90430 A WORLD BANK STUDY Improving Basic Services for the Bottom Forty Percent LESSONS FROM ETHIOPIA Qaiser M. Khan, Jean-Paul Faguet, Christopher Gaukler, and Wendmsyamregne Mekasha Improving Basic Services for the Bottom Forty Percent A WORLD BANK STUDY Improving Basic Services for the Bottom Forty Percent Lessons from Ethiopia Qaiser M. Khan, Jean-Paul Faguet, Christopher Gaukler, and Wendmsyamregne Mekasha © 2014 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 17 16 15 14 World Bank Studies are published to communicate the results of the Bank’s work to the development com- munity with the least possible delay. The manuscript of this paper therefore has not been prepared in accordance with the procedures appropriate to formally edited texts. This work is a product of the staff of The World Bank with external contributions. The findings, inter- pretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: Khan, Qaiser M., Jean-Paul Faguet, Christopher Gaukler, and Wendmsyamregne Mekasha. 2014. Improving Basic Services for the Bottom Forty Percent: Lessons from Ethiopia. World Bank Studies. Washington, DC: World Bank. doi:10.1596/978-1-4648-0331-4. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Responsibility for the views and opinions expressed in the adaptation rests solely with the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use a compo- nent of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not lim- ited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-0331-4 ISBN (electronic): 978-1-4648-0333-8 DOI: 10.1596/978-1-4648-0331-4 Cover photo: © Arne Hoel / World Bank. Further permission required for reuse. Cover design: Debra Naylor, Naylor Design Library of Congress Cataloging-in-Publication Data has been requested. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Contents Foreword xi Acknowledgments xiii About the Authors xv Executive Summary xvii Abbreviations xxiii Introduction and Background 1 Context and Structure of This Book 3 The Promoting Basic Services Program 3 Notes 8 Chapter 1 Conceptual Framework and Methodology 9 Notes 13 Chapter 2 Citizen Direct Voice and Accountability 15 Decentralization and the Quality and Appropriateness of Local Services 16 Financial Transparency and Accountability 17 Social Accountability 18 Grievance Redress Mechanisms 18 Notes 19 Chapter 3 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture 21 A Positive Association between Woreda Block Spending and Results in Education 21 A Positive Association between Woreda Block Grant Spending and Results in Health 22 A Positive Association between Woreda Block Grant Spending and Results in Agriculture 25 Notes 28 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  v vi Contents Chapter 4 How Equitable Is Decentralized Spending at the Woreda Level? 29 Woreda Block Grant Spending on Health Improves Outcomes, Especially for the Rural Poor 29 Incidence Analysis of Woreda Block Grant Spending on Health and Education 32 Catalytic Effect of Spending on Extension Services by Plot Size 33 Gender Equity Analysis for Woreda Block Grant Spending 33 Notes 35 Chapter 5 The Federal System’s Role in Helping Lagging Areas and Groups 37 Regionally Disaggregated Analysis of Woreda Block Grant Spending 37 Ethnically Disaggregated Analysis of Woreda Block Grant Spending 39 Notes 41 Conclusions and Recommendations 43 Appendix A Detailed Methodology and Data Sources 45 Appendix B Financial Transparency and Social Accountability under the PBS Program 53 Appendix C The Center and the Periphery in Ethiopia: The Evolution of Today’s Federal State 55 Appendix D Detailed Regression Results 59 Bibliography 105 Boxes I.1 The Indirect Route of Accountability Through Policy Makers to Service Providers 5 1.1 Some Caveats on Interpreting the Results of This Study 10 2.1 The Direct Route of Accountability—Citizen Direct Voice and Accountability Models in Ethiopia 15 3.1 Ethiopia’s Health Extension Program and Health Extension Workers 22 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Contents vii Figures I.1 The Triangle of Accountability in Service Delivery 4 1.1 Conceptual Model of the Results Chain of PBS Spending 9 1.2 Woreda-Level Expenditure in the Five PBS Sectors, 2011 12 3.1 Growth in Extension Services and Woredas’ Expenditure on Agriculture, 2008–11 25 3.2 Production by Crop, 2011 27 4.1 Percent Improvement in Selected Health Indicators Among Rural People between 2005 and 2011, by Wealth Quintile 30 5.1 Woreda Total Recurrent Expenditure Per Capita, 2010 38 Tables I.1 Federal Block Grants and PBS Program Disbursements 6 I.2 Total Woreda Recurrent Expenditures by Region 7 3.1 Association of Log of Per Capita Education Expenditure with Log of Education Outcomes 22 3.2 Association of Log of Per Capita Health Expenditure with Health Outcomes 24 3.3 Effect on Farmers’ Use of Extension Services of ETB 1 Per Capita Spending on Agricultural Extension Workers 26 3.4 Association of Log of Per Capita Spending on Agricultural Extension Workers with Log of Yield 28 4.1 Predicted Probabilities for “Successful” Health Outcomes by Place of Residence and Wealth Quintile if an HEW Visited the Household in the Past Year 31 4.2 Incidence of Woreda-Level Spending by Wealth Quintile 32 4.3 Effect of ETB 1 Spending on Agricultural Extension Workers on Probability of Using Improved Farming Techniques, by Plot-Size Quintile, 2011 33 4.5 Association of Log of Education Expenditure with Log of Net Enrollment Rate and Log of Net Intake Rate, by Gender 34 4.5 Predicted Probabilities of Improved Farming Techniques Being Used on a Field, by Gender of Household Head, 2011 35 5.1 Woredas Spending More Than 110 Percent of the National Average on Basic Services, by Region, 2010 38 5.2 Selected Ethnicities’ Influence on Total Woreda Recurrent Expenditure Per Capita 40 A.1 Variables in the Poverty and Social Impact Analysis Database, by Administrative Level 48 C.1 Ethiopian Regional Population Distribution and Ethnic Composition 56 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 viii Contents D.1 Stage 1: Education—Linear Regression Association of Per Capita   Education Expenditures with Net Enrollment Rate 59 D.2 Stage 1: Education—Quadratic Regression Association of   Per Capita Education E­ xpenditures with Net Enrollment Rate 60 D.3 Stage 1: Education—Log Linear Regression Association of Per   Capita Education E­ xpenditures with Net Enrollment Rate 61 D.4 Stage 1: Education—Linear Regression Association of Per   Capita Education Expenditures with Pupil-Teacher Ratio 62 D.5 Stage 1: Education—Quadratic Regression Association of   Per Capita Education E­ xpenditures with Pupil-Teacher Ratio 63 D.6 Stage 1: Education—Log Linear Regression Association of Per   Capita Education E­ xpenditures with Pupil-Teacher Ratio 64 D.7 Stage 1: Health—Linear Regression Association of Per   Capita Health Expenditures with Penta3 Vaccination Rate 65 D.8 Stage 1: Health—Quadratic Regression Association of   Per Capita Health Expenditures with Penta3 Vaccination Rate 66 D.9 Stage 1: Health—Log Linear Regression Association of   Per Capita Health Expenditures with Penta3 Vaccination Rate 67 D.10 Stage 1: Health—Linear Regression Association of Per Capita   Health Expenditures with Antenatal Care 68 D.11 Stage 1: Health—Quadratic Regression Association of Per   Capita Health E­ xpenditures with Antenatal Care 69 D.12 Stage 1: Health—Log Linear Regression Association of Per   Capita Health Expenditures with Antenatal Care 70 D.13 Stage 1: Health—Linear Regression Association of Per Capita   Health Expenditures with Contraceptive Acceptance Rate 71 D.14 Stage 1: Health—Quadratic Regression Association of Per Capita   Health Expenditures with Contraceptive Acceptance Rate 72 D.15 Stage 1: Health—Log Linear Regression Association of Per Capita   Health Expenditures with Contraceptive Acceptance Rate 73 D. 16 Stage 1: Health—Linear Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant 74 D.17 Stage 1: Health—Quadratic Regression Association of Per Capita   Health Expenditures with Delivery by Skilled Birth Attendant 75 D.18 Stage 1: Health—Log Linear Regression Association of Per Capita   Health Expenditures with Delivery by Skilled Birth Attendant 76 D.19 Stage 1: Agriculture—Log Linear Regression Association   of Per Capita Agriculture ­Expenditures with Cereal Yield 77 D.20 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Pulses Yield   of Per Capita Agriculture E 79 D.21 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Root Crop Yield   of Per Capita Agriculture E 81 D.22 Stage 1: Agriculture—Log Linear Regression Association Expenditures with Vegetable Yield   of Per Capita Agriculture ­ 83 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Contents ix D.23 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Oil Seeds Yield   of Per Capita Agriculture E 85 D.24 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Enset Yield   of Per Capita Agriculture E 87 D.25 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Fruit Yield   of Per Capita Agriculture E 89 D.26 Stage 1: Agriculture—Log Linear Regression Association ­ xpenditures with Coffee Yield   of Per Capita Agriculture E 91 D.27 Stage 3: Agriculture—Predicted Probabilities for Farmers’   Use of Any Improved ­ Technique, 2011 93 D.28 Gender Issues: Linear Regression Association of Per Capita   Education Expenditures with Net Enrollment Rate for Males 95 D.29 Gender Issues: Quadratic Regression Association of Per Capita   Education E­ xpenditures with Net Enrollment Rate for Males 96 D.30 Gender Issues: Log Linear Regression Association of Per Capita   Education Expenditures with Net Enrollment Rate for Males 96 D.31 Gender Issues: Linear Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Females 97 D.32 Gender Issues: Quadratic Regression Association of Per Capita   Education Expenditures with Net Enrollment Rate for Females 97 D.33 Gender Issues: Log Linear Regression Association of Per Capita   Education Expenditures with Net Enrollment Rate for Females 98 D.34 Gender Issues: Linear Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Males 98 D.35 Gender Issues: Quadratic Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Males 99 D.36 Gender Issues: Log Linear Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Males 99 D.37 Gender Issues: Linear Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Females 100 D.38 Gender Issues: Quadratic Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Females 100 D.39 Gender Issues: Log Linear Regression Association of Per Capita   Education Expenditures with Net Intake Rate for Females 101 D.40 Gender Issues: Predicted Probabilities for Farmers’ Use of Any   Improved Technique by Gender, 2011 101 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Foreword When Ethiopia embarked on a program of improving service delivery by decen- tralizing power to the third, or woreda, level, we were initially skeptical. In the World Development Report 2004: Making Services Work for Poor People, we asserted that the delivery of basic services requires two things: First, citizens must be able to hold politicians accountable for allocating resources and monitoring results; second, politicians in turn must be able to hold service providers—teach- ers, doctors, nurses—accountable for quality delivery of services. We called these the two legs of the “long route of accountability.” While Ethiopia’s strong party affiliations and military traditions would make the second leg of the long route work well, the limited political competition at the national level potentially could weaken the first leg. Furthermore, whereas in principle decentralizing to the subnational level would make these accountability relationships work bet- ter—citizens would be closer to political leaders, and politicians could better monitor service providers—the limited capacity in some of the woredas could act as a brake on progress. Our skepticism was unwarranted. Ethiopia has experienced a remarkable improvement in service delivery outcomes. Child mortality has fallen from 123 per thousand in 2005 to 88 in 2010, and the primary net enrollment rate rose from 68 percent in 2004/05 to 82 percent in 2009/10. This richly documented and analytically rigorous book explains why. While political competition at the national level remains limited, competition at the local level is healthy, with officials’ career advancement depending on their being able to meet develop- ment and service delivery targets. The government has introduced social account- ability mechanisms such as grievance redress through the Ethiopian Institution of the Ombudsman, financial transparency (the local budgets are posted in com- mon areas), and citizen feedback on budget preparation and implementation. The government has also strengthened the second leg of the long route of accountability by recruiting teachers, health workers, and agricultural extension workers locally, and rates of provider absenteeism are now lower than in other countries in Africa. Greater citizen participation and the intrinsic motivation of people working in their local areas may have helped overcome the initially weak administrative capacity. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   xi xii Foreword The relative success of Ethiopia’s decentralized service delivery program is not a reason to be complacent, however. The achievements have arisen from a low base; Ethiopia remains one of the poorest countries in the world. The progress on indicators that people care about—students’ learning and overall health status— has not been impressive. As these indicators are more specific to the individual, their achievement will require an even sharper focus on citizen voice and par- ticipation. Ethiopia has demonstrated how much can be achieved by opening up the system to greater accountability “from below.” Just think how much more can be achieved by going further—and allowing poor people to participate fully in the delivery of services that are so important to them and their children. Shantayanan Devarajan and Ritva Reinikka Co-Director, World Development Report 2004 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Acknowledgments This book was written by Qaiser M. Khan (Lead Economist, World Bank), Jean- Paul Faguet (Professor for the Political Economy of Development, London School of Economics), Christopher Gaukler (Consultant, World Bank) and Wendm­ syamregne Mekasha (Senior Social Protection, Specialist, World Bank). This book is based on a World Bank report (84215-ET) developed with fund- ing from the World Bank and a Trust Fund for Enhanced Supervision of Ethiopia’s Promoting Basic Services program, supported mostly by the UK Department for International Development (DFID). That report benefited from suggestions pro- vided by the peer reviewers—Robert Chase (Lead Economist, World Bank), Nazmul Chaudhury (Lead Economist, World Bank), Andrew Dabalen (Lead Poverty Specialist, World Bank), and Philip O’Keefe (Lead Economist, World Bank)—and other colleagues present at the review meeting, including DFID’s Ethiopia Team, Andrew Goodland (Senior Agriculture Economist, World Bank), Ruth Hill (Senior Economist, World Bank), and Thanh Thi Mai (Senior Education Economist, the World Bank). Before the review meeting, Yoseph Abdissa (Senior Social Protection Specialist, World Bank), Colin Andrews (Senior Social Protection Specialist, World Bank), Carlo del Ninno (Senior Economist, World Bank), G.N.V. Ramana (Lead Public Health Specialist, World Bank), and Huihui Wang (Senior Health Economist, World Bank) also provided comments on the report. The authors would also like to acknowledge the technical guidance and advice received from Deon Filmer (Lead Economist, World Bank) and Lynne Sherburne-Benz (Sector Manager, World Bank). Members of the Ethiopian gov- ernment who provided invaluable assistance include Ato Temesgen Walelign (Development Planning and Research Directorate); Ato Degu Lakew (Government Accounts); Ato Alemayhu Gebretsadik, Ato Biratu Yigezu, and Ato Habekristos Beyene (Central Statistical Agency); Ato Feta Zeberga (Ministry of Agriculture); Ato Asmelash Mersa (Ministry of Education); and Ato Wondimu Ayele (Ministry of Health and Economic Development). Support from Ato Getachew Negera, Head of Channel One Program Coordinating Unit (Ministry of Finance), is also gratefully acknowledged. Any errors and omissions are the responsibility of the authors. We thank Shantayanan Devarajan and Ritva Reinikka, Co-Director, World Development Report 2004, for writing the foreword to this book. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   xiii xiv Acknowledgments Kavita Watsa and Chandrani Ray provided advice and assistance in the pro- duction of this book, which was edited by Kelly Cassaday and benefited from the support of Sofia Said, Team Assistant, World Bank. Gelila Wodeneh, Communications Officer, the World Bank also provided support and advice. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 About the Authors Qaiser M. Khan is Lead Economist working on Africa. He has worked on social protection and basic services for the poor at the World Bank since 1989, all over Africa, the Middle East, and Asia, and he has written considerably on the subject. Before joining the World Bank, he taught economics in the United States and worked for a management consulting firm out of New York and for a nongovern- mental organization in Bangladesh. He has a PhD and two master’s degrees from the University of Pennsylvania in Philadelphia, Pennsylvania, United States, and a bachelor’s degree from Colby College in Waterville, Maine, United States. Jean-Paul Faguet is Professor of the Political Economy of Development at the London School of Economics. He is also Chair of the Decentralization Task Force of the Initiative for Policy Dialogue at Columbia University. His research blends quantitative and qualitative methods to investigate the institutions and organiza- tional forms that underpin rapid development. He has published extensively in the academic literature, including Governance from Below: Decentralization and Popular Democracy in Bolivia (University of Michigan Press), which won the W.J.M. Mackenzie Prize for best political science book of 2012. He also recently edited a special issue of World Development on “Decentralization and Governance” (2013), and co-edited Descentralización y democratización en Bolivia: La historia del Estado débil, la sociedad rebelde y el anhelo de democracia (with M. Zuazo and G. Bonifaz; Friedrich Ebert Stiftung, 2012). His teaching and research focus on comparative political economy, new institutional economics, economic develop- ment, and economic history. Faguet has a PhD from the London School of Economics, a master’s degree from the JFK School at Harvard, and his bachelor’s degree from Princeton University. Christopher Gaukler is a consultant for the World Bank’s Africa Social Protection Unit and has been working on Ethiopia since 2006. He focuses on monitoring and evaluation for the Basic Services team. Between 2011 and 2012, Gaukler worked as a monitoring and evaluation specialist for the government of South Sudan in the Ministry of Gender, Child, and Social Welfare. He has a master’s degree from the School of Advanced International Studies at the Johns Hopkins University and an undergraduate degree from the University of Virginia. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   xv xvi About the Authors Wendmsyamregne Mekasha (Wendm) is a Senior Social Protection Specialist in the World Bank office based in Addis Ababa, Ethiopia. He works on program monitoring and evaluation, building nationwide management information sys- tems, and designing and implementing various analytical works including impact/ process evaluation for programs and projects. Mekasha received his bachelor’s degree in mathematics from Addis Ababa University (Ethiopia), his master’s degree in agricultural economics from Haromaya University (Ethiopia), and his master’s of liberal arts in general management from Harvard University (United States). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Executive Summary Ethiopia, like most developing countries, has opted to deliver services such as basic education, primary health care, agricultural extension advice, water, and rural roads through a highly decentralized system (Manor 1999; Treisman 2007). That choice is based on several decades of theoretical analysis examining how a decentralized government might respond better to diverse local needs and pro- vide public goods more efficiently than a highly centralized government.1 Ethiopia primarily manages the delivery of basic services at the woreda (dis- trict) level. Those services are financed predominantly through intergovernmental fiscal transfers (IGFTs) from the federal to the regional and then the woreda administrations, although some woredas raise a small amount of revenue to sup- port local services. Since 2006, development partners and the government have cofinanced block grants for decentralized services through the Promoting Basic Services (PBS) Program. Aside from funding the delivery of services, the program supports measures to improve the quality of services and local governments’ capacity to deliver them by strengthening accountability and citizen voice.2 Objectives This study attempts to determine the extent to which spending at the woreda level on basic services is associated with key policy outputs and human outcomes. Woreda-level block grants primarily support locally recruited staff who provide basic services in five sectors (health, education, agriculture, water and sanitation, and rural roads). Because of limitations in the data, the analysis focuses on health, education, and agriculture. A parallel objective of the study is to assess the inci- dence of these expenditures by wealth quintile, in line with the World Bank’s objective of achieving shared growth by reaching the bottom 40 percent. A final objective is to investigate whether the allocation of woreda-level block grants reflects the constitutional objective of providing additional resources to histori- cally marginalized populations. In other words, how effective and equitable is spending on basic services at the woreda level? The study also looks at the effec- tiveness of efforts to improve financial transparency and accountability, social accountability, and grievance redress mechanisms, because of their strong influ- ence on the availability and quality of basic services. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   xvii xviii Executive Summary Overall Findings on the Effectiveness of a Decentralized Approach to Service Delivery Ethiopia’s model for delivering basic services appears to be succeeding and to confirm that services improve when service providers are more accountable to citizens. As discussed in the World Development Report 2004 (World Bank 2003), accountability for delivering basic services can take an indirect, long route, in which citizens influence service providers through government, or a more direct, short route between service providers and citizens. When the long, indirect route of accountability is ineffective, service delivery can suffer, especially among poor or marginalized citizens who find it challenging to express their views to policy makers. In Ethiopia, the indirect route of accountability works well precisely because of decentralization. Service providers are strictly accountable to local governments for producing results, but in turn, the local authorities are held accountable by the regional and federal governments. A degree of local competition for power and influence helps to induce local authorities and service provides to remain open to feedback from citizens and take responsibility for results. The direct route of accountability has been reinforced by measures to strengthen financial transpar- ency and accountability (educating citizens on local budgets and publicly provid- ing information on budgets and service delivery goals), social accountability (improving citizens’ opportunities to provide feedback directly to local administra- tors and service providers), and impartial procedures to redress grievances (insti- tuting the independent Ethiopian Institution of the Ombudsman, for example). Woreda-level spending, financed through IGFTs and supported by the PBS Program, has been a very effective strategy for Ethiopia to attain its Millennium Development Goals (MDGs). Spending on health and education accounts for 80 percent of PBS-financed spending by the woredas, which goes to pay for health extension workers (HEWs) and teachers. Although the link between numbers of personnel hired and services delivered must be treated with caution (and is being explored in detail in another study), this study finds evidence that woreda-level spending in health and education is effective. Owing to the intervention of HEWs, the use of health services has increased, especially among the poorest quintiles. Every additional US$1 of per capita spending by the woredas on health is associated with a 7.5 percent increase in the contraceptive prevalence rate and an 12.4 percent increase in deliveries by skilled birth attendants (Wang et al., forthcoming) (two interventions that can reduce maternal mortality dramatically), as well as a 4 percent increase in ante- natal care (which can reduce infant and child mortality significantly). For educa- tion, an increase of US$1 per capita in spending by each woreda is associated with a 3.7 percent increase in the net primary enrollment rate within that woreda. Similar results are seen for the pupil-teacher ratio. Finally, the effect of woreda-level spending on agricultural extension workers is associated with higher yields for major crops, including cereals, vegetables, enset, coffee, and fruit. Spending on agricultural extension workers increases the Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Executive Summary xix probability that farmers, regardless of the size of their plots, will use improved farming techniques. Education, health, and agriculture account for 97 percent of woreda spending, which in turn constitutes 97 percent of PBS-financed IGFTs. This is comple- mented by support for greater engagement among citizens, improvements in local capacity to manage resources, and better access to information on national and local budgeting and development objectives. While it is difficult to provide precise estimates of the impact of the latter activities, the direction of their effect is clear: Spending efficiency is improved through better capacity, more transpar- ency, and greater accountability to citizens. In interpreting these results, it is important to bear in mind that the chain of causality from woreda spending to results is direct for education: Spending on teachers directly drives enrollments. For health and agriculture, on the other hand, spending at the woreda level on health and agricultural extension workers is best described as catalytic, because it increases the effectiveness of systemwide spending (federal capital investments in medical infrastructure, for example, or national research to develop improved varieties for farmers). Yet overall, these findings demonstrate the power of the PBS-financed decentralized approach for improving access to basic services, encouraging broadly shared development, and propelling Ethiopia rapidly toward the MDGs. Improving Accountability and Citizen Voice in Providing Basic Services The social accountability component of the PBS Program has increased the opportunities for constructive collaboration between citizens and the state to improve basic public services in pilot areas. Even though woredas must operate under federal guidelines, they still exercise a significant amount of discretion that can affect the quality of life and services. More than 84 percent of those surveyed in pilot areas responded positively to social accountability initiatives, which had increased citizens’ awareness of their rights, responsibilities, and entitlements to basic services. After service providers and users drew up joint service improve- ment plans, basic services improved, and so did the quality of the engagement between citizens and service providers. Through the financial transparency and accountability component of the PBS Program, citizens have become more aware of the government budgeting process, and they are advocating more effectively for their rights as a result. The PBS Program has also improved the efficiency of resource use by improving financial management and procurement capacity at the woreda level. Grievance redress mechanisms are another means of ensuring that citizens’ voices are heard with respect to government services. A grievance redress mech- anism provides the opportunity for an impartial third party to review a transac- tion that has taken place between the government and a citizen or a group of citizens. Through dialogue and technical and financial support, the PBS Program is strengthening the Ethiopian Institution of the Ombudsman and the regional Grievance Handling Offices, which offer these services. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 xx Executive Summary Addressing Equity Issues in Providing Basic Services Overall, the benefits of PBS-financed spending at the woreda level on health, education, and agriculture accrued to all income levels. Woreda-level spending on health and education is particularly pro-poor: 58 percent goes to the two bottom wealth quintiles.3 In agriculture, woreda-level spending (primarily for agricul- tural extension workers) drives increases in output and the adoption of new, improved methods across all asset quintiles, although the magnitude was smaller for the bottom quintile, perhaps because of a lack of financing to purchase pro- ductivity-enhancing inputs. Spending on basic services also appears to be reaching females, especially spending on health and education. In fact, health spending is undeniably pro- female, given that much of it promotes women’s access to services that have a strong impact on reducing maternal mortality—including contraception, ante- natal care, and assisted deliveries. Education is the only sector for which expenditure can be associated with data on results, disaggregated by gender, albeit for only two indicators—net enrollment rate and net intake rate at the primary level. The coefficient for female primary school students is slightly higher than for males for both net enrollment and intake, although not sig- nificant. The important finding is that no bias against females was found in expenditure on education, the sector that receives the bulk of PBS Program funds. The results on agriculture tell a somewhat different and less clear story. Eighty-seven percent of Ethiopia’s farmers are male, and access to extension services was found to favor males. It is not certain whether this bias is driven by gender differences in crop choices, the quality of land farmed by men compared to women, or some other inherent gender bias. Further analysis is required to clarify these issues and develop policy responses based on the evidence. A final important question related to equity is whether PBS Program resources channeled to the woredas are reaching Ethiopia’s historically dis- advantaged regions and ethnic groups. The answer appears to be that in terms of basic service expenditure per capita, the current system broadly favors Ethiopia’s historically disadvantaged regions compared to the histori- cally dominant ones. More than 50 percent of the woredas in Gambella and 30 percent in Beneshangul-Gemuz—the two most disadvantaged regions of the country—spend more than 110 percent of the national average on the basic service sectors. Spending also appears to favor some historically disad- vantaged ethnic groups, in accordance with constitutional mandates. Five majority-Anyiwak woredas are noteworthy for receiving the most public resources of all woredas in the nation. The exception occurs among the Somali groups. Federal transfers to Somali Region appear to be fine, but the region transfers a much smaller share to the woredas—49 percent, compared to the national average of 73 percent (excluding Addis Ababa) over the four years studied here.4 Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Executive Summary xxi Conclusions and Recommendations Decentralized spending at the woreda level is both effective and pro-poor. The estimates provided here imply that the returns to this spending are far from decreasing, which means that Ethiopia has scope to increase spending and speed its attainment of the MDGs. The current approach also appears to be helping some of Ethiopia’s histori- cally disadvantaged areas and ethnic groups to catch up with the rest of the country. Expenditure to provide basic services at the woreda level is broadly equal across Ethiopia’s woredas, with the striking exception of a small number of woredas that are concentrated in the country’s most disadvantaged regions and receive significantly greater resources. Resource flows are lowest among the more developed, historically dominant regions. In contrast to the predictions of some public management theories, the decen- tralized provision of services in Ethiopia is not increasing regional, ethnic, or gender inequalities in investment inputs or service outputs. Indeed, the opposite seems to be true for education and health, where the impact of PBS-financed IGFT resources was disproportionately high among the bottom two quintiles and women. In sum, support for decentralized services in Ethiopia appears to be an effective use of development partners’ resources from both an efficiency and equity perspective. The only exception to these findings is agriculture, for which the impact of PBS-financed IGFT expenditure was smaller for the bottom quintile. In this instance, a wide array of factors is likely to be at work, not the least of which could be poor farmers’ inability to pay for productivity-enhancing inputs, land of poor quality, or water management issues that could not be controlled for in the analysis (except for rainfall). Although quantitative evidence is unavailable at present, descriptive evidence from the first phase of the Ethiopia Social Accountability Program implies that structured feedback sessions involving citizens and service providers are strength- ening citizens’ participation in pilot areas. That evidence, together with the strong guidance emerging from governance and accountability theory, argues for the continued application of social accountability tools and the development of policies to sustain their use in the Ethiopian context. This conclusion can be verified when the impact assessment for the second phase of the Ethiopia Social Accountability Program becomes available. The difficulty of conducting subnational empirical work on Ethiopia cannot be overstated. Creating the database for this study required considerable effort and improvisation on the part of the research team. The resulting standardized database of woreda-level expenditures and characteristics, which will be made public, is a major output of this study. The data will be augmented with new data collected under the third phase of the PBS Program and used in future research related to the program. It is hoped that in time this dataset will become a useful tool for researchers and students elsewhere in Africa and beyond. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 xxii Executive Summary Notes 1. See, for example, Tiebout (1956), Oates (1972), Besley and Coate (2003), and Faguet (2000, 2012). 2. The design of the PBS Program is influenced by the accountability triangle between government, service providers, and citizens (see World Bank 2003 and figure I.1 in the main text). 3. The Ethiopia section of an ongoing, multicountry study (Woldehanna, Tsehaye, and Hill, forthcoming) also finds that spending on primary education in Ethiopia (which is mostly managed by the woredas) is pro-poor—more so than overall spending on education (33 percent goes to the two bottom wealth quintiles). 4. Although spending at the woreda level in Somali Region appears to be lower than required to compensate for decades of underinvestment in basic services, the data for Somali woredas may be understated; some expenditures normally paid by woredas in other regions are paid regionally due to security and capacity constraints. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Abbreviations AGSS Agriculture Sample Survey AIDS acquired immune deficiency syndrome ANC antenatal care CAR contraceptive acceptance rate CSA Central Statistical Agency DelSBA deliveries by skilled birth attendants DHS Demographic and Health Survey EFY Ethiopia fiscal year EIO Ethiopian Institution of the Ombudsman ETB Ethiopian birr FBG Federal Block Grant FTA financial transparency and accountability GDP gross domestic product HEW health extension worker HIV human immunodeficiency virus IDA International Development Association IGFT intergovernmental fiscal transfer MDG Millennium Development Goal NER net enrollment rate NIR net intake rate PBS  Promoting Basic Services (formerly, Protection of Basic Services) Program PTR pupil-teacher ratio SNNP Southern Nations, Nationalities, and Peoples (Region) WASH Water, Sanitation, and Hygiene WBG Woreda Block Grant Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   xxiii Introduction and Background Among Ethiopia’s recent impressive development results, its rapid improvement in basic service delivery indicators is outstanding. The Overseas Development Institute reports that Ethiopia has moved faster toward the Millennium Development Goals (MDGs) than all but two other developing countries (ODI 2010). According to the latest Ethiopia Demographic and Health Survey (DHS) data, child mortality fell from 123 per thousand in 2005 to 88 in 2010, and the primary net enrollment rate rose from 68 percent in 2004/05 to 82 percent in 2009/10. Progress in delivering basic services over the past 10 years has been coupled with impressive economic growth: Gross development product (GDP) grew by 11 percent per annum on average between 2004/05 and 2009/10, according to official estimates. Initially led by agriculture, growth has become more broad based, with rising contributions from the mining, services, and manufacturing sectors. Ethiopia’s growth may have slowed more recently, but it remains among the highest of any country in the world. The share of the population living in absolute poverty is declining as well: It fell from 38.7 percent in 2004/05 to 29.6 percent in 2011, according to official data. All of these results spring from a long-term, concerted government commit- ment to pro-poor development. Building on the Sustainable Development and Poverty Reduction Program 2002–04/05 and the Plan for Accelerated and Sustained Development to End Poverty 2005–10, the government’s current development plan, the Growth and Transformation Plan, aims to achieve the MDGs by 2015 and middle-income status for Ethiopia by 2020–23. Ethiopia achieved the MDG-4 (Child Mortality) target in 2014, ahead of schedule, and appears to be on track to reach the other MDGs by 2015, according to United Nations progress reports. Under its constitution and legal framework, the government is committed to maintaining the integrity and capacity of the nation’s decentralized administra- tions down to the woreda level.1 Within Ethiopia’s federal structure, the central government provides unearmarked block grants through regional governments, Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  1 2 Introduction and Background which in turn provide block grants to woreda administrations to deliver services to citizens. This administrative and financial structure has provided timely, predictable financing to sup- port a steady and impressive increase in basic services throughout the country. But has Ethiopia’s decentralized approach to delivering basic services been cost effective? For decades, analysts have debated whether decentralization can make government more responsive to diverse local needs and more efficient in providing public goods.2 The job of providing basic services generally falls to local governments, so building their capacity is an important part of improving basic services. For shared growth and development to occur, it is vital that such services reach the poorest 40 percent of the population. Ethiopia’s fiscal architecture makes it a rich, promising empirical setting for investigating the effectiveness of decentralization as a tool for improving both the quality and distribution of public sector outputs. The evidence that Ethiopia can offer is especially welcome because so much of the empirical evidence that has accumulated on decentralization over the past four decades has been inconclusive. Consider the evidence from the broadest empirical surveys. Rondinelli, Cheema, and Nellis (1983) note that decentralization has usually disappointed its advocates. Most developing countries have encountered serious administrative problems in implementing decentraliza- tion. The few comprehensive evaluations that have been conducted of the costs and benefits of decentralization report limited success in some countries but none in others. A decade and a half later, surveys by Piriou-Sall (1998), Manor (1999), and Smoke (2001) were slightly more positive but contained caveats about the strength of the evidence in favor of decentral- ization. Manor notes that the evidence, though extensive, is incomplete, but concludes that “while decentralization…is no panacea, it has many virtues and is worth pursuing.” Smoke, by contrast, finds the evidence mixed and anecdotal, and asks whether there is any empirical justification for pursuing decentralization. Given the sheer size of the literature, the lack of progress toward an overarching conclusion is surprising. It is worth noting that more recent empirical studies, which are often technically more sophisticated thanks to the enormous improvement in data from developing countries over recent decades, are generally more positive about decentralization’s potential. At least five recent studies address the link between decentralization and substantive outcomes directly and with rigorous quantitative evidence. Escaleras and Register (2012) find that fiscal decentralization is associated with lower natural disaster death rates, implying more effec- tive preparation and/or responses to natural disasters by countries with decentralized gov- ernments. Clark (2009) applies regression discontinuity to a natural experiment from Great Britain and concludes that schools opting out of the centralized education regime—in effect decentralizing themselves—enjoy large increases in student achievement. Galiani, Gertler, and Schargrodsky (2008) find that moving the control of schools from central to provincial governments in Argentina had a positive impact on students’ test scores. The poorest, how- ever, did not gain from decentralization and indeed may have lost. Barankay and Lockwood (2007) report that greater decentralization of education to Swiss cantons is associated with higher educational attainment, especially for boys. By adding empirical evidence from Ethiopia, a low-income country where large programs in decentralization and public invest- ment are yielding potentially significant results, this study makes an important contribution to the literature. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Introduction and Background 3 Context and Structure of This Book This book is based on a report issued as part of a programmatic knowledge series for Ethiopia on improving services for the poorest citizens.3 Future reports will elucidate the findings pre- sented here in several ways. For example, they will present the results of additional studies of the quality of service delivery in health and education, drawing on data from ongoing surveys; offer a detailed assessment of the impact of the Promoting Basic Services (PBS) Program, based on multiround surveys over the next three years; and assess the impact of the social accountability component of PBS (the baseline survey is complete, and a full report is due in two years). The findings reported here will be updated next year, when results from the mini Demographic and Health Survey (DHS) are available, but they are being issued now to pro- vide input into the government’s next Growth and Transformation Plan. This book begins by describing the PBS Program and detailing how intergovernmental fiscal transfers (IGFTs) are channeled through Ethiopia’s decentralized federal structure to support the provision of basic services. The sections that follow focus on the approach used for the analysis, along with the governance and accountability mechanisms that are integral to the PBS Program. Findings on the effectiveness of local spending (IGFTs) for achieving develop- ment results in education, health, and agriculture are reported in detail. Subsequent sections examine the equity impacts of the PBS Program with respect to income, gender, and histori- cally disadvantaged regions and ethnic groups. The book concludes with a summary of the results and a series of recommendations. Four appendixes provide additional details on the methodology and context for the study. The Promoting Basic Services Program Through the PBS Program, the Government of Ethiopia and development partners cofinance block grants to support the provision of basic services at the local level (the government’s share of the financing is larger and rising). The block grants are distributed from the federal government to the regional governments using a formula that adjusts for the size of the popu- lation in each region and for need, based on the current level of development and historic lack of access to services. The funds are transmitted from the regions to the woredas using regional formulas based on similar criteria. Aside from providing block grants to woredas, the PBS Program supports various measures designed to improve the quality of basic services and local governments’ capacity to manage them. The program’s design is influenced by the concept of the “accountability triangle” between policy makers, citizens, and service providers, described in the World Development Report 2004: Making Services Work for Poor People (figure I.1). The program’s design recog- nizes that the long “indirect route” of accountability, in which citizens appeal to policy makers (government) to hold service providers to account, must be complemented by a shorter “direct route,” in which citizens can directly hold service providers accountable for results. In Ethiopia, the indirect route of accountability works well precisely because of decentral- ization (box I.1). Service providers are strictly accountable to local governments for producing results, but, in turn, the local authorities are held accountable by the regional and federal governments for delivering basic services and reaching the delivery targets set in the Growth and Transformation Plan. The teachers, health workers, and agricultural extension workers Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 4 Introduction and Background Figure I.1  The Triangle of Accountability in Service Delivery Policy makers Stronger Voice Stronger Compact Poor people Providers More Choice and Participation Source: World Bank 2003. who deliver those services are recruited and managed by the woreda leadership. Local competition for these leadership positions is significant, and officeholders are keen to reach their service delivery targets. This arrangement encourages local authorities to listen to citizens and take responsibility for results. On the other hand, the direct route of accountability is reinforced by measures that strengthen financial transparency and accountability (educating citizens on local budgets and service delivery goals), social accountability (improving citizens’ opportunities to provide feedback directly to local administrators and service providers), and impartial procedures to redress grievances (instituting the inde- pendent Ethiopian Institution of the Ombudsman, for example). Recognizing the critical role of good information to improve development outcomes, the PBS Program also finances a range of surveys and other data col- lection and management efforts. Capacity building includes training in financial management and procurement for local woredas, in addition to other types of training, based on demand. Woreda-level block grants primarily support locally recruited staff in five decentralized sectors (education, health, agriculture, water and sanitation, and rural roads). With some minor exceptions, woredas have very little revenue of their own, and they receive no other regular and predictable transfers that can be used to support staff. For that reason, total woreda spending in sectors relevant to the PBS Program is used here as a proxy for the block grants cofinanced by PBS partners and the government. Health sector spending focuses on health extension workers (HEWs), education spending on teachers, agriculture spend- ing on agricultural extension workers (also referred to as development agents), water sector spending on recurrent costs for water systems, and road spending on road maintenance staff. In health, education, and agriculture there is a one-to-one correlation between woreda-level block grant spending and HEWs, teachers, and agricultural extension workers. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Introduction and Background 5 Box I.1  The Indirect Route of Accountability Through Policy Makers to Service Providers In Ethiopia’s strong push to achieve middle-income status by the mid-2020s, the federal, re- gional, and woreda governments all assign high priority to attaining development results quickly, especially in the basic service sectors. The Ethiopian state has high expectations of performance from civil service staff at the decentralized levels of government. These indi- viduals, who are responsible for moving their communities toward specific development out- comes, are at the heart of the relationship between policy makers and service providers. Policymakers operating at the local level are held accountable to their higher-ups, and the Ethiopian government expects officials at every level to deliver results. The accountability between policy makers and service providers is one of the three sides of the “accountability triangle” depicted in the World Development Report 2004. Policy makers have some degree of control over providers’ results, and well-intentioned policy makers can put into place the ap- propriate mechanisms and incentives to drive positive outcomes. Two elements of Ethiopia’s system influence the degree of accountability and deserve note. First, decentralization has created woreda offices that are positions of inherent prestige, conferring a certain level of power and influence in the community. To some extent, competi- tion for these desirable positions leads local authorities to feel responsible for results and to remain open to feedback from citizens. Second, the relationship between local policy makers, service providers, and the community is strengthened by recruiting individuals locally: Teach- ers, health workers, and other local residents hired to provide services are apt to feel an intrin- sic motivation to serve their community. According the World Development Report 2004, “a critical element in the policymaker-pro- vider relationship is information.” A woreda’s improvements in the enrollment rate or the pro- vision of antenatal care, for example, create an impression of the relative success or failure of the woreda’s officials. Similarly, knowledge of provider absenteeism enables woreda officials to hold specific service providers to account. The financial transparency and social account- ability components of the PBS Program, as well as efforts to strengthen monitoring, evalua- tion, and related reporting from each sector, are providing additional opportunities to im- prove the accountability of service providers. The PBS Program is the primary means by which the World Bank and other development partners assist Ethiopia to achieve the MDGs. At the program’s inception in 2006, Ethiopia had just started to register noticeable improvements in human development indicators, albeit from low absolute levels. The Project Appraisal Document for the first phase of PBS clearly stated the challenge of maintaining that nascent progress in the face of political and economic uncer- tainty: “These gains represent the first steps on a steep development trajectory that Ethiopia will need to sustain if it is to have a chance to meet any of the MDGs” (World Bank 2012a). Eight years on, it is obvious that through the PBS Program, Ethiopia is meeting that challenge, and more. The current five-year third phase of PBS costs US$6.4 billion, of which about half is financed by the government. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 6 Introduction and Background Table I.1 shows the evolution of Federal Block Grants (FBGs) provided to the regions by the federal government between 2005/06 and 2012/13; Table I.2 shows the evolution of Woreda Block Grants (WBGs) provided by the regions to the woredas. While helping to achieve many of the eight MDGs, the PBS Program focuses principally on three—achieving universal primary school education (Goal 2), reducing child mortality (Goal 4), and improving maternal health (Goal 5)—and also has a direct impact on the eradi- cation of extreme poverty and hunger (Goal 1) and the promotion of gender equality (Goal 3). The program takes advantage of Ethiopia’s well-developed system of fiscal decen- tralization, which was well under way when the program began, to channel resources to the Table I.1  Federal Block Grants and PBS Program Disbursements 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 Federal Block Grants (FBGs) to the regions (ETB million) 7,071.5 9,365.0 13,532.5 16,554.8 19,555.7 25,555.8 30,576.4 35,555.3 Federal govern- ment expendi- tures (Treasury source only) (ETB million) 21,856.5 26,976.2 34,717.4 41,053.8 53,063.7 67,769.6 83,470.2 89,064.2 FBGs as a share of federal government expenditures 32.4 34.7 39.0 40.3 36.9 37.7 36.6 39.9 PBS donors’ contri- bution to FBGs (US$ million) 91.0 446.7 477.9 437.2 453.9 542.6 329.5 444.5 Average exchange rate (ETB/US$) 8.6810 8.7943 9.2441 10.4205 12.8909 16.1081 17.7686 18.1947 PBS donors’ contri- bution to FBGs (ETB million) 790.0 3,928.8 4,417.7 4,556.3 5,851.7 8,741.0 5,855.1 8,087.4 PBS donors’ contribution to FBGs (%) 11.2 42.0 32.6 27.5 29.9 34.2 19.1 22.7 IDA’s contribution to FBGs (%) 11.2 9.4 12.4 14.7 14.8 12.9 12.1 9.5 FBGs as a share of total regional expenditures 76.2 88.5 85.3 86.2 79.8 80.4 62.8 54.4 Source: Ministry of Finance and Economic Development for the data on FBGs and federal government expenditures, PBS donors for data on PBS disbursements, and National Bank of Ethiopia for exchange rate data. Note: ETB = Ethiopian birr; IDA = International Development Association; PBS = Promoting Basic Services. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Introduction and Background 7 Table I.2  Total Woreda Recurrent Expenditures by Region ETB million Region 2005/06 2006/07 2007/08 2008/09 2009/10 2010/11 2011/12 2012/13 Tigray 457.3 556.9 797.6 941.9 1,040.0 1,426.7 1,837.8 2,265.0 Afar 126.9 119.2 191.2 247.9 278.3 371.3 405.5 481.4 Amhara 1,385.4 1,363.1 2,160.2 2,956.2 3,136.5 3,787.2 5,522.7 6,429.4 Oromia 2,056.7 2,109.8 3,070.7 4,312.9 4,401.9 5,866.5 6,957.0 8,887.9 Somali 217.8 232.0 367.2 282.0 459.3 626.9 972.3 1,123.0 Beneshangul- Gemuz 100.3 124.1 148.2 167.1 216.7 293.1 452.7 488.7 SNNP 1,250.4 1,539.2 1,910.1 2,097.9 3,012.6 3,950.5 5,161.1 6,279.5 Gambella 69.4 84.2 127.8 140.1 151.3 202.1 286.9 351.0 Harari 0.0 24.2 24.2 33.9 38.3 48.8 60.1 73.7 Dire Dawa 0.0 14.4 27.8 28.3 31.7 31.8 41.4 52.2 Addis Ababa 507.2 534.9 656.9 1,187.0 1,178.9 1,447.3 1,922.9 2,581.0 All regions 6,171.3 6,702.2 9,481.8 12,395.1 13,945.3 18,052.0 23,620.3 29,012.9 All regions (excluding Addis Ababa) 5,664.1 6,167.3 8,825.0 11,208.1 12,766.5 16,604.8 21,697.4 26,431.8 Source: World Bank based on Ministry of Finance and Economic Development data. woreda level to finance recurrent expenditures and efficiently deliver and expand services in the five sectors mentioned earlier. At the local level, this fund- ing is used to deploy frontline development workers and in this sense is linked to meeting MDG targets. HEWs, for example, promote preventive behaviors and provide basic health interventions, such as vaccinations and family planning information, which serve to reduce child mortality and improve maternal health. The hiring of additional teachers is related directly to increasing net enrollment rates in order to achieve universal primary education. Funds transferred to the woredas in the first two phases of the PBS Program helped to hire more than 100,000 new primary school teachers, 38,000 HEWs, and 45,000 agricultural extension agents, dramatically expanding access to their respective services. Despite the impressive scale of the PBS Program, prior to this study no empirical analysis was done to assess the effects of PBS spending on outcomes, partly because of the difficulty of obtaining data on local expenditures and out- comes over time. One of this study’s signal accomplishments is that it has com- piled comprehensive expenditure data through time and consolidated it into a database with output indicators for health and education for the same time period—a prerequisite for the analysis that follows. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 8 Introduction and Background Notes 1. Appendix C briefly describes the evolution of the modern Ethiopian Federal State. 2. See, for example, Tiebout (1956), Oates (1972), Besley and Coate (2003), and Faguet (2000, 2012). 3. World Bank (2014). This programmatic knowledge series complements another pro- grammatic knowledge series in Ethiopia, led by the Poverty Reduction Economic Management network, which will include a poverty assessment, public expenditure reviews, and various reports on growth and employment. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 CHAPTER 1 Conceptual Framework and Methodology A primary objective of this study is to assess the relationship between expendi- ture at the woreda level to provide basic services in five sectors and key outputs and outcomes for those sectors, focusing particularly on health, education, and agriculture (for which the best data are available).1 As noted, woreda-level spending in those three sectors is associated strongly with key service outputs, such as numbers of teachers, health extension workers (HEWs), and agricultural extension workers. This study focuses on such outcomes as the net enrollment rate, rates of vac- cination and contraceptive use, and agricultural yields of a variety of crops (see figure 1.1 for a diagrammatic representation of the Promoting Basic Services [PBS] results chain). An important consideration is that the link between Figure 1.1 Conceptual Model of the Results Chain of PBS Spending Inputs Delivery Outputs Outcomes Spending for salar- Teachers and Direct impact: ies of teachers and health/agricultural • Net enrollment ratio health/agricultural extension workers • Pupil-teacher ratio • Gains in literacy extension workers hired • Increased life expectancy Catalytic impact: • Lower infant mortality • Number of children • Increased agricultural vaccinated income • Women receiving antenatal care • Number of people using contraception • Yield from agricultura l crops Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  9 10 Conceptual Framework and Methodology spending on staff and results cannot be one to one; service availability and use need to be assessed as well. Unfortunately such data are not available for the most part, with the very limited exception of the health sector, where Demographic and Health Survey (DHS) data show that the use of health ser- vices, especially by the poor, is catalyzed by HEWs (Wang et al., forthcoming). Other caveats concern the quality of the data and local effects of spending out- side the PBS Program (box 1.1). A parallel objective of this study is to assess the incidence of local expendi- tures by wealth quintile, in line with the World Bank’s objective of achieving shared growth by targeting the bottom 40 percent. Another objective is to inves- tigate the allocation of block grants at the woreda level in relation to Ethiopia’s constitutional objective of providing additional resources to historically under- served ethnic groups. A final objective, related to the triangle of accountability shown in figure I.1, is to assess the components of the PBS Program that are designed to strengthen citizen voice in improving basic services. This study makes use of a database comprising woreda-level recurrent expen- ditures and information on outcomes for a variety of health and education indi- cators at the woreda level between 2008 and 2011. These data are complement- ed by Agriculture Sample Survey (AGSS) data and information from the 2007 census on the demographic characteristics of each woreda, including its overall Box 1.1  Some Caveats on Interpreting the Results of This Study Please note that while the results reported in the text are for log-linear regressions, we also estimated quadratic and linear regressions in addition to probits. The detailed results are in appendix D. Clearly, the local development results achieved in the sectors studied here are affected by spending from other levels of government, such as capital spending and nonsalary recurrent costs. Examples include the cost of learning materials (financed separately in education) and the costs of facilities above the health center level (financed separately in health). PBS Pro- gram spending in agriculture and roads includes only spending on extension workers and road maintenance staff. In other words, it is important to keep in mind that the results pre- sented here for PBS spending—in health, for example—show the catalytic effect of health extension workers; the contributions of capital spending, spending on medicines, and other types of expenditures are taken as given. To cite another example, the results of PBS spend- ing in agriculture would not be possible without the contribution of other spending, public and private, and reflect only the catalytic effects of agricultural extension workers. Another important caveat is that, as always, the results presented here are subject to the quality of the data used in the analysis. The management information systems for health and especially education are fairly reliable and improving (not least through PBS support), al- though much room for improvement remains. Fortunately, the results for health and educa- tion can be confirmed by national-level analysis of DHS data. The analysis for agriculture uses Agriculture Sample Survey data, which are also considered reliable. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Conceptual Framework and Methodology 11 population, ethnic composition, the percentage of the population that is rural and urban, and other variables. Econometric modeling is used to assess the asso- ciation of increased local spending with the expansion of basic services and out- comes, and by extension the efficiency of the PBS Program in meeting its devel- opment objective. The study also verifies the woreda-level findings in health and education using the household data from DHS 2005 and 2011, including the use of limited dependent variable (primarily probit) regressions to predict the impact on household behavior of interactions with HEWs. Cross-time pooled regressions with the log of the outcome variable of interest on the left-hand side were run to evaluate the result of the log per capita sector expenditure, controlling for shares of rural/urban population and ethnicity (which constitute a good proxy for historical lags in development). A variable for time was included to isolate the time-series effects from other effects. The main text presents results from the log-linear regressions, which are preferred because they eliminate the effects of extreme values and allow for declining returns to scale. Linear regressions and quadratic regressions were also estimated (see appendix A). The indicators evaluated are drawn mostly from the PBS Results Framework. No reliable data on outcomes at the woreda level were available for agriculture, water supply, and roads, so alternative approaches were used. For agriculture, zonal data on outcomes were available to assess the effectiveness of spending on agriculture, taking the average per capita woreda spending on agriculture as a proxy for services offered by agricultural extension workers. More severe data constraints for water supply and roads made it necessary to resort to desk reviews.2 Conducting subnational empirical work on Ethiopia involves numerous chal- lenges and constraints. Relatively little data is collected, the data are often of poor quality, and few attempts are made to systematize the results into any obviously comparable framework. A few illustrations are telling. Until very recently, fiscal data on subnational expenditures in health, education, agricul- ture, water, and roads were available only for Ethiopia Fiscal Year (EFY) 2003. Their geographic identifying codes and names do not match those used for census data, in which geographic codes and names vary in unpredictable but pervasive ways from fiscal data. The last census counted some 740 woredas, zones, and regions, but the fiscal dataset includes more than 850. Consolidating these two sources yielded a dataset of 989 subnational units, 250 more than in the census. The analysis consists of three key stages. Stage I follows the approach of Faguet (2012) and Faguet and Sánchez (2013), examining the relationship between woreda-level spending in each sector (figure 1.2) and results in those sectors. Ideally first differences would be used on the outcome left-hand-side variables, but the few years of data currently available make this procedure impractical. Instead, trends are estimated by placing levels on the left-hand side. Stage II is an incidence analysis of woreda-level spending by quintile, using the wealth quintile breakdown from the DHS survey. Improvements in health and education outcomes are allocated to each quintile by pro-rating the expenditure increases to each quintile by the improvement in outcomes achieved for that Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 12 Conceptual Framework and Methodology Figure 1.2  Woreda-Level Expenditure in the Five PBS Sectors, 2011 Water Roads expenditure expenditure 2% 1% Health expenditure 17% Agriculture Education expenditure expenditure 18% 62% Source: World Bank calculations using data collected from regional and Woreda bureaus of the Ministry of Finance and Economic Development. quintile, controlling for the average improvement for all groups.3 The results indicator for education was the net enrollment rate, which is directly affected by the number of teachers recruited for the woreda. For health, the results indicator is an average of four indicators that are affected directly by locally recruited HEWs: increased use of contraception, increased rates of immunization, increased use of antenatal care, and increased use of skilled birth attendants. In stage III, limited dependent variable regressions are used to examine the link between woreda expenditure and sectoral results for different wealth quin- tiles4 to estimate probabilities that households will act in a certain manner. This three-stage analysis was limited to sectors with reliable data on woreda-level results and corresponding household survey data on health and education. For education, the third-stage test was considered unnecessary because of the direct linkage between woreda-level spending, which is mostly for teachers, and the net enrollment rate. The third-stage test was needed for health, however, because the linkage between spending on HEWs and results is less direct. In that case, the link between contact with HEWs and health sector outcomes was tested using DHS data. Given that no output data were available for agriculture at the woreda level, the effect of agricultural extension services was analyzed by plot-size quintile. No national or local analysis could be done for water supply; a recent national census found that the data were unreliable, and in any event the sector accounts for only 2 percent of woreda-level expenditure. Nor was in-depth analysis possible for roads (which account for only 1 percent of woreda-level expenditure). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Conceptual Framework and Methodology 13 Notes 1. Note that this study does not look at the quality or effectiveness of service delivery, however. Phase three of the PBS Program supports the collection of data on service delivery indicators for health and education, which will be used in future studies of those issues. 2. Normally, DHS has good water supply data, but definitional changes between the two rounds of DHS in Ethiopia have made the water supply data incompatible. 3. This methodology was adopted because there were no data on direct use of different services by quintile. This approach assumes a link between woreda-recruited staff and results. 4. Education and health quintiles were based on DHS quintiles. Quintile estimates for agriculture used land-size quintiles. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 CHAPTER 2 Citizen Direct Voice and Accountability Before discussing the effectiveness and equity analyses of woreda-level spending to provide basic services, it is important to document how the context of accountability is changing in Ethiopia with respect to service provision. Mechanisms that enhance financial transparency, increase social accountability, and permit grievances to be addressed all strengthen citizens’ capacity to share their concerns effectively with service providers and local authorities (box 2.1). Note that one element of direct accountability that is not common in Ethiopia outside larger urban centers is the market for basic services, because private pro- viders are concentrated in major centers. Box 2.1  The Direct Route of Accountability—Citizen Direct Voice and Accountability Models in Ethiopia Decentralization brings service providers under the control of local governments and their constituents.a Staffing levels and allocations are centrally approved, but personnel are locally hired and managed—and in theory more directly accountable to the community. For ac- countability to occur, however, local residents must be able to understand the issues sur- rounding service delivery and the options for voicing their concerns. The Promoting Basic Services (PBS) Program is testing and in some cases scaling up complementary strategies to strengthen citizens’ voice and access to information. For example, the project’s financial transparency and accountability component makes information about local government budgets and spending publicly available in a simple, clear format. Through budget literacy training, citizens learn how to provide feedback to local authorities on budgets and the provision of basic services. The project’s social accountability component uses structured social accountability tools such as community score cards, citizen report cards, participatory budgeting, and inter- face (direct) meetings between service providers and service users to make services better, more accessible, and more effective. These tools are being used in more than 340 woredas by about 4.5 million service users. A rigorous, independent evaluation based on a randomized box continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   15 16 Citizen Direct Voice and Accountability Box 2.1  The Direct Route of Accountability—Citizen Direct Voice and Accountability Models in Ethiopia (continued) controlled trial is under way to examine the impact of these tools and determine whether they should be made available nationally. The project also plans to strengthen and expand the use of grievance redress mecha- nisms. A grievance redress mechanism provides the opportunity for an impartial third party to review a transaction that has taken place between the government and a citizen or a group of citizens, who assert that a government service or benefit has been denied, or some other harm has occurred. The Ethiopian Institution of the Ombudsman (EIO) and the regional Griev- ance Handling Offices offer these services. The EIO is an independent entity that has six re- gional branches and is accountable to the Federal Parliament. Grievance Handling Offices, established in all nine regional states, examine the validity of citizens’ grievances and find ways to rectify them. They are accountable to the respective chief administrator in their re- gional state and receive technical support and capacity building from the EIO. a. See Faguet (2004) for an in-depth discussion of the theory surrounding this topic, and Faguet and Wietzke (2006) for practical applications. Decentralization and the Quality and Appropriateness of Local Services Until woreda-level data from the period before decentralization are available, it will not be possible to estimate the effects of decentralization per se (in the sense of the transfer of expenditure and decision-making authority from upper to lower tiers of government) on financial flows, service provision, or substantive outcomes. These are the questions on which the decentralization literature typi- cally focuses, and when data permit, they will be addressed in future reports. Other, more micro-level questions can be examined with the data assembled for this study, however. They concern the effects of woreda-level discretion and decision making on the quality of services and their appropriateness to local conditions. These questions are distinct from “big discretion” issues, such as how to divide resources between investment and expenditure, or how to allocate spending between schools, hospitals, and roads. “Small discretion” questions con- cern such issues as how to target a vaccination campaign, or when and exactly how to maintain a road, or which side of a road or stream is the best location for a new school. Decisions of this kind, which can increase the efficiency and reduce the cost of public services by tailoring services to highly specific local conditions and needs, are the bread and butter of local governments. Small discretion has other important effects on service delivery. Centrally established staffing norms determine the local funding available for particular sectors, but the actual hiring of staff occurs at the woreda level. Local hiring can influence the attendance of teachers, health extension workers (HEWs), and agricultural extension workers, because the hiring and paying authority is present in the immediate area where they work. Absenteeism is near 50 percent in many Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Citizen Direct Voice and Accountability 17 countries that centrally recruit and deploy teachers. The limited evidence avail- able for Ethiopia shows that teacher absenteeism is about 15 percent.1 The same study finds that only 10.4 percent of respondents said that health staff were frequently absent, including the HEWs who serve the immediate community. Nearly all respondents (96 percent) reported that agricultural extension workers were present in their kebele (subdistrict), and 95 percent found them to be use- ful. This is the effect of small discretion. More services are delivered because more service providers are on the job, and if they are not, evidence indicates that a significant percentage of citizens will complain to their local government about it (Kamurase and Alibhai 2014). In sum, the limited data show that decentraliza- tion increases accountability, and that even if a small number of citizens make their voices heard, they can have a positive effect on the quality of service delivery. Ethiopia’s decentralization law also allows citizens from different ethnic or language groups to request their own woreda and education in their own lan- guage. Such actions appear bound to introduce a greater measure of local accountability and increase learning effectiveness. Financial Transparency and Accountability Over the last five years, financial transparency and accountability (FTA) tools designed under the PBS Program have been rolled out to disclose information on regional and local budgets, service delivery targets, and accomplishments. More than 90 percent of Ethiopia’s woredas and city administrations now post such information publicly. Aside from posting information, regions also use radio and television programs, brochures, t-shirts, and other means to disseminate informa- tion. These actions have revolutionized transparency, as this key information was not released to the public before 2006. The increase in transparency has not been even across the country. A recent survey (Selam Development Consultants 2013, 34–35) finds that 42.5 percent of respondents have seen FTA templates posted in their jurisdictions. It is encour- aging that citizens are not merely passively receiving this information but are responding to it. More than 230,000 citizens across all woredas and city admin- istrations have received budget literacy training to improve their understanding of the budget process and service delivery issues. As a result, over time more citi- zens are discussing budget and spending priorities with their local officials, pro- viding feedback on budget execution, and monitoring project implementation. The survey mentioned earlier finds that 37 percent of respondents have dis- cussed budget information with their woreda or city officials. Another 26 percent have discussed the information with other citizens. Nor is this communication one sided. Some 37 percent of survey respondents indicated that woreda and city officials had sought citizens’ views on budgetary issues, and 43 percent revealed that officials had also sought citizens’ views on the quality of public services. (See appendix B for more detail.) Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 18 Citizen Direct Voice and Accountability Social Accountability The social accountability component of the PBS Program complements FTA initiatives by supporting civil society organizations and other means for citizens to provide feedback to local administrators and service providers.2 In an earlier phase of PBS, 86 woredas tested tools such as community score cards, citizen report cards, and participatory budgeting. They also promoted interface meetings between citizens and local authorities to provide feedback on service delivery. Following a positive evaluation of those pilot efforts, the current phase of PBS features an expanded social accountability component.3 The evaluation finds that more than 84 percent of the individuals surveyed in pilot areas responded posi- tively to social accountability initiatives, which had increased citizens’ awareness of their rights, responsibilities, and entitlements to basic services. After service providers and users drew up joint service improvement plans, basic services improved, and so did the quality of the engagement between citizens and service providers. These findings bear out the prediction of the model shown in figure I.1. During the current phase of the PBS Program, participating woredas are con- tinuing to use the tools piloted in the first phase. Interface meetings between users and providers of services, along with woreda and kebele officials, continue as well, in addition to the development of agreed joint action plans monitored by joint committees of service users and providers. In 224 woredas, 49 Social Accountability Implementing Partners are working to assist public service pro- viders to deliver better-quality services in education, health, agriculture, water and sanitation, and rural roads, in response to feedback from communities and citizen groups. Other social accountability tools will be considered carefully, based on their value added, including participatory planning and budgeting, bud- get tracking, gender-responsive budgeting, and service charters. (See appendix B for details.) The government is exploring strategies to sustain and strengthen these social accountability initiatives. Initial thoughts on such a policy highlight the need to strengthen linkages and synergies between social accountability and FTA, evalu- ate the implementation of recommendations on linkages, and develop more medium- and long-term options based on those evaluations. Grievance Redress Mechanisms A study conducted under the auspices of the PBS Program in 2011 determined that Ethiopia had established several grievance redress mechanisms in different regions under various programs (Randolph and Edjeta 2011). The mechanisms varied significantly in their legal underpinnings, procedures, the government entity to which they were responsible, accountability, and the finality and enforcement of grievance findings. Through dialogue and technical and financial support, the PBS Program aims to strengthen the Grievance Handling Offices at the regional state level by improving public awareness of the services they provide, delivering technical Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Citizen Direct Voice and Accountability 19 assistance to develop a common standard of grievance redress procedures, and developing the capacity of grievance handling officers. The PBS Program also supports the opening of Grievance Handling Offices in all regional states, as well as branches of the Ethiopian Institution of the Ombudsman (EIO). Currently the PBS Program finances capacity building and training workshops conducted by EIO for regional EIO branch offices and regional grievance handling officers. It also supports studies to aid in standardizing and improving the grievance redress system across the country. This important contribution supports and expands the forums where ordinary citizens can air their concerns. Notes 1. Based on initial results from a baseline survey conducted for an assessment of the social accountability component of the PBS Program. Data on absenteeism include not just absence from school but also absence from the classroom. See Kamurase and Alibhai (2014). 2. Under PBS, social accountability activities are financed through a programmatic mul- tidonor trust fund administered by the International Development Association (IDA), though without IDA resources. 3. IPE Global (2010). The assessment was not a formal evaluation but relied more on recapitulative data. As mentioned, the new phase of PBS was designed to include a full, rigorous evaluation, for which a baseline survey has been completed. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 CHAPTER 3 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture A Positive Association between Woreda Block Spending and Results in Education Over half of the resources channeled to local governments under the Promoting Basic Services (PBS) Program are used to hire primary school teachers.1 Demographic and Health Survey (DHS) data indicate that the net enrollment rate for primary school climbed from 68 percent to 82 percent between 2005 and 2011, and the completion rate rose from 34 to 49 percent. The analysis that follows sought to determine if there was an association between per capita expenditures on primary education at the woreda level and the net enrollment rate (NER) and pupil-teacher ratio (PTR). The third phase of the PBS Program is expected to achieve specific improvements in both of these education-related indicators, which are included in the results framework. The analysis did not consider nonsalary recurrent costs in education, which are covered by other levels of government.2 The exclusion of those costs is not expected to affect the results of the analysis to a great extent, given that teacher costs constitute more than 90 percent of recurrent costs in basic education. The model does consider the impact of capital costs such as school buildings, how- ever, which can drive enrollments. Capital expenditure is based on per capita expenditures at the regional level, where most capital spending occurs. Table 3.1 presents the main coefficients of interest from the log-linear regression analysis. Both net enrollment rate and pupil-teacher ratio show a strong and significant relationship with woreda-level per capita recurrent expenditure, after controlling for the effects of rural/urban percentage and ethnicity. For every additional Ethiopian birr (ETB) per capita of woreda education spending, net enrollment rate increases by 0.20 percent. These results are all significant at the 1 percent level and robust to changes in specification. The results for pupil-teacher ratio are similar. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   21 22 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture Table 3.1  Association of Log of Per Capita Education Expenditure with Log of Education Outcomes Dependent Independent variable variable/indicator Coefficient/(SE) Significance Log of expenditure Log of net enrollment 0.2757 *** rate (.0280) Log of pupil-teacher –0.2229 *** ratio (.0203) Source: Based on Poverty and Social Impact Analysis Database from 2008 to 2011. Note: Number of observations is 2,583 for NER and 2,695 for PTR. Standard errors given in parentheses; *** indicates significance at 1percent level. A Positive Association between Woreda Block Grant Spending and Results in Health About 20 percent of PBS Program resources channeled to the local level are used in the health sector, mainly to hire frontline health extension workers (HEWs) (box 3.1). HEWs do not provide extensive curative services; their assignment is to support public health in several important ways. They promote healthy lifestyle practices in the community and refer complicated cases, such as difficult pregnan- cies or severe child malnutrition, to a health center, where they can be treated by health professionals with more extensive training. They also provide services that can be scheduled periodically, the most important being immunizations, family planning (insertion of contraceptive implants), and antenatal care. HEWs also mobilize pregnant women to seek care from skilled birth attendants—nurses, Box 3.1  Ethiopia’s Health Extension Program and Health Extension Workers The Health Extension Program is an innovative, community-based program to deliver health services. The program’s underlying assumption is that health can be enhanced in communi- ties by raising awareness, diffusing knowledge through training, and creating model families, which exemplify sound health practices (completing immunization schedules, for example) and can share their knowledge and behavior. The program has infrastructure and human staffing components. Every village with 5,000 residents is supposed to construct a health post and recruit two female health extension workers (HEWs) from the community. After one year of training in which they learn to imple- ment 16 health extension packages, these women return to their community as frontline health care staff. The 16 packages are as follows: • Disease prevention and control (3): – Human immunodeficiency virus (HIV)/acquired immune deficiency syndrome (AIDS) and other sexually transmitted infections; tuberculosis prevention and control – Malaria prevention and control – First aid emergency measures box continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture 23 Box 3.1  Ethiopia’s Health Extension Program and Health Extension Workers (continued) • Family health (5): – Maternal and child health – Family planning – Immunization – Nutrition – Adolescent reproductive health • Health education and communication • Hygiene and environmental sanitation (7): – Excreta disposal – Solid and liquid waste disposal – Water supply and safety measures – Food hygiene and safety measures – Healthy home environment – Control of insects and rodents – Personal hygiene. The HEWs’ main task is to increase awareness of disease prevention strategies. They may also supervise the intake of medication for tuberculosis and antiretroviral treatment for HIV/AIDS, conduct rapid diagnostic tests for malaria and administer malaria drugs, attend uncomplicated childbirths, and collect vital statistics; they are not allowed to administer antibiotics. A critical aspect of their work is to refer patients to nearby health centers as needed. The Health Extension Program was piloted in five regions in 2002/03, with encouraging results. By mid-2008/09, the Federal Ministry of Health had successfully deployed over 30,190 HEWs throughout Ethiopia. Various evaluations have found the Health Extension Program to have very tangible effects on rural people’s awareness of disease prevention, family health, hygiene, and environmental sanitation. The program now covers all of the country. health officers, trained midwives, or physicians—whose positions are also funded from the local health budget, although they account for a smaller percentage of expenditure. In other low-income countries, decentralized expenditure has been shown to improve health outcomes (Faguet and Ali 2009). Does this hold in Ethiopia? Woreda spending on health was analyzed with respect to four health-related outcomes: the Penta 3 vaccination rate; the percentage of pregnant women who receive antenatal care (ANC); the contraceptive acceptance rate (CAR); and the percentage of deliveries by skilled birth attendants (DelSBA). The first two out- comes are included in the results framework for the third phase of the PBS Program, while the latter two were included in the framework for the second phase. Each outcome is directly related to local spending on health. Control Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 24 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture Table 3.2  Association of Log of Per Capita Health Expenditure with Health Outcomes Indicator Coefficient/(SE) Significance Log of expenditure Log of Penta 3 vaccinations −0.0578 ** (0.0276) Log of Penta 3 vaccinations on 0.1116 *** expenditure lagged one year (.0374) Log of antenatal care 0.0861 ** (0.0347) Log of contraceptive acceptance −0.0064 NS rate (0.0410) Log of contraceptive acceptance 0.1161 *** on expenditure lagged one year (.0491) Log of deliveries by skilled birth 0.2669 *** attendants (0.0747) Source: Based on Poverty and Social Impact Analysis Database from 2008 to 2011. Note: Number of observations is 1,664 for Penta 3, 2,277 for antenatal care, 2,243 for contraceptive acceptance rate, and 2,154 for deliveries by skilled birth attendants. Standard errors given in parentheses; *** indicates significance at 1 percent level, ** at 5 percent level. variables included the per capita capital expenditure (at the regional level).3 Table 3.2 presents the results of the log-linear regressions. Increased health expenditures by woredas appear to improve the rates of Penta 3 vaccination, antenatal care, contraceptive use, and deliveries by skilled attendants. All of these results are significant, and all are robust to changes in specification. As with education, the model for health considers only per capita woreda-level spending on HEWs and health center staff; it does not consider nonsalary and capital costs, although clearly they can affect the results presented here. For example, HEWs can convince mothers to immunize their children, but the vaccines (financed separately from the health workers) must also be available. The indicators included in the model directly reflect the responsibilities of HEWs, who account for most of the local recurrent health expenditure. HEWs mobilize the community during immunization campaigns. They are the frontline providers of antenatal care and the primary spokespersons in the community on the importance of contraception in family planning. Technically speaking, HEWs are not skilled birth attendants, but their referrals to those professionals influence the proportion of women who give birth with the assistance of a skilled attendant. A smaller percentage of local health expenditure is directed to health centers. By federal mandate, each of the 3,000 health centers must be staffed by one to three health officers, depending on the characteristics of the host community. Each health officer is backed by a team of about four nurses and has at least one trained midwife. Each of the indicators assessed here can potentially be influ- enced by HEWs as well as health center staff, but except for deliveries by skilled Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture 25 birth attendants, most of the effects found in the analysis would be expected to arise from HEWs. Deliveries by skilled attendants would be influenced more equally by HEWs and health center staff. A Positive Association between Woreda Block Grant Spending and Results in Agriculture Ethiopia relies heavily on agriculture, which accounts for nearly half of gross domestic product (GDP) and employs around 80 percent of the population. The PBS Program supports agriculture by financing recurrent costs at the woreda level, mostly consisting of the costs of employing agricultural extension workers. Extension workers are trained to teach community members about the benefits of improved farming techniques, such as the appropriate use of improved seed and fertilizer, the importance of irrigation and erosion preven- tion, and other practices. About 20 percent of PBS funds go to the agricultural sector. In agriculture, just as in health and education, expenditure levels increased rapidly between 2008 and 2011. Concomitant with this increase, the percentage of fields that use extension services has more than doubled. Figure 3.1 shows this trend for both variables. The agricultural data presented here and used in much of the analysis in the rest of this section are collected by the Central Statistical Agency as part of its annual Agriculture Sample Survey (AGSS). This survey obtains information from around 45,000 households and includes data from 500,000 fields in Ethiopia’s nine regions as well as the city-state of Dire Dawa. The data are aggregated at the zonal level to be more representative of the underlying population. Figure 3.1  Growth in Extension Services and Woredas’ Expenditure on Agriculture, 2008–11 14 30 Percent of fields using extension 12 25 10 20 ETB per capita 8 services 15 6 10 4 2 5 0 0 2008 2009 2010 2011 Year Percent fields using extension services Average agriculture expenditure per capita Source: World Bank, using AGSS data. Note: ETB = Ethiopian birr. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 26 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture Table 3.3  Effect on Farmers’ Use of Extension Services of ETB 1 Per Capita Spending on Agricultural Extension Workers Indicator Coefficient/(SE) Significance Field using extension services 0.0008 ** (0.0004) Field using improved seed 0.0002 * (0.0001) Field using fertilizer 0.0007 ** (0.0003) Source: Based on probit models using cross-time pooled AGSS data from 2008 to 2011. Note: AGSS data aggregated to the zonal level. Number of observations is 191. Controls include current and previous years’ deviations from average rainfall (calculated as the average between 1996 and 2011), zonal poverty rate, percentage of the zone’s population that is rural, and the ethnic groupings used in previous regressions. Standard errors are given in parentheses; ** indicates significance at 5 percent level, * at 10 percent level. Zonal spending on agriculture significantly affects the use of a variety of improved farming techniques, as seen in the result of cross-time pooled regressions on AGSS data between 2008 and 2011 (table 3.3). The regres- sions for agriculture, like those for health and education, control for the per- centage of the population that is rural and the ethnic composition of the zone. For each zone, deviation from the average rainfall (1996–2011) in the current and previous year is included as a predictor. In all regressions where an agri- cultural variable is included on the left-hand side, both the current year’s as well as the previous year’s deviation from average rainfall is a significant predictor. Although the effects appear to be relatively small, several aspects of improved farming are significantly and positively associated with zonal per capita spending on agriculture. For every additional US$1 per capita spent, for example, the probability that a field in a given zone will benefit from extension services increases by about 0.2 percent (assuming an exchange rate of ETB 20 to US$1). Coefficients for the use of fertilizer and improved seed are smaller, but these practices remain positively and significantly associated with spending. In contrast, irrigation is not associated with higher spending. This result may be related to the higher capital costs associated with irrigating fields, and it would therefore depend more on zonal expenditure in the water sector than in agriculture. Figure 3.2 shows the overall basket of crops produced by private farmers in 2011 by the amount produced (in quintals). Cereals (including barley, teff, wheat, sorghum, maize, oats, and rice) make up almost 75 percent of production. A further 15 percent consists of pulses (such as beans, chickpeas, and lentils) and root crops (such as potatoes, carrots, and onions). Enset, fruit crops, and coffee, which can be grown only in certain geographical areas, represent a smaller pro- portion of overall production. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture 27 Figure 3.2 Production by Crop, 2011 Fruit Crops 2% Coffee 1% Enset 3% Other 1% Oilseeds 3% Vegetables 3% Root Crops 6% Pulses 9% Cereals 72% Source: World Bank estimates, AGSS, 2011. Note: “Other” includes hops and chat. The figure does not include sugar, which is not often grown on private land. The overall objective of spending on agriculture is to increase the productivity of farmers’ fields (measured by yield, expressed here as the number of quintals produced per hectare of cultivated land). Cross-time pooled regressions between 2008 and 2011, with the log of yield of a specific category of crop (cereals, pulses, and so on) as the dependent variable, show strong positive relationships with log per capita recurrent agricultural spending and crop yields. Once again, control variables included each zone’s percentage of rural population, overall poverty rate, the current and past years’ deviation in rainfall from the average, and ethnic composition (this last variable helps to account for geographical heterogeneity in crop production). Of the eight categories of crops grown in Ethiopia, five—cereals, vegetables, enset, fruit, and coffee, which represent about 85 percent of national agricultural production—show a positive and significant relationship with agricultural spend- ing (table 3.4). These results, combined with the positive effect of spending on agricultural extension services, indicate that PBS Program funds play an impor- tant role in increasing farmers’ productivity. As discussed, that role should be viewed as catalytic, because increased productivity also depends on other public and private spending, as well as investments in infrastructure, which are not reflected in this analysis. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 28 Effectiveness of Woreda Block Grant Spending on Education, Health, and Agriculture Table 3.4  Association of Log of Per Capita Spending on Agricultural Extension Workers with Log of Yield Dependent variable/indicator Independent variable (yield in quintals/ha) Coefficient/(SE) Significance Log of expenditure Log of cereal yield 0.128 *** (0.0489) Log of pulse yield −0.020 NS (0.0645) Log of root crops yield 0.320 NS (0.2266) Log of vegetable yield 0.582 *** (0.1330) Log of oilseed yield −0.118 NS (0.1900) Log of enset yield 2.397 *** (0.5766) Log of fruit yield 1.790 *** (0.2669) Log of coffee yield 1.267 *** (0.2342) Source: Cross-time pooled AGSS data from 2008 to 2011. Note: AGSS data aggregated to the zonal level. Number of observations: 167 (cereals), 159 (pulses), 152 (root crops), 167 (vegetables), 151 (oilseeds), 90 (enset), 162 (fruits), and 137 (coffee). Standard errors given in parentheses; *** indicates significance at 1 percent level; NS is not significant. Notes 1. Secondary education expenditure is split between the woreda and the region, so it was not included in the analysis. 2. For example, improvements in the quality of education are supported under the General Education Quality Improvement Project, funded by the Government of Ethiopia, World Bank, DFID, United States Agency for International Development, and many others. 3. As noted in the discussion on methodology, other control variables included the rural/ urban percentage and ethnicity. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 CHAPTER 4 How Equitable Is Decentralized Spending at the Woreda Level? The analyses in the previous sections find that decentralized spending on education, health, and agriculture at the woreda level is efficient, contributing positively and significantly to major development objectives in each sector. But has that spending also been equitable? This section examines several facets of the equity question, including the extent to which investments and outcomes may have varied by income group, gender, region, or ethnic group. The analysis should shed some empirical light on whether decentralized spending on service provision has fur- thered the government’s objective of shared growth by targeting individuals in the two lowest wealth quintiles, and whether it has supported the constitutional objec- tive of providing additional resources to historically underserved populations. Woreda Block Grant Spending on Health Improves Outcomes, Especially for the Rural Poor As mentioned, Ethiopia has met the Millennium Development Goals (MDG) of reducing the mortality rate by two-thirds in children under five years. This impressive achievement is due in no small part to rising local health budgets that pay for staff who provide the services and promote the behaviors that reduce under-five mortality. Similar improvements have been seen for other health indi- cators. Because Ethiopia conducted its Demographic and Health Survey (DHS) one year prior to the Promoting Basic Services (PBS) Program and then five years into its implementation, the data can reveal how the program may have affected individuals grouped by income level and location (in a rural or urban area). As most PBS resources flow to rural areas, this breakdown makes it possible to view the effects of health extension workers (HEWs) more clearly. Figure 4.1 demonstrates the substantial improvement seen among rural peo- ple for all six health indicators between 2005 and 2011. It also shows that—in the majority of cases—the improvement was greatest among the bottom two wealth quintiles.1 Contraceptive acceptance rates rose over 200 percent for the bottom two wealth quintiles, more than twice as much as in the top two Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   29 30 How Equitable Is Decentralized Spending at the Woreda Level? Figure 4.1  Percent Improvement in Selected Health Indicators Among Rural People Between 2005 and 2011, by Wealth Quintile Percent improvement between 2005 and 2011 DHS 280 230 180 130 80 30 −20 n EW EW te ) y io ld lit ra at so ta rH rH cin ce or ar ,o ,o an m ye ac ife ife el sv e 5 iv w w ev 1– sle rf id id pr y( de ,m ,m ea e lit Un iv M se se ta pt ur ur or ce ,n ,n m ra or or ild nt ct ct Ch Co Do Do C: y: er AN liv De Poorest Second Middle Fourth Richest Source: World Bank, based on DHS 2005, 2011 data. Note: ANC = Antenatal care; HEW = health extension worker. quintiles. Similarly, for measles vaccination rates, antenatal care, and delivery by a skilled birth attendant, the poorest quintile showed the largest improvement. All of these outcomes are highly influenced by the presence of health workers in the field, which is financed by intergovernmental fiscal transfers (IGFTs) through the PBS Program. The poorest quintiles also saw improvements in child and under-five mortal- ity, more so than the richest quintile, but to a lesser degree than other quintiles. Outcomes for these indicators, unlike those for the other indicators, are particu- larly susceptible to factors outside the control of local recurrent health expendi- ture. Food security, nutritional status, and a mother’s education all play a larger role than local health spending in driving these outcomes and can have an out- sized effect among poorer individuals. It is telling that improvements in child mortality are larger than improvements in under-five mortality. The implication is that services—in the form of the health extension packages offered by HEWs (see box 3.1)—target children more than infants. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 How Equitable Is Decentralized Spending at the Woreda Level? 31 The probability of improvement in some of these outcomes can be estimated based on whether a household has had contact with an HEW (using 2011 DHS data on contacts with HEWs). For example, the 2011 survey asked whether a “family planning specialist or HEW” had visited the household in the past year. As many of the outcomes fit nicely into a binomial response (use of contracep- tion versus no use of contraception, for instance), it is possible run a probit model in which the response to the question about an HEW visit is the right-hand-side variable and the left-hand side is either “success” or “failure” in achieving the health outcome. This approach assumes that an HEW can deliver multiple mes- sages per contact, and not (for example) just a message about family planning. The model coefficients are then used to determine predicted probabilities for each specific outcome. In fact, a household was more likely to achieve an outcome if an HEW had visited (table 4.1). This association holds across all outcomes for urban and rural locations, and for three of the four outcomes examined by wealth quintile, despite problems related to the small samples for some quintiles.2 For example, pregnant women in the poorest rural households were 15 percent more likely to receive antenatal care and 12 percent more likely to vaccinate their child against measles if they had received a visit from an HEW than if they had not. The richer the household, the more likely it was to achieve many of the health outcomes examined here, probably because of unobserved factors, the foremost of which is probably women’s education. Table 4.1  Predicted Probabilities for “Successful” Health Outcomes by Place of Residence and Wealth Quintile if an HEW Visited the Household in the Past Year Delivery by ANC with doctor, doctor, nurse, Income Contraceptive Measles nurse, midwife, midwife, Residence quintile use vaccination or HEW or HEW Urban Total 0.071*** 0.176*** 0.124*** 0.110*** Rural Total 0.094*** 0.144*** 0.107*** 0.017** Poorest 0.043** 0.115*** 0.148*** NS Second 0.078*** 0.106** 0.073*** NS Middle 0.069*** 0.123*** 0.097*** NS Fourth 0.108*** 0.181*** 0.096*** 0.022* Richest 0.127** 0.189* 0.130** NS Source: Based on probit models. Note: Number of observations varies by quintile, location, and outcome (see appendix A for details); *** indicates significance at 1% level, ** at 5 percent, and * at 10 percent. NS is not significant. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 32 How Equitable Is Decentralized Spending at the Woreda Level? Incidence Analysis of Woreda Block Grant Spending on Health and Education In examining the equity effects of the PBS Program, an important step was to analyze the incidence of health and education spending at the woreda level by wealth quintile. The most common form of incidence analysis relies on facility and service information by income group, but data constraints eliminated that approach. Instead, the analysis relied on DHS 2005 and 2011 data. The incidence of increase in per capita expenditures in the health and education sectors over 2005–11 was allocated by quintile based on the ratio of the improvement in results in that quintile to the average improvement in results (see appendix A, equation [A.4]). For education, one-result variable (net enrollment rate) was con- sidered, whereas for health, the analysis used an average of the four results that can be affected by HEWs (vaccination, contraceptive use, antenatal care, and deliveries by skilled birth attendants). Given that the HEW system started between the two rounds of the DHS, the data constitute something of a natural experiment on the impact of HEWs. As noted, data on the use of HEWs’ services from DHS 2011 indicate that HEWs have had a proportionately very strong effect on the four results just listed, particularly among individuals in the bottom quintiles. For education, woreda-level spending appears to be pro-poor. The bottom 40 percent benefited from 56 percent of that expenditure, and the top quintile benefited from 13 percent. The incidence of per capita woreda-level spending on education was more than 2.5 times higher for the bottom quintile than for the top quintile (table 4.2). For health, a similar pattern emerged: 63 percent of the health expenditure at the woreda level accrued to the bottom 40 per- cent, whereas 10 percent accrued to the top quintile.3 For someone in the bot- tom quintile, the benefit incidence of per capita woreda-level health spending was more than three times higher than it was for someone in the top quintile. A forthcoming study of multiple countries,4 using a standard methodology, corroborates the results obtained with the less-orthodox methodology used here. It also finds that woreda-level spending (financed through the PBS Program) is significantly more pro-poor than overall spending in education and health. Specifically, the overall incidence of all public education spending on the bottom 40 percent was 33 percent (it was 34 percent for health). Table 4.2  Incidence of Woreda-Level Spending by Wealth Quintile Multiple by which Bottom 40% Top 20% bottom quintile share (%) share (%) exceeds top quintile Education spending 56 13 2.7 Health spending 63 10 3.4 Combined education and health spending 58 12 2.9 Source: World Bank estimates based on several sources according to the methodology in appendix A of this book. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 How Equitable Is Decentralized Spending at the Woreda Level? 33 Catalytic Effect of Spending on Extension Services by Plot Size Does the catalytic effect of spending on agricultural extension services vary by plot size? This question was explored using Agriculture Sample Survey (AGSS) data grouped by plot-size quintile. Regression analysis was done for each quintile to estimate the effect of each birr of spending on agricultural extension services on the probability that farmers in that quintile would use improved techniques (defined as improved seed, irrigation, or fertilizer). Table 4.3 clearly shows that agricultural extension spending at the woreda level increases the use of improved farming techniques across all plot-size quin- tiles. The magnitude is relatively low in the bottom quintile, which may reflect financial constraints on purchasing inputs and otherwise investing in improved techniques. If so, this constraint could be addressed by improving access to rural finance. Gender Equity Analysis for Woreda Block Grant Spending Another major question related to equity is whether there is a gender bias in PBS expenditures on health, education, and agriculture. In health, the answer is an obvious yes. Women in the community, along with children, are the pri- mary focus of outreach by HEWs and benefit from most of the packages they deliver (box 3.1). Spending on health by the woredas can also be considered pro-woman because so much of that spending is positively associated with increasing women’s access to services that have a strong impact on reducing maternal mortality—including contraception, antenatal care, and assisted deliveries. A more precise way to answer the gender equity question is to look at the association of expenditure with data on results disaggregated by gender, but such data are available only for education, and even then only for certain indicators. Table 4.4 provides key coefficients for education recurrent expenditure on net enrollment rate and net intake rate (NIR) at the primary level, disaggregated by Table 4.3  Effect of ETB 1 Spending on Agricultural Extension Workers on Probability of Using Improved Farming Techniques, by Plot-Size Quintile, 2011 Plot-size quintile Any improved technique Smallest and poorest 0.000322*** Second smallest 0.000344*** Middle 0.000524*** Fourth 0.000667*** Largest and richest 0.000818*** Source: Based on probit models using AGSS data. Note: Cross-sectional data with 303,242 observations (which are plots, not total household land holdings). Standard errors given in parentheses; *** indicates significance at the 1 percent level. NS is not significant. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 34 How Equitable Is Decentralized Spending at the Woreda Level? Table 4.4  Association of Log of Education Expenditure with Log of Net Enrollment Rate and Log of Net Intake Rate, by Gender Dependent variable/ Independent variable indicator Coefficient/(SE) Significance Log of expenditure Log of NER grades 1–8, male 0.115 *** (0.0116) Log of NER grades 1–8, female 0.158 *** (0.0137) Log of NIR grades 1–8, male 0.390 *** (0.0240) Log of NIR grades 1–8, female 0.430 *** (0.0269) Source: Based on Poverty and Social Impact Analysis Database from 2008 to 2011. Note: Number of observations is 2,583 for NER and 2,464 for NIR; *** indicates significance at 1 percent level. NS is not significant. NER = net enrollment rate; NIR = net intake rate. gender. The coefficient for female primary school students is slightly higher than for males for both the net enrollment and intake rates, but this difference is not significant. The important finding here is that there is no negative bias against females in expenditure on education, the sector that receives the bulk of PBS Program funds. By contrast, evidence from the agricultural sector implies an apparent bias in outcomes (female farmers are less likely to benefit from extension services), which was also found in a study by the International Food Policy Research Institute (Ragasa et al. 2012). The reasons for this bias are not clear, but it could reflect a wide range of circumstances, such as a tendency for extension workers to target male farmers (87 percent of Ethiopia’s farmer are male), a tendency of women farmers to grow crops of different types or quality than those grown by men, or a tendency of female farmers to have smaller holdings. Table 4.5 shows the probability that improved farming techniques will be used on a field, disaggregated by the gender of the head of the household. About 13 percent of fields in the sample are farmed by female-headed households; such fields are significantly less likely to benefit from extension services, though the effect is relatively small. Again, in interpreting these results, it is important to bear in mind that many other factors also influence the adoption of improved farming techniques, the sample size for female-headed households is small, and the quality of the data could be improved. Further analysis is required to deter- mine whether extension workers are indeed preferentially targeting fields owned by men and to develop policy responses based on the evidence. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 How Equitable Is Decentralized Spending at the Woreda Level? 35 Table 4.5  Predicted Probabilities of Improved Farming Techniques Being Used on a Field, by Gender of Household Head, 2011 Probability that an improved Household head technique is used on a field Male 0.000027*** Female −0.000023*** Source: Based on probit models using AGSS data. Note: Number of observations is 303,242. Standard errors given in parenthesis; *** indicates significance at 1 percent level. Notes 1. Owing to the small urban population in Ethiopia, the samples for the wealth quin- tiles contained too few observations to be representative. 2. For example, only 1,000 deliveries attended by a doctor, nurse, midwife, or HEW were reported, making it difficult to obtain reliable estimates when the data were disag- gregated by wealth quintile. 3. Thirty-five percent of woreda health and education expenditures accrued to the bot- tom quintile. 4. See Woldehanna, Tsehaye, and Hill (forthcoming). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 CHAPTER 5 The Federal System’s Role in Helping Lagging Areas and Groups The Government of Ethiopia has made the development of historically disad- vantaged groups and areas a priority. Its system of fiscal decentralization trans- fers funds to local governments based on formulas applied at the federal and regional levels, and level of development is a primary element of those formulas. The following sections examine the allocation of subnational expenditures by geographical area and ethnic group. Regionally Disaggregated Analysis of Woreda Block Grant Spending Figure 5.1 is a simple scatterplot of per capita subnational expenditures across Ethiopia’s 718 woredas in 2010, in which each woreda is represented by a dot. Identification numbers assigned by region are on the x-axis, which naturally groups woredas by region, demarcated in red. The dashed yellow line shows the average expenditure in Ethiopia across all woreda, at Ethiopian birr (ETB) 183 per capita. Most woredas fall into a spending band of ETB 150–250 per capita. The lower outliers cluster near this band, whereas some outliers range as high as ETB 1,000 per capita. Taking ETB 385 per capita, 110 percent above the national average, as the cut-off. Table 5.1 provides data on the higher outliers. Seven of the woredas that spend the most on basic services, including the five highest observations nationwide, are in Gambella. More than 50 percent of the woredas in Gambella spend more than 110 percent of the national average on the basic service sectors, and so do six woredas in Beneshangul-Gemuz, repre- senting 30 percent of the woredas in that region. Gambella and Beneshangul- Gemuz are among the regions that historically have been the most economically deprived. Another six woredas where per capita spending on basic services is higher than the average are in Tigray. Among the much larger regions that have dominated government and the economy since the days of the empire, Amhara has only two woredas where spending exceeds ETB 385 per capita, and Oromia Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   37 38 The Federal System’s Role in Helping Lagging Areas and Groups Figure 5.1  Woreda Total Recurrent Expenditure Per Capita, 2010 1,200 1,000 Total recurrent expenditure per capita., 2003 Gambella 800 Afar Beneshangul Gemuz Oromia Tigray Somali Amhara Southern 600 Nations 400 Harar 200 Avg = ETB 183/capita 0 Dire Dawa 0 100 200 300 400 500 600 700 800 900 Woreda number Source: World Bank estimates based on data described in appendix A of this book. Table 5.1  Woredas Spending More Than 110 Percent of the National Average on Basic Services, by Region, 2010 No. woredas above Percent of woredas 110% national above 70% national Region No. woredas average average (%) Tigray 46 6 13 Afar 30 0 0 Amhara 137 2 2 Oromia 278 6 2 Somali 48 0 0 Beneshangul-Gemuz 20 6 30 SNNP 144 2 1 Gambella 13 7 54 Harari 1 0 0 Dire Dawa 1 0 0 Source: World Bank. Note: SNNP = Southern Nations, Nationalities, and Peoples (Region). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 The Federal System’s Role in Helping Lagging Areas and Groups 39 has six. The observations for the lowest spending levels appear in Amhara, Oromia, and Somali. In terms of expenditure per capita, then, the current sys- tem broadly favors Ethiopia’s historically disadvantaged regions at the expense of the historically dominant ones. At the same time, table 5.1 shows that the federal system also prioritizes underserved areas in regions with the larger, more established ethnic groups, such as the Amhara, Oromo, and Tigrayan. The next section looks in greater detail at Woreda Block Grant spending by ethnic group. Ethnically Disaggregated Analysis of Woreda Block Grant Spending Ethiopia’s complex history explains some of the regional differences in develop- ment.1 In the late nineteenth century, areas consisting of today’s Beneshangul- Gemuz, Gambella, Southern Nations and Nationalities, Afar, Oromia, and Somali Regions were brought under the feudal system of the Ethiopian Empire, based in the highlands of north-central Ethiopia. Relations between the newly integrated areas and the historic center of the empire were troubled. Owing to the structural weakness of the center, successive Ethiopian governments did not command effective control over the periphery. Similarly, exploitive economic policies pro- mulgated by the center resulted in visible marginalization and relative underde- velopment at the periphery of the empire. This dynamic persisted through the socialist era. It was not until 1991, when the Ethiopian People’s Revolutionary Democratic Front came into power, that ethnic federalism took root in Ethiopia. A primary objective of this new government was to reverse the policy of eth- nic homogenization and promote “equitable and broadly shared” economic growth among all regions. Achieving this objective involved affirmative action in the form of increased budget allocations to historically less developed regions. Preferential treatment of previously disadvantaged ethnic groups within Ethiopia is helping to create a more equitable base for development. The small lagging groups appear to be faring well, yet at the woreda level, the largest lagging group—the Somali—is not receiving the additional resources per capita war- ranted by its development status. The reason for this disparity appears to be that during the four years covered in this analysis, the Somali Region transferred only 49 percent of its Federal Block Grant resources to woredas, compared to the average of 73 percent for all other regions.2 Ethiopia has 91 registered linguistic groups and 93 registered ethnicities, mak- ing it impractical to include a full set of dummies for either of those variables, let alone both, in the econometric estimations. Most of these language and ethnic groups account for less than 1 percent of the population. Given that the five largest ethnicities are heavily concentrated in their five home regions, controlling for region in the regression models is substantially similar to controlling for eth- nicity. For that reason, the econometric models focus on smaller ethnic groups, particularly a subset of historically disadvantaged groups. Using demographic data from the 2007 census, the analysis focuses on the Nuwer and Anyiwak3 from Gambella. All other ethnicities, representing less than 1.5 percent of the population, are combined into a separate group (“other small”). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 40 The Federal System’s Role in Helping Lagging Areas and Groups Total recurrent expenditure per capita at the woreda level between 2008 and 2011 is then compared for these categories versus the largest ethnic groupings in the country, omitted from the estimations, such as Amhara, Oromo, and Tigrayan. Other estimations not reported here substituted linguistic dummies for ethnic dummies, with no significant change in the results. Dummies for Ethiopia’s 12 regions also mirror the main findings with respect to large versus small ethnic/ linguistic groups, and they did not substantially change any of the other findings. Table 5.2 provides the key results. It shows that ethnicities with a smaller proportion of the population receive a larger share of recurrent expenditure, particularly the Anyiwak from Gambella, with ETB 770 more per capita than larger groups. The Nuwer receive substantially more per capita as well—ETB 142. As the positive significant coefficient for the “other small” category shows, ethnic groups in that category, which are more likely to be at a relative historical disadvantage in development, receive a significantly higher proportion of recur- rent expenditure. Table 5.2  Selected Ethnicities’ Influence on Total Woreda Recurrent Expenditure Per Capita Indicator: Ethnic percentage Coefficient/(SE) Significance Nuwer 141.5 *** (36.92) Anyiwak 769.5 *** (43.82) Gemuz 280.3 *** (39.03) Konso −10.5 NS (71.85) Somalia −66.5 *** (14.26) Afar −0.5 NS (17.56) Other small 32.9 ** (12.94) Historically developed groups (Oromo, Amhara, 24.9 ** Gurage, Tigray) (10.79) Source: Based on Poverty and Social Impact Analysis Database from 2008 to 2011. Note: “Other small” includes: Agew Awi, Agew Hamyra, Alaba, Argoba, Ari, Arborie, Bacha, Basketo, Bench, Berta, Bodi, Brayle, Burji, Bena, Chara, Dasenech, Dawuro, Debase Gewada, Derashe, Dime, Dizi, Donga, Fedashe, Gamo, Gebato, Gedeo, Gedicho, Gidole, Goffa, Guagu, Hareri, Hamer, Irob, Kefficho, Kembata, Konta, Komo, Koro, Koyego, Kunama, Karo, Kusumie, Malie, Mao, Mareko, Mashola, Mere, Meenite, Messengo, Mejenger, Mossiye, Mursi, Murle, Nao, Nyanegatome, Oida, Qebena, Qechem, Qewama, Shekecho, Sheko, Shinasha, Silte, Surma, She, Timebaro, Tsemay, Upo, Werji, Yem, Zeyese, and Zelmam; *** indicates significance at the 1% level, ** at 5%. NS is not significant. a. Data for Somali woredas may be understated, as some expenditures normally paid by woredas in other regions are paid regionally due to security and capacity constraints. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 The Federal System’s Role in Helping Lagging Areas and Groups 41 A central question posed in this study is whether woredas with a high propor- tion of historically disadvantaged ethnic groups are favored or disfavored by fiscal transfers. The higher distributions to such groups reported here indicate that PBS Program funding in fact targets their development needs. Notes 1. For more details see appendix C. 2. The average excludes Addis Ababa, a large city where woredas need to do less with regard to basic service provision and other development targets. Note that data on spending at the woreda level may be understated in Somali Region; some expendi- tures normally paid by woredas in other regions are paid regionally due to security and capacity constraints. 3. Nuwer is also spelled as Nuer. Note as well that the census spelling for Anyiwak is used here; alternative forms include Anuak and Anyiwa, among others. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Conclusions and Recommendations The evidence assembled here implies that decentralized spending at the woreda level is both effective and pro-poor. The estimates imply that the returns to this spending are far from decreasing, which means that Ethiopia has scope to increase spending and speed its achievement of the Millennium Development Goals (MDGs). In agriculture, the impact of decentralized spending could be improved by providing better access to rural finance for the poorest producers. Some caveats are needed with respect to these findings. In education, spend- ing on materials and other quality drivers is not included in Woreda Block Grants. Teachers’ salaries, which are included, usually constitute 80–90 percent of education spending. In health, spending in the woredas does not drive the results but should be regarded more as a catalyst to increase the effectiveness of systemwide spending. The agricultural extension workers supported through the Woreda Block Grants play a similarly catalytic role. Some historically disadvantaged areas are significantly favored under the cur- rent spending framework. Expenditure is broadly equal across Ethiopia’s wore- das, with the striking exception of a small number of woredas concentrated in the country’s most disadvantaged regions, which receive significantly greater resources. Woredas receiving the least resources are concentrated in the more developed, historically dominant regions. Four majority-Anyiwak woredas are noteworthy for receiving more public resources than all other woredas in the nation. At the very least, the additional resources appear to help disadvantaged groups catch up. The lack of data from the period prior to decentralization prevents an analysis of the effects of decentralization per se on financial flows, sectoral outputs, or substantive outputs. Even so, data gathered at the woreda level as decentraliza- tion unfolds can contribute to the debate about the effectiveness of woreda- controlled service provision. Based on the data available to date, this study finds that contrary to the predictions of some public management theories, the decen- tralized provision of services in Ethiopia is not increasing regional, ethnic, or gender inequalities in investment inputs or service outputs. Indeed, the opposite seems to be true for education and health, where the impact of Promoting Basic Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   43 44 Conclusions and Recommendations Services (PBS)-financed intergovernmental fiscal transfer (IGFT) resources was disproportionately high among the bottom two quintiles and women. In sum, support for decentralized services in Ethiopia appears to be an effective use of development partners’ resources from both an efficiency and equity perspective. The only exception to these findings is agriculture, for which the impact of PBS-financed IGFT expenditure was smaller for the bottom quintile. In this instance, a wide array of factors is likely to be at work, especially poor farmers’ inability to buy inputs or the poor quality of their land. Although quantitative evidence is unavailable at present, descriptive evidence from the first phase of the Ethiopia Social Accountability Program implies that structured feedback sessions involving citizens and service providers are strength- ening citizens’ participation in pilot areas. That evidence, together with the strong guidance emerging from governance and accountability theory, would appear to favor the continued application of social accountability tools and the development of policies to sustain their use in the Ethiopian context. This con- clusion will be verified when the impact assessment for the second phase of the Ethiopia Social Accountability Program becomes available (the only data avail- able now are from the baseline survey). A final important point is that the difficulty of conducting subnational empiri- cal work on Ethiopia cannot be overstated. Creating the database for this study required considerable effort and improvisation on the part of the research team. The resulting standardized database of woreda-level expenditures and character- istics, which will be made public, is a major output of this study. The data will be augmented with new data collected under the third phase of the PBS Program and used in future research related to the program. It is hoped that in time this dataset will become a useful tool for researchers and students elsewhere in Africa and beyond. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Appendix A Detailed Methodology and Data Sources Methodology This study relies on a database of woreda-level recurrent expenditure and out- come information for a variety of health and education indicators between 2008 and 2011. The data are complemented by information from the 2007 census on the demographic characteristics of each woreda, including population, shares of population by ethnic group and by rural or urban residence, and other variables. Econometric modeling is used to assess the impact of increased local spending on the expansion of basic services and outcomes, and by extension the efficiency of the PBS Program in meeting its development objective. Cross-time pooled regressions with the outcome variable of interest on the left-hand side were run to evaluate the impact of per capita sector expenditure, controlling for rural/urban percentage and ethnicity, over the four years. The outcome indicators evaluated are mostly drawn from the Promoting Basic Services (PBS) Results Framework. No reliable outcome data at the local level could be gathered for agriculture, water supply, or roads, however, so alternative approaches were used (zonal data on outcomes for agriculture, and desk reviews for water supply and roads). Note that expenditure on water supply and roads accounts for 3 percent of woreda-level spending; agriculture, education, and health, which are fully analyzed, account for the remaining 97 percent. National surveys (DHS [Demographic and Health Survey] and Agriculture Sample Survey [AGSS]) complement the analysis from the woreda-level database. The analysis proceeded in three stages. Stage I follows the approach of Faguet (2012) and Faguet and Sánchez (2013), examining the relationship between woreda-level spending (a proxy for service delivery, as discussed in the main text) in each sector on results in those sectors. That is, lnOmt = α + ζlnEmt + βRm + δCm + ηlnKmt + τt + εmt(A.1) where lnO is different outcome variables, such as the net enrollment rate or use of antenatal care, capturing key outcomes in each sector; E is yearly expenditure Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   45 46 Detailed Methodology and Data Sources per capita in the relevant sector; K is capital expenditure per capita; R is the percentage of rural population in each woreda; τ is a variable for the year to control for the time-series effect in this cross-time pooled dataset; and C is a vector of demographic controls, capturing the percentage of certain historically disadvantaged ethnic groups that are the focus of the analysis—all indexed by woreda m and, for expenditure, by year t. R and C are taken from the census and do not change over time. The above was also estimated using the quadratic form to check for decreasing marginal returns. Omt = α + ζEmt + λEmt2 + βRm + δCm + ηKmt+τt + εmt(A.2) As Yang (2012, 1) indicates: Interpreting the results of an analysis with log transformed data may be challeng- ing…. A log transformation is often useful for data which exhibit right skewness (positively skewed), and for data where the variability of residuals increases for larger values of the dependent variable. When a variable is log transformed, note that simply taking the anti-log of your parameters will not properly back transform into the original metric used. To properly back transform into the original scale we need to understand some details about the log-normal distribution. In probability theory, a log-normal distribution is a continuous probability distribution of a ran- dom variable whose logarithm is normally distributed. Finally, a linear version was also tested. Omt = α + ζEmt + βRm + δCm + ηKmt +τt + εmt(A.3) For all three equations, it is expected δO/δE≥0 and δO/δK≥0. Similarly, δlnO/ δlnE≥0 (the first partial derivative), while in equation (A.2) δ2O/δ2E≤0 (the sec- ond partial derivative). The expected results imply a positive association of the dependent variable with the independent variable (per capita woreda expendi- tures in the sector). Equation (A.2) also tests whether the association is positive with a declining marginal impact. The form estimated in equation (A.2) allows us to verify if the returns to scale are declining. The form in equation (A.1) reduces the effect of extreme values on the results. Stage II conducts an incidence analysis of woreda-level spending by quintile, using the wealth quintile breakdown from the DHS survey. The incidence analy- sis was done by allocating improvements in health and education outcomes to each quintile by pro-rating the expenditure increases to each quintile by the improvement in outcomes achieved for that quintile, controlling for the average improvement for all groups. That is, ΔEi = (ΔIi/ΔI)*ΔE(A.4) where ∆Ei is the per capita increase in expenditure for quintile i, ∆E is the overall increase in expenditure per capita, ∆Ii is the change in outcome change for quin- tile i, and ∆I the outcome change for all quintiles. The results indicator for Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Detailed Methodology and Data Sources 47 education was the net enrollment rate, which is directly affected by teachers recruited locally by the woreda. For health, the results indicator is an average of four indicators that are directly affected by locally recruited HEWs: increased use of contraception, increased rates of immunization, increased use of antenatal care, and increased use of skilled birth attendants. Stage III examines the link between woreda expenditure and sectoral results for different wealth quintiles, using DHS household survey data and limited dependent variable estimations as follows: Pi,j = f   (Ѳ, α)(A.5) In equation (A.5), Pi,j is the probability of improvement for household j in quin- tile i, and α indicates whether the household has had contact with a health worker. Ѳ is a “yes” or “no” variable, with “yes” being coded 1 and “no” coded 0. This three-stage analysis was limited to sectors with reliable data on woreda- level results and corresponding household survey data on health and education. In the case of education, the third-stage test was not considered necessary due the direct linkage between woreda-level spending, which is mostly for teachers, and the net enrollment rate. It was needed for health, however, because the link- age is less direct. Woreda-level spending is on HEWs; therefore, the link between contact with health extension workers (HEWs) and health sector outcomes was tested using DHS data. As discussed in the main text, data limitations prevented incidence analysis for agriculture, but the effect of agricultural extension services was analyzed by plot-size quintile. No in-depth analysis was possible for water supply or roads. The Poverty and Social Impact Analysis Database The database constructed for this study includes data from five ministries— Health, Education, Agriculture, Water and Energy, and Finance and Economic Development—as well as the Central Statistical Agency and the Disaster Risk Management and Food Security Sector of the Ministry of Agriculture. It also includes woreda-level poverty rates calculated by the Central Statistical Agency with World Bank technical assistance. As standardization of woreda codes and the transliteration of Amharic names into Latin script is not yet consistent across ministries, much time was dedicated to presenting information on woredas from various sources in one consistent format, based on the Population and Housing Census of Ethiopia (2007) codes. The woreda database includes information on expenditures by sector and year; key results for health and education; and information on ethnicity, the fre- quency of droughts, and other control variables. In addition, regional data on per capita capital expenditures and zonal data on crop yields have been linked to the woreda data. The plan is to update the database yearly, incorporating new results for each sector as they become available, and adding other useful data, such as local-level capital expenditure through time. The database has already been revised to incorporate education data broken down into grades 1–4 and 5–8, Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 48 Detailed Methodology and Data Sources which were not available for the analysis presented here. Other researchers are invited to supplement the database with data collected as part of their own research programs. The analysis of these data is a public good, and they should be publicly available for researchers in Africa and elsewhere. Table A.1 lists variables in the database along with the corresponding administrative levels and years. Table A.1  Variables in the Poverty and Social Impact Analysis Database, by Administrative Level Sector Variable Administrative level Years Demographic Population: Total Woreda, zone, region 2007 Population: Urban Woreda, zone, region 2007 Population: Rural Woreda, zone, region 2007 Religiona Woreda 2007 b Mother tongue Woreda 2007 Ethnicityc Woreda 2007 Expenditure Approved recurrent budget: Agriculture, Education, Woreda 2008, 2009, 2010, Health, Water 2011 Adjusted recurrent budget: Agriculture, Education, Woreda 2008, 2009, 2010, Health, Water 2011 Annual recurrent expenditure: Agriculture, Woreda 2008, 2009, 2010, Education, Health, Water 2011 Over/under recurrent expenditure: Agriculture, Woreda 2008, 2009, 2010, Education, Health, Water 2011 Capital expenditure Region 2008, 2009, 2010, 2011 Meteorological Number of years of drought between 1974 and 2007 Woreda NA Deviation from average rainfall between 1986 and Zone 2008, 2009, 2010, 2011 2011 Average rainfall between 1986 and 2011 Zone 2008, 2009, 2010, 2011 Poverty Poverty rate Woreda, zone 2011 Standard error of poverty rate Woreda 2011 Gini coefficient Woreda, zone 2011 Depth of poverty Woreda, zone 2011 Significantly poorer than average Woreda 2011 Water and Safe water access Woreda, zone 2011 sanitation Latrine access Woreda, zone 2011 Safe water usage Woreda, zone 2011 Safe water access at health facilities Woreda, zone 2011 Safe water access at schools Woreda, zone 2011 Any latrines at schools Woreda, zone 2011 Improved latrines at schools Woreda, zone 2011 table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Detailed Methodology and Data Sources 49 Table A.1  Variables in the Poverty and Social Impact Analysis Database, by Administrative Level (continued) Sector Variable Administrative level Years Education Gross enrollment rate: Grades 1–8, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 1–8, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 1–8, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 1–4, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 1–4, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 1–4, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 5–8, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 5–8, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Gross enrollment rate: Grades 5–8, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–8, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–8, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–8, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–4, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–4, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 1–4, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 5–8, total Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 5–8, male Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net enrollment rate: Grades 5–8, female Woreda, zone, region 2007, 2008, 2009, 2010, 2011 Net intake rate: Grades 1–8, total Woreda 2007, 2008, 2009, 2010, 2011 Net intake rate: Grades 1–8, male Woreda 2007, 2008, 2009, 2010, 2011 Net intake rate: Grades 1–8, female Woreda 2007, 2008, 2009, 2010, 2011 Apparent intake rate: Grades 1–8, total Woreda 2007, 2008, 2009, 2010, 2011 table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 50 Detailed Methodology and Data Sources Table A.1  Variables in the Poverty and Social Impact Analysis Database, by Administrative Level (continued) Sector Variable Administrative level Years Apparent intake rate: Grades 1–8, male Woreda 2007, 2008, 2009, 2010, 2011 Apparent intake rate: Grades 1–8, female Woreda 2007, 2008, 2009, 2010, 2011 Primary schools Woreda 2008, 2009, 2010, 2011 Pupil-teacher ratio Woreda 2007, 2008, 2009, 2010, 2011 Section-student ratio Woreda 2008, 2009, 2010, 2011 Dropout rate: Grade 1, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 1, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 1, female Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 2, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 2, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 2, female Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 3, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 3, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 3, female Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 4, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 4, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 4, female Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 5, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 5, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 5, female Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 6, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 6, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 6, female Woreda 2007, 2008, 2009, 2010 table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Detailed Methodology and Data Sources 51 Table A.1  Variables in the Poverty and Social Impact Analysis Database, by Administrative Level (continued) Sector Variable Administrative level Years Dropout rate: Grade 7, total Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 7, male Woreda 2007, 2008, 2009, 2010 Dropout rate: Grade 7, female Woreda 2007, 2008, 2009, 2010 Repetition rate: Grades 1–8, total Woreda, zone, region 2007, 2008, 2009, 2010 Repetition rate: Grades 1–8, male Woreda, zone, region 2007, 2008, 2009, 2010 Repetition rate: Grades 1–8, female Woreda, zone, region 2007, 2008, 2009, 2010 Health Number of health posts Woreda 2011 Number of health centers Woreda 2011 Contraceptive acceptance rate Woreda 2008, 2010, 2011 Antenatal care Woreda 2008, 2010, 2011 Deliveries by skilled birth attendants Woreda 2008, 2010, 2011 Prevention of mother-to-child transmission tested Woreda 2008, 2010, 2011 Penta3 vaccination rate Woreda 2008, 2010, 2011 Measles vaccination rate Woreda 2008, 2010, 2011 Fully immunized rate Woreda 2008, 2010, 2011 Tuberculosis detection rate Woreda 2008, 2010, 2011 Tuberculosis treatment rate Woreda 2010, 2011 Tuberculosis cure rate Woreda 2010, 2011 Agriculture Total cultivatable area (hectares) Zone 2008, 2009, 2010, 2011 Total production (quintals) Zone 2008, 2009, 2010, 2011 Number of fields Zone 2008, 2009, 2010, 2011 Source of irrigationd Zone 2008, 2009, 2010, 2011 Number of fields irrigated Zone 2008, 2009, 2010, 2011 Number of fields using extension services Zone 2008, 2009, 2010, 2011 Number of fields using fertilizer Zone 2008, 2009, 2010, 2011 Number of fields using improved seed Zone 2008, 2009, 2010, 2011 Cereals: Area cultivated (hectares)e Zone 2008, 2009, 2010, 2011 Cereals: Production (quintals) Zone 2008, 2009, 2010, 2011 Pulses: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 52 Detailed Methodology and Data Sources Table A.1  Variables in the Poverty and Social Impact Analysis Database, by Administrative Level (continued) Sector Variable Administrative level Years Pulses: Production (quintals) Zone 2008, 2009, 2010, 2011 Oilseeds: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Oilseeds: Production (quintals) Zone 2008, 2009, 2010, 2011 Vegetables: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Vegetables: Production (quintals) Zone 2008, 2009, 2010, 2011 Root crops: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Root crops: Production (quintals) Zone 2008, 2009, 2010, 2011 Fruit: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Fruit: Production (quintals) Zone 2008, 2009, 2010, 2011 Coffee: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Coffee: Production (quintals) Zone 2008, 2009, 2010, 2011 Chat: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Chat: Production (quintals) Zone 2008, 2009, 2010, 2011 Enset: Area cultivated (hectares) Zone 2008, 2009, 2010, 2011 Enset: Production (quintals) Zone 2008, 2009, 2010, 2011 Source: World Bank. Data sources for variables include: Annual Agriculture Sample Surveys, Central Statistical Agency, 2008–11; Education Management Information System Administrative Data, Ministry of Education, 2007–11; Ethiopia Demographic and Health Surveys, 2005 and 2011; Health Management Information System administrative data, Ministry of Health, 2007–11; Poverty Map with woreda and zonal poverty rates, Central Statistical Agency, 2012; Population and Housing Census of Ethiopia (using 10 percent sample), Central Statistical Agency, 2007; Water, Sanitation, and Hygiene (WASH) Census, Ministry of Water and Energy, 2012; Woreda Annual Progress Reports, Ministry of Health, 2007–11; and woreda-level recurrent expenditures in basic services, Ministry of Finance and Economic Development, 2007–11. a. Orthodox, Protestant, Catholic, Muslim, Traditionalist, and other. b. Afarigna, Agew Awinigigna, Agew Kamyrigna, Alabigna, Amarigna, Anyiwakgna, Arbogigna, Arigna, Arborigna, Bachagna, Basketigna, Benchigna, Bertagna, Bodigna, Brayligna, Burjigna, Benagna, Charigna, Dasenechgna, Dawurogna, Debosgna, Derashigna, Dimegna, Dizigna, Dongigna, Demegna, Felashigna, Fedashigna, Gamogna, Gebatogna, Gedeogna, Gedichogna, Gedoligna, Goffigna, Gumuzigna, Guragiegna, Guagugna, Hadiyigna, Hareriegna, Hamerigna, Irobigna, Kefficho, Kembatigna, Kontigna, Komigna, Konsogna, Koregna, Koygogna, Koyrigna, Kunamigna, Karogna, Kusumegna, Maliegna, Maogna, Marekogna, Mashiligna, Merigna, Me’enigna, Messengogna, Mejengerigna, Mossigna, Mursygna, Murlegna, Naogna, Nuwerigna, Nyangatomigna, Oromigna, Oydigna, Qebenigna, Qechemigna, Qewamigna, Shekacho, Shekogna, Shinashigna, Sidamigna, Shetagna, Somaligna, Surmigna, Shegna, Tigrigna, Timbarogna, Tsemayigna, Welatigna, Wergigna, Yemsagna, Zeysegna, and other. c. Affar, Agew Awi, Agew Hamyra, Alaba, Amara, Anyiwak, Argoba, Ari, Arborie, Bacha, Basketo, Bench, Berta, Bodi, Brayle, Burji, Bena, Chara, Dasenech, Dawuro, Debase Gewada, Derashe, Dime, Dizi, Donga, Fedashe, Gamo, Gebato, Gedeo, Gedicho, Gidole, Goffa, Gumuz, Guragie, Guagu, Hadiya, Hareri, Hamer, Irob, Kefficho, Kembata, Konta, Komo, Konso, Koro, Koyego, Kunama, Karo, Kusumie, Malie, Mao, Mareko, Mashola, Mere, Meenite, Messengo, Mejenger, Mossiye, Mursi, Murle, Nao, Nuwer, Nyanegatome, Oromo, Oida, Qebena, Qechem, Qewama, Shekecho, Sheko, Shinasha, Sidama, Silte, Somali, Surma, She, Tigrie, Timebaro, Tsemay, Upo, Welaita, Werji, Yem, Zeyese, Zelman, and other. d. River, lake, pond, harvesting, other. e. Yield for each type of agricultural product is the quotient of its production and area. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Appendix B Financial Transparency and Social Accountability under the PBS Program The World Development Report 2004: Making Services Work for Poor People advo- cated transparency and accountability mechanisms as effective tools for increas- ing the effectiveness of development projects focused on service delivery. Accordingly, when the Promoting Basic Services (PBS) program started its first phase in 2006, it included a financial transparency and accountability (FTA) component that provided approximately US$11 million to promote the disclo- sure of budget information to the general public at the local level. It also includ- ed a social accountability component that provided about US$6 million to pilot innovative methods of improving citizens’ knowledge of their rights and respon- sibilities with respect to public service delivery. Social accountability activities were implemented through civil society organizations operating where the pilots were located. The government’s initial unfamiliarity with FTA meant that activities were implemented slowly at first, but with additional support, four prototype FTA tools were developed, validated, and disseminated to regions by the end of the program’s first phase. These tools included laypersons’ guides to the budget pro- cess and three templates for disseminating information on the budget and expen- ditures at local level.1 The baseline FTA Perceptions Survey was completed in June 2009. Social accountability activities reached 80 woredas (approximately 12 per- cent of the country) under the first phase of the program. Activities focused on four basic service sectors: education, health, agriculture, and water and sanita- tion. The 12 civil society organizations engaged for this work used internationally proven best practices that were new to Ethiopia (the human rights-based approach to development, community score cards, citizen report cards, and par- ticipatory budgeting) as well as focus group and interface discussions and partici- patory planning and monitoring. Although the activities were new to most of the participating organizations, the program achieved its expected outcomes (the adoption of social accountability best practices and the creation of a learning initiative among participating citizens). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   53 54 Financial Transparency and Social Accountability under the PBS Program Given the potential demonstrated by the FTA and social accountability initia- tives in phase one, phase two included substantial resources for their scaling up and institutionalization. Regions started posting their budget appropriations to woredas on their websites. The FTA tools developed in phase one were extended to all regions, translated to regional languages, customized for local conditions in other ways, and disseminated to woredas. Budget literacy training, which had reached 1,500 individuals under the first phase, was expanded to reach 171,017. More generally, FTA was institutionalized in the public finance system. By the end of the second phase, approximately 94 percent of woredas were posting budget and expenditure data using the customized templates, and 53 percent of service facilities on average used the service delivery templates. Forty-three per- cent of citizens confirmed that woreda officials had sought their views concern- ing the quality of basic public services. The percentage of citizens reporting knowledge of the woreda budget increased from 13 percent in 2008 to 20 per- cent in 2011. The scaled-up social accountability component for phase two was based on findings from an independent evaluation of the pilot. A key difference in the second phase was the shift away from individual service facilities to a sectoral approach that engaged institutions and key actors in the community under a more comprehensive strategy. Activities were piloted in 86 woredas, in collabora- tion with 45 civil society organizations, with strong results. Community mem- bers received training, tools, and other mechanisms to assess service delivery and budget use and develop joint action plans with service users to improve the quality of services. Building on that foundation, FTA activities (budgeted at almost US$12.5 mil- lion) under the third phase of PBS focus on improving the quality of budget and expenditure information provided to the public by refining and simplifying the FTA tools. Continued budget literacy training will expand the awareness of bud- get processes and issues among citizens and woreda and kebele council members. The program will engage with other sectors to institutionalize FTA by disclosing their budget and service information to citizens. Social accountability activities (budgeted at US$20.5 million under phase three) will be extended to 170 wore- das, bringing the total number of woredas with such activities to 344, represent- ing about 4.5 million service users. Impact will be measured through a random- ized controlled trial. Existing methodologies will be refined, and new social accountability tools will be carefully considered based on their value added, including participatory planning and budgeting, budget tracking, gender-respon- sive budgeting, and service charters. Overall, the progression of FTA and social accountability activities under the PBS Program has brought about much greater transparency and accountability in publicly provided basic services. Note 1. Citizens did not use the tools in the program’s first phase, but the groundwork was laid for their use. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Appendix C The Center and the Periphery in Ethiopia: The Evolution of Today’s Federal State Ethiopia’s ethnic composition reflects its turbulent history. During the medieval period, Ethiopia primarily comprised the Tigray, Agaw, and Amhara peoples. Menelik II’s ascension to the throne in 1889 initiated a period of territorial expansion, radiating from the province of Shoa in present-day Amhara Region. The areas now known as Beneshangul-Gemuz, Gambella, Southern Nations and Nationalities, Afar, Oromia, and Somali Regions came under the Ethiopian Empire’s feudal system. After the Battle of Adwa in 1896 and Europe’s recogni- tion of Ethiopian statehood, border treaties were concluded with the surround- ing colonial powers, and the modern Ethiopian state was born. A Legacy of Troubled Relations between the Center and Periphery Relations between the newly integrated areas and the historic center of the empire were troubled. Menelik sent governors from the center to administer the new territories, but owing to the center’s structural weaknesses, successive Ethiopian governments did not effectively control those areas. Exploitive eco- nomic policies imposed on those areas led to their visible marginalization and relative underdevelopment, and resulted in limited integration among the border regions within Ethiopia (Mulugeta 2002). Yet the center tightened its hold over the periphery. According to noted his- torian Bahru Zewde, “The period after 1941 witnessed the apogee of absolutism. The tentative beginnings in this direction of the pre-1935 years matured into untrammeled autocracy. The power of the state reached a limit unprecedented in Ethiopian history” (Zewde 1991). The revised constitution of 1955 solidified the absolute powers of the emperor, claiming “His dignity…inviolable and His power…indisputable.” It also entrenched Amharic as the sole official language and the Ethiopian Orthodox Church as the national religion. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   55 56 The Center and the Periphery in Ethiopia: The Evolution of Today’s Federal State Socialist rule (1974–91) did not diminish the center’s hold over the periphery or change the prevailing economic policies of exploitation. Despite the regime’s appeal to a socialist ideology, the Derg was identified with an “Amhara suppressor” by the nationalist liberation movements (Weldemariam 2011). Any conduct promoting ethnic individualism, and thereby challenging the state’s integrity, was outlawed. The 1991 victory of the Ethiopian People’s Revolutionary Democratic Front over the Derg reversed more than one hundred years of ethnic homogeniza- tion. As Christopher Clapham says, “The overthrow of the Mengistu govern- ment in May 1991 amounted to more than the collapse of a particular regime. It effectively marked the failure of a project, dating back to Menelik’s accession in 1889 of creating a ‘modern’ and centralized Ethiopian state around a Shoan core” (Clapham 1994). Ethnic federalism now came to the fore in the Transitional Charter, which allowed the right to self-determination for the country’s various “nations and nationalities.” Like the Transitional Charter, the new constitution of 1995 recog- nized the rights of ethnic self-determination up to secession. It also created a federal government with nine regional states divided along ethno-linguistic lines: Tigray; Afar; Amhara; Oromia; Somali; Beneshangul-Gemuz; South Nations, Nationalities, and Peoples; Gambella; and Harari (table C.1). Table C.1 Ethiopian Regional Population Distribution and Ethnic Composition Estimated size in Region Population square kilometers Ethnic composition (%) Tigray 4,316,988 50,078 96.6 Tigray, 1.6 Amhara, 1.8 others Afar 1,390273 96,707 90.0 Afar, 5.2 Amhara, 1.6 Argobba, 1.2 Tigray, 2.0 others Amhara 17,221,976 159,173 91.5 Amhara, 4.9 Agaw, 2.6 Oromo, 1.0 others Oromia 27,158,471 353,006 87.8 Oromo, 7.2 Amhara, 0.9 Gurage, 4.1 others Somali 4,445,219 279,252 96.2 Somali, 2.3 Oromo, 0.7 Amhara, 0.8 others SNNP 14,929,548 112,343 19.4 Sidama, 10.6 Welayta, 8.0 Hadiya, 7.5 Gurage, 7.0 Gamo, 5.4 Kafficho, 5.4 Silt’e, 36.7 others Gambella 307,096 25,802 47.3 Nuer, 21.7 Anyiwak, 8.0 Amhara, 5.2 Opo and Komo, 4.4 Majangir, 13.4 others Beneshangul- 784,354 49,289 25.4 Berta, 21.7 Amhara, 20.9 Gemuz, Gemuz 13.6 Oromo, 7.7 Shinasha, 4.2 Agaw-Awi, 6.5 others Harari 183,415 311 56.4 Oromo, 22.8 Amhara, 8.7 Harari, 4.4 Gurage, 3.9 Somali, 1.5 Tigray, 2.5 others Source: Data from 2007 census conducted by the Central Statistical Agency. Note: SNNP = Southern Nations, Nationalities, and Peoples (Region). Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 The Center and the Periphery in Ethiopia: The Evolution of Today’s Federal State 57 A primary focus of the new government was economic integration and equi- table development. As Meles Zenawi, President of the Federal Democratic Republic of Ethiopia, declared in 1997, “It is only through fast economic growth that is broadly shared by the population that we can hope for sustainable peace. And, therefore, one of the most important pillars of our program is fast economic growth that is equitable and broadly shared among the population” (Meles 1997). The government adopted a policy of affirmative action toward historically disadvantaged regions, whereby Beneshangul-Gemuz, Gambella, Afar, and Somali would receive preferential treatment in budget allocations and enrollment in higher education. Centralization’s legacy is that developing regions are not yet fully integrated into the economy of the Ethiopian state, and few of their resi- dents participate in running the regional administrative structures. The gradual emergence of local administrative officials in the regions, establishment of a more equitable base for development, and better investment in education, health, social and physical infrastructure, and other areas all attest to the positive out- comes of Ethiopia’s federalization and affirmative action policy (Adegehe 2009). The Modern Ethiopian Federal State Decentralization of political, administrative, and fiscal authority to regional and local governments has been fundamental to the affirmative action strategy. The Government of Ethiopia’s strong commitment to decentralization and building a federal state is enshrined in the 1995 Federal Constitution. Decentralization should be seen as a work in progress, in which the underlying institutional arrangements for success are evolving and continue to require focused support. The first wave of decentralization occurred during the transitional period (1991–94) as state powers devolved to geographically defined ethno-linguistic groups and legislation was passed to create regional and woreda (district) councils. As needed, regions could establish zones as intermediaries between regional and district administrations. In addition to gaining the right to self-determination, the new regional units assumed a range of executive, legislative, and judicial powers. Within their borders, they exercised jurisdiction over social and economic devel- opment as well as basic service delivery. Regions were mandated to create the internal institutional arrangements to perform those functions, including a coun- cil, executive committee, judicial administration office, public prosecution office, audit office, police and security office, and service and development committee. Proclamation No. 7/1992 stipulated the sources of revenue for regional gov- erning units, which included tax revenues derived within their jurisdictions, fiscal transfers from the central government, domestic borrowing, and others. The lat- ter category was specified in Proclamation No. 33/1992. Capacity constraints prevented regional governments from performing their revenue-collecting assignments, however, and they remained highly dependent on central govern- ment grants to meet their new spending obligations in the social sectors. Despite these rather elaborate governing arrangements, the new regional gov- ernments remained subordinate to the central government. The regional councils Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 58 The Center and the Periphery in Ethiopia: The Evolution of Today’s Federal State were accountable to citizens living within their regional borders, yet legally they were also responsible to the central government’s Council of Representatives. Promulgation of the 1995 Federal Constitution signified another stage in Ethiopia’s decentralization. The constitution affirmed the roles and functions of federal versus regional government. The federal government was still charged with a broad range of functions and responsibilities (such as fiscal and monetary policy and international trade), but the regions and woredas gained responsibility for ensuring the delivery of basic services in their respective jurisdictions. The federal government retained authority to set policies and standards in each of the major social service delivery sectors. In 2002, decentralization was extended to the woreda governments, which became largely responsible for service delivery. Woredas receive block grants from their respective regional governments; the size of Woreda Block Grants is determined by formulas set by the regional gov- ernments, using methodologies broadly similar to those applied to federal– regional grants. Decentralization to the woreda level is crucial for achieving an equitable dis- tribution of prosperity. Ethiopian history, sociology, and ethnography all suggest that citizen-state relationships are framed by alternative norms of political cul- ture, power, and state-society relations, which have evolved locally, in the context of historical processes of state formation (World Bank 2013). With decentraliza- tion to the woreda level, budgets allocated for local expenditures have increased dramatically. Indicators of human development have similarly improved (net enrollment and vaccination rates are just two examples). Decentralization and affirmative action are mitigating the effects of economic policies that prevailed for more than a century, resulting in more broadly shared economic prosperity. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Appendix D Detailed Regression Results Table D.1  Stage 1: Education—Linear Regression Association of Per Capita Education Expenditures with Net Enrollment Rate Linear Regression Dependent variable: Net Enrollment Rate Independent variables: OLS Yearly expenditure per capita in Education 0.045 (0.0091) Yearly regional per capita Capital Expenditure 174.331 (1,580.6640) Percentage of rural population −12.283* (6.7861) Ethnicity controls  Nuwer −7.387 (24.1886)  Anyiwak −11.089 (27.8701)  Gumuz −16.048 (24.4713)  Konso −48.182 (44.7173)   Hist. Adv. −0.967 (7.2677)  Somali −62.222*** (10.1186)  Affar 77.755*** (11.4409)   Other Small −13.178 (8.5432) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   59 60 Detailed Regression Results Table D.1  Stage 1: Education—Linear Regression Association of Per Capita Education Expenditures with Net Enrollment Rate (continued) Time controls   T zero −3.548*** (1.2457)   T one −4.863*** (0.9834)   T two −0.301 (0.6894)   T three Omitted  Constant 104.449*** (9.4504) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.2  Stage 1: Education—Quadratic Regression Association of Per Capita Education ­Expenditures with Net Enrollment Rate Quadratic Regression Dependent variable: Net Enrollment Rate Independent variables: OLS Yearly expenditure per capita in Education 0.025 (0.0223) Yearly expenditure per capita in Education Squared −0.00005 (0.00003) Yearly regional per capita Capital Expenditure 55.277 (1,583.27) Percentage of rural population −10.394 (6.8810) Ethnicity controls  Nuwer −8.569 (24.1212)  Anyiwak −13.453 (27.8232)  Gumuz −18.196 (24.4323)  Konso −48.041 (44.5685)   Hist. Adv. −1.193 (7.2454)  Somali −61.075*** (10.1151)  Affar −77.279*** (11.4079)   Other Small −13.489 (8.5172) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 61 Table D.2  Stage 1: Education—Quadratic Regression Association of Per Capita Education ­Expenditures with Net Enrollment Rate (continued) Time controls   T zero −2.438* (1.4520)   T one −3.952*** (1.1585)   T two 0.087 (0.7375)   T three Omitted  Constant 100.071*** (9.8686) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.3  Stage 1: Education—Log Linear Regression Association of Per Capita Education ­Expenditures with Net Enrollment Rate Log Linear Regression Dependent variable: Log Net Enrollment Rate Independent variables: OLS Log yearly expenditure per capita in Education 0.276*** (0.0280) Log yearly regional per capita Capital Expenditure 0.040 (0.0293) Percentage of rural population 0.110 (0.0830) Ethnicity controls  Nuwer −0.042 (0.2949)  Anyiwak −0.545* (0.3428)  Gumuz −0.130 (0.3081)  Konso −0.467 (0.5273)   Hist. Adv. 0.004 (0.0861)  Somali −0.845*** (0.1267)  Affar −1.506*** (0.1490)   Other Small −0.132 (0.0120) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 62 Detailed Regression Results Table D.3  Stage 1: Education—Log Linear Regression Association of Per Capita Education ­Expenditures with Net Enrollment Rate (continued) Time controls   T zero 0.144*** (0.0266)   T one 0.091*** (0.0203)   T two 0.062 (0.0120)   T three Omitted Constant 3.263*** (0.2725) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.4  Stage 1: Education—Linear Regression Association of Per Capita Education Expenditures with Pupil-Teacher Ratio Linear Regression Dependent variable: Pupil-Teacher Ratio Independent variables: OLS Yearly expenditure per capita in Education −0.024 (0.0199) Yearly regional per capita Capital Expenditure 8,000.899*** (2,800.951) Percentage of rural population 15.676 (4.7215) Ethnicity controls  Nuwer −3.408 (13.6348)  Anyiwak −32.619* (17.6828)  Gumuz −21.557 (14.7049)  Konso −18.551 (26.2358)   Hist. Adv. −15.588*** (4.1635)  Somali 39.245*** (6.1928)  Affar −26.338*** (7.1540)   Other Small −7.671 (4.7664) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 63 Table D.4  Stage 1: Education—Linear Regression Association of Per Capita Education Expenditures with Pupil-Teacher Ratio (continued) Time controls   T zero 6.381** (2.6909)   T one 1.990 (2.1899)   T two 4.038** (1.6336)   T three Omitted  Constant 43.085*** (8.1009) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.5  Stage 1: Education—Quadratic Regression Association of Per Capita Education ­Expenditures with Pupil-Teacher Ratio Quadratic Regression Dependent variable: Pupil-Teacher Ratio Independent variables: OLS Yearly expenditure per capita in Education −0.110*** (0.0403) Yearly expenditure per capita in Education Squared 0.0002** (0.00006) Yearly regional per capita Capital Expenditure 7,472.043*** (2,805.786) Percentage of rural population 12.328** (4.9101) Ethnicity controls  Nuwer −0.908 (13.6528)  Anyiwak −34.057** (17.6673)  Gumuz −17.969 (14.7562)  Konso −18.807 (26.1966)   Hist. Adv. −14.821*** (4.1692)  Somali 35.936*** (6.3310) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 64 Detailed Regression Results Table D.5  Stage 1: Education—Quadratic Regression Association of Per Capita Education ­Expenditures with Pupil-Teacher Ratio (continued)  Affar −28.423*** (7.1947)   Other Small −7.151 (4.7640) Time controls   T zero 3.476 (2.9395)   T one −0.288 (2.3785)   T two 3.067* (1.6798)   T three Omitted Constant 53.924*** (9.2286) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.6  Stage 1: Education—Log Linear Regression Association of Per Capita Education ­Expenditures with Pupil-Teacher Ratio Log Linear Regression Dependent variable: Pupil-Teacher Ratio Independent variables: OLS Log yearly expenditure per capita in Education −0.223*** (0.0203) Log yearly regional per capita Capital Expenditure 0.063*** (0.0158) Percentage of rural population 0.202*** (0.0363) Ethnicity controls  Nuwer 0.234 (0.1176)  Anyiwak −0.600*** (0.1424)  Gumuz −0.252 (0.1299)  Konso −0.315 (0.2126)   Hist. Adv. −0.227*** (0.0329)  Somali 0.146** (0.0528) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 65 Table D.6   Stage 1: Education—Log Linear Regression Association of Per Capita Education ­Expenditures with Pupil-Teacher Ratio (continued)  Affar −0.556*** (0.0658)   Other Small −0.129*** (0.0389) Time controls   T zero −0.044 (0.0193)   T one −0.096*** (0.0153)   T two 0.007 (0.0099)   T three Omitted Constant 5.434*** (0.1460) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.7  Stage 1: Health—Linear Regression Association of Per Capita Health Expenditures with Penta3 Vaccination Rate Linear Regression Dependent variable: Penta3 Vaccination Rate Independent variables: OLS Yearly expenditure per capita in Health −0.063*** (0.0233) Yearly regional per capita Capital Expenditure −1,447.151 (1,315.3060) Percentage of rural population 8.677*** (1.8676) Ethnicity controls  Nuwer −64.966*** (6.4810)  Anyiwak −53.901*** (8.3259)  Gumuz −44.426*** (7.3255)  Konso −50.230** (24.1339)   Hist. Adv. −9.565*** (2.0970)  Somali −51.458*** (2.7817) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 66 Detailed Regression Results Table D.7  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Penta3 Vaccination Rate (continued)  Affar −47.295*** (3.2447)   Other Small −13.755*** (2.4127) Time controls   T zero −9.362*** (1.2211)   T one −3.007** (1.2684)   T two −6.915*** (0.8858)   T three Omitted Constant 91.697*** (3.1549) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.8  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Penta3 Vaccination Rate Quadratic Regression Dependent variable: Penta3 Vaccination Rate Independent variables: OLS Yearly expenditure per capita in Health −0.122*** (0.0395) Yearly expenditure per capita in Health Squared 0.0003* (0.0001) Yearly regional per capita Capital Expenditure −1431.498 (1,315.529) Percentage of rural population 8.152*** (1.8899) Ethnicity controls  Nuwer −64.123*** (6.4999)  Anyiwak −55.78*** (8.3911)  Gumuz −43.354*** (7.3510)  Konso −49.085** (24.1561)   Hist. Adv. −9.581*** (2.0974) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 67 Table D.8  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Penta3 Vaccination Rate (continued)  Somali −51.333*** (2.7833)  Affar −46.810*** (3.2567)   Other Small −13.546*** (2.4160) Time controls   T zero −9.963*** (1.2633)   T one −3.535*** (1.2994)   T two −7.166*** (0.8955)   T three Omitted Constant 93.646*** (3.3278) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.9  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expendi- tures with Penta3 Vaccination Rate Log Linear Regression Dependent variable: Penta3 Vaccination Rate Independent variables: OLS Log yearly expenditure per capita in Health −0.058** (0.0276) Log yearly regional per capita Capital Expenditure −0.003 (0.0262) Percentage of rural population 0.182*** (0.0512) Ethnicity controls  Nuwer −1.919*** (0.1910)  Anyiwak −1.018*** (0.2379)  Gumuz −0.701*** (0.2067)  Konso −0.626 (0.6551)   Hist. Adv. −0.141** (0.0595) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 68 Detailed Regression Results Table D.9  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Penta3 Vaccination Rate (continued)  Somali −1.048*** (0.0796)  Affar −1.026*** (0.0843)   Other Small −0.220*** (0.0697) Time controls   T zero −0.170*** (0.0404)   T one −0.095** (0.0428)   T two −0.131 *** (0.0296)   T three Omitted Constant 4.613*** (0.2042) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.10  Stage 1: Health—Linear Regression Association of Per Capita Health Expendi- tures with Antenatal Care Linear Regression Dependent variable: Antenatal Care Independent variables: OLS Yearly expenditure per capita in Health 0.026 (0.0313) Yearly regional per capita Capital Expenditure −249.548 (1,711.0580) Percentage of rural population −10.849*** (2.4200) Ethnicity controls  Nuwer −56.384*** (8.5451)  Anyiwak −44.982*** (10.2385)  Gumuz −59.973*** (9.1047)  Konso −42.666 (30.2839)   Hist. Adv. −13.034*** (2.5008) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 69 Table D.10  Stage 1: Health—Linear Regression Association of Per Capita Health Expenditures with Antenatal Care (continued)  Somali −41.722*** (3.4255)  Affar −59.354*** (4.0043)   Other Small −13.486*** (2.8967) Time controls   T zero −13.198*** (1.6319)   T one −6.886*** (1.6671)   T two −4.589*** (1.2176)   T three Omitted Constant 96.300*** (4.0269) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.11  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Antenatal Care Quadratic Regression Dependent variable: Antenatal Care Independent variables: OLS Yearly expenditure per capita in Health 0.086 (0.0532) Yearly expenditure per capita in Health Squared −0.0003 (0.0002) Yearly regional per capita Capital Expenditure −238.134 (1,711.9540) Percentage of rural population −10.305*** (2.4525) Ethnicity controls  Nuwer −57.340*** (8.5779)  Anyiwak −43.322*** (10.3145)  Gumuz −61.090*** (9.1462)  Konso −43.918 (30.3276) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 70 Detailed Regression Results Table D.11  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Antenatal Care (continued)   Hist. Adv. −13.080*** (2.5029)  Somali −41.852*** (3.4293)  Affar −59.895*** (4.0262)   Other Small −13.747*** (2.9050) Time controls   T zero −12.542*** (1.6965)   T one −6.326*** (1.7131)   T two −4.323*** (1.2316)   T three Omitted Constant 94.305*** (4.2712) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.12  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Antenatal Care Log Linear Regression Dependent variable: Antenatal Care Independent variables: OLS Log yearly expenditure per capita in Health 0.086** (0.0347) Log yearly regional per capita Capital Expenditure −0.021 (0.0330) Percentage of rural population −0.016 (0.0671) Ethnicity controls  Nuwer −2.710*** (0.2509)  Anyiwak −0.718*** (0.2905)  Gumuz −1.1760*** (0.2727)  Konso −0.635 (0.8460) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 71 Table D.12  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Antenatal Care (continued)   Hist. Adv. −0.178** (0.0683)  Somali −0.852*** (0.1017)  Affar −1.768*** (0.1280)   Other Small −0.230*** (0.0815) Time controls   T zero −0.197*** (0.0485)   T one −0.134*** (0.0485)   T two −0.107 *** (0.0339)   T three Omitted Constant 4.330*** (0.1601) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.13  Stage 1: Health—Linear Regression Association of Per Capita Health Expendi- tures with Contraceptive Acceptance Rate Linear Regression Dependent variable: Contraceptive Acceptance rate Independent variables: OLS Yearly expenditure per capita in Health −0.052* (0.0287) Yearly regional per capita Capital Expenditure −1,371.507 (1,583.1580) Percentage of rural population −11.900*** (2.2773) Ethnicity controls  Nuwer −64.900*** (7.8045)  Anyiwak −45.951*** (9.5779)  Gumuz −57.212*** (8.3646)  Konso −10.295 (28.6759) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 72 Detailed Regression Results Table D.13  Stage 1: Health—Linear Regression Association of Per Capita Health Expenditures with Contraceptive Acceptance Rate (continued)   Hist. Adv. −5.363** (2.3429)  Somali −61.431*** (3.302)  Affar −53.855*** (3.7462)   Other Small −12.126*** (2.7181) Time controls   T zero −14.651*** (1.4963)   T one −8.743*** (1.4947)   T two −3.260*** (1.0739)   T three Omitted Constant 83.648*** (3.7609) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.14  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Contraceptive Acceptance Rate Quadratic Regression Dependent variable: Contraceptive Acceptance rate Independent variables: OLS Yearly expenditure per capita in Health −0.012 (0.0493) Yearly expenditure per capita in Health Squared −0.00023 (0.0002) Yearly regional per capita Capital Expenditure −1,351.924 (1,583.894) Percentage of rural population −11.527*** (2.3092) Ethnicity controls  Nuwer 65.492*** (7.8317)  Anyiwak −45.000*** (9.6311)  Gumuz −57.931*** (8.4008) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 73 Table D.14  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Contraceptive Acceptance Rate (continued)  Konso −11.154 (28.7083)   Hist. Adv. −5.396** (2.3444)  Somali −61.430*** (3.3037)  Affar −54.207*** (3.7654)   Other Small −12.297*** (2.7251) Time controls   T zero −14.187*** (1.5666)   T one −8.350*** (1.5451)   T two −3.075*** (1.0896)   T three Omitted Constant 82.290*** (4.0021) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.15  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Contraceptive Acceptance Rate Log Linear Regression Dependent variable: Contraceptive Acceptance rate Independent variables: OLS Log yearly expenditure per capita in Health −0.006 (0.0410) Log yearly regional per capita Capital Expenditure −0.056 (0.0390) Percentage of rural population −0.136 (0.0796) Ethnicity controls  Nuwer −4.864*** (0.2766)  Anyiwak −1.567*** (0.3427)  Gumuz −1.552*** (0.3164) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 74 Detailed Regression Results Table D.15  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Contraceptive Acceptance Rate (continued)  Konso −0.232 (1.0094)   Hist. Adv. −0.116 (0.0805)  Somali −2.671*** (0.1225)  Affar −1.971*** (0.1507)   Other Small −0.309*** (0.0963) Time controls   T zero −0.308*** (0.0561)   T one −0.221*** (0.0559)   T two −0.0058** (0.0371)   T three Omitted Constant 4.744*** (0.3058) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D. 16  Stage 1: Health—Linear Regression Association of Per Capita Health Expendi- tures with Delivery by Skilled Birth Attendant Linear Regression Dependent variable: Delivery by Skilled Birth Attendant Independent variables: OLS Yearly expenditure per capita in Health 0.066*** (0.0287) Yearly regional per capita Capital Expenditure 328.600 (1,211.0980) Percentage of rural population −22.745*** (1.7421) Ethnicity controls  Nuwer −15.830*** (5.9881)  Anyiwak −19.264*** (7.4207)  Gumuz −17.340*** (6.2405) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 75 Table D. 16  Stage 1: Health—Linear Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant (continued)  Konso −3.805 (20.9835)   Hist. Adv. −4.326** (1.7805)  Somali −10.021*** (2.5271)  Affar −9.327*** (2.8013)   Other Small −5.143** (2.0568) Time controls   T zero −3.308*** (1.1900)   T one −4.626*** (1.2333)   T two 1.012 (0.9168)   T three Omitted Constant 40.980*** (2.8922) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.17  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant Quadratic Regression Dependent variable: Delivery by Skilled Birth Attendant Independent variables: OLS Yearly expenditure per capita in Health 0.068* (0.0389) Yearly expenditure per capita in Health Squared −0.00001 (0.0001) Yearly regional per capita Capital Expenditure 330.274 (1,211.889) Percentage of rural population −22.726*** (1.7685) Ethnicity controls  Nuwer −15.862*** (6.0100)  Anyiwak −19.204** (7.4858) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 76 Detailed Regression Results Table D.17  Stage 1: Health—Quadratic Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant (continued)  Gumuz −17.374*** (6.2652)  Konso −3.847 (21.0059)   Hist. Adv. −4.327** (1.7815)  Somali −10.024*** (2.5288)  Affar −9.344*** (2.8147)   Other Small −5.152** (2.0616) Time controls   T zero −3.285*** (1.2422)   T one −4.607*** (1.2694)   T two 1.021 (0.9270)   T three Omitted Constant 40.911*** (3.0919) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.18  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant Log Linear Regression Dependent variable: Delivery by Skilled Birth Attendant Independent variables: OLS Log yearly expenditure per capita in Health 0.267*** (0.0747) Log yearly regional per capita Capital Expenditure −0.071 (0.0698) Percentage of rural population −0.720*** (0.1456) Ethnicity controls  Nuwer −2.329*** (0.5238)  Anyiwak −1.914*** (0.6249) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 77 Table D.18  Stage 1: Health—Log Linear Regression Association of Per Capita Health Expenditures with Delivery by Skilled Birth Attendant (continued)  Gumuz −0.777** (0.5637)  Konso −0.372 (1.7985)   Hist. Adv. 0.033 (0.1450)  Somali −0.620*** (0.2224)  Affar −1.310*** (0.2378)   Other Small −1.106 (0.1726) Time controls   T zero −0.027 (0.0981)   T one −0.362*** (0.0976)   T two −0.042 (0.0700)   T three Omitted Constant 2.818*** (0.5497) Note: Probit regressions with robust standard errors in parentheses; Predicted probabilities with unconditional standard errors in parentheses; *, **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.19  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Cereal Yield Log Linear Regression Dependent variable: Cereal Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 0.128*** (0. 0489) Percentage of rural population −0.414* (0.2205) Rainfall controls   Average rainfall 0.000 (0.0008)   Deviation from average rainfall 0.001** (0.0007)   Deviation from average rainfall (lagged) 0.002*** (0.0007) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 78 Detailed Regression Results Table D.19  Stage 1: Agriculture—Log Linear Regression Association of per Capita Agriculture ­ expenditures with Cereal Yield (continued) Regional dummies  Tigray 4.576 (9.0103)  Afar 3.544 (8.5370)  Amhara 3.522 (8.9863)  Oromia 3.951 (8.9503)  Somali 11.307 (9.2046)   Beneshangul Gemuz 17.258 (18.0467)  SNNP 3.502 (8.9883)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall 0.000 (0.0009)   Afar × average rainfall Omitted   Amhara × average rainfall 0.000 (0.0009)   Oromia × average rainfall 0.000 (0.0008)   Somali × average rainfall −0.010*** (0.0027)   Beneshangul Gemuz × average rainfall −0.011 (0.0128)   SNNP × average rainfall 0.000 (0.0008)   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate −0.020 (0.0182)   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.003 (0.0058)   Oromia × Poverty rate 0.002 (0.0063)   Somali × Poverty rate −0.078*** (0.0200) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 79 Table D.19  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Pulses Yield (continued)   Beneshangul Gemuz × Poverty rate −0.036 (0.0438)   SNNP × Poverty rate 0.000 (0.0025)   Gambella × Poverty rate 0.102 (0.2498) Constant −1.510 (8.9315) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. Dire Dawa and Hara omitted due to small scale of agriculture in these regions. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Table D.20  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Pulses Yield Log Linear Regression Dependent variable: Pulses Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture −0.020 (0.0645) Percentage of rural population 0.105 (0.2456) Rainfall controls   Average rainfall 0.001 (0.0014)   Deviation from average rainfall 0.001 (0.0010)   Deviation from average rainfall (lagged) 0.001 (0.0010) Regional dummies  Tigray 3.165 (2.126)  Afar 0.918 (1.1132)  Amhara 2.115 (1.9126)  Oromia 3.184* (1.8845)  Somali 3.569 (3.0569)   Beneshangul Gemuz 39.151* (20.4730) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 80 Detailed Regression Results Table D.20  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Pulses Yield (continued)  SNNP 2.462 (1.9117)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall −0.001 (0.0015)   Afar × average rainfall Omitted   Amhara × average rainfall 0.000 (0.0014)   Oromia × average rainfall −0.001 (0.0014)   Somali × average rainfall −0.002 (0.0036)   Beneshangul Gemuz × average rainfall −0.030* (0.0166)   SNNP × average rainfall −0.001 (0.0014)   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate −0.012 (0.0196)   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.008 (0.0063)   Oromia × Poverty rate −0.010 (0.0070)   Somali × Poverty rate −0.027 (0.0265)   Beneshangul Gemuz × Poverty rate −0.126** (0.0563)   SNNP × Poverty rate 0.002 (0.0027)   Gambella × Poverty rate Omitted Constant −0.213 (1.8785) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP =Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 81 Table D.21  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Root Crop Yield Log Linear Regression Dependent variable: Root Crop Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 0.320 (0.2266) Percentage of rural population −0.082 (0.7218) Rainfall controls   Average rainfall 0.000 (0.0032)   Deviation from average rainfall 0.002 (0.0039)   Deviation from average rainfall (lagged) 0.010 (0.0039) Regional dummies  Tigray 0.739 (5.2598)  Afar Omitted  Amhara 0.314 (4.4270)  Oromia 0.579 (4.4138)  Somali −1.041 (5.5139)   Beneshangul Gemuz 4.842 (43.2237)  SNNP 0.626 (4.4447)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall 0.000 (0.0036)   Afar × average rainfall Omitted   Amhara × average rainfall 0.000 (0.0033)   Oromia × average rainfall 0.000 (0.0033)   Somali × average rainfall 0.003 (0.0069)   Beneshangul Gemuz × average rainfall −0.003 (0.0351) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 82 Detailed Regression Results Table D.21  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Root Crop Yield (continued)   SNNP × average rainfall 0.000 (0.0033)   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate 0.012 (0.0618)   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.008 (0.0170)   Oromia × Poverty rate −0.009 (0.0211)   Somali × Poverty rate Omitted   Beneshangul Gemuz × Poverty rate −0.006 (0.1203)   SNNP × Poverty rate −0.016** (0.0074)   Gambella × Poverty rate Omitted Constant 2.431 (4.5185) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 83 Table D.22  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Vegetable Yield Log Linear Regression Dependent variable: Vegetable Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 0.582*** (0.1330) Percentage of rural population −0.110 (0.5548) Rainfall controls   Average rainfall 0.001 (0.0021)   Deviation from average rainfall 0.001 (0.0020)   Deviation from average rainfall (lagged) −0.003 (0.0021) Regional dummies  Tigray 32.703 (22.4893)  Afar 34.322 (22.5041)  Amhara 32.568 (22.4326)  Oromia 33.773 (22.3465)  Somali 10.227 (22.9840)   Beneshangul Gemuz 35.983 (43.4964)  SNNP 31.791 (22.4365)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall −0.002 (0.0024)   Afar × average rainfall −0.005 (0.0042)   Amhara × average rainfall −0.002 (0.0022)   Oromia × average rainfall −0.002 (0.0021)   Somali × average rainfall 0.031*** (0.0067) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 84 Detailed Regression Results Table D.22  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Vegetable Yield (continued)   Beneshangul Gemuz × average rainfall −0.005 (0.0306)   SNNP × average rainfall 0.001 (0.0021)   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate 0.004 (0.0618)   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.003 (0.0145)   Oromia × Poverty rate −0.034** (0.0159)   Somali × Poverty rate 0.185*** (0.0490)   Beneshangul Gemuz × Poverty rate −0.036 (0.1049)   SNNP × Poverty rate −0.004 (0.0064)   Gambella × Poverty rate 0.919 (0.6237) Constant −29.575 (22.3062) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 85 Table D.23  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Oil Seeds Yield Log Linear Regression Dependent variable: Oil seeds Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture −0.118 (0.1899) Percentage of rural population −0.229 (0.6220) Rainfall controls   Average rainfall −0.008*** (0.0024)   Deviation from average rainfall −0.001 (0.0034)   Deviation from average rainfall (lagged) −0.001 (0.0036) Regional dummies  Tigray −43.523* (25.2816)  Afar −38.156 (24.0540)  Amhara −42.608* (25.2609)  Oromia −42.904* (25.1993)  Somali −41.315* (24.3306)   Beneshangul Gemuz −52.981 (44.3822)  SNNP −44.061* (25.2585)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfal 0.008*** (0.0027)   Afar × average rainfall Omitted   Amhara × average rainfall 0.008*** (0.0025)   Oromia × average rainfall 0.008*** (0.0025)   Somali × average rainfall Omitted   Beneshangul Gemuz × average rainfall 0.017*** (0.0025) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 86 Detailed Regression Results Table D.23  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Oil Seeds Yield (continued)   SNNP × average rainfall 0.009*** (0.0025)   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate 0.017 (0.0450)   Afar × Poverty rate Omitted   Amhara × Poverty rate 0.001 (0.0144)   Oromia × Poverty rate 0.007 (0.0186)   Somali × Poverty rate Omitted   Beneshangul Gemuz × Poverty rate 0.012 (0.1041)   SNNP × Poverty rate 0.000 (0.7084)   Gambella × Poverty rate −1.044* (0.7084) Constant 45.731* (25.1903) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 87 Table D.24  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Enset Yield Log Linear Regression Dependent variable: Enset Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 2.397*** (0. 5766) Percentage of rural population −1.091 (1.7449) Rainfall controls   Average rainfall 0.000 (0.0011)   Deviation from average rainfall 0.010 (0.0168)   Deviation from average rainfall (lagged) 0.068*** (0.0157) Regional dummies  Tigray Omitted  Afar Omitted  Amhara Omitted  Oromia 8.654*** (3.0691)  Somali Omitted   Beneshangul Gemuz Omitted  SNNP 1.357 (1.4055)  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall Omitted   Afar × average rainfall Omitted   Amhara × average rainfall Omitted   Oromia × average rainfall −0.005*** (0.0016)   Somali × average rainfall Omitted   Beneshangul Gemuz × average rainfall Omitted   SNNP × average rainfall Omitted   Gambella × average rainfall Omitted table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 88 Detailed Regression Results Table D.24  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Enset Yield (continued) Regional dummies × Poverty rate   Tigray × Poverty rate Omitted   Afar × Poverty rate Omitted   Amhara × Poverty rate Omitted   Oromia × Poverty rate −0.034 (0.0527)   Somali × Poverty rate Omitted   Beneshangul Gemuz × Poverty rate Omitted   SNNP × Poverty rate 0.020*** (4.3094)   Gambella × Poverty rate −1.044* (0.7084) Constant −13.452*** (4.3094) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; *, **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 89 Table D.25  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Fruit Yield Log Linear Regression Dependent variable: Fruit Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 1.790*** (0.2669) Percentage of rural population −0.794 (0.9149) Rainfall controls   Average rainfall 0.000 (0.0006)   Deviation from average rainfall 0.002 (0.0048)   Deviation from average rainfall (lagged) 0.028*** (0.0050) Regional dummies  Tigray −2.459 (3.3924)  Afar 4.850* (2.8878)  Amhara −0.309 (0.9558)  Oromia 2.933** (1.2536)  Somali −5.032 (6.4944)   Beneshangul Gemuz −108.598** (55.213)  SNNP Omitted  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall 0.001 (0.0019)   Afar × average rainfall −0.009 (0.0059)   Amhara × average rainfall 0.001 (0.0013)   Oromia × average rainfall −0.002** (0.0008)   Somali × average rainfall 0.005 (0.0090) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 90 Detailed Regression Results Table D.25  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Fruit Yield (continued)   Beneshangul Gemuz × average rainfall 0.087* (0.0448)   SNNP × average rainfall Omitted   Gambella × average rainfall Omitted Regional dummies × Poverty rate   Tigray × Poverty rate 0.002 (0.0665)   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.023 (0.0213)   Oromia × Poverty rate −0.019 (0.0267)   Somali × Poverty rate 0.113 (0.0727)   Beneshangul Gemuz × Poverty rate 0.288* (0.1537)   SNNP × Poverty rate 0.014 (0.0093)   Gambella × Poverty rate Omitted Constant −4.019*** (1.4968) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 91 Table D.26  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agricul- ture ­Expenditures with Coffee Yield Log Linear Regression Dependent variable: Coffee Yield Independent variables: OLS Log yearly expenditure per capita in Agriculture 1.267*** (0. 2342) Percentage of rural population −0.777 (0.7845) Rainfall controls   Average rainfall 0.000 (0.0005)   Deviation from average rainfall −0.001 (0.0048)   Deviation from average rainfall (lagged) 0.029*** (0.0049) Regional dummies  Tigray Omitted  Afar Omitted  Amhara −0.632 (0.8110)  Oromia 2.014* (1.0696)  Somali 1.43** (0.6380)   Beneshangul Gemuz −18.760 (46.6977)  SNNP Omitted  Gambella Omitted Regional dummies × average rainfall   Tigray × average rainfall Omitted   Afar × average rainfall Omitted   Amhara × average rainfall 0.000 (0.0011)   Oromia × average rainfall −0.001*** (0.0007)   Somali × average rainfall Omitted   Beneshangul Gemuz × average rainfall 0.014 (0.0379)   SNNP × average rainfall Omitted   Gambella × average rainfall Omitted table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 92 Detailed Regression Results Table D.26  Stage 1: Agriculture—Log Linear Regression Association of Per Capita Agriculture ­Expenditures with Coffee Yield (continued) Regional dummies × Poverty rate   Tigray × Poverty rate Omitted   Afar × Poverty rate Omitted   Amhara × Poverty rate −0.012 (0.0180)   Oromia × Poverty rate 0.004 (0.0228)   Somali × Poverty rate Omitted   Beneshangul Gemuz × Poverty rate 0.068 (0.1300)   SNNP × Poverty rate 0.007 (0.0082)   Gambella × Poverty rate Omitted Constant −4.095*** (1.3591) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares; SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 93 Table D.27  Stage 3: Agriculture—Predicted Probabilities for Farmers' Use of Any Improved ­ Technique, 2011 Dependent variable: Any Improved Technique, 2011 Independent variables: Probit Predicted Probability Yearly expenditure per capita in Agriculture 0.0033 (0.00029) Poverty rate 0.0075** (0.00295) Rainfall controls   Average rainfall −0.00078*** (0.00007)   Deviation from average rainfall 2011 −0.00306*** (0.00046)   Deviation from average rainfall 2010 0.0090*** (0.00043) Regional dummies  Tigray omitted  Afar −5.1478*** (0.58951)  Amhara −2.8538*** (0.13734)  Oromia −0.02614 (0.14042)  Somali −1.1050 (2.79749)   Beneshangul Gemuz −28.8422*** (5.38058)  SNNP −0.9599 (0.14272)  Gambella 40.0622 Regional dummies × average rainfall   Tigray × average rainfall omitted   Afar × average rainfall omitted   Amhara × average rainfall 0.0031*** (0.00007)   Oromia × average rainfall −0.00001 (0.00007)   Somali × average rainfall −0.0025 (0.00408)   Beneshangul Gemuz × average rainfall 0.0234*** (0.00434) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 94 Detailed Regression Results Table D.27  Stage 3: Agriculture—Predicted Probabilities for Farmers' Use of Any Improved ­ Technique, 2011 (continued)   SNNP × average rainfall 0.0005*** (0.00008)   Gambella × average rainfall 0.0022*** 0.00066 Regional dummies × Poverty rate   Tigray × Poverty rate omitted   Afar × Poverty rate 0.1298*** (0.01668)   Amhara × Poverty rate −0.0027 (0.00304)   Oromia × Poverty rate −0.0207*** (0.00307)   Somali × Poverty rate −0.0016 (0.02692)   Beneshangul Gemuz × Poverty rate 0.0446*** (0.01563)   SNNP × Poverty rate −0.0010 (0.00298)   Gambella × Poverty rate −1.4040 Quintile  Smallest/Poorest omitted 0.0003*** (0.00002)  Second/Poorer 0.0445*** 0.0003*** (0.01386) (0.00003)  Middle/Middle 0.3727*** 0.0005*** (0.01282) (0.00004)  Fourth/Richer 0.5961*** 0.0006*** (0.01265) (0.00005)  Largest/Richest 0.8211*** 0.0008*** (0.01280) (0.00007) Constant −1.7225*** (0.15902) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 95 Table D.28  Gender Issues: Linear Regression Association of Per Capita Education Expendi- tures with Net Enrollment Rate for Males Linear Regression Dependent variable: Net Enrollment Rate—Male Independent variables: OLS Yearly expenditure per capita in Education 0.0438*** (0.0070) Percentage of rural population −7.011 (5.71317) Ethnicity controls  Nuwer 6.7239 (6.35115)  Anyiwak −17.9619 (13.65227)  Gumuz 2.1839 (7.19484)  Konso −34.8711* (19.43414)   Other Small 0.7764 (3.70825) Constant 84.8515*** (5.20259) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 96 Detailed Regression Results Table D.29   Gender Issues: Quadratic Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Males Quadratic Regression Dependent variable: Net Enrollment Rate—Male Independent variables: OLS Yearly expenditure per capita in Education 0.095*** (0.0180) Yearly expenditure per capita in Education Squared −0.0001*** (0.00005) Percentage of rural population −6.474 (5.6910) Ethnicity controls  Nuwer 5.652 (6.4082)  Anyiwak −11.147 (12.9711)  Gumuz 0.707 (7.0657)  Konso −34.690* (19.6667)   Other Small 0.632 (3.6941) Constant 81.002*** (5.2059) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.30   Gender Issues: Log Linear Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Males Log Linear Regression Dependent variable: Log Net Enrollment Rate—Male Independent variables: OLS Log of yearly expenditure per capita in Education 0.115*** (0.0116) Constant 3.752*** (0.0627) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 97 Table D.31  Gender Issues: Linear Regression Association of Per Capita Education Expendi- tures with Net Enrollment Rate for Females Dependent variable: Net Enrollment Rate—Female Independent variables: OLS Yearly expenditure per capita in Education 0.0476*** (0.00756) Percentage of rural population −12.2364** (5.53352) Ethnicity controls  Nuwer −10.2088*** (3.26873)  Anyiwak −21.4472*** (12.98825)  Gumuz −22.8067*** (8.07086)  Konso −37.6488*** (13.07193)   Other Small −8.2162** (3.91541) Constant 87.8880*** (5.09835) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.32  Gender Issues: Quadratic Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Females Quadratic Regression Dependent variable: Net Enrollment Rate—Female Independent variables: OLS Yearly expenditure per capita in Education 0.095*** (0.0180) Yearly expenditure per capita in Education Squared −0.0001*** (0.00005) Percentage of rural population −11.682** (5.5100) Ethnicity controls  Nuwer −11.325*** (3.3056)  Anyiwak −14.655 (12.271)  Gumuz −24.307*** (7.9683)  Konso −37.462*** (13.3247) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 98 Detailed Regression Results Table D.32   Gender Issues: Quadratic Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Females (continued)   Other Small −8.355** (3.8904) Constant 81.002*** (5.2059) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.33  Gender Issues: Log Linear Regression Association of Per Capita Education Expenditures with Net Enrollment Rate for Females Log Linear Regression Dependent variable: Log Net Enrollment Rate—Female Independent variables: OLS Log of yearly expenditure per capita in Education 0.158*** (0.0137) Constant 3.505*** (0.0746) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.34  Gender Issues: Linear Regression Association of Per Capita Education Expendi- tures with Net Intake Rate for Males Dependent variable: Net Intake Rate—Male Independent variables: OLS Yearly expenditure per capita in Education 0.1670*** (0.01702) Percentage of rural population 31.6473*** (5.71317) Ethnicity controls  Nuwer −74.3077*** (5.43405)  Anyiwak −95.4275*** (20.37484)  Gumuz −54.9533*** (11.42498)  Konso −27.4514 (22.54912)   Other Small −3.7305 (4.52154) Constant 43.4391*** (6.55315) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 99 Table D.35  Gender Issues: Quadratic Regression Association of Per Capita Education Expenditures with Net Intake Rate for Males Quadratic Regression Dependent variable: Net Intake Rate—Male Independent variables: OLS Yearly expenditure per capita in Education 0.418*** (0.0275 Yearly expenditure per capita in Education Squared −0.0006*** (0.00007) Percentage of rural population 34.070*** (6.8756) Ethnicity controls  Nuwer −81.015*** (4.4688)  Anyiwak −64.507*** (11.3181)  Gumuz −61.676*** (11.3097)  Konso −29.614 (22.6827)   Other Small −4.308 (4.5489) Constant 24.648*** (6.1562) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.36  Gender Issues: Log Linear Regression Association of Per Capita Education Expenditures with Net Intake Rate for Males Log Linear Regression Dependent variable: Log Net Intake Rate—Male Independent variables: OLS Log of yearly expenditure per capita in Education 0.390*** (0.0240) Constant 2.420*** (0.1202) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 100 Detailed Regression Results Table D.37   Gender Issues: Linear Regression Association of Per Capita Education Expendi- tures with Net Intake Rate for Females Dependent variable: Net Intake Rate—Female Independent variables: OLS Yearly expenditure per capita in Education 0.16698*** (0.01642) Percentage of rural population 0.0476*** (6.51252) Ethnicity controls  Nuwer −71.7297*** (5.09329)  Anyiwak −93.7788*** (20.3585)  Gumuz −61.4025*** (9.81233)  Konso −18.8002 (18.30648)   Other Small −10.58377** (4.31694) Constant 44.9443*** (6.10655) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.38   Gender Issues: Quadratic Regression Association of Per Capita Education Expenditures with Net Intake Rate for Females Quadratic Regression Dependent variable: Net Intake Rate—Female Independent variables: OLS Yearly expenditure per capita in Education 0.404*** (0.0273) Yearly expenditure per capita in Education Squared −0.0006*** (.00007) Percentage of rural population 28.551*** (6.2956) Ethnicity controls  Nuwer −78.070*** (4.2173)  Anyiwak −64.674*** (11.6827)  Gumuz −67.738*** (9.8523)  Konso −20.859 (18.5272) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 101 Table D.38   Gender Issues: Quadratic Regression Association of Per Capita Education Expendi- tures with Net Intake Rate for Females (continued)   Other Small −11.122** (4.3080) Constant 27.221*** (5.7160) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.39  Gender Issues: Log Linear Regression Association of Per Capita Education Expenditures with Net Intake Rate for Females Log Linear Regression Dependent variable: Log Net Intake Rate—Female Independent variables: OLS Log of yearly expenditure per capita in Education 0.430*** (.0269) Constant 2.183*** (0.1354) Note: Cross-time pooled regressions with robust standard errors; standard errors in parentheses; * , **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. OLS = ordinary least squares. Table D.40  Gender Issues: Predicted Probabilities for Farmers' Use of Any Improved Technique by Gender, 2011 Dependent variable: Any Improved Technique by gender, 2011 Independent variables: Probit Predicted Probability Yearly expenditure per capita in Agriculture 0.0019 (0.00032) Poverty rate 0.0089*** (0.00285) Rainfall controls   Average rainfall −0.00060*** (0.00007)   Deviation from average rainfall 2011 −0.00379*** (0.00046)   Deviation from average rainfall 2010 0.0090*** (0.00043) Regional dummies  Tigray omitted  Afar −5.9148*** (0.57195)  Amhara −2.8286*** (0.13357)  Oromia 0.32604 (0.13683) table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 102 Detailed Regression Results Table D.40  Gender Issues: Predicted Probabilities for Farmers' Use of Any Improved Technique by Gender, 2011 (continued)  Somali −1.5086 (2.71172)   Beneshangul Gemuz −33.9580*** (5.17708)  SNNP −0.0528 (0.14926)  Gambella 43.2032 Regional dummies × average rainfall   Tigray × average rainfall omitted   Afar × average rainfall omitted   Amhara × average rainfall 0.0031*** (0.00007)   Oromia × average rainfall −0.00036 (0.00007)   Somali × average rainfall −0.0058 (0.00395)   Beneshangul Gemuz × average rainfall 0.0277*** (0.00418)   SNNP × average rainfall 0.0028*** (0.00008)   Gambella × average rainfall 0.0013*** 0.00081 Regional dummies × Poverty rate   Tigray × Poverty rate omitted   Afar × Poverty rate 0.1515*** (0.01644)   Amhara × Poverty rate −0.0071 (0.00295)   Oromia × Poverty rate −0.0207*** (0.00297)   Somali × Poverty rate −0.0369 (0.02617)   Beneshangul Gemuz × Poverty rate 0.0528*** (0.01504)   SNNP × Poverty rate −0.0040 (0.00298)   Gambella × Poverty rate −1.4770 table continues next page Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4  Detailed Regression Results 103 Table D.40   Gender Issues: Predicted Probabilities for Farmers' Use of Any Improved Technique by Gender, 2011 (continued) Sex  Male 0.0002*** 0.00003*** (0.00000)  Female −0.0001*** −0.00002*** (0.00001) (0.00000) Constant −1.2469*** (0.15362) Note: Probit regressions with robust standard errors in parentheses; Predicted probabilities with unconditional standard errors in parentheses. *, **, *** = coefficients significant at the 10 percent, 5 percent and 1 percent levels. SNNP = Southern Nations, Nationalities, and Peoples (Region). Dire Dawa and Hara omitted due to small scale of agriculture in these regions. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Bibliography Adegehe, Asnake. 2009. “Federalism and Ethnic Conflict in Ethiopia: A Comparative Study of the Somali and Benishangul-Gumuz Regions.” Doctoral thesis, Department of Political Science, Faculty of Social and Behavioural Sciences, Leiden University. Assefa, Fisshea. 2010. Federalism and the Accommodation of Diversity in Ethiopia. 3rd ed. Addis Ababa: Eclipse Press. Barankay, I., and B. Lockwood. 2007. “Decentralization and the Productive Efficiency of Government: Evidence from Swiss Cantons.” Journal of Public Economics 91 (5–6): 1197–218. Batina, R. G., and T. Ihori. 2005. Public Goods: Theories and Evidence. New York: Springer. Besley, T., and S. Coate. 2003. “Centralized versus Decentralized Provision of Local Public Goods: A Political Economy Approach.” Journal of Public Economics 87 (12): 2611–37. Bevan, Phillipa, Catherine Dom, and Alula Pankhurst. 2010. “Long-term Perspectives of Development Impacts in Rural Ethiopia: Wide Stage One Final Report.” Mokoro Limited, Oxford. Boone, P., and J. P. Faguet. 1998. “Multilateral Aid, Politics, and Poverty: Past Failures and Future Challenges.” Chapter 2 in The Global Crisis in Foreign Aid, edited by R. Grant and J. Nijman, 11–26. Syracuse: Syracuse University Press. Cassette, A., and S. Paty. 2010. “Fiscal Decentralization and the Size of Government: A European Country Empirical Analysis.” Public Choice 143 (1–2): 173–89. Ceballos, M., and D. Hoyos. 2004. “Tendencias del comportamiento electoraly descentral- ización en los municipios de Colombia, 1988–2000.” Crisis States Programme Working Paper No. 57, London School of Economics. Clapham, Christopher. 1994. “Ethnicity and the National Question in Ethiopia.” In Conflict and Peace in the Horn of Africa: Federalism and Its Alternatives, edited by Peter Woodward and Murray Forsyth, 27–40. Aldershot: Dartmouth Publishing. Clark, D. 2009. “The Performance and Competitive Effects of School Autonomy.” Journal of Political Economy 117 (4): 745–82. CSA (Central Statistical Agency of Ethiopia). “Voice of the Central Statistical Agency.” Various issues. Addis Ababa. Devarajan, S., S. Khemani, and S. Shah. 2009. “The Politics of Partial Decentralization.” In Does Decentralization Enhance Service Delivery and Poverty Reduction? edited by E. Ahmad and G. Brosio. Cheltenham: Edward Elgar. Dowding, K., and P. John. 1994. “Tiebout: A Survey of the Empirical Literature.” Urban Studies 31 (4/5): 767–97. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4   105 106 Bibliography Dom, Catherine, and Stephen Lister, with Manos Antoninis. 2010. “An Analysis of Decentralization in Ethiopia.” Mokoro Limited, Oxford. Eldon, Jack, Abebe, Cunningham, and Bibby. 2011. “Options for Support to Service Delivery in Ethiopia’s Developing Regional States.” Escaleras, M., and C. A. Register. 2012. “Fiscal Decentralization and Natural Hazard Risks.” Public Choice 151 (1): 165–83. Faguet, J. P. 2000. “Decentralization and Local Government Performance: Improving Public Service Provision in Bolivia.” Revista de Economía del Rosario 3 (1): 127–76. ———. 2004. “Why So Much Centralization? A Model of Primitive Centripetal Accumulation.” LSE-STICERD Development Economics Discussion Paper No. 43, London School of Economics. ———. 2012. Decentralization and Popular Democracy: Governance from Below in Bolivia. Ann Arbor: University of Michigan. ———. 2014. “Decentralization and Governance.” World Development 53: 2–13. Faguet, J. P., and Z. Ali. 2009. “Making Reform Work: Institutions, Dispositions, and the Improving Health of Bangladesh.” World Development 37 (1): 208–18. Faguet, J. P., and F. Sánchez. 2013. “Decentralization and Access to Social Services in Colombia.” Public Choice. doi: 10.1007/s11127-013-0077-7. Faguet, J. P., and F. B. Wietzke. 2006. “Social Funds and Decentralization: Optimal Institutional Design.” Public Administration and Development 26 (4): 303–15. Galiani, S., P. Gertler, and D. Schargrodsky. 2008. “School Decentralization: Helping the Good Get Better, but Leaving the Poor Behind.” Journal of Public Economics 92 (10– 11): 2106–20. Garcia, Marito, and Andrew Sunil Rajkumar. 2008. “Achieving Better Service Delivery through Decentralization in Ethiopia.” African Human Development Series Working Paper No. 131, World Bank, Washington, DC. Gonçalves, S. 2013. “The Effects of Participatory Budgeting on Municipal Expenditures and Infant Mortality in Brazil.” World Development. http://dx.doi.org/10.1016/j. worlddev.2013.01.009. Granados, C., and F. Sánchez. 2014. “Water Reforms, Decentralization, and Child Mortality in Colombia, 1990–2005.” World Development 53 (C): 68–79. IPE Global. 2010. “Evaluation and Design of the Social Accountability Component of the Protection of Basic Services Project.” New Delhi. Kamurase, Alex, and Aly Salman Alibhai. 2014. “Social Accountability for Ethiopia PBS.” Presentation for session on Tricks and Tools of Task Team Leaders, Sustainable Development Network Forum, World Bank, Washington, DC, February. Krishan, Pramila, and Manasa Patnam. 2013. “Neighbours and Extension Agents in Ethiopia: Who Matters More?” American Journal of Agricultural Economics 96 (1): 308–27. Manor, J. 1999. The Political Economy of Democratic Decentralization. Washington, DC: World Bank. Meles, Zenawi. 1997. “Premier’s Speech at Butare National University in Rwanda.” Ethiopian Herald, December 13. Mulugeta, Allehone. 2002. “Issues of Security and Conflict, in the Ethiopian Frontiers: Notes on State Policies and Strategies.” Report of Ethiopia National Workshop, Conflict in the Horn: Prevention and Resolution, Organization for Social Science Research in Eastern and Southern Africa (OSSREA), Addis Ababa. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Bibliography 107 Oates, W. 1972. Fiscal Federalism. New York: Harcourt Brace. ODI (Overseas Development Institute). 2010. “Millennium Development Goals Report Card: Measuring Progress Across Countries.” London. Ostrom, E., L. Schroeder, and S. Wynne. 1993. Institutional Incentives and Sustainable Development: Infrastructure Policies in Perspective. Boulder: Westview. Ostrom, E., and G. P. Whitaker. 1973. “Does Local Community Control of Police Make a Difference? Some Preliminary Findings.” American Journal of Political Science 17 (1): 48–76. Piriou-Sall, S. 1998. “Decentralisation and Rural Development: A Review of Evidence.” Washington, DC. Processed. Prohl, S., and F. Schneider. 2009. “Does Decentralization Reduce Government Size? A Quantitative Study of the Decentralization Hypothesis.” Public Finance Review 37 (6): 639–64. Prud’homme, R. 1995. “On the Dangers of Decentralisation.” World Bank Research Observer 10 (2): 210–26. Putnam, R. D. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University. Ragasa, Catherine, Guush Berhane, Fanayae Tadesse, and Alemayehu Seyoum Taffesse. 2012. “Gender Differences in Access to Extension Services and Agriculture Productivity.” ESSP Working Paper No. 49, Ethiopia Strategy Support Program (ESSP), Ethiopia Development Research Institute (EDRI), and International Food Policy Research Institute (IFPRI), Washington, DC. Randolph, Robert C., and Buli Edjeta. 2011. “Study on Strengthening Grievance Redress Mechanisms for the Protection of the Basic Services (PBS) Program in Ethiopia.” Addis Ababa. Rondinelli, D. A., G. S. Cheema, and J. Nellis. 1983. “Decentralisation in Developing Countries: A Review of Recent Experience.” Staff Working Paper No. 581, World Bank, Washington, DC. Samoff, J. 1990. “Decentralisation: The Politics of Interventionism.” Development and Change 21 (3): 513–30. Selam Development Consultants. 2013. “Financial Transparency and Accountability Implementation Assessment Report.” Addis Ababa. Smoke, P. 2001. “Fiscal Decentralisation in Developing Countries: A Review of Current Concepts and Practice.” Democracy, Governance, and Human Rights Programme Paper No. 2, United Nations Research Institute for Social Development, Geneva. Solomon, Negussie. 2008. “Fiscal Federalism in the Ethiopian Ethnic-based Federal System.” Forum of Federations, Addis Ababa. Tiebout, C. M. 1956. “A Pure Theory of Local Expenditures.” Journal of Political Economy 64 (5): 416–24. Treisman, D. 2007. The Architecture of Government: Rethinking Political Decentralization. New York: Cambridge University. Vaillancourt, Francois. 2013. “Sharing Central Revenues with Regions: An Examination of Principles and Practices in the Context of the PBS (Ethiopia) Program.” World Bank, Addis Ababa. Vaughan, Sarah. 2012. “Understanding Ethiopia’s Transformation: Vision and Strategies for Development.” World Bank, Addis Ababa. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 108 Bibliography Wang, Huihui, G. N. V. Ramana, Jie Huang, and Bekele Chaka. Forthcoming. “Ethiopia Universal Health Care Case Study.” World Bank, Washington, DC. Weldemariam, Alemayehu F. 2011. “Greater Ethiopia: Evolution of a Pluralist Politico- Legal System in a Pluralist Polity.” http://dx.doi.org/10.2139/ssrn.1909803. Woldehanna, Tassew, Eyasu Tsehaye, and Ruth Hill. Forthcoming. “The Distributional Impact of Fiscal Policy in Ethiopia.” Contribution to a Global Study of Fiscal Incidence led by Nora Lustig, World Bank, Washington, DC. World Bank. 2003. World Development Report 2004: Making Services Work for Poor People. Washington, DC: World Bank. ———. 2010. “Financial Transparency and Accountability Initiative in Ethiopia: Progress and Way Forward.” Addis Ababa. ———. 2011a. “An Assessment of Public Finance Management Systems in Woreda Governments.” Addis Ababa. ———. 2011b. Citizens and Service: Assessing Use of Social Accountability Approaches in the Human Development Sectors. Directions in Development. Washington, DC: World Bank. ———. 2012a. “Project Appraisal Document for Ethiopia Promoting Basic Services Project Phase III.” Report No. 69689-ET, World Bank, Washington, DC. ———. 2012b. “Serving the Ethiopian People: The Origins and Evolution of the PBS Programme.” Working Paper No. 74851, World Bank, Washington, DC. ———. 2013. “Understanding Incentives and Strengthening Services: The Political Economy of Decentralized Delivery in Ethiopia.” World Bank, Washington, DC. ———. 2014. “Improving Basic Services for the Bottom Forty Percent: Results of the Poverty and Social Impact Assessment of Decentralized Basic Service Delivery in Ethiopia.” Report No: 84215-ET, World Bank, Washington, DC. Yang, Jing. 2012. “Interpreting Coefficients in Regression with Log-Transformed Variables.” StatNews No. 83 (June). Cornell Statistical Consulting Unit, Ithaca. Zewde, Bahru. 1991. A History of Modern Ethiopia 1855–1974. Addis Ababa: Addis Ababa University Press. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 Environmental Benefits Statement The World Bank is committed to reducing its environmental footprint. In support of this commitment, the Publishing and Knowledge Division leverages electronic publishing options and print-on-demand technology, which is located in regional hubs worldwide. Together, these initiatives enable print runs to be lowered and shipping distances decreased, resulting in reduced paper consumption, chemical use, greenhouse gas emissions, and waste. The Publishing and Knowledge Division follows the recommended standards for paper use set by the Green Press Initiative. Whenever possible, books are printed on 50 percent to 100 percent postconsumer recycled paper, and at least 50 percent of the fiber in our book paper is either unbleached or bleached using Totally Chlorine Free (TCF), Processed Chlorine Free (PCF), or Enhanced Elemental Chlorine Free (EECF) processes. More information about the Bank’s environmental philosophy can be found at http://crinfo.worldbank.org/wbcrinfo/node/4. Improving Basic Services for the Bottom Forty Percent  •  http://dx.doi.org/10.1596/978-1-4648-0331-4 E thiopia’s model for delivering basic services is succeeding. Accountability for delivering basic services can take a long, indirect route whereby citizens influence service providers through the government or a short, more direct route between service providers and citizens. The indirect route of accountability works well precisely in Ethiopia because of decentralization. Service providers are strictly accountable to local governments for producing results, but, in turn, local authorities are held accountable by regional and federal governments. The direct route of accountability has been reinforced by measures to strengthen financial transparency and accountability, social accountability, and impartial procedures to redress grievances. Spending at the woreda (district) level in Ethiopia has been a very effective strategy for attaining the country’s Millennium Development Goals. Woreda health and education pays for health exten- sion workers and teachers. Bottom Forty Percent finds that woreda-level spending in health and education has been effective. Use of health services has increased, especially among the poorest quintiles. Woreda-level spending on agricultural extension workers is associated with higher yields for major crops. Spending on these workers increases the probability that farmers will use improved farming techniques. Spending efficiency is improved through better capacity, more transparency, and greater accountability to citizens. The World Bank Studies series is available for free download online through the Open Knowledge Repository (https://openknowledge.worldbank.org). ISBN 978-1-4648-0331-4 SKU 210331