatively important, including identifying whether there is a pivotal variable, as defined above. Second, we are using a newly compiled data set that reports observations on a wide range of wellbeing indicators for quintiles within developing countries, rather than just overall averages for each country. As a consequence, our model gives equal weight to the wellbeing outcomes of the rich and the poor within a country. Third, our measure of material wellbeing is based on a household survey recording each household’s possessions. We make no reference to per capita income or wealth: instead, our model employs a basic measure of visible material prosperity at the household rather than the personal level. Appropriate for analysing wellbeing in developing countries, this measure is based on the basic assets a household possesses. These assets are basic enough for differences in quality across countries not to be a major worry. Our chapter differs from previous research on aid effectiveness in two main ways. The first is in the use of subnational data to assess aid impact, rather than national averages. This allows us to assess simultaneously the impacts of aid across population subgroups, rich and poor. These impacts have not previously been addressed in empirical research, and the fact that they have not is indicative of a huge void in the aid literature. The second difference is that our chapter provides estimates of a structural model of aid and wellbeing outcomes. This also has not previously been attempted, and allows us to identify a range of direct and indirect aid impacts. There is obvious relevance here for MDG achievement, especially regarding the intra-country allocation of wellbeing aid most conducive to achieving these goals. To outline this chapter, we first review briefly strands of the wellbeing and aid effectiveness literatures,1 before outlining the ways our variables are defined and measured. Next we present descriptive statistics obtained from the data we use, outlining the econometric methods employed to analyse these data, and discuss the results obtained from this analysis.

Aid and the MDGs

57

Then we report results from a number of simulations, mainly involving the effects predicted by our model of increasing aid or increasing its effectiveness. Among the increases under consideration is a doubling of aid levels from those that were provided in the periods under review. Among the chapter’s findings is that child mortality is the central variable, in that a decrease in child mortality leads to the largest beneficial changes in the other MDG or MDG-similar variables under consideration. It is also the variable on which aid has the largest quantitative impact. Our examination further finds that while aid is effective overall, the poorest subgroups within each country are typically not the principal beneficiaries of these inflows.

A brief literature review Economists have long been aware of the importance of links between the various wellbeing dimensions and their implications for poverty. There is a large literature tracing the theory and evidence relating to the ways the income of a population is connected to standards of education and health, and also to fertility. Average standards of education and health are elements of human capital that are likely to determine a region’s overall productivity level, and hence its per capita income. Moreover, with decreasing returns to scale, higher fertility and population growth will result in lower labour productivity. On the other hand, a household’s decisions about human capital investment and the number of children to produce may depend on its current income level, especially with imperfect capital markets (Becker 1981). While the literature on links between wellbeing indicators offers a number of important insights, we believe that it embodies a number of limitations. First, empirical studies relating to the connections between different dimensions of wellbeing focus typically on a single link in the chain. There are studies of the impact of a region’s education on its income (for example, Teulings and van Rens 2003), of income on education (for example, Fernandez and Rogerson 1997), of health on income (for example, Pritchett and Summers 1993), of income on health (for example, Bloom et al. 2004), of fertility on income (for example, Ahlburg 1996) and of income on fertility (for example, Strulik and Siddiqui 2002).2 Many of these studies present careful and compelling evidence on their chosen area of research, but taken as a whole they embody certain limitations. The heterogeneity of statistical methodologies and data sets across these chapters means that they do not shed any collective light on the relative importance of the different causal links in the overall


wellbeing achievement process. It would be useful to know, for example, if any one link is particularly strong, and therefore a potential focus for wellbeing policy interventions. Moreover, while authors are aware of the likely simultaneity of different wellbeing indicators, the focus on a single link in the chain means they never venture beyond an instrumental variables approach to estimation. Such an approach neglects the correlation of errors across equations for different indicators, which may be of interest in itself as well as affecting the statistical efficiency of the estimates. Second, most existing cross-country studies use data on the average value of the wellbeing indicators in each country. The main aim of most empirical economic research has been to explain correlations in these indicators at the national level. Researchers in education and health sciences have often been more sensitive to the drawbacks of such an approach.3 They point out that using mean income places a large weight on the income of the rich, because income distributions are left-skewed, so the mean figure reported for a country is higher than the median. Looking at the link between variations in mean income and, say, variations in infant mortality might be misleading, because high infant mortality is a consequence of the poverty of middle- and low-income groups in a developing country. One way of addressing this problem might be to include a measure of income distribution in the empirical model; however, a more direct approach would be to measure separately the income and health status of the rich and poor within a country. Economists have also long had an interest in aid effectiveness. Dating back to the 1950s, the literature on this topic has been dominated by empirical studies looking at links between aid and per capita income growth. Studies conducted from the mid-1990s generally conclude that per capita growth in recipient countries would be lower in the absence of aid. Some studies find that this outcome is contingent on the quality of recipient country policies, while others point to other contingencies such as recipient country structural vulnerability, institutions, climatic conditions and political stability. The best-known and arguably the most influential aid growth study is Burnside and Dollar (2000). Others include Guillaumont and Chauvet (2001); Hansen and Tarp (2001); Hudson and Mosley (2001); Chauvet and Guillaumont (2002); Collier and Hoeffler (2004); Dalgaard et al. (2004).4 While the recent aid-growth literature has provided a number of useful insights, it must be recalled that in principle aid is primarily about enhancing wellbeing in developing countries, especially that of the poorest people. Recognizing that wellbeing has both economic and

Aid and the MDGs

59

non-economic dimensions, this involves not only increases in incomes but other outcomes, especially in health and education.5 The MDGs are a reaffirmation of these points, given the role that aid is expected to play in their achievement and the range of target variables they identify. Levels of achievement in many wellbeing dimensions tend to be linked. Countries that achieve higher incomes tend to achieve better health and higher levels of education, although this rule of thumb can break down if one looks at specific country groups or subnational data (McGillivray 2005). To this extent one might be able to infer from the recent aid-growth literature that aid has also led to improvements in health and education. However, it would be better to depend not on inferences but on direct evidence, especially given the non-uniformity of income, health and education outcomes. The case for such evidence grows if one accepts the widespread belief that per capita income growth alone is insufficient to attain the MDGs. Another stream within the aid effectiveness literature looks inter alia at the impact of aid on various categories of government expenditure6 This literature is based on a recognition that most aid goes typically to the public sector of recipient countries as an intended augmentation of public expenditure. Better-known studies of this type include Heller (1975); Pack and Pack (1990, 1993); Gang and Khan (1991); Franco-Rodriguez et al. (1998); Feyzioglu et al. (1998). Most of these studies estimate a system of equations that allows for the simultaneous determination of expenditures. The categories of expenditure include those relevant to achievement in non-economic wellbeing dimensions, including health and education. While results do vary between studies, aid is typically associated with increases in expenditure on health and education. Other studies have reported a positive association between aid and broader expenditure categories that include spending on health and education.7 While these results are encouraging, there remains the problem of identifying connections between changes in these expenditures and actual wellbeing outcomes.8 Without considerable additional information this is a highly speculative exercise, given the complexity of links between wellbeing-related expenditures and outcomes such as life expectancy, infant mortality and literacy. Similarly, World Bank (2003) is one of many studies to show that changes in public expenditure on health and education are often weakly related with health and education outcomes. One of the possible reasons is that the efficiency of these expenditures varies among countries. Increasing these expenditures appears to be a necessary but not sufficient condition for improving health and education outcomes. We return to this important issue below.


A much more recent stream in the aid effectiveness literature is not subject to the preceding criticism, as the studies within it do actually look at links between aid and wellbeing outcomes. These studies include Boone (1996); Kosack (2003); Mosley et al. (2004); Gomanee et al. (2005a); and Gomanee et al. (2005b). Each looks at the impact of aid on the infant mortality rate and, with the exception of Boone (1996), the human development index (HDI).9 The majority finding of these studies is that aid is positively associated with national wellbeing outcomes in health and possibly also education. This is perhaps the most encouraging result emanating from the aid literature, given the objective that such transfers are intended to achieve. But this research is subject to a fundamental criticism, one that it shares with the two other streams just discussed: it employs national data, either averaged or aggregated in some way. The primary beneficiaries, in terms of wellbeing outcomes, might not be the poorest within aid-receiving countries. Such an outcome is counter to the principles of aid, and to what the MDGs are intended to achieve. This speculation is fuelled (although clearly not confirmed) by the fact that within many developing countries public spending on health and education, including that on primary health and primary education, has a pro-rich orientation (World Bank 2003). What is required is information on the impact of aid by population subgroup. The literature on aid effectiveness, discussed above, ignores this question.

Data definition and measurement The wellbeing indicators that are the focus of our econometric analysis are taken from the World Bank’s HNP Poverty Data (World Bank n.d.), which aggregates household survey data from 55 countries. Of these countries, 48 are included in our analysis; they are listed in Table 3.1.10 As can be seen from the table, the year of measurement varies slightly from one country to another. An innovative characteristic of the HNP (health, nutrition, population) data set is the way it measures material wellbeing. Its material wellbeing measure is based on the presence or absence of various durable assets in the household, and of certain characteristics of the household’s dwelling place. The assets in question vary from one country to another, depending on the material possessions specific to a certain culture. Every household in the country survey is ascribed a value of zero or one for each asset or dwelling attribute, depending on whether that asset or attribute is present in the household. A household-specific prosperity index is then constructed as the weighted sum of all the binary asset variables.

Aid and the MDGs Table 3.1

61

Countries included in the analysis Survey year

Bangladesh Benin Bolivia Brazil Burkina Faso Cambodia Cameroon C.A.R. Chad Colombia Comoros Côte d’Ivoire Dom. Rep. Egypt Ethiopia Gabon Ghana Guatemala Guinea Haiti India Indonesia Jordan Kenya

2000 2001 1998 1996 1999 2000 1998 1995 1997 2000 1996 1994 1996 2000 2000 2000 1998 1999 1999 2000 1999 1997 1997 1998

Survey year Madagascar Malawi Mali Mauritania Morocco Mozambique Namibia Nepal Nicaragua Niger Nigeria Pakistan Paraguay Peru Philippines Rwanda S. Africa Tanzania Togo Uganda Vietnam Yemen Zambia Zimbabwe

1997 2000 2001 2001 1992 1997 2000 2001 2001 1998 1990 1990 1990 2000 1998 2000 1998 1999 1998 2001 2000 1997 2002 1999

The weights are the coefficients in the first principal component of the whole set of asset variables, scaled so as to sum to unity. (A few of the weights are negative, and in these cases one might conclude that the presence of that characteristic is a sign of poverty.) Households are then ranked by the index and divided into quintiles; average health and education statistics are reported for the households in each quintile. We wish to construct a cross-country measure of material wellbeing. The asset indices reported in the HNP data set are not appropriate for this purpose, because they are based on country-specific sets of assets. Nevertheless, there is a subset of eight assets and attributes common to all countries in the database.11 These are: the presence of an electricity supply; possession of a radio, of a television, of a refrigerator, of a car; access to a flush toilet; use of a ‘bush or field latrine’ (a euphemism for the complete absence of sanitary facilities); and the presence of a dirt or

62 David Fielding, Mark McGillivray and Sebastián Torres Table 3.2

Descriptive statistics for the asset weights

Asset

Mean

Median/mean

Std dev./mean

Electricity Radio Television Refrigerator Car Flush toilet Bush/field latrine (−) Dirt/sand floor (−)

0.149 0.095 0.144 0.146 0.090 0.097 0.128 0.149

1.03 1.01 1.04 1.00 1.07 0.99 1.02 1.07

0.18 0.27 0.11 0.16 0.26 0.38 0.43 0.31

Notes: The numbers in the table are subject to rounding error.

sand floor in the house. The last two of these characteristics are signs of poverty and take a negative weight in all countries. If we look at the relative importance of each of these characteristics in each country, we find very little variation from one country to another. Table 3.2 reports the cross-country means of the weights on the eight characteristics (scaled so that these mean weights sum to unity; subject to rounding error), along with the ratios of each median and standard deviation to its respective mean. The table shows that the standard deviations are quite small, and that the medians are close to the means, indicating an approximately symmetrical distribution. Therefore, we will construct a cross-country asset measure for the kth quintile of the nth country as follows: asskn = h sh · zhkn

(3.1)

where h = 1, . . . , 8 indexes the assets, sh is the weight on the hth asset, taken from the first column of Table 3.2 (h sh = 1) and zhkn indicates the fraction of households in the quintile possessing the asset. In the case of ‘bush latrines’ and dirt floors, zhkn indicates the fraction of houses without the characteristic. As can be seen from Table 3.2, there is not a great deal of variation in the sh , so results from an alternative definition of material wellbeing with ∀hsh = 0.125 yields results very similar to the ones reported below. Our other four endogenous wellbeing outcome indicators capture average levels of sanitation, education, fertility and health of each quintile in each country. Sanitation (wtrkn ) is measured as the fraction of the quintile with access to piped water. (One alternative modelling strategy is to aggregate the lavatorial elements of the assets index with wtr.

Aid and the MDGs 60 ass

150 wtr

40

100

20

50 0.00 sch

0.25

0.50

0.75

1.00

fer

0.00

0.25

0.50

0.75

63

1.00

40

30 20

20

10

mor

0.00

0.25

0.50

0.75

1.00

0

2

4

6

8

50 25

⫺0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Figure 3.1 Frequencies of values of the five wellbeing indicators

However, the resulting measure of sanitation leads to a model that fits the data more poorly. It seems that high-quality lavatory facilities are more an indication of material prosperity than they are of good sanitation.) In the HNP data set, among the measures of educational attainment is the fraction of adults aged 15–49 who have completed grade 5. We denote this measure as schkn . Fertility (ferkn ) is measured in the database as the average number of live births per woman aged 15– 49. A wide range of family health indicators are reported, though not all are reported for every country. In the results reported below, we use the mortality rate for children under five years (morkn ). To summarize, our wellbeing outcome indicators are asskn , wtrkn , schkn , ferkn and morkn . The distributions of these five variables are illustrated in Figure 3.1. We did also consider alternative definitions of material wellbeing, education and health, using (i) uniform asset weights to define material wellbeing, (ii) the fraction of women reading a newspaper at least once a week to measure education and (iii) the mortality rate for children under 12 months to measure family health. The seven alternative regression specifications combining the different measures produced results similar to the ones reported below. Alternative measures of sanitation (for example, the fraction of households with access to a


purpose-built latrine or flush lavatory) produce results broadly similar to the ones here, but with rather higher standard errors. It can be seen from Figure 3.1 that the wellbeing indicators are not normally distributed. In the case of ass, wtr and sch there are some observations close to the theoretical bound where all of the households in the quintile possess all of the assets (or, at the other extreme, none of them) in the material wellbeing index; some observations record quintiles where all or none of the households have access to piped water or primary education. This problem does not occur for the other two variables, fer and mor, but in the latter case the distribution does seem to be left-skewed. For this reason the ass, wtr and sch equations are fitted in Probit form, as described below. The fer and mor equations will be fitted in log-linear form. In order to identify the impact of one wellbeing indicator on another, we need to include a range of exogenous conditioning variables in our regression equations. Restrictions on the coefficients of the conditioning variables will permit us to identify the links between the wellbeing indicators. Note that these exogenous national characteristics vary across countries but not across quintiles within a country. We include in our model variables to capture factors relating to (i) geography, (ii) history and (iii) culture. Data sources for these variables are listed in the Appendix. Included in (i) are: the country’s surface area in square kilometres (size), a logarithmic measure of the value of its natural resources in US dollars (natres), a dummy for whether it has a maritime coastline (coast), its mean annual temperature in 0.1 degrees centigrade (temp), and the fraction of the population at risk from malarial infection (malfal). Given its significance in previous studies (for example, Easterly and Levine 1997), we also include a dummy for countries in Africa (africa). Included in (ii) are dummy variables for whether the country was colonized by Great Britain (britain) or by France (france). Included in (iii) are the fraction of the population that are Christian (chrs), the fraction that are Muslim (mus) and an index of ethno-linguistic fractionalization (ethno). We introduce into our analysis a measure of the level of aid to a country, aid, to test whether aid impacts on the above-mentioned endogenous wellbeing variables. This variable is measured as the average ratio of net ODA disbursements to GDP in the five years up to and including the measurement of the wellbeing indicators. Data are taken from the OECD–DAC online database (OECD 2005). There are many reasons why aid might influence these variables, including: (i) its economy-wide impact on growth; (ii) its impact on government expenditure on health,

Aid and the MDGs

65

education and water (in cases where these expenditures are productive); (iii) aid-funded projects that boost private incomes and support the provision of public goods and services; (iv) donor initiatives to improve the productivity of the above expenditures through the provision of technical assistance and capacity-building; and (v) conditions attached to aid inflows requiring recipients to pay more attention to health, education and water in their own policy agendas. On the basis of these factors, combined with evidence from the existing literature discussed above that the effects mentioned in points (i) and (ii) are in general positive, our expectation a priori is that aid will have a positive impact on the endogenous wellbeing variables. We have no firm expectations a priori regarding the relative impacts of aid on each population subgroup. One might expect that the primary beneficiaries of aid are the poorest quintiles, especially when the interaction between donors and recipient governments encourages the latter to give the wellbeing of the poorest a greater weight in policy. However, given the range and complexity of aid effects, it remains uncertain whether these groups will benefit most from aid. Finally, we need an instrument for our potentially endogenous aid variable, in line with the recent literature on aid. The chosen instrument, discom, is the ratio of ODA disbursements to ODA commitments over the five years before the first year of measurement of aid. An ODA commitment is the amount of funds donors make available to recipients for disbursement. Recipients that do not fully disburse commitments in the current year can be expected to receive less aid in the subsequent year. More generally, if a smaller proportion of the ODA commitment is disbursed in period t − 1, less aid will be provided in period t. All data required to calculate discom were taken from OECD (2005).

Empirical analysis Descriptive statistics Table 3.3 provides data on the unconditional correlations of the wellbeing indicators, again disaggregating by quintile. The signs on individual correlation coefficients are what one might expect. Assets, water and education (the ‘goods’) are positively correlated; fertility and child mortality (the ‘bads’) are also positively correlated. Correlations across these two pairs are always negative. The correlations are generally highest for quintiles 4–5 (the richest) and lowest for quintiles 1–2 (the poorest). This suggests that the variation in outcomes for richer households is more systematic, and may be more closely correlated with observable

66 Table 3.3 Summary statistics Means

ass

wtr

sch

fer

mor

Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5

0.1595 0.2511 0.3379 0.4481 0.6534

0.0411 0.1054 0.1551 0.2820 0.5277

0.3075 0.4083 0.4974 0.6110 0.7851

6.0625 5.3417 4.8542 4.2646 3.2146

0.1781 0.1682 0.1556 0.1294 0.0876

Standard deviations Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5

0.1175 0.1610 0.2013 0.2250 0.1822

0.1322 0.2289 0.2711 0.3150 0.3109

0.2224 0.2592 0.2691 0.2580 0.1642

1.3731 1.4158 1.5492 1.5444 1.1897

0.0914 0.1013 0.1023 0.0929 0.0582

Quintile 1 correlations

ass

wtr

sch

fer

wtr sch fer mor

0.7932 0.4772 −0.3617 −0.5057

0.2931 −0.1536 −0.3641

−0.2518 −0.5494

0.4271

Quintile 2 correlations wtr sch fer mor

0.7703 0.6525 −0.5455 −0.6679

0.3558 −0.3125 −0.4482

−0.4803 −0.6694

0.6669


0.8205 0.7197 −0.6858 −0.7425

0.4443 −0.4690 −0.5123

−0.6550 −0.7393

0.8126


0.8391 0.7498 −0.7453 −0.7874

0.5373 −0.5862 −0.6281

−0.7558 −0.7893

0.8747


0.8285 0.7128 −0.6536 −0.8270

0.5697 −0.5301 −0.7640

−0.7281 −0.8186

0.8613

Aid and the MDGs

67

independent characteristics; the variation among poorer households may have a larger stochastic element. These characteristics indicate that in our econometric model it would be unwise to try to impose a priori any structure on the covariance matrix of residuals for each wellbeing indicator and each quintile. Variances and covariances are unlikely to be uniform across quintiles, let alone across indicators. Outcomes at the upper end of the assets distribution are likely to be somewhat more predictable than those at the lower end. Model structure The descriptive statistics suggest strong interrelations between our four wellbeing indicators. However, the descriptive statistics also suggest that conditional variances are unlikely to be constant across indicators or across quintiles, and it would be unwise to make any assumptions a priori about the corresponding covariances. So our model will take the following general form. Let the jth wellbeing indicator for the kth quintile in the nth country (j = 5, k = 5, n = 48) be denoted yjkn . Then our regression equation for the jth indicator is yjkn = F(αjk + i=j βij · yikn + p ϕjp · xnp + θj ln aidn ) + ujkn

(3.2a)

for j = (ass, wtr, sch) and ln (yjkn ) = αjk + i=j βij · yikn + p ϕjp · xnp + θj ln aidn + ujkn

(3.2b)

for j = (fer, mor). F(.) is the Normal cumulative density function. xnp is the value of the pth exogenous conditioning variable in the nth country and ujkn is a residual. Our aid equation is ln (aidn ) = αAID + p ϕAIDp xnp + θAID · ln discomn + uAIDn

(3.3)

A priori restrictions on the φjp -coefficients allow us to identify (most of) the βij -coefficients that capture the interactions between the five wellbeing indicators. We allow the conditional cross-country mean of each wellbeing indicator, αjk , to vary across quintiles, so that we are in fact fitting a fixed-effects model. We have 5 × 5 × 48 = 1200 observations of yjkn , and hence 1200 observations of the residuals ujkn . We do not wish to assume any restriction on the correlation of residuals across indicators or across quintiles, so the model is fitted by stacking 26 regression equations – one for each j and each k, plus one for aid – and estimating the coefficients in each equation simultaneously by 3SLS. With only 48

68 David Fielding, Mark McGillivray and Sebastián Torres Table 3.4 Variable definitions and model structure y-variables ass wtr sch fer mor aid

The assets index The fraction of households with access to piped water The fraction of household members with primary education Live births per woman Under–5 mortality rate Ratio of aid to GDP

x-variables for: africa dummy = 1 if in Africa britain dummy = 1 if colonized by Britain france dummy = 1 if colonized by France ethno ethno-linguistic fractionalization index ln(siz) log country surface area ln(nat) log natural resource capital value coast dummy = 1 if country has a coastline chrs fraction of the population that is Christian musl fraction of the population that is Muslim temp temperature (in 0.1 degrees C) tems tmp2 /100 malfal fraction of population at risk from malaria

Appearing in the equations ass wtr sch fer mor ass wtr sch fer mor ass wtr sch fer mor ass wtr ass ass ass

wtr wtr wtr

mor

sch

fer

sch

fer mor mor mor

Note: All of the x variables also appear in the aid equation, along with the disbursement– commitment ratio discom.

countries, we do not have enough degrees of freedom to allow the slope coefficients (βij , φjp ) to vary across quintiles, so each of these should be interpreted as the mean effect of a particular explanatory variable across all countries and all quintiles. It is possible to fit a quintile-specific model, but with 48 observations, standard errors on individual coefficients are so high as to preclude much economic interpretation. Identification of the β-coefficients requires some a priori restrictions on the φ-coefficients. These restrictions, summarized in Table 3.4, are as follows. First, some of the geographical characteristics are unlikely to have a direct impact on anything other than material resources (ass, wtr) through an effect on factor productivity. These characteristics are country size (size), natural resource wealth (natres) and ethno-linguistic fractionalization (ethno). Similarly, other geographical characteristics

Aid and the MDGs

69

are unlikely to have a direct impact on anything other than health. These characteristics are temperature (temp)12 and malaria risk (malfal). Whether a country has a coastline (coast) might affect health and material prosperity, but it is unlikely to affect education or fertility directly, and so it can be excluded from the equations for these two indicators. These restrictions together allow us to identify the effects of material wellbeing (ass), water access (wtr) and health (mor) in each of the other four equations, except that the effects of ass on wtr and of wtr on ass are unidentified. The effects of fertility (fer) and education (sch) in the assets, water and health equations are identified by assuming that religious adherence, as captured by chrs and musl, has no direct effect on assets and health. However, it might affect attitudes towards contraception or the value of education (especially female education), and so have a role in determining fer and sch. The other effects we do not attempt to identify – because of an absence of any obvious instrument – are of fer in the sch equation and of sch in the fer equation. This will turn out to be important when we come to interpret the coefficients in these two equations. Equation (3.3) allows us to identify the effects of aid in all five of the wellbeing indicator equations. The θj-coefficients will then indicate the impact of an increase in aid on each indicator, holding the other indicators constant. That is, they will determine the partial derivatives ∂yj /∂ ln (aid). (However, in the case of ass, wtr and sch the partial elasticities will not be constant because of the logistic transformation.) In addition, we can solve the equations and calculate the total derivative effects dyj /d ln (aid). Again, these derivatives will not be constant, but we are able to use the model to predict the consequence for each indicator in each quintile in each country of an increase in aid to that country by a certain percentage amount, starting at the observed levels of aid and of the wellbeing indicators. Regression results The regression results for the four wellbeing indicators are reported in Tables 3.5 and 3.6. Note that the first three equations are in Probit form,13 so that the coefficients tell us, approximately, how proportionate changes in the explanatory variables translate into proportionate changes in y/(1 − y). The other equations are log-linear, so that coefficient values translate into proportionate changes in y. The first part of Table 3.5 shows the β- and φ-coefficients in the logistic ass equation, along with corresponding standard errors and t-ratios. The significant φ-coefficients are those on size, coast, africa and ethno. Ceteris paribus,

70 Table 3.5 Fitted regression coefficients coeff.

Std error

t-ratio

ass equation natres ethno size coast africa britain france sch fer mor aid

0.0130 −0.2025 0.0489 0.1025 −0.1659 −0.0047 0.1067 0.8900 0.1068 −0.4146 0.0697

0.0225 0.0988 0.0222 0.0474 0.0719 0.0615 0.0845 0.1793 0.0780 0.0448 0.0393

0.5786 −2.0500 2.2032 2.1640 −2.3071 −0.0770 1.2635 4.9640 1.3701 −9.2497 1.7732

wtr equation natres ethno size coast africa britain france sch fer mor aid

−0.0561 −0.5033 0.2920 0.0463 −0.2788 0.5123 0.5081 −0.4526 0.7836 −1.5108 0.4437

0.0658 0.3358 0.0600 0.1343 0.2609 0.2357 0.2799 0.3975 0.2159 0.1526 0.1182

−0.8528 −1.4989 4.8688 0.3451 −1.0687 2.1734 1.8156 −1.1384 3.6301 −9.9025 3.7524

sch equation chr mus africa britain france ass wtr mor aid

0.3867 −0.2141 0.1136 0.5751 0.0545 2.3295 −0.7767 −0.3486 0.0423

0.1150 0.0873 0.1030 0.0725 0.0691 0.2288 0.1050 0.0546 0.0436

3.3640 −2.4526 1.1029 7.9383 0.7890 10.1799 −7.3970 −6.3894 0.9696

fer equation chr mus africa britain france ass wtr mor aid

0.4504 0.2654 −0.1506 0.1188 0.0578 −0.8459 0.1393 0.3700 0.1037

0.0421 0.0375 0.0533 0.0518 0.0428 0.1167 0.0802 0.0313 0.0318

10.7058 7.0833 −2.8236 2.2931 1.3514 −7.2497 1.7360 11.8133 3.2605 (Continued)


71

(Continued) coeff.

Std error

t-ratio

mor equation temp tems/100 malfal coast africa britain france ass wtr sch fer aid

−0.4122 0.0766 0.4375 −0.1305 0.1770 0.1439 0.0809 0.0930 −0.5388 −1.0602 0.6854 −0.1208

0.2360 0.0583 0.0705 0.0389 0.0840 0.0817 0.0878 0.2642 0.1097 0.1619 0.0931 0.0464

−1.7462 1.3152 6.2065 −3.3551 2.1078 1.7618 0.9213 0.3519 −4.9106 −6.5503 7.3647 −2.6060

aid equation natres ethno size coast chr mus temp tems/100 malfal britain africa france discom

0.2672 0.1513 −0.3787 −0.2652 0.1304 0.4902 −1.6383 0.2680 0.6492 0.2348 −0.1308 −0.1795 0.4436

0.0675 0.4695 0.1083 0.1609 0.4248 0.2555 1.4365 0.3412 0.3164 0.2717 0.3356 0.3613 0.2031

3.9566 0.3222 −3.4955 −1.6480 0.3069 1.9188 −1.1405 0.7854 2.0520 0.8644 −0.3898 −0.4969 2.1843

large countries with a maritime coastline can be expected to have a higher level of material prosperity. Ethno-linguistic diversity and location in Africa have a negative impact on material prosperity, as in Easterly and Levine (1997). Two of the three identified β-coefficients are large and statistically significant. As expected, better standards of education (higher sch) and health (lower mor) lead to higher levels of material prosperity: this is the human-capital effect. An increase of 1 percentage point in the fraction of household members with primary education can be expected to raise ass/(1 − ass) by just under 1 per cent. A 1 per cent reduction in child mortality can be expected to lower ass/(1 − ass) by just under 0.5 per cent. These effects do not take into account any feedback from the effects of higher material prosperity on education and health, which is discussed later.

72 Table 3.6 Main econometric results (a) Regression R2 statistics

ass wtr sch fer mor aid

Quintile 1

Quintile 2

Quintile 3

Quintile 4

Quintile 5

0.4600 0.4440 0.6469 0.4410 0.6912

0.6805 0.3202 0.7838 0.6534 0.8528

0.7397 0.3234 0.8545 0.7957 0.8770 0.4320

0.7980 0.5124 0.8592 0.8098 0.8866

0.7166 0.6704 0.8480 0.7719 0.9113

(b) Regression standard errors

ass wtr sch fer mor aid

Quintile 1

Quintile 2

Quintile 3

Quintile 4

Quintile 5

0.0901 0.1004 0.1319 0.2076 0.3159

0.0930 0.1984 0.1234 0.1871 0.2457

0.1038 0.2376 0.1048 0.1660 0.2582 0.7862

0.1010 0.2223 0.0966 0.1760 0.2681

0.0970 0.1809 0.0641 0.1758 0.2279

(c) Cross-quintile averages of cross-variable residual correlations

wtr sch fer mor

ass

wtr

sch

fer

0.6363 −0.4511 0.3470 0.1632

−0.0231 0.0993 0.3121

−0.2774 0.3273

−0.5908

(d) Within-variable residual correlations

ass

wtr

sch

Quintile 1

Quintile 2

Quintile 3

Quintile 4

Quintile 2 Quintile 3 Quintile 4 Quintile 5

0.8807 0.6502 0.2000 0.0148

0.8756 0.4934 0.2487

0.7617 0.4746

0.8141


0.7876 0.7975 0.5862 0.2343

0.9107 0.7083 0.4440

0.7817 0.4716

0.6498

Quintile 2 Quintile 3

0.8947 0.7603

0.9082 (Continued)


fer

mor

73

(Continued) Quintile 1

Quintile 2

Quintile 3

Quintile 4

Quintile 4 Quintile 5

0.5552 0.4061

0.6364 0.4832

0.8128 0.6135

0.7768


0.8232 0.6786 0.5435 0.1627

0.8673 0.7500 0.3731

0.8059 0.5271

0.6933


0.7586 0.6003 0.5321 0.3678

0.7424 0.6263 0.3120

0.7643 0.5103

0.5475

The coefficient on fer is positive. Although the coefficient is statistically insignificant, its sign contradicts both received wisdom and the negative unconditional correlation between ass and fer. One explanation for the positive coefficient (or at least, for an effect offsetting the usual negative impact of high fertility, leading to a coefficient close to zero) is that for a given level of education and health, higher fertility leads to larger households, and larger households are able to acquire more assets. It is not possible to test this hypothesis directly, because household size is not reported in the data set. This is magnified if there are scale economies in some types of household production. None of this implies that higher fertility is good for material prosperity in equilibrium, because – as we see shortly – higher fertility could be bad for education and health, and therefore bad for material prosperity overall. The second part of Table 3.5 reports the results for the wtr equation. Here, the significant φ-coefficients – all positive – are those on size, britain and france. The first of these might reflect the fact that ass is not identified in the wtr equation: it is likely that access to piped water is more extensive in countries where households have access to other material assets. The second two suggest some positive influence of British and French colonization. There are significant β-coefficients on fertility and mortality, but not on schooling. Households with poorer health outcomes are less likely to be able to afford piped water. A 1 per cent increase in the mortality variable is associated with a fall in w/(1 − w) of around 1.5 per cent. More surprisingly, a 1 per cent increase in fertility is associated with an increase in w/(1 − w) of around 0.75 per cent. One explanation for this


effect is household economies of scale, as discussed above. Another is that households with an unexpectedly large number of children value sanitation more at the margin, and use resources that would otherwise be spent elsewhere to gain access to clean water. The third part of Table 3.5 reports the results for the sch equation. Here, the statistically significant φ-coefficients are those on chrs, musl and britain. Ceteris paribus, countries with a relatively large Christian population and those colonized by Britain can expect to have relatively high education levels. All three of the identified β-coefficients are large and statistically significant. On average, more prosperous households invest in more education: a rise in the assets index by 1 percentage point is associated with a level of sch/(1 − sch) about 2 per cent higher. But for a given level of assets, healthier households also invest in more education. A 1 per cent reduction in mor is associated with a level of sch/(1 − sch) that is about 0.3 per cent higher. Why is this so? Surely education and healthcare make competing claims on household resources? One reason for this is that caring for the sick and dying takes up time that would otherwise be spent learning. Another is that a high rate of child mortality reflects the poor health status of the parents, in whose education few resources have been invested because sickness reduces the returns to schooling. Unfortunately, there is no information in the data set that would shed light on which of these reasons is more important. By contrast, an increase in wtr by 1 percentage point reduces sch/(1 − sch) by about 0.75 per cent. This may reflect that fact that, ceteris paribus, a higher level of expenditure on sanitation leaves fewer resources for education. Another potential explanation is that, for a given level of material prosperity and health, households that already have access to clean water are more likely to choose a large number of children, and that a higher fertility rate is associated with less education. The impact of wtr in the fer equation is discussed below; remember that fer is not identified in the sch equation. The fourth part of Table 3.5 reports the results for the fer equation. Here, the statistically significant φ-coefficients are those on chrs, musl, africa and britain. Fertility is higher in countries with large Christian and Muslim populations, especially if they were British colonies, but, given these characteristics, lower in Africa than elsewhere. There is a large and significant β-coefficient on mor. That is, a higher level of child mortality leads to a higher fertility rate: to some extent, parents will seek to replace the children they have lost. A 1 per cent increase in mortality leads to an increase in fertility of around 0.3 per cent. There is also a negative coefficient on ass, a 1 per cent increase in the assets index being associated with a reduction in fertility of just under 1 per cent. On average, families

Aid and the MDGs

75

with more material assets have fewer children. However, a percentage point increase in the fraction of houses with piped water increases fertility slightly. For a given level of assets and a given child mortality rate, better sanitation facilities are associated with a decision to have more children. This suggests that better sanitation reduces the costs of bearing and raising an infant. It also ties in with the negative coefficient on wtr in the sch equation, discussed above. The fifth part of Table 3.5 reports the results for the mor equation. Here, the statistically significant φ-coefficients are those on coast, africa, britain and malfal. Mortality rates are higher in countries with a climate favourable to malaria-bearing mosquitoes but lower in a maritime environment. Given these characteristics, they are still higher in Africa. (Here is the part of explanation for the negative Africa dummy in growth regressions: Africa is not just unusually inefficient, it is also unusually unhealthy.) Three of the four β-coefficients are statistically significant. Higher levels of sanitation and education are associated with lower mortality rates. The first effect reflects the importance of sanitation for health: a percentage point increase in the fraction of houses with piped water reduces mortality by around 0.5 per cent. The second reflects either the complementarity of investment in education and investment in health, or a beneficial effect of education on household hygiene and therefore health outcomes. A percentage point increase in the fraction of households with primary education leads to a reduction in mortality of around 1 per cent. Finally, a 1 per cent increase in fertility increases child mortality by around 0.67 per cent. A higher birthrate increases the risks facing each individual child. Conditional on these effects, variations in the level of household assets have no impact on mortality. Table 3.6 presents some descriptive statistics for the Table 3.5 model. Parts (a) and (b) of Table 3.6 show that the model explains a relatively small fraction of the sample variation in the characteristics of households at the bottom of their national asset distributions (quintiles 1 and 2), and a relatively large fraction of the corresponding variation for their more prosperous neighbours (quintiles 4 and 5). This difference is manifested in a systematic pattern in the R2 statistics for our 20 regressions, and consequently in a systematic pattern in the corresponding equation standard errors, which are lower for the higher-asset quintiles. Non-modelled country-specific effects play a larger role in determining the outcomes for the poor than they do in determining the outcomes for the rich. The reasons for this discrepancy are an important subject for future study. Parts (c) and (d) of Table 3.6 report some of the residual correlations from the fitted model. Part (c) reports the average value of between-indicator

76 David Fielding, Mark McGillivray and Sebastián Torres Table 3.7 Equilibrium effects on each variable of 1 standard error shock to each equation Equation

ass increase Std deviation wtr increase Std deviation sch increase Std deviation fer decrease Std deviation mor decrease Std deviation Sum

ass ↑ 1 s.e.

wtr ↑ 1 s.e.

sch ↑ 1 s.e.

fer ↓ 1 s.e.

mor ↓ 1 s.e.

0.1988 0.0658 0.0732 0.0669 0.1776 0.0933 0.3153 0.1225 0.4254 0.2015 1.1903

−0.0157 0.0194 0.2055 0.0577 −0.0630 0.0371 −0.0336 0.0392 0.0224 0.0621 0.1156

0.1250 0.0690 0.0890 0.0733 0.2339 0.0963 0.2660 0.1267 0.4666 0.2052 1.1805

0.0524 0.0309 0.0429 0.0363 0.0676 0.0401 0.3404 0.0581 0.3232 0.0905 0.8265

0.1334 0.0678 0.1892 0.1252 0.1414 0.0824 0.3650 0.1238 0.7529 0.2116 1.5819

Note: Each column relates to a shock to a particular equation, and each row to the impact on a particular variable. Figures in bold record the mean effect across all countries and quintiles; figures in italic record the corresponding standard deviations.

correlations. These are generally quite small, so there is no substantial unexplained co-movement in our wellbeing indicators. However, in part (d) we see a positive correlation coefficient for the individual dependent variables across quintiles, which suggests that random variations in country-specific characteristics do play a role in determining outcomes for each particular wellbeing indicator, conditional on the observed levels of the others. Some implications of the results The φ-coefficients in Table 3.5 indicate the partial derivatives of each wellbeing indicator with respect to the others. They generally show beneficial associations: that is, improvements in one indicator are associated typically with improvements in the others. (Remember, of course, that in the case of fertility and mortality, improvements correspond to a fall in the indicator.) However, there are a number of exceptions to this general pattern. In particular, wider access to piped water appears to be associated with higher fertility rates and less education. Table 3.7 provides some information on the relative importance of the different interactions. It shows the impact on every variable in the system (ass, wtr, sch, fer, mor) of a shock to each equation in turn.

Aid and the MDGs

77

The effects reported are equilibrium effects, allowing for all of the interactions between the different wellbeing indicators. The shocks are normalized on the standard errors in the Table 3.5 equations. This normalization is an important part of the interpretation of the results: we are using the equation standard errors as an indication of the conditional variation that we can typically expect in each variable. In all cases we consider ‘beneficial’ shocks, so the shocks to the ass, wtr and sch equations are positive, while the shocks to the fer and mor equations are negative. The effects recorded are the predicted increases in ass, wtr and sch, and the predicted reductions in fer and mor. Each column in the table corresponds to a certain shock, while each row corresponds to a certain variable affected by the different shocks. In this context, a shock can be interpreted as an idiosyncratic effect that makes one of the wellbeing indicators for a particular quintile in a particular country better than one could expect, given the characteristics of that quintile and that country. The nonlinearity of the model means the effects depend on the initial values of the variables. For the purposes of Table 3.7, we compute the effects in 240 different cases; in each case, as starting values we use the observations actually recorded for one point in our data set. The main figures in Table 3.7 are means for the 240 observations; the table also includes the corresponding standard deviations. It can be seen from the table that, in general, the shocks with the largest beneficial effect are those to the mortality equation. The effects of shocks to the ass and sch equations are a little smaller – about three-quarters the size of the mor effects on average. The effects of shocks to the fer equation are slightly smaller again. However, the outstanding feature in the table is the impact of shocks to the wtr equation. As we have already seen, improvements in access to piped water are associated with higher fertility and lower standards of schooling. This means that in equilibrium, a beneficial shock to the wtr equation actually worsens the schooling and fertility outcomes on average, and as a consequence also the material assets outcome: the relevant figures in the wtr column in Table 3.7 are negative. Nevertheless, the detrimental feedback between wtr, sch and fer is offset by beneficial feedback between wtr and mor, so all of the equilibrium effects of a shock to wtr on the other variables are close to zero on average. Improving access to piped water may be valuable in its own right, but it does not seem to facilitate other dimensions of wellbeing, on average. By contrast, anything that helps to generate idiosyncratic improvements in one of the other wellbeing indicators will, on average, improve all of the rest; these effects are largest for improvements in mortality. The conflicting feedback effects in wtr also explain why the figures


in the wtr row in Table 3.7 are smaller than those in other rows: beneficial shocks to other variables do not improve water access so much, on average. The impact of aid In this subsection we investigate what the fitted model tells us about the changes in wellbeing outcomes we can expect to see if the amount of aid allocation to each country increases. However, all of the following remarks apply equally as predictions of the consequences of an improvement in the ‘effectiveness’ of a given amount of aid, which would be captured by a proportional change in the size of the aid coefficient in each equation. While the impact of aid on wellbeing outcomes is an interesting topic in its own right, we are specifically interested in whether there is empirical support for the international community’s strategy of achieving the MDGs through, inter alia, increased aid flows. Such support does appear to exist, based on the results we now report and discuss. Table 3.5 shows that there is a statistically significant direct (partial derivative) effect of aid on 4 out of the 5 wellbeing indicators; the exception is schooling. Ceteris paribus, a 1 per cent increase in aid will raise ass/(1 − ass) by somewhat less than 0.1 per cent and wtr/(1 − wtr) by about 0.4 per cent; it will reduce mor by a little over 0.1 per cent. Household assets, sanitation and mortality do improve significantly if aid rises. However, there is also a 0.1 per cent increase in fer: the extra income associated with aid inflows tends to raise birthrates on average. To the extent that higher fertility is detrimental to other wellbeing objectives, this effect will tend to undermine the effectiveness of aid flows. The estimates of these direct effects do not take into account any of the beneficial or detrimental interactions between the wellbeing indicators. In order to see how the interactions are likely to affect the consequences of an increase in aid, we need to solve out the Table 3.5 model. Again, it is important to note that the nonlinearity of the equations means that the impact of an increase in aid will not be uniform across all quintiles and countries. The magnitude of the response of the different wellbeing indicators will depend on their initial values. For this reason, Table 3.8 shows the predicted change in each of our wellbeing indicators with a doubling of aid in each quintile and in each country. In this table, each figure measures the difference in percentage terms14 between the current value of the indicator and the value predicted with a doubling of aid in the solved-out model. The table shows improvements in ass, wtr, sch and mor in all 240 cases. Schooling improves as a result of improvements in

Table 3.8

Predicted percentage change in each variable for a 100 per cent increase in aid Assets Q1

Q2

Bangladesh 2.8 6.0 Benin 1.7 3.4 Bolivia 5.7 10.1 Brazil 5.2 10.4 Burkina 1.0 1.6 Cambodia 3.6 6.3 Cameroon 2.7 7.2 C.A.R. 1.9 3.6 Chad 1.7 2.0 Colombia 9.0 7.7 Comoros 2.5 3.7 Côte d’Ivoire 2.2 4.7 Dom. Rep. 8.4 10.0 Egypt 6.9 9.7 Ethiopia 1.5 2.0 Gabon 11.9 10.4 Ghana 3.9 7.8 Guatemala 4.8 7.2 Guinea 1.4 2.3 Haiti 2.8 5.4 India 2.6 7.0 Indonesia 6.3 13.4 Jordan 8.3 5.9 Kenya 6.8 9.4

Water Q5

Q1

Q2

Schooling

Q3

Q4

Q3

Q4

Q5

Q1

11.2 5.4 8.4 8.1 2.5 9.1 12.4 6.4 2.9 5.2 8.5 7.1 8.3 8.7 2.3 8.2 8.8 9.8 4.2 8.6 11.5 11.7 4.5 9.2

10.7 5.7 1.5 7.7 17.2 24.6 25.9 4.2 9.5 7.5 0.7 5.7 8.4 20.8 27.4 2.6 4.7 2.3 8.8 27.5 27.0 15.8 4.3 5.9 5.4 4.1 0.6 9.0 16.9 25.2 24.9 8.5 3.5 10.1 0.4 2.0 5.6 8.0 26.5 1.5 10.3 5.0 2.7 14.6 24.2 28.8 20.2 4.1 9.8 5.5 0.6 7.7 17.3 21.5 25.3 4.0 8.0 9.4 0.9 6.5 13.8 20.7 25.4 3.1 4.7 9.5 3.2 4.1 7.2 14.6 29.1 1.9 2.8 2.2 22.8 27.6 22.0 8.8 8.8 6.4 11.4 9.1 0.9 3.3 8.5 23.7 23.8 3.2 9.1 7.3 1.9 12.0 20.8 27.5 24.8 2.8 5.6 3.0 2.4 9.9 16.6 25.4 17.4 12.2 5.7 2.1 23.9 25.5 19.7 9.4 1.0 4.2 3.8 12.3 0.5 1.4 1.2 3.3 14.9 2.3 7.4 4.2 9.4 13.4 16.4 18.7 20.8 14.6 7.2 4.8 2.3 12.1 12.8 20.5 22.0 6.4 6.7 3.5 10.9 25.5 25.8 25.7 18.1 4.5 8.6 9.2 0.1 1.7 5.2 13.6 24.2 1.6 11.1 6.3 2.4 12.8 16.8 25.5 24.8 3.0 8.7 4.7 0.4 4.8 9.8 20.3 24.4 4.5 7.0 3.1 1.1 16.3 21.4 26.8 13.1 7.5 3.6 1.8 24.3 7.7 5.9 4.5 0.6 4.5 8.2 5.1 3.2 12.2 24.1 26.8 21.0 10.0

Fertility

Q2

Q3

Q4

Q5

Q1

Q2

6.9 3.9 7.2 11.2 2.0 4.8 8.3 4.4 2.0 3.5 3.6 4.2 10.1 6.7 2.5 10.0 7.6 4.0 2.5 4.2 9.6 14.5 4.5 10.1

11.5 8.5 1.4 0.1 −5.3 6.1 8.4 3.3 2.2 −0.7 4.4 1.8 1.4 −4.6 −11.5 5.7 1.1 0.2 −4.7 −12.5 2.5 3.5 7.5 3.6 2.5 6.3 7.1 2.3 −0.8 −5.1 13.0 7.6 1.2 0.4 −7.3 6.2 5.4 6.9 1.7 −1.3 2.4 3.2 5.9 2.4 1.9 1.7 1.3 0.8 −9.5 −7.0 9.4 10.0 6.5 1.0 −0.7 5.3 5.9 3.6 1.4 −2.8 6.0 1.4 0.4 −10.2 −11.5 6.5 4.5 1.6 −6.2 −10.6 2.9 4.6 12.8 2.6 1.9 5.8 4.3 0.7 −15.9 −12.1 8.0 3.3 0.6 −2.3 −8.0 7.1 2.4 0.4 −3.0 −6.5 4.4 8.8 6.1 3.1 1.5 7.5 9.0 2.5 0.6 −3.6 13.5 5.8 0.6 0.2 −7.6 10.9 3.2 1.7 −5.4 −18.0 3.2 2.3 1.1 −7.8 −3.9 6.1 4.1 1.3 −7.5 −11.1

Mortality Q3

Q4

Q5

Q1

Q2

Q3

Q4

Q5

−14.2 −4.3 −8.2 −7.7 1.0 −9.6 −16.2 −5.9 0.5 −2.8 −9.4 −6.7 −8.1 −9.0 1.3 −7.9 −9.3 −11.0 −1.8 −9.2 −14.9 −14.7 −1.6 −9.7

−12.6 −11.0 −1.9 −3.0 −0.8 −11.7 −10.9 −7.7 −2.4 1.1 −14.1 −9.7 −3.4 −3.9 −1.4 −6.4 −5.7 −5.2 −9.6 −13.4 −8.6 −6.0 0.0 −7.8

−3.6 −6.6 1.9 −1.0 −11.6 −2.8 −3.2 −10.4 −10.3 2.0 −9.7 −6.4 0.6 2.3 −15.8 −1.1 −1.8 0.0 −9.7 −4.7 −2.0 0.2 2.8 −2.5

−13.3 −9.8 −22.0 −20.4 −7.6 −14.4 −12.4 −10.8 −10.4 −33.2 −11.4 −11.2 −28.8 −29.4 −9.1 −38.8 −17.7 −20.7 −7.9 −12.2 −12.9 −20.0 −30.9 −25.2

−22.9 −15.8 −37.8 −32.6 −9.7 −24.3 −25.7 −17.0 −11.2 −31.0 −14.2 −20.7 −31.3 −35.6 −10.3 −33.6 −27.7 −30.1 −10.7 −21.7 −25.7 −43.6 −19.5 −32.4

−38.5 −21.8 −32.5 −28.1 −13.1 −33.8 −41.5 −25.8 −14.3 −23.5 −28.6 −29.2 −28.5 −31.2 −10.9 −28.0 −29.3 −36.4 −16.6 −31.0 −37.1 −40.5 −15.7 −33.6

−38.3 −35.1 −19.7 −24.7 −16.6 −38.5 −34.6 −29.9 −20.8 −13.5 −40.4 −35.2 −25.4 −20.3 −15.7 −26.7 −26.1 −27.8 −30.8 −39.9 −30.6 −29.7 −12.9 −31.7

−25.8 −30.4 −10.7 −22.3 −37.6 −23.2 −24.9 −35.6 −36.5 −12.4 −33.8 −29.3 −17.6 −8.8 −39.7 −20.7 −21.7 −18.2 −33.7 −27.0 −23.1 −16.8 −7.8 −22.3

(Continued)

Table 3.8

(Continued) Assets Q1

Water

Q2

Q3

Q4

Q5

Q1

Q2

Schooling Q3

Q4

Q5

Q1

Fertility

Q2 Q3

Q4

Q5

Q1

Mortality

Q2

Madagascar 2.2 3.6 5.4 8.6 7.4 1.9 7.5 14.5 26.8 27.4 3.4 4.0 4.6 5.5 3.4 1.1 −1.2 Malawi 1.8 3.8 5.1 7.3 7.4 0.2 2.7 3.4 7.9 19.7 4.2 6.3 7.7 8.6 3.9 1.3 −2.2 Mali 1.2 1.8 2.5 4.0 9.8 1.0 4.4 7.0 12.1 29.2 1.7 1.9 2.0 2.9 6.3 3.2 2.2 Mauritania 3.9 5.5 6.6 9.8 7.1 18.2 25.8 25.6 27.6 15.2 1.1 1.4 2.5 6.5 6.3 −0.8 −3.3 Morocco 3.8 6.5 10.8 9.5 3.8 16.5 26.7 27.3 25.9 5.5 1.3 2.5 8.2 6.8 4.0 −0.6 −5.0 Mozambique 1.1 2.6 3.7 6.4 10.4 0.1 2.8 5.1 13.7 28.6 1.9 3.3 4.1 5.6 7.3 3.3 0.7 Namibia 8.8 9.1 7.3 5.4 2.1 28.6 28.4 20.1 18.2 2.7 4.9 5.4 4.0 1.3 0.4 −9.0 −9.6 Nepal 2.7 4.4 6.7 8.8 6.7 2.2 9.7 20.0 24.8 26.2 4.3 4.8 5.1 6.4 2.0 0.1 −2.6 Nicaragua 6.5 9.2 6.4 4.4 1.6 24.6 26.4 18.1 13.3 1.6 3.5 5.8 3.3 2.0 1.1 −5.3 −9.7 Niger 1.0 1.1 1.7 2.6 9.4 0.6 0.9 2.0 4.9 28.5 1.0 1.1 1.4 2.1 5.9 3.9 3.6 Nigeria 1.3 3.0 6.0 7.7 5.6 0.0 0.4 1.7 5.6 14.1 3.2 5.7 9.2 8.7 1.8 2.3 −0.8 Pakistan 3.1 5.0 7.4 8.9 5.3 6.8 14.5 22.4 28.6 16.2 3.2 4.0 5.2 5.1 2.4 −0.2 −3.2 Paraguay 9.4 9.5 7.6 4.5 1.7 27.6 25.7 27.4 16.6 2.3 6.3 6.4 3.4 1.3 0.9 −10.3 −10.3 Peru 10.6 11.4 6.0 4.3 1.6 9.8 24.3 26.2 21.5 3.3 12.1 9.8 2.0 1.0 0.9 −13.2 −14.0 Philippines 13.0 9.6 6.4 4.7 3.1 6.8 20.0 22.6 24.4 17.6 16.7 7.6 2.5 0.5 0.4 −17.9 −10.6 Rwanda 1.4 2.7 4.5 7.6 11.0 0.0 0.4 0.9 2.1 13.0 2.9 4.7 6.9 11.0 10.9 2.4 0.0 South Africa 8.4 6.9 5.9 5.2 2.3 5.6 17.0 24.6 24.7 6.6 10.1 3.6 0.8 0.3 0.3 −9.5 −5.2 Tanzania 4.9 7.2 8.9 10.6 7.5 3.9 12.5 11.1 20.5 24.8 7.1 7.9 10.4 9.6 3.4 −3.9 −7.5 Togo 2.0 4.0 7.3 12.6 7.9 0.6 4.1 6.7 13.7 21.4 2.9 5.2 9.0 14.2 4.6 1.7 −1.9 Uganda 2.4 4.5 6.7 8.1 6.5 1.0 7.1 11.3 18.7 25.1 5.0 5.9 7.1 6.0 1.5 0.1 −3.1 Vietnam 11.4 8.0 4.6 3.5 1.6 20.5 27.1 13.0 9.6 1.5 10.8 4.4 3.0 2.3 1.6 −14.2 −7.6 Yemen 3.9 6.9 9.6 9.7 5.1 5.2 22.3 29.4 28.1 12.8 4.5 4.5 6.1 6.2 2.9 −1.5 −6.1 Zambia 3.8 6.5 9.0 7.9 5.2 2.2 11.2 12.8 16.8 23.3 6.4 7.3 9.5 5.3 0.4 −2.2 −6.4 Zimbabwe 9.6 9.3 8.0 6.7 4.7 7.1 20.4 20.2 23.8 21.8 12.4 7.4 4.8 2.0 0.4 −12.1 −10.2 Average

4.7

6.2

7

7.1

5.7

6.6 12.8 15.4 18.6 17.9

5.3 5.6

5.8

5

3

−2.9

−5.3

Q3

Q4

Q5

Q1

Q2

Q3

Q4

Q5

−3.9 −4.4 1.1 −5.0 −12.8 −1.1 −6.2 −6.1 −4.7 2.8 −6.0 −7.1 −6.8 −4.2 −4.7 −3.1 −3.4 −10.5 −7.8 −6.4 −2.1 −10.5 −10.3 −7.3

−8.9 −7.6 −1.3 −10.8 −10.5 −5.6 −2.5 −9.5 −1.5 1.1 −8.0 −9.2 −1.4 −1.5 −1.8 −8.7 −2.4 −12.8 −16.8 −7.9 −0.3 −10.6 −7.2 −4.8

−6.6 −6.3 −10.9 −6.7 −1.2 −12.0 2.7 −5.0 3.0 −10.2 −2.7 −2.9 2.9 3.0 0.5 −13.2 2.2 −6.5 −7.3 −4.4 2.8 −2.6 −2.3 −1.7

−12.0 −12.0 −8.3 −19.6 −18.7 −8.1 −34.4 −13.7 −28.4 −7.0 −10.1 −15.2 −36.2 −34.6 −40.8 −9.7 −27.9 −20.2 −10.4 −13.9 −39.6 −16.7 −17.5 −32.7

−17.2 −17.7 −11.0 −25.5 −28.2 −12.7 −35.1 −20.0 −34.5 −7.4 −14.9 −22.3 −35.2 −40.4 −33.6 −13.3 −24.2 −27.9 −17.0 −20.1 −32.1 −28.7 −25.9 −33.3

−23.2 −20.9 −13.3 −27.6 −39.6 −15.9 −27.1 −28.1 −24.3 −8.9 −22.7 −30.2 −30.7 −27.0 −25.8 −17.9 −24.2 −31.7 −26.3 −25.7 −19.6 −36.9 −31.5 −28.6

−34.0 −26.2 −18.5 −36.7 −35.9 −24.9 −20.8 −34.2 −18.3 −12.2 −25.4 −34.7 −19.2 −21.6 −22.9 −26.5 −23.2 −37.3 −41.2 −29.5 −15.9 −36.5 −27.2 −26.0

−30.5 −26.7 −37.3 −27.2 −16.0 −38.8 −8.2 −27.5 −8.1 −36.2 −19.2 −21.2 −8.4 −8.9 −17.6 −35.0 −10.5 −29.1 −29.1 −25.9 −8.8 −19.7 −22.5 −21.3

−6.5

−6.4

−3.7 −19

−24.2 −26.6 −27.4 −23.2

Aid and the MDGs

81

the other wellbeing indicators, even though aid has no significant direct effect on this variable. The largest improvements in sch (14 per cent or more) are predicted for quintile 1 in the Philippines and Gabon. The largest improvements in wtr (28 per cent or more) are predicted for quintile 1 in the Philippines, quintiles 3–4 in the Yemen, quintile 4 in Cambodia and quintile 5 in Niger. The largest improvements in mor (40 per cent or more) are predicted for quintile 1 in the Philippines, quintile 3 in Indonesia and quintile 4 in Comoros and Togo. The overall effects on fertility are quite varied. As we have already seen, the direct effect of aid on fertility is positive. However, this is offset by a strong bidirectional link between fertility and mortality. To the extent that aid reduces child mortality, it induces a virtuous spiral in which lower mortality reduces the number of children a family needs to have, which in turn improves child health. So the predicted equilibrium effect of aid is positive in some cases and negative in others. In the average country, fertility falls slightly. Table 3.9 shows the average impact of the 100 per cent increase in aid by quintile, along with corresponding figures for 10 per cent and 50 per cent increases. Either quintile 3 or quintile 4 benefits most on average from the 100 per cent increase with respect to all wellbeing outcomes. Quintile 1, the poorest, does worst from the 10 per cent and 50 per cent increases with respect to all five wellbeing indicators under question; quintiles 3 and 4 do best. Moreover, in the majority of instances the standard deviations for quintile 1 are larger than standard deviations for other quintiles. Overall, the poorest in each country benefit least from increases in aid, and their experience is more varied. Table 3.10 further summarizes the predicted consequences of an increase in aid. Like Table 3.9, Table 3.10 shows, for each wellbeing indicator, the average change predicted with a 100 per cent increase in aid. But this time, mean effects are calculated for a group of absolutely rather than relatively poor (and not-so-poor) quintiles. That is, our first aggregate is formed from a group of quintiles with poor outcomes relative to the whole sample rather than to the respective countries in which they are located. Also, poverty is defined in terms of observed values of the wellbeing indicators, with a different aggregation for each indicator. Two further aggregates are constructed for each indicator, one for a ‘middle’ group and one for a ‘high’ group. For assets, water and schooling, the ‘low’ group consists of quintiles for which the existing measure is less than 20 per cent, the ‘middle’ group consists of quintiles for which the existing measure is between 20 per cent and 50 per cent, and the ‘high’

Percentage increases in each wellbeing indicator with 10, 50 and 100 per cent increases in aid, disaggregated by

Assets 10% increase in aid 50% increase in aid 100% increase in aid Water 10% increase in aid 50% increase in aid 100% increase in aid Schooling 10% increase in aid 50% increase in aid 100% increase in aid Fertility 10% increase in aid 50% increase in aid 100% increase in aid Mortality 10% increase in aid 50% increase in aid 100% increase in aid

Quintile 1

Quintile 2

Quintile 3

Quintile 4

Quintile 5

Mean

Mean

Mean

Mean

Mean

Std dev.

Std dev.

Std dev.

Std dev.

82

Table 3.9 quintiles

Std dev.

0.6 2.6 4.7

0.4 1.9 3.3

0.8 3.6 6.2

0.5 1.9 3.0

1.0 4.1 7.0

0.4 1.6 2.6

1.0 4.2 7.1

0.4 1.5 2.5

0.8 3.4 5.7

0.4 1.8 2.9

0.7 3.4 6.6

1.1 4.8 8.6

1.5 6.9 12.8

1.3 5.4 9.2

1.8 8.4 15.4

1.2 4.9 8.3

2.3 10.5 18.6

1.1 4.6 7.8

2.5 10.7 17.9

1.2 5.2 9.1

0.7 3.1 5.3

0.5 2.1 3.7

0.8 3.4 5.6

0.5 1.9 2.9

0.9 3.5 5.8

0.5 1.9 3.1

0.8 3.1 5.0

0.5 2.0 3.2

0.5 1.9 3.0

0.5 1.9 2.9

−0.3 −1.5 −2.9

0.7 3.3 5.8

−0.7 −3.1 −5.3

0.8 3.1 5.0

−0.9 −3.8 −6.5

0.7 2.7 4.4

−0.9 −3.9 −6.4

0.7 2.7 4.5

−0.6 −2.4 −3.7

0.8 3.0 5.0

−2.4 −10.8 −19.1

1.3 5.6 10.0

−3.2 −14.0 −24.2

1.4 5.6 9.2

−3.6 −15.4 −26.6

1.1 4.7 7.9

−3.8 −16.1 −27.4

1.1 4.7 8.0

−3.4 −14.0 −23.2

1.3 5.6 9.5

Notes: Figures in bold indicate mean percentage changes in each variable as a result of the increase in aid (that is, percentage point changes for ass, wtr and sch, and percentage growth for fer and mor). Figures in italic indicate corresponding standard deviations.

Aid and the MDGs

83

Table 3.10 Percentage increases in each wellbeing indicator with 10, 50 and 100 per cent increases in aid, disaggregated by initial wellbeing levels Aid increase ass change low initial ass (

Achieving the Millennium Development Goals (Studies in Development Economics and Policy)

Millennium Development Goals Report 2006

Health Inequality and Development (Studies in Development Economics and Policy)

Entrepreneurship and Economic Development (Studies in Development Economics and Policy)

The First Millennium Development Goals Report, IRAN

Targeting Development (Routledge Studies in Development Economics)

Advancing Development: Core Themes in Global Economics (Studies in Development Economics and Policy)

Development Studies and Colonial Policy

Development Finance in the Global Economy: The Road Ahead (Studies in Development Economics and Policy)

Labour Relations in Development (Routledge Studies in Development Economics)

Millennium Development Goals for Health in Europe and Central Asia: Relevance and Policy Implications

Achieving sustainable development

Feminist Visions of Development: Gender Analysis and Policy (Routledge Studies in Development Economics)

Financial Sector Development and the Millennium Development Goals (World Bank Working Papers) (World Bank Working Papers)

Are We on Track to Achieve the Millennium Development Goals? Annual World Bank Conference on Development Economics—Europe 2005

2002 Annual Review of Development Effectiveness: Achieving Development Outcomes : The Millennium Challenge

2002 Annual Review of Development Effectiveness: Achieving Development Outcomes : The Millennium Challenge

The Rise of China and India: Impacts, Prospects and Implications (Studies in Development Economics and Policy)

Ideas, Policies and Economic Development in the Americas (Routledge Studies in Development Economics)

Contemporary Issues in Development Economics

The World Bank and Urban Development: From Projects to Policy (Routledge Studies in Development and Society)

The African Economy: Policy, Institutions and the Future (Routledge Studies in Development Economics, 13)

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Development Policy in the Twenty-First Century: Beyond the Post-Washington Consensus (Routledge Studies in Development Economics)

Social Policy and Economic Development in the Nordic Countries (Social Policy in a Development Context)

Microfinance Perils and Prospects (Routledge Studies in Development Economics)

Hunger and Public Action (Wider Studies in Development Economics)

New Directions in Development Economics: Growth, Environmental Concerns and Government in the 1990s (Routledge Studies in Development Economics, 3)

Achieving the Millennium Development Goals (Studies in Development Economics and Policy)

Millennium Development Goals Report 2006

Health Inequality and Development (Studies in Development Economics and Policy)

Entrepreneurship and Economic Development (Studies in Development Economics and Policy)

The First Millennium Development Goals Report, IRAN

Targeting Development (Routledge Studies in Development Economics)

Advancing Development: Core Themes in Global Economics (Studies in Development Economics and Policy)

Development Studies and Colonial Policy

Development Finance in the Global Economy: The Road Ahead (Studies in Development Economics and Policy)

Labour Relations in Development (Routledge Studies in Development Economics)

Millennium Development Goals for Health in Europe and Central Asia: Relevance and Policy Implications

Achieving sustainable development

Feminist Visions of Development: Gender Analysis and Policy (Routledge Studies in Development Economics)

Financial Sector Development and the Millennium Development Goals (World Bank Working Papers) (World Bank Working Papers)

Are We on Track to Achieve the Millennium Development Goals? Annual World Bank Conference on Development Economics—Europe 2005

2002 Annual Review of Development Effectiveness: Achieving Development Outcomes : The Millennium Challenge

2002 Annual Review of Development Effectiveness: Achieving Development Outcomes : The Millennium Challenge

The Rise of China and India: Impacts, Prospects and Implications (Studies in Development Economics and Policy)

Ideas, Policies and Economic Development in the Americas (Routledge Studies in Development Economics)

Contemporary Issues in Development Economics

The World Bank and Urban Development: From Projects to Policy (Routledge Studies in Development and Society)

The African Economy: Policy, Institutions and the Future (Routledge Studies in Development Economics, 13)

Arms Trade and Economic Development: Theory and Policy in Offsets (Studies in Defence Economics)

Development Policy in the Twenty-First Century: Beyond the Post-Washington Consensus (Routledge Studies in Development Economics)

Social Policy and Economic Development in the Nordic Countries (Social Policy in a Development Context)

Microfinance Perils and Prospects (Routledge Studies in Development Economics)

Hunger and Public Action (Wider Studies in Development Economics)

New Directions in Development Economics: Growth, Environmental Concerns and Government in the 1990s (Routledge Studies in Development Economics, 3)

Recommend Documents