42552 W O R L D B A N K W O R K I N G P A P E R N O . 1 3 1 A F R I C A H U M A N D E V E L O P M E N T S E R I E S Achieving Better Service Delivery Through Decentralization in Ethiopia Marito Garcia Andrew Sunil Rajkumar THE WORLD BANK W O R L D B A N K W O R K I N G P A P E R N O . 1 3 1 Achieving Better Service Delivery Through Decentralization in Ethiopia Marito Garcia Andrew Sunil Rajkumar Africa Region Human Development Department THE WORLD BANK Washington, D.C. Copyright © 2008 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First Printing: January 2008 printed on recycled paper 1 2 3 4 5 11 10 09 08 World Bank Working Papers are published to communicate the results of the Bank's work to the development community 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. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed herein are those of the author(s) and do not necessarily reflect the views of the International Bank for Reconstruction and Development/The World Bank and its affiliated organizations, or those of the Executive Direc- tors of The World Bank or the governments they represent. TheWorld Bank does not guarantee the accuracy of the data included in this work.The bound- aries,colors,denominations,and other information shown on any map in this work do not imply any judgment on the part of The World Bank of the legal status of any territory or the endorse- ment or acceptance of such boundaries. The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/The World Bank encourages dissemination of its work and will normally grant permission promptly to reproduce portions of the work. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, Tel: 978-750-8400, Fax: 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, Fax: 202-522-2422, email: pubrights@worldbank.org. ISBN-13: 978-0-8213-7382-8 eISBN: 978-0-8213-7383-5 ISSN: 1726-5878 DOI: 10.1596/978-0-8213-7382-8 Marito Garcia is a Lead Economist in the Africa Region Human Development Department of the World Bank. Andrew Sunil Rajkumar is an Economist in the same department. Library of Congress Cataloging-in-Publication Data has been requested Contents Foreword. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Acronyms and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv 1. Improvements in Health and Education Services . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Changes in Outcomes over the Past 15 Years . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Was Decentralization Responsible? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2. Decentralization and the Delivery of Basic Services . . . . . . . . . . . . . . . . . . . . . . . . . 7 Phasing in Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 A Framework for Understanding Service Delivery Outcomes . . . . . . . . . . . . . . . . . . 9 3. The Scope of Decentralization and Intergovernmental Fiscal Transfers . . . . . . . 13 Subnational Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 Decentralization of Functions and Responsibilities . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Financing Social Services through Intergovernmental Transfers . . . . . . . . . . . . . . . 18 4. How Did Decentralization to Woreda Level Affect the Delivery of Social Services?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Increases in Woreda-Level Spending Following Decentralization. . . . . . . . . . . . . . . 38 Decentralization Appears to Have Improved the Distribution of Expenditures Across Woredas with Worse-off Woredas Benefiting the Most . . . . . 40 After Woreda-Level Decentralization Began, Some Evidence of Lagging Woredas Partly Catching Up in Social Service Delivery. . . . . . . . . . . . . . 44 Teacher Redeployment Across Woredas and Within Woredas . . . . . . . . . . . . . . . . . . 46 Regression Analysis: Linking Expenditures with Outcomes . . . . . . . . . . . . . . . . . . . 51 Decentralization's High Potential to Improve Service Delivery. . . . . . . . . . . . . . . . . 53 5. Making Decentralization Work: Overcoming Constraints in Decentralized Service Delivery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Critical Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Regional Innovations for Recruiting and Retaining Workers . . . . . . . . . . . . . . . . . . 74 Empowering Citizens and Communities to Improve Services and Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 Improving Accountability Mechanisms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 iii iv Contents APPENDIXES A Calculating Block Grants Allocations From Federal Government to Regions and Regions to Woredas Using the Three-Parameter Formula . . . . . . . . . . . . . . 91 B The "Fiscal Equalization" and "Unit Cost" Approaches for Block Grant Allocations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 C Regression Results on Education Outcomes Before and After Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 D Estimating the Effects of Decentralization on the Delivery of Human Development Services in Ethiopia. . . . . . . . . . . . . . . . . . . . . . . . . . . 103 E Methodology and Technical Details for Data Analysis in Chapter 4 . . . . . . . . 107 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 LIST OF TABLES 1.1. Index of Real Government Expenditures and Spending as Percentage of GDP, 1999­2005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2. User Satisfaction with Government Health Services, 2005 . . . . . . . . . . . . . . . . . . 6 1.3. User Satisfaction with School Services, 2005. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1. Trends in Service Delivery, 1995/96­2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.1. Assignment of Expenditure and Revenue Responsibilities for Education, Health, and Water and Sanitation, by Tier of Government, circa 2005 . . . . . . . 16 3.2. A Wide Variation in Per Capita Block Grant Transfers to Regions . . . . . . . . . . . 20 3.3. Two Different Approaches for Allocating Federal Resources to Regions . . . . . . 22 3.4. Regional Budgets and Share of Budgets Transferred to Woredas in Four Regions, 2005/06 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.5. Block Grant Allocations in Oromiya, 2002/03­2004/05 . . . . . . . . . . . . . . . . . . . 30 3.6. Block Grant Allocations in SNNPR, 2002/03­2004/05 . . . . . . . . . . . . . . . . . . . . 30 3.7. Trends in Regional Expenditure, 1993/4­2005/06 . . . . . . . . . . . . . . . . . . . . . . . . 31 3.8. Aggregate Fiscal Performance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.1. Woreda-Level Spending on Education and Health in SNNPR, by Category of Spending, 2001 and 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.2. Woreda-Level Per Capita Spending on Education and Health in Oromiya, by Category of Spending, 2001 and 2004. . . . . . . . . . . . . . . . . . . . . 43 4.3. Woreda-Level Spending on Education and Health in SNNPR, by Type of Woreda, 2001 and 2004. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.4. Recurrent Expenditure Per Primary Student in SNNPR, 2001 and 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.5. Recurrent Expenditure Per Secondary Student in SNNPR, 2001 and 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.6. Education Outcomes in SNNPR, 2001 and 2004 . . . . . . . . . . . . . . . . . . . . . . . . . 45 Contents v 4.7. Primary Education Outcomes in Oromiya, 2001 and 2004. . . . . . . . . . . . . . . . . 45 5.1. Block Grant Transfers as a Proportion of Federal Revenues, 1996/97­2005/06 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 5.2. Number of Regional and Woreda-Level Personnel in Oromiya Before and After Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 5.3. Staffing in Sector Offices in Delanta Dawnt Woreda, Amhara Region, 2003/04. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 5.4. Staff Vacancies in Selected Woredas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 5.5. Community Contributions to Education in Boset Woreda, Oromiya Region, 2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 5.6. Community Contributions to Education in Boritcha Woreda, SNNPR, 2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 A1. Per Capita Block Grant Transfers and Population . . . . . . . . . . . . . . . . . . . . . . . . 94 C1. School-Level Regressions Explaining Educational Outcomes Before and After Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 C2. Woreda-Level Regressions Explaining Educational Outcomes Before and After Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 D1. Effects of Changes in Expenditures on Educational Outcomes in the Primary Education Subsector: Results of Woreda-Level Regressions for SNNPR. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 LIST OF FIGURES 1.1. Primary Enrollment Increased Rapidly Beginning in the Mid-1990s. . . . . . . . . . 3 1.2. Increases in Primary School Enrollment Since 1995 Occurred Despite Only Modest Increases in Spending on Education as a Percentage of GDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.1. Decentralization has Devolved Responsibility to Subnational Levels of Government . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.2. Accountability for Providing Services Can Follow a Long Route or a Short Route. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 3.1. Federal Transfers to Regions Using Block Grants have been Rising but Not as Fast as Federal Discretionary Spending . . . . . . . . . . . . . . . . . . . . . . . 19 3.2. A Very Close Inverse Relationship Between a Region's Population and its Per Capita Transfer from the Federal Government in 2005/06. . . . . . . . 21 3.3. Following Decentralization, Transfers from Regional Governments to Woredas and Zones Increased, Except in SNNPR, Where They Were Already High . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.4. Regional Budgets With and Without Special Purpose Grants. . . . . . . . . . . . . . . 32 3.5. Regional Revenue as a Share of General Government Revenue, 2002/03­2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 vi Contents 3.6. Real Per Capita Government Spending on Education Rose After 2000, but Much of the Increase went to Tertiary Education . . . . . . . . . . . . . . . . . . . . . 33 3.7. Overall Real Per Capita Spending on Health Increased Between 2000/01 and 2004/05, but Real Per Capita Subnational Government Spending Declined. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.8. Woredas' Share of Regional Recurrent Budgets for Amhara, Oromiya, SNNPR, and Tigray (the Four Main Decentralizing Regions) Together . . . . . . 36 4.1. Expenditures of Woredas Increased After Decentralization. . . . . . . . . . . . . . . . . 39 4.2. Recurrent Woreda-Level Spending Rose in All Sectors in SNNPR Following Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.3. The Share of Total Recurrent Spending for Primary and Secondary Education in SNNPR Rose Following Decentralization . . . . . . . . . . . . . . . . . . . 41 4.4. Gross Enrollment Rates have Improved in All Categories of Woredas in SNNPR Since Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.5. Grade 8 Pass Rates in SNNPR Have Increased Since Decentralization, with Improvement in Pastoral Areas Greater than in Urban Areas. . . . . . . . . . . 47 4.6. Repetition Rates have Fallen Sharply in Urban Areas and Pastoral Areas in SNNPR Since Decentralization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.7. Teachers have been Redeployed from Urban to Non-Urban Woredas Since Decentralization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 4.8. Teacher-Section Ratios Tended to Equalize Since Decentralization Began in SNNPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.9. The Gap in Pupil-Teacher Ratios Across Woredas has Narrowed Since Decentralization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.10. No Reallocation of Teachers or Narrowing of Gaps in Pupil-Teacher Ratios Appears to Have Taken Place in Oromiya Since Decentralization. . . . . . 51 4.11. Differences in Pupil-Teacher Ratios Across Schools Within Woredas in SNNPR have Narrowed Since Decentralization . . . . . . . . . . . . . . . . . . . . . . . . 52 5.1. Per Capita Block Grants to the Regions Averaged just Birr 82 Per Capita in 2004/05 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 5.2. Inadequate Financing Represents an Even Greater Constraint at the Woreda Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.3. Capital Spending by Subnational Governments as a Share of Total Spending has been Declining, Especially in the Social Sectors . . . . . . . . . 64 5.4. Since Decentralization to the Woredas, There has been an Acceleration in the Production of Physical Capital Stock Related to Education and Health Service Delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 5.5. Nursing and Medical Students in Ethiopia Perceive that the Assignment of Posts is Subject to Manipulation . . . . . . . . . . . . . . . . . . . . . . 74 5.6. Community Contributions Represent a Much Larger Source of School Financing Than Incentive Awards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Contents vii 5.7. Enrollment by Girls Rose in BESO Schools. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 5.8. School Attendance is Higher in Woredas in which Primary Education Continues to Grade 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 LIST OF BOXES 3.1. The "Unit Cost" Approach to Block Grant Allocation . . . . . . . . . . . . . . . . . . . . . 25 3.2. Performance Agreements in SNNPR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.3. Devolution of Power in Theory and in Practice . . . . . . . . . . . . . . . . . . . . . . . . . . 27 5.1. Improving the Legal Framework for Decentralization in Tigray . . . . . . . . . . . . 62 5.2. Improving Financial Management Reporting at the Woreda Level. . . . . . . . . . . 69 5.3. Community Participation in Service Delivery at Work: Ethiopia's Basic Education Strategic Objective (BESO) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 5.4. Boosting Girls' Participation in Primary Education Through Girls' Advisory Committees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82 5.5. Community Participation in Water Supply, Sanitation, and Health Education Schemes in Oromiya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 5.6. Using Citizen Report Cards in Ethiopia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 Foreword O ver the last 10 years, several African countries have made notable progress in meeting the Millenium Development Goals (MDGs). Among them is Ethiopia where in 2005 nearly 75 percent of children enrolled in primary schools, a result of an impressive 9 per- cent annual growth in enrollment since 1993. Other indicators also registered improve- ments: immunization coverage for measles rose to 57 percent in 2005 from 40 percent in 1995. The percentage of Ethiopians with access to clean water rose from 19 percent in 1995 to 36 percent in 2005. These improvements happened at a time of massive decentralization in Ethiopia--first from the federal to region, and subsequently into woredas (districts). This work presents an account of how decentralization has supported the delivery of basic services, including education and health to woredas. Because the authors were able to piece together detailed fiscal data from federal, region and woreda levels, as well as education and health data before and after the decentralization, the results provided insights into what has gone into the decisionmaking processes that contributed to the outcomes particularly in the educa- tion sector. It provides a glimpse of how the fiscal and administrative autonomy responds to the needs of their constituencies to improve service delivery. This work provides evidence to the observation that the devolution of power and resources from the federal and regional governments to the woredas contributed to improvements in the delivery of basic services particularly in education. The independent surveys of beneficiaries by NGO groups in Ethiopia, perceive the service coverage and qual- ity to have improved, especially in education. One of the most important results provided by the authors is that the decentralization has disproportionately favored the more remote woredas, the food insecure and pastoral woredas. This is a very important finding from a strategic standpoint because it indicates that decentralization itself is pro-poor, and helps those lagging areas in the country. The aggregate woreda spending for education for pastoral woredas increased dramatically with decentralization, faster than other woredas. Education budgets in food insecure woredas rose much higher than in food secure woredas, and narrowing the gap in educational outcomes. Surely decentralization is only one among many other factors that contributed to the improvement in service delivery in Ethiopia. This work has also documented the changes in sectoral policies, for example the policy of using local languages at the lower primary level which improved school participation. Likewise, the growing role of communities, parent-teachers' associations (PTAs), and other local organizations, may have also con- tributed to better service delivery. In preparing this volume the authors worked extensively with national experts and schol- ars not only from the capital city but also with regional specialists. This provided ample opportunities for both learning and capacity building for analytical work in this area, to understand the impact of decentralized service delivery on the population.Such capacity will no doubt support future work to help finely tune the decentralization processes,and improve the ability to find solutions to improve service delivery in a decentralized system. Yaw Ansu Director, Human Development Department Africa Region The World Bank ix Acknowledgments T his book expands on a report prepared for the World Bank Africa Region's Human Development Department under the general guidance of Laura Frigenti (Sector Manager, AFTH3), Ishac Diwan (Country Director, AFC06), Yaw Ansu (Sector Director, AFTHD) and Trina Haque (Lead Economist, AFTH3). The book benefited from the analyses and reports of Jemal Mohammed Omer (AFTP2), Andrew Dabalen (ECSPE), Mirafe Marcos (SASPR) and consultants Mohammed Mussa, Gizaw Molla, Yigremew Adal, Jean-Paul Faguet, Nada Eissa, Mike Ingram, Tor Halvorsen, Robert Langley Smith, Mammo Kebede, Rahel Kassahun, Ephraim Kebede, and Caroline Pöschl. The Ethiopia Human Development country team led by Trina Haque, and comprised of Getahun Gebru, Gary Theisen, Endashaw Tadesse,William Wiseman, Dr. Gebreselassie Okubaghzi, and Harold Alderman provided useful comments and insights throughout. The authors are grateful for the invaluable comments received from the Ethiopia country team. In particular, they would like to thank Jeni Klugman, Anwar Bach-Baouab, Mesfin Girma Bezagawaw, Jemal Omer, Christine Lao Pena, and Agnes Soucat. The book has benefited from comments by the peer reviewers Shantayanan Devarajan (Chief Economist, SASVP) and Elizabeth King (Sector Manager, DECRG). The authors would like to express their special appreciation and gratitude to Woizerit Samia of the Central Statistical Authority, Ministry of Finance and Economic Develop- ment, and her team for facilitating access to data. Likewise, thanks are due to the Bureau of Finance and Economic Development as well as the Bureaus of Education, Health and Capacity Building of the regional governments of SNNPR and Oromiya for their assistance in obtaining data for this book. xi Acronyms and Abbreviations AED Academy for Education Development ATS Average Teacher Salary BA Bachelors of Arts BESO Basic Education Strategic Objective BOFED Bureau of Finance and Economic Development CSAP Community School Activity Program CGPP Community Government Partnership Program DPT 3 Diphtheria, Pertussis, Tetanus--3 doses of vaccines FDR Four Main Decentralization Regions FSP Food Security Program FY Fiscal Year GER Gross Enrollment Rate GDP Gross Domestic Product GOE Government of Ethiopia HICES Household Income and Consumption Expenditure Survey HIV/AIDS Human Immunodeficiency Virus/Acquired Immunodeficiency Syndrome IMCI Integrated Management of Childhood Illnesses JRM Joint Review Mission MA Masters of Arts MDGs Millennium Development Goals MEFF Macroeconomic and Fiscal Framework MoFED Ministry of Finance and Economic Development NGO Non-governmental Organization PANE Poverty Action Network of Ethiopia PEPs Public Expenditure Programs PSCAP Public Sector Capacity Project PSNP Productivity Safety Nets Program PTA Parent-Teacher Association PTR Pupil-Teacher Ratio SDPRP Sustainable Development and Poverty Reduction Program SNNPR Southern Nations, Nationalities and Peoples Region SPG Special Purpose Grant UNICEF United Nations Children Fund USAID United States Agency for International Development WLI World Learning Incorporated WOFED Woreda Office of Finance and Economic Development xiii Executive Summary E thiopia has made major strides in improving its human development indicators in the past 15 years, achieving significant increases in the coverage of basic education and health services in a short period of time. Improvements took place during a period of massive decentralization of fiscal resources, to the regions in 1994 and to woredas in 2002/03. Decentralization Appears to Have Contributed to Improved Service Delivery . . . The devolution of power and resources from the federal and regional governments to woredas appears to have improved the delivery of basic services. This is clearly shown in the Southern Nations, Nationalities and Peoples Region (SNNPR) and to a lesser degree in the Oromiya Region according to an analysis of woreda-level data before and after decentral- ization in both regions. Surveys of beneficiaries reveal that they perceive that service cov- erage and quality have improved. Beneficiary satisfaction has increased markedly in education, and less conspicuously in water and health services. In SNNPR, the decentralization to woredas in 2002/03 tended to narrow differences in per capita expenditures on education and health across woredas. Decentralization dispro- portionately favored remote (more than 50 kilometers from a zonal capital), food-insecure, and pastoral woredas, suggesting that decentralization has been pro-poor. Between 2001 (pre-decentralization) and 2004 (post-decentralization), for example, the aggregate budget for education in SNNPR rose 44 percent in remote woredas and just 9 percent in woredas located less than 50 kilometers from a zonal capital. A similar trend occurred for health. Aggregate spending on education and health by pastoral woredas rose 67­86 percent between 2001 and 2004, while spending by non-pastoral woredas rose by just 16­23 per- cent. Food-insecure woredas also gained disproportionately from decentralization, with education budgets rising about 30 percent in food-insecure woredas compared to 18 per- cent in food-secure woredas. Decentralization also narrowed the gap in educational outcomes between disadvan- taged and better-off woredas, especially in SNNPR. Pastoral, food-insecure, and remote woredas gained in terms of the educational outcomes examined (gross enrollment rates, Grade 8 examination pass rates, repetition rates, pupil-teacher ratios, and teacher-section ratios). These results mirror changes in spending patterns, which also disproportionately favored lagging areas in SNNPR. The changes in education service delivery may also be partly influenced by better deployment of teachers with decentralization. In SNNPR the redeployment of teachers, especially from towns to nearby rural areas, resulted in a more even distribution of teach- ers across woredas, thus partly alleviating problems in schools with very high pupil-teacher ratios. These changes were influenced by the strong sector policy on teacher deployment and placement, and supported by a budget transfer system using the "unit cost" approach adopted by SNNPR. The decentralization is only one among the myriad of inter-related factors that may have affected service delivery. Sectoral policies in education and health may also have improved service delivery. The policy of using local languages at the lower primary level may have increased school participation, for example. The growing role of communities, parent-teacher associations (PTAs), and other local organizations, which can be measured only qualitatively, may also have contributed to better service delivery. Household surveys have also shown that demand for education has increased over the past decade, as shown by the rising rates of returns for education. Decentralization has also enabled the regions to experiment with innovative service delivery mechanisms such as the use of performance agreements in SNNPR. xv xvi Executive Summary . . . But Weaknesses in the System Threaten the Achievement of Human Development Goals Although social indicators have improved, at the current pace, targets for primary school completion,child and maternal mortality,and clean water and sanitation will not be attained. To increase the pace of improvement, policymakers need to address several weaknesses. Woredas Need More Funding to Adequately Undertake Service Delivery A major constraint is the inadequate funding for basic services provided to lower tiers of government. Woreda administrations, which are responsible for meeting service provision targets, rely almost exclusively on unconditional block grants from regional governments. About 90 percent of these grants is spent on salaries and operational costs, leaving little for other investments essential to reaching the Millennium Development Goals (MDGs). The Capacity of Local Governments Needs to Be Increased Many woredas lack skilled personnel. Administrative and technical posts created to handle administrative responsibilities have not been fully filled or have been filled by untrained personnel with limited capacity. Many woredas also have poor infrastructure to support transformation, with water, electricity, and communication networks either nonexistent or old and obsolete. Progress is being made on these fronts, as capacity building efforts take hold and upgrading of woredas occurs. This progress needs to continue, supported by increases in recurrent and capital budgets for woredas. In addition, the role of capacity building programs, such as the Public Sector Capacity Building Project (PSCAP), is key to deepening decentralization. Need for Clarity in Assignment of Roles Across Tiers of Government Detailed expenditure assignments by level of government have not been clearly defined.The lack of clarity creates a disconnect between responsibilities and finances at lower tiers of gov- ernment, severely so at the woreda level; leads to duplicated effort and lack of coordination; and deprives each level of government of a point of reference against which to budget in order to meet service delivery objectives. Efforts are being made to address this situation, and regional legal frameworks specifying expenditure assignments are being set up; these efforts need to be supported, and continued progress needs to be made in this area. Citizens Do Not Hold Service Providers Accountable Before the start of decentralization, the role of citizens was one of the weakest links in the chain of service delivery in Ethiopia and one of the rationales for decentralization was to strengthen this. Within the decentralized system, the kebele council is the forum where cit- izens could play a role in enforcing accountability of service delivery. But the system is still evolving and initiatives to further strengthen this formal channel of citizens' participation should be welcomed. Opportunity Costs Prevent Many Households from Using Services Many poor households in Ethiopia find it difficult to use education and health facilities because the opportunity costs of doing so are very high. Many children are out of school because they work in household enterprises (mostly farms). Not only distance to schools Executive Summary xvii and health facilities, but also distance to water and firewood, reduces their use of education and health services. Even if the other hurdles to improving service delivery are overcome, it may not be possible to reach the MDG targets if the opportunity costs of many house- holds are not reduced. Policy Recommendations To accelerate the achievement of human development outcomes, several aspects of decen- tralization and service delivery need to be improved. Increase the budget envelop of woredas commensurate to their assigned responsibili- ties for delivering basic services. Lack of adequate resources limits the woredas' ability to empower themselves and improve service delivery. Increasing the block grants from the Federal government to the regions and from regions to woredas would help redress this sit- uation. In this regard, the commitment by the Federal government to increase total block grant transfers to the regions from 5.5 billion Birr in 2004/05 to 21.56 billion Birr in 2009/10, after reaching 9.4 billion Birr in 2006/07, is very encouraging. Implement an appropriate tax-collection and revenue-sharing mechanism to provide incentives to woredas to diversify and expand their sources of revenue. To improve woredas' resources for service delivery, a revenue-sharing and revenue-assignment mechanism needs to be in place to serve as an incentive for woredas to diversify and expand their sources of revenue. Currently, woredas collect revenues on behalf of the regions; they are given revenue targets which are not always realistic. If they col- lect less revenue than the target level, they are often penalized. Conversely, revenues in excess of the targets are often"offset"against the woredas' block grant allocations so that woredas in effect do not retain any excess revenues, and so have no incentive to collect revenues beyond a certain level. Guidelines on revenue-sharing mecha- nisms and revenue assignments are needed. Adopt a unit cost or other needs-based approach for determining intergovernmental fiscal transfers. The data presented in this report indicates that the "unit cost" approach used until recently in SNNPR to allocate block grants to woredas did a better job of improving the efficiency of resource allocation across woredas. The "unit cost"approach was also much better at addressing service delivery needs than the "three-parameter" approach used in several other regions in Ethiopia. The "three-parameter" approach is also flawed in that it has a built-in bias against large localities. This is especially apparent at the Federal level, where it was used until 2006/07 to determine block grant transfer levels to the regions. Until 2006/07, there was a tight inverse relationship between a region's per-capita transfer and its pop- ulation--with factors other than the population, such as a region's development level, appearing to have little influence on the final per-capita transfer level. Improve local government capacity by scaling up implementation of the Public Sector Capacity Building Project to increase the use of resources for capacity building at the woreda level. In this project, utilization of funds is "demand-driven", with regions making applications for funds as needed. The utilization rate has been low, and one of the first priorities should be to increase this rate by boosting the demand for these funds on the part of regions. Improve the capacity and incentives of frontline providers (teachers, health staff, and woreda bureaus). Human resources are critical for service delivery. The regions need xviii Executive Summary to create incentive mechanisms to attract and retain qualified staff at the woreda level. Regions have been experimenting with incentives such as hardship allowances for teachers and health staff posted in remote or arid lands, and training opportu- nities for staff in remote areas. Align sector policies with fiscal decentralization policies to improve impact on service delivery. The experience of SNNPR has shown that improvements in efficiency in education service delivery could be achieved with better deployment of teachers. This was made possible by two policies working in synergy: (a) the regional educa- tion sector policy on deployment and placement of teachers from urban schools with low pupil-teacher ratios to schools with high pupil-teacher ratios and (b) the policy of allocating block grants to woredas using the "unit cost" approach which supported the budgets for such teacher reallocation, and which explicitly aims to allocate resources across woredas in line with needs. Clarify the functional roles between regions and woredas. Legal frameworks and func- tional assignments between regions and woredas should be clearly identified, and woredas should be given decisionmaking power over areas for which they have functional responsibility. Enhance the role of communities. Community involvement in service delivery has been increasing in Ethiopia. In many regions, communities help out by providing financial contributions and by building schools and health posts. Increasingly, com- munity organizations like the parent-teacher associations (PTAs) are assisting in mobilization and have helped increase enrolment rates in many parts of the coun- try. Regional experiments involving PTAs in the management of schools have been successful in improving service delivery. In this respect it is recommended that such involvement of PTAs in the management of schools be promoted systematically across all regions. Strengthen transparency and accountability in service delivery. SNNPR has successfully used performance agreements between the region and woredas. The agreements allow for a more transparent monitoring of inputs and outputs in service delivery between regions and woredas, providing a mechanism for ensuring accountability and effec- tive resource management. Such agreements should be adopted throughout the country. More formally, one should ensure that citizen voice can be strengthened within the purview of the Kebele Council. Furthermore, citizens can be involved in local management of services through PTAs and local health boards. PTAs can coor- dinate the use of community scorecards of school performance while participatory planning and monitoring by citizens can be used to strengthen the accountability of woreda and kebele councils. Citizen report cards, communicating the importance of service quality and coverage, can be an input into a system of evaluation of local gov- ernments against agreed performance contracts. In addition, ongoing initiatives to post woreda and regional budgets in public places, and to present budgets in formats that are readily understood by citizens, should be strengthened and encouraged. It is essential to promote the sharing of experiences from all these experiments across all regions. Use public information campaigns to improve education and health outcomes. Child health outcomes could be improved significantly by providing nutrition information to mothers or helping households understand the value of boiling water, use of oral Executive Summary xix rehydration salts, and hand washing. Decentralization and citizen empowerment should magnify the effects of public information campaigns. Woredas and munici- palities could use health extension workers and community or municipal radios to disseminate knowledge on child health. Systematically evaluate key reforms. Ethiopia is in the process of implementing a major transformation of its political economy. Far-reaching reforms have been adopted in school curricula, the structure of the education system, school fees, the structure of the health system, user fees, and modes of delivery of primary health services through health extension workers. However, no systematic attempt has been made to evaluate many of these interventions to understand what works well and what does not. Rigorous evaluation of these reforms is needed to ensure that the proposed scaling up is informed by lessons learned to date from local as well as international experience. CHAPTER 1 Improvements in Health and Education Services T his report seeks to identify changes in human development outcomes in a period of deepening decentralization and to suggest how the country's decentralized gov- ernance structure could be improved to increase access to, as well as the quality of, relevant services. A key message of the report is that the decentralized governance structure helped facil- itate improvements in service delivery and human development outcomes but that weak- nesses in that structure can derail these gains. The report argues that while policymakers, providers, and citizens must work together to strengthen accountability mechanisms, there is a particular need to strengthen local government and enhance the role of service beneficiaries. The report focuses on key actors and their roles in accelerating progress toward achiev- ing the MDGs in Ethiopia. It complements rather than duplicates other recent studies. The education and health country status reports for Ethiopia (World Bank 2005a, 2005b) are rich in both content and policy recommendations. They examine in greater detail the determinants of education and health outcomes and propose many insightful policy options, using both administrative and household survey data. The Poverty Assessment (World Bank 2005c) draws on and enriches the findings in the country status reports by linking these outcomes to poverty. It extends the analysis to cover more general questions of economic opportunities and welfare. This study also draws on the Pilot Citizens Report Card prepared by the Poverty Action Network of Ethiopia (PANE 2005), the Woreda-City Government Benchmarking Survey conducted by GTZ-Selam Development Consultants (2005), an inter- governmental fiscal review by Heymans and Mussa (2004), a Fiduciary Assessment report (Government of Ethiopia 2005b), and an Oxfam study on implementation of decentraliza- tion policy in Ethiopia (Adal 2005). 1 2 World Bank Working Paper The report is organized as follows. This chapter reports on trends in human development indicators and government social expenditures in Ethiopia over the past 15 years. Chapter 2 describes the phasing in of Ethiopia's decentralization plan and presents a framework of mechanisms of accountability for providing services. Chapter 3 examines the scope of decen- tralization in Ethiopia and the use of intergovernmental transfers. Chapter 4 presents some of the data collected for this report on changes in budget allocation behavior at the woreda level before and after decentralization and the effects on allocations for the social sectors. It also examines changes in key service delivery outcomes at the woreda level. Chapter 5 exam- ines some of the constraints to improved service delivery and proposes ways of increasing resources at decentralized levels, of improving the capacity of local governments and service providers, and of involving citizens in service delivery in a more robust and effective way. It stresses the critical need to clarify functional expenditure assignments and rules governing the allocation of block grants, to establish guidelines for communities' role in service delivery, and to empower communities by providing them with more information. Changes in Outcomes over the Past 15 Years During the past 15 years, Ethiopia has witnessed rapid improvements in health and educa- tion outcomes. Enrollment rates at all levels have risen substantially, albeit from a low base (Figure 1.1). The total number of primary schools has risen steadily over the last 40 years, with a recent sharp acceleration starting in 2004. School enrollment surged from 1993 onwards. Coverage in health services also increased, although to a smaller extent than in educa- tion. For most Ethiopians, health facilities are closer than they were 10 years ago, and their rate of usage is higher, with the annual number of per capita visits to health facilities increas- ing from 0.3 in 1993 to 0.4 in 2005. Progress was also made in child health outcomes. Immunization rates for DPT 3 remain low, but they reached 60 percent by 2005, up from 46 percent in 1995. 80 percent of the population is now immunized against polio. Coverage of water services increased, with the proportion of households with access to safe drinking water (water from a tap or covered well) rising from 19.1 percent in 1996 to 35.9 percent in 2004 (GOE/MoFED 2005a). Over the same period, the proportion of house- holds with access to improved sanitation (flush toilet or pit latrine) rose from 13.0 percent to 30.6 percent. Was Decentralization Responsible? Basic coverage accelerated during the period of decentralization, but without further analy- sis it is not possible to conclude that decentralization was responsible for the change, as decentralization was part of a package of government policies and actions that included school construction, teacher hiring, provision for local language instruction at the primary level, and adjustment of the school calendar (especially in rural areas) to accommodate the agricultural cycle of farm households. Government spending on education and health rose, both as a share of GDP and as a share of total government spending (Table 1.1). Public education expenditures almost tripled between 1999/2000 and 2005/2006 in real terms, rising from 3.1 percent to 5.9 per- Achieving Better Service Delivery Through Decentralization in Ethiopia 3 Figure 1.1. Primary Enrollment Increased Rapidly Beginning in the Mid-1990s 14,000 18 16 12,000 14 10,000 ('000s) 12 ('000s) 8,000 10 schools 8 enrollment 6,000 primary 6 of 4,000 primary 4 number 2,000 2 0 1968 1973 1978 1983 1988 1993 1998 2003 0 primary enrollment number of primary schools Source: World Bank staff estimates from survey and administrative data; World Bank (2005a); data from the Annual Abstracts on education. cent of GDP over this period. Over the same period, real public health expenditures almost doubled, rising from 1.1 percent to 2.0 percent of GDP. The share of government spending going to education almost doubled between the col- lapse of the Derg government in 1990 and the beginning of decentralization to the regions in 1993/94 (Figure 1.2). Given the relationship between spending on education and primary school enrollment, it is tempting to conclude that most of the growth in primary enroll- ment was driven by increases in public spending on education. Such a conclusion would be misguided, however, for two reasons. First, some of the growth in spending was on secondary and tertiary education, where spending has risen more rapidly than with primary education. Second, a closer look at the relationship between spending and primary enrollment indicates that observed increases in government spending are insufficient to account for the rapid rise in enrollment, which rose much more rapidly than government spending (Figure 1.2). In short, there is a momentum to enrollment growth that cannot be explained by the increase in public spending alone. 4 World Bank Working Paper Table 1.1. Index of Real Government Expenditures and Spending as Percentage of GDP, 1999­2005 2000/01 2001/02 2002/03 2003/04 2004/05 2005/06 1999/2000 Pre- Pre- Pre- Pre- Pre- Pre- Item Actual Actual Actual Actual Actual Actual Actual Index of Real Expenditures (1999/2000 = 100) Total 100 97.8 103.0 114.1 107.1 121.7 129.2 government expenditure Defense 100 51.0 43.4 33.8 32.6 36.3 33.4 Debt servicing 100 101.5 101.8 107.3 87.5 76.7 71.2 Agriculture 100 156.7 148.7 173.3 223.5 342.3 368.0 and natural resources Roads 100 216.6 203.3 225.7 201.0 264.6 369.5 Education 100 138.6 168.8 197.6 230.8 252.2 294.2 Health 100 236.9 220.7 187.2 149.2 191.1 190.1 Other sectors 100 96.5 121.6 156.1 122.0 116.0 107.9 Percentage of GDP Total 26.7 23.0 21.6 29.1 27.2 29.3 33.4 government expenditure Defense 10.6 4.8 3.6 3.4 3.3 3.5 3.4 Debt servicing 1.7 1.6 1.4 1.8 1.5 1.2 1.2 Agriculture 2.0 2.7 2.3 3.3 4.2 6.2 7.1 and natural resources Roads 1.4 2.6 2.2 3.0 2.6 3.3 5.0 Education 2.6 3.1 3.4 4.8 5.6 5.8 7.3 Health 0.8 1.7 1.4 1.5 1.2 1.4 1.5 Other sectors 7.6 6.4 7.2 11.3 8.8 7.9 7.9 Source: Data provided by MoFED. One source of this additional momentum is the change in policies, especially the teach- ing of primary school in local languages and the adoption of a flexible calendar in rural areas that takes into consideration the agricultural cycle. Another source may have been the shift in demand for education and health that is likely to have taken place among the house- holds and individuals most likely to invest in health and education. The rate of return to education in Ethiopia was about 6 percent in the early 1990s (Suleiman 1995; Weir and Knight 1996; Krishnan 1996). More recent studies on the demand for education, based on surveys conducted in early 2001, show that returns had risen to 8­12 percent (World Bank 2005a). The end of civil strife, the return of law and order, and the establishment of a gov- ernment of national unity may have convinced many households and individuals that the new social order was irreversible in the medium to long term. These perceptions may have Achieving Better Service Delivery Through Decentralization in Ethiopia 5 Figure 1.2. Increases in Primary School Enrollment Since 1995 Occurred Despite Only Modest Increases in Spending on Education as a Percentage of GDP 80 18 70 Enrollment 16 (Grades 1-8) 14 60 primary spending) 12 50 education ratio, 10 on public 40 8 total 30 6 spending and enrollment Education spending 20 (% of GDP) 4 GDP Gross 10 Public of Education spending 2 (% of Total Public Spending) (% 0 0 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 Source: World Bank (2005a). encouraged many households to make long-term investments in their own education and health and that of their children. This shift in demand is unlikely to have been a major cause of the increase in enrollment, however, because the opportunity costs of education remain formidable, as discussed in Chapter 5. Consultations with a wide array of Ethiopians--from woreda officials to community residents--reveal several advantages of decentralization. According to the people interviewed for this study, locally relevant solutions to problems are now found promptly through more democratic and representative processes. Plans are locally drafted and approved, budgets are managed locally, and control is more effective. Budget shortfalls are dealt with more fairly under decentralization, and the relationship between tiers of government is character- ized more by partnerships than the more remote and bureaucratic structures that prevailed before decentralization (Halvorsen, Smith, and Shenkut 2005). Beneficiary satisfaction is improving. Three consumer satisfaction assessments--the Welfare Monitoring Survey 2004, the Citizens' Report Card (PANE 2005), and the Ethiopia Participatory Poverty Assessment 2004/05 (GOE/MoFED 2005b)--point to moderately improving service provision in health, education, and water services. Similar findings were obtained in the Woreda-City Government Benchmarking Survey (GTZ-Selam Development Consultants 2005). The Citizens' Report Card provides feedback from users of public services on access to, quality of, and adequacy of services. It is based on a sample of 3,829 households in 10 woredas and municipalities in Afar, Dire Dawa, Oromiya, Tigray, and the Southern Nations, Nationalities and Peoples Region (SNNPR). A surprising result from this survey is the moderate level of satisfaction with health services, given the relatively poor coverage 6 World Bank Working Paper Table 1.2. User Satisfaction with Government Health Services, 2005 Measure Dire Dawa Oromiya SNNPR Tigray Time taken to attend 63 54 57 79 Behavior of staff/doctor 60 56 58 80 Overall satisfaction 46 50 46 54 Note: Figures reflect percentage of respondents that are completely satisfied. Source: PANE (2005). and quality of care of the Table 1.3. User Satisfaction with School Services, 2005 facilities (Table 1.2). About Measure Dire Dawa Oromiya SNNPR Tigray half of respondents reported Quality of teaching 76 68 75 87 being satisfied with the ser- School building 80 51 58 68 vices received at health facil- Toilet 69 58 75 49 ities. Satisfaction was highest Behavior of teachers 73 71 78 90 in Tigray, where 80 percent Overall satisfaction 72 59 56 41 of clients reported being sat- isfied with the behavior of Note: Figures reflect percentage of respondents that are com- pletely satisfied. doctors. Source: PANE (2005). Satisfaction with the quality and behavior of tea- chers is also surprisingly high among users, across the country, possibly as a result of improvement in the qualifications of lower primary level teachers, 96 percent of whom now have Teachers Training Institutes certificates and a rising proportion of whom have College of Teachers Education diplomas (Table 1.3). Almost two-thirds of respon- dents were satisfied with school infrastructure, one of the key areas of improvement in service delivery at the woreda level. For most households, schools are within 5 kilometers from home. In some woredas in Tigray, however, about a third of children travel more than 5 kilometers to school. While textbooks and teaching aids are usually available in schools, toilet and water facilities are severely lacking, with just 10 percent of schools in Tigray, 25 percent in SNNPR, and 33 percent in Oromiya having toilets. CHAPTER 2 Decentralization and the Delivery of Basic Services E thiopia has made considerable progress toward meeting the Millennium Develop- ment Goals (MDGs) (Table 2.1). In 2005 nearly 75 percent of Ethiopian children were enrolled in primary schools, the result of an impressive 9 percent annual growth in enrollment since 1993. Similar expansion was achieved at the secondary and post- secondary levels. Immunization coverage for measles stood at 57 percent, up from 40 percent in 1995. Stunting fell from 66 percent in 1996 to 47 percent in 2005, while under- five mortality declined from 21 percent in 1990 to 14 percent in 2003. The percentage of Ethiopians with access to clean water rose from 19 percent in 1995 to 36 percent in 2005. Phasing in Decentralization Improvements in health and education outcomes in the past 15 years occurred at a time of massive decentralization in Ethiopia. Ethiopia introduced decentralization as the strategic tool for empowering citizens and devolving power to lower levels, following the Constitu- tion. This, in turn, was expected to establish a conducive environment for enhancing the delivery of basic services. The decentralization process was implemented in two phases. The first phase created a four-tier governance structure, consisting of the center, the regions (nine ethnic-based states plus the cities of Addis Ababa and Dire Dawa), the zones, and the woredas.1 During this phase, the regional governments were given responsibility for delivering all education (except tertiary and secondary teacher training) and health services. To facilitate these 1. Ethiopia has 700 woredas, with an average population of just over 110,000. 7 8 World Bank Working Paper Table 2.1. Trends in Service Delivery, 1995/96­2004/05 (percent) Percentage Percentage Year Change Change Indicator 1995/96 1997/98 1999/00 2004/05 1996­2004 2000­2004 Primary (Grades 1­8) Gross enrollment rate 37 52 61 74 37 13 Literacy rate 26 27 29 38 12 9 Under-five child immunization Measles 39 47 48 57 18 9 BCG 35 47 46 55 20 9 Access to safe drinking 19 24 28 36 17 8 water Source: PASDEP (Government of Ethiopia/MoFED, 2005a). functions, thousands of civil servants were redeployed from the center to the regions. The zones and the woredas are lower-level tiers of the regional governments, with woredas below the level of zones. During the second half of the 1990s, the Government, together with development partners, conducted a number of studies to identify the factors that hindered public sector efficiency, grassroots empowerment, and accountability. An important lesson learned from these studies was that woredas had very limited fiscal or administrative autonomy with which to respond to the needs of their constituencies. In response to this finding, in 2002/03 the Government initiated the second phase of decentralization with a series of legal, fiscal, and administrative reforms beginning with four of the largest regions (Amhara, Oromiya, SNNPR, and Tigray), which together account for 87 percent of Ethiopia's population. Under this phase, some control was also devolved to the woreda level, and eventually urban administrations with woreda status and responsibil- ities were created in urban areas. (Important note: Throughout this document, the term "woreda" includes urban administrations by definition, unless otherwise stated.) Within the four regions, woredas now manage about 45 percent of regional public expenditures. This phase of decentralization seeks to empower communities to engage in development inter- ventions, improve local democratic governance, and enhance the scope and quality of the delivery of basic services at the local level. The new governance framework establishes the foundations for consolidating account- ability mechanisms for better services. It delegates responsibility for service delivery to local governments (Figure 2.1). Clarification of these responsibilities continues to be worked out across tiers of government. The new framework also devolves substantial control over real resources (personnel, assets, and finances, through block grants) to subnational governments (regions and woredas). It provides a platform for citizen participation in politics and the enforcement of agreements. The opening of service delivery to private and nongovernmen- tal actors and the empowering of civil society, though just beginning, are promising steps. While decentralization has empowered local decision makers to set priorities in line with local demand, fiscal and human resources remain a major impediment. Woredas have Achieving Better Service Delivery Through Decentralization in Ethiopia 9 had to manage carefully to meet basic recurrent requirements, mainly salaries, leaving few resources for capital investment, a much needed requirement for expanding services. These constraints are analyzed in Chapter 5. While human development outcomes have improved, far too many children remain out of school, are in poor health, and die from easily preventable diseases. In education, arguably the most successful effort, at least 4 million children are not served by any school, and many children who have access to a school cannot attend or are not learning, for a vari- ety of reasons. Ethiopia's indicators still lag behind those observed in other poor countries, and progress in many areas--service quality, equity of access, complementary services-- has not been as rapid or has deteriorated. In both education and health, the quality of inputs (staff, books, drugs) is inadequate; in water, sanitation, and electricity, huge inequities of access between rich and poor remain. Figure 2.1. Decentralization has Devolved Responsibility to Subnational Levels of Government Federal (center) Finance, delegation Region Politicians / Policymakers Voice (elections) Finance, delegation, monitoring Voice Compact (elections) Woreda (regional) Politicians / Policymakers Voice Compact (woreda) (elections) Client power Citizens/Clients (woreda) Providers Client power (regional) A Framework for Understanding Service Delivery Outcomes In the past, analysis of outcomes focused on the interactions between individual and house- hold variables (preferences, effort, prices, income) and the availability, quality, and financing arrangements within sectors. This focus promoted technical fixes and larger budgets, both of which, while important, were found insufficient. 10 World Bank Working Paper To broaden the focus of the analysis, the World Development Report 2004: Making Ser- vices Work for Poor People called for a more encompassing approach. The key message of that report is that improving outcomes requires making services work for all people, especially the poor. Doing so calls for understanding, establishing, and strengthening a service delivery framework, or chain. The framework adopted in the World Development Report 2004 focuses on the behavior of actors within the chain, who are linked by accountability mechanisms (Figure 2.2). The chain is defined by the roles of four actors: politicians/policymakers; organizational providers (public, private, nongovernmental organizations); frontline professionals (doctors, nurses, teachers, and so forth); and citizens/clients (patients, students, parents, voters). The accountability relationships are defined by rules and arrangements that include delegation, financing, performance, information, and enforcement. Figure 2.2. Accountability for Providing Services Can Follow a Long Route or a Short Route The State Politicians Policymakers Compact Voice Citizens/Clients Providers Coalitions/inclusion Client power Management Non-poor Poor Frontline Organizations Services Source: World Development Report 2004. The framework identifies a long route and a short route of influence. Along the long route, citizens use political institutions to hold politicians and policymakers accountable, and politicians and policymakers enter into agreements with providers to deliver services to citizens. Citizens delegate policymaking and the allocation of resources across compet- ing needs to politicians and policymakers, who use providers to implement these policies. Politicians and policymakers provide providers with policy priorities and resources, with which providers manage the frontline professionals who run services. Along the short route, the influence of citizens/clients on providers is more direct. Citizens/clients can switch providers, they can participate in decisionmaking, and they can use better information to monitor the actions of providers and enforce compliance. Achieving Better Service Delivery Through Decentralization in Ethiopia 11 All of these relationships are important; any weak link in the chain can lead to service delivery failure. In an ideal world, the benefits from services will be maximized if both the long and the short routes function: politicians/policymakers respond to citizen/client demands by establishing strong agreements, which citizens/clients (together with politicians/ policymakers) monitor and enforce by exercising power over providers. How well these accountability mechanisms, especially the power of citizens/clients over politicians/ policymakers and providers, work often depends on a country's political economy. The framework depicted in Figure 2.2 suggests that along the long route, accountabil- ity is best accomplished in an environment in which citizens/clients are principals--that is, actors who delegate policymaking to politicians/policymakers but wield the power to remove politicians from office for poor performance. To retain the support of citizens/clients in every electoral cycle, politicians/policymakers establish and enforce strong agreements with providers. The long route thus has a better chance of working in a well-functioning democratic environment. For most of its history, Ethiopia has not had a competitive democratic system. Under both the monarchy and communist rule, citizens had no voice in Ethiopia. Since 1991 Ethiopia has had a constitutional democracy. Three national elections have been held, the second in the midst of a war. The third, held in May 2005, appeared to usher in more reforms, although post-election violence has slowed the otherwise orderly democratic process. Ethiopia's electoral system is still new, and it has not always been viewed as completely free or competitive. But there is now space for new and hitherto excluded groups to orga- nize and to voice their demands for better representation and access to services. Whether the full potential of the emerging long route of accountability will be maximized remains to be seen. More far-reaching has been the administrative and fiscal dimensions of the decentral- ization process. The motivation for Ethiopia's decentralization effort was--in line with the Constitution--to improve on the record of centralized control and allocation of resources, which ignored local preferences and neglected entire regions. Decentralization appears to be an attempt to implement and reap the benefits of the accountability framework by reducing the distance between citizens and politicians/policymakers. CHAPTER 3 The Scope of Decentralization and Intergovernmental Fiscal Transfers R egional constitutions and urban proclamations have defined the broad functional assignments of local governments since decentralization. Regional governments are still fine-tuning the detailed assignments of both expenditures and revenues. This chapter describes the kinds of services and functions assigned to lower tiers of government. It examines the relationships between woreda-level governments (woreda administrations) and governments at the zonal, region, and federal levels. It also describes the financing of these services, reviewing trends in intergovernmental fiscal transfers from the federal government to the regions and from the regions to the woredas (including urban administrations by definition). Subnational Structures Decentralization in Ethiopia established nine regions (Afar, Amhara, Benishangul-Gumuz, Gambella, Harari, Oromiya, SNNPR, Somali, and Tigray) and two city administrations (Addis Ababa and Dire Dawa). The regions are divided into zones, woredas/urban admin- istrations, and kebeles (village areas, with an average population of 5,000). The city admin- istrations of Addis Ababa and Dire Dawa have different structures but are considered the equivalent of regions. Over the past two years, the zones in most regions have become arms of the regional government. The woredas are considered the key local government units in each region, with significant responsibility for providing basic services. The local government landscape follows similar patterns across regions. There are 140 local governments (128 rural woredas and 12 urban administrations) in Amhara, 284 local 13 14 World Bank Working Paper governments (248 rural woredas and 36 urban administrations) in Oromiya, 156 local gov- ernments (137 rural woredas and 19 urban administrations) in SNNPR, and 47 local gov- ernments (34 rural woredas and 13 urban administrations) in Tigray. However, these numbers are constantly changing; the number of woredas has been steadily increasing over the last few years because of frequent splitting of existing woredas to create new ones. Regional Governments Each region has at its apex a regional council, with regional council members representing each woreda directly elected. Councils have legislative and executive authority to direct the internal affairs of the levels of government under their jurisdiction. Administrative func- tions are executed by regional bureaus, structured on a sector basis. In each region, a cab- inet, under a regional president, performs the executive role. The regions are governed by city proclamations and other local government legislation, but all are also in the process of updating this to deal with complex demands of the decentralization structure. The role of regions and zones in relation to local governments varies from region to region. Woreda Administrations The tripartite structure of council, executive committee, and sector bureaus is replicated at the woreda level. Woreda councils consist of directly elected representatives from each kebele in the woreda. The woreda councils have dual accountability, upward to the regional cabinet/council and downward to the electorate. Woreda cabinets consist of about a dozen members, drawn from elected representatives and heads of sectoral offices. The main con- stitutional powers and duties of the woreda council and its executive are preparing and approving the annual woreda development plans and budgets and monitoring their imple- mentation; setting certain tax rates and collecting local taxes and levies (principally land use taxes, agricultural income taxes, sales taxes, and user fees); remitting a portion of the local tax take to the zone; administering the fiscal resources available to the woreda (from own sources and transfers); constructing and maintaining low-grade rural tracks, water points, and woreda-level administrative infrastructure (offices, houses); administering pri- mary schools and health institutions; managing agricultural development activities; and protecting natural resources. Urban Administrations Urban administrations--often also called urban local government administrations or sim- ply urban governments--have the same status as woredas. They perform state and munici- pal functions. State functions include health, education, and agricultural services. Municipal functions include preparation, approval, and implementation of development plans; assess- ment and collection of allowable municipal revenues; provision of internal roads and bridges; provision of markets, slaughter houses, terminals, public gardens, recreational areas, and other public facilities; regulation of cleanliness and provision of solid waste, water, sewerage, and drainage services; management of urban land and provision of urban land services; and delivery of miscellaneous services, including fire protection, libraries, public toilets, street lighting, nursery schools, and ambulance services. Achieving Better Service Delivery Through Decentralization in Ethiopia 15 In addition to its 84 urban administrations, Ethiopia has 863 municipalities/towns. These municipalities are under the jurisdiction of the woredas and generally carry out municipal functions. They do not have woreda status and do not carry out state functions. They typically do not receive transfers from the regional or other governments, relying on own sources of revenue. It is expected that an increasing number of municipalities will seek to become urban administrations with woreda status in the future. Kebele Administrations The kebeles do not enjoy the same constitutional formality as regions. Their administrations consist of an elected council (in principle of 100 members), an executive committee of five to seven citizens, and a social court. The main responsibilities of the kebele council and exec- utive committee are preparing an annual development plan, ensuring the collection of land and agricultural income tax, organizing local labor and in-kind contributions to develop- ment activities, and resolving conflicts within the community through the social courts. Decentralization of Functions and Responsibilities Decentralization devolved expenditure and revenue responsibilities for education, health, and water and sanitation to different tiers of government (Table 3.1). The definition of functions is evolving, although across regions the trend is toward decentralizing basic service delivery functions to the woredas and assigning higher order functions--such as provision of hospital care--to regional governments. Tertiary services, such as tertiary education, standard setting, and operation of specialized hospitals, remain the responsibility of the federal government. Allocation of Responsibilities for Education Woredas are responsible for establishing and administering basic education services, includ- ing primary schools (Grades 1­8), the first cycle of secondary school (Grades 9­10), and adult education. Their responsibilities include supervising these schools; printing and distributing primary school textbooks; and establishing and administering primary boarding schools. In the delivery of basic services, the woredas are tasked with planning and developing short- and long-term plans for education in their jurisdictions and ensuring that plans are implemented, by both government and nongovernmental schools, according to the standards set at the regional and federal levels. This function includes planning the location of new schools in kebeles ensuring equity in access to education; strengthening educational super- vision; enhancing community participation by supporting citizen participation in educational administration; and encouraging and supporting parent-teacher associations (PTAs). In principle, the woredas are responsible for hiring elementary school teachers, deploy- ing them to specific schools, and paying their salaries. In collaboration with the Woreda Education and Training Board, they make and implement decisions about disciplinary measures regarding school heads. In SNNPR the zones and special woredas are responsi- ble for translating the curriculum into local languages and providing technical support and supervision to the woreda bureaus. Most zones are made up of several woredas; special woredas are different in that they are single woredas that also have the status of a zone. 16 World Bank Working Paper Table 3.1. Assignment of Expenditure and Revenue Responsibilities for Education, Health, and Water and Sanitation, by Tier of Government, circa 2005 Tier of Expenditure and functional government responsibilities Revenue sources Federal National standard setting for education, Custom duties; export and import health, and water and sanitation taxes or levies; income and enter- Tertiary education prise taxes; taxes on proceeds of national lottery, stamp duties, Specialized/referral hospitals income from federal monopolies, and fees from transport services (road, air, water, and rail).1 Regional Standard setting for primary and sec- Grants from federal government ondary education and regional health (70­80 percent of revenues) Vocational and technical training, Income taxes (personal, sales, teacher training institutes, and corporate, profit, property, and medium-level colleges other) Hospitals and referral hospitals Fees on agricultural land, licensing, Control and prevention of HIV/AIDS royalty, forest resources, water Immunization use, and other activities Coordination and execution of civil ser- Fees on health services, such as vice reform programs at regional level drugs Policy setting on regional water resource development and protection Second cycle of secondary education (Grades 11­12) Woreda Coordination of school management and Grants from regional government cost-sharing activities (80­90 percent of revenues)2 Coordination of primary preventive and Personal income tax, agricultural curative health care activities income tax, rural land use tax, rental income tax, licenses and Establishment and administration of fees primary schools (Grades 1­8) and fist cycle of secondary education (Grades 9­10) Implementation of health extension Construction and administration of health stations and health posts and administration of clinics Control and prevention of HIV/AIDS and malaria Immunization Coordination and management of drink- ing water supplies Notes: 1. The federal government shares revenues from profit, excise, and personal income taxes; divi- dend taxes; and income derived from mining, petroleum, and gas operations with the regions. Currently there are no export taxes, but the Federal government has the mandate to impose them. 2. Many municipalities do not receive block grants but are self-financing. Source: Government of Ethiopia (2005a); Amhara National Regional State (2005). Achieving Better Service Delivery Through Decentralization in Ethiopia 17 The zones and regions are responsible for establishing and administering the second cycle of secondary education (Grades 11­12), technical and vocational schools, special schools, teacher training institutes, and medium-level colleges. Regions are also responsible for printing and distributing books for secondary schools; administering boarding schools in pastoral areas; formulating regional education policy and strategy; preparing curriculum for primary schools; administering educational radio broadcast centers; providing techni- cal and material support; training primary school teachers, and ensuring that national edu- cation sector service standards are met. In large regions like Oromiya and SNNPR, the zones have much more responsibilities than in other regions and are performing functions assigned to regions. In regions with large zones, the zones coordinate the distribution of textbooks and other educational materials, and provide support to woredas. In many regions, zones are assigned functions that cannot be implemented by woredas because of capacity constraints. These constraints, arising from the lack of clarity in responsibilities and expenditure assignments, are described more fully in Chapter 5. Regional governments also coordinate the efforts of other levels of government, NGOs, and other public organizations toward meeting educational objectives in the region. They work with the zones as well as the woredas in deploying teachers, especially graduates of regional institutions, to woredas. The federal government is responsible for establishing and administering tertiary insti- tutions; developing national education policy; establishing national education standards and monitoring observance by regions; supporting regional curriculum development efforts for primary education, secondary education, and technical and vocational educa- tion; providing technical and professional support to regional education bureaus; setting standards for national examinations and standards of certificates; establishing professional standards for teachers; and supporting the training provided by the regions. In coopera- tion with relevant bodies, the federal government has devised mechanisms to enhance edu- cational support to girls, women, and those from minority groups. Allocation of Responsibilities for Health The woredas are responsible for coordinating primary preventive and curative health care and implementing health extension; constructing and administering health stations and health posts; administering clinics; and preventing and controlling HIV/AIDS and malaria. Their responsibilities include the hiring of health staff assigned to health stations, health posts, and clinics. The regions are responsible for formulating regional health policy; coordinating health extension activities; establishing and administering training institutions and junior colleges for junior health professionals; training health professionals; providing technical support to zones and woredas; establishing and administering health exami- nation centers; building and maintaining referral hospitals; coordinating the control of communicable diseases; purchasing and distributing medical equipment and medicines; and preventing and controlling HIV/AIDS. In large regions the zones are responsible for providing technical support to woredas; establishing and administering zonal and dis- trict hospitals; constructing and administering pharmacies; and preventing and con- trolling HIV/AIDS. 18 World Bank Working Paper The federal government is responsible for establishing and administering specialized and referral hospitals. It is also responsible for setting national health policy; establishing national health standards and basic criteria for health; and monitoring implementation across regions. Allocation of Responsibilities for Water Services The woredas are tasked with providing support to communities to administer water schemes; develop springs; build hand-dug wells; handle minor and moderate maintenance of water schemes; and promote indigenous irrigation schemes. The zones are responsible for providing the woredas with training and technical sup- port; constructing water supply projects in small towns; and designing, building, and main- taining medium-size springs and wells. The regions are responsible for formulating water development and conservation strat- egy; providing training on operation and maintenance of watering schemes; evaluating the need for, designing, building, and maintaining boreholes; setting standards for different water- ing scheme services; evaluating the need for and building town water supply facilities; and eval- uating the need for, designing, and building large- and medium-scale irrigation projects. The federal government is responsible for setting national standards for water and san- itation, including the use of water, lakes, and rivers linking and crossing Ethiopia's national boundaries. Financing Social Services Through Intergovernmental Transfers Public financing of basic social services is implemented through intergovernmental fiscal transfers from the federal government to the regions as well as from regional governments to woredas. In implementing the decentralization strategy, the Government has developed a body of rules that govern transfers from one tier to the next. The most important source of financing of the regional and woreda-level governments is the block grant transfer from the next higher level of government. This section describes the features of the allocation rules and recent trends in their implementation. Fiscal Transfers from Federal Governments to the Regions A major transfer of federal resources to the regions takes place through the federal grant sys- tem, which consists of block grants and specific-purpose grants. Most of the transfer takes place as block grants. In 2003/04, the Federal government tried to introduce a particular type of formula (described below) to make allocations of block grants to each region. However, the formula--when applied to the entire pool of resources available for block grants--was found to provide inadequate allocations to some regions, especially the ones with increas- ing recurrent costs and expanding numbers of public facilities. Therefore, it was decided that in every year from 2003/04 onwards, a two-step allocation process would be used. First, each region would obtain the same recurrent allocation as in 2002/03; this would be a min- imum allocation. Second, after making these initial minimum allocations, the formula would be used to distribute additional resources across regions. Achieving Better Service Delivery Through Decentralization in Ethiopia 19 From 2003/04 to 2006/07, the above two-step allocation process was used to make Fed- eral block grant allocations across regions. The formula used for the second step of this allo- cation process is called the "three-parameter" block grant allocation formula. It takes into consideration the population size (65 percent), poverty and development level (25 percent), and an index of revenue effort and sectoral performance (10 percent) of each region. One problem is that the underlying data (on poverty and development level, etc.) used to apply the formula were not updated since 2003/04, even though the formula was used to make allocations every year from 2003/04 to 2006/07. The method for applying this formula is given in Appendix A. Note that some adjustments were made to the formula described to account for recurring budget deficits. The aim of the federal block grant transfer is to address the vertical imbalances in rev- enue versus expenditure assignments between the federal and regional administrations. Figure 3.1. Federal Transfers to Regions Using Block Grants have been Rising but Not as Fast as Federal Discretionary Spending 14.0 12.0 10.0 Federal discretionary Birr 8.0 budget 6.0 Billion Treasury transfer to regions 4.0 2.0 0.0 /05 1998/99 1999/00 2000/01 2001/02 2002/03 2003/04 2004 Source: Data provided by MoFED. The trends in absolute amounts of federal block grant transfers to regions can be seen in Figure 3.1 for the period from 1998/99 to 2004/05. It shows that these transfers are increasing, although not rising as fast as the federal discretionary budget. On the other hand, the faster growth in the federal discretionary budget can be attributed to a large increase in funding for Special Purpose Grants, which are included under the federal discretionary spending category even though they are executed at the regional level. On a per capita basis, the block grant allocations to regions in 2005/06 are given in Table 3.2 below. The data show a wide variation in per capita transfers to regions ranging from close to Birr 95 in Amhara, Oromiya and SNNPR; to Birr 603 in Gambella. 20 World Bank Working Paper Table 3.2. A Wide Variation in Per Capita Block Grant Transfers to Regions Per-Capita Per-Capita Per-Capita Budget Block Grant Block Grant Regional from Own Transfer Transfer from Own Revenues and from Treasury and Revenue Treasury-Source Population Treasury External Per Capita Block Grant (millions) Sources Sources 2005/06 Transfers 2005/06 2005/06 (A) 2005/06 (B) (C) (A+C) 2005/06 Tigray 4.28 117 127 70 187 Afar 1.37 214 226 29 243 Amhara 18.87 81 93 20 101 Oromyia 26.18 89 95 24 113 Somali 4.27 111 117 14 125 Benshangul-G 0.62 335 353 55 390 SNNP 14.70 93 98 18 111 Gambella 0.24 590 603 45 636 Harari 0.19 510 530 111 621 Diredawa 0.39 306 316 84 390 Note: Addis Ababa is excluded from the above table, since it is mostly self-sufficient and does not receive any block grant transfer from Treasury sources. All figures other than in the "Population" column are in Birr. Source: Data provided by MoFED. The three regions receiving the lowest per capita transfers in 2005/06 were also the largest regions in terms of population: Oromiya, Amhara, and SNNPR. This illustrates the built-in "bias" that the "three-parameter" formula has against larger regions, if one looks at per capita allocations. Because of the design of the formula, there is a very tight inverse rela- tionship between a region's population and its per capita transfer (Figure 3.2). The rela- tionship is so tight that factors other than population, such as the development level of a region, appear to have little influence on the final level of per capita transfer of a region. The reasons for this, and some of the problems associated with the formula, are discussed in Appendix A. To illustrate the main point of the previous paragraph, Figure 3.2 depicts the fact that the very small region of Harari received a much higher per-capita block grant transfer in 2005/06 than the much larger region of SNNPR. Yet Harari--a predominantly urban region--is gen- erally considered to be much richer and more developed than SNNPR. According to the 1999/2000 Household Income, Consumption and Expenditure Survey (HICES), for exam- ple, the proportion of people living in absolute poverty and in moderate poverty were much lower in Harari (26 and 42 percent respectively) than in SNNPR (51 and 69 percent). The "three-parameter" formula has a number of problems besides a bias against large and poor regions. In particular, the formula does not take into account the greater needs that some regions may have over others, after controlling for differences in population and development level. Achieving Better Service Delivery Through Decentralization in Ethiopia 21 Figure 3.2. A Very Close Inverse Relationship Between a Region's Population and its Per Capita Transfer from the Federal Government in 2005/06 700 2005/06 600 500 Region, to 400 (Birr) Transfer 300 200 Federal 100 Capita Per 0 0 5 10 15 20 25 30 Average Population of Region, 2005/06 (millions) Source: Data provided by MoFED. For the abovementioned reasons, starting from 2007/08, the Federal government is gradually switching over to a new formula based on a "fiscal equalization" approach. Among other things, this new approach is "needs-based" in the sense that recurrent funding--which accounts for the bulk of total funding--is allocated in higher quantities to regions with higher demand and provision levels of public services. For example, ceteris paribus, regions with higher enrollment rates receive higher per capita recurrent alloca- tions. At the same time, capital funding is equity-oriented in the sense that regions that are more "backward," with lower public service provision levels, are favored in the allocations so that they have the means to advance by accelerating their production of the appropri- ate types of capital stock. In addition, adjustments are made to the total (recurrent plus capital) allocations to adjust for intra-regional differences in revenue-raising ability, and to take into account economies of scale as well as higher unit provision costs in more sparsely populated regions. More details on this approach are given in Appendix B. The "fiscal equalization" approach is being phased in gradually. The allocated Federal transfer to each region in the 2007/08 budget has been derived by calculating the weighted average of: (i) the allocation to that region as computed by the new "fiscal equalization" approach; and (ii) the allocation that would prevail if that region were to obtain the same share of the total transfers (to all regions combined) in 2007/08 as it did in 2006/07. The weights given to these two elements are 25 percent and 75 percent respectively. In short, the 2007/08 allocations give a weight of 25 percent to the new "fiscal equalization" approach and a weight of 75 percent to the old approach that was based on the "three-parameter" formula. 22 World Bank Working Paper Similar calculations will be made to compute the allocations to regions in 2008/09, 2009/10, and 2010/11, except that the weight given to the "fiscal equalization" approach will rise successively from 25 percent in 2007/08 to 50 percent in 2008/09, then to 75 per- cent in 2009/10, and then to 100 percent in 2010/11. From 2010/11 onwards, the alloca- tions will thus be based fully on the "fiscal equalization" approach. Table 3.3 lists the share of total allocated Federal transfers going to each region in 2007/08, under the hypothetical scenario where the allocations are made purely according to the "fiscal equalization" approach. Also listed are the shares going to each region in the 2006/07 budget; these were allocated according to the old "three-parameter"-based approach. Finally, the table also provides figures on the population share of each region. The table illustrates the in-built bias of the old "three-parameter"-based approach against the smaller regions. The 2006/07 budget allocated a share of total transfers to each of the two smallest regions--Gambella and Harari--that was several times larger than their popula- tion share. If the allocations were based entirely on the new "fiscal equalization" approach, then the shares of the total pool of resources going to the smaller regions would be signifi- cantly lower, as the right-hand column of the table shows. However, these shares would still be larger than the population shares of the smaller regions, to an extent justified by economies of scale. Table 3.3. Two Different Approaches for Allocating Federal Resources to Regions Share of Total Share of Total Federal Population Federal Transfers Transfers Going to Each Share of Each Going to Each Region in 2007/08 If New Region in Region in 2006/07 "Fiscal Equalization" 2007/08 Budget Formula Was Fully Applied Tigray 6.04% 7.58% 6.38% Afar 1.96% 3.81% 2.51% Amhara 26.59% 23.16% 26.49% Oromyia 36.76% 32.25% 33.72% Somali 6.02% 6.72% 6.68% Benshangul-Gumuz 0.87% 2.62% 1.42% SNNP 20.62% 19.13% 20.72% Gambella 0.34% 1.92% 0.92% Harari 0.26% 1.24% 0.50% Diredawa 0.54% 1.56% 0.66% Note: The 2006/07 shares are calculated from figures that include transfers from Treasury sources as well as from external loans and assistance (grants). Addis Ababa is excluded from the table, since it is mostly self-sufficient and does not receive any block grant transfer from Treasury sources. Source: MoFED and Ethiopia House of Federation (2007). Fiscal Transfers from Regions to Woredas With the decentralization to local governments in 2002, block grants to local governments were introduced in four regions: Amhara, Oromiya, SNNPR, and Tigray (Figure 3.3), with Achieving Better Service Delivery Through Decentralization in Ethiopia 23 Figure 3.3. Following Decentralization, Transfers from Regional Governments to Woredas and Zones Increased, Except in SNNPR, Where They Were Already High 90% Share 80% as (Incl. 70% 60% 2002/03 Zones) Budget Budgets 50% 2003/04 and 40% 2004/05 Zonal Regional 30% 2005/06 and Woredas20% Total 10% of Woreda 0% Tigray Amhara Oromiya SNNPR Source: Data provided by MoFED. other regions soon following suit. The aim was to replace the incremental budget alloca- tion method with a transparent general purpose grant that gives autonomy to local gov- ernments. Initially after 2002, to make block grant allocations from regions to woredas, regions adopted the "three-parameter" formula used at the time for Federal block grant allocations to the regions, sometimes with some minor modifications. In this formula, block grants were allocated to each woreda on the basis of population, development level, and revenue collection effort. In many regions, urban administrations with woreda status were created as indepen- dent administrative units two years or more after the start of woreda-level decentralization. In the initial stages of decentralization to woredas, most urban administrations did not exist as separate administrative entities; they were a part of larger woredas that incorporated the urban areas (towns) as well as surrounding rural areas. Around 2004 many towns were sep- arated from their "parent" woredas and given the status of urban administrations with sep- arate administrative boundaries and with woreda status. They are often subject to different rules than other (rural) woredas, as discussed below. While some regions have continued to use the "three-parameter" formula for making block grant allocations to rural woredas,2 others have experimented with different approaches. One advantage of Ethiopia's decentralized system is that regions are allowed to choose their own approach for making block grant allocations, as long as this is done in a non- discretionary, formula-based manner. 2. For most regions (with exceptions such as SNNPR), the discussion in this section applies to rural woredas only, and not to urban administrations. See below for a discussion of transfers to urban administrations. 24 World Bank Working Paper SNNPR pioneered the "unit cost" approach for making block grant transfers to woredas. It used the "three-parameter" approach in 2002/03, the first year of decentralization to the woreda-level, but began using the "unit cost" approach from 2003/04 onwards. This is pri- marily a "needs-based" approach, and is similar in some ways to the "fiscal equalization" approach, which is one of the distribution methods currently under consideration for the Federal block allocations to the regions. As in the case of the "fiscal equalization" approach, SNNPR's "unit cost" approach allocates recurrent funding--which accounts for the bulk of total funding--to woredas in a "needs-based" manner, with higher per capita recurrent allo- cations generally going to woredas with higher per capita public service provision levels. More details are given in Box 3.1 and Appendix B, which explains some of the similarities and dif- ferences between SNNPR's "unit cost" approach and the "fiscal equalization" approach. SNNPR's "unit cost" approach has received favorable reviews, and several other regions have subsequently also adopted it, or have been considering adopting it. Oromiya began decentralization to the woreda-level with the "three-parameter" approach in 2002/03, and continued until 2004/05. But from 2005/06 onwards, it decided to adopt the "unit cost" approach. Amhara also has now adopted the "unit cost" approach. Benshangul-Gumuz has been experimenting with a type of approach modeled very closely on the "fiscal equaliza- tion" approach being phased in for Federal block grant allocations. Within SNNPR, the region's innovative switch to the "unit cost" approach in 2003/04 was not favored by all parties. When introduced, the move was justified internally on the grounds that it would be temporary and subject to revision as needed within a few years. The political climate favored a modification of the existing "unit cost" approach in 2006/07, and the region now uses a two-step approach to distribute block grants. First, block grants are allocated by the region to the zones using the "three-parameter" formula (giving weights of 60 percent, 25 percent, and 15 percent to the population, development index, and rev- enue effort of each zone, respectively). Next, each zone allocates block grants to the woredas within that zone, using a formula of its own choosing--with most using the "unit cost" approach to make this allocation. The involvement of the zones in the new system of allo- cating block grants is in recognition of the high degree of ethnic diversity in SNNPR, with each zone representing one ethnic group, or a small number of ethnic groups. In several regions, urban administrations are not treated like other (rural) woredas in the block grant allocation process and are subject to different rules when determining transfer amounts from regions, even though they officially have woreda status. Urban administrations perform "state functions" as well as "municipal functions." The "state functions" are financed by transfers from the regions and are similar to the func- tions performed by the rural woredas. The amount of the transfer in most regions is deter- mined on an ad hoc basis, primarily to finance the recurrent costs of education, health, and other sectors. Recurrent needs are identified based on on-going expenditure obligations and the total transfer amount is negotiated with the region--in contrast to the formula- based approach used for block grant transfers to rural woredas. However, there are differences in approach. In some regions such as SNNPR, the "state functions" of the urban administrations are financed by block grant transfers following a similar formula as for the rural woredas. In SNNPR, however, the transfers for the urban administrations come from a separate pool of funding, which is divided across the 19 dif- ferent urban administrations in the region. In making budget allocations, the region inde- pendently decides how much funding to allocate the pool for the urban administrations, Achieving Better Service Delivery Through Decentralization in Ethiopia 25 Box 3.1. The "Unit Cost" Approach to Block Grant Allocation Concerns with the "three-parameter" formula for allocating block grant resources, includ- ing the inability of many districts to meet their expenditure needs, the poor understanding of the formula by many, and the poor quality of information on which to base allocations, led SNNPR to try out the "unit cost" approach to make allocations, starting from 2003/04. The approach was used for block grant allocations to rural woredas as well as to urban administrations, but allocations for each of these came from a different "pool" (see Appen- dix B for full details on all this). Although many deemed the "unit cost" approach success- ful, a new hybrid approach was introduced in 2006/07, for reasons explained elsewhere in this section. To implement the "unit cost" approach, SNNPR's Bureau of Finance and Economic Develop- ment would initially split the total envelope for block grant allocations to woredas (rural woredas or urban administrations) into two parts: (I) one allocated in a "needs-based" man- ner in line with woredas' recurrent needs (this accounted for the bulk of the total funding); and (II) one allocated in an equity-based manner for woredas' capital investment needs. From (I), the aim was to provide higher per capita funding levels to woredas with well-developed services so that they could have sufficient staff and operational budgets for efficient func- tioning. By contrast, from (II), the aim was to provide higher per capita allocations to woredas with low levels of public service­related infrastructure (for example, number of schools and health posts), enabling them to expand services through accelerated investment. The split between capital and recurrent parts of the envelope was based on historical information and subject to annual revision. Recurrent allocations were based on specific performance targets for each subsector for the following fiscal year (number of primary students to be enrolled, number of health posts to be managed). Starting from 2004/05, these targets were set as part of "performance agree- ments" signed between woredas and the region, whereby the woreda commits to attain the targets (see Box 3.2 on performance agreements in SNNPR). Recurrent funding allocations for each sub-sector, for each woreda, were calculated from these performance targets as well as target values for "cost drivers," or key measures determining unit cost. For exam- ple, each woreda's recurrent allocation for primary education was based on its expected pri- mary enrollment as well as an explicitly stated target value for each of the following cost drivers, among others: (i) the pupil-teacher ratio (PTR) in the woreda; (ii) the average teacher salary in the woreda. The target values for these cost drivers were allowed to differ across woredas, in recognition of past patterns, but they were expected to converge over time to the national average, for each cost driver--leading, over time, to a convergence across woredas of unit costs (for example, of per student primary spending in the case of the pri- mary education subsector). For example, if a woreda had a PTR that is lower than the national average in any one fiscal year, then the target PTR assigned to it in the following fiscal year's recurrent allocation process would be higher than in the current year, but lower than the national average. The aim was to equalize woreda PTRs as well as other cost drivers across woredas over time. Ultimately, each woreda's total allocation was transferred as a block grant and it was free to use this funding as it wished, the only expectation being that it should meet the performance targets (such as for primary enrollment) that it had signed on to in the performance agree- ments. It was not obliged to meet the targets for the cost drivers (except when required by the performance agreement--see Appendix B). However, these cost-driver targets were explicitly stated and in practice each woreda had an incentive to try and meet them, regardless. In the case of its target PTR, for example, a woreda was free to hire more than the number of teach- ers that was implied by its target PTR, but doing so would require it to dig into funding allo- cated by the formula for other uses. This was unlikely to happen since most woredas tended to be strapped for resources. 26 World Bank Working Paper Box 3.2. Performance Agreements in SNNPR From 2003/04 to 2005/06, SNNPR used the "unit cost" approach to make block grant alloca- tions to woredas (see Box 3.1). In 2004/05 and 2005/06, SNNPR signed annual "performance agreements" with woredas. In these agreements, woredas would commit to attaining specified numerical performance targets in each key sector, in the following fiscal year. The calculation of the woreda's block grant allocation for the following fiscal year would, in turn, be based on these agreed-to performance targets. For example, key performance targets in the primary education subsector were primary enroll- ment and the number of new schools to be constructed. Each woreda would agree to attain spe- cific numerical targets for each of these, and would receive funding allocations accordingly. One difficulty encountered by the region was how to ensure that woredas adhered to their performance agreements. One approach used to encourage adherence was to publicize the names of the woredas who had performed poorly, as well as those who had performed well, with regard to the performance agreements. Regional officials also stated that they had been looking into ways of implementing a legal framework to enforce the agreements. SNNPR has now switched to a two-step process where the region allocates block grants to zones, and each zone then allocates block grants to woredas (see above). A new system of per- formance agreements is being established, whereby the region signs performance agreements with the zones, and each zone signs agreements with its woredas. and how much to the pool for the rural woredas. The result is that there may have been some differential treatment of urban administrations as compared to rural woredas, which the region has justified by pointing to urban-rural differences in the prices of goods and services and in the costs of living, among other things. The municipal functions of urban administrations everywhere are financed exclusively from local revenues (through taxes, including land lease income, fees, and user charges). Municipal functions receive no regular financial support from higher levels of government. There are also municipalities without the status of urban administrations, which may raise their own revenues to perform municipal functions. These do not perform "state" func- tions, and do not receive any revenues from higher-level governments; also, there is no tax- sharing program with municipalities. Woredas sometimes receive financial assistance from specific donor or NGO projects; from federal grants, such as the Road Fund, the Food Security Program, and the Public Sector Capacity Building Program; and from contingency assistance, selected cash trans- fers and asset transfers from the region. These fund flows--some of which are off-budget-- are often highly unpredictable and arbitrary. The share of regional budgets allocated to the woredas varies across regions, ranging from 52 to 81 percent (Table 3.4). Allocations also vary across sectors and sub-sectors, and also across and within regions. In 2002/03, 46 percent of the regional budget in Oromiya was allocated to the woredas; 50 percent of the total treasury and regional budget went to regional and zonal sector bureaus (Table 3.5). The situation changed in 2002/03 and 2004/05, when the shares of the budget going to woredas increased to 51 percent. The significant portion of the recurrent budgets allocated to social services (for example, health and education) reflects mainly teacher salaries. Achieving Better Service Delivery Through Decentralization in Ethiopia 27 Box 3.3. Devolution of Power in Theory and in Practice Government decentralization is a trend that is happening around the world. Yet, in many coun- tries, practice often does not follow theory, and there is often less real decentralization on the ground than there should be according to a country's legal framework. This is partly because implementation takes time, partly because there is resistance from certain segments of soci- ety, and sometimes because the capacity to take on new functions or responsibilities may not exist. Although Ethiopia has made great strides in implementing real decentralization of power to lower-level governments, this has occurred in some spheres more than in others. Overall, regions appear to have a large amount of autonomy and decisionmaking power. One manifestation of this is that regions are allowed to determine what formula to use to distrib- ute block grant resources across woredas. The only constraint is that these resources are expected to be allocated in a rules-based and non-discretionary manner, following a prede- termined formula or set of clear, quantitative criteria. As a result, different regions have adopted different approaches for distributing block grant resources, effectively carrying out a "natural experiment" that enables a comparison of the effectiveness of different approaches (see Chapter 4). At the woreda level, many argue that there is much less autonomy and decisionmaking power. This statement is difficult to prove conclusively and opposing arguments can certainly be made. But support for the statement comes from two recent studies: According to a report by Oxfam (see Adal 2005), woreda decisions are often largely based on "sectoral guidelines from regions and party leadership", and plans are based more on "regional targets" than on "community needs". According to a World Bank (2005d) study on empowerment in Ethiopia, local government officials have an "upward view for inputs into decisionmaking and accountability", look- ing "toward their political parties or horizontally toward officials in the state bureaucracy for decisionmaking guidance". According to the World Bank study, this continued reliance on higher levels of government can be explained, in part, by low capacity on the part of woreda officials. If this argument is accepted, then autonomy and decisionmaking powers on the part of woreda administrations can be expected to improve over time, as capacity at the lower levels improves. It should not be forgotten that decentralization to the woreda level only began in 2002/03, and the expe- rience of many countries shows it takes time for true autonomy to take hold at the lower lev- els after decentralization has begun. By contrast, decentralization to the regions began with a big push in 1993/94. Trends in Regional Spending and Regional Revenues The devolution of expenditure responsibilities started in 1993/94 with a big push, when 45 percent of the overall government budget was transferred to the regions. Since then, regional spending has been growing at an annual average of 13 percent (see Table 3.7), with a slowdown in the late 1990's during the conflict. The recurrent budget tripled in ten years, from 2.3 billion to 7.9 billion and now accounts for 72 percent of the regional budget. Regions are major implementers of primary service delivery programs and currently man- age about 80 percent of the pro-poor recurrent budget. Federally mandated programs that are financed from Special Purpose Grants (SPGs)-- including the Food Security, Productive Safety Net, PSCAP and HIV/AIDS programs--are in fact executed by regional administrations, even though they officially appear only in the Fed- eral budget. The SPGs account for about 21 percent of total regional spending, if the latter is 28 World Bank Working Paper Table 3.4. Regional Budgets and Share of Budgets Transferred to Woredas in Four Regions, 2005/06 Regional budget (including woreda- Woreda share of regional level budgets), in million Birr budget, in percent Tigray Amhara Oromiya SNNPR Tigray Amhara Oromiya SNNPR Administration & Gen. Serv. 168.9 447.8 648.6 472.2 67.8 60.1 61.9 81.2 Organ of the State 42.9 93.3 160.6 97.7 71.0 82.8 76.2 83.9 Justice 34.2 48.5 85.6 45.6 39.2 24.4 42.1 87.3 Public Order 36.3 148.5 176.4 100.1 76.7 49.5 42.0 90.5 General Service 29.7 106.5 150.1 129.2 71.7 80.3 66.8 87.4 Economic Services 165.8 573.7 912.2 365.8 51.3 57.9 27.4 68.5 Agriculture & Rural Development 90.2 381.0 489.5 268.7 63.3 75.1 46.4 82.5 Natural Resource 36.2 123.6 143.7 14.7 43.2 28.5 7.1 0.0 o/w Water 30.6 95.4 143.8 12.3 41.4 29.9 7.0 0.0 Construction 23.1 39.3 237.7 19.7 26.5 8.3 2.7 15.4 o/w Road Construction 9.9 27.1 134.8 12.9 58.5 9.4 1.4 0.0 o/w Urban Development 7.5 12.2 98.3 4.7 0.0 5.9 0.0 64.2 Social Services 425.6 953.8 1,534.6 794.8 61.6 75.8 63.0 87.3 Education 303.2 687.6 1,148.9 639.8 66.9 78.4 69.5 89.4 o/w Primary & Junior Education 205.9 363.4 441.5 425.3 86.9 100.0 96.4 99.0 o/w Secondary School Education 27.8 85.2 118.1 48.9 68.1 76.0 44.9 100.0 o/w Technical Education 12.9 36.0 92.9 31.3 19.9 0.0 0.0 0.0 o/w Higher Education 53.4 19.1 52.0 21.3 0.0 0.0 0.0 4.4 Health 107.9 228.9 300.3 149.2 50.0 71.9 40.1 80.3 o/w Prim. Health, Clinic & Hth Stations 50.5 122.0 81.7 97.5 92.8 91.7 68.1 88.1 Achieving o/w Hospitals 40.1 44.1 82.4 13.9 0.0 0.0 9.3 0.0 o/w Malaria Prevention 10.8 11.9 12.8 1.6 38.7 95.8 24.2 0.0 Better o/w Nurses Training 0.0 4.8 1.9 0.0 0.0 0.0 0.0 0.0 Social Affaires 3.1 7.2 0.0 1.5 9.8 25.7 0.0 0.3 Service DPPC 5.1 9.9 22.2 1.4 48.4 27.5 83.2 15.7 Food Security 1.4 0.0 5.3 0.7 38.1 0.0 50.7 0.0 Delivery HIV/AIDS Office 1.3 1.3 1.2 4.5 65.4 60.1 0.0 100.0 Miscellaneous 78.1 159.1 15.0 72.1 58.1 66.0 0.0 75.0 Through Total Recurrent and Capital Budget 840.9 2,135.8 3,116.8 1,710.0 60.5 66.9 52.0 81.1 Source: Data provided by MoFED. Decentralization in Ethiopia 29 30 World Bank Working Paper Table 3.5. Block Grant Allocations in Oromiya, 2002/03­2004/05 Budget Source 2002/03 2003/04 2004/05 Million Birr % Million Birr % Million Birr % Woredas 1,045 46 1,260 51 1,384 51 Regional bureaus 1,138 50 1,176 47 1,285 47 Contingencies 81 4 45 2 38 2 Total 2,264 100 2,482 100 2,707 100 Source: Mussa (2005a). Table 3.6. Block Grant Allocations in SNNPR, 2002/03­2004/05 Level 2002/03 (million Birr) 2003/04 (million Birr) 2004/05 (million Birr) Rural woredas 901 (72%) 883 (64%) 931 (61%) Zones 100 (9%) 98 (7%) 103 (7%) Urban administrations 0.0 64 (5%) 120 (8%) Regional bureaus 237 (19%) 320 (23%) 359 (24%) Total 1,239 1,366 1,514 Source: Mussa (2005b). defined as spending from regional budgets plus SPG-financed (and Federal-mandate) pro- grams that are executed by regional governments.3 The magnitude of resources financed from SPGs in 2005/06 was about Birr 3.7 billion, of which the Food Security SPG accounts for more than half. Total resources handled by regional governments, including the SPG-financed programs, amounted to about Birr 16.9 billion. In the past three years, the total amount of SPG-financed spending increased sub- stantially due to a doubling of Food Security grant funding as well as the implementation of the PSCAP and Productive Safety Net Programme SPGs (see Figure 3.4). Regions are heavily reliant on transfers from the Federal government because of their limited share of total government revenues. Regions on average collect 18 percent of total consolidated government revenue (including Federal revenue), even though regional bud- gets in total are about 38 percent of the total consolidated (including the Federal) govern- ment budget. Figure 3.5 provides more details on the main sources of regional revenue. Regions collect about half of all direct taxes, 15 percent of all indirect taxes, and 22 per- cent of all non-tax revenue. Among the direct taxes, regions collect 100 percent of taxes on agriculture income and capital gains, as well as of rural and urban land use fees. Regarding regional revenues, one positive development has been a decline in the coef- ficient of variation4 of regions' expenditure financed from own sources, from 0.69 in 2001/02 to 0.43 in 2005/06. This means that there is now less variation among the regions in the extent to which their expenditure is financed from own revenue sources. Contribu- tory factors include improved revenue mobilization in poorer regions. 3. These figures and information, as well as some of the other information in this chapter, come from the informative World Bank study "Ethiopia: Review of Public Finance" (World Bank, 2006a). 4. The coefficient of variation is the standard deviation of a distribution divided by the mean of the distribution. It is a commonly used measure of variation. Achieving Better Service Delivery Through Decentralization in Ethiopia 31 Table 3.7. Trends in Regional Expenditure, 1993/4­2005/06 93/94­ 98/99­ 97/98 02/03 Pre-actual Pre-actual (annual (annual 2002/03 2003/04 2004/05 2005/06 (In millions of Birr) average) average) Pre-actual Pre-actual Pre-actual Budget Regional Government Recurrent 2323.6 3726.5 4755.5 5794.8 6154.7 7914.7 Expenditure Capital Expenditure2 1356.3 1673.3 1470.3 1918.2 2672.9 5173.7 Total Expenditure 3679.9 5399.8 6225.8 7713.0 8827.6 13088.4 Consolidated Government Recurrent 5809.4 12091.9 13527.1 11960.8 13035.7 16001.3 Expenditure Recurrent 3520.1 5530.1 7077.5 7730.4 8383.4 11030.4 Expenditure (excl non- discretionary)1 Capital Expenditure2 3457.3 4836.8 6313.4 8271.6 11515.0 18697.3 Total Expenditure 9266.7 16928.6 19840.5 20232.4 24550.7 34698.7 Total Expenditure 6977.3 10366.9 13390.9 16002.0 19898.4 29727.7 (excl. non- discretionary) Regional Expenditures in percent of Conso- lidated Government Expenditures - Including Non- Discretionary Regional Share of 39.7 31.9 31.4 38.1 36.0 37.7 Total Spending Regional Share of 40.0 30.8 35.2 48.4 47.2 49.5 Recurrent Spending Regional Share of 39.2 34.6 23.3 23.2 23.2 27.7 Capital Spending - Excluding Non- Discretionary Regional Share 52.7 52.1 46.5 48.2 44.4 44.0 of Total Spending Regional Share of 66.0 67.4 67.2 75.0 73.4 71.8 Recurrent Spending Regional Share 39.2 34.6 23.3 23.2 23.2 27.7 of Capital Spending Notes: 1. Non-discretionary includes debt service, defense, and relief aid. 2. Pre-actuals for regional capital expenditure do not include external funds due to lack of reporting from regions until accounts are closed. Source: Data provided by MoFED. 32 World Bank Working Paper Figure 3.4. Regional Budgets With and Without Special Purpose Grants (million of Birr) 25000 20000 15000 10000 5000 0 2002/03 2003/04 2004/05 2005/06 Regional Budget SPG transfer Source: Data provided by MoFED. Figure 3.5. Regional Revenue as a Share of General Government Revenue, 2002/03­2004/05 Direct taxes Personal income tax Rental income tax Business profit tax Agri. income tax Capital gain tax Rural land use Urban land lease Dom. indirect taxes Sales/excise taxes Service sales taxes Stamp sales & duty Non-tax revenue Charges & fees Sale of goods & services Reimb. & prop. sales Miscellaneous Revenue Total revenue 0 20 40 60 80 100 Source: Data provided by MoFED. Achieving Better Service Delivery Through Decentralization in Ethiopia 33 Changes in Federal and Regional Spending on Education and Health Real public per capita spending on education more than doubled between 2000 and 2005, from about Birr 30 to about Birr 65 (Figure 3.6). But much of this increase was due to the increase in federal expenditures, which reflected the rapid expansion in spending on tertiary education. Per capita education and health expenditures assigned to the regions and woredas have remained roughly constant since the decentralization to the woreda-level in 2002. Over- all, over the last several years extending before 2002, the nominal rate of growth of total regional (including woreda) spending has been roughly in line with inflation, reflecting the relatively slow growth rate of the main revenue source for regions--Federal block grant trans- fers (Figure 3.1). This appears inadequate, given the regions' (and woredas') responsibility for improving the access and quality of primary-level services. Real education spending by subnational governments (excluding Addis Ababa), which are responsible for providing primary and secondary education, was 79 percent higher in 2004/05 than in 1997/97­1998/99. However, given the 127 percent increase in primary enrollment and the 134 percent increase in secondary enrollment, the increases were insuf- ficient to maintain pupil-teacher or pupil-section (the number of students per class5) ratios, both of which rose. Figure 3.6. Real Per Capita Government Spending on Education Rose After 2000, but Much of the Increase went to Tertiary Education 80.0 70.0 60.0 prices Addis Ababa City Government 50.0 2004 40.0 Subnational Governments Excl. Addis Ababa capita, per 30.0 Federal Birr 20.0 real 10.0 0.0 98 0 6/97 199 1997/ 1998/99 1999/0 2000/01 2001/02 2002/03 2003/04 2004/05 Source: Data provided by MoFED. 5. In Ethiopia as in many countries, students at each grade within a school are divided into one or more classes/sections. A class/section is a group of students who sit together and together undergo instruc- tion in all core subjects. A "class" is not quite the same as a "classroom," although in most cases a school would need to have at least one classroom for each class/section. 34 World Bank Working Paper With resources increasing at a slower pace than enrollment and the real wages of teach- ers rising, the pupil-teacher ratio rose from 42:1 in 1996/97 to 66:1 in 2004/05 at the primary level and from 36:1 to 58:1 at the secondary level. Over the same period, pupil-section ratios rose from 57:1 to 69:1 at the primary level and from 66:1 to 93:1 at the secondary level. These ratios are very high by international standards. In the health sector, real spending at the subnational level over the past several years rose by less than the increase in the population, resulting in a decline in real per capita health spending at this level (Figure 3.7). Real health spending by subnational governments excluding Addis Ababa rose about 13 percent in 2004/05 relative to 1996/97­1998/99. Over the same period, the population increased by about 17 percent, resulting in a slight decline in real per capita health spending at this level, from Birr 10.1 to Birr 9.7. This slight decline in real per-capita health spending on the part of subnational gov- ernments needs to be evaluated alongside the very significant growth in public health spending on basic services that has taken place via channels outside of subnational gov- ernment budgets. The amount of off-budget spending on basic health services (for exam- ple, local provision by NGOs) has grown substantially over the last few years, as has the quantity of goods that are procured and paid for centrally but distributed at the local level. Nevertheless, as in the case of education, the trend in real per-capita subnational spending on health appears to reflect an inadequate growth rate in Federal block grant transfers to the regions, in the face of growing regional commitments to basic service delivery. This is likely to change over the next few years, if the Federal government fulfils its commitments to substantially increase block grant transfers to the regions (see Chapter 5). Figure 3.7. Overall Real Per Capita Spending on Health Increased Between 2000/01 and 2004/05, but Real Per Capita Subnational Government Spending Declined 20.0 18.0 escir 16.0 p 14.0 Addis Ababa City 2004,ati 12.0 Government Subnational Governments 10.0 pacre Excl. Addis Ababa 8.0 Federal prri 6.0 Blaer 4.0 2.0 0.0 6/97 98 98/99 9/00 /01 /03 3/04 04/05 199 1997/ 19 199 2000 2001/02 2002 200 20 Source: Data provided by MoFED. Achieving Better Service Delivery Through Decentralization in Ethiopia 35 Performance of Regional Budgets Regional government revenue tends to be unpredictable with a volatile performance record across years (see Table 3.8). On average over the past three years, regions' actual revenue fell short of plans by more than ten percent. In FY2004/05, regions managed to collect only 71 percent of their planned revenue. In particular in Addis Ababa, non-tax (mainly lease) revenue fell short by more than 50 percent. Excluding Addis Ababa, the rest of the regions managed to collect 81 percent of planned revenue in 2004/05. It is not surprising that recurrent spending is more predictable than capital spending. The former comprises predominantly wage and salaries and actual regional spending has been close to budgeted levels, even better than the general government. In the three of the four years under consideration, actual regional spending was more than 90 percent of their budgeted lev- els. However, under-spending of the capital budget at the regional level is a major concern, especially for funding from non-Treasury sources (project-specific donor funding). Table 3.8. Aggregate Fiscal Performance (in percent of budget) 2001/02 2002/03 2003/04 2004/05 Revenue (domestic) General government 82 88 96 87 Regional government 76 85 83 71 Recurrent expenditure General government 85 104 86 90 Regional government 98 87 93 95 Capital expenditure General government 79 92 92 90 Regional government 61 49 55 52 Treasury source only 95 93 75 65 (regional governments) Note: The "General Government" refers to the combined Federal and regional governments. Source: Data from MoFED. The Structure of Woreda-Level Budgets Responsibility for social service delivery has been devolved to woredas with education (pre- dominantly primary) being the single most devolved category of sector expenditure. In the 2005/06 budgets for the four main decentralizing regions (see Figure 3.8), woredas admin- ister 81 percent of the regional education recurrent budget, while the ratio for health is about 65 percent. The pace of devolution varies across sectors, as the figure shows. On average, capital spending accounts for only 16 percent of the district-level budgets. Shares of the capital budgets accounted for by woredas by sector vary from 11 percent for education to 27 percent for the economic sectors, for the four main decentralizing regions (FDRs--Amhara, Oromiya, SNNPR, and Tigray). The woreda-level budget is dominated by wage and salary costs, which account for about 78 percent of total recurrent spending in the FDRs. Woredas are facing increasing 36 World Bank Working Paper Figure 3.8. Woredas' Share of Regional Recurrent Budgets for Amhara, Oromiya, SNNPR, and Tigray (the Four Main Decentralizing Regions) Together 2004/05 2005/06 90 80 70 60 50 40 30 20 10 0 Total Admin. & Economic Agri. & Social Education Health Recurrent Gen. Serv. Service Rural Dev't Service Source: Data provided by MOFED. pressure from the fast-growing wage bill. Indeed, in the FDRs, the budget for salary and allowances rose by 33 percent in 2005/06 compared to an increase in the overall woreda budget of 17 percent. As a result, district-level budgets for operational (non-personnel) spending fell in absolute terms (by Birr 50 million) as well as in relative terms. The decline in operating budgets and the pattern of accelerating wage bills is not uni- form across sectors. The education sector has been the most affected with the wage share rising from 70 percent in 2004/05 to 92 percent in 2005/06 in the FDRs. For the rest of the sectors, the increase in wage share ranges from 2 to 3 percent only. CHAPTER 4 How Did Decentralization to Woreda Level Affect the Delivery of Social Services? T his chapter examines the implementation of decentralization to the woreda level in two of the largest regions in Ethiopia, SSNPR and Oromiya. The two regions' 39 million people represent roughly 55 percent of the country's population. One benefit of decentralization and devolution of power to the regions is that the regions are allowed to determine the formula and criteria used to distribute block grant resources to woredas, with the constraint that these resources are expected to be distributed in a rules- based, objective manner. As a result, different regions have adopted different approaches for distributing block grant resources, effectively carrying out a "natural experiment" that enables a comparison of the effectiveness of different approaches. One of the reasons for choosing SNNPR and Oromiya for the analysis is to examine the impact of two contrasting methods of block grants allocations--until the fiscal year 2005/06 the former used the "unit cost" approach while the latter used the original "three-parameter" approach until the year 2004/05. (See Chapter 3 for a description of both approaches.6) The report team obtained longitudinal data on budgets and expenditures as well as outcomes and inputs in education and health at the woreda level. Data were obtained before decentralization (2001) and after decentralization (2004) for all 123 woredas (including urban administrations) in SNNPR and 69 of the 197 woredas (including urban administra- tions) in Oromiya.7 Indicators of socioeconomic situation were created, based on population, 6. As Chapter 3 notes, several regions treat urban administrations differently from other woredas. SNNPR's "unit cost" approach was used to determine block grant allocations to urban administrations, but these came from a different "pool" than that used to distribute block grant resources to rural woredas. In Oromiya, however, urban administrations have been treated in the same manner as other woredas with allocations for both coming from the same pool. 7. These figures correspond to the situation in 2004. The total number of woredas is now more than 123 in SNNPR and more than 197 in Oromiya, due to the creation of new woredas (via the splitting of existing woredas)--see below for more on this. 37 38 World Bank Working Paper agricultural and animal production, pastoral areas, food security status, physical infrastruc- ture, area, urbanization, and distance to main urban centers. School-level scores on Grade 8 examinations were obtained for the two periods. Because of the limitations in the data, the analysis could be conducted more extensively for SNNPR than for Oromiya, and more for education rather than health services. The expenditure data obtained represent expenditures from Treasury sources only, and appearing in woreda budgets. The following are not captured by the present dataset: (i) woreda on-budget spending funded directly from foreign assistance and loans; (ii) off-budget spending; (iii) spending by zones and by the regional government itself; and (iv) woreda spending funded from special-purpose Federal grants or SPGs (for exam- ple, PSCAP, Food Security Program and Productive Safety Nets Program grants). These data gaps are recognized as limitations to the present analysis. The omission of the SPGs from the expenditure data is not likely to significantly affect the results of the current analysis. In 2001, total SPG funding was almost zero. In 2004, before a large acceleration in the growth of SPG funding, the main SPGs were PSCAP (mainly for capacity building at the district level); anti-HIV/AIDS grants; and Food Security Program (FSP) grants.8 Of these, the latter--totaling 1 billion Birr--accounted for the bulk of the total, and funded mainly agricultural packages for households as well as resettlements with a small amountgoingtootherrelatedactivities;itdidnotprovidecashtransfers.Thus,thereisunlikely to have been much "leakage" of FSP or other SPG funds to education or health expenditures.9 One difficulty with quantitative comparisons of the pre- and post-decentralization periods of analysis (2001 and 2004) is that there were many cases where a woreda split into two or even three--thus creating one or even two new localities with separate administra- tive boundaries. Each of these new localities became a new woreda--either a rural woreda or an urban administration with woreda status; urban administrations did not exist before decentralization to the woreda level. For consistency, all comparisons between 2001 and 2004 of expenditure or outcome measures were done using either: (i) pre-decentralization woreda boundaries for both periods of time, or (ii) post-decentralization boundaries for both time periods. The former approach, for example, treats all newly created woredas in 2004 as a part of the original woredas to which they belonged, rather than as separate units. More details on this and other technical aspects are given in Appendix E. Increases in Woreda-Level Spending Following Decentralization Did decentralization10 change expenditure patterns at the woreda level? Woreda expendi- tures capture virtually all government spending on primary service delivery (primary edu- cation and health) both before and after decentralization. Spending from donor and NGO sources which are not flowing through the woreda budget are not available and thus not 8. The Productive Safety Net Program (PSNP) is currently a major SPG program, but it began only in 2005. 9. Some "leakage" could still have occurred. For some household beneficiaries of the FSP, FSP expen- diture on in-kind items for the household could have displaced private household funds (from household income) that would otherwise have gone towards purchase of these same items. Some private household funds would have been freed up, and some of this could have gone towards education or health expenditures. 10. Only recurrent expenditures are examined, as capital spending did not go through woreda bud- gets before decentralization (it was funded through zonal budgets). Achieving Better Service Delivery Through Decentralization in Ethiopia 39 captured in this analysis. Before decentralization, spending on primary service delivery went through woreda budgets even though woredas had no control over the amount or composition of this spending. Recurrent expenditure by woreda administrations rose significantly with decentralization (Figure 4.1). Mean recurrent expenditure nearly doubled between 2001 and 2004, in both SNNPR and Oromiya. A means test of expenditure per capita suggests that decentralization significantly changed public expenditure at the woreda level, as a result of: (i) the transfer to woredas of some functional responsibilities; (ii) much higher woreda administrative spending to enable woreda administrations to handle their new responsibilities; and (iii) an increase in Figure 4.1. Expenditures of Woredas Increased After Decentralization a) Expenditures 20,000,000 Pre-D 16,000,000 Post-D 12,000,000 8,000,000 4,000,000 0 0 20 40 60 80 100 120 Number of Woredas b) Expediture per capita Expenditure per capita by Woreda 200 180 Pre-D 160 Post-D 140 120 100 80 60 40 20 0 0 20 40 60 80 100 120 Number of Woredas Note: "Pre-decentralization boundaries" (see the introduction to this chapter) were used to calculate all expenditure figures for the above and other diagrams. Source of data: SNNPR Bureau of Finance and Economic Development (BOFED). 40 World Bank Working Paper per capita recurrent spending on some items that had been going through woreda budgets before decentralization. In terms of sectoral allocation, woredas have consistently allocated the largest shares of their budgets to education and health, which accounted for almost 55 percent of their budgets in 2004 (41 percent to education and 12 percent to health on average in SNNPR; 45 percent to education and 9 percent to health on average in Oromiya). In SNNPR woredas, per capita spending rose in all sectors between 2001 and 2004 (Figure 4.2). Recurrent spending on pri- mary education, which represented more than 85 percent of recurrent spending on educa- tion before decentralization, rose to 90 percent of education expenditures, suggesting the importance of basic education in resource allocation (Figure 4.3). Per capita health spending rose following decentralization, although it remains low. Decentralization Appears to Have Improved the Distribution of Expenditures Across Woredas with Worse-off Woredas Benefiting the Most The increase in spending on education and health was accompanied by a narrowing of the dif- ferences in per capita spending across woredas in both SNNPR and Oromiya (Tables 4.1 and 4.2). At the same time, the differences in per student education spending across woredas also narrowed in both regions. The decline in the coefficient of variation for most of the spend- Figure 4.2. Recurrent Woreda-Level Spending Rose in All Sectors in SNNPR Following Decentralization (n=104 woredas) Recurrent Expenditure by Sector Other Sectors 2001 Administration 2004 Agriculture and Rural Health Education 0 5 10 15 20 25 real Birr per capita, 2004 prices Note: All calculations regarding expenditures, in all the diagrams, were done using "pre-decentralization boundaries" (see the introduction to this chapter). This explains why there are 104 woredas in total rather than 123 (which would be the case if "post-decentralization" boundaries were used). Source of data: SNNPR BOFED. Achieving Better Service Delivery Through Decentralization in Ethiopia 41 Figure 4.3. The Share of Total Recurrent Spending for Primary and Secondary Education in SNNPR Rose Following Decentralization (n 104 woredas) Others 2001 2004 Secondary Primary 0% 20% 40% 60% 80% 100% Percent of Total Note: Calculations assume "pre-decentralization" woreda boundaries. Source of data: SNNPR BOFED. ing measures suggests that per capita health spending and per student education spending in many lagging woredas are catching up with that in better off woredas. (Unlike the standard deviation or variance, the coefficient of variation is comparable across distributions and is thus the appropriate variability measure to use here. See Appendix E for more details.) There is generally more variability in the spending measures among the SNNPR woredas than in the Oromiya woredas in the sample. This reflects the fact that there is much less vari- ability within the Oromiya sample than within the SNNPR group of woredas to begin with, in characteristics such as extent of food insecurity and whether or not a woreda is pastoral. Indeed, unlike the SNNPR sample, the Oromiya sample does not have any pastoral woredas. The narrowing in the dispersion of various spending measures after decentralization (Tables 4.1 and 4.2) is likely attributable to the equalization effects that can be attained through formula-based allocation of resources through block grants. They mirror the experience of countries around the world, that formula-based allocations are typically fairer than arbitrary or ad hoc allocations. As described earlier, until the fiscal year 2005/06, SNNPR uses the "unit cost" approach for block grant allocations while Oromiya used the "three-parameter" approach until 2004/05. In the longer run, the "unit cost" approach tends to equalize per student spending across woredas (see Box 3.1), while the "three-parameter" approach tends to equalize per capita spending across woredas (because the bulk of the fund- ing is allocated in proportion to woreda population). In the first two years since decentral- ization to woredas was introduced in 2003, the data show that both these types of equalization have occurred for several education-spending measures in SNNPR and Oromiya. This implies that the distribution of funding across woredas in the two regions was highly unequal to begin with, before decentralization when the allocations were made on an ad hoc basis. The differences between the unit cost approach and the three-parameter approach are explained in detail in Appendix A and Appendix B. 42 World Bank Working Paper Table 4.1. Woreda-Level Spending on Education and Health in SNNPR, by Category of Spending, 2001 and 2004 (n 104 woredas, 2004 Birr prices) Mean Mean Coefficient of Coefficient of Spending Spending Variation Variation Spending Category 2001 2004 20012 20042 Recurrent primary education 121 143 0.52 0.41 per primary student3 Total recurrent education 140 157 0.44 0.36 per student4 Recurrent primary education 16.0 20.8 0.56 0.41 per capita Total recurrent education 19.3 23.7 0.61 0.54 per capita Recurrent health per capita 3.58 4.25 1.08 0.63 Notes: 1. Calculations assume "pre-decentralization" woreda boundaries. 2. All pastoral woredas were excluded from the calculations for the coefficients of variation, since they are very different from other woredas, for the following reasons: (i) they have very low popula- tion densities and therefore high unit costs (low pupil-section and pupil-teacher ratios); (ii) teachers in pastoral woredas earn a 40 percent increment over their base salaries as a hardship allowance; (iii) there is a special fund providing additional resources just to the pastoral woredas. When the coefficients of variation were calculated for pastoral woredas, they fell between 2001 and 2004 for all spending measures, as is the case for the non-pastoral woredas (reported in the table). 3. Only includes students in public primary schools. 4. This measure incorporates both primary and lower secondary school recurrent spending (i.e. for Grades 9 and 10); its denominator is the total number of primary and lower secondary school students. Note that this measure is not entirely comparable between 2001 and 2004, because there has not always been consistency on the question of which government level (woreda, zone or region) funds lower secondary schooling in a locality. The difficulty arises in part because there are many "combination schools" with both lower and upper secondary schooling, and often with teachers teaching at both levels. Woredas are supposed to be responsible in principle for lower secondary schooling, and the region for upper secondary schooling. But at the "combination schools," it is difficult to ensure that each level of government funds (from its own budget) only the type of schooling for which it is officially responsible. By contrast, funding for all public primary schools consistently passed through woreda budgets only in 2001 as well as in 2004. Source of data: SNNPR BOFED and SNNPR Bureau of Education. Increase in Spending in Remote, Pastoral, and Food-Insecure Woredas DecentralizationinSNNPRdisproportionatelyfavoredremote(morethan50kilometers from a zonal capital), food-insecure, and pastoral woredas (Tables 4.3 to 4.5). The aggregate budget for education, for example, rose 36 percent in remote woredas and just 3 percent in woredas located less than 50 kilometers from a zonal capital. A similar trend occurred for health. Aggregate spending on education and health by pastoral woredas rose by 57­75 percent between 2001 and 2004, while spending by non-pastoral woredas rose by just 9­16 percent. Remote woredas also gained disproportionately from decentralization, with per student expenditures for primary education rising about 20 percent as opposed to just 16 percent for non-remote woredas. The percentage change in per student secondary expenditure exceeded 95 percent in remote and food-insecure woredas, while it was just 49 percent in non-remote woredas and just 4 percent in food-secure woredas. Achieving Better Service Delivery Through Decentralization in Ethiopia 43 Table 4.2. Woreda-Level Per Capita Spending on Education and Health in Oromiya, by Category of Spending, 2001 and 2004 (n 69 woredas, 2004 Birr prices) Mean Mean Coefficient of Coefficient of Spending Spending Variation Variation Spending Category 2001 2004 2001 2004 Recurrent primary education 132 149 0.47 0.30 per primary student Total recurrent education 154 163 0.52 0.32 per student Recurrent primary education 14.1 20.3 0.53 0.41 per capita Total recurrent education 17.2 23.3 0.63 0.45 per capita Recurrent health per capita 2.77 3.28 1.34 0.58 Notes: Calculations assume "pre-decentralization" woreda boundaries. The sample for Oromiya did not have any pastoral woredas (see footnote 2 of previous table). See also other notes of previous table, which also apply here. Source of data: Oromiya BOFED and Oromiya Bureau of Education. Table 4.3. Woreda-Level Spending on Education and Health in SNNPR, by Type of Woreda, 2001 and 2004 (n 104 woredas) (real thousands of Birr, 2004 prices) Education Health 2001 Total 2004 Total 2001 Total 2004 Total Budget Budget Budget Budget (average (average (average (average across all across all Percent across all across all Percent Type of Woreda woredas) woredas) Change woredas) woredas) Change Remote (more than 1,897 2,576 36 314 398 27 50 kilometers from a zonal capital) Nonremote 3,681 3,778 3 501 512 2 (less than 50 kilometers from a zonal capital) Food insecure 3,073 3,760 22 453 503 11 Food secure 2,051 2,283 11 346 393 14 Pastoral 897 1,574 75 253 398 57 Nonpastoral 2,774 3,227 16 419 458 9 Note: Calculations assume "pre-decentralization" woreda boundaries Source of data: SNNPR BOFED. 44 World Bank Working Paper Table 4.4. Recurrent Expenditure Per Primary Student in SNNPR, 2001 and 2004 (n 104 woredas) (real Birr, 2004 prices) Type of Woreda 2001 2004 Percentage Change Remote (more than 50 kilometers 114 137 20 from a zonal capital) Nonremote (less than 50 kilometers 129 150 16 from a zonal capital) Food insecure 120 140 16 Food secure 121 146 21 Pastoral 236 286 21 Nonpastoral 119 141 18 Note: Calculations assume "pre-decentralization" woreda boundaries. Source of data: SNNPR BOFED and SNNPR Bureau of Education. Table 4.5. Recurrent Expenditure Per Secondary Student in SNNPR, 2001 and 2004 (n 104 woredas) (real Birr, 2004 prices) Type of Woreda 2001 2004 Percentage Change Remote (more than 50 kilometers 103 202 96 from a zonal capital) Nonremote (less than 50 kilometers 134 199 49 from a zonal capital) Food insecure 96 210 119 Food secure 168 175 4 Note: Calculations assume "pre-decentralization" woreda boundaries. Source of data: SNNPR BOFED and SNNPR Bureau of Education. After Woreda-Level Decentralization Began, Some Evidence of Lagging Woredas Partly Catching Up in Social Service Delivery Gross primary education enrollment rates (GERs) have soared in both SNNPR and Oromiya within the three years since decentralization to woredas began in 2002/03, continuing a nationwide trend that started several years earlier. Woreda-level data show an improvement in GERs over this three-year period of about 10 percent in SNNPR and 15 percent in the Oromiya subsample of woredas (Tables 4.6 and 4.7). This rate of improvement is high even by international standards. Over the same period, educational quality indicators largely declined, with the exception of repetition rates in SNNPR which fell from 11.9 percent to 10 percent (Tables 4.6 and 4.7). Repetition rates rose in Oromiya, as did pupil-teacher ratios and pupil-section ratios in both regions. This reflects the rapidly rising demand for primary schooling observed throughout Ethiopia, with supply often not being able to keep pace. More interestingly, the coefficient of variation (Tables 4.6 and 4.7) in education outcomes was, in most cases, smaller in 2004 (after decentralization) compared to 2001 (before decen- Achieving Better Service Delivery Through Decentralization in Ethiopia 45 Table 4.6. Education Outcomes in SNNPR, 2001 and 2004 (n 123 woredas) Coefficient of Coefficient of Item Mean 2001 Mean 2004 Variation, 2001 Variation, 2004 Primary gross enrollment ratio 62.6% 72.9% 0.57 0.47 Proportion passing Grade 8 71.3% 74.2% 0.23 0.19 exams Primary repetition rate 11.9% 10.0% 0.48 0.48 Primary pupil-teacher ratio 63.4 67.2 0.45 0.36 Primary pupil-section ratio 76.8 80.2 0.26 0.26 Primary teacher-section ratio 1.21 1.19 0.34 0.18 Note: The calculations assume "post-decentralization" boundaries and each locality covered by an urban administration--each urban district--is thus treated as a separate unit. Source of data: SNNPR Bureau of Education. Table 4.7. Primary Education Outcomes in Oromiya, 2001 and 2004 (n 69 woredas) Coefficient of Coefficient of Item Mean 2001 Mean 2004 Variation, 2001 Variation, 2004 Gross enrollment rate 53.6% 68.7% 0.27 0.22 Repetition rate 3.6% 4.9% 0.71 0.44 Pupil-teacher ratio 59.35 72.50 0.30 0.31 Pupil-section ratio 64.14 73.24 0.18 0.17 Teacher-section ratio 1.13 1.06 0.20 0.17 Note: The above figures were calculated for a subsample of 69 out of 197 woredas in Oromiya region (using "post-decentralization boundaries--see the introduction to this chapter for an explanation of this, and see also the footnote for the previous table"). The means are close to the overall regional means from official statistics. Source: Oromiya Bureau of Education. tralization). Unlike the standard deviation or variance, the coefficient of variation is compa- rable across distributions and is thus the appropriate variability measure to use here. The result indicates that education outcomes across woredas have become more equal, with some evi- dence of catching up by lagging woredas. This mirrors similar findings for spending measures, as reported above for both SNNPR (Table 4.1) and Oromiya (Table 4.2). The "catch-up" trend for lagging woredas is, however, less evident in Oromiya compared to SNNPR. Indeed, in some cases the equalization of outcomes across woredas may be due more to a deterioration in woredas that are ahead, than to "catching up" by lagging woredas. This is especially true where there has been a sharp deterioration of indicators, as in the case of repetition rates and pupil-teacher ratios in Oromiya. The above findings are confirmed by an analysis of educational outcomes by sub- groupings of woredas for SNNPR. This exercise was not possible for Oromiya due to data limitations. The analysis for SNNPR shows that between 2001 and 2004--i.e., before and after decentralization--there has been significant "catch-up" by lagging woredas in sev- eral educational outcomes (Figures 4.3 to 4.5). These improvements mirror the changes in 46 World Bank Working Paper spending patterns analyzed above, which also disproportionately favored lagging woredas in SNNPR (Tables 4.3 to 4.5). Pastoral woredas in SNNPR seem to have benefited most since the start of the woreda- level decentralization, perhaps because they started from a low base (Figures 4.4 to 4.6). These woredas enjoyed the greatest improvement in primary repetition rates and Grade 8 examination pass rates. Pastoral woredas still fall behind in overall enrollment and have a long way to go before they reach the 65 to 80 percent gross enrollment rate achieved by the rest of the region, but they have been improving more rapidly than other woredas. Again, this mirrors the finding above that the pastoral woredas benefited disproportionately in terms of increases in spending. These findings are further confirmed by regression analysis (see Appendix C), which also shows that there has been equalization of outcomes at the school level within woredas in SNNPR. More specifically, since the start of decentralization lagging schools have been "catch- ing up" in educational outcome indicators such as repetition rates and Grade 8 examination pass rates. The figures show that remote and food-insecure woredas have been catching up in terms of gross enrollment ratios, but Grade 8 exam pass rates have not changed. These find- ings suggest that many other factors affect student performance, such as family socioeconomic status and other demand-side factors. Teacher Redeployment Across Woredas and Within Woredas One of the most important features of decentralization was the devolution of management of resources, including human resources, to the woredas, which are responsible for deploying Figure 4.4. Gross Enrollment Rates have Improved in All Categories of Woredas in SNNPR Since Decentralization (n 123 woredas) 140% 120% 100% Rate 80% 2001 2004 60% Enrollment 40% Gross 20% 0% Towns Pastoral Food-Secure Food- Remote Nonremote (Urban Woredas Woredas Insecure Other Than Other Than Adminis- Other Than Woredas Pastoral Pastoral and trations) Pastoral Other Than Towns Pastoral Note: Calculations assume "post-decentralization" boundaries. Source of data: SNNPR Bureau of Education. Achieving Better Service Delivery Through Decentralization in Ethiopia 47 Figure 4.5. Grade 8 Pass Rates in SNNPR Have Increased Since Decentralization, with Improvement in Pastoral Areas Greater than in Urban Areas (n 123 woredas) 90% 80% 8 70% Grade 60% 50% 2001 Passing 2004 40% 30% 20% Percentage 10% 0% Towns Pastoral Food-Secure Food- Remote Nonremote (Urban Woredas Woredas Insecure Other Than Other Than Adminis- Other Than Woredas Pastoral Pastoral and trations) Pastoral Other Than Towns Pastoral Source of data: SNNPR Bureau of Education. Figure 4.6. Repetition Rates have Fallen Sharply in Urban Areas and Pastoral Areas in SNNPR Since Decentralization (n 123 woredas) 20% 18% 16% 14% Rate 12% 2001 10% 2004 8% Repetition 6% 4% 2% 0% Towns (Urban Pastoral Food-Secure Food-Insecure Remote Other Nonremote Adminis- Woredas Woredas Other Woredas Other Than Pastoral Other Than trations) Than Pastoral Than Pastoral Pastoral and Towns Source of data: SNNPR Bureau of Education. 48 World Bank Working Paper Figure 4.7. Teachers have been Redeployed from Urban to Non-Urban Woredas Since Decentralization (n 123 woredas) 7.0 6.0 2001 5.0 2004 Residents 4.0 1,000 3.0 per 2.0 1.0 Teachers 0.0 Towns (Urban Rural Woredas Nonremote Remote Rural Pastoral Adminis- Surrounding Rural Woredas Woredas Other Woredas trations) Towns Other Than Than Pastoral Those Woredas Surrounding Towns and Pastoral Woredas Source of data: SNNPR Bureau of Education. primary school teachers. Since decentralization, there has been a substantial redeployment of teachers in SNNPR, with the number of teachers in rural areas and nearby towns rising and the number of teachers in urban areas falling (Figure 4.7). A separate study (World Bank 2005a) observed that before decentralization, Ethiopia (except for Tigray) had a high degree of randomness in teacher deployment in government primary schools. The redeployment of teachers from less-crowded schools in urban areas to overcrowded schools in rural areas increased pupil-teacher and reduced teacher-section ratios (ratio of number of teachers to number of sections) substantially in towns in SNNPR (Figure 4.8 to 4.9), with both ratios approaching the average levels in other woreda categories. The pupil-teacher ratio in towns (urban administrations) rose from 38 in 2001 to 54 in 2004, while the teacher-section ratio in towns fell from 2.0 in 2001 to 1.4 in 2004. This signifi- cant change in the teacher-section ratio in towns is a strong indicator of increased efficiency in teacher use since it indicates a rise in the average teaching load in towns to levels close to those in rural areas.11 Despite the redeployment, pupil-teacher ratios in rural remote areas rose, because the increase in supply was insufficient to meet the large increase in demand. An analysis of the available data show clearly that these trends began in 2002 and not before, around the time that decentralization to the woreda level started. 11. In theory, a lower teacher-section ratio means one or more of the following: (i) a higher average teaching load; (ii) less average class time by students; (iii) increase in (or start in) rotation by some teach- ers between different classes, so that they handle two or more classes at the same time. Indications are that the first factor was the key one at play in the towns in SNNPR. Achieving Better Service Delivery Through Decentralization in Ethiopia 49 Figure 4.8. Teacher-Section Ratios Tended to Equalize Since Decentralization Began in SNNPR (n 123 woredas) 2.5 Sections of 2.0 Number to 1.5 2001 2004 1.0 Teachers of 0.5 Number of 0.0 Ratio Towns (Urban Pastoral Food-Secure Food- Remote Nonremote Adminis- Woredas Woredas Insecure Other Than Other Than trations) Other Than Woredas Pastoral Pastoral and Pastoral Other Than Towns Pastoral Source of data: SNNPR Bureau of Education. Figure 4.9. The Gap in Pupil-Teacher Ratios Across Woredas has Narrowed Since Decentralization (n 123 woredas) 90 80 70 Ratio 60 50 2001 40 2004 30 Pupil-Teacher 20 10 0 Towns (Urban Pastoral Food-Secure Food- Remote Other Nonremote Adminis- Woredas Woredas Insecure Than Pastoral Other Than trations) Other Than Woredas Pastoral and Pastoral Other Than Towns Pastoral Source of data: SNNPR Bureau of Education. 50 World Bank Working Paper It is unusual to see huge teacher reallocation happen in a short period of time, and with visible impact on pupil-teacher ratios across different areas. In SNNPR, several factors are likely to have contributed to this change: Firstly, starting at around the time of decentralization in 2002/03, the Regional Council with full support of the Regional Education Bureau started an explicit pol- icy of reallocation of teachers from towns (with very low pupil-teacher ratios) to sur- rounding rural localities. An appropriate directive was issued, and zonal councils were also supportive. Secondly, the introduction of the "unit cost" block grant formula in SNNPR facil- itated the equalization of pupil-teacher ratios. This is an explicit aim of the formula and is in line with the key overall objective of equalizing per student costs across woredas (see Box 3.1). Allocations to urban administrations were based on the same "unit cost" approach as allocations to rural woredas, and SNNPR differs in this sense from some other regions which allocate funding to urban administrations in a discretionary manner.12 Thirdly, the "unit cost" approach had the effect of creating higher awareness and better planning by woredas and the region in the allocation of teachers, thereby facil- itating the process of equalization of pupil-teacher ratios. Lastly, equalization of pupil-teacher ratios was achieved not just by reallocation of existing teachers away from towns (urban administrations), but also by appropri- ate placement of newly hired teachers in areas with high pupil-teacher ratios. The incentives given by the "unit cost" approach are especially important in this regard. In Oromiya, this study finds no indication of teacher redeployment after decentralization. In fact Table 4.7 shows that the coefficient of variation of the distribution of woreda pupil-teacher ratios actually increased slightly after decentralization to the woredas began. Figure 4.10 shows that pupil-teacher ratios (PTRs) rose substantially in both high-PTR woredas and in low-PTR woredas between 2001 and 2004. This finding from the Oromiya data illustrates that the introduction of the simple "three- parameter" block grant formula may not be sufficient to improve allocation of inputs such as teachers across woredas. On the other hand, SNNPR managed to substantially improve the efficiency of teacher allocation through the introduction of a block grant formula (using the unit cost approach) more suited for the purpose, and which was facilitated by an explicit sector policy on teacher reallocation supported at the highest levels in the region. The "three-parameter" approach is especially unsuitable for equalizing pupil-teacher ratios given that the enrollment rate is not an explicit factor in the formula--higher enroll- ments do not mean higher block grant allocations. Redeployment of teachers took place not only between woredas but also within woredas in SNNPR. As a result, differences in pupil-teacher ratios across schools within woredas nar- rowed (Figure 4.11). The pupil-teacher ratio (PTR) of schools below the median PTR within 12. A caveat here, as mentioned in Chapter 3, is that all urban administrations in SNNPR got alloca- tions from a common pool, and all rural woredas also got allocations from a common pool--but these two pools were different. In making budget allocations, the region would decide independently how much funding to allocate to each of these pools. Regardless, for many "cost drivers" (see Box 3.1) such as the average pupil-teacher ratio, the aim was still to equalize them across and between rural woredas and towns (urban administrations). Achieving Better Service Delivery Through Decentralization in Ethiopia 51 Figure 4.10. No Reallocation of Teachers or Narrowing of Gaps in Pupil-Teacher Ratios Appears to Have Taken Place in Oromiya Since Decentralization (n 69 woredas) 100 90 Ratio 80 70 60 50 2001 2004 Pupil-Teacher 40 30 20 Overall 10 0 Woredas Below Median of Woredas Above Median of Distribution of Woreda PTRs Distribution of Woreda PTRs Source of data: Oromiya Bureau of Education. each woreda rose from 58 to 64 since the start of decentralization, while that of schools above the median within each woreda fell, from 107 to 78. The available data show that this "nar- rowing" trend of within-woreda PTRs was not observed before decentralization; between 2000 and 2001, there was a rise in the PTR of schools below the median within-district PTR (from 56 to 58), as well as of schools above the median within-district PTR (from 104 to 107). These findings are confirmed by the regression analysis presented in Appendix C. In all the regressions, the coefficient of the main independent variable of interest--the original 2001 value of the relevant educational indicator--is negative and statistically significant, in both the school-level and the woreda-level regressions. The regression results with the school pupil-teacher ratio as the dependent variable was based on analyses of 2,244 schools in the 123 woredas in the region. They show that the differences in pupil-section ratios narrowed-- that is, teachers were redeployed from schools with higher pupil-teacher ratio to schools with lower pupil-teacher ratios in order to reduce the imbalance in teacher allocation. What this implies is that while the overall number of teachers may not have increased enough to accommodate the rapid growth in enrollment, the redeployment of teachers from less crowded schools to more crowded schools in SNNPR has likely improved the efficiency of the teaching force. This may be one factor contributing to the decline in repetition rates observed in most woredas in the region. Regression Analysis: Linking Expenditures with Outcomes Summarizing the evidence presented so far in this chapter, with decentralization to woredas in 2002/03, the gap in education outcomes between the lagging woredas and the better-off 52 World Bank Working Paper Figure 4.11. Differences in Pupil-Teacher Ratios Across Schools Within Woredas in SNNPR have Narrowed Since Decentralization (n 123 woredas) 120 100 Ratio 80 60 Pupil-Teacher 40 2001 2004 20 Overall 0 Schools Below Median of Schools Above Median of Distribution of School PTRs Distribution of School PTRs Within Each Woreda Within Each Woreda Source of data: SNNPR Bureau of Education. woredas in SNNPR tended to narrow. There has been substantial equalization of pupil- teacher ratios across woredas, achieved partly by extensive teacher reallocation from urban to rural areas. At the same time, the differences in per capita and per student education spend- ing across woredas have been narrowing, likely attributable mainly to the introduction of a formula-driven approach to determining block grant allocations. The picture in Oromiya is somewhat different. There has also been substantial equal- ization of per capita and per student education spending across woredas but this has not been accompanied by an actual improvement of allocation across woredas of the main edu- cational budgetary input--the teachers. Furthermore, there is evidence of "catching up" of lagging woredas in gross enrollment rates but not in other indicators such as repetition rates. The analysis in this section focuses on the evidence presented for SNNPR. Is there a clear relationship between spending and outcomes? What could explain the narrowing of per capita spending and educational outcomes across woredas? To explore this question, woreda-level regressions analyzing the effects of expenditure on education and health outcomes, controlling for variables related to infrastructure, level of development, population size, and location, were conducted. In addition, similar regressions were conducted with measures of teacher quantity--such as the pupil-teacher ratio and the number of teachers per 1,000 population--as right-hand side variables, to directly examine the impact of these variables on educational outcomes. The results, presented in Appendix D, show that after controlling for other factors, increases in the levels of expenditure on education had a strong effect on increases in enroll- ment rates and in Grade 8 pass rates. A strong positive relationship is observed between changes in the level of teacher inputs and changes in enrollment rates as well as in Grade 8 pass rates. In an earlier study (World Bank 2005a), teacher inputs in Ethiopia were found to a Achieving Better Service Delivery Through Decentralization in Ethiopia 53 strong impact on Grade 8 pass rates at the level of the individual student. That study did not find any relationship between woreda-level spending and Grade 8 pass rates--although it should be noted it was based on cross section data, compared to the present study which uses longitudinal data. Primary gross enrollment rates rose with education expenditure, as did Grade 8 pass rates. Grade 8 pass rates also improved with per student primary expenditure. Primary gross enroll- ment rates rose with the number of teachers per 1,000 population while Grade 8 pass rates were found to rise as the pupil-teacher ratio declined. The results related to Grade 8 pass rates were obtained with samples restricted to poorly performing woredas--those with especially low Grade 8 pass rates to begin with (see Appendix D for more details). The regression results add weight to the hypothesis that there is a relationship between the observed "catching up" of lagging woredas in educational outcomes on the one hand, and the equalization of spending measures as well as of pupil-teacher ratios on the other. The latter is likely attributable in large part to the introduction of formula-driven block grant allocations since the start of decentralization, especially the use of the "unit cost" approach. In turn, one can conclude that there is evidence that decentralization in SNNPR influenced the improvements observed with the lagging woredas. The regression analysis also confirms many of the earlier results from the descrip- tive analysis on the relative changes in education outcomes and in spending measures for specific groups of woredas. For example, the regressions show that increases in gross enrollment rates are positively related to the distance of a woreda from the zonal capital, mirroring the earlier finding of especially large increases in enrollment rates in remote woredas since decentralization started (see Figure 4.4). The changes in education expen- diture measures in remote woredas (see Tables 4.3 to 4.5) add weight to the hypothesis that there is a relationship between the large increases in spending and the large rises in enrollment rates in remote woredas. The analysis in this chapter recognizes a number of limitations. Many other factors that cannot be measured have important effects on the outcomes of interest. Education outcomes are affected by sector policies such as reallocation of teachers, as well as by the contributions of communities and leadership in the region and the woreda. Communi- ties have been contributing material, funding, and support to schools through the con- struction and management of schools through PTAs. Likewise, the effects of household socio-economic factors cannot be easily accounted for in an analysis that uses the woreda as the unit of investigation. Decentralization's High Potential to Improve Service Delivery Based on the available evidence, decentralization has already contributed to positive changes in service delivery in education and health in SNNPR. Total enrollment rates and Grade 8 pass rates in poorly performing woredas rose with education expenditure growth, and both also rose with increases in teacher inputs. Repetition rates in poorly performing woredas fell with increases in teacher inputs. The data show that since decentralization started, there has been an equalization of spending measures and pupil-teacher ratios across woredas on the one hand, and an equalization of educational outcomes--with "catching up" of lagging woredas--on the other. The regression results imply that there is a relationship between these two. 54 World Bank Working Paper These results imply also that there is a discernible level effect of increasing expenditure on improving enrollment. Many woredas increased their education budgets, which appears to have increased enrollment. As seen in Chapter 5, the results were achieved in an environment with constraints in both capacity and monitoring. This indicates that, even with improving capacity, better outcomes can be achieved. The implementation of decentralization with formula-driven block grant allocations in SNNPR has clearly shifted the bias in expenditures toward the lagging and disadvan- taged woredas. Lagging food-insecure, pastoral, and remote woredas have begun to catch up in terms of outcomes and have benefited disproportionately in terms of per capita spending. One possible explanation for the equalizing effects of decentralization observed in SNNPR is the use of the unit cost approach to determine how to allocate block grants from regions across woredas. In contrast to the three-parameter formula used in most other regions, the unit cost approach helped to reduce the gap in per student teacher allocations and pupil- teacher ratios across woredas. It also provides incentives for increasing efficiency in the deploy- ment of teachers. Because teachers account for a large proportion of the recurrent education spending of woredas, in an environment of limited budgets, sustainable improvements in service delivery cannot occur without improving the deployment of teachers from less crowded classrooms to more crowded classrooms. Equally important for the teacher reallocations in SNNPR was the implementation of a specific sector policy favoring teacher reallocation from towns (urban administrations) to rural woredas. This policy emanated from the highest levels--the Regional Council--and had the full support of the Regional Educational Bureau as well as the zonal councils. In Oromiya, there has been equalization of per capita and per student education spend- ing across woredas, but no equalization of pupil-teacher ratios nor any real teacher reallo- cation. The latter is likely attributable to: (i) the use of the three-parameter block grant formula that does not purposely target the equalization of pupil-teacher ratios; as well as (ii) the lack of a specific policy favoring teacher reallocation. Gross enrollment rates have risen and their distribution has become more equal across woredas. However, repetition rates in Oromiya have also risen, and this appears to have occurred disproportionately in woredas with low repetition rates to begin with. Thus the pic- ture in terms of improvement of educational outcomes is mixed, just as the picture in terms of educational inputs--such as teachers and spending measures--is mixed. Certainly, the path taken by SNNPR is clearly a more favorable one. It could be argued that the equalization across districts of spending and outcomes, observed in SNNPR and to a more limited extent in Oromiya, is largely attributable to a more equal distribution of resources across woredas, and could have occurred without an increase in accountability and autonomy at the local level. Indeed, some argue that decentralization to the district level is still very new, and that--following the experience of other countries-- it would take time for significant increases to be seen in true local-level accountability/ autonomy (see Box 3.3). Even if local-level accountability and autonomy are still limited, however, decentralization has clearly been very beneficial in obliging the use of a formula- based approach to distributing resources across woredas. This appears to have resulted in measurable benefits, especially in SNNPR. In theory, a formula-based approach to resource allocation across woredas could have been possible even without district-level decentral- ization. Yet, decentralization made this approach feasible in practice. Achieving Better Service Delivery Through Decentralization in Ethiopia 55 It should also be pointed out that the observed narrowing of school pupil-teacher ratios within woredas, which began after district-level decentralization in SNNPR, cannot be explained by the introduction of a formula-based approach to resource allocation across woredas. Here, there is a stronger rationale for an explanation pointing to increased autonomy and accountability at the local level, due to decentralization. Decentralization appears to have already improved service delivery in poorer locations, at least in SNNPR. These changes are occurring even where capacity has not reached any- where near the woredas' potential. With further improvement in woreda capacity, and with the use of appropriate formula for block grant allocations as well as appropriate sector poli- cies, decentralization may contribute significantly to achieving the targets for improving health and education in Ethiopia. CHAPTER 5 Making Decentralization Work Overcoming Constraints in Decentralized Service Delivery D ecentralization has shown the potential for facilitating the delivery of basic social services through modest increases in the control of resources, increased spend- ing for social services, and service delivery being more responsive and more flex- ible. The analysis in the previous chapter has shown that the beneficial effects can be fully realized with adequate levels of fiscal transfers to woredas, and with the use of appropri- ate methods of distributing these resources across woredas. In addition, optimal results can be realized if sector policies are aligned with the goals of decentralization as amply demonstrated in SNNPR. Critical Constraints Despite these improvements, however, critical constraints continue to stymie the efficient and effective delivery of services. Failure to address these constraints could negate the achievements made to date. This chapter identifies some of these constraints and proposes options for overcoming them. Inadequate Resources at the Woreda Level Lower tiers of government in Ethiopia have substantial responsibility for meeting tar- gets in education, health, and water and sanitation. But financing from block grants has not been sufficient to allow them to provide these basic services. In 2004/05 about 26 per- cent of the federal budget was transferred to the regions as unconditional grants (35 per- cent when special purpose grants are included) (Figure 5.1). These transfers provided 57 58 World Bank Working Paper Figure 5.1. Per Capita Block Grants to the Regions Averaged just Birr 82 Per Capita in 2004/05 Total transfer Total block grants Per capita transfer Per capita transfer (millions of Birr) (Birr) 2,000 500 1,800 450 1,600 400 1,400 350 1,200 300 1,000 250 800 200 600 150 400 100 200 50 0 0 la Afar bel arari Tigray H Amhara Oromia Somali SNNP reDawa Gam Di Benshangul/Gumuz Region Source: Data provided by MoFED. the average region with just 82 Birr per capita to undertake all the development activi- ties within its mandate. Adding regions' own resources to the pool increases spending to just 108 Birr per capita. The constraints imposed by inadequate financing are even tighter at the woreda level (Figure 5.2). In the Four Main Decentralizing Regions, the regions transfer between 51 per- cent (Oromiya) and 78 percent (SNNP) of the unconditional block grants they receive from the federal government. The average woreda in SNNPR and Oromiya allocates 45 percent of its recurrent resources to education, 90 percent of which goes to salaries (Dom 2004). During the initial period of decentralization to the woredas, 34 of 123 woredas in SSNPR lacked adequate budgets. According to woreda-level officials, this lack of funds made it difficult for them to expand or improve the quality of services. Woredas finan- cial resources have been improving, as transfers from regions have increased. Moreover, not all woredas face the same budget predicament. Nevertheless, the lack of adequate budgets remains a serious constraint. Although the overall national expenditure envelop has increased substantially over the past seven to eight years, grants from the federal government to subnational governments Achieving Better Service Delivery Through Decentralization in Ethiopia 59 Figure 5.2. Inadequate Financing Represents an even Greater Constraint at the Woreda Level Total transfers Per capita transfer Total block grant transfer Per capita transfer (millions of Birr) (Birr) 1000 90 900 80 800 70 700 60 600 50 500 40 400 30 300 20 200 100 10 0 0 Tigray Amhara Oromia SNNP Regions Note: Data are for 2004/05. Source: Data Provided by MoFED. through block grant transfers have not risen as rapidly, although this pattern is now chang- ing (see below). As a proportion of federal revenues, block grant transfers to regions fell, from about 73 percent in 2000/01 to about 46 percent in 2004/05 (Table 5.1).13 This appears to indicate a need for much larger block grant transfers, to enable subnational governments to fulfill their commitments to the delivery of basic services at the lower levels in the face of rapidly increasing demands. However, the figures do not take into account the sharp increase in Federal funding for Special Purpose Grants which appear only on the Federal budget but are executed at the regional level (see Figure 5.4). It should also be noted that a significant portion of the increase in Federal spending has gone towards basic services; for example, the new rural electrification program took up about 6 percent of total Federal revenue (including grants) in 2005/06. In any case, the Federal government has made commitments to increase block grant transfers to the regions substantially over the next few years, with support from the multi- 13. Debt repayments, interest payments, external assistance and defense spending were subtracted from total Federal revenue (including grants) before making these percentage calculations (see Table 5.1). 60 World Bank Working Paper donor Protecting Basic Services operation which began in June 2006. This trend has already begun. Total block grant transfers to the regions in 2006/07 were 9.4 billion Birr, as compared to 5.5 billion Birr in 2004/05. According to the latest draft MEFF, block grant transfers will rise by 4 billion Birr a year for the next three years, reaching 21.56 billion Birr in 2009/10. If these commitments are fulfilled, the trend shown in Table 5.1 where the vertical allocation of resources has been shifting against the regions would be firmly reversed. Table 5.1. Block Grant Transfers as a Proportion of Federal Revenues, 1996/97­2005/06 (millions of Birr) 1997/98 2000/01 2001/02 2002/03 2003/04 2004/05 2005/06 Federal Revenues 7,649 11,132 9,845 12,808 14,834 16,576 18,865 Including Grants Federal Revenues 4,459 4,874 5,172 6,358 10,603 11,924 14,216 net of Interest Payments, Debt Repayments, External Assistance and Defense (A) Block Grant 3,233 3,548 3,996 4,523 5,021 5,498 7,071 Transfers (B) B as percentage of A 73% 73% 77% 71% 47% 46% 50% Note: All figures reflect actual revenues and expenditures. Source: Data provided by MoFED. Difficulty in Planning Due to Inadequate Information In principle, budget planning involves elements of a "bottom-up" participatory approach, in which communities and woreda offices of education, health, and water and sanitation set priorities and goals, which are then approved by the woreda council before submission to the regional government. In practice, most regions do not know the size of their grant before the end of the third quarter or the beginning of the fourth quarter. As a result, most the planning is mechanical and often turns into a wish list exercise. The poor information base and uncertainty regarding the budget envelop also undermines the credibility of long- term planning, even though most frontline service providers do produce three-year devel- opment plans, and it undermines budgeting for capital expenditures. If basic services in education, health, and water and sanitation are to improve, the resources available to woredas must be adequate and predictable (World Bank 2004a). Need to Improve Clarity in Expenditure Assignments and Responsibilities The expenditure assignments between levels of government have not yet been fully defined, although much progress has been made in this regard. It is not realistic to expect all issues of expenditure assignment and the corresponding responsibilities to have been sorted out less than five years after the start of decentralization to the woreda level. However, expen- Achieving Better Service Delivery Through Decentralization in Ethiopia 61 diture assignments must be clearer and serve as the basis for the allocation of resources across tiers of government. For accountability mechanisms to be effective and service delivery to improve, delega- tion of responsibilities must be matched by adequate finances especially at the local gov- ernment level, and expenditure assignments of the different levels of government need to be fully and clearly defined. This does not yet happen to a sufficient degree in Ethiopia. Expenditure assignments are in many ways the starting point for fiscal frameworks, because they represent the functional parameters for financial flows. Lack of clear assignments can lead to duplicated effort and lack of coordination. It also makes it difficult to assign budgets in order to meet functional objectives. In the regions that have instituted woreda-level decentralization, practices regarding assigning responsibilities to various levels of government have been evolving. The regions, zones, and woredas are clear about the broad areas of recurrent expenditure managed by each level of government in education and health. Responsibilities in other areas are less clear. In some regions, for example, regional, zonal and local authorities are all responsible for "prevention and control of HIV/AIDS", with no indication of exactly what is expected of them, how they are supposed to complement each other, or how much each level is supposed to contribute. Not only are these distinctions weak or absent, there is no clear link to fiscal capacity and responsibility. Functional and fiscal assignments need to be clearly tied and assignments costed as accurately as possible in order to eliminate any disjuncture between functional accountability and fiscal arrangements. Within each of the broadly assigned areas of responsibility, the exact roles of each level of government and institution need to be made clearer (Box 5.1). While woredas' func- tional assignments indicate that they are responsible for recruitment and retrenchment, for example, in practice the zones have often overruled their decisions about retrenchment (Heymans and Mussa 2004), although there are indications that this may be changing at least in some regions. Substantial progress is being made, however, in defining expenditure assignments of different levels of government, especially in the four larger regions which began decen- tralization first (Amhara, Oromiya, SNNPR and Tigray). These regions revised their constitutions in 2001; these constitutions are the basis for decentralization of power to woredas. The next step would be for each region to adopt a legal framework clearly spelling out expenditure assignments and the division of responsibilities between different levels of government. To ensure that this would happen and as a guide to the regions, a draft prototype legal framework was developed in March 2006, as a model on which regional legal frameworks were supposed to be based. Among other things, the draft framework provided details and regulations on the structure and composition of different tiers of administration; assign- ments at the regional, woreda and kebele level for different service delivery functions; con- duct of official meetings; powers and functions of different actors; financial provisions; inter governmental relationships; human resource and management issues; and miscella- neous provisions. Following this, each of the regions is now taking measures to finalize its legal frame- work for decentralization to woredas. Tigray has made significant progress in provid- ing such a framework (see Box 5.1). Amhara has produced a draft regulation laying out 62 World Bank Working Paper a legal framework, which has been submitted to the regional administration for further comments and approval. In SNNP, a workshop on this legal framework has been con- ducted by the Bureau of Capacity Building; selected woreda staff were invited to give feedback and comments. In Oromiya, work is under way on this legal framework, and a consultant has been recruited for this purpose. Furthermore, all of the four larger regions have adopted municipal proclamations which lay out responsibilities and assignments of municipalities. Even without an appropriate legal framework for decentralization to woredas, regions usually have written functional assignments for different sectors, but these are not always operational. Hiring of certain types of personnel, for instance, is the responsibility of the woredas, but in some instances zones hire staff when the woredas cannot find candidates in their area. In principle, firing and transferring staff are the responsibilities of the woredas, but their decisions are sometimes overruled by the regional bureaus or zones, without con- sultation. The region sometimes arbitrarily makes decisions about transferring health workers and teachers who serve in rural woredas, which creates tension between the woredas and the region. According to some woreda officials, the unplanned transfer of frontline workers to the woreda has caused budget deficits and made it impossible for the woreda to pay salaries. But all of this can be expected to improve soon, as long as regions finalize the establishment of the legal frameworks for decentralization to woredas, and then enforce these frameworks. Box 5.1. Improving the Legal Framework for Decentralization in Tigray The Tigray region has made significant progress in providing a legal framework for decentral- ization. In early 2006 the Regional Council proclaimed two acts, "Strengthening Decentraliza- tion" and "Organization and Tasks and Responsibilities of Woreda Executive Bodies". These acts go a long way toward clarifying the regional legal and policy framework for decentralization. They formalize existing practice, legally establishing the woreda block grant mechanism and linking woreda functional assignments to resources. They provide further clarity on other points, such as woreda functional assignments; woreda participatory planning and its link with the regional strategic planning policies; and regional support to and oversight of woredas. They also break new ground in a number of important dimensions, such as provid- ing a framework for assigning revenues to woredas and allowing for flexibility in woredas' administrative structures and staffing policies. Particularly difficult is the issue of who is responsible for the capital budget. Regions increasingly pass on responsibility for capital investment to the woredas. In FY 2004/2005, for instance, about 49 percent of education capital investment by the regions was bud- geted at the woreda level (Dom 2004). Giving woredas some responsibility for planning and financing development makes sense, but in practice they often control little of the capital budget. Reluctance to devolve more control to the woredas is rooted in several ten- sions in the system. One is the thorny issue of how best to finance capital spending. Should it come from the block grant? Should there be a special purpose grant for capital spend- ing? If woredas are required to finance capital expenditures from the block grant, will there be any assurance that funds will not be diverted to more pressing recurrent expenditure Achieving Better Service Delivery Through Decentralization in Ethiopia 63 needs, given the unconditional nature of the block grant? Another difficult issue is the fear by federal and regional governments that planning and managing complex capital invest- ments may be beyond the capacity of many woredas. Not surprisingly these tensions remain unresolved and have tended to favor the creation of special purpose grants, even though the signals in the government's financial reporting system (most directly from the execution of capital budgets) seem to suggest that caution is needed in providing special purpose grants. A very visible example of the ambiguity in responsibility for capital investment man- agement is secondary education (Dom 2004). Woredas are responsible for the first cycle of secondary education (Grades 9­10). While this makes sense in large and populated woredas, where enrollment potential is high, in smaller woredas demand may be insufficient. Allow- ing woredas to make their own decisions about secondary education may lead to the prolif- eration of small and costly schools. Despite this problem, all regions have encouraged woredas to increase their capital expenditures. But pressure on woredas to meet their recurrent expenditure obligations has limited their scope for capital investments. The woredas are also reluctant to commit funds to capital expenses early on, because they are legally prohibited from changing such allo- cations back to the recurrent side of the budget. The current system risks perpetuating this trend in that the assumption is that most capital investment falls within the ambit of fed- eral and regional responsibilities. Furthermore, many woredas have been limited by the lack of adequate staff to deal with the challenge of capital budgeting and capital investment implementation. A more complex capital budget allocation issue is how externally funded programs are treated in the envelop of resources allocated by block grants to woredas. At the Federal level, the full amount expected to be spent by each region from external project-specific (and on-budget) funding is subtracted or "offset" from the region's block grant allocation. Thus, each additional Birr that a region gets in project-specific funding results in a one Birr reduction in its block grant allocation, that is, there is 100 percent offsetting. At the regional level, there is variation in the extent of "offsetting" of project-specific funding from block grant allocations to woredas. In some cases, there is full (100 percent) offsetting, while in other cases there is partial or no offsetting from woreda block grant allocations. Offsetting can be problematic because it results, in effect, in high fungibility of external aid; provid- ing project-specific funding becomes almost like providing untied funding. In addition, offsetting reduces the incentive for regions or woredas to accept external project-specific funding; given the choice between project-specific funding and an equivalent amount of untied block grant funding, many regions/woredas would accept the latter. This is because project-specific funding flows are often unpredictable, and the projects can often impose a significant bureaucratic burden. Declining Share of Spending on Capital Expenditures by Subnational Governments As decentralization to woredas went into full swing in 2002/3 in the four largest regions, there was a strong belief that many woredas were not investing much in capital and that the newly found autonomy would lead local governments to spend on recurrent expen- ditures, in particular salaries. In fact, capital spending by subnational governments 64 World Bank Working Paper Figure 5.3. Capital Spending by Subnational Governments as a Share of Total Spending has been Declining, Especially in the Social Sectors 100.0% Education 90.0% 80.0% Health 70.0% 60.0% Agriculture and 50.0% Natural Resources (incl. Water) 40.0% Labour and Social Welfare 30.0% 20.0% All Social Sectors 10.0% 0.0% Economic Development /97 2/03 Sectors 1996 1997/981998/991999/002000/012001/02200 2003/042004/05 Source of data: Data provided by MoFED. (regional bureaus, zones and woredas) as a share of total spending has been declining steadily over the past several years, a trend that started before decentralization to the woredas began (Figure 5.3). The declining shares are particularly evident in the social sec- tors, including education and health. The evidence suggests that low capital spending by woredas may often be due more to a lack of resources than a suboptimal desire to spend on capital relative to recurrent spend- ing. Regional guidelines as to appropriate capital budget shares for the woredas may be use- ful, but they are unlikely to work unless adequate resources are transferred. Concerned about suboptimal capital spending for budget items handled by the woredas, in 2004 the Oromiya regional government issued a guideline that woredas should spend 8­10 percent of their budgets on capital items. Despite this, in 2005/06 just 7 percent of total woredas budgets in Oromiya were allocated to capital items. Yet in the same year, woredas allocated 16 percent of their total budgets to capital items in Tigray. This higher share of capital spending in Tigray was probably facilitated in turn by the much higher per capita block grant allocation to woredas in Tigray than in Oromiya. As a result of the latter, 2005/06, the per capita total budget (recurrent plus capital) of woredas in Tigray was 113 Birr, as opposed to 63 Birr in Oromiya. Overall, the capital share of total subnational spending has been falling for several years before 2002/03, but has remained roughly constant since then. Similarly, real per capita Achieving Better Service Delivery Through Decentralization in Ethiopia65 capital subnational spending has also remained more or less constant since 2002/03. For the social sectors, the capital share of total subnational government spending (excluding Addis Ababa) has remained at about 17 percent since 2002/03 (Figure 5.3). For the eco- nomic development sectors, the capital share rose from 75.3 percent in 2002/03 to 82.3 per- cent in 2004/05 before falling to 76.5 percent in 2005/06. Despite a fall until 2002/03 and then stagnation in real per capita capital spending by subnational governments, there appears to have been acceleration in physical investment in capital stock related to primary service delivery over the same period. In the education sector, between 2002/03 and 2004/05, the number of new primary schools created per 10 million people was 5.9, up from 4.2 in the two years preceding decentralization to the woreda level (Figure 5.4). More new secondary schools and new health posts were created per 1 million people between 2002/03 and 2004/05 than between 1999/00 and 2001/02. Figure 5.4. Since Decentralization to the Woredas, There Has Been an Acceleration in the Production of Physical Capital Stock Related to Education and Health Service Delivery 16 14 over period 1999/00 to 12 2001/02 10 over period 2002/03 to 2004/05 8 6 4 2 0 Increase in Increase in Increase in Average number Average number number of number of number of of sections per of sections per primary schools secondary health posts per primary school secondary per 10 million schools per 10 million school population 1 million population population Source: Official Data from Ministries of Education and Health. Furthermore, there is some evidence that the productivity of these capital investments has risen since decentralization. One measure of the productivity of the capital stock in the education sector is the average number of sections per school. This figure rose from 9.2 in 1999/2000­2001/02 to 10.3 in 2002/03­2004/05. One explanation for the rise is the increase in the number of double-shift schools during the period. There is one caveat to note, how- ever. Many of the new schools and health posts have been built by communities with limited government assistance, and there is some concern that some of these may be of substandard quality. 66 World Bank Working Paper In sum, declines in capital spending in the education or health sector--in real per capita terms or as a share of total spending--is not necessarily an indication of inadequate capital spending. Nor does it imply a misalignment between capital and recurrent spending. The acceleration in the production of new schools and health posts is likely to be a result of various factors, including shifts to low-cost construction and substantial increases in community contributions. The Ministry of Education has estimated annual community contributions in the sector to be about Birr 11 per capita in 2005/06. One should also note that the above analysis is based on data on spending that occurs through subnational government budgets. It does not capture the large amount of off- budget capital spending on local service delivery. This includes not just the community contributions that have just been referred to, but also the increasing amount of local spend- ing undertaken by NGOs, especially in the health sector. One problem is that there is evidence that many woredas have limited capacity to exe- cute capital spending, especially where donor funds are involved. Building capacity at the woreda level to execute capital investment is critical. Need for Mechanisms to Address the Differences in Revenue Generating Capacity of Woredas The woreda budget consists of transfers from the region and own revenues. The current prac- tice is to fix a target revenue collection for each woreda and use that as a basis for determin- ing the level of block grant transfers to the woreda. More specifically, a predetermined formula is used to determine the block grant allocation to each woreda. From this allocation, the targeted revenue collection for that woreda is subtracted to derive the transfer from the region to that woreda. This practice has been challenged by woredas, on the grounds that the targets are often not feasible or that the woredas may not be able to meet the targets in some circumstances. A recent review found that the targets were negotiated in Oromiya, but SNNPR's policy was to generally avoid topping up the transfer element of the block grant in favor of defaulting woredas. Instead, the region generally lent funds needed to make up for shortfalls. This practice negatively affected service delivery in Wonego woreda in SNNPR, where the non-salary component of the education recurrent budget was cut by 50 percent in 2003/04, hampering the delivery of primary school materials (Dom 2004). In another woreda, Alaba, there were shortfalls in teacher salaries, training, textbooks, some office supplies, and supervision budgets. The regional governments do not yet have clear mechanisms for sharing revenues with the woredas. Woredas collect revenue on behalf of the regions, and the collected rev- enue is part of each woreda's block grant. In Tigray an act on revenue-sharing mecha- nisms between the region and the woredas has been approved by the regional council and has been officially issued in Negaret Gazeta as a proclamation. Other regions have attempted to draft acts on revenue-sharing mechanisms, but these drafts have not yet become official documents. Revenue sources are broadly categorized as direct taxes, indirect taxes, and non-tax revenues (more detailed breakdowns could not be obtained). Direct taxes include personal income taxes, rental income taxes, and business income taxes; indirect taxes include taxes on works contracts, stamp duties, income from work permits, court fees, and business reg- istration fees. Direct taxes are the major source of revenue, followed by non-tax revenues. Achieving Better Service Delivery Through Decentralization in Ethiopia 67 Woredas obtain their revenue from land, agricultural, business, and income taxes, as well as from license fees and fines for minor crimes. City governments have similar sources of revenue, but they collect revenue from about 20 taxes, including property taxes, market transaction taxes, and charges on services rendered, such as water provision. The revenue collected by municipalities is used to carry out municipal functions; in the case of short- falls, they do not receive support from the region. The revenue collected is fully managed by the municipalities, which do not transfer resources to the treasury. No revenue-sharing mechanism is in place in SNNPR, but a study is underway on rev- enue potential and capacity that is expected to lead to revenue-sharing legislation for the region. The woreda administrations vary greatly in institutional and fiscal capacity, ranging from those with substantial own-revenue bases (mostly in peri-urban areas such as Awassa Zuria) to ones without much of a fiscal base and very low capacity (such as Boritcha, a recently created woreda in a rural area that is highly food-insecure). In 2002/03, Awassa Zuria received 56 percent of its budget from the regional government and raised 44 percent of its budget through local revenue. In contrast, Boritcha's own-revenue contribution was 10 percent; 90 percent of its budget came from the regional block grant. In 2004, Oromiya attempted to introduce a revenue-sharing mechanism between the regional government and woreda administrations. It identified revenue as a key issue, acknowledging that some local areas make considerable contributions not only to local rev- enue but also to the regional treasury funds. For example, in Aweday, in eastern Hararghe, the center of Ethiopia's tchat (a tobacco-like cash crop) industry, substantial collections could increase overall regional revenues. The region seems to have made some progress in allowing local structures to retain at least part of their surpluses (the amount of own revenue collected in excess of the expected revenue collection, if any), but this approach has not been fully consolidated and surpluses are still largely offset against the block grant. The region does recognize that local use of sur- pluses without being offset by the block grant can become a strong incentive for local rev- enue generation. The prospects for local governments improving their revenue performance in the short term are constrained by the limited capacity of local government officials to effectively administer a revenue system; the prevalence of weak administration procedures; poor record keeping, and inefficient and irregular updating of tax rolls; inefficient collec- tion procedures; and inadequate information about tax bases. Even where local governments can keep all or most of their own revenue surpluses, the treatment of own revenues in most regions is problematic. In effect, if a woreda takes steps to improve its own revenue collection, it is raising expectations of its own revenue collection in future years, thereby reducing its block grant transfer from the region in future years. In this respect, the "fiscal equalization" approach under consideration for Federal block grant transfers to the regions is superior since woredas would, in effect, be allowed to "keep" all their own revenues; the approach would not reduce woredas' incen- tive to collect own revenues. (See Appendix B for more on this; Chapter 3 describes briefly the "fiscal equalization" approach.) Need to Improve Monitoring and Enforcement Except for urban municipalities, which are expected to be self-financing, most woredas finance the services for which they are responsible through the unconditional block 68 World Bank Working Paper grant. The unconditional nature of the grant has the desirable feature of granting the woredas significant discretion to pursue local priorities in budget allocation. Until now, the federal and regional governments relied on informal understandings and consen- sus on national goals to guide spending priorities. However, within this flexibility enjoyed by local governments, in most regions there is no institutionalized system to encourage good performance. Currently, SNNPR is the only region that has initiated a process designed to pro- mote good performance; it has introduced "performance agreements" (see Box 3.2). The region is also planning to introduce a system that will require each woreda administra- tion to development a three-year sectoral public expenditure program (PEP), which would link consolidated (regional and woreda) sector financial requirements to agreed upon/expected sector performance. The PEPs would be embedded within a regional Medium-Term Expenditure and Financial Framework, and the first year of the PEP would be the basis for preparation of the following year's annual fiscal plan and budget. This approach would provide woredas with firmer "pre-budget ceilings" for the first step of the woreda budgeting cycle. The development of a binding three-year sectoral PEP by all regions is the missing link between regional and woreda budget making, and consolidated sector envelop projections at the national level. This is clearly a long-term agenda. It will take SNNPR time to develop and implement the PEP approach and even more time for other regions to see the benefits of such a system. Need to Boost Administrative Capacity at the Woreda Level Compounding the problem of inadequate financing is weak local government capacity to deliver services. Woreda capacity has been improving, as more and more staff at the zonal and regional levels are redeployed to the woredas and training programs for local level staff are completed. Despite progress, however, lack of sufficient capacity remains as one of the key obstacles to improving services to achieve better outcomes in social sectors. This problem assumes three forms. One is the lack of complementary infrastructure (water, electricity, phones) and equipment (vehicles, computers) to raise staff productiv- ity. Another is the shortage of trained staff. Many woredas report having too few people to perform tasks that would improve their efficiency. The woredas are short on managerial talent; frontline professionals (teachers, doctors, nurses, water specialists); and people with specialized skills in procurement, financial management, and accounting and reporting. The problems posed by these skill shortages are compounded by weak support systems. While streamlining of woreda-level financial management and accounting is under way and regions are moving toward a system of payments for all woreda offices through a single trea- sury system (Box 5.2), timely and complete reporting remains a challenge. Accounting and financial reporting is manually performed at the woreda level, even in SNNPR, the most advanced region with regard to expenditure management and control reforms; this slows the sharing of financial information. At the woreda level, sector managers yearn for simple and robust management tools, such as a basic information sheet for each sector that com- bines physical and financial data in a simple summary that allows finance and sector man- agers to monitor budget execution and its impact during the year. Currently, there is no easy way to translate detailed data on budgets and expenditures into summary information on Achieving Better Service Delivery Through Decentralization in Ethiopia 69 Box 5.2. Improving Financial Management Reporting at the Woreda Level Once the woreda budget is notified, sectors may request adjustments once a year. All such adjustments, including requests to transfer resources from one line item to another within the same category of expenditure (for example, recurrent non-salary), are supposed to be approved by the woreda Cabinet. Woredas follow the rules for budget transfer at the regional and federal level. These rules do not allow transfers from non-salary to salary budgets or from the capital to the recurrent budget. Authorized adjustments are supposed to be made only after review of the first six months of budget execution. Woreda Cabinets are reportedly actively monitoring budget execution and supervising sector offices' implementation of their plans as a basis for such adjustments. Streamlining woreda-level financial management is under way in all of the regions visited. In SNNPR all financial management tasks have been reorganized under the Woreda Office of Finance and Economic Development (WOFED), which "serves" all other woreda offices (single treasury system). WOFED receives sector offices' budget release requests for the coming month and notifies them of their monthly envelop. Thereafter it performs all financial transactions necessary for sector offices to perform the activities planned for the month. This system was adopted in 2004/05; hence, it is too early to assess whether it serves offices well. The concen- tration of financial management expertise should facilitate timely and complete financial management reporting. Source: Dom (2004). financial performance (on a sector by sector or program by program basis). No formal mech- anism is in place through which regional Bureaus of Finance and Economic Development provide regionally consolidated sectoral financial performance to sectoral bureaus, although in regions where the expenditure management and control reforms are sufficiently advanced (such as SNNPR), officials have indicated a willingness to establish such mechanisms. With decentralization at the woreda-level in 2002/03, the number of staff at the woreda level increased dramatically. In Oromiya, for example, the number of woreda-level staff (excluding teachers and health professionals) rose 54 percent (Table 5.2). Table 5.2. Number of Regional and Woreda-Level Personnel in Oromiya Before and After Decentralization Item 2001/02 (actual) 2004/05 (plan) Percent Increase Regional level 3,090 3,340 8.1 Woreda level 49,591 76,151 53.6 Teachers 50,880 50,880 0 Health workers 6,519 6,519 0 Source: Mussa (2005a). The devolution of responsibility to the woredas has significantly increased demands for staffing woreda administrations, bureaus, and frontline services. Four years into the woreda- level decentralization, a large number of positions remain vacant (Tables 5.3 and 5.4). The GTZ-Selam Development Consultants (2005) Woreda-Municipality Benchmarking Study indicates that up to 25 percent of positions in six woredas in Oromiya remained unfilled in 70 World Bank Working Paper 2005. The problem appears acute in Tigray and SNNPR. A study by Adal (2005) of four woredas indicates that: (i) in Harshin woreda, Somali region, the capacity building office had just 11 percent of the employees it needed, and the woreda had just 50 percent of the teach- ers and 14 percent of the health office staff it needed; (ii) in Bolosso Sorre woreda, in SNNPR, the health sector had only 60 percent of the staff it needed; and (iii) in Delanta Dawnt woreda, Amhara region, education offices had just 39 percent and the health offices only 29 percent of the staff needed. A recent review of the civil service (Mussa 2005c) concludes that the reasons for severe staff shortages include unattractive work environment/location (remoteness and lack of ser- vices, amenities, and opportunities); low pay; and the shortage of qualified personnel in the market. The study indicates that low salaries, the lack of fringe benefits, job dissatisfaction, and demand from outside the civil service are among the factors affecting staff shortages. Table 5.3. Staffing in Sector Offices in Delanta Dawnt Woreda, Amhara Region, 2003/04 Positions Positions Positions Percent Vacant Sector Office Approved Budgeted Vacant Positions Capacity building 57 29 28 49 Education 27 25 2 7 Health 14 5 9 64 Woreda administration 38 26 12 32 General administration 4 3 1 25 and security Woreda court 26 24 2 8 Justice 8 4 4 50 Agriculture and rural 264 143 121 46 development Information 39 23 16 41 Finance and economic 34 16 18 53 development Mass organization and 6 2 4 67 participation Youth, culture, and sport 17 3 14 82 Small-scale industry 3 2 1 33 Police 10 5 5 50 Prisons 11 11 0 0 Total 558 321 237 43 Source: Adal (2005). The study notes the traditions of stability of tenure and risk aversion in the Ethiopian civil service system. That situation is changing, however, and turnover is now high. Turnover and scarcity of staff are high in those professional and technical areas where sup- ply is low and demand high, particularly in the health and water sectors. One of the staff deployment strategies of decentralization was to transfer 40­60 per- cent of regional staff to the woredas. Doing so has proved problematic. Staff have been Achieving Better Service Delivery Through Decentralization in Ethiopia 71 Table 5.4. Staff Vacancies in Selected Woredas Region Woreda Total Staff Vacancies Percent Vacant Positions Amhara Fogera 1,175 221 19 Habru 1,084 184 17 Kutaber 871 209 24 Sekela 693 109 16 Sekota 957 257 27 Weldia 490 64 13 Total 5,270 1,044 20 Mean 878 174 20 Oromiya Boji 715 178 25 Cheliya 789 222 28 Gubakorcha 702 340 48 Guduru 714 149 21 Kurfa Chele 637 121 19 Mettu 522 47 9 Total 4,079 1,057 26 Mean 680 176 26 Tigray Gulomakda 785 62 8 Kafte Humera 644 39 6 Kelete Awlaelo 612 8 1 Laelaye Adiyabo 607 114 19 Atsbi Womberta 635 10 2 Total 3,283 233 7 Mean 657 47 7 SNNPR Basketo 404 98 24 Kedida Gamila 1,461 4 1 Konso 1,163 188 16 Kucha 574 228 40 Shebedino 1,218 104 9 Yiregachifee 263 50 19 Total 5,083 672 13 Mean 847 112 13 Total for all regions 17,715 3,006 17 Mean for all regions 770 131 17 Source: GTZ-Selam Development Consultants (2005). reluctant to go to the woredas, where career ladders are less attractive than at higher levels of government. The structure is inadequate, particularly for the positions opened, to absorb qualified and highly paid personnel. Attempts have been made to attract and retain staff, including offering them priority in training, establishing hardship allowances, creating a system of differential rates for counting service years, and offering contractual or tempo- rary employment. 72 World Bank Working Paper The GTZ-Selam Development Consultants (2005) Benchmarking Survey identifies three main causes for the failure of the woredas to recruit sufficient numbers of staff: low attractiveness of the location, particularly in remote areas; unattractive salary; and the lack of candidates who met the criteria for the jobs with open vacancies. Need to Improve Personnel Management of Frontline Service Providers Personnel management and compensation policies of frontline providers are key to the success of service delivery reforms. But personnel management is one of the least devel- oped areas in the region-to-woreda decentralization. This challenge has to be addressed urgently to avoid slowing the momentum of decentralized service provision. One of the most important issues raised by decentralization is which level of govern- ment is responsible for deploying teachers and health staff. Currently, responsibility for teacher deployment is shared by the regional government and woreda administrations. Regional Bureaus of Education recruit teachers with degrees (for secondary schools) and teachers with diplomas (for Grades 5­8), whom they deploy to woredas, in principle accord- ing to the woredas' requests. Woredas recruit and deploy teachers for Grades 1­4 themselves. Similar shared and overlapping responsibilities exist in the deployment of doctors, health officers, and nurses. About 20,000 health extension workers will be trained and posted to woredas throughout the country between 2004 and 2009. While the activities for which these health extension workers are recruited are clearly the responsibility of the woredas, their train- ing and deployment is an initiative of the federal government (Dom 2004). There is general agreement on the desirability of increasing the deployment of teach- ers and health workers to woredas to make progress toward national goals, such as improv- ing school quality and obtaining better balance of personnel between rural and urban areas. Responsibility for managing the personnel, much less financing their deployments, is not clear, however. Salary scales and promotions are handled by the region. In principle, woredas have dis- cretion to set salaries; in practice they often adopt the scales and guidelines of he regions and federal government. Shedding staff remains difficult, and the woredas often have to confer with the regional authorities to take such actions. Turnover of teachers is high, threatening the operation of schools and the improve- ment of school quality (JRM Education 2005). In Oromiya teachers typically remain on the job for a maximum of just four to five years. In many schools a third of all teaching positions remain vacant and a significant percentage of the staff, particularly in the second cycle of primary school, is under-qualified. In SNNPR while nearly 95 percent of the teach- ers in the first primary cycle are certificate holders, one-third of all teachers in the second cycle are under-qualified, and the figure rises to two-thirds in secondary schools. The efficiency of deployment of staff across schools and health facilities needs to be addressed. A recent analysis (World Bank 2005a) finds that teachers are often deployed randomly, that is, based on factors other than school enrollment. The study using 2001/02 data indicates 41 percent randomness in teacher deployment in SNNPR; 8 percent in Tigray; 22 percent in Amhara; and 28 percent in Oromiya. The low figure for Tigray is comparable to that in the best-performing countries, suggesting that practices in the region may hold lessons for other regions on ways to manage the deployment of teachers. The findings for the country as a whole, however, suggest that the management of teacher Achieving Better Service Delivery Through Decentralization in Ethiopia 73 deployment is weak, not only across schools but also between the two cycles of primary schooling within schools. These results are consistent with the reality that while teacher deployment rules may exist, outcomes are often decided at the local level by officials who ultimately implement rules in response to the priorities they perceive. The health sector suffers from high rates of attrition and turnover. Two major reasons for this are poor working conditions and weak enforcement of staff policies. In principle, most nursing and medical students are funded by the government (Federal and regional) and required to fulfill a service contract of specified duration (usually three to four years). They are posted to service centers on the basis of a lottery, which in theory cannot be influ- enced, except for medical reasons. In practice, there are questions as to whether this process is done in a genuinely unbiased manner, and there are allegations that the lottery may be subject to manipulation (see below). A survey of final-year nursing and medical students around the country reveals that 51 percent believe that a first posting is determined primarily by personal contacts, 56 per- cent believe they can get an undesirable first posting changed, and 43 percent admit that they may look for a job outside the health sector because pay and opportunities are better (Serneels and others 2004--see Figure 5.5).14 In a study based on focus group discussions in Addis Ababa and four provincial towns, participants generally believed that the lottery system for posting new graduates is "corrupt and that assignments are not truly random" (Lindelow, Serneels, and Lemma 2004). They believed that individuals with the right con- nections are able to turn down their initial assignment and wait for a better one; that there are cases where students have swapped their postings, sometimes with an accompanying financial transfer; and that many prospective health sector workers find that they earn too little and therefore often leave the sector. The findings of such surveys and studies should be interpreted with care because peo- ple's perceptions about a situation--especially as communicated in interviews--do not always reflect the reality of what is taking place on the ground. Furthermore, focus group assessments are neither objective nor representative, by design. Nevertheless, the findings of the survey and the focus group study that have just been described raise questions that need to be answered. At the very least, they point to the need for government action in undertaking an in-depth and critical review of the situation. Significant premiums are needed to attract nurses and doctors to rural areas. Eighty percent of medical and nursing students report that they would take up a rural post for wage premiums of 43­60 percent above the urban rate. Yet, according to regressions done using data from the 2001 Child Labor Force Survey, there is no rural wage premium; doc- tors and nurses in rural areas were found to earn less than those in urban areas, after con- trolling for other factors such as the level of education. These results are based on data collected a few years ago, however, and some steps are being taken to address the situation, as described in the next section. All of these problems are compounded by a degree of "brain drain" of qualified people away from the public sector and to the private sector, or outside the country. Substantial additional investment in human resources will be needed to reduce this, and to address 14. The survey included a total of 219 nursing and 90 medical students in eight institutions all over the country (Addis Ababa University; Gonder University in Amhara; Jimma University in Oromiya; Asella Nursing School in Oromiya; Yirgalem School of Nursing in SNNPR; Selihorn Nursing School in SNNPR; St. Luke's Catholic College in Oromiya; and Medco Biomedical College in Addis Ababa). 74 World Bank Working Paper Figure 5.5. Nursing and Medical Students in Ethiopia Perceive that the Assignment of Posts is Subject to Manipulation The obligation of working for the public sector if the government pays for education is a fair procedure My marks in class have an effect on the post I get offered in the lottery The first public sector posting depends primarily on your contacts The allocation of doctors and nurses to posts through a lottery is a fair procedure 0% 10% 20% 30% 40% 50% 60% 70% 80% 90%100% strongly agree agree neither agree nor disagree disagree strongly disagree Source: Serneels and others (2004). some of the other problems mentioned above such as the low wage premiums currently earned by workers in rural areas. This investment is especially needed at the regional and woreda levels, reinforcing the need for much more block grant resources to be passed down from the Federal government to the regions, as discussed above. Regional Innovations for Recruiting and Retaining Workers Regional governments have been testing innovations for recruiting and retaining different categories of employees (Mussa 2005a, 2005b). In Oromiya salary and non-salary incentives are offered in remote woredas which are facing difficulties in retaining staff. The region has already begun introducing new incentive mechanisms for teachers and health professionals where staff turnover is high. Among the planned incentives is the possibility of transferring professional staff from remote woredas to more urban areas after a certain period of service. The region is also adopting the federal government's policy that provides a 20­40 percent hardship allowance in woredas with severe climatic conditions. In this regard, the region has been paying 20­30 percent of salary as hardship allowance to all personnel in 20 woredas. There is some demand for hardship allowances to be paid in some other woredas as well. The Region is working to respond to this demand phase by phase. In the Tigray region, there have been proposals to introduce a fast-track career mecha- nism for certain types of public employees. As an incentive to staff, capacity building programs for woreda sector offices--distance education and regular education at the diploma, BA, and MA levels--were introduced. To ensure that staff enrolled in regular education programs at regional and national universities return, they are required to sign an agreement that obligates them to serve for a given number of years to compensate for the cost of their education. Achieving Better Service Delivery Through Decentralization in Ethiopia 75 In 2003/04 some woredas in the region introduced an incentive mechanism for school teachers. Based on performance of the schools, the teachers receive a bonus at the end of the year. The scheme has enhanced competition between schools. The Amhara region provides allowances for teachers, health professionals, and con- struction workers. For health professionals, "responsibility allowances," which range from Birr 50 per month for head nurses to Birr 200 for medical directors, were introduced in 2001/02. Teacher allowances have been provided since 1998/99. They range from Birr 90 for a head teacher in a first-cycle primary school (Grade 1­4) to Birr 125 for a head master of a teacher-training center. In 2001/02, the allowance for construction engineers was increased by about 40 percent of the previous professional allowance. Personnel based in remote and disadvantaged woredas also receive priority in obtaining training opportunities. In SNNPR the staffing structure stratifies woredas into four categories, based on the degree of hardship. In 2003/04 incentive mechanisms were introduced for 9 climatically harsh woredas and 17 disadvantaged woredas. The incentives include the following: Hardship allowance for nine climatically harsh woredas. Allowances of 40 percent of salary are given to all employees in the five woredas with the worst climate; allowances of 35 percent of salary are given to all employees in four other woredas. All medical coverage is provided for employees working in the nine woredas and their families. Workers in these woredas also "receive" two extra years of service on top of their actual number of years of service. Incentives for professionals in 17 disadvantaged and 9 climatically harsh woredas. Professionals receive three-step salary increases, 100 percent medical coverage for themselves and their family members, and two extra years of service on the top of their actual number of years of service. Training opportunities for health workers and teachers in remote woredas. Some employees receive opportunities to pursue BA and MA studies. Employees sign con- tracts with the civil service to ensure that they return to work after they complete their studies and serve for as long a period as they spent at a college or university. In December 2006, the Ministry of Capacity Building issued guidelines for the regions on how to ensure appropriate allocation and retention of staff, especially with regard to rural and hardship areas. These include: (i) guidelines on the amount of additional allowance for civil servants in hardship areas (e.g. 30 percent to 40 percent on top of salary for those in hot climate areas; 10 percent additional for those in cold climate areas); and (ii) priority to be given to employees that have remained for three years or more in hardship areas, in decisions regarding transfers of employees. The four larger regions have agreed to these guidelines. Empowering Citizens and Communities to Improve Services and Outcomes Given the crucial position of the woredas in the service delivery chain, the functional rela- tionships between the woredas and the regional and Federal governments have to be strengthened and more clearly defined. The starting point should be a clearer definition of specific, well-defined outcomes the woredas are expected to achieve and the provision of greater assistance from the regional and Federal governments in terms of funding and 76 World Bank Working Paper capacity building. But the benefits of decentralization will not be maximized unless another crucial group--citizens, the community, users of services--is pulled into the accountability mechanisms. Ethiopia's development framework (the Sustainable Development and Poverty Reduction Program) and discourse espouse the spirit and language of participation. Citi- zens are acknowledged as playing a crucial role in mobilizing their communities, holding local governments accountable, and accelerating the process of democratization. Official pronouncements are matched by numerous local organizations, including community- level organizations, in the education, health, and water sectors. For the social sectors, the most advanced and detailed proposal for involving citizen par- ticipation in service delivery was prepared for the education sector (see Government of Ethiopia 2002). The proposal provides a road map for decentralizing education management down to the school level and identifies the roles of all actors in the sector, starting with the Federal Ministry of Education down to the woreda-level education offices; Kebele Education Training Boards; and communities, teachers, and students. Among the more radical pro- posals in the guidelines are those that call for woredas to provide block grants to schools, in order to encourage competition between schools and reward well-performing schools; and also to communities, to encourage them to participate actively in school management. Before the start of decentralization, the role of citizens was one of the weakest links in the chain of service delivery in Ethiopia. One of the rationales for decentralization was to strengthen the role of citizens. Within the decentralized system, the kebele council is the forum where citizens could play a role in enforcing accountability of service delivery. How- ever, the system is still evolving and initiatives to further strengthen this formal channel of citizens' participation should be welcomed. There is mounting evidence, including within Ethiopia, that huge gains in outcomes can be made if citizens' roles are enhanced through participation and provision of infor- mation and choice. The role of the kebele councils is being strengthened by efforts--with participation by all levels of government in Ethiopia--to enhance transparency and accountability as part of the Protecting Basic Services operation. These efforts include ini- tiatives to post woreda and regional budgets in public places, and to present budgets in formats that are readily understood by citizens. The results of recent micro-studies provide insights into the extent of community participation in service delivery in Ethiopia. The role of citizens and communities is expressed through three main avenues: direct financial or in-kind contributions toward the establishment or maintenance of a facility, management of schools through the PTA or payment of teachers, and citizen voice in the local kebele and woreda councils, to bal- ance the local executive political leadership and ensure that service providers are account- able to the community. Findings from a recent micro-survey of woredas in Amhara, Oromiya, and SNNPR show that in general communities are willing to contribute to primary education (Mussa, Dinku, and Abdo 2005). In Amhara communities have made substantial contributions toward expanding education by building and rehabilitating schools. Data from the Regional Education Bureau indicate that the total amount of community contributions has increased considerably in recent years. According to the bureau's annual report for 2004/05, com- munity contributions in that year totaled about Birr 118 million (Birr 39 million in cash, Birr 52 million in labor, and Birr 27 million in materials), up significantly from 2003/04, Achieving Better Service Delivery Through Decentralization in Ethiopia 77 when community contributions totaled Birr 26 million. The contributions were used to build new schools, satellite rooms, classrooms, fences, and teacher housing, and also to fund the purchase of school furniture. Community contributions represented roughly 19 percent of the total primary education budget for the region in 2004/05. In areas of Oromiya where Basic Education Strategic Objective (BESO) programs oper- ate, communities built 2,515 new classrooms and rehabilitated another 2,575; constructed 110 new and rehabilitated 1,220 teachers' houses; hired 1,917 teachers; and contributed in other ways--with contributions totaling Birr 46 million across the region in 2003/04. This is equiv- alent to Birr 12.7 per student--a non-negligible sum, given that the Government of Ethiopia's (2002) guideline suggests that woredas should provide the equivalent of Birr 10 per student as a government contribution to meet elementary school operational costs (Dom 2004). In Bure woreda in the Amhara region, once parents and other community members agreed on the need for contributions to expand their schools, the amount of the contri- bution per household was left to be determined by individual families. Contributions are collected based on individual pledges. Parents contribute Birr 10­200 per school year, purely on a voluntary basis. Very poor families are exempted from contributing, as they are in other regions. Within a kebele, about 10 percent of residents decline to contribute, for various reasons. In the Lay Gaynt woreda in the Amhara region, minimum rates were determined in advance and collected accordingly. The amount of contribution was between Birr 5 and Birr 10 per school year. Very poor households and the landless were exempted from any kind of contribution. In both cases, contributions for education were linked to participa- tion in food and cash for work schemes, and participants were expected to make their con- tribution before collecting their wages. The types of contributions from communities are exemplified by two woredas, Boset in Oromiya and Boritcha in SNNPR. In Boset, it was found that communities and parents made considerable contributions (in cash and kind) for a new school building; expansion and rehabilitation of existing schools; construction and rehabilitation of office rooms and residential units for teachers, fences, latrines, and other structures. In a few cases, contri- butions also paid the salaries of school guards, assistant teachers, and facilitators. The amount of the contribution per household in Boset depends on various factors, including budget requirements for planned activities and community capacity. In most cases, parents and community members collectively determine the minimum amount of cash contributions per household (not per student). The contribution thus varies across households, kebeles, and woredas. Communities also hire teachers locally when there is a shortage of teachers, and salaries can be negotiated between the communities and the teachers. Although Boset is a food-insecure woreda in Oromiya, the community has shown a willingness to contribute (Table 5.5). In 2004/05, 22 first-cycle primary schools were constructed and about 64 teach- ers hired and paid by the community. In primary schools where the Woreda Education Office could not hire teachers due to budget constraints, the community discussed and decided to hire teachers. In 2004/05 the Woreda Education Office hired 64 qualified teach- ers on behalf of the communities, for a total of Birr 128,000. In addition, four first-cycle primary schools were built, at a cost of Birr 83,200, Birr 26,800 of which was covered by the community and the remaining Birr 56,400 of which was covered by UNICEF. The costs of other activities were paid for entirely by the community. 78 World Bank Working Paper Table 5.5. Community Contributions to Education in Boset Woreda, Oromiya Region, 2004/05 (Birr, except where otherwise indicated) Activity Number Total Cost Cash Labor Material Total Construction of schools 4 83,200 12,000 4,800 10,000 26,800 Maintenance of schools 8 1,190 500 690 -- 1,190 Expansion of classrooms 96 745,864 115,977 597,225 32,661 745,864 Construction of latrines 5 2,566 1,625 941 0 2,566 Payment of teacher salaries 64 128,000 128,000 0 0 128,000 Total 904,420 As percent of total woreda 16 budget Source: Mussa, Dinku and Abdo (2005). In the Boritcha woreda in SNNPR, the expansion and maintenance of classrooms and the purchases of desks and other materials was funded entirely by contributions from the community (Table 5.6). The community also financed the purchase of stationery and duplicating paper for school exams, transport, and allowances. The contribution of com- munities in Boricha in 2004/05 is estimated at Birr 1.45 million Birr, or 19 percent of the woreda budget for education. Another promising source of evidence of the benefits of citizen participation in ser- vice delivery comes from a local school management initiative currently underway in Amhara, Benshangul-Gumuz, and SNNPR. The program, currently in its second phase, is a collaboration between communities, woreda administrations, and USAID through an NGO called World Learning. The core objective of the program is to increase community ownership and management of primary schools by increasing effective participation of communities and fostering partnerships with local governments. The most visible benefit of enabling citizen participation in the experimental pro- gram has been large financial and in-kind contributions for improving school condi- tions. These contributions are substantial: contributions to all 940 schools totaled $820,106 in 2004, almost four times the contribution (about $224,000) of the BESO project (Figure 5.6). The substantial community financial contributions have been used to construct class- rooms and pedagogical centers; improve infrastructure (desks, other furniture); and reha- bilitate and build new teacher housing. These efforts have improved the quality of the schools, which has attracted more children and helped retain those already enrolled. The efforts have also increased school completion rates. Progress has been enhanced by campaigns by PTAs, Kebele Education Training Boards, and girls' advisory committees. Girls' advisory committees in participating schools have helped increase girls' enroll- ment (Box 5.4). These committees educate the community on the value of girls' education and discourage early marriage and abduction of girls, which often deny girls the opportu- nity to complete schooling (Figure 5.7). The success of the committees has motivated some education officials to form them in schools that are partnering with BESO. Community participation is also having the spillover effect of promoting local political support for the implementation of national policies, such as the MDGs. With Achieving Better Service Delivery Through Decentralization in Ethiopia 79 Table 5.6. Community Contributions to Education in Boritcha Woreda, SNNPR, 2004/05 (Birr, except where otherwise indicated) Woreda Community Contribution Education Total Activity Number Office Cash Labor Material Community Construction of 4 300,000 48,000 48,000 new schools Expansion 90 rooms 0 270,000 32,400 48,600 351,000 of school Maintenance of 20 rooms 0 30,000 36,000 54,000 120,000 school Purchase of desks 1,125 0 21,094 25313 37965 84,375 Construction of 62 0 12,400 14,800 22,320 49,600 housing for teachers Construction of 15 km 0 15,625 18,750 28,125 62,500 school fencing Construction of 18 0 13,500 16,200 2,430 54,000 pit latrines Construction of 10 0 2,5000 3,000 4,5000 10,000 library Purchase of 8 0 30,000 36,000 54,000 120,000 teaching materials Purchase of 85 0 4,250 5,100 12,750 17,000 blackboards Creation of 120 has. 0 49,800 0 0 49,800 school garden Collection 20 pits 60,000 22,500 27,000 40,500 90,000 of water Total 1,056,275 Community 19 contribution as percent of woreda budget Source: Mussa, Dinku and Abdo (2005). their enhanced roles in governance through participation and sense of empowerment, communities have taken part in monitoring the actions of frontline service providers. For instance, although communities cannot fire teachers, parents can and do report teachers who are absent or engage in socially unacceptable behavior to woreda education officials, who reflect these reports in teachers' performance records and evaluations (USAID/WLI 2004). In the Arsi zone of the Oromiya region, communities are demonstrating the capacity to manage complex water projects. A key reason for the success of the water scheme is the active involvement of the community in managing the resource and the flexibility it has to levy fees and sell extra water outside the community (Box 5.5). 80 World Bank Working Paper The Role of Communities in School Management The active participation of communities through PTAs has helped increased enrollment rates and improved the enrollment and retention of girls. The BESO Programme illustrates how communities help in school management (see Box 5.3). PTAs are involved in plan- ning and budgeting processes and in monitoring teaching. They deal with disciplinary problems of teachers and students and consult parents on a continuous basis about their children's school attendance. As community organizations, PTAs are the driving force for community mobilization, participation, and contributions. A PTA typically consists of five parents and two teachers. To build the capacity of PTAs, which is weak, the Woreda Education Offices and the BESO Box 5.3. Community Participation in Service Delivery at Work: Ethiopia's Basic Education Strategic Objective (BESO) The program: The first phase of the Basic Education Strategic Objective (BESO I) was initiated in SNPPR, as the Community School Activity Program (CSAP). It targeted 720 schools between 1997 and 2000. The lessons learned from this phrase of the project were then scaled up to cover schools in two additional regions, Amhara and Benshangul-Gumuz. The second phase (BESO II) began in 2002, as the Community Government Partnership Program (CGPP). By June 2004 the program covered 940 primary schools in 110 woredas. It is expected to cover 1,800 schools in the three regions before the end of the five-year second phase (2002­07). How it works: The schools participating in the partnership are selected by the Regional Educa- tion Bureaus and Woreda Education Offices. Once they are identified, World Learning contacts the Woreda Education Office and the schools to begin a series of consultations that clarify the duration and objectives of the partnerships and the role of the community. Each school is in the partnership for a maximum of two years. School incentive awards: Once a school joins the partnership, it receives an incentive award of about $1,500 paid over the two years that the school is in the partnership. The school incen- tive award is paid in three phases against completion of actions agreed on by the PTA and the community. The awards are provided to schools and communities to stimulate community involvement. Strategic objectives: The strategic objective of the program is to strengthen local community support for primary education by increasing the effectiveness of PTAs and local education orga- nizations, such as the Kebele Education Training Boards. This is done by strengthening the capacities of PTAs, the Kebele Education Training Boards, and the Woreda Education Offices; promoting alternative education by establishing non-formal education centers; using parapro- fessional teachers and flexible timetables; and promoting equal access to schooling for boys and girls through campaigns discouraging early marriage for girls. School development agents and zonal coordinators: A key instrument in building capacity and maintaining the sustainability of local participation in school management is the establishment of school development agents and zonal coordinators--experienced teachers or school direc- tors who have been seconded to the World Learning Ethiopia office by the Woreda Education Offices or Regional Bureau of Education. Each school development agent works with at least 12 schools; each zonal coordinator is responsible for 10 woredas and 85 schools. The responsi- bilities of the school development agents include building the capacity of PTAs by identifying their training needs and delivering the necessary training, fostering community-government partnerships, developing programs geared toward building community ownership of schools, and motivating parents to send their children to school. Source: Government of Ethiopia and AED/BESO II (2005a, 2005b). Achieving Better Service Delivery Through Decentralization in Ethiopia 81 Figure 5.6. Community Contributions Represent a Much Larger Source of School Financing Than Incentive Awards 100% 80% 60% contribution of 40% Share 20% 0% Amhara Bens hang-Gumuz SNNPR Total Region and Total (Source: USAID/WLI, 2004) Community contributions School incentive awards Figure 5.7. Enrollment by Girls Rose in BESO Schools 120,000 100,000 80,000 60,000 40,000 20,000 0 Amhara Benshang-Gumuz SNNPR Total Pre-program (2001/02) 57,261 3,704 24,880 87,844 Most recent program year (2003/04) 65,617 5,131 29,702 102,449 Participating Regions and Years (Source: USAID/WLI, 2004) 82 World Bank Working Paper Box 5.4. Boosting Girls' Participation in Primary Education Through Girls' Advisory Committees To increase girls' enrollment rates, most schools in the BESO program have established girls' advisory committees. Using their local knowledge, these committees implement the most appro- priate strategies for increasing girls' participation in their communities. Some activities are designed to increase participation, while others are geared toward empowering girls. Activities related to increasing girls' participation include: Assessing the causes of low enrollment and educational performance by girls and designing appropriate interventions to improve them. Organizing community forums to disseminate information on the value of educating girls, and encouraging parents to send school-age girls to school and keep them there. The forums include influential people, such as religious, traditional, and political leaders. Home visits to households in the company of community leaders are also sometimes made. Providing tutoring and make-up classes for girls. Encouraging parents to reduce the amount of domestic chores girls must perform and to give equal support to boys and girls at home. Organizing school registration periods and, if necessary, extending them for girls. Providing public recognition of parents who send girls to school as a way to motivate others to do the same. Activities related to empowering girls include: Providing moral support to girls and presenting them with awards for demonstrated progress. Arranging for class and school competition programs that involve girls, and presenting awards, in the presence of parents, to girls who do well. Providing continuous follow-up and encouragement through girls' advisory committees, women teachers, and other female professionals in the area. Assigning leadership roles to girls--as class monitors, school club leaders, and chairpersons in meetings. project have conducted some training on planning; reporting; community participation and contributions; and record keeping. The roles and responsibilities of PTAs vary across regions. In the Amhara region, PTAs are not responsible for managing the collection or use of the money contributed by the community. In contrast, PTAs in Oromiya and SNNPR are responsible for these activities. In most communities, only households with children in school are expected to contribute to the school, but some communities require all households to contribute. In some schools, the community contribution is counted on a per student basis, while in others it is counted on a per parent basis. The Challenge of Community Participation in the Context of Decentralization Community involvement in service delivery has been increasing in Ethiopia. Communi- ties have a sense of ownership of their schools, and their contributions are large in finan- cial terms, even compared with woreda budgets. Community participation in supporting local schools places a heavy burden on communities, however. Even in food-insecure woredas, participation is high, albeit not uniform across regions. Achieving Better Service Delivery Through Decentralization in Ethiopia 83 Box 5.5. Community Participation in Water Supply, Sanitation, and Health Education Schemes in Oromiya The Hittosa water supply project is a joint project of WaterAid (an NGO), the water department of the Arsi zone, and the community of Hittosa. The scheme aims to provide safe drinking water for 28 villages and 3 small towns. Four other such schemes exist in the zone. Each village forms an independent water committee, which works with an autonomous water management board at the woreda level. An executive committee coordinates the village com- mittees and the water administration office. The communities contribute fees and labor to the scheme and are allowed to sell surplus water to other villages and users. The income and fees from the water sales are used to maintain the scheme and pay attractive and regular staff salaries. In return, employees provide reliable services and keep good records, especially of financial transactions. The scheme has worked because it is well designed, the leadership is responsible to the community, there is a well-established community structure and manage- ment system (water committees), and staff are motivated. Source: Kassa (2003). One of the main reasons for the increased burden on communities is the inadequate financing of local governments since decentralization. Woreda budgets are not sufficient to recruit teachers, maintain schools, and provide learning materials. With the growing trend toward greater community financing of these services, the capacity of communities is reaching its limit. It is becoming common, particularly in remote areas, for communi- ties to contribute toward the construction of first-cycle primary schools; in some instances, they hire teachers locally and pay salaries. Since the continuity and sustainability of these salaries are not guaranteed, turnover is high, and Woreda Education Offices cannot replace such teachers, since the positions are not budgeted. The drive to improve the quality and effectiveness of service delivery needs to fully take into account--given their scope--community contributions and participation in the plan- ning and budgeting processes. It is essential to ensure that the burden is shared equitably across regions, woredas and kebeles. The burden on disadvantaged communities, includ- ing food-insecure woredas, should be reduced, with larger allocations of woreda resources. Clear guidelines should be established on how the burden should be shared by communi- ties, woreda administrations, and other parties, including donors. The role of community associations in promoting effective participation can be enhanced with training, as in the case of the BESO program. The training of PTAs is a key strategy for promoting community participation in the delivery of services. Improving Services by Providing Information Information is essential for assessing the performance of service providers and the actions of politicians and policymakers. It is crucial in making decisions about the benefits of a pol- icy. Information also helps individuals make private choices about health, education, and water and sanitation. Two key determinants of child health outcomes in Ethiopia are maternal knowledge and access to information. This conclusion is consistent with the finding of Christiansen and Alderman (2004), who show that maternal knowledge can explain much of the 84 World Bank Working Paper observed differences in child malnutrition. Unlike their study, which uses the educa- tion of the most educated female in the household, the new analysis done for this study using household data from the Welfare Monitoring Surveys of 1995 and 2000, matches children and their mothers, for the period 1995­2000. It uses year dummies to control for time variation in order to improve the ability to identify the effects of explanatory variables that are potentially endogenous in the cross-section. It is possible, for exam- ple, to conflate the effect of distance to a health center with preferences for health cen- ters. If some parents have very strong preferences for using health clinics and therefore choose to live near a health center, the estimated effect of government intervention reflects those preferences as well. With a given distribution of preferences for health centers, expanding access to other areas may not generate similar effects. The advan- tage of time variation is that it allows the effect of changes in distance over time (that would occur through the opening or closing of health clinics) to be examined. By con- trolling for region and time (through the interaction of region and year dummies), this report identifies the determinants of health outcomes using within-region and over- time variation in the explanatory variables. In the basic regression of child malnutrition, a literate mother can reduce her child's malnutrition by 0.5 standard deviations (23 percent) relative to illiterate mothers. One interpretation, derived from the health capital model, is that literate mothers face a lower cost of investment in health capital. Among mothers who report using health facilities, it is not education but the combination of education and information that generates the strongest impact on child health. A literate mother with access to infor- mation (as proxied by access to radio and television) increases her child's nutritional status by 15 percent. Mother's education (measured by literacy) is the single most important determinant of child mortality. Literacy reduces infant mortality by 2.6 percentage points (nearly all of the reported deaths for the sample) and child morbidity by nearly 14 percentage points (34 percent). Mother's literacy matters more for boys than for girls, however, whereas the number of boys' deaths declines by 0.13 (9.3 percent), the number of girls' death falls by 0.09 (6.5 percent). A second story that emerges is the role of internal pollution. Clean cooking oil has a consistently negative effect on child survival. Although not precise in the still birth regres- sions, the results suggest that the impact of clean cooking is of a similar magnitude as the effect of literacy. Like mother's literacy, clean cooking oil has a greater effect on boys than girls: indoor pollution has twice as great an effect on child morbidity among boys (-0.11, or 8 percent) than among girls (0.055, or 4 percent). The Government's introduction of health extension workers into communities through the Health Extension Program is a way to improve household information about basic health inputs. The program focuses on the prevention of childhood and maternal diseases. It will be especially critical in implementing the Integrated Management of Childhood Ill- nesses (IMCI), which is being implemented in all regions as the key strategy for reducing under-five mortality and morbidity. The strategy supports interventions of the type con- sidered in a recent note on reducing child mortality in the Lancet (Jones and others 2003). The results suggest that increasing maternal information and reducing internal pollution (through the use of clean cooking oil) may be fruitful strategies for reducing deaths among infants and children. Achieving Better Service Delivery Through Decentralization in Ethiopia 85 Improving Accountability Mechanisms Communities make significant financial contributions to maintaining the continuity and quality of services in Ethiopia, participating directly in the management of the services, and offering specific solutions to difficult challenges (such as keeping girls in school). But cit- izen participation in Ethiopia is a weak link in the chain of service, primarily because there are no clear guidelines for communities' role and many communities lack information. Several sets of actions could improve citizen participation and involve citizens in service delivery in a more robust and effective way. Institutionalizing Citizen Report Cards Citizen report cards can be an effective way of enabling citizens to monitor the quality of services and the prudent and transparent use of resources (Box 5.6). They can help enforce agreements between the federal and regional government on the one hand and with the local government on the other. Regional governments can use woredas' scores on the report cards, together with other sources of information, to enforce performance contracts on specific targets. Ethiopia currently lacks a mechanism for introducing, implementing, and publicizing the results of woreda report cards. A good beginning would be to scale up the recent activ- ities of PANE, a consortium of NGOs, which has completed a report on a pilot citizen report card. This report assesses the performance of a sample of woredas in implementing the Sustainable Development and Poverty Reduction Program goals on education, health, and water and sanitation. Strengthening Citizen Voice Through Other Social Accountability Mechanisms to Apply Social Pressures on Providers Other promising approaches and mechanisms to enhancing clients' voice could also be tested in rural and urban environments Ethiopia. Several approaches that have been suc- cessful in some regions or in other countries could be piloted in selected woredas and kebe- les and scaled up as experience grows. These include: Citizen monitoring of woreda performance indicators that have been agreed on with the regions (such as the performance agreements implemented in SNNPR-- see Box 3.2). Community scorecards of school performance in coordination with PTAs. Citizen participatory planning and monitoring to strengthen the accountability of woreda and kebele councils for service delivery. One of the key objectives of these mechanisms is to enhance community involvement in performance assessment and to coordinate channels of accountability to ensure better ser- vice delivery. This could also be achieved through the formal channel of the kebele where citizens are also represented in the council. 86 World Bank Working Paper Box 5.6. Using Citizen Report Cards in Ethiopia Citizen report cards allow users of public services to provide feedback to service providers regarding their performance. By identifying the key constraints (coverage, quality, reliability, and so forth) that limit their use of public services, citizens provide a clear message on the weaknesses of public services. A pilot citizen report card in Ethiopia sampled 3,829 households in Afar, Dire Dawa, Oromiya, SNNPR, and Tigray. Dire Dawa was chosen as the representative urban subsample; only rural households were sampled in the other regions. Households were asked to assess key basic public services typically delivered by the woredas, including drinking water, health and sani- tation, education, and agricultural extension services. The report cards revealed the following key findings: Drinking water Nearly three-fourths of all households lack potable water, and pollution of river water is a major concern. A significant proportion of respondents report that communities are involved in maintain- ing water sources. Most of this involvement is in the form of contributed labor. Many households express willingness to pay for improved water services. Health and sanitation Most households report improvement in access to health care. Most households report that immunization services have expanded, but many cite lack of medicines, long distances, and poor quality of services as major reasons why they prefer private over public services. The cost of obtaining treatment in government facilities is high. Very few respondents who use public services report receiving free medicine. Education Teacher absenteeism is low, and the quality of teaching is not a major concern. The shortage of drinking water and functioning toilets in schools is a major problem. Community involvement in school management is informal. Few parents belong to PTAs. The cost of education is significant and varies across regions. One-third of parents make extra payments to school administrators, and 17 percent of those paying extra fees report being coerced to do so. Source: PANE (2005). Launching Public Information Campaigns Some key actions for improving education and health outcomes can be accomplished through public information campaigns. For instance, many children are out of school because their families do not permit them to attend, perhaps because they do not recog- nize the value of education. Child health outcomes could be improved substantially by pro- viding nutrition information to mothers or helping households understand the value of boiling water, especially for children. Public campaigns against smoking, drinking and driving, and unsafe sex have been suc- cessful in other countries. In Ethiopia the gains from public information--in health, in edu- cation, and in water and sanitation--would likely be large. Woredas and municipalities Achieving Better Service Delivery Through Decentralization in Ethiopia 87 could use health extension workers in conjunction with community or municipal radios to disseminate information on child health and the value of education. Involving com- munities in the management of services in health, education, and water could increase accountability, improve ownership, and provide an avenue for transmitting information on specific issues, such as the value of girls' education. Reducing the Opportunity Costs of Sending Children to School Even if improvements in two of the key constraints--weak local government and inadequate citizen participation--occur, achieving the MDGs will not be feasible unless the huge oppor- tunity costs facing poor households are removed or substantially reduced. These opportu- nity costs are evident in the tradeoff between income generating activities and the use of services. Earlier studies (World Bank 2006b) indicate that about 45 percent of primary school­ age children in Ethiopia work full time. Adding to this figure children who work part time pushes the rate of working children to 60 percent. Children of all ages participate in work, and the participation rate increases with age. When asked why their children work, 47 per- cent of parents reply by referring to high opportunity costs of sending their children to school. About 16 percent report that their children are not in school because they help with Figure 5.8. School Attendance is Higher in Woredas in which Primary Education Continues to Grade 8 0.7 0.6 0.5 Rate 0.4 0.3 grades 7-8 Attendance 0.2 grades 1-6 0.1 0 13 14 15 16 17 18 Age Source: World Bank staff estimates from the 2000 Welfare Monitoring Survey. 88 World Bank Working Paper domestic work. Another 14 percent report lack of a nearby school, 9 percent report that they cannot afford school, and 8 percent report that their children need to supplement family income. The significance of opportunity costs could be even larger than these fig- ures indicate, because more than 80 percent of children report that they work in order to supplement household income or assist in household chores (Figure 5.8). In 1994 the Government changed the structure of schooling from six years of primary education to eight. For a number of reasons (financial resources, political factors at the regional level, and so forth), the pace of change to the new structure was uneven across the country. Some communities adopted the new system quickly; others could not. In woredas where the full cycle of primary education is not available, children aged 13­18 years have lower attendance rates than children in woredas where eight years of primary education are provided. These results may indicate that parents whose children have little or no prospect of completing the full cycle of primary education have less incentive to send them to school. High opportunity costs make it difficult for households to use and benefit from ser- vices. One consequence of high opportunity costs is child labor. Another is high child mor- tality and morbidity. Differences in the opportunity costs faced by households lead to inequalities in outcomes. If not addressed through public policy, these differences lead to persistent inequalities, which act as a drag on progress toward the MDGs. APPENDIXES APPENDIX A Calculating Block Grants Allocations From Federal Government to Regions and Regions to Woredas Using the Three-Parameter Formula F rom 2003/04 to 2006/07, a two-step allocation process was used to make Federal block grant allocations across regions. First, each region would obtain the same recurrent allocation as in 2002/03; this would be a minimum allocation. Second, after making these initial minimum allocations, the "three-parameter" formula would be used to distribute additional resources across regions. This section describes this second step and the way in which the "three-parameter" formula was applied to this additional pool of resources. As explained in Chapter 3, starting from 2007/08, the Federal government is grad- ually shifting to the new "fiscal equalization" approach. The "three-parameter" approach has also been used in some regions to distribute block grant resources to woredas, but regions have also been generally shifting to other approaches. The material in this section also applies to the "three-parameter" approach as it has been applied by regions to calculate the size of block grant transfers to each woreda. The "three-parameter" approach uses three parameters: population, level of develop- ment, and revenue collection effort. The allocation to each locality (region/woreda) is based on the following formula: Gt = Popi w1 + Devi w2 + Revpeeri w3 100 10 10 Pop 10 i Dev i Revper i i=1 i=1 i=1 where: (i) Gi is the percentage share of grant entitlement for locality i; (ii) Popi is the pop- ulation of locality i; (iii) Devi is an Index of development level of locality i (with a higher value indicating a less developed locality; (iv) Re vperi is an index of revenue raising effort 91 92 World Bank Working Paper and sectoral output performance indicator of locality i; and (v) wj are weights assigned to the indicators. Key features: The total available funding is in effect split into three portions--allocated respectively based on the "population," "development level" and "own revenue effort" of localities--with predetermined percentage shares of the total going to each (65 percent, 25 percent and 10 percent for the allocations from the Federal government to the regions). The "population" portion is allocated across localities on an equal per-capita basis (e.g. if one locality has twice the population of a second locality, then it gets twice the allocation). The "development level" portion is meant to address equity; it is allocated across localities such that less developed localities get larger allocations. The "own revenue effort" portion is allocated across localities such that those with greater revenues and sectoral performance get larger allocations. The "development level" portion is independent of population size and locality size. So two localities with equal development level and own revenue collection effort would get equal total allocations, regardless of differences in population size. Localities are free to use their total allocation in any way they choose, in theory. In practice, in some regions, woredas are given guidelines as to minimum shares that should go to capital spending. For Federal allocations to regions, there is 100 percent "offsetting" of expected external donor funding (for specific projects) from block grant allocations. This works as follows: (i) The allocations to each region following the above three-parameter formula are made from a pooled basket of funds that consists of two types of funds: the total amount of (untied) block grant funding available, as well as the total amount of (tied) funding that is expected to be available for regional donor-funded projects. (With some exceptions, all donor funding to regions for specific projects are supposed to go through the Federal gov- ernment.) (ii) The previous step gives the total allocation to each region, from untied block grant (Treasury) sources as well as from tied donor-funded sources for specific projects. The allocation of block grant (untied) Federal funding for each region is calculated by tak- ing the total allocation to that region (as calculated from the formula), and subtracting from this the amount of donor specific-project funding for that region. In other words, there is 100 percent offsetting of external donor funding from block grant allocations. In the case of Federal block grant transfers to regions, there is no offsetting of each region's own revenue collection from the region's block grant allocation. In the case of the transfers from regions to woredas, there is variation across regions in terms of what items are "offset" from block grant allocation. There are two types of items that are often fully or partially offset from the block grant allocation to each woreda: (I) the expected amount of (on-budget) spending that the woreda is expected to undertake from donor specific-project funding; and (II) the amount that the woreda is expected to collect in own revenues. Typically, there is full (100 percent) offsetting of item (II) from block grant allocations, following the procedure explained in the previous paragraph. There is variation across regions in the extent of offsetting of item (I)--see next paragraph. Achieving Better Service Delivery Through Decentralization in Ethiopia 93 In some regions, there is full offsetting (100 percent) of item (I) from the block grant allocation to each woreda, while in other regions there is partial (less than 100 percent) or no offsetting from block grant allocations. "Partial offsetting" means that a woreda's block grant allocation is offset by a predetermined percentage share of its expected expenditure from donor funding for specific projects. The principle behind the offsetting of specific-project spending is that the total entitle- ment of funding for each region (or woreda if applicable) should be determined by the block grant allocation formula, regardless of whether this funding is untied (in effect, in the form of block grants) or tied (as in the case of donor funding for specific projects). Issues in the Application of the "Three-Parameter" Formula The formula is not sensitive to the service delivery needs of different localities, possibly cre- ating a disincentive to expanding service delivery. As an example, compare two localities with: (i) the same population, development level and own revenue effort, but where: (ii) one locality has substantially higher enrollment than the other, because of greater efforts made to increase enrollment. Both will receive the same per capita allocation and total allocation. However, spending per student in the higher-enrollment locality would likely be lower than in the other locality and thus would be less able to fulfill its needs. In turn, this allocation system may reduce incentive to enroll more students. Small localities benefit disproportionately in per capita terms, in a manner that may not be equitable; a small and rich locality could get a larger per capita allocation than a large and poor locality. The "population" portion of the total available funding provides an equal per capita amount to all localities, and so any difference between localities in per capita allocations stems from the "development level" and "own revenue effort" portion. The "development level" portion is problematic because it provides an equal amount of funding to two different localities that are equally developed, even if one has a much larger population than the other. This implies a lower per capita allocation for the larger locality. A large and poor locality will get a larger total allocation from the "development level" portion than a small but rich locality. But the per capita allocation of the large locality will often be less than that of the small one, even though the large region is poorer. The per capita Federal block grant transfers to the regions in Table A1 illustrate this point. As an example, SNNPR--which has one of the highest regional poverty rates-- received a much lower per capita transfer in 2005/06 than the relatively rich region of Harari, due to the low population size in Harari. Indeed, as Figure 3.2 in Chapter 3 shows, there is a very close inverse relationship between a region's population and its per capita Federal transfer from the Federal government. One justification for this type of relationship is that there are fixed costs that every region has to face, such as the costs of maintaining a Bureau of Finance, a Bureau of Edu- cation, and so on. Another is that the cost of service delivery (and therefore per capita needs) is higher in regions with lower population density. The upshot is that the "three-parameter" formula is being used to address these dif- ferent objectives simultaneously. In particular, the "development level" portion is attempt- ing to address equity concerns, fixed cost concerns, and possibly the greater costs of service delivery in less densely populated areas, in ways that do not meet these objectives. 94 World Bank Working Paper Table A1. Per Capita Block Grant Transfers and Population Proportion of people living Proportion of in moderate people living poverty according in absolute 2005/06 to 1999/2000 poverty according per capita HICES to 1999/2000 Population transfer from household household 2005/06 federal survey survey (millions) government (Birr) Tigray 78.9 61.4 4.28 118 Afar 70.1 56.0 1.37 181 Amhara 62.3 41.8 18.87 90 Oromia 60.6 39.9 26.18 79 Somali 65.2 37.9 4.27 101 Benshangul/Gumuz 70.6 54.0 0.62 312 SNNP 69.4 50.9 14.70 84 Gambella 71.1 50.5 0.24 547 Harari 42.0 25.8 0.19 441 Dire Dawa 52.6 33.1 0.39 254 APPENDIX B The "Fiscal Equalization" and "Unit Cost" Approaches for Block Grant Allocations General Descriptions and Comparison of the "Fiscal Equalization" and "Unit Cost" Approaches Aside from the "three-parameter" approach, the report mentions two other approaches for allocating block grants to localities (regions/woredas), the "fiscal equalization" approach (which is currently partially used for allocating Federal block grant transfers to the regions, as explained in Chapter 3) and the "unit cost" approach (used by several regions, notably SNNPR until 2005/06, to allocate block grants to woredas). Over the next four years, the "fis- cal equalization" approach will be gradually phased in for the allocation of Federal block grant transfers to regions; from 2010/11 onwards, these allocations will be based entirely on the "fiscal equalization" approach. Although there are important differences between the "fiscal equalization" and "unit cost" approaches, they share some common principles: Both approaches are primarily "needs-based" in the sense that recurrent funding-- which accounts for the bulk of total funding--is allocated in larger quantities to localities with higher demand and provision levels of public services (localities with higher "needs"). Recurrent funding allocations are made separately for each sector/subsector (primary education, secondary education, health, and so forth). In both approaches, capital funding is allocated in a more equity-oriented manner, with larger amounts provided to localities that are more "backward," in effect, those with lower amounts of capital stock related to public service provision (for exam- ple, less schools per capita, health posts per capita). The rationale is that these local- ities are most in need of additional funding so as to be able to accelerate investment 95 96 World Bank Working Paper in the appropriate types of capital stock, and to be thus able to "catch up" with other localities. Both approaches--in common with the "three-parameter" approach--make adjustments to take into account own revenue collections by the localities, but they do so in different manners (see below). Both approaches--in common with the "three-parameter" approach, and in keeping with the principle of "block" (or untied) grants--are merely methods for calculating the total amount of the block grant allocation to each locality. Ultimately, each locality is generally free to use its total amount of available funding as it wishes; it is not obliged to adhere to the sectoral sub-allocations used in the calculation of its total block grant allocation. (So if $x were allocated to a locality for recurrent spending on primary education, for example, actual spending on this subsector need not be $x.) There may, however, be a constraint imposed that total recurrent spending (across all sectors and subsectors) may not exceed the total recurrent allocation. In the case of SNNPR, there has also been the expectation that woredas and urban administrations must meet the tar- gets agreed to in the performance agreements (see below), and must use their available funding accordingly. The two approaches differ somewhat in the manner by which the above two principles are adhered to. With the recurrent allocations, a key difference lies in the assumptions regarding the unit cost of service delivery. With the primary education subsector, for example, for both approaches, recurrent funding is allocated based on the expected total primary enrollment of each locality (they are both "needs-based" in this sense). However: With the "fiscal equalization" approach, recurrent allocations are initially made to each subsector under the assumption of equal unit cost for all localities. (For exam- ple, recurrent allocations for primary education for the following fiscal year are such that each locality gets the same amount per student that is expected to be enrolled. The amount allocated per student is roughly equal to the total primary education expenditure across all localities, divided by the total primary enrollment.) However, at the end of the calculation process, the total block grant allocation is adjusted by a factor that depends on population and other variables, to correct for differences between localities in economies of scale and remoteness. With the "unit cost" approach as practiced in SNNPR until 2005/06, recurrent allocations to each subsector were made under the assumption that all woredas would achieve equal unit cost in the longer run (over several years). In the shorter run, the recurrent allocation to a woreda for a particular subsector and fiscal year was based on a target unit cost that was set somewhere in between the unit cost of that woreda in previous years and the regional average unit cost. The expectation was that woredas with higher-than-average unit costs should gradually lower their unit costs over time--and that woredas with lower-than-average unit costs should gradually raise their unit costs over time. In each successive year, the target unit costs across woredas would thus get closer and closer to each other, converging Achieving Better Service Delivery Through Decentralization in Ethiopia 97 over time. The target unit costs were, in turn, computed from target values of "cost drivers" (see below). With the "unit cost" approach as used by SNNPR, a "total sector ceiling" was applied to each subsector and sector. For each subsector/sector, allocations for each woreda were scaled up/down (usually down) so that they added up (across woredas) to the total ceiling allowed for that subsector/sector, for that fiscal year. There are also some differences between the two approaches in making capital alloca- tions, even though both follow similar principles as outlined above. Even though the aim in both cases is to allocate more to localities that are "backward," there are some differ- ences in how this is defined and measured. Finally, the two approaches differ in the way by which they take into account own rev- enue collection, as follows: With the "unit cost" approach as practiced by SNNPR until 2005/06, the compu- tations described above for recurrent and capital funding were used to compute the total allocation/entitlement to each woreda (woreda/urban administration), but this was not equal to the actual transfer. To derive the actual transfer level for each woreda for a fiscal year, the woreda's expected own revenue collection for that year was subtracted from its total allocation/entitlement. A woreda's expected own rev- enue for a fiscal year was, in turn, estimated from its actual revenue collected in previous fiscal years. The "three-parameter" approach also subtracts a woreda's expected own revenue from its total allocation to determine its actual transfer. With the "fiscal equalization" approach, a locality's expected own revenue is not subtracted from its total allocation to determine its actual transfer. Each locality's transfer is equal to its allocation (with the latter determined using the recurrent and capital allocation processes described above). However, adjustments are made to the total allocations to take into account differences in own-revenue-raising capacity. Allocations are adjusted downwards for localities with higher-than- average capacity to collect own revenues (for example, a larger per-capita tax base). For regions with lower-than-average capacity to collect own revenues, allocations are adjusted upwards. These adjustments are made independently of the actual amounts of revenue collected in previous years; they only take into account fac- tors such as the size of the tax bases which determine the capacity to collect own revenue. The treatment by the "unit cost" and "three-parameter" approaches of own revenues is problematic because it reduces localities' incentives to widen their tax bases and to otherwise improve own revenue collection. In effect, if a locality takes steps to improve its own revenue collection, it is raising expectations of its own revenue collection in future years, thereby reducing its block grant transfer from the higher-level (regional/ federal) government in future years. In this respect, the "fiscal equalization" approach is superior since it does not reduce localities' incentive to collect own revenues. (With the "unit cost" and "three-parameter" approach, matters are worsened by the fact that, in many regions, local governments are not allowed to "keep" for themselves any own 98 World Bank Working Paper revenue collected in excess of their expected own revenue collection. This gives them almost no incentive to collect any surplus revenue.) More Details on the Unit Cost Approach This section describes key details of the "unit cost" approach as used in SNNPR until 2005/06. The "recurrent" portion of the total available block grant resources was allocated to woredas according to recurrent needs calculated separately for each key sector (such as education, health, agriculture), and based on specific targets for each subsector for the fol- lowing fiscal year (such as number of primary students to be enrolled, number of health posts to be managed). Starting from 2004/05 in SNNPR, these targets were set as part of "performance agreements" signed between districts and the region, whereby the district commits to attain the targets. In each key subsector s, the recurrent allocation for woreda i was roughly given by Recurrent Allocation to Woreda Performmance Target for Target Woreda UUnit Cost i for Subsectoors = Following Fiscal Year (e.g. Primary Education) × (e.g. Target Woreda Expenditure) (e.g. Primary Enrolment) PeerPrimary Student --with all of these three figures specific to each woreda, for every subsector, in every fiscal year. As mentioned above, a woreda's target unit cost for a fiscal year was set at a level some- where in between the woreda's actual unit cost in previous years and the regional average unit cost. Thus, target unit costs across woredas would get closer and closer to each other in successive years, converging over time. Target unit costs (for each district and subsector) were, in turn, computed from tar- get values for "cost drivers"; "cost drivers" are key measures and ratios (such as the pupil- teacher ratio or PTR in the case of education) that determine the unit cost. These target "cost driver" values were set individually for each woreda, with different values set for each woreda. The chosen target values were clearly specified for each district; each woreda knew exactly how its allocations were computed and which target "cost driver" values were used in the computations. The setting of target values for the "cost drivers" was based on the principle that, in the longer run, all woredas should have the same values for "cost drivers" (for example, PTRs). (This would, in turn, ensure that all woredas have the same unit costs in the longer run.) A woreda's target value for a "cost driver" (for example, the PTR in the dis- trict) for a fiscal year was set somewhere in between actual values in previous fiscal years and the regional average value for this "cost driver." As in the case of the unit costs, the expectation was that woredas with higher-than-average values for a "cost driver" (such as PTRs) should gradually take measures to lower these values, while woredas with lower- than-average values for "cost drivers" should do the opposite. In each successive year, the target values for the "cost drivers" would thus get closer and closer to each other, converging over time. In primary education, for example, the target woreda unit cost (target expenditure per primary student) was the sum of: (i) target teacher salary expenditure per primary student; Achieving Better Service Delivery Through Decentralization in Ethiopia 99 (ii) target non-salary expenditure per primary student; and (iii) target operational (non- salary) cost per primary student. The biggest of these is the target teacher salary expendi- ture per primary student, which was given by: per Target teacher expenditure primary studennt = [ target average teacher salary ] × [taarget PTR ] Again, all of these target figures were set individually for each woreda, and they were spe- cific to each woreda, in every fiscal year. Over time, the target values set for the average teacher salary and the PTR would converge across woredas. Ultimately, of course, woredas were not obliged to meet their target values for the "cost drivers", unless specifically stated so in the performance agreements. Each woreda's total allocation was transferred as a block grant, and it was free to use this funding as it wishes, the only expectation being that it should meet the performance targets (for example, for primary enrollment) that it had signed onto in the performance agreements. However, the target values for the "cost drivers" were explicitly stated to each woreda, and in practice each woreda had an incentive to try and meet them, regardless. This can be illustrated by examining incentives regarding the woreda average teacher salary (ATS) and the woreda pupil-teacher ratio (PTR), key cost drivers. A woreda with higher-than-average ATS (i.e. ATS higher than the regional average) would get a target ATS that was lower than its own ATS. It was free to maintain an ATS higher than its tar- get ATS, but doing so would require it to dig into funding allocated by the formula for other uses. This was unlikely to happen since most woreda tended to be strapped for resources. By contrast, a woreda with lower-than-average ATS would get a target ATS that was higher than its own ATS. It would be given a funding allocation in accordance with the target ATS, which would be sufficient for it to start raising its own ATS to meet the target value (thereby raising teaching quality). The rationale is similar in the case of the woreda PTR. A woreda was free to hire more than the number of teachers that was implied by its target PTR. Yet, doing so would require it to dig into funding allocated by the formula for other uses, which it was not very likely to do given the very limited amount of total resources available to most woreda with lower-than-average PTRs (too many teachers) thus had an incentive to gradually raise their PTRs. Woredas with higher-than-average PTRs (too few teachers) on the other hand, had tar- get PTRs that were lower than their own PTRs. They were given funding allocations deemed sufficient for them to hire a sufficiently large number of additional teachers, so as to attain their target PTRs. Even with the larger funding allocations, the region believed that some woredas would not hire the additional number of teachers needed; and so the performance agreements-- introduced in 2004/05--included the district PTR as a specified maximum target not to be breached. High-PTR woredas would thus, in effect, sign onto an agreement to hire enough new teachers so as to significantly reduce their PTRs. In this respect, the PTR dif- fers from most other cost drivers, which are not stated in the performance agreements. The PTR appears in the performance agreements, but only as a maximum target that is not to be breached. 100 World Bank Working Paper Rural woredas and urban administrations both received block grant resources from the region in SNNPR, but the allocations for each of these subgroups came from differ- ent pools. In making budget allocations, the region first decided how much funding to allocate to each of these pools, and how much to allocate to the regional bureaus. Next, the block grant allocations were made across urban administrations based on the total amount available from the "urban administrations" pool; a similar procedure was carried out for the "rural woredas" pool. The result was that there may have been some differen- tial treatment of urban administrations versus rural woredas, which the region justified by pointing to urban-rural differences in the prices of goods and services and in the costs of living, among other things. Regardless, for many "cost drivers" such as the average pupil-teacher ratio, the aim was still to equalize them across (and between) rural woredas and urban districts. APPENDIX C Regression Results on Education Outcomes Before and After Decentralization Table C1. School-Level Regressions Explaining Educational Outcomes Before and After Decentralization Dependent Variable Change in Change in Change Change Change in Ratio Grade 8 Repetition in Pupil- in Pupil- of Numbers of Independent Pass Rate Rate Teacher Section Teachers to Variable (%) (%) Ratio Ratio Sections Grade 8 pass rate, -0.81 2000/01 (%) (-13.15) Repetition rate -0.8 2000/01 (%) (-35.26) Pupil-teacher ratio, -0.81 2000/01 (-34.64) Pupil-section ratio, -0.54 2000/01 (-32.7) Ratio of number of -0.89 teachers to number (-40.75) of sections, 2000/01 Constant 66.24 9.67 52.6 32.7 0.93 (10.76) (30.97) (49.65) (16.57) (24.45) Number of observations 373 2244 2244 2244 2244 R2 0.7 0.58 0.78 0.38 0.64 Note: Dummy coefficients for each woreda are included but not reported. Dependent variables are changes between 2001 and 2004. Numbers in parentheses are t-values. 101 102 World Bank Working Paper Table C2. Woreda-Level Regressions Explaining Educational Outcomes Before and After Decentralization Dependent Variable Change in Change in Change Change Change in Ratio Grade 8 Repetition in Pupil- in Pupil- of Numbers of Independent Pass Rate Rate Teacher Section Teachers to Variable (%) (%) Ratio Ratio Sections Grade 8 pass rate, -0.78 2000/01 (%) (-10.68) Repetition rate -0.33 2000/01 (%) (-6.6) Pupil-teacher ratio, -0.35 2000/01 (-5.85) Pupil-section ratio, -0.16 2000/01 (-3.23) Ratio of number of -0.64 teachers to number (-13.34) of sections, 2000/01 Constant 57.84 4.66 22.4 12.6 0.76 (10.58) (22.4) (6.03) (3.18) (13.98) Number of observations 123 123 123 123 123 R2 0.46 0.23 0.48 0.08 0.76 Note: Each woreda is a unit of observation; the analysis uses data defined using "post- decentralization" boundaries (see the introduction to Chapter 4 for an explanation of this.) Dependent variables are changes between 2001 and 2004. Numbers in parentheses are t-values. APPENDIX D Estimating the Effects of Decentralization on the Delivery of Human Development Services in Ethiopia T o fully investigate the effects of decentralization on public services in Ethiopia, it is useful to start with outcome variables of interest and public expenditure. Did these variables change during the period in which decentralized took place? If outcome variables did not change, decentralization may not have been responsible for the effects often attributed to it. If expenditure did not change during the decentralizing period, it is possible that both the center and the periphery made identical choices or that the center did not really decentralize but only appeared to do so. Hence the first thing to do is to estab- lish that reform did in fact lead to nontrivial changes in both outcomes and financial flows over the period in question. If they did, did they improve or deteriorate? How did expen- diture patterns change following decentralization, both in aggregate and by sector? How were these changes related to changes in expenditure? Three hypotheses were used to test the effect of decentralization on the delivery of human development services: i. O1 =O . This simple t-test determines whether the means of outcome variables,O, ­ ­ ­ 2 changed in statistically significant ways over the period in question. The variables are subscripted by period, where 1 denotes 2001/02 (before decentralization) and 2 denotes 2002/03 (after decentralization). Significance implies that decentralization coincided with changes in outcome variables of interest, such as enrollment rates and Grade 8 test scores. Data limitations allowed this test to be performed only for the education sector. ii. E­1 = E­2. This t-test determines whether the means of expenditures variables, E­, changed in statistically significant ways with decentralization. The variables are sub- scripted in the same way as above. National aggregates as well as disaggregated data for education, health, agriculture, administration, and own revenues, are examined. 103 104 World Bank Working Paper iii. The differences in education outcome variables are set on the left-hand side of the equation to estimate the model Oj = + E j + Fj + Sj + Gj + j , where the decentralization variable is Ej = E2 - E1, that is, the degree to which decen- j j tralization changed education expenditure woreda by woreda; F measures resource availability at the woreda level; S is a vector of socioeconomic and demographic char- acteristics; and G is a vector of geographic characteristics, all indexed by woreda j. If decentralization improved educational outcomes--enrollment rates and test results-- by giving woredas more discretion over greater resources, one would expect to be positive and significant. But it is also possible that outcomes improved not due to higher expenditures but to better administration and targeting of a constant or lower level of expenditure. Table D1 reports regression results that test the above hypotheses. Woreda-level primary subsector regressions using data from SNNPR for 2001 and 2004 were run with two depen- dent variables: (i) the change in gross enrollment rate (between 2001 and 2004); and (ii) the change in the Grade 8 examination pass rate. The main purpose is to examine the relationship of education outcomes with each of the following: (i) changes in spending; and (ii) changes in quantity of the main educational input, i.e. teachers. Only selected regres- sion results are reported in the above table. Regressions (1-a) and (1-b) report results with the change in the gross enrollment rate (GER) as the dependent variable. The relevant teacher quantity measure here is the number of teachers per 1,000 residents rather than the pupil-teacher ratio, which is an endogenous variable whose value would rise if the enrollment rate (the dependent variable) rises. Regres- sion (1-a) shows that this variable is statistically significant and of the expected sign, implying that increased teacher quantity leads to improved enrollment rates. In regression (1-b) where the GER is again the dependent variable, the teacher quantity measure is excluded, and a spending measure is included: the change in the total educational spending per capita. While a more appropriate measure would have been the change in per capita spending in just the primary subsector, this would have limited the number of obser- vations substantially because a breakdown into primary versus other types of educational spending is not available for many woredas. The regression shows that the percentage change in the spending measure is statistically and positively associated with the change in the GER, at 10 percent level. Other independent variables in these regressions also provide useful information, sug- gesting that some groups of disadvantaged woredas have experienced larger changes in GERs than others. The sign of the "distance from zonal capital" is positive and statistically signifi- cant, indicating large GER improvements on the part of remote woredas, and confirms the findings from earlier descriptive analysis of improvements for the remote woredas with decen- tralization. Another indicator of development level is the degree of road access of woredas. The regressions show that woredas that are disadvantaged in this regard--i.e. that have fewer kilometers of seasoned roads per capita than others--had stronger GER improvements than other woredas. Regressions (2-a), (2-b) and (2-c) provide results with the change in the Grade 8 pass rate as the dependent variable. The sample used is restricted to woredas that started off with low pass rates to begin with--those whose average Grade 8 pass rate in 2001 was less than Achieving Better Service Delivery Through Decentralization in Ethiopia 105 Table D1. Effects of Changes in Expenditures on Educational Outcomes in the Primary Education Subsector: Results of Woreda-Level Regressions for SNNPR Change in Gross Enrollment Change in Grade 8 Percentage Rate 2001­2004 (%) Pass Rate 2001­20042 (%) (I-a) (1-b) (2-a) (2-b) (2-c) Gross Enrollment 0.12** 0.14* Rate 2001 (2.74) (1.86) Grade 8 -0.69** -0.96** -0.99** Percentage (-4.56) (-4.86) (-4.15) Pass Rate 2001 Repetition Rate 2001 Percentage Change 0.67** 6.36** in Total (1.97) (2.50) Education Spending Per Capita, 2001­2004 Percentage Change 1.46** in Primary (4.21) Education Spending Per Primary Student, 2001­2004 Change in Pupil- -0.002* Teacher Ratio, (-1.82) 2001­2004 Change in Number 8.16** of Teachers Per (2.49) 1000 Residents, 2001­2004 Change in Gross -0.04 Enrollment Rate, (-0.14) 2001-2004 Dummy for Food 1.51 1.55 5.99* 9.35** 8.11* Insecure (0.69) (0.66) (1.79) (2.23) (1.82) Woredas Dummy for 5.82 7.09 -2.17 -6.93 -2.02 Pastoral Woredas (1.30) (1.22) (-0.45) (-1.35) (-0.22) Dummy for Rural 2.72 6.20* 5.11 8.83** Woredas (0.77) (1.73) (1.42) (2.23) Distance from Zonal 0.024** 0.02** -0.017 -0.03 -0.03 Capital (km) (2.31) (2.15) (-0.65) (-0.84) (-0.89) Total Kms of -9.1** -8.7** 22.0** 24.0** 25.8** Seasoned Roads (-2.38) (-2.37) (5.14) (4.55) (3.77) in Woreda, Per 1,000 Residents (continued) 106 World Bank Working Paper Table D1. Effects of Changes in Expenditures on Educational Outcomes in the Primary Education Subsector: Results of Woreda-Level Regressions for SNNPR Change in Gross Enrollment Change in Grade 8 Percentage Rate 2001­2004 (%) Pass Rate 2001­20042 (%) (I-a) (1-b) (2-a) (2-b) (2-c) Constant 1.55 -1.55 42.7** 53.6** 67.2** (0.33) (-0.21) (4.36) (4.89) (4.5) Number of 92 80 60 54 40 Observations (Woredas) R-squared 0.39 0.3 0.47 0.46 0.47 *, **: statistical significance at 10 percent and 5 percent levels, respectively. Notes: 1. OLS regressions reported with robust standard errors; t-statistics in parentheses. 2. The sample for the regressions with the change in the Grade 8 pass rate as the dependent vari- able was restricted to the woredas with an average Grade 8 pass rate of less than 80 percent in 2001. 3. All regressions use data only covering primary schools. 4. All regressions use data defined using "pre-decentralization" boundaries (see the introduction to Chapter 4 as well as Appendix 4.C for an explanation of this). 80 percent. The results support the case that pass rates are much easier to improve when they are already low. The teacher quantity measure used here is the commonly used pupil-teacher ratio. Regression (2-a) shows that lower changes in Grade 8 pass rates are associated with larger changes in pupil-teacher ratios. Regressions (2-b) and (2-c) show a positive and statistically significant relationship between changes in Grade 8 pass rates and percentage changes in two different spending measures: total educational spending per capita and educational spending per primary student. The Grade 8 pass rate regressions also show that two disadvantaged groups benefit, again: food insecure woredas and to some extent rural woredas. However, the woredas with greater road access apparently benefited more from changes in the pass rate than other woredas. APPENDIX E Methodology and Technical Details for Data Analysis in Chapter 4 T he data used for the analysis in Chapter 4 were obtained directly from the regional governments of SNNPR and Oromiya. Data for 2001 and 2004 on the key variables were obtained at the woreda level. In addition, data were obtained for SNNPR at the primary school level, on test scores, number of teachers, number of students, number of sections and repetition rates. Woredas were classified as food-insecure if they had any food insecure households eli- gible for participation in the Productive Safety Nets Program or PSNP (regardless of what proportion of all woreda households were eligible). All food-insecure households are sup- posed to be eligible for PSNP food and cash transfers; as part of the PSNP, effort is spent on identifying food insecure segments of the population. As mentioned in the introduction to Chapter 4, between 2001 and 2004 there were many cases where a woreda (which will be referred to below as the "mother woreda") split into two or even three--thus creating one or even two new localities with separate administrative boundaries. Each of these new localities (each referred to below as a "daughter woreda/urban locality") became either a rural woreda or an urban administration with woreda status. Urban administrations did not exist before decentralization to the district-level. All quantitative comparisons between 2001 and 2004 were done using either "pre- decentralization woreda boundaries" or "post-decentralization boundaries," for both time periods. (The chapter specifies which approach was used for which portion of the analy- sis.) When "pre-decentralization boundaries" were used, it was necessary to aggregate 2004 data (for example, budget/expenditure data) for the "daughter" woredas, so as to be com- parable with the 2001 data for the "mother" woreda. "Post-decentralization boundaries" were used only for comparisons of outcome data (for example, test scores, pupil-teacher ratios) between 2001 and 2004. To derive outcome 107 108 World Bank Working Paper data for 2001 when using "post-decentralization boundaries," it was necessary to compute outcomes for the geographical areas in 2001 that would later become the "daughter" woredas in 2004. This was done by computing outcomes for the schools within each rele- vant geographical area. For example, between 2001 and 2004, Awassa woreda split into two parts: Awassa Zuria (rural) woreda and Awassa Town, the latter with its own separate urban administration. Computing outcomes for 2004 was straightforward; they were simply based on 2004 data for Awassa Zuria and Awassa Town. But computing outcomes for 2001 was less straightforward. To do this, it was necessary to use 2001 data and make separate computations for: (i) the schools that would become a part of Awassa Zuria by 2004; and (ii) the schools that would become a part of Awassa Town by 2004. Several places in the chapter refer to the "coefficient of variation." The coefficient of variation of a distribution is simply the standard deviation of the distribution divided by the mean. The use of the mean in the denominator corrects for any scale effect, and makes it possible to use the coefficient of variation to compare the variability of two very differ- ent distributions. By contrast, the standard deviation is not comparable across different distributions, and (unlike the coefficient of variation) is not independent of the unit of measurement used. References Adal, Yigremew. 2005. A Study on the Implementation of the Decentralization Policy in Ethiopia. Addis Ababa: Oxfam. Amhara National Regional State. 2005. "Intergovernmental Relationships: The Experience of the Amhara Regional State." Paper presented at the Government of Ethiopia-Donor IGR Workshop, March 23, Addis Ababa. Central Statistical Authority (CSA). 2005. "Intergovernmental Relationships: The Experience of the Amhara Regional State." Presentation at the Government of Ethiopia-Donor IGR Workshop, March 23, Addis Ababa. Christiansen, L., and H. Alderman. 2004. "Child Malnutrition in Ethiopia: Can Maternal Knowledge Augment the role of Income?" Economic Development and Cultural Change 52 (2):287­312. Dom, Catherine. 2004. "Decentralized Policy-Based Budgeting for Education." Study pre- pared for the Joint Review Mission of the Education Sector Development Program, Addis Ababa. Ethiopia House of Federation. 2007. "The New Federal Budget Grant Distribution For- mula." The Secretariat of the House of Federation. May, Addis Ababa. GovernmentofEthiopia.2002."GuidelineforOrganizationofEducationManagement,Com- munity Participation and Educational Finance." Ministry of Education, Addis Ababa. ------. 2005a. "Ethiopia: Fiscal Decentralization Strategy." Paper presented at the Gov- ernment of Ethiopia-Donor IGR Workshop, March 23, Addis Ababa. ------. 2005b. "Ethiopia Fiduciary Assessment." Produced with Task Team funded by the UK Department for International Development and the European Union. Government of Ethiopia and AED/BESO II. 2005a. "Second National Learning Assessment of Grade 4 Students." National Organization for Examinations and AED/BESO II, Addis Ababa. 109 110 World Bank Working Paper Government of Ethiopia and AED/BESO II. 2005b. "Second National Learning Assessment of Grade 8 Students." National Organization for Examinations and AED/BESO II, Addis Ababa. Government of Ethiopia/Ministry of Finance and Economic Development. 2005a. "Ethiopia: Building on Progress: A Plan for Accelerated and Sustained Development to End Poverty (PASDEP)." Addis Ababa. ------. 2005b. "Ethiopia Participatory Poverty Assessment 2004­05." Addis Ababa. GTZ-Selam Development Consultants. 2005. "Woreda-City Government Benchmarking Survey Part II." Addis Ababa. Halvorsen Tor and Robert L. Smith. 2005. "Delivery of Health and Education Services in Ethiopia: The Role of National Institutions." Presentation at the 8th Annual Norwegian Education Trust Fund Seminar. October 13th and 14th. Oslo, Norway. Heymans, C., and M. Mussa. 2004. "The Ethiopia IGR: Intergovernmental Fiscal Reforms in Ethiopia: Trends and Issues." World Bank, Washington, D.C. Joint Review Mission Report for Education. 2005. Addis Ababa. Jones, Gareth, Richard Steketee, Robert Black, Zulfiqar Bhutta, Saul Morris, and the Bel- lagio Child Survival Study Group. 2003. "How Many Child Deaths Can We Prevent This Year?" Lancet 362 (July). Kassa, Adane. 2003. "Hittosa Water Supply, Sanitation and Health Education Management, Arsi Zonal Administration, Oromiya Regional National Government, Ethiopia." Addis Ababa. Krishnan, P. 1996. "Private and Social Rates of Returns to Schooling." PHRD Study 9. Oxford CSAE, Oxford University, Oxford, UK. Lindelow, M., P. Serneels, and T. Lemma. 2004. "The Performance of Health Workers in Ethiopia: Results from Qualitative Research." World Bank, Washington, D.C. Mussa, M. 2004. "A Study of Capital Budget Utilization in the Education Sector." Joint Review Mission 2004 Preparatory Study. Addis Ababa, Ethiopia. ------. 2005a. "Progress of Decentralization and Capacity Building Programs in Oromiya Region." Addis Ababa. ------. 2005b. "Progress of Decentralization and Capacity Building Programs, SNNPR." Addis Ababa. ------. 2005c. "Civil Service Employment and Pay in Ethiopia." World Bank, Wash- ington, D.C. Mussa, M., Dinku, and Abdo. 2005. "Community Participation in Primary Education." Report to the Joint Review Mission for Education 2005. Addis Ababa. PANE (Poverty Action Network of Ethiopia). 2005. "Pro-Poor Services in Ethiopia. A Pilot Citizen Report Card." Addis Ababa. Resal Ethiopia. 2001. "Public Expenditure Management Aspects of Food Security." Final Report. Addis Ababa. Serneels, P., M. Lindelow, J. Garcia-Montalvo, and A. Barr. 2004. "An Honorable Calling? Findings from the First Wave of a Cohort Study with Final-Year Nursing and Medical Students in Ethiopia." Oxford University, Oxford, United Kingdom. SNNPR (Southern Nations, Nationalities and Peoples Region). 2005. "Overview of Inter- regional Government Fiscal Relation in the SNNPR." Presentation at the Government of Ethiopia-Donor IGR Workshop, Addis Ababa, March 23. Achieving Better Service Delivery Through Decentralization in Ethiopia 111 Suleiman, A. 1995. "Evaluating the Efficiency of Farmers in Ethiopia." Ethiopian Journal of Economics 35(2):47­66. USAID/WLI (U.S. Agency for International Development/World Learning Incorporated) 2004. "Basic Education Strategic Objective 2: Community Government Partnership Program." Annual Report. Washington, D.C. Weir, S., and J. Knight. 1996. "Household Demand for Schooling (Rural Survey)." PHRD Study 1. Oxford University, Centre for the Study of African Economies, Oxford. World Bank. 2004a. "Ethiopia: Draft Budget Update, 2004/05." Background note for the 2004 Joint Budget and Aid Review. A World Bank/DBS Joint Assessment, Washing- ton, D.C. ------. 2005a. "Education in Ethiopia: Strengthening the Foundation for Sustainable Progress." Report 28037-ET. Washington, D.C. ------. 2005b. "Ethiopia: A Country Status Report on Health and Poverty." 2 vols. Report 28963-ET, Washington, D.C. ------. 2005c. "Well-Being and Poverty in Ethiopia: The Role of Agriculture and Agency." Report 29468-ET, Washington, D.C. ------. 2005d. "Empowerment in Ethiopia: Issues and Options." Draft. Part of the Ethiopia Institutional Governance Review. ------. 2006a. "Ethiopia: Review of Public Finance." Draft. Washington, D.C. ------. 2006b. "Youth in Africa's Labor Market." Vol. II. Country Case Studies. Wash- ington, D.C. Eco-Audit Environmental Benefits Statement The World Bank is committed to preserving Endangered Forests and natural resources. We print World Bank Working Papers and Country Studies on 100 percent postconsumer recy- cled paper, processed chlorine free. The World Bank has formally agreed to follow the rec- ommended standards for paper usage set by Green Press Initiative--a nonprofit program supporting publishers in using fiber that is not sourced from Endangered Forests. For more information, visit www.greenpressinitiative.org. In 2007, the printing of these books on recycled paper saved the following: Trees* Solid Waste Water Net Greenhouse Gases Total Energy 264 12,419 96,126 23,289 184 mil. * Pounds Gallons Pounds CO2 Equivalent BTUs 40" in height and 6-8" in diameter Achieving Better Service Delivery Through Decentralization in Ethiopia is part of the World Bank Working Paper series. These papers are published to communicate the results of the Bank's ongoing research and to stimulate public discus- sion. Ethiopia has made major strides in improving its human development indicators in the past 15 years, achieving significant increases in the coverage of basic education and health services in a short period of time. Improvements took place during a period of massive decentralization of fiscal resources, to the regions in 1994 and to woredas in 2002­03. The devolution of power and resources from the federal and regional governments to woredas appears to have improved the delivery of basic services. Surveys of beneficiaries reveal that they perceive that service coverage and quality have improved. Beneficiary satisfaction has increased markedly in education, and less conspicuously in water and health services. In the south, the decentralization to woredas in 2002­03 tended to narrow differences in per capita expenditures on education and health across woredas. Decentralization disproportionately favored woredas that are remote (more than 50 kilometers from a zonal capital), food-insecure, and pastoral, suggesting that decentralization has been pro-poor. Decentralization also narrowed the gap in educational outcomes between disadvantaged and better-off woredas, especially in the south. Pastoral, food-insecure, and remote woredas gained in terms of the educational outcomes examined (gross enrollment rates, grade 8 examination pass rates, repetition rates, pupil-teacher ratios, and teacher-sec- tion ratios). World Bank Working Papers are available individually or on standing order. Also available online through the World Bank e-Library (www.worldbank.org/elibrary). ISBN 978-0-8213-7382-8 THE WORLD BANK 1818 H Street, NW Washington, DC 20433 USA Telephone: 202 473-1000 Internet: www.worldbank.org SKU 17382 E-mail: feedback@worldbank.org