A W O R L D B A N K C O U N T R Y S T U D Y Education in Rwanda Rebalancing Resources to Accelerate Post-Conflict Development and Poverty Reduction THE WORLD BANK A W O R L D B A N K C O U N T R Y S T U D Y Education in Rwanda Rebalancing Resources to Accelerate Post-Conflict Development and Poverty Reduction THE WORLD BANK Washington, D.C. Copyright © 2004 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: October 2003 printed on recycled paper 1 2 3 4 05 04 03 World Bank Country Studies are among the many reports originally prepared for internal use as part of the continuing analysis by the Bank of the economic and related conditions of its develop- ing member countries and to facilitate its dialogues with the governments. Some of the reports are published in this series with the least possible delay for the use of governments, and the academic, business, financial, and development communities. The typescript of this paper therefore has not been prepared in accordance with the procedures appropriate to journal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank cannot guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply on the part of the World Bank any judgment of the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. The World Bank encourages dissemination of its work and normally will grant permission for use. Permission to photocopy items for internal or personal use, for the internal or personal use of specific clients, or for educational classroom use, is granted by the World Bank, provided that the appropriate fee is paid. Please contact the Copyright Clearance Center before photocopying items. Copyright Clearance Center, Inc. 222 Rosewood Drive Danvers, MA 01923, U.S.A. Tel: 978-750-8400 · Fax: 978-750-4470. For permission to reprint individual articles or chapters, please fax your request with complete information to the Republication Department, Copyright Clearance Center, fax 978-750-4470. All other queries on rights and licenses should be addressed to the World Bank at the address above, or faxed to 202-522-2422. ISBN: 0-8213-5610-0 eISBN: 0-8213-5611-9 ISSN: 0253-2123 Library of Congress Cataloging-in-Publication Data Education in Rwanda: rebalancing resources to accelerate post-conflict development and poverty reduction. p. cm. -- (A World Bank country study) Summary in French. Includes bibliographical references. ISBN 0-8213-5610-0 1. Education--Rwanda. 2. Education--Rwanda--Statistics. I. World Bank. II. Series. LA2090.R95E37 2003 370'.967571--dc21 2003057688 TABLE OF CONTENTS Foreword, by Romain Murenzi xi Foreword, by Birger Fredriksen xiii Abstract xv Acknowledgments xvii Abbreviations, Acronyms, and Data Notes xix Executive Summary 1 Résume Analytique 9 1. Rwanda: Demography, Economy, and the Government Budget 19 The Demographic Context 21 Macroeconomic Conditions 24 Overall Patterns of Government Finance 26 Public Spending on Education 27 Conclusion 29 2. Enrollment Trends and Patterns of Student Flow 31 Aggregate Enrollments by Level of Education 31 Trends in Educational Coverage 34 From Cross-Sectional Indicators of Coverage to Student Flow Patterns 37 Student Flow Patterns in Primary Schooling 38 Student Flow Patterns in Secondary Schooling 47 Policy Perspectives on Management of Student Flow 49 Conclusion 53 3. Education Finance 55 National Spending on Education 55 A Closer Look at Public Spending on Education 59 Public Spending per Student 66 Policy Implications 71 Conclusion 73 4. Socioeconomic Disparities in Education 75 Overview of Participation Rates 76 Educational Participation Rates among Orphans 79 Disparities in Student Flow Patterns 81 Distribution of Public Spending on Education 85 Policy Implications 88 Conclusion 90 5. Service Delivery in Primary Education 91 Overview of the Supply of Services 91 Teacher Allocation across Schools 96 iii IV TABLE OF CONTENTS Economies of Scale in Service Delivery 100 Student Learning 103 Policy Implications and Conclusion 108 6. Service Delivery in Secondary Education 111 Overview of the Supply of Services 112 Teacher Qualifications, Utilization, and Deployment 118 Economies of Scale in Service Delivery 124 Examination Results and Their Correlates 128 Policy Implications 131 Conclusion 132 7. Higher Education 135 Structure of the System 136 Study Abroad on Government Scholarships 143 Staffing Patterns in Public and Private Institutions of Higher Education 147 Costs of Service Delivery and Student Finance 151 Student Flow Efficiency and Output of Graduates 158 Policy Implications 161 Conclusion 164 8. Education and the Labor Market 165 Employment Structure, Educational Attainment of Workers, and Returns to Education 166 Output of Graduates and Their Absorption into the Workforce 170 Policy Implications 174 Conclusion 176 Statistical Appendix 177 References 223 TABLES 1-1: Population Size, Rwanda, 1970­2002, and Poverty Rates and Selected Health Indicators, 1970­2000 22 1-2: Orphans and Children Living apart from Biological Parents, Rwanda, circa 2000 23 1-3: Economic Performance of Rwanda and Selected Countries, 1980­2000 25 1-4: Government Revenue, Rwanda, 1981­2001 26 1-5: Government Spending, Rwanda, 1981­2001 28 1-6: Government Spending on Education, Rwanda, 1981­2001 29 2-1: Gross Enrollment Ratios (GERs) by Level of Education, Rwanda, 1991­92 and 2000­01 35 2-2: Coverage of the Education System in Rwanda and Selected Countries, Late 1990s­2000 37 2-3: Percentage Entry Rate to Grade 1, Rwanda, 1991­92 and 2000­01 39 2-4: Percentage Survival Rates in Primary Schooling, Rwanda, 1991­92 and 2000­01 41 TABLE OF CONTENTS V 2-5: Percentage Repetition Rates in Primary Schooling, Rwanda, 1990­91 and 2000­01 43 2-6: Summary Indices of Student Flow Efficiency in Primary Schooling, Rwanda, 1990­91 and 2000­01 44 2-7: Secondary School Transition and Survival Rates, Rwanda, 1991­92 and 2000­01 47 2-8: Percentage Repetition Rates in Secondary Schooling, Rwanda, 1990­91 and 2000­01 48 2-9: Summary Index of Student Flow Efficiency in Secondary Schooling, Rwanda, 1990­91 and 2000­01 49 2-10: Distribution of Primary Schools and New First-Graders by Highest Grade of Instruction Offered by the School, Rwanda, 2000­01 50 2-11: Grade-Specific Enrollments in Primary and Secondary Education, Rwanda, 2000­01 51 3-1: Level and Distribution of Public Spending on Education, Rwanda, 1982­2001 56 3-2: Current Public Spending on Education through the Ministry of Education and Other Ministries or Government Agencies, Rwanda, 1999­2001 58 3-3: Household Spending on Education, Rwanda, circa 2000 59 3-4: Distribution of Current Public Spending on Education, Rwanda, 1999­2001 60 3-5: Numbers of Teachers and School-Level Administrative Staff in Government-Financed Primary and Secondary Schools, Rwanda, 1999 61 3-6: Distribution of Public School Staff by Salary Grade and Average Salaries, Rwanda, 1999 62 3-7: Estimated Current Public Spending on Education by Function and Level, Rwanda, 1999 63 3-8: Functional Distribution of Current Public Spending on Education, Rwanda, 1999 65 3-9: Public Spending per Student by Level of Education, Public Sector, Rwanda, 1999 67 3-10: Level and Composition of per Pupil Public Spending in Public Primary and Secondary Schools, Rwanda, 1999 68 3-11: Current Public Spending per Student, Rwanda, 1999, and Selected Country Groups, 1990s 69 3-12: Public Sector Teacher Salaries and Pupil-Teacher Ratios, Rwanda, 1999, and Selected Country Groups, 1990s 70 3-13: Take-Home Pay of Public Sector Teachers and Other Workers in Relation to Per Capita GDP, Rwanda, 2001 72 4-1: Gross Enrollment Ratios by Province, Locality, Gender, and Income Group, Rwanda, 1992­2000 77 4-2: Percentage Share of Salaried Workers among Parents of Students at Two Public Higher Education Institutions and in the Population of Rwanda, 2000 79 4-3: Percentage of Children Ages 7­12 Enrolled in Primary School by Orphanhood Status, Rwanda, 1998­99 80 4.4: Prevalence of Orphanhood among Primary, Secondary, and Higher Education Students, Rwanda, circa 2000 81 4-5: Selected Student Flow Indicators by Gender, Locality, and Income Group, Rwanda, circa 2000 82 VI TABLE OF CONTENTS 4-6: Regression-Predicted Indicators of School Progression during Two Consecutive Years in a Cohort of Children Who Were Ages 7­12 in 1998, by Population Group, Rwanda, 1998­2000 84 4-7: Share of Public Spending on Education Benefiting the Poorest and Richest Population Quintiles, Rwanda, 2000, and Selected Sub-Saharan African Countries, 1990s 88 4-8: Gender Differences in Performance on National End-of-Cycle Examinations, Rwanda, circa 2002 90 5-1: Percentage Distribution of Primary Schools and Pupils by Sector, Rwanda, 2000 92 5-2: Characteristics of State, Libre Subsidié, and Private Primary Schools, Rwanda, 2000 93 5-3: Characteristics of Primary School Teachers, Rwanda, 1999 95 5-4: Accessibility of Primary Schools and Problems with Schooling Reported by Currently Enrolled Pupils, Rwanda, 2001 96 5-5: Regression Estimates of the Relation between Numbers of Teachers and Pupils, Public Primary Schools, Rwanda, 2000 98 5-6: Regression Estimates of the Relation between Numbers of Teachers and Pupils in Public Primary Schools by Province, Rwanda, 2000 99 5-7: Regression Estimates of the Relation between Total Personnel Costs and Enrollments across Schools, Rwanda, 2000 102 5-8: Percentage Distribution of Primary Schools and Cost per Pupil by Size of Enrollment, Rwanda, 2000 103 5-9: Primary School Leaving Examination Results by Type of School, Rwanda, 1999 104 5-10: Correlates of School-Level Pass Rates on the Primary School Leaving Examination, Rwanda, 1999 106 5-11: Regression Estimates of Provincial Differences in Examination Pass Rates, Controlling for Differences in per Pupil Spending on Personnel, Rwanda, 1999 108 6-1: Number and Percentage Distribution of Types of Secondary School, Rwanda, 2001­02 113 6-2: Selected Characteristics of State, Libre Subsidié, and Private Secondary Schools, Rwanda, 1999­2001 114 6-3: Percentage Distribution of Schools by Level and Number of Instructional Streams Offered, Rwanda, 2000­01 115 6-4: Percentage Distribution of Secondary Students by Cycle and Stream and by School Type, Rwanda, 2000­01 116 6-5: Number of Schools Offering Upper Secondary Programs and Average Enrollments per School by Field, Rwanda, 2000­01 117 6-6: Characteristics of Teachers by Type of Secondary School, Rwanda, 1999­2000 119 6-7: Distribution of Teachers by Level of Classes Taught, and Teaching Loads and Student-Teacher Ratios, Public Secondary Schools, Rwanda, 1999­2000 120 6-8: Distribution of State and Libre Subsidié Secondary School Teachers by Educational Attainment and Level of Classes Taught, Rwanda, 1999­2000 121 6-9: Regression Estimates of the Relation between Numbers of Teachers and Students, Public Secondary Schools, Rwanda, 1999­2000 122 TABLE OF CONTENTS VII 6-10: Relation between Total Cost of Personnel and Enrollments in Public Secondary Schools, Rwanda, 1999­2000 125 6-11: Size Distribution of Enrollments in Public Secondary Schools and Simulated Cost per Student, Rwanda, 1999­2000 127 6-12: Results on the National Examination at the End of the Tronc Commun Cycle, Rwanda, 1999­2000 128 6-13: Correlates of School-Level Performance on the National Examination at the End of the Tronc Commun Cycle, Rwanda, 1999­2000 130 7-1: Number of Institutions of Higher Education, Overall Enrollments, and Share of Students in Private Sector, Rwanda, 1960s­Present 137 7-2: Institutional Distribution of Students in Higher Education, Share of Female Students, and Scores of Entrants on National Examination at End of Secondary Cycle, Rwanda, 2000­01 139 7-3: Number and Distribution of Students in Public and Private Higher Education Institutions, Rwanda, 2000­01 141 7-4: Trends in Enrollments and Distribution by Field, Université Nationale du Rwanda, Selected Periods 142 7-5: Higher Education Students on Overseas Government Scholarships, Rwanda, 1967­2002 143 7-6: Number and Distribution of Rwandan Students on Overseas Government Scholarships by Field and Level of Study, 2000­01 145 7-7: Number of Rwandan Students on Overseas Government Scholarships and Their Host Countries, 1984­85 and 1999­2002 146 7-8: Number and Composition of Higher Education Faculty by Institution, Rwanda, Selected Years 148 7-9: Number of Nationals on the Université Nationale du Rwanda (UNR) Faculty Studying Overseas on Government Scholarships as of February 2002 149 7-10: Number of Faculty and Actual and Simulated Staffing Ratios in Public and Private Higher Education Institutions, Rwanda, 2000­01 150 7-11: Student-Faculty Ratios by Field of Study, Université Nationale du Rwanda (UNR), 2000­01 152 7-12: Cost of Service Delivery per Student in Public Higher Education Institutions, and Distribution by Field of Study at the UNR, 2000 154 7-13: Per Student Costs of Service Delivery, Université Nationale du Rwanda, 2000 155 7-14: Number and Share of Students Receiving Bursaries and Various Student Services, Selected Public Institutions, Rwanda, 1994­2001 156 7-15: Average Annual Value of Student Bursaries, Cash Transfers, and Deductions for Student Services in Public Higher Education, Rwanda, 2000­01 157 7-16: Fees and Percentage of Students Receiving Government Bursaries, Private Higher Education Institutions, Rwanda, 1998­2002 159 7-17: Promotion, Repetition, and Survival Rates, Selected Higher Education Institutions, Rwanda, circa 2000 160 7-18: Number of Graduates by Institution and Broad Field of Specialization, Rwanda, 1980s and circa 2000 162 VIII TABLE OF CONTENTS 8-1: Selected Data on Population, Labor Force, Employment, and Related Indicators, Rwanda, 1991 and 2000 166 8-2: Distribution of Employment by Sector, Rwanda, 1991 and 2000 167 8-3: Number and Distribution of Workers by Type of Employment and Job, Rwanda, 1991 and 2000 168 8-4: Percentage Distribution of the Employed Population by Educational Attainment, Rwanda, 1991 and 2000 169 8-5: Average Salaries and Years of Schooling of Wage Earners in the Formal and Informal Sectors, Rwanda, 2000 169 8-6: Rates of Return to Education by Level and Sector of Employment, Rwanda, 2000 170 8-7: Percentage Distribution of Wage Earners by Educational Attainment Relative to Workers in the Same Jobs, Rwanda, 2000 172 8-8: Unemployment Rates by Educational Level and Age, Rwanda, 1991 and 2000 173 A1-1: Government Revenue and Expenditure, Rwanda, 1980­2001 178 A1-2: Current and Capital Public Spending on Education, Rwanda, 1981­2001 179 A2-1: Enrollments by Level of Education, Rwanda, Selected Years, 1970­2001 180 A3-1: Public Spending on Education by Level, Rwanda, 1971­2001 181 A3-2: Public Spending on Education, Rwanda, Selected Years, 1985­2001 183 A3-3: Itemized Household Spending on Education, Rwanda, circa 2000 184 A3-4: Current Spending on Education by Level of Education and Function, Rwanda, 1999­2001 185 A3-5: Per Student Spending in Public Primary and Secondary Schools, Rwanda, 1999 186 A3-6: Mincerian Earnings Function, Excluding Wage Earners Working for Educational Institutions, Rwanda, 2001 187 A4-1: School Attendance Status of a Cohort of Children Ages 7­12 in 1998, Rwanda 188 A4-2: School Progression Rates during Two Consecutive Years among Children Ages 7­12 by Orphanhood Status, Rwanda, 1998­2000 190 A4-3: Regression Estimates of School Attendance Status in a Cohort of Children Ages 7­12 in 1998, Rwanda, 1998­2000 191 A4-4: Number and Characteristics of Candidates Passing End of-Cycle Examinations for Primary and Secondary Education, Rwanda, 2001 and 2002 192 A5-1: Teacher's Educational Attainment, Qualifications, and Remuneration, Rwanda, 2001 193 A5-2: Pupil-Teacher Ratios in Public (State and Libre Subsidié) Primary Schools by Province, Rwanda, 2000 195 A5-3: Regression Estimates of the Relation between Numbers of Teachers and Pupils in Public (State and Libre Subsidié) Primary Schools, with Provincial Dummy Variables, Rwanda, 2000 196 A5-4: Regression Estimates of the Correlates of School-Level Pass Rates on the National Primary School Leaving Examination, Public Schools, Rwanda, 1999 197 TABLE OF CONTENTS IX A6-1: Distribution of Secondary Schools by Level and Type of Instruction Offered, Rwanda, 2000­01 199 A6-2: Number of Secondary School Teachers by Qualifications and Level of Classes Taught, Rwanda, 1999 200 A6-3: Regression Estimate of the Relation between Numbers of Teachers and Students, Public Secondary Schools, Rwanda, 1999­2000 200 A6-4: Correlates of End-of-Cycle Tronc Commun National Examination Results at School Level, Rwanda, 1999 201 A7-1: Enrollments in Higher Education by Institution, Rwanda, 1963­2002 202 A7-2: Enrollments in Public and Private Higher Education Institutions by Gender, Rwanda, Selected Years, 1984­2002 204 A7-3: Distribution of Students by Field of Study, Selected Higher Education Institutions, Rwanda, 2000­01 206 A7-4: Enrollments by Field of Study, Université Nationale du Rwanda, Selected Years, 1982­2002 208 A7-5: Numbers of Rwandan Students on Government Overseas Scholarships and Enrolled Locally, Selected Years, 1967­2002 210 A7-6: Number of Rwandan Students on Government Overseas Scholarships by Host Country, Selected Years, 1984­2002 211 A7-7: Academic Fees, Welfare, and Travel Costs Paid by the Rwandan Government for Rwandan Students on Government Overseas Scholarships, circa 2002 213 A7-8: Number of Higher Education Faculty by Institution and Nationality, Rwanda, 1985­2001 215 A7-9: Current and Proposed Arrangements regarding Student Bursaries at the Université Nationale du Rwanda 217 A8-1: Mincerian Earning Functions among Wage Earners, Rwanda, 1999­2001 218 A8-2: Mincerian Earning Functions among Wage Earners by Sector of Employment, Rwanda, 1999­2001 220 FIGURES 1-1: Adult HIV/AIDS Prevalence Rates and Share of Orphans among Children Ages 7­14, Rwanda and other East African Countries, Circa 2000 24 2-1: Enrollment Trends in Primary Schooling, Rwanda, 1971­2001 32 2-2: Enrollment Trends in Secondary Education, Rwanda, 1996­2001 33 2-3: Enrollment Trends in Higher Education, Rwanda, 1975­2002 34 2-4: Relation between Educational Attainment and Probability of Being Literate, Rwanda, 2001 38 2-5: Grade-Specific Enrollment Rates in Primary Schooling, Rwanda, 1991­92 and 2000­01 42 2-6: Primary School Student Flow Indicators in Rwanda and in Selected Sub-Saharan African Countries, circa 2000 46 2-7: Efficiency of Student Flow in Rwanda and in Selected Sub-Saharan African Countries, circa 2000 46 3-1: Relation between Primary School Completion Rates and Public Spending on Primary Education, Low-Income Countries, circa 2000 57 X TABLE OF CONTENTS 3-2: Functional Distribution of Current Public Spending on Education, Rwanda, 1999 66 4-1: Gross Enrollment Ratios in Lower and Upper Secondary Education by Income Quintile, Rwanda, 2000 78 4-2: Primary School Enrollment Rates in a Cohort of Children across Income Groups, Rwanda, 2000 83 4-3: Share of Cumulative Public Spending on Education Benefiting the 10 Percent Best Educated in a Generation, Rwanda and Other Sub-Saharan African Countries, circa 1998 86 5-1: Institutional Composition of Primary Schools by Province, Rwanda, 2002 92 5-2: Pupil-Teacher Ratios in Public Primary Schools in Rwanda and Other Sub-Saharan African Countries, circa 1999 94 5-3: Relation between Number of Pupils and Number of Teachers, Public Primary Schools, Rwanda, 2000 97 5-4: R2 Values of Regressions Relating Number of Teachers to Number of Pupils across Schools, Rwanda and Selected Sub-Saharan African Countries, circa 2000 98 5-5: Relation between Personnel Cost per Pupil and School Size, Public Schools, Rwanda, 1999 101 5-6: Economies of Scale in Public Primary Education, Rwanda, 2000 102 5-7: Relation between Examination Results and per Pupil Cost of Personnel, Public Schools, Rwanda, 1999 105 6-1: Number and Institutional Composition of Secondary Schools by Province, Rwanda, 2000­01 112 6-2: Relation between Numbers of Students and Teachers, State and Libre Subsidié Schools, Rwanda, 2000 122 6-3: R2 Values of Regressions Relating Numbers of Teachers and Students across Schools by Level of Education and School Type, Rwanda, circa 2000 123 6-4: Relation between Number of Students and Cost per Student, Tronc Commun and Upper Secondary Cycles, Rwanda, 1999 124 6-5: Simulations of Economies of Scale in Secondary Education, Tronc Commun and Upper Secondary Cycles, Rwanda, 1999­2000 126 6-6: Relation between Personnel Costs per Student and Results of National Examination at End of the Tronc Commun Cycle, Public Secondary Schools, Rwanda, 1999­2000 129 7-1: Arrangements for Student Finance in Public Higher Education in Rwanda up to January 2002 156 8-1: Relation between Per Capita Income and Coverage in Higher Education, Low-Income Countries, circa 1998, and Rwanda, 1997­98 and 2001­02 174 A4-1: Cumulative Shares of Public Spending on Education Benefiting a Hypothetical Cohort, by Educational Attainment 222 BOX 1-1: Rwanda at a Glance 20 FOREWORD T his Country Study represents an important contribution to the strategic planning process for the development of education in Rwanda. It provides informative analyses of national trends up to 2001 and offers revealing comparisons with other countries in similar situations. The report has already been used to inform Rwanda's draft Education Sector Strategic Plan of March 2003 and was discussed with interest during the joint review of the education sector held in Kigali in April 2003. I am sure that the study will continue to provide an important reference point for the ongoing Sector Wide Approach Program (SWAP) in the education planning and management process. By way of an update, it is important to note that several new policy initiatives have been launched since the data for the report were collected in 2001. In particular, the 2002 Education Sector Policy makes clear the government's commitment to providing quality basic education to all Rwandan children. We have proposed steps, including the abolition of school fees, to make this goal a reality. We have also taken action to reduce the high repetition rates noted in the study, and I am pleased to report that they have declined from 34 percent in 2000­01 to 17 percent in 2002­03. We hope that this salutary trend will continue. To meet the high pressures from the rising numbers of primary school graduates and the need to provide young people with the basic skills required for life and work, we intend to put in place strategies to ensure that lower secondary school education is offered to as many children as possi- ble. Great efforts are being made to raise the quality of education through the provision of trained teachers and pedagogical materials. We have introduced a radical reform of higher education financing to reduce government subsidies to university students, which will release additional resources for investment in the primary and secondary sectors. Finally, the teaching of science and technology, especially information communication technology, will receive special attention at all levels of our education system. All in all, I am confident that Rwanda will continue to make good progress toward achieving the United Nations Millennium Development Goals in education and that we will be able to pro- vide the country with an educated and skilled workforce to build a prosperous and peaceful future for all. I would like to take this opportunity to thank all those who have in one way or another made this study a success. Special gratitude goes to the World Bank team led by Jee-Peng Tan, lead economist, and the Rwandan team led by Claver Yisa, director of planning, for their commendable work. Prof. Romain Murenzi Minister of Education, Science, Technology, and Scientific Research Kigali, Rwanda June 2003 xi FOREWORD T his study is one in a series of country status reports (CSRs) on education being prepared by World Bank staff in collaboration with national teams from various Sub-Saharan African coun- tries. The immediate objective of the CSRs is to enhance the knowledge base for policy develop- ment in the education sector. More broadly, CSRs create a basis for engaging a diverse audience in dialogue on education sector policies and for developing a shared vision for the future. These processes have become increasingly important as governments renew their commitment to reduce poverty and the international donor community pledges to provide the needed financial assistance. In this new dynamic of international development assistance, the World Bank has oriented its corporate mission toward supporting governments in the fight against poverty, making that task its defining priority. In practice the strategy is articulated through a two-pronged approach: sup- port for policies that accelerate economic growth, combined with explicit measures to promote a more equitable distribution of the benefits from growth. Human development, through invest- ments in education as well as in health, plays a central role in both processes. This fact receives recognition in the emphasis placed on the two sectors in the debt relief arrangements under the Heavily Indebted Poor Countries (HIPC) Initiative. Governments have also invariably treated education and health as priority sectors in their Poverty Reduction Strategy Papers (PRSPs). In education, two specific goals stand out: ensuring that all children receive basic schooling of adequate quality, and managing the production of graduates at post-basic levels, in terms of num- bers and skills mix, so as to match the demand for skilled labor. To fulfill these objectives, an infu- sion of resources, from both domestic and international sources, may well be needed in a number of countries. In all countries, an equally important issue is to ensure that currently available resources are used efficiently and equitably. Some countries have already embarked on reforms to improve performance. In the 1990s, for example, policies to encourage more effective use of teachers through multigrade teaching and double shifting were put in place to boost coverage. More remains to be done, particularly in tackling the structural constraints that encumber the education system. The challenges include defining an appropriate division of responsibility between the public and private sectors for financ- ing and delivering education services; setting sustainable levels of teacher remuneration; and creat- ing institutional arrangements for effective system management. Reforms in these areas are key to boosting learning outcomes, which in turn will be crucial for progress toward Education For All by 2015. A first step is to develop a country-specific knowledge base that sheds light on the key struc- tural weaknesses in the education system. In a large number of Sub-Saharan African countries, that knowledge base is sparse, reflecting the systematic neglect of analytical work in the past. To recapi- talize it, the World Bank has initiated the preparation of CSRs, using a standard format to consoli- date the available information in a policy-relevant manner. Two features characterize the CSR, one pertaining to its technical content, the other to the process by which it is prepared. With regard to content, five aspects are worth mentioning. First the CSR pays close attention to issues of equity and the distribution of public resources for educa- tion, given the importance of these topics in the HIPC and PRSP context. Second, the analysis relies on commonly available administrative data, as well as household surveys--an approach that has helped improve the consistency and robustness of the statistical results. Third, the CSR focus- es on outcomes by emphasizing indicators beyond the usual gross and net enrollment ratios. In particular, it documents the schooling careers of children as reflected in the shares of each cohort that enter grade 1 and attain the various grades on the educational ladder. This more detailed approach has highlighted socioeconomic, gender, and geographic disparities in schooling and has clarified the sources of the disparities. Fourth, the CSR uses school-level data to assess the scope for improving service delivery to the poor. Finally, the CSR makes use of data on student learning, xiii XIV FOREWORD where available, to gauge the performance of the education system in this important domain and to identify cost-effective measures for progress. As for process, the CSR is a product of a partnership between the World Bank and a national team from the country that is the subject of the report. This arrangement promotes training and capacity building and leads to a shared understanding of the policy issues and the context in which they present themselves. It is important to note that CSRs are diagnostic documents whose pur- pose is to help identify policy questions rather than to offer solutions and make recommendations. The process of policy development is more appropriately led by the national team, and the avail- ability of a CSR provides a good foundation for disseminating the findings and stimulating a broad national dialogue on the way forward. The document also serves as a basis for preparing a rational and defensible plan for sector development to inform the country's medium-term budget planning exercise. These processes are already under way in Rwanda, as exemplified by the use of the CSR at the April 2003 joint review meeting that engaged Rwandan policymakers and donor representatives in assessing the country's proposed sector strategic development plan. On the World Bank's part, the document is facilitating ongoing discussions on the design of a new budget support credit as part of the Bank's country assistance strategy. The publication of the CSR for Rwanda is intended to institutionalize our collective knowl- edge on education in Rwanda and to share that knowledge as widely as possible. It is my hope that as new knowledge emerges in the course of implementing the country's poverty reduction strategy, the CSR will be updated to track progress and distill lessons learned that may have broader application in other African countries. Birger Fredriksen Senior Education Advisor Africa Region The World Bank ABSTRACT T en years after the 1994 genocide in which an estimated 10 percent of the country's popula- tion perished, Rwanda's devastated education system is now back on its feet. Classrooms have been repaired and new ones built; teachers who fled the mayhem have been reintegrated into the teaching force; arrears in teacher pay have been cleared up; a Genocide Fund has been created specifically to assist orphans; the higher education system has been diversified; and new arrange- ments for student finance at the tertiary level have been introduced. These successes notwith- standing, the task of transforming the rapid recovery into sustained progress has only just begun. A priority will be to ensure that all Rwandan children are able to complete a full course of primary schooling of reasonable quality and that expansion at postprimary levels proceeds at a pace com- mensurate with the labor market's capacity to absorb highly educated job seekers. Achieving these goals will require a rebalancing of public spending in favor of primary education--a shift that will call for continued heavy reliance on private financing at postprimary levels and reforms to reduce unit costs and student bursaries in public higher education. In addition, efforts must be made to direct public subsidies to the most vulnerable children, such as double orphans, so that they can afford to attend school. Policies to improve services delivery will also be critical to ensure that teacher deployment is consistent across schools according to the size of enrollments, that peda- gogical materials are available in classrooms, and that instructional hours are increased, particularly in the first three grades, in combination with possible reforms in teacher recruitment practices. Progress will depend, too, on better supervision and incentives designed to encourage more effec- tive classroom practices and so improve student flow and student learning throughout the system. xv ACKNOWLEDGMENTS T his report is the result of a collaborative effort by the government of Rwanda and the World Bank to deepen understanding of the current status of education in Rwanda. The Rwandan government team was led by Claver Yisa, Director of Research and Planning of Education in the Ministry of Education (MINEDUC), under the overall leadership of Prof. Romain Murenzi, Min- ister of Education, and Eugène Munyakayanza, formerly MINEDUC's Secretary General and now its Minister of State for Primary and Secondary Education. The World Bank team was led by Jee- Peng Tan under the overall guidance of the following Africa Region managers: Emmanuel Mbi, Laura Frigenti, and Arvil Van Adams. The Rwandan team initially consisted of Faustin Habineza, Joseph Matsiko, Herman Musa- hara, and Ernest Rutungisha. As the work progressed, additional support was received from vari- ous individuals working in the Ministry of Education, including Aviti Bagabo, Sampson Kagorora, Leonard Manzi, Catherine Mukankuranga, Susy Ndaruhutse, Joseph Le Strat, and Patrick Rwabidadi. Many people outside the ministry also offered invaluable assistance with data collec- tion and preparation. They included François Rwambonera, Conseil protestant du Rwanda; A. Aloys Guillaume, Sécretariat national de l'enseignement catholique; Joseph Murekeraho and his staff at the National Examination Council of Rwanda; Straton Nsanzabaganwa, Ministère de l'ad- ministration local et des affaires sociales; John Ruzibuka and François Katangulia, Office national de la population; Jacques Gashaka and Omar Sarr, Enquête intégrale sur les conditions de vie des ménages au Rwanda; Jean-Baptiste Rulindamanywa, Camille Rwemayire, and Evariste Nzabanita, Ministère de la fonction public et du travail; and Patrick Jondoh, UNICEF. The World Bank team consisted of Gérard Lassibille, Jean-Bernard Rasera, and Kiong Hock Lee. Valuable assistance was provided by Susan Opper, task team leader for education in Rwanda, and by peer reviewers Robert Prouty and Quentin Wodon. In addition to the contributions made by the people mentioned above, helpful feedback and support were received from the following interlocutors in Rwanda: Narcissse Musabeyezu, Direc- tor of Primary Education; Emma Rubagumya, Director of Secondary Education; Callixte Kayisire, Director of Higher Education; John Rutayisire, Director of the National Curriculum Develop- ment Centre; Béatrice Mukabaranga, Kigali Institute of Education; and Silas Lwakabamba, Kigali Institute of Science, Technology and Management, and from the following World Bank col- leagues: Jaap Bregman, Edward Brown, Birger Fredriksen, Raju Kalidindi, Toni Kayonga, Alain Mingat, Chukwuma Obidegwu, Mamy Rakotomalala, and Guido Rurangwa. Contributions from Desmond Bermingham, Michael Delens, and William O'Hara of the U.K. Department for Inter- national Development are also gratefully acknowledged. In the course of preparing the report, members of the World Bank team visited Rwanda, and members of the Rwandan team traveled to Washington, D.C. These visits were greatly facilitated by the cheerful and reliable logistical support offered by Bathilde Jyulijyesage, Antoinette Kamanzi, John Muhemedi, Nellie Sew Kwan Kan, Jacqueline Nijimbere, Anna Rutagengwaova, and Marie Jeanne Uwanyarwaya. Nancy Levine edited the document, Kofi Edoh translated the Executive Summary into French, and Julia Anderson, Elsie Maka, and Stuart Tucker managed the processing of the report for publication. Mike McRory supplied the cover photo. The work on this report was financed by the World Bank and by the government of Norway through its Norwegian Education Trust Fund for the advancement of educational development in Africa. xvii ABBREVIATIONS, ACRONYMS, AND DATA NOTES AGSER average grade-specific enrollment rate ARR average repetition rate CA certificat d'aptitude CERAI centres de l'enseignement rural et artisanal intégré CERAR centres de l'enseignement rural et artisanal de Rwanda CESK Centre d'enseignement supérieur de Kigali CFATS Centre de formation des adjoints techniques de la statistique CPR Conseil protestant du Rwanda CSR cohort survival rate CWIQ Core Welfare Indicators Questionnaire (in French, QUID) DHS Demographic and Health Surveys EFA Education For All EICV Enquête intégrale sur les conditions de vie des ménages au Rwanda (Household Living Conditions Survey) EPLM Ecole pratique des langues modernes ES école secondaire ESGI Ecole supérieure de gestion et d'information ESM Ecole supérieure militaire ESTI Ecole des sciences et techniques de l'information FARG Fonds national pour l'assistance aux rescapés du génocide et des massacres au Rwanda FRw Rwandan franc FTB Faculté de théologie de Butare GDP gross domestic product GER gross enrollment ratio GNP gross national product GSK Grande séminaire de Kabgayi HIPC Heavily Indebted Poor Countries (Initiative) HIV/AIDS human immunodeficiency virus/acquired immune deficiency syndrome IAMSEA Institut Africain et Mauricien de statistiques et d'économie appliquée IDA International Development Association IFB Institute of Finance and Banking (formerly the Institut supérieur des finances publiques) IGER implied gross enrollment ratio IMF International Monetary Fund IPN Institut pédagogique national ISAE Institut supérieur d'agronomie et d'élevage ISCED International Standard Classification of Education ISCPA Institut supérieur catholique de pédagogie appliquée de Nkumba ISFP Institut supérieur des finances publiques (now the Institute of Finance and Banking) ISPG Institut supérieur de pédagogie de Gitwe KHI Kigali Health Institute KIE Kigali Institute of Education KIST Kigali Institute of Science, Technology and Management MDGs Millennium Development Goals MICS Multiple Indicator Cluster Survey MINALOC Ministère de l'administration local et des affaires sociales xix XX ABBREVIATIONS,ACRONYMS, AND DATA NOTES MINEDUC Ministry of Education NER net enrollment ratio NGO nongovernmental organization NPV net present value NUR National University of Rwanda (in French, Université nationale du Rwan- da, UNR) OECD Organisation for Economic Co-operation and Development PRS poverty reduction strategy PRSP poverty reduction strategy paper PSPP public spending per pupil PTA parent-teacher association QUID Questionnaire unifié sur les indicateurs de développement (in English, CWIQ) RPA Rwandan Patriotic Army RR repetition rate SF section familiale SIMA Statistical Information Management and Analysis database (World Bank) SNEC Sécretariat nationale de l'enseignement catholique UAAC Université adventiste d'Afrique centrale ULK Université libre de Kigali UNAIDS Joint United Nations Programme on HIV/AIDS UNDP United Nations Development Programme UNESCO United Nations Educational, Scientific, and Cultural Organization UNICEF United Nations Children's Fund UNILAK Université laïque de Kigali UNR Université nationale du Rwanda (in English, National University of Rwanda, NUR) DATA NOTES Unless otherwise indicated, all dollar figures are current U.S. dollars. Fiscal year: January 1­December 31 CURRENCY EQUIVALENTS (as of January 2002) Currency unit = Rwandan franc (FRw) FRw 1.00 = US$0.0022 US$1 = FRw 456.81 EXECUTIVE SUMMARY R wanda's recent history has been marred by the genocide of 1994 in which at least 800,000 people--about 10 percent of the population--lost their lives. Stability and secu- rity have been restored, and recovery has been under way for several years now. Rebuild- ing the stock of human capital is an important part of the recovery process. The Rwandan gov- ernment has worked hard to make up for lost time in broadening access to education and enhancing the quality of services. Indeed, progress in education is among the core objectives of the government's economic and social development strategy, as articulated in its recent poverty reduction strategy paper (PRSP). The importance accorded to education in Rwanda is consistent with recent developments on the international stage that have propelled the education sector into the limelight. Two of the eight Millennium Development Goals (MDGs) agreed on in the United Nations Millennium Declaration of 2000 have to do with education. Goal 2 is to ensure that by 2015 children everywhere--boys and girls alike--are able to complete a full course of primary schooling, and goal 3 calls for eliminating gender disparities in primary and secondary education, ideally by 2005 and at all levels by 2015. The Monterrey International Conference on Financing for Develop- ment, held in March 2002, arrived at a consensus that reaching the goals will require action by both rich and poor countries. Rich countries must, among other measures, boost foreign aid to poor countries, and poor countries must put in place policies and governance structures that will ensure effective use of resources to achieve the goals. On the country level as well, the principle of linking provision of resources with results on the ground is beginning to permeate discussion of development policies. Education is one among many sectors with a claim on scarce public resources, and the strength of its claim depends increasingly on its ability to deliver tangible results. This context presents clear challenges for the managers of Rwanda's education system. To attract increased resources, whether from interna- tional sources or from within the country, the sector must show evidence of good stewardship of the resources it already receives. 1 2 WORLD BANK COUNTRY STUDY This Country Study looks at the prospects for improvement in the education sector and the policies that are required to ensure that the education sector develops in an efficient, equitable, and fiscally sustainable direction. It examines the gaps in domestic financing that external resources might usefully fill and discusses the outcomes that might be agreed on as a focus for policy direction, as well as to create proper accountability structures in the system. Purpose and Scope The study is based on data up to 2001, the latest year for which it was possible to gather the necessary statistical information. The education system has not, of course, been standing still meanwhile, and so the snapshot picture captured by these data does not track recent develop- ments. Among these are the actions being taken by the Rwandan government to address con- straints on progress in the education sector. To give a few instances, the latest school census returns, for 2003, suggest that there has been significant progress in reducing grade repetition in primary school. In higher education, reforms in financial arrangements have been launched. One objective is to reduce the costs of government-sponsored study abroad by, for example, redirecting students to lower-cost host countries such as South Africa and India. Initiatives to manage the costs of in-country study at the tertiary level include treating student bursaries as loans rather than as outright grants and creating the Student Financing Agency to institutionalize collection of repayments. This report--a stock-taking exercise based on the situation at a given point in time--is best seen as a diagnostic document intended to contribute toward an understanding of the current performance of Rwanda's education system, the constraints on the system's progress, and the tradeoffs that may have to be made in the coming years. Building a shared understanding of the issues is an integral part of policy development. Many governments, including Rwanda's, are already engaged in consultations with partners and civil society as part of the process of defining their poverty reduction strategies. In education the importance of consultation can hardly be exaggerated. The success of policies hinges on how they are implemented, and this in turn depends greatly on how well the problems are understood and on the extent of agreement on the proposed solutions. The groups involved in any reform, and whose agreement needs to be sought, are many and diverse, including as they do policymakers, education planners, teachers, school managers, parents, and students. This report is, accordingly, aimed at a broad audience. In the first instance, it is directed toward Rwanda's policymakers in the education sector and toward education practitioners and researchers. It should also be of interest to policymakers and analysts in other parts of the gov- ernment, particularly those charged with managing the country's overall development strategy and with aligning public spending with that strategy. The study contains information and analyses that will serve as useful tools for the government's development partners--from donors to citizens at large--as they engage actively in discussing and articulating the country's vision for the education sector and in designing and implementing policies to put that vision into effect. The report does not pretend to address the full gamut of issues that Rwanda's policymakers may face. Its coverage is limited to key economic factors that are especially relevant in the PRSP context--cost, finance, service delivery, and educational outcomes--and it focuses in particular on those factors that lend themselves to quantification. Although admittedly incomplete, this treat- ment yields a picture of the broad structural characteristics of the system, the implicit patterns of resource allocation, and the effectiveness of service delivery. Progress to Date in the Education Sector Despite the untold havoc caused by the 1994 genocide, the Rwandan education system has recovered remarkably well, at least quantitatively. EDUCATON IN RWANDA 3 Increased Enrollments The most impressive aspect of the system's recovery is the rapid growth of enrollments. Only five years after the genocide, the number of children in primary school had already surpassed the number that would have been enrolled had the system expanded at historical rates of increase. The current gross enrollment ratio of 107 percent exceeds the corresponding ratio for the aver- age low-income country in Africa today. In secondary education the number of students has grown 20 percent a year since 1996, implying that the system is now nearly three times as large as it was then. Although the gross enrollment ratio at the secondary level remains below the aver- age for low-income Sub-Saharan Africa (13 percent compared with 20 percent), the gap would have been wider had the system stagnated after the genocide. In higher education, enrollments rose even more rapidly--from 3,400 students in 1990­91 to almost 17,000 by 2001­02, a nearly fourfold increase over a decade. The tertiary system's coverage is now comparable to the average of about 200 students per 100,000 population for low-income Sub-Saharan Africa. Diversified Public-Private School System As the system has expanded, it has done so in ways that have nudged it toward a good balance between the public and private sectors. At the base of the education pyramid, the government has made consistently strong efforts to extend the coverage of the public system. As a consequence, the share of enrollments in private primary schools has remained modest, at less than 1 percent. At the secondary level, enrollments grew as fast in the public as in the private sector in the postgenocide years, and the share of students attending private schools has remained steady at about 40 percent--down from 62 percent in the 1980s, but still much higher than the average of 20 percent for low-income Sub-Saharan Africa. In higher education the private sector grew in fits and starts in the two decades before 2000, but its share is unmistakably increasing over time, having reached about 38 percent in 2001­02, from about 8 percent at the start of the 1980s. The diversity of postsecondary institutions and the mix of public sector and private sector providers is a strength of the system, giving it the flexibility to meet the growing demand for places at that level. Good Basis for Further Development The expansion of the system has been taking place within a structure consisting of a six-year pri- mary cycle, a three-year tronc commun (lower secondary) cycle, a three-year upper secondary cycle, and, typically, a four-year cycle of higher education. A meritocratic examination system for selection of entrants to the various levels of schooling is in place. The 6-3-3-4 education struc- ture, along with the examination-based selection mechanism, provides a sensibly configured sys- tem for managing the sector's expansion. Studies show that Rwandan children who complete six years of primary schooling usually remain permanently literate and numerate as adults. Making completion of primary school uni- versal would help build the human capital base needed for broadly based economic and social development, and the system is already structured in line with this objective. In postprimary schooling, the selection arrangements for progression from one level to the next give managers of the system administrative levers for calibrating the pace of expansion of enrollments according to the availability of resources and the absorptive capacity of the market for highly educated labor. Relatively Broad Access Rwanda's education system compares favorably with that of other low-income countries in Africa with respect to access by different socioeconomic groups, especially at the primary level. (The gender, geographic, and income-group disparities that emerge mainly at higher levels are discussed in the next section.) School participation rates are relatively high, and noteworthy efforts have been made to ensure access by orphans, a particularly disadvantaged group. The 1994 genocide left Rwanda with one of the highest orphanhood rates in the world; nearly 40 percent of the 4 WORLD BANK COUNTRY STUDY children who were 7 to 14 years old in 2000 had lost at least one parent. Yet the gap in enroll- ments between orphans and other children is noticeable only among the most vulnerable children-- those who have lost both parents (double orphans) or who live apart from their parents. Appar- ently, there exist relatively well-developed safety nets that have managed to ensure high primary school participation rates even among orphans. In secondary education orphans are at least as well represented as nonorphans. This remarkable outcome owes much to the government's deci- sion in 1998 to establish the tax-funded Genocide Fund as a mechanism for assisting orphans in secondary school. The Challenges Ahead In the coming years, Rwanda will face second-generation problems that are beginning to emerge following the system's successful rebound from the devastation of the genocide. Priorities are accordingly shifting, from implementing emergency measures to reestablish the functioning of the system, to charting an appropriate and fiscally sustainable course for the sector's long-term devel- opment. Here, we summarize some of the main systemic issues that policymakers will likely have to grapple with in the foreseeable future. Managing Student Flow and Output of Graduates Primary education. To realize the goal of giving all children the opportunity to complete a full course of primary schooling, it is necessary to raise entry rates to grade 1 and to increase the likelihood that children will continue to the end of the cycle. In Rwanda, where entry rates have been historically high, at 90 percent, the challenge is to identify and help the last 10 percent of the population that is yet to enroll in primary school. With regard to survival to the end of the cycle, Rwanda's performance is respectable: its estimated 73 percent compares well with the rate in other low-income countries and with its own record of 44 percent in 1990­91. Yet this rosy situation is unlikely to persist, given the exceptionally high rate of grade repetition in the system-- about 34 percent in 2000­01, or more than three times that of a decade earlier. In line with what is widely believed (based on cross-country experience) to be good practice, it would make sense for the country to aim for a medium-term target of, say, a 10 percent improvement in the repetition rate; to implement measures designed to rationalize policies and practices regard- ing grade-to-grade promotion; and to take steps to improve learning outcomes and so mini- mize the need for grade repetition. As noted earlier, the government has begun implementing such policies, and the latest data show that repetition rates have fallen to about 17 percent in 2002­03. Secondary and higher education. The pressures to expand access beyond primary education are beginning to mount as larger cohorts of children complete the lower cycles of schooling. In designing policies governing the progress of students from level to level, it is useful to distinguish between the tronc commun and upper secondary cycles. The former can be thought of as a con- tinuation of the primary cycle. From this viewpoint, an appropriate medium-term objective would be to universalize access to the lower secondary cycle, to the extent that resources permit. The upper secondary cycle is more aptly viewed as a preparatory phase for higher education. Because graduates from upper secondary and higher education are being groomed at great expense for jobs in the modern economy, it is important to ensure that the number and skills mix of the graduates are compatible with the prospective demand for skilled labor. In the immediate postconflict period, pervasive shortages of highly educated workers were felt throughout the gov- ernment and in the private sector. These shortages stimulated a rapid growth in enrollments, particularly in higher education. Now that the stock of human capital is gradually being replen- ished, in part through the return of the Rwandan diaspora, and signs of potential graduate unem- ployment are beginning to emerge, it would be wise to review the situation and adapt admission policies in upper secondary and higher education accordingly. EDUCATON IN RWANDA 5 Mobilizing and Making Effective Use of Resources Support for educational development has been and continues to be strong in Rwanda. In the past few years the convergence of international and domestic commitment to education has helped boost public spending on education, which in 2001 reached an all-time high of 5.5 percent of gross domestic product (GDP). Yet closer examination shows that the increase has been dominated by spending on capital investments. Although such investments have obviously been needed to rehabilitate devastated facilities and expand capacity, at some point adequate resources will have to be directed toward sustaining the smooth day-to-day functioning of the system. In this regard, the record is not as satisfactory: current spending on education has stagnated at 1980 levels--just over 3 percent of GDP. Even at this level of expenditure, the education sector already claims more than a quarter of the government's total current budget (net of debt interest payments); the share is even higher if education-related spending channeled through other government organs is counted. Increased donor funding for the education sector, provided flexibly through budget support, could help rebalance spending in favor of current costs. But even in the best of circumstances, and looking at all sources of funding, the scope for rapid and large increases in the sector's share of public cur- rent spending is probably limited. Making better use of existing resources and using them to leverage private contributions must therefore be a centerpiece of efforts to advance the sector's goals. Making better use of public resources. There is room for improving allocation of public funds among levels of education and categories of expenditure. Primary education, which is so funda- mental to development, now receives only about 45 percent of public current spending on educa- tion, whereas higher education gets nearly 40 percent. The strong focus on higher education, which currently serves only 2 percent of the popula- tion in the relevant age group, has predictably inequitable results. First, Rwanda's unit costs in higher education are among the highest in the world today and are about 75 times the unit costs in primary education. Second, the best-educated 10 percent in a cohort claims more than 70 percent of the cumulative public spending on education received by that cohort. It is thus no surprise that Rwanda's system is one of the least structurally equitable in Sub-Saharan Africa. Redirecting spending toward primary education is the only way to rectify the situation. A second direction for improving the use of public funds is to restructure its functional distri- bution. To see the potential for improvement, it is helpful to break spending down conceptually into three categories: overhead, service delivery at the facility level, and student welfare. Overall, the system devotes barely two-thirds of spending to service delivery, and it is natural to wonder whether the objectives of the system are being achieved with this pattern of allocation. On closer examination, the main reason for the pattern in higher education is that spending at that level is heavily skewed toward student welfare services and bursaries for local and overseas study. In recent years these items have accounted for nearly half the government's current spend- ing on the subsector. The distribution is less skewed at the secondary level, but even there, provi- sion of food for students claims between 5 and 6 percent of current spending. At the primary level, nothing is currently spent on school feeding programs or other welfare services, but some improvements in allocation can still be made, notably by channeling resources away from man- agement overhead and toward better deployment of classroom teachers and increased provision of pedagogical materials. Mobilizing private resources. Households are the main source of nongovernment spending on education. Their overall contributions are significant, equivalent to roughly 40 percent of govern- ment expenditure on the sector. More can be done, however, to shift the weight of the house- hold contribution toward postprimary levels of education. As of this writing, the government has introduced new policies for student finance in higher education that will eventually require stu- dents to repay their government-financed bursaries. The government also continues to encourage 6 WORLD BANK COUNTRY STUDY privately financed institutions in secondary and higher education so as to ensure that the private sector remains a key source of funding for expansion of places at these levels. Using Explicit Criteria to Deploy Teachers across Schools Because the bulk of resources for primary and secondary education reaches the school or the classroom in the form of teachers, the deployment of teachers is an important aspect of how well public resources serve the education system's intended beneficiaries. In a well-managed system, the number of teachers that a school receives would be closely related to the size of enrollments. In Rwanda the relation is not as tight as might be expected on the basis of experience in other low-income African countries. A primary school serving 500 pupils, for example, may receive as few as 5 teachers or as many as 15. This diversity in allocation implies that resources are inequitably distributed across schools, giving rise to highly disparate learning environments. A few public primary schools are so well endowed that their teachers handle only one teaching shift instead of the two required of most teachers in the public system. Not surprisingly, the better- endowed schools are typically located in Kigali Ville, the capital, while the worst-endowed ones are often concentrated in certain provinces such as Byumba, Gisenyi, Kigali Rural, Kibungo, and Kibuye. Province-level differences, however, explain only a small part of the pattern; our analysis suggests that within-province differences are even more pronounced. The corollary is that teacher deployment within provinces can probably be better managed to produce greater consistency in the availability of teachers across schools. Balancing the Accessibility of Schools against Scale Economy Considerations The spatial distribution of schools is important because it determines the accessibility of the facili- ties to the intended beneficiaries. Yet the closer that schools are to students' homes, the more likely they are to serve small catchment areas, and the higher are their unit costs. The problem is to strike a good balance between accessibility and the cost of service delivery. As with other issues, the answer depends on the level of education. At the primary level, children are too young to travel far, so the issue of accessibility is highly pertinent. Yet the current structure of Rwanda's primary school network tends to emphasize size over proximity: children in half the households covered in a recent survey take more than 30 minutes to reach the nearest primary school, and the school they attend is likely to be among the three-quarters of the country's public schools that enroll more than 450 pupils. Large schools, however, confer few economies of scale; unit costs are relatively flat beyond enrollments of about 450. Consideration could therefore be given to extending the primary school network so as to situate schools closer to pupils' homes. This would reduce both the physical barriers to school participation and the opportunity cost of attending. At postprimary levels, students are able to travel farther, making physical distance less of a constraint on enrollment, and economies of scale begin to set in because specialized teaching arrangements become increasingly common. In Rwanda schools providing instruction at the tronc commun level tend to be small and to have high unit costs. Per student costs in schools with 400 or more students are less than two-thirds those in schools enrolling between 100 and 200 students and four-fifths those in schools enrolling between 200 and 300 students. Yet schools that enroll 300 or fewer students account for more than 55 percent of enrollments. Our calcula- tions show that moving to a situation in which no school enrolls fewer than 400 students would reduce the unit cost of service delivery by 20 percent systemwide--arguably, not a trivial saving in a context of scarce resources. As enrollments expand in the coming years, it is important to accommodate the increase by raising enrollments in existing lower secondary schools to an eco- nomic size rather than by building more new schools for small catchment areas. In upper secondary education, as in higher education, enrollments are relatively small and are likely to grow slowly as long as the absorptive capacity of the modern sector labor market remains limited. It is therefore important for schools and institutions to take advantage of scale economies EDUCATON IN RWANDA 7 in service delivery, which may require some consolidation of course offerings. Only 14 percent of students at the upper secondary level attend a school serving 400 or more students, and students at each school tend to be spread thinly across fields of specialization. There is also evidence of duplication among public institutions and between the public and private sectors in some fields of study. These results suggest that in managing the expansion of these levels, policymakers might seek to minimize proliferation of course offerings, particularly where demand is weak and the potential for scale economies in service delivery is limited. Managing Classroom Conditions and Processes to Enhance Student Learning A crucial test of the performance of an education system is its effectiveness in transforming the resources at its disposal into learning outcomes. In both primary and secondary education there seems to be room for improvement: the performance of schools, as measured (however imper- fectly) by examination scores, relates only weakly to the amount of resources they receive. This evidence raises questions about the effectiveness of the mix of school inputs that support service delivery and, more important, the incentives for performance. Inputs in primary education. In an ideal, well-funded world, pupils would receive sufficient instructional time in small classes taught by well-qualified teachers and with ample supplies of books and other learning materials. In resource-constrained environments, tough tradeoffs have to be made, and policymakers must find the mix of inputs that works best under the constraints. In Rwanda the input mix in public primary schools favors teacher qualifications at the expense of class size and instructional time. The country's pupil-teacher ratio of 57 to 1 is currently among the highest in the world, and it translates into a situation in which most teachers in the first three grades of the cycle have to teach two shifts of pupils. As a result, instructional time for pupils in these grades averages only about 500 hours a year. (The average is 1,000 hours in the upper grades in Rwanda, and other low-income countries have set a target of 850 to 1,000 hours for primary instruction on the basis of international best practice.) Reducing the pupil-teacher ratio by hiring more teachers would help address the problem, but in a tight budgetary situation this will require changes in the input mix. Adjusting the educational qualifications for recruitment of teachers is one option for manag- ing the tradeoff. All new primary school teachers are currently hired from among those who have completed upper secondary education. Although recruiting at this level has advantages in theory and may be appropriate as a long-run objective, in the current context of Rwandan schools, such teachers are no more effective than those with only lower secondary education and are more costly. Given the differences in wages between the two levels of teacher qualification, it may be worth relaxing the recruitment criteria so as to make it financially feasible and sustainable to reduce the pupil-teacher ratio. Inputs in secondary education. At the secondary level, the issue of teacher qualifications is also highly relevant. Most teachers in the tronc commun cycle satisfy the minimum qualification requirement (that is, they have at least an upper secondary school diploma), and some even have a university degree. By contrast, half of all teachers in upper secondary schools are probably underqualified, having only an upper secondary school diploma themselves. Any strategy to improve learning outcomes must seek to raise the educational profile of teachers in that cycle. The obvious solution is to set and apply higher standards for recruitment, but as a stopgap reme- dy, policymakers might consider redeploying the existing cadre of secondary school teachers between the tronc commun and upper secondary levels according to their qualifications. Incentives. Although choosing the right input mix is important, it is by no means the entire answer to poor performance. Because schooling is a social process involving multiple actors, out- comes ultimately depend on how people behave. Putting in place the right incentives to align behavior with the goal of enhancing student learning is a key challenge in managing for results. The potential interventions in this regard are as diverse as the contexts in which people live and work, but the principle of establishing clear accountabilities and matching them with spending 8 WORLD BANK COUNTRY STUDY and management authority at all levels is probably relevant everywhere. Cross-country experience also highlights the importance of defining and monitoring tangible indicators of progress as a way of increasing the incentives for better performance. Minimizing Barriers to Access As noted above, socioeconomic gaps in enrollments are narrower in Rwanda than in other low- income countries. But this should not be a reason for complacency. More can be done to reach the most vulnerable children, including double orphans, street children who do not benefit from systematic adult supervision, children living in rural areas, and children from the poorest 40 per- cent of households. Double orphans are easily identified, and systematic efforts can and must be made to improve their prospects in life. For the other groups, diverse types of assistance are likely to be needed. Exploratory and pilot interventions are probably appropriate in seeking the best way to serve them. In secondary and higher education, participation rates among children from the top income quintile far outstrip those in the rest of the population. Girls are as likely as boys to enroll in sec- ondary school, but they lag significantly behind in higher education. The study's findings suggest that although financial assistance might help the lagging groups in the same way as it is helping orphans in secondary education under the Genocide Fund, this kind of intervention probably needs to be combined with efforts to improve learning outcomes. Such measures are important because in a meritocratic selection system based on examination results, enlarging the representa- tion of lagging groups is possible only to the extent that these groups are able to compete for the coveted places in postprimary education. Conclusion Nearly a decade after the 1994 genocide, Rwanda's leaders can look back with pride and satisfac- tion at the country's record of achievements in education. A devastated system has been put back on its feet. Classrooms have been repaired and new ones built to accommodate the growing numbers of students; teachers who fled the mayhem and have returned have been reintegrated into the teaching force; arrears in teacher pay have been cleared up; the Genocide Fund has been established specifically to assist orphans; and a diversified system has been created in higher edu- cation. Yet the task ahead remains daunting, as the recovery phase gives way to the work of nurtur- ing the long-term development of the education sector. Concerns about efficiency, equity, and fiscal sustainability will inevitably become increasingly relevant as the country seeks to advance educational progress in a resource-constrained environment. It is hoped that the findings of this study will contribute to the discussion by forging a common understanding of the issues and by drawing attention to the emerging challenges. RÉSUME ANALYTIQUE L 'histoire récente du Rwanda a été perturbée par un génocide en 1994. Au cours de ce génocide, au moins 800.000 personnes, soit environ 10 pour cent de la population, ont perdu la vie. La stabilité et la sécurité ont été restaurées et le processus de redressement est en cours depuis plusieurs années. La reconstruction du potentiel de capital humain est une partie importante de ce processus. Par ailleurs, le gouvernement s'est également efforcé de compenser le temps perdu dans l'élargissement de l'accès à l'éducation et le renforcement de la qualité des services. En effet, le progrès en éducation figure parmi les objectifs essentiels de la stratégie de développement économique et social du gouvernement comme articulé dans son récent docu- ment de stratégie de réduction de la pauvreté (DSRP). L'importance accordée à l'Education au Rwanda est consistante avec les récents développe- ments qui, sur le plan international, a placé le secteur de l'Education sous les projecteurs de l'ac- tualité. Au titre de la déclaration des Nations Unies pour le Millénium 2000, deux des huit Objectifs de Développement (ODM) retenus traitent de l'Education. L'objectif 2 est de s'assurer que d'ici 2015, partout, les garçons comme les filles pourront achever une scolarité primaire com- plète, et l'objectif 3 réclame l'élimination des disparités entre les genres dans l'enseignement pri- maire et secondaire, idéalement en 2005 au plus tard. Selon la conférence internationale de Mon- terrey de 2002 sur le financement du développement, il se dégage aussi un consensus que la réalisation des objectifs nécessitera l'action conjointe des pays riches et pauvres. Les pays riches doivent accroître l'aide aux pays pauvres (entre autres mesures) qui, à leur tour, doivent mettre en place des politiques appropriées et des structures de bonne gouvernance pour assurer l'utilisation efficace des ressources afin d'atteindre les objectifs visés. Au niveau du pays, le principe de lier les ressources aux résultats sur le terrain commence à imprégner la discussion sur les politiques de développement. Le secteur éducatif est l'un des nom- breux secteurs qui réclament des ressources publiques rares et la force de sa revendication dépend de plus en plus de sa capacité de fournir des résultats tangibles. Ce contexte présente des défis 9 10 WORLD BANK COUNTRY STUDY clairs aux responsables du système éducatif au Rwanda. Pour attirer des ressources accrues aux niveaux national et international, le secteur doit mettre en évidence la bonne gestion des ressources déjà disponibles. Cette étude examine les perspectives d'amélioration du secteur éducatif et des politiques appropriées en vue de s'assurer que le secteur se développe de façon efficace, équitable et fiscale- ment durable. Elle examine les lacunes de financement au plan national que les ressources externes pourraient utilement combler, et discute des résultats qu'on pourrait accepter comme base d'orientation des politiques, ainsi que de la création de structures de responsabilité propre dans le système. But et portée Cette étude est basée sur des données collectées jusqu'en 2001, dernière année pour laquelle il a été possible d'obtenir des informations statistiques appropriées. Le système éducatif n'est pas resté, bien sûr, statique; le tableau que ces données présentent ne couvre donc pas les récents développements intervenus dans le secteur. Parmi ces derniers, les actions prises par le Gouvernement rwandais pour s'attaquer à certaines des contraintes dans le développement du secteur éducatif. A titre d'exemples, les résultats du dernier recensement scolaire de 2003 ont démontré que des progrès remarquables ont été réalisés dans la réduction du redoublement dans l'enseignement primaire. Des réformes ont été introduites dans les finances de l'Enseignement Supérieur. L'un des objectifs vise à réduire les coûts des études à l'étranger financés par le Gouvernement: par exemple, des étudiants ont été redirigés vers des pays moins chers comme l'Afrique du Sud et l'Inde. Des initiatives de gestion des coûts des études faites sur place ont été également prises: des bourses ont été transformées en prêts, contrairement aux subventions d'antan; une Agence de Financement des Etudiants, visant à institutionnaliser le remboursement des prêts, a été mise en place. Ce rapport, basé sur la situation en un moment donné, est mieux perçu comme un document diagnostic destiné à contribuer à une compréhension de la performance actuelle du système édu- catif au Rwanda, les contraintes qui freinent son progrès, et les choix à faire dans la planification du développement du système au cours des années à venir. L'instauration d'un consensus sur les questions constitue une part intégrale d'une politique de développement. Bon nombre de Gou- vernements, y compris celui du Rwanda, se sont déjà engagés dans des consultations avec les partenaires et la société civile comme partie du processus de définition de leurs stratégies de réduction de la pauvreté. Dans le secteur de l'éducation, l'importance de la concertation peut à peine être exagérée. Le succès des politiques dépend de la façon dont elles sont mises en oeuvre et ceci dépend à son tour, à un haut degré, de la bonne compréhension des problèmes et du degré d'entente sur les solutions proposées. Les groupes impliqués dans toute réforme et ceux dont l'avis est nécessaire sont nombreux et variés; on compte parmi eux les décideurs, les planificateurs de l'éducation, les enseignants, les directeurs d'école, les parents et les étudiants. Ce rapport vise en conséquence un grand public. En premier lieu, il s'adresse aux décideurs, aux praticiens et chercheurs du secteur de l'éducation au Rwanda. Il devrait également être d'in- térêt aux décideurs et analystes des autres secteurs du Gouvernement, en particulier, ceux chargés de contrôler la stratégie globale de développement du pays et d'aligner des dépenses publiques sur cette stratégie. Cette étude comporte l'information et les analyses qui serviront d'outils utiles aux partenaires au développement du Gouvernement--des donateurs aux citoyens en général--au moment où ils s'engagent activement dans les discussions et articulations de la vision du pays pour le secteur de l'éducation et dans la conception et la mise en oeuvre des politiques en vue de réaliser cette vision. Ce rapport ne prétend pas aborder toute la gamme de questions auxquelles les décideurs du Rwanda pourraient faire face. Il est limité aux aspects économiques clés qui sont particulièrement appropriés dans le contexte de DSRP. Ceux-ci concernent le coût, les finances, la prestation de service et les résultats de l'éducation, en particulier les aspects qui se prêtent à la quantification. EDUCATON IN RWANDA 11 Tout en admettant son caractère incomplet, le document fournit une image des grandes carac- téristiques structurelles du système, du modèle implicite d'attribution des ressources et de l'effi- cacité de la prestation de service. Progrès de l'enseignement jusqu'à ce jour En dépit des ravages indescriptibles provoqués par le génocide de 1994, le système éducatif rwandais a récupéré remarquablement bien, au moins en termes quantitatifs. Augmentation des inscriptions L'aspect le plus impressionnant du redressement du système est le rythme croissant des inscriptions après le génocide. Cinq (5) ans seulement après l'événement, le nombre d'enfants inscrits à l'é- cole primaire avait déjà dépassé le nombre qui aurait été inscrit si le système avait connu une croissance aux taux de progression historiques. Actuellement à 107 pour cent, le taux brut d'in- scription dépasse le taux correspondant pour la moyenne des pays à faible revenu en Afrique à ce jour. Dans l'enseignement secondaire, le nombre d'élèves a augmenté de 20 pour cent par an depuis 1996, impliquant que le système est maintenant presque trois fois plus grand que ce qu'il était précédemment. Tandis que le taux brut d'inscription au niveau secondaire demeure en- dessous de la moyenne pour l'Afrique subsaharienne des pays à faible revenu (13 pour cent au lieu de 20 pour cent), l'écart aurait été encore plus grand si le système avait stagné après le géno- cide. Dans l'enseignement supérieur, les inscriptions ont grimpé plus rapidement encore, allant de 3.400 étudiants en 1990­1991 à presque 17.000 en 2001­2002, une augmentation qui a presque quadruplé en l'espace d'une décennie. La couverture du système de l'Enseignement Supérieur est maintenant comparable à la moyenne d'environ 200 étudiants pour une population de 100.000 en Afrique subsaharienne à faible revenu. Diversification du système éducatif Le système s'est développé tout en maintenant un bon équilibre entre les secteurs public et privé. A la base de la pyramide de l'enseignement, le Gouvernement a fait un effort considérable en vue d'étendre la couverture du secteur public. Comme conséquence, la plupart des inscriptions dans les écoles privées sont demeurées modestes à moins de 1 pour cent. Au niveau secondaire, les inscriptions ont augmenté aussi rapidement dans le secteur public que privé dans les années d'après génocide, et la proportion des étudiants inscrits dans les écoles privées est demeurée stable à environ 40 pour cent, ce qui est plus bas que les 62 pour cent des années 80, mais toujours beaucoup plus élevé que les 20 pour cent en moyenne dans les pays à faible revenu en Afrique subsaharienne. Dans l'enseignement supérieur, le secteur privé s'est développé tant bien que mal pendant les deux décennies jusqu'en 2000, mais la proportion d'étudiants inscrits a augmenté indubitablement avec le temps, soit environ 38 pour cent en 2001­2002 au lieu d'environ 8 pour cent au début des années 80. La diversité des établissements post-secondaires et le mélange de prestataires publics et privés du secteur constituent une force du système, le dotant d'une flexibil- ité pour satisfaire la demande croissante de places à ce niveau d'études. Une base solide pour un développement ultérieur L'expansion du système s'est effectuée dans une structure qui prévoit un cycle primaire de six ans, un cycle de tronc commun de trois ans (cycle secondaire inférieur), un cycle secondaire supérieur de trois ans et un cycle de quatre ans d'Enseignement Supérieur. Un système d'évaluation basé sur le mérite a été mis en place pour contrôler la sélection des étudiants des divers cycles d'études. La structure éducative 6-3-3-4, avec le mécanisme de sélection basé sur les examens constitue un système raisonnablement structuré pour contrôler l'expansion du secteur. Des études indiquent que les enfants rwandais qui ont achevé les six ans d'école primaire savent, presque toujours de façon permanente, lire et compter même à l'âge adulte. Réaliser l'éducation primaire universelle contribuerait à la reconstruction de la base du capital humain 12 WORLD BANK COUNTRY STUDY nécessaire pour un développement économique et social élargi, et le système est déjà structuré en conformité avec cet objectif. Dans l'enseignement post-primaire, les arrangements de la sélection pour la progression, d'un niveau à l'autre donne aux responsables du système les leviers adminis- tratifs pour adapter le rythme de l'expansion des inscriptions à la disponibilité des ressources et à la capacité d'absorption par le marché d'une main d'oeuvre hautement qualifiée. Accès relativement élargi Le système rwandais se compare favorablement à celui des autres pays à faible revenu en Afrique, en termes d'accès des différents groupes socio-économiques, particulièrement au niveau de l'en- seignement primaire. Les disparités de genre, de géographie, de revenu qui émergent principale- ment dans l'Enseignement Supérieur sont discutées dans la section suivante. Les taux de fréquen- tation de l'école sont relativement élevés, et des efforts remarquables ont été faits en vue de garantir l'accès aux orphelins, un groupe particulièrement défavorisé. Le génocide de 1994 a laissé le Rwanda avec un des taux d'orphelins les plus élevés au monde, avec presque 40 pour cent des enfants âgés de 7 à 14 ans en 2000 ayant perdu au moins un parent. L'écart des inscriptions entre les orphelins et les autres enfants n'est remarquable que parmi les enfants les plus vulnérables (ceux qui ont perdu les deux parents ou ceux qui vivent séparés de leurs parents par exemple). Apparemment cette situation est due à l'existence de filets de sécurité relativement bien développés qui ont permis d'assurer des taux de fréquentation élevés à l'école primaire, même parmi les orphelins. Dans l'enseignement secondaire, les orphelins sont au moins aussi bien représentés que les non-orphelins. Ce résultat remarquable doit beaucoup à la décision du Gou- vernement en 1998 de créer le Fonds de Génocide, alimenté par une taxe, comme mécanisme d'aide aux orphelins dans l'enseignement secondaire. Les défis de l'avenir Dans les années à venir, le Rwanda devra faire face aux défis des problèmes de deuxième généra- tion, qui commencent déjà à se poser, suite à l'heureux rebondissement du système après la dévastation du génocide. L'accent se déplace ainsi d'une situation où les mesures urgentes pour rétablir le fonctionnement du système étaient prioritaires, à une situation où la préoccupation consiste à donner une direction appropriée et fiscalement soutenable au développement à long terme du secteur. Un sommaire de certaines des questions principales au niveau systémique aux- quelles les décideurs devront probablement s'attaquer à dans un proche avenir est présenté ci- dessous. Gestion du flux d'étudiants et production de diplômés du système Enseignement primaire. Pour atteindre l'objectif qui consiste à donner à tous les enfants la chance de terminer leur cycle primaire, il est nécessaire d'accroître les taux d'accès en première année et de renforcer la probabilité que les enfants continuent jusqu'à la fin du cycle. Au Rwanda, les taux d'entrée ont été historiquement élevés et atteignent 90 pour cent. Ainsi le défi ici consiste à identifier et à aider les derniers 10 pour cent de la population qui refusent d'aller à l'école. En ce qui concerne la survie jusqu'à la fin du cycle, la performance du Rwanda est respectable: le taux actuel, estimé à 73 pour cent, se compare favorablement au taux d'autres pays à faible revenu et aux 44 pour cent du Rwanda dans les années 1990­1991. Cependant, il est peu probable que ce bon résultat persiste; en effet, le taux de redoublement de classe dans le système est particulièrement élevé--environ 34 pour cent en 2000­2001, soit plus de 3 fois le taux de la décennie précédente. Sur la base de ce qui est généralement admis comme une bonne pratique selon les expériences dans d'autres pays, il serait raisonnable que le Rwanda vise une moyenne d'environ 10 pour cent et mette en place des mesures en vue de rationaliser les politiques et les pratiques relatives au passage d'un niveau à un autre, et d'améliorer les résultats des élèves afin de réduire au minimum le besoin de faire doubler les classes. Comme indiqué plus haut, le Gouvernement a commencé à réaliser de telles EDUCATON IN RWANDA 13 politiques et les dernières statistiques indiquent que les taux de redoublement ont diminué à environ 17 pour cent en 2002­2003. Enseignement secondaire et supérieur. Les pressions en vue de l'accès à l'Enseignement Sec- ondaire et Supérieur commencent déjà à se faire sentir; en effet, des cohortes plus importantes d'enfants achèvent maintenant les cycles inférieurs d'instruction. Pour développer des politiques appropriées régissant la progression des étudiants d'un niveau à l'autre, il est utile de distinguer le cycle de tronc commun du cycle secondaire supérieur. Le premier peut être considéré comme une continuation du cycle primaire; ainsi, dans la mesure où les ressources le permettent, il serait approprié de viser l'accès universel à ce cycle comme objectif à moyen terme. Le cycle secondaire supérieur quant à lui est considéré à juste titre comme une phase prépara- toire à l'enseignement supérieur. Les diplômés de ces niveaux sont formés à grands frais pour des emplois dans l'économie moderne; il est donc important de s'assurer que le nombre de diplômés formés et la diversité de leurs qualifications sont compatibles avec la demande prospective de main d'oeuvre qualifiée. Dans le contexte post-conflit immédiat, de grandes pénuries de travailleurs hautement qualifiés étaient senties dans l'administration et dans le secteur privé, et ces pénuries ont stimulé une croissance rapide des inscriptions, en particulier dans l'Enseignement Supérieur. Maintenant que le potentiel de capital humain s'est graduellement reconstitué en par- tie par l'immigration de la diaspora rwandaise et que des signes de chômage potentiel des diplômés commencent à apparaître, il serait sage de réexaminer la situation et d'adapter en con- séquence les politiques d'admission dans l'enseignement secondaire supérieur et l'enseignement supérieur. Mobilisation et utilisation efficiente des ressources L'appui au développement de l'enseignement a été et continue d'être solide. Au cours des dernières années, la convergence de l'engagement national et des engagements internationaux ont contribué à faire monter les dépenses publiques pour l'enseignement au Rwanda au taux historique jamais atteint de 5,5 pour cent du PIB en 2001. Cependant, un examen attentif révèle que l'augmentation a été dominée par des dépenses d'investissement en infrastructure. Bien que de tels investissements aient été nécessaires pour remettre en état les installations détruites et en augmenter la capacité, tout le système à un certain moment aurait dû requérir des ressources adéquates pour soutenir son fonctionnement régulier. A cet égard, le système a moins réussi; en effet, les dépenses actuelles pour l'enseignement sont demeurées au niveau de celles des années 1980, juste au-dessus de 3 pour cent du PIB. Il est néanmoins important de noter que même à ce niveau des dépenses, le secteur bénéficie déjà de plus du quart du budget actuel du Gouvernement (moins le paiement des intérêts de la dette); la contribution du Gouvernement serait encore plus élevée si on incluait les dépenses liées à l'en- seignement drainées par d'autres organes du Gouvernement. L'augmentation du financement des donateurs du secteur, pour une utilisation flexible à travers l'appui au budget, pourrait aider à rééquilibrer les dépense en faveur des coûts récurrents. Cependant, même dans le meilleur des cas, les chances pour une augmentation importante et rapide des ressources du secteur dans les dépenses publiques récurrentes sont limitées. Faire un meilleur usage des ressources existantes et s'en servir pour susciter des contributions privées doit donc être une pièce maîtresse des efforts pour faire aboutir les objectifs du secteur. Une meilleure utilisation des ressources publiques. Il serait nécessaire d'améliorer les allocations aux divers niveaux d'enseignement et aux catégories de dépenses. L'enseignement primaire, qui si fondamental pour le développement, ne reçoit que 45 pour cent des dépenses récurrentes publiques consacrées à l'éducation, alors que l'enseignement supérieur en obtient presque 40 pour cent. La très forte concentration sur l'enseignement supérieur combinée avec le fait que l'enseigne- ment supérieur ne représente présentement que 2 pour cent de la population concernée produit un résultat inéquitable prévisible: les coûts unitaires de l'enseignement supérieur au Rwanda fig- 14 WORLD BANK COUNTRY STUDY urent parmi les plus élevés au monde (environ 75 fois ceux de l'enseignement primaire) et les 10 pour cent les mieux formés dans une promotion absorbent plus de 70 pour cent des dépenses publiques cumulatives à l'éducation reçues par la cohorte. Il n'est donc pas surprenant que le système rwandais soit l'un des moins structurellement équitables en Afrique subsaharienne. La réorientation des dépenses vers l'enseignement primaire est le seul moyen de rétablir la situation. La restructuration de la distribution fonctionnelle constitue une deuxième direction en vue d'améliorer l'utilisation des fonds publics. Il est utile de conceptualiser les dépenses sous trois rubriques: frais généraux, prestation de service au niveau des écoles et bien-être de l'étudiant. De façon générale, le système consacre à peine les deux tiers des dépenses à la prestation de service; il est donc normal de se demander si les objectifs du système sont atteints avec ce modèle d'allocation. A l'examen approfondi, la raison principale de ce modèle s'explique par le fait que les dépens- es pour l'Enseignement Supérieur sont fortement orientées vers les services d'assistance sociale de l'étudiant et les bourses d'étude (pour des études dans le pays et à l'étranger) qui, ces dernières années, ont représenté presque la moitié des dépenses courantes du Gouvernement pour le sous secteur. La distribution est moins déviée au niveau secondaire; même ici les dépenses pour l'ali- mentation des élèves représentent 5 à 6 pour cent des dépenses récurrentes. Au niveau primaire aucune dépense n'est actuellement faite pour la restauration à l'école ou pour d'autres services d'assistance sociale, mais on peut également améliorer les allocations, notamment en diminuant les frais généraux de gestion en faveur d'un meilleur déploiement des enseignants et de l'acquisi- tion de plus importantes quantité de matériels didactiques. Mobilisation des ressources privées. Les ménages constituent la source la plus importantes des dépenses non gouvernementales sur l'éducation. Leurs contributions globales sont considérables, équivalentes approximativement à 40 pour cent des dépenses publiques du financement de l'édu- cation. On pourrait mieux faire sur le plan de l'éducation, cependant, pour re-orienter les dépens- es des ménages vers les niveaux de l'éducation post-primaire. A la rédaction de ce rapport, le Gouvernement a présenté de nouvelles politiques pour le financement de l'Enseignement Supérieur qui feront rembourser les bourses d'études octroyées aux étudiants sous forme de prêt financé par le Gouvernement. Le Gouvernement continue également d'encourager les établisse- ments du secondaire et du supérieur privés afin de s'assurer que le secteur privé demeure une source principale pour l'expansion de l'offre de places à ses niveaux. Des critères explicites pour le déploiement des enseignants dans les écoles Puisque la majorité des ressources destinées au financement de l'éducation primaire et secondaire parvient à l'école ou à la salle de classe sous forme d'enseignants, le déploiement des enseignants constitue un aspect important de la gestion des ressources publiques pour servir le système édu- catif des bénéficiaires ciblés. Dans un système bien géré, le nombre d'enseignants d'une école aurait une relation étroite avec l'importance des inscriptions. Au Rwanda, cette relation n'est pas aussi étroite que l'on pourrait s'y attendre en se basant sur l'expérience d'autres pays africains à faible revenu. Une école primaire accueillant 500 élèves, par exemple, pourrait recevoir aussi peu que 5 enseignants ou un effectif aussi important que 15. Cette diversité dans l'attribution implique que des ressources sont inéquitablement distribuées aux écoles et que par conséquent les conditions d'apprentissage sont fortement disparates. Quelques écoles publiques sont si bien dotées qu'elles peuvent permettre à leurs enseignants d'assurer seulement tour d'enseignement au lieu des deux habituellement exigés de la plupart des enseignants dans le système public. Comme on pourrait s'y attendre, les écoles les mieux dotées sont typiquement situées dans Kigali Ville; les mal loties sont souvent concentrées dans des provinces telles que Byumba, Gisenyi, Kigali-Rural, Kibungo et Kibuye. Les différences de niveau au sein des provinces n'expliquent cependant qu'une petite partie de la situation. Notre analyse suggère que les différences à l'intérieur d'une province sont même plus prononcées. Le corollaire est que le processus du déploiement des enseignants dans chaque province peut probablement être mieux géré pour produire une plus grande uniformité dans la disponibilité des enseignants au sein des écoles. EDUCATON IN RWANDA 15 Equilibrer l'accessibilité des écoles avec les considérations d'économies d'échelle L'implantation spatiale des écoles est importante parce qu'elle détermine leur accessibilité aux bénéficiaires cibles. Cependant, plus les écoles sont situées à proximité des habitations des élèves, plus il est probable qu'elles servent de petits secteurs de recrutement scolaire, et donc plus élevés sont les coûts unitaires. Comment trouver alors un bon équilibre entre l'accessibilité et le coût de la prestation de service? Comme ailleurs, la réponse dépend du niveau de l'enseignement. Au niveau primaire, les enfants sont trop jeunes pour faire de grands déplacements; la ques- tion de l'accessibilité devient alors fortement pertinente. Pourtant le réseau actuel des écoles pri- maires au Rwanda tend à privilégier la taille par rapport à la proximité: les enfants dans la moitié des ménages, selon une enquête récente, mettent plus de 30 minutes pour atteindre les écoles les plus proches; et ces écoles sont susceptibles d'être parmi les trois quarts des écoles publiques du pays qui accueillent plus de 450 élèves. L'accent mis sur les écoles de grande taille permet peu d'économies d'échelle cependant, puisque les coûts unitaires sont relativement uniformes au-delà de l'inscription d'environ 450 élèves. Dans l'enseignement primaire, on pourrait donc accorder une attention particulière à l'extension du réseau des écoles afin de les situer plus près des habita- tions des élèves. Le résultat devrait être la réduction de la barrière physique à la fréquentation scolaire ainsi que la réduction du coût occasionné par cette fréquentation. Aux niveaux post-primaires, les élèves peuvent voyager plus loin, et la distance physique devient moins une contrainte qui limite les inscriptions; par la même occasion, les économies d'échelle com- mencent à apparaître parce que les installations pour les enseignements spécialisés deviennent de plus en plus communs. L'équilibre entre l'accessibilité et les économies d'échelle pourrait ainsi être dif- féremment réalisé. Au Rwanda, une part considérable des écoles du cycle de tronc commun tend à être des écoles de petite taille qui fonctionnent à des coûts unitaires élevés. Les écoles de 400 élèves coûtent deux tiers moins cher par élève que celles qui accueillent 100 à 200 élèves, et elles coûtent les quatre cinquièmes de celles qui accueillent 200 à 300 élèves. Pourtant les écoles de ces deux dernières catégories accueillent plus de 55 pour cent des inscrits. Nos calculs montrent qu'une situa- tion où aucune école n'inscrirait pas moins de 400 élèves réduirait le coût unitaire de la prestation de service de 20 pour cent au niveau de l'ensemble du système, probablement une économie non négligeable dans un contexte de ressources rares. Ainsi, comme les inscriptions augmenteront dans les années à venir, il est important d'harmoniser la croissance en augmentant les inscriptions des écoles existantes jusqu'à ce qu'elles atteignent une taille économiquement acceptable, plutôt qu'en construisant plus de nouvelles écoles pour de petits secteurs de recrutement. Dans l'enseignement secondaire supérieur comme dans l'enseignement supérieur, les inscrip- tions sont relativement faibles et sont susceptibles de se développer lentement aussi longtemps que la capacité d'absorption du marché du travail du secteur moderne demeure limitée. Cette tendance implique qu'il est même plus important pour les écoles et institutions de tirer profit des économies d'échelle dans la prestation des services et ceci peut exiger une certaine consolidation de leurs offres de cours. Au Rwanda, seuls 14 pour cent des élèves du secondaire supérieur vont à une école de 400 étudiants ou plus, et les étudiants de chaque école ont tendance à se disperser dans plusieurs domaines de spécialisation. Dans l'enseignement supérieur, il y a également une évidence de duplication entre les institutions publiques d'une part, et entre les secteurs public et privé d'autre part, dans certains domaines d'étude. Ces résultats suggèrent qu'en contrôlant l'ex- pansion de ces niveaux, les décideurs pourraient chercher à réduire au minimum la prolifération des offres de cours, particulièrement là où la demande est faible et le potentiel pour des économies d'échelle dans la prestation de service limité. Gérer les conditions et les processus de classe en vue d'améliorer l'apprentissage de l'étudiant. Un test crucial de performance du système éducatif est son efficacité dans la transformation des ressources mises à sa disposition en résultats scolaires. Dans l'enseignement primaire et secondaire, il apparaît évident que des possibilités d'améliorer la performance des écoles existent; en effet, la performance des écoles, mesurée (même de façon imparfaite) par les résultats aux examens, n'a 16 WORLD BANK COUNTRY STUDY qu'un rapport très faible avec le niveau des ressources qu'elles reçoivent. Ceci soulève deux types de question: l'efficacité de la combinaison des ressources qui soutiennent la prestation des servic- es; et de façon très importante, les incitations à la performance. Intrants de l'enseignement primaire. La situation idéale, sans contrainte financière, serait que les élèves disposent d'un temps d'instruction suffisant, soient enseignés dans de petites classes par des enseignants compétents et aient accès à un approvisionnement suffisant en livres et autres matériels didactiques. Mais dans un environnement de ressources limitées, des choix difficiles doivent se faire et les décideurs doivent trouver les combinaisons de ressources qui fonctionnent le mieux sous ces contraintes. Au Rwanda, la combinaison des ressources dans les écoles primaires publiques favorise la qualification des enseignants aux dépens de la taille de la classe et du temps d'instruction. Le ratio élève-enseignant de 57 à 1 du pays se situe actuellement parmi les plus élevés au monde et se traduit en une situation dans laquelle la plupart des enseignants des trois premiers niveaux du cycle d'enseignement sont obligés d'enseigner deux vagues d'élèves. En con- séquence, le temps d'instruction des élèves dans ces niveaux s'élève en moyenne à environ cinq cents (500) heures par an, au lieu de mille (1000) heures dans les niveaux supérieurs et de la fourchette de huit cent cinquante (850) à mille (1000) heures que d'autres pays à faible revenu visent présentement en se basant sur les meilleures pratiques internationales. Réduire le ratio élève-enseignant (ce qui nécessiterait le recrutement d'un plus grand nombre d'enseignants) aiderait à résoudre le problème, mais dans un contexte de budgets serrés, ceci exigera un ajuste- ment de la combinaison des ressources. L'ajustement du niveau de recrutement des enseignants est une option possible pour gérer ce choix. Actuellement tous les nouveaux enseignants sont recrutés parmi ceux qui ont achevé l'en- seignement secondaire supérieur. Tandis que le recrutement à ce niveau a ses avantages en principe (et peut être approprié comme objectif à long terme), la réalité est que de tels enseignants ne sont pas plus efficaces que ceux qui n'ont achevé que l'enseignement secondaire inférieur, dans le con- texte actuel des écoles rwandaises. Mais compte tenu des différences de salaires entre les deux types d'enseignants, cela vaudrait peut-être la peine d'assouplir les critères de recrutement afin de réaliser la réduction du ratio élèves/enseignant de façon financièrement faisable et durable. Intrants de l'enseignement secondaire. Dans l'enseignement secondaire, la question de la qual- ification des enseignants est aussi hautement pertinente. Dans le cycle du tronc commun, la plu- part des enseignants satisfont à la qualification minimum (ce qui veut dire qu'ils ont au moins un diplôme secondaire supérieur) et certains ont même un diplôme d'université. En revanche, la moitié des enseignants dans les écoles secondaires supérieures sont probablement sous qualifiés, n'ayant eux-mêmes qu'un diplôme secondaire supérieur. Toute stratégie pour améliorer les résul- tats de l'apprentissage doit donc chercher à relever le profil éducationnel des enseignants du cycle secondaire supérieur. La solution évidente est de fixer et d'appliquer des normes claires pour le recrutement; mais si le problème doit être résolu plus rapidement, les décideurs pourraient chercher à rationaliser le déploiement du cadre actuel des enseignants de cycle secondaire en les réaffectant (dans la mesure du possible) entre le tronc commun et le cycle secondaire supérieur selon leur qualification. Mesure d'encouragement. Bien que le choix de la bonne combinaison des ressources soit important, il ne constitue en aucun cas toute l'explication d'une faible performance. Puisque la scolarisation est un processus social qui implique des acteurs multiples, les résultats dépendent finalement du comportement des gens. Mettre en place les bonnes incitations pour aligner les comportements sur l'objectif qui consiste à relever le niveau de l'élève constitue ainsi un défi clé dans la gestion des résultats. Les interventions potentielles à cet égard sont aussi diverses que le contexte dans lequel les gens vivent et travaillent, mais le principe d'établir des responsabilités claires et de les associer avec l'autorité de dépense et de gestion à tous les niveaux est probable- ment approprié partout. Les expériences à travers le pays ont également mis en évidence l'impor- tance de la définition et du contrôle des indicateurs tangibles de progrès comme moyen pour augmenter les incitations en vue d'une meilleure performance. EDUCATON IN RWANDA 17 Réduire au minimum les barrières à l'accès Comme précédemment indiqué, les écarts socio-économiques dans les inscriptions sont plus réduits au Rwanda que dans d'autres pays à faible revenu. Ce résultat ne devrait cependant pas pousser à la complaisance. On peut et on devrait faire davantage pour atteindre les enfants les plus vulnérables, y compris les orphelins de père et de mère, les enfants de la rue qui ne bénéfi- cient pas de la surveillance systématique des adultes, les enfants vivant dans les zones rurales et ceux des 40 pour cent des ménages les plus pauvres. Les doubles orphelins--manifestement les plus vulnérables--sont facilement identifiables et des efforts systématiques peuvent et doivent être faits pour améliorer leurs perspectives dans la vie. Pour les autres groupes, l'aide nécessaire devrait être diversifiée. Des interventions pilotes et exploratoires sont probablement appropriées pour trouver la meilleure manière d'être à leur service. Dans l'enseignement secondaire supérieur, les taux de fréquentation des enfants du quintile supérieur dépassent de loin ceux du reste de la population. Les filles ont les mêmes chances que les garçons de s'inscrire dans l'enseignement secondaire, mais elles traînent sensiblement derrière les garçons dans l'enseignement supérieur. En ce qui concerne le choix des interventions pour réduire les disparités socio-économiques des taux de fréquentation, les conclusions du rapport suggèrent que l'aide financière pourrait aider les groupes à la traîne de la même manière qu'elle aide les orphelins dans l'enseignement secondaire dans le cadre du Fonds de Génocide; cepen- dant, l'intervention devrait probablement être associée à des efforts d'amélioration des résultats des études. De tels efforts sont importants car, dans un système de sélection basée sur le mérite, élargir la représentation des groupes à la traîne n'est possible que si ces groupes sont capables de rivaliser pour les places convoitées dans l'enseignement post-primaire. Conclusion Presque une décennie après le génocide, les dirigeants du Rwanda peuvent être fiers et satisfaits de tout ce qu'ils ont pu réaliser dans le domaine de l'éducation. Ils ont remis un système dévasté sur pied: des salles de classe ont été réparées et de nouvelles ont été construites pour faire face au nombre croissant d'élèves; les enseignants qui ont fui la violence et sont rentrés ont été réintégrés dans l'enseignement; les arriérés de salaire des enseignants ont été régularisés; le Fonds du Géno- cide a été créé spécifiquement pour aider les orphelins; et un système diversifié a été mis en place dans l'enseignement supérieur. Cependant, la tâche restante demeure impressionnante au moment où la phase du redresse- ment cède la place au travail de consolidation et de développement à plus long terme du secteur. Les soucis d'efficacité, d'équité et de stabilité fiscale deviendront inévitablement de plus en plus réels au moment où le pays cherche à faire progresser le système éducatif dans un environnement de ressources limitées. Il faut espérer que les résultats de cette étude pourront contribuer à la discussion en créant une compréhension commune des problèmes et en attirant l'attention sur certains des défis émergeants. CHAPTER 1 RWANDA: DEMOGRAPHY, ECONOMY, AND THE GOVERNMENT BUDGET T he recent poverty reduction strategy paper (PRSP) for Rwanda envisions a key role for education in the government's efforts to improve the social and economic well-being of the population (Rwanda 2002a). The emphasis on education is consistent with a growing international consensus--encapsulated in the United Nations Millennium Declaration of 2000 and the Millennium Development Goals (MDGs)--that progress in education cannot be left out of the fight against poverty. This confluence of public opinion is already leading to support for increased funding for education, both within Rwanda and on the part of international donors.1 Yet one can expect that any increase in resources will be contingent on the effective use of existing resources. The consensus also carries with it the implicit expectation of tangible progress toward full coverage of the primary-school-age population, improved student learning, and a better fit between the amount and mix of skills produced by the school system and the demands of the labor market. The recent developments present exciting opportunities for educational development in Rwanda, but they also pose serious challenges. To set the stage for an examination of the choices and issues that face Rwanda's education system, this chapter presents a brief overview of the char- acteristics of the population--highlighting in particular the prevalence of orphanhood among school-age children and youths--as well as the country's macroeconomic conditions, including historical and cross-country perspectives, and its public finances. Box 1-1 provides summary back- ground information on Rwanda. 1. An indication of the potential for increased external funding is the World Bank's launch, on June 12, 2002, of the Education For All (EFA) Fast Track Initiative on behalf of a new EFA Partnership. The part- nership includes various bilateral donors active in education, as well as the European Union, the United Nations Educational, Scientific, and Cultural Organization (UNESCO), the United Nations Children's Fund (UNICEF), and the multilateral development banks. On the domestic side, countries benefiting from the Heavily Indebted Poor Countries (HIPC) Initiative have made a commitment to allocate 40 percent of the debt relief monies for investment in education (World Bank 2003). 19 20 WORLD BANK COUNTRY STUDY Box 1-1: RWANDA AT A GLANCE Geography Known poetically as the "Land of a Thousand Hills," Rwanda is a landlocked nation bordered on the north by Uganda, on the south by Burundi, on the east by Tanzania, and on the west by the Democratic Republic of Congo. The country enjoys a mild tropical climate with two rainy seasons, February to April, and November to January. A habitat that extends over parts of Rwanda, Uganda, and the Democratic Republic of Congo is home to the only 650 mountain gorillas left in the world today.With an area of 26,338 square kilometers (10,169 square miles) and an estimated 310 inhabitants per square kilometer (based on the results of the August 2002 census), Rwanda is one of the most densely settled countries in Africa.The pressure of people on scarce land poses a constant threat to social harmony and the physical environment. People According to preliminary results from the 2002 census, Rwanda's population numbers 8.16 million.The official languages are Kinyarwanda, French, and English; local dialects include Igikiga, Bufumbwa, Rutwa, and Hutu. Economy With a per capita gross domestic product (GDP) of only US$242 in 2000, Rwanda is one of the poorest coun- tries in the world. Its agriculture-dominated economy, consisting mainly of small and increasingly fragmented farms producing to meet subsistence needs, has changed little over time. Coffee and tea continue to be the country's principal exports.The manufacturing sector accounts for 20 percent of GDP and is dominated by import-substituting industries. In the 1960s and 1970s prudent financial policies, coupled with generous exter- nal aid and favorable terms of trade, fostered sustained growth, but in the 1980s, when the price of coffee plummeted, growth became erratic. GDP then declined sharply during the four years of civil war that culmi- nated in the genocide of 1994. Recently, GDP has again been growing steadily under a program of improved tax collection, accelerated privatization of state enterprises, and continued improvements in export crop and food production. Politics, civil war, and genocide Rwanda gained its independence from Belgium in July 1962. A massacre of the minority Tutsis by the majority Hutus began in 1959, sending hundreds of thousands of Tutsis into exile in neighboring countries. Since then, Rwanda has been through one episode of violence after another. Widespread massacres of Tutsis again took place during 1990­93. In an attempt to stem ethnic strife, the Arusha Peace Agreement was signed on August 4, 1993, but the death of President Juvénal Habyarimana in April 1994 sparked a genocide of unprecedented swiftness that, by one conservative estimate, left up to 800,000 Tutsis and moderate Hutus dead at the hands of Hutu militia.The Rwandan Patriotic Army (RPA) launched a counterstrike and captured Kigali, the capital city, on July 4, 1994. Over 2 million refugees fled to Tanzania and the Democratic Republic of Congo (then called Zaire). On July 19, 1994, a "Government of National Unity" was established, and a fragile process of recovery began to take hold.The country continued to experience large movements of people. Following unrest in the eastern part of Zaire, more than 600,000 refugees moved back to Rwanda in November 1996.This massive wave was followed at the end of December 1996 by the return of another 500,000 from Tanzania.At present, it is esti- mated that fewer than 100,000 Rwandans, thought to be remnants of the defeated Hutu militia, remain outside the country. Relations with the World Bank Rwanda became a member of the World Bank on September 30, 1963, and received its first World Bank credit in 1970 for a roads and highway project. Since then, it has received policy advice, technical assistance, and proj- ect and program financing in support of the country's economic development.The World Bank Group's EDUCATION IN RWANDA 21 International Development Association (IDA) and the International Monetary Fund (IMF) have agreed to sup- port a comprehensive debt reduction package for Rwanda under the Enhanced Heavily Indebted Poor Countries (HIPC) Initiative.The package will save Rwanda a cumulative total of about US$810 million in debt service over the coming years and will reduce by 71 percent the net present value of Rwanda's debt outstand- ing as of end-1999. Sources: Based on the following online resources accessed during May 2002: Central Intelligence Agency, The World Factbook: Rwanda, ; Embassy of Rwanda,"Gen- eral Information:The People," ;Valerie Percival and Thomas Homer-Dixon, 1995,"Environmental Scarcity and Violent Conflict:The Case of Rwanda," Occasional Paper, Project on Environment, Population and Security (Washington, D.C.:American Association for the Advance- ment of Science and the University of Toronto), ; Republic of Rwanda,"History of a People," ; SIL International,"Languages of Rwanda," ; U.S. Department of State, Bureau of African Affairs,"Background Note: Rwanda," November 2001, ; World Bank,"Countries: Rwanda: History," ;World Bank,"Rwanda to Receive US$810 Million in Debt Service Relief:The World Bank and IMF Support Debt Relief for Rwanda under the Enhanced HIPC Initiative," News Release 2001/192/S, . The Demographic Context Rwanda's population grew from 3.6 million in 1970 to 8.2 million in 2002 (see table 1-1). At the country's historical rates of population growth--averaging 3.8 percent a year in the 1970s and more than 3 percent a year in the 1980s--the population in 2002 would have been much bigger had it not been for the horrific genocide in 1994, which claimed at least 800,000 lives. The data in the table document a still incomplete recovery from the massacre: The urban share of population has nearly doubled, to 10.5 in 2000 from the 5.5 percent recorded in 1991, perhaps reflecting the congregation of displaced persons in the capital and in other cities where relief services may be more easily accessed. The share of the population that cannot afford a minimally adequate diet has swelled from 51 percent in 1991 to 60 percent in 2000. The under-five mortality rate, a good indicator of overall health conditions, has deteriorated by nearly 30 percent since 1991. Although women are having fewer children--an average of 5.8 births per woman over the reproductive lifespan in 2000, compared with 8.3 in 1978--the rate remains high and provides one explanation for Rwanda's exceptionally high level of maternal mortality. For comparison, maternal mortality in neighboring Tanzania and Uganda, estimated using data from similar surveys and a comparable methodology, was found to be only about half as high (Rwanda 2002b). As in most of Africa, HIV/AIDS poses a serious health threat. According to the Joint United Nations Programme on HIV/AIDS (UNAIDS), the prevalence rate among Rwandan adults ages 15­49 exceeded 11 percent in 1999. Although the rate is not as high as in, for example, South Africa, Zambia, or Zimbabwe, containing the spread of the disease remains a challenge--one that calls for a determined and sustained response from the gov- ernment and its partners. Incidence of Orphanhood Orphanhood is a characteristic feature of the demographic makeup of many countries affected by the HIV/AIDS epidemic, but in Rwanda the problem has been greatly exacerbated by the 1994 genocide. Table 1-2 presents data based on the 2000 Multiple Indicator Cluster Survey (MICS) 22 WORLD BANK COUNTRY STUDY TABLE 1-1: POPULATION SIZE, RWANDA, 1970­2002,AND POVERTY RATES AND SELECTED HEALTH INDICATORS, 1970­2000 Census year Indicator 1970 1978 1991 2000 Total population (millions) 3.6 4.8 7.2 8.2a Population increase from year in previous column (% per year) -- 3.8 3.1 1.2b Primary-school-age (ages 7­12) population as % of total -- -- 17.7 15.7 Urban share of population (%) 3.5 4.6 5.5 10.5 Percentage of population living in povertyc -- -- 51.2 60.3 Urban areas -- -- -- 19.4 Rural areas -- -- -- 65.7 Selected health and demographic indicators HIV/AIDS prevalence rate among adults ages 15­49 (%) -- -- -- 11.2 Under-five mortality (deaths per 1,000 live births) 210 -- 151 196 Maternal mortality (deaths per 100,000 live births)d -- -- -- 1,071 Total fertility rate (births per woman over reproductive lifespan) -- 8.3 6.2 5.8 -- Not available. a. Refers to the population in 2002 (based on preliminary returns from the 2002 census). b. Refers to the annual growth rate between 1991 and 2002. c.Those unable to afford a food basket yielding a minimum caloric food intake and unable to satisfy minimum nonfood needs. Minimum caloric food intake is defined as 2,500 kilocalories per adult equivalent per day in the 2000 survey; the threshold in the earlier survey is unknown.The data for 1991 refer to the results from a survey conducted in 1993, as reported in World Development Indicators 2002. d. Refers to estimate for 1995­2000, based on the 2000 Demographic and Health Survey for Rwanda (Rwanda 2002b). Sources: For population, 1970, 1978, and 1991; urban share of population, 1970, 1978, and 1991; share of school- age population, 1970, 1978, and 1991; and intercensal population growth rates, 1970­91: Rwanda (1994b). For urban share of population and share of population living in poverty, 2000: Rwanda (2000b). For HIV/AIDS infec- tion rate: UNAIDS, at . For share of population living in poverty, 1993; under-five mortality rate and total fertility rate, 1978 and 1991; population and share of school-age population, 2000: United Nations Population Division estimates and projections as reported in the World Bank's Statistical Information Manage- ment and Analysis (SIMA) database and in World Development Indicators 2002. For total fertility rate, 2000: Rwanda DHS report (Rwanda 2000b). Estimated population growth, 1991­2002, is computed using population data shown in the table. sponsored by the United Nations Children's Fund (UNICEF), which collected information on more than 9,000 children below age 15 in the sampled households. An average of 28.5 percent of these children had lost at least one parent, usually the father. Among 10-to-14-year-olds, the share was much higher, exceeding 40 percent. Five percent of the children in the sample had lost both parents; again, the share among older children was significantly higher. Cross-Country Comparison of the Incidence of Orphanhood Figure 1-1 shows relevant cross-country data on orphanhood rates for around 2000. The rate in Rwanda among children ages 7­14 was estimated at nearly 40 percent. This figure is more than 1.5 times as high as the rates in Uganda and Zimbabwe, more than 2.5 times those in Malawi, Mozambique, and South Africa, and nearly 4 times those in Kenya and Tanzania.2 Although the 2. Estimated orphanhood rates in such countries as Uganda and Mozambique reach 20 percent or close to it. It is thus likely that Rwanda's orphanhood rate would be at least as high. EDUCATION IN RWANDA 23 or yb e es siz umber) 2,541 1,862 1,633 1,793 1,251 9,080 1,594 1,284 471 Sample opulationP figur (n ya wa ents separate, par (%) 2.7 11.7 14.3 17.1 20.4 12.0 105 202 96 Living om mparable fr co y highl both 2000 (%) is or ents 9.9 A 25.1 34.4 40.5 44.7 28.5 261 482 210 par girls dead One CIRC and , A ys bo (%) AND ents W 0.5 3.2 6.3 7.9 Both 10.3 5.0 26 92 49 ,R par among dead ARENTSP one (%) ent orphanhood AL 9.5 21.8 28.1 32.6 34.4 23.5 235 391 162 of otal,T par dead dead IOLOGIC ent incidence B (%) M par 8.8 the .yev ORF atherF 18.7 24.1 27.4 28.6 20.1 208 332 134 sur One dead T Because MICS AR AP (%) 0.7 3.1 4.0 5.2 5.8 3.4 28 59 27 eights. 2000 w Mother IVING dead the L om fr sampling a yb, data HILDREN C earsy vincialo on in, pr yb based AND Cluster oup gr nationwide ega (thousands) eighted w RPHANS Indicator yb wn. 0­14 O earsy e calculations ar sho oups, population in, not 1-2: Multiple 2000,,yev gr e oup Authors' age gr centagesreP ar ABLE T omrF 0­3 4­6 7­9 10­12 13­14 All Sur ega 0­6 7­12 13­14 a. Source: Estimated gender 24 WORLD BANK COUNTRY STUDY FIGURE 1-1: ADULT HIV/AIDS PREVALENCE RATES AND SHARE OF ORPHANS AMONG CHILDREN AGES 7­14, RWANDA AND OTHER EAST AFRICAN COUNTRIES, CIRCA 2000 50 40 30 Percent 20 10 0 Kenya Africa Uganda Malawi Zambia Tanzania Rwanda Zimbabwe Mozambique South HIV prevalence, 2000 Percentage of orphans among children ages 7Ð14 Note: Orphans refers to children ages 7­14 who have lost at least one parent. Data refer to 1992 for Malawi, 1996 for Tanzania, 1997 for Mozambique, 1998 for Kenya, South Africa, and Zambia, 1999 for Zimbabwe, and 2000 for Rwanda and Uganda. Source: For HIV/AIDS prevalence rates, UNAIDS, at ; for orphanhood rates for Rwanda, table 1-2 and for other countries, Ainsworth and Filmer (2002). HIV/AIDS epidemic has created many orphans in all these countries, the much higher rate in Rwanda leaves no doubt as to the devastating legacy of the 1994 genocide. The high prevalence of orphanhood in Rwanda has obvious implications for the education system. Orphans have few or no means to invest in their schooling without significant help from others. The sheer size of the problem calls for action on a commensurate scale, probably with government leadership in support of efforts by nongovernmental organizations (NGOs), local communities, and concerned individuals. In 1998 the government set up a Genocide Fund to mobilize resources to provide systematic financial support to orphans.3 The fund was established by law and is capitalized by an allocation of 5 percent of domestic revenues. The first disburse- ments were made in 2000 to about 33,000 students. The effort clearly has a long way to go, since the fund currently supports only orphans in secondary school. As in any scheme providing financial aid, an important challenge will be to make sure that funds reach their intended benefi- ciaries, particularly considering that cases of fraudulent claims have already been reported. Beyond the financial cost, an even more daunting task is that of educating and nurturing large numbers of traumatized children who have lost the most significant adults in their lives. The challenge will undoubtedly require that frontline workers--in this case, teachers and school administrators--go far beyond the ordinary duties of classroom teaching to address the psycho- logical needs of their charges as well. Macroeconomic Conditions Table 1-3 shows Rwanda's per capita gross domestic product (GDP) in 1980 and 2000 and the average annual growth rate of GDP in the two decades separating those years. For comparison, the table also reports similar data for countries in Africa and Asia with per capita GDP comparable to 3. The French name of the Genocide Fund is Fonds national pour l'assistance aux rescapés du génocide et des massacres au Rwanda (FARG). EDUCATION IN RWANDA 25 TABLE 1-3: ECONOMIC PERFORMANCE OF RWANDA AND SELECTED COUNTRIES, 1980­2000 GDP per capita as multiple Average GDP of Rwanda's growth rate, 1980­ Country 1980 2000 2000 (% per year) Rwanda 1.0 1.0 1.1 (Per capita GDP in constant (321) (242) 1995 US$) Sierra Leone 0.91 0.61 ­1.2 Malí 0.95 1.19 2.2 Nigeria 0.98 1.05 1.8 Níger 1.02 0.84 0.9 Kenya 1.05 1.36 2.8 Madagascar 1.07 1.02 1.1 Benin 1.13 1.71 3.6 Lesotho 1.14 2.28 4.0 Gambia,The 1.17 1.53 3.5 Ghana 1.23 1.71 3.2 Nepal 0.46 1.00 4.8 Bangladesh 0.70 1.54 4.7 India 0.70 1.90 5.5 Pakistan 0.99 2.14 5.0 Source: World Bank Statistical Information Management and Analysis (SIMA) database, as of June 5, 2002. or lower than Rwanda's in the 1980s. Rwanda's record is dismal in comparative perspective. GDP expanded at only 1.1 percent a year between 1980 and 2000, and per capita GDP in 2000 was only about three-quarters as high as in 1980. The performance of the economy over the two decades was comparable to that of Madagascar, better than that of Sierra Leone, and significantly worse than in Kenya, Benin, Lesotho, The Gambia, and Ghana. Rwanda's slow growth comes into even sharper relief in light of the record of growth in low-income Asia, whose economies expanded at average rates of around 5 percent a year during the period. By 2000, Nepal, which started with a per capita GDP less than half that of Rwanda's in 1980, had caught up with Rwanda, while Bangladesh, which lagged Rwanda by 30 percent in 1980, had achieved a per capita GDP 50 percent higher than Rwanda's. Although it is beyond the scope of this report to evaluate Rwanda's growth prospects, the com- parative data suggest that more rapid growth should be possible. Faster growth would make it easier to reduce poverty--Devarajan (2001) estimates that a 1 percent increase in the growth rate could reduce the poverty rate by between 1 and 2 percent--and improve educational and other social outcomes. It is widely recognized, however, that the trickle-down benefits of growth alone are inad- equate (Devarajan 2001). Going one step further, Devarajan and Renikka (2002) argue that even increases in public spending would have limited impact and that progress is made by targeting pub- lic spending to fund the right types of services for the right groups of people; by ensuring that the money reaches frontline service providers; by improving the incentives for service providers to serve their clients; and by addressing the constraints, financial or otherwise, that households may face in making use of the services offered. Subsequent chapters will address some of these issues. 26 WORLD BANK COUNTRY STUDY Overall Patterns of Government Finance The time-series data cited in this section show the expected sharp discontinuities around 1994, the year of the genocide. Because the data for the mid-1990s are probably less comparable with those for other years, we begin with the 1980s to enable observation of long-term trends during the two decades. On the revenue side (table 1-4), revenues, including grants, ranged between 11 and 14 percent of GDP during the 1980s. In the 1990s, they were generally higher-- between 13 and 18 percent--except in 1994. Because grants make up a significant and generally rising share of these revenues (40 percent of total revenues in 2001), the volume of revenue mobilized domestically has grown only slowly in the two decades, from about 8.5 percent of GDP in the early 1980s to about 10 percent by the late 1990s and early 2000s. Compared with other countries, this level of domestic revenue generation is on the modest side: the correspon- ding averages around 2000 were 15.8 percent of GDP for 30 low-income African countries, TABLE 1-4: GOVERNMENT REVENUE, RWANDA, 1981­2001 Total government revenue, including Government revenue, excluding grants grants As share of GDP As share of GDP Grants as Billions of Index Billions of Index share of total Year current FRw % (1981 = 100) current FRw % (1981 = 100) revenue (%) 1981 14.9 12.2 100.0 10.4 8.5 100.0 30.1 1982 15.2 11.6 95.5 11.7 8.9 104.6 23.5 1983 15.4 10.9 89.1 11.6 8.1 95.6 25.1 1984 17.9 11.3 92.5 14.3 9.0 105.5 20.3 1985 21.2 12.2 100.1 17.2 9.9 116.2 18.9 1986 23.5 13.8 113.5 19.8 11.6 136.9 15.7 1987 23.1 13.5 110.8 19.7 11.5 135.3 14.7 1988 22.9 12.5 102.5 17.1 9.4 110.0 25.0 1989 24.4 12.7 104.1 19.2 9.9 116.8 21.6 1990 27.5 12.9 105.6 21.6 10.1 118.8 21.4 1991 36.1 15.1 123.8 25.0 10.4 122.8 30.7 1992 44.3 16.2 133.2 27.6 10.1 118.7 37.8 1993 44.0 15.6 128.1 25.9 9.2 107.7 41.2 1994 7.5 4.6 37.4 6.0 3.6 42.8 20.1 1995 61.5 18.1 149.0 23.1 6.8 80.2 62.4 1996 70.8 16.7 137.1 39.4 9.3 109.2 44.3 1997 95.8 17.2 141.0 58.0 10.4 122.2 39.4 1998 99.0 15.9 130.9 66.0 10.6 124.9 33.3 1999 101.9 15.8 129.9 63.5 9.9 115.9 37.7 2000 132.3 19.0 156.0 68.5 9.8 115.6 48.2 2001 139.5 18.2 149.6 83.9 11.0 128.8 39.9 Note: FRw, Rwandan francs; GDP, gross domestic product. Source: Authors' calculations, based on data from the World Bank Africa Live Database (April 24, 2002, version). See also appendix table A1-1. EDUCATION IN RWANDA 27 17.4 percent for 11 non-African low-income countries, and 11.6 percent for the seven countries other than Rwanda with per capita GDP between US$200 and US$300.4 A World Bank (1999) study on development prospects in Madagascar (which has a per capita GDP comparable to Rwanda's) is instructive in this regard: it suggests that Rwanda would need to raise domestic revenue generation to around 15 percent to finance its poverty reduction agenda. We turn next to the expenditure side (table 1-5). As a percentage of GDP, total public current and capital spending in the late 1990s and early 2000s was about the same as in the 1980s--that is, around 20 percent. The ratio of public spending to total revenues was lower in the late 1990s and early 2000s than in the 1980s and early 1990s, a trend that is consistent with rising total revenue in the past two decades. Current spending net of interest--one measure of the cost of running the gov- ernment--has remained relatively stable throughout the two decades, although the data for the most recent years, 1998 to 2001, show a slight increase over the averages in the 1980s. Interest on the public debt is not as burdensome as in some other low-income countries; it accounted for around 4­6 percent of total spending, or about 1 percent of GDP, throughout the period, except for the years around 1994. The interest paid on external debt has been smaller in recent years than in the 1980s and the first half of the 1990s. It reached a historical low of less than 25 percent in 2001. Public Spending on Education In this section we examine the trend in public spending on education in the overall context of public finances described above. Particular attention is given to whether the emerging modest rise in public spending in recent years has translated into a similar tendency in spending on education. We then look at how the increase, if any, has been distributed between capital investment and the running cost of service delivery. As table 1-6 shows, total public spending on education was noticeably higher from 1999 onward than in the 1980s, reaching 5.5 percent of GDP and more than a quarter of total government spending in 2001, compared with an average of 3.4 percent of GDP and less than a fifth of total government spending in the mid-1980s. Expressed as a share of domestic govern- ment revenues, the increase in spending on education is even more striking: by 2001, education spending accounted for more than half of such revenues, compared with an average of 33 percent in the mid-1980s. The increase in spending on education reflects rising capital investments in the sector rather than increased allocations for service delivery. In the three years 1999, 2000, and 2001--the years with the highest historical levels of total spending on education relative to GDP--the current share of total spending averaged 73 percent, compared with more than 94 percent in the mid-1980s. As a result, current spending on education as a share of GDP in the late 1990s and early 2000s was only slightly better than the average of 3.1 percent during the 1980s. The shift in spending toward capital investment after 1994 is consistent with the efforts to rebuild educational infra- structure destroyed during the genocide. Yet at some point, one would expect current spending to begin growing as well, given that the capital investment is presumably meant to translate into expanded service delivery for a larger population of students. Although it is not our purpose to make a detailed assessment of the prospects for an increase in current spending on education, a few patterns and developments are noteworthy. An increase in current spending on education would depend on three underlying parameters and their expect- ed trends: domestic revenues, external resource flows, and the share of education in public cur- rent spending. The first and last parameters are unlikely to rise dramatically, given their historical trends in Rwanda and the fact that education already claims more than a quarter of total public current spending net of debt interest payments. That leaves, as a source of higher spending on education, increased external support. 4. The data are based on the World Bank Africa Live Database (April 24, 2002, version). The averages for the low-income African samples exclude data for Angola, Eritrea, Lesotho, and Nigeria, where, for various reasons, government revenues as a percentage of GDP are exceptionally high, exceeding 30 percent. 28 WORLD BANK COUNTRY STUDY TABLE 1-5: GOVERNMENT SPENDING, RWANDA, 1981­2001 Current spending Total spending net of interest As share of GDP payments Interest on all public debt As ratio Share of Billions of total As % As % of interest on current of Index government of total As % of total As % of external Year FRw % (1981 = 100) revenue spending GDP spending GDP debt (%) 1981 22.6 18.4 100 1.51 57.8 10.6 1.7 0.3 57.9 1982 25.7 19.6 106 1.68 54.3 10.6 2.3 0.5 47.1 1983 28.6 20.1 109 1.85 53.6 10.8 2.5 0.5 66.6 1984 23.9 15.0 81 1.33 62.1 9.3 4.4 0.7 67.0 1985 28.4 16.3 89 1.34 56.1 9.2 4.3 0.7 68.9 1986 31.6 18.6 101 1.34 57.7 10.7 4.4 0.8 69.7 1987 37.3 21.7 118 1.61 57.5 12.5 4.9 1.1 67.9 1988 34.6 18.9 103 1.52 60.8 11.5 5.9 1.1 66.2 1989 35.7 18.5 101 1.46 60.1 11.1 5.7 1.1 63.5 1990 42.1 19.7 107 1.53 64.9 12.8 5.1 1.0 70.0 1991 55.6 23.2 126 1.54 62.1 14.4 7.7 1.8 76.3 1992 65.8 24.1 131 1.49 61.6 14.9 7.9 1.9 73.7 1993 65.2 23.1 126 1.48 57.4 13.3 9.4 2.2 73.8 1994 26.6 16.1 87 3.53 55.7 8.9 27.8 4.5 66.2 1995 69.5 20.5 111 1.13 49.3 10.1 11.3 2.3 51.4 1996 95.3 22.5 122 1.35 51.4 11.5 7.2 1.6 42.0 1997 109.6 19.6 107 1.14 52.2 10.2 6.2 1.2 44.3 1998 117.4 18.9 103 1.19 59.3 11.2 4.9 0.9 40.6 1999 126.9 19.7 107 1.25 63.5 12.5 4.6 0.9 40.7 2000 131.8 18.9 103 1.00 62.8 11.9 4.9 0.9 30.8 2001 164.2 21.4 116 1.18 60.5 13.0 4.0 0.9 24.2 Note: FRw, Rwandan francs; GDP, gross domestic product. Source: Authors' calculations, based on data from the World Bank Africa Live Database (April 24, 2002, version). See also appendix table A1-1. The prospects for greater external support appear promising, although still tentative at this stage. Like other countries, Rwanda benefited from debt relief under the Heavily Indebted Poor Countries (HIPC) Initiative, and the expectation is that it, like the other countries, will use the resources thus freed to increase spending on education.5 Another reason for optimism about the availability of increased donor funding for education is the growing international consensus on and support for the Millennium Development Goals (MDGs), one of which is completion of primary education by all children by 2015. The MDGs are challenging donors to increase external funding 5. It is recognized, however, that Rwanda is already spending a large proportion of its domestic budget on education and that the health sector is, in relative terms, seriously underfunded. Hence, the health sector may receive priority over education in benefiting from the HIPC money. EDUCATION IN RWANDA 29 TABLE 1-6: GOVERNMENT SPENDING ON EDUCATION, RWANDA, 1981­2001 Total spending on education Current spending on education As % of As % of As % of As % of government total public current spending As % of total revenues net spending on As % of net of debt Year GDP spending of grants education GDP interest payment 1981 -- -- -- -- 2.7 25.2 1982 -- -- -- -- 3.0 27.8 1983 -- -- -- -- 3.2 29.8 1984 3.3 22.1 36.8 97.2 3.2 34.5 1985 3.1 18.9 31.2 94.3 2.9 31.8 1986 3.5 18.7 29.7 93.8 3.2 30.3 1987 3.5 16.1 30.5 92.4 3.2 25.9 1988 -- -- -- -- 3.1 27.0 1989 -- -- -- -- 3.0 26.5 1990 -- -- -- -- 2.9 22.9 1991­95 -- -- -- -- -- -- 1996 3.2 14.5 35.0 62.6 2.0 17.6 1997 3.4 17.5 33.1 59.1 2.0 19.9 1998 3.1 16.7 29.6 71.1 2.2 20.0 1999 4.3 21.9 43.8 78.1 3.4 27.0 2000 4.0 21.0 40.3 80.9 3.2 27.0 2001 5.5 25.6 50.2 60.2 3.3 25.5 -- Not available. Note: GDP, gross domestic product. Source: Authors' calculations, based on data from the World Bank Africa Live Database (April 24, 2002, version). See also appendix table A1-1. for education; one early outcome has been the launching in June 2002 of the Education For All (EFA) Fast Track Initiative. The confluence of international opinion and domestic prioritization in favor of education is likely to enhance the availability of resources for the sector. But given the keen competition for resources, both among countries and across sectors within Rwanda, the reality is likely to consist of a gradual growth in current spending on education rather than a massive rise over a short period. Whatever the magnitude of future increases, an ever-present task is to ensure that all resources, especially those currently available to the sector, are used as effectively as possible. The chapters that follow examine various aspects of the effectiveness issue. Here, we simply note that there is evidence at both the country level and the school level suggesting the potential for improvement in this regard. Conclusion The idea that education can play a key role in the government's poverty reduction strategy cur- rently enjoys support within Rwanda, as well as in the international donor community. To deliver on this expectation, the education sector faces the triple challenge of ensuring that no child is 30 WORLD BANK COUNTRY STUDY denied the chance to complete primary school; that students at all levels make adequate progress in acquiring the knowledge and skills to equip them for adult life; and that the mix and number of graduates from the education system correspond to the economy's demand for skilled labor. In Rwanda the legacy of the 1994 genocide and the continuing spread of HIV/AIDS add to the difficulty of an already demanding task. Putting more money into the sector is probably a necessary, although by no means sufficient, part of meeting the challenge. Public spending on education in Rwanda reached a historical high in 2001, but because much of the increase after 1994 has gone into capital investment, current spending to support the day-to-day running of the system remains at about the level prevailing in the 1980s--slightly more than 3 percent of GDP. In low-income African countries that already put all their children through primary school, total current spending on education at all levels currently averages about 4.5 percent of GDP. Raising public spending to this level will not be easy in Rwanda, given the country's historically small tax base and the large proportion (about 25 percent) of the government's discretionary spending that already goes for education. Donor funding can potentially help supplement government spending. Although the prospects are good in light of international commitment to the MDGs, the competition for such resources is keen, and the amount available may cover only a part--and possibly not a large part--of the potential shortfalls in funding for the sector. Even as increased funding for education is sought, it is important to recognize that it is not the only, or even the most important, part of a strategy for enhancing the contribution of educa- tion to the country's social and economic development. Making good use of currently available resources is at least as important, and in Rwanda's tight public budget context, efforts in this direction can hardly be ignored. Tough questions arise in this regard. Are public funds financing the right types of services, and do they benefit the intended population groups? Are the alloca- tions reaching down to the school level, where teachers and school administrators are ultimately responsible for delivering services? Are teachers and their managers given the right incentives to deliver services effectively, and do students and their families receive appropriate incentives, where needed, to encourage school participation? In the next chapter, we begin our examination of these questions by providing a broad overview of the coverage of the education system, including enrollment trends and student flow patterns. CHAPTER 2 ENROLLMENT TRENDS AND PATTERNS OF STUDENT FLOW D espite severe setbacks following the 1994 genocide, Rwanda's education system has made a remarkable recovery, in numerical if not in qualitative terms, particularly at the primary level. Yet the rebound in the space of just seven years is already beginning to give rise to what might be termed second-generation challenges, including the risk of backsliding in the face of resource and capacity constraints. Two of the most pressing of these challenges relate to (a) the potential erosion in educational quality as enrollments expand rapidly in a context of difficult tradeoffs in resource use and (b) the inevitable buildup of pressures to expand postpri- mary education, as growing numbers graduate from the primary and secondary cycles. Aggregate Enrollments by Level of Education Rwanda's education system has a 6-3-3-4 structure: six years of primary schooling, three years of lower secondary schooling (tronc commun), three years of upper secondary schooling, and an average of four years of higher education. This section documents the numbers of students enrolled at the various levels and the growth in enrollments over time. A consistent time series exists for primary and higher education from the 1970s. For secondary education the series cov- ers only the postgenocide period because the data from earlier years suffer from reporting prob- lems and from inconsistencies in definition. Primary Schooling Primary school enrollments stagnated at around 0.4 million pupils in the first half of the 1970s and then began to climb steadily at an average rate of 3.9 percent a year between 1975 and 1993 (see figure 2-1).1 1. The duration of the primary cycle changed twice during the period covered by figure 2-1, going from six to eight years in 1980­81 and back to six years in 1992­93. (The first change was formalized by law in 1985 and the second in 1991.) To ensure comparability in the data for the entire period, the enrollments in years when the eight-year cycle was in effect were adjusted downward by about 12 percent, reflecting the share of students in the last two grades in the eight-year cycle. 31 32 WORLD BANK COUNTRY STUDY FIGURE 2-1: ENROLLMENT TRENDS IN PRIMARY SCHOOLING, RWANDA, 1971­2001 1.6 1.4 Extrapolation for 1994-2001 based on pregenocide trend 1.2 1.0 Trend line, 1976-93 pupils of 0.8 0.6 Millions 0.4 Genocide, 1994 0.2 Actual enrollments 0.0 1970 1975 1980 1985 1990 1995 2000 Source: See appendix table A2-1. Total enrollments stood at 1.2 million children in 1993, but when schools reopened after the genocide, there were nearly 30 percent fewer pupils. Amazingly, enrollments rebounded immedi- ately and have been growing vigorously ever since. By 2001, the number of children in the sys- tem had reached 1.5 million, surpassing what might have been expected had the system expanded at the same rate as in the pregenocide years. The rapid pace of enrollment growth is striking, perhaps reflecting the government's deliberate efforts to expand basic educational opportunities as part of the nation's recovery plan. Throughout the period, private schooling has played a minimal role, accounting for less than 0.8 percent of primary school enrollments throughout the 1990s. The historical growth in enrollments is therefore largely a public sector phenomenon, involving schools financed almost entirely by the government and managed either by the government itself or, in the case of what are termed libre subsidié schools, by churches and other organizations. Secondary Education The available data for secondary education refer to the years since 1996.2 Figure 2-2 shows the number of students in the formal secondary school system and the share of the private sector. From a base of only 50,000 students in 1996, the system grew to more than 141,000 students by 2001, implying a growth rate of more than 20 percent a year, much faster than the 7 percent a year growth rate at the primary level during the same period.3 The share of students in private 2. As shown in appendix table A2-1, the available data for secondary education reflect enrollments that are not always consistently aggregated across the state, libre subsidié, and private sectors. (Libre subsidié schools are funded by the government but are run by nongovernmental organizations.) Furthermore, the system's structure changed during the 1980s and early 1990s in that the three-year postprimary vocational training offered in the centres de l'enseignement rural et artisanal intégré (CERAI), the girls-only section familiale (SF), and the boys-only centres de l'enseignement rural et artisanal de Rwanda (CERAR) was dis- continued after 1992. The big jump in reported enrollments in general secondary education for 1992 (62,701, compared with just 45,098 for the previous year) suggests that some of the enrollments may have included students who were previously enrolled in the CERAI, SF, and CERAR programs. These features of the data make it impossible to construct a consistent time series except for the years since 1996. 3. If the CERAI, SF, and CERAR courses are considered part of secondary schooling, total enrollments in secondary education in 1988­91 (just before these programs were discontinued) would average about 66,000. Taking this as the base would imply that access to some form of secondary schooling was expanding at the rate of about 7 percent a year between the period 1988­91 and 2001--which is still an impressive pace. EDUCATION IN RWANDA 33 FIGURE 2-2: ENROLLMENT TRENDS IN SECONDARY EDUCATION, RWANDA, 1996­2001 160 60 (%) Private share (%) 140 50 students) 120 schools of 40 100 private 80 30 in Enrollment 60 (thousands 20 40 students 10 20 of Enrollment 0 0 1996 1997 1998 1999 2000 2001 Share Source: See appendix table A2-1. schools has remained relatively stable at about 45 percent, suggesting that enrollments grew equally fast in the private and public sectors. The private sector share in Rwanda is much greater than the average of about 20 percent for a sample of 25 African countries for which relevant data are available; of these, only Botswana, Mauritius, Tanzania, and Zimbabwe have more of their secondary students in the private sector than does Rwanda. Higher Education Enrollments in higher education have also grown rapidly (see figure 2-3). From a base of some 1,200 students in 1980­81 in a single public institution, the Université nationale du Rwanda, and a few private institutions, the system's enrollment had by 2001­02 expanded to nearly 17,000 students spread across six public and eight private institutions.4 Enrollments grew at about 10.6 percent a year between 1980­81 and 1990­91, but growth in the postgenocide years quickened to an average rate of about 29.0 percent a year between 1996­97 and 2001­02. Throughout the past three decades, the private sector in higher education has expanded faster than the public sector. The share of enrollments in the private sector increased from 9.5 percent in 1980­81 to 22.6 percent in 1989­90 and rose even more in the postgenocide period, particu- larly in the past few years, reaching nearly 38 percent by 2001­02. 4. The private institutions in operation in 1980­81 were as follows: Institut Africain et Mauricien de statistiques et d'économie appliquée, Grand séminaire de Nyakibanda, Faculté de théologie de Butare, Grand séminaire de Kabgayi, and Centre d'enseignement supérieur de Kigali. Currently, the public institu- tions are the Université nationale du Rwanda, Kigali Institute of Science, Technology and Management, Kigali Institute of Education, Kigali Health Institute, Institut supérieur d'agronomie et d'élevage, and Insti- tut supérieur des finances publiques. The private institutions include the Université libre de Kigali, Univer- sité laïque de Kigali, Institut supérieur de pédagogie de Gitwe, Université adventiste d'Afrique centrale, Grand séminaire de Nyakibanda, Faculté de théologie de Butaré, Grand séminaire de Kabgayi, and Centre d'enseignement supérieur de Kigali. Some private institutions that existed in earlier years have closed or have merged with other institutions; these include the Institut Africain et Mauricien de statistiques et d'économie appliquée, Institut supérieur catholique de pédagogie appliquée de Nkumba, and Ecole supérieure de ges- tion et d'information. 34 WORLD BANK COUNTRY STUDY FIGURE 2.3: ENROLLMENT TRENDS IN HIGHER EDUCATION, RWANDA, 1975­2002 20,000 40 (%) 17,500 35 15,000 30 sector Private share (%) Enrollments 12,500 25 private 10,000 20 in enrollments 7,500 15 Total 5,000 10 students of 2,500 5 Genocide, 1994 0 0 Share 1975-76 1980-81 1985-86 1990-91 1995-96 2000-01 Note: Breaks in the series denote years when classes were canceled (années blanches) or for which data are unavailable. Source: See appendix table A2-1. Trends in Educational Coverage The next question concerns how the expanding enrollments are related to the coverage of the school-age population by the education system. Coverage is typically measured using gross and net enrollment ratios. The gross enrollment ratio (GER) is defined as the ratio between all stu- dents enrolled in a given cycle of education and the population in the official age range for that cycle. The net enrollment ratio (NER) is defined in a similar way, except that the numerator includes only students in the official age range for that cycle of schooling. In Rwanda the age range for primary schooling is 7­12 years and for secondary schooling, 13­18 years. For higher education, the standard practice is to define the age range as the five-year bracket following the last year of the range for secondary schooling (ages 19­23 in the case of Rwanda). At this level of education, the inconsistency between the ages of the enrolled and reference populations tends to be wider than in the lower levels of schooling, reflecting the cumulative effects of late entry and grade repetition earlier in the system, as well as the likelihood that many secondary school gradu- ates will work for a few years before enrolling in higher education. Instead of the gross enroll- ment ratio, some analysts therefore prefer the number of students per 100,000 population as a measure of coverage in higher education. Coverage in 1991­92 and in 2000­01 In this section we focus mainly on the gross enrollment ratio to simplify the presentation and facilitate cross-country comparison.5 Table 2-1 shows estimates for 1991­92 and 2000­01 based 5. Education systems in which children begin their schooling on time and few repeat a grade can expect to have gross and net enrollment ratios that are practically the same. When the ratios do diverge, both late entry and grade repetition play a role, although it is impossible to ascertain their relative contribution merely on the basis of the size of the divergence between the two ratios. The further up the educational ladder one goes, the more problematic is the net enrollment ratio as a measure of coverage because of the growing inconsistency in the age ranges of the numerator and denominator populations stemming from the cumula- tive effects of late entry and grade repetition in earlier cycles of schooling. If, for example, 20 percent of the students in each entering class in secondary school are one year older than the official entry age for this cycle (because they started their primary schooling late or repeated a grade) and there is no grade repetition in secondary school, the net enrollment ratio would suggest a level of coverage that is about 20 percent lower than the reality, simply because the over-age students have been excluded from the calculation. EDUCATION IN RWANDA 35 TABLE 2-1: GROSS ENROLLMENT RATIOS (GERS) BY LEVEL OF EDUCATION, RWANDA, 1991­92 AND 2000­01 Educational level and type of estimate 1991­92a 2000­01b Primary Survey-based GER estimate (%) DHS 73.9 106.6 QUID -- 108.3 Census-based GER estimate (%) 72.5 107.3 Number of pupils in grades 1­6 (thousands)c 962.3 1,428.7 Population ages 7­12 (thousands) 1,327.2 1,332.0 Secondary Survey-based GER estimate (%) DHS 20.4 17.0 QUID -- 10.8 Census-based GER estimate (%) 21.8 12.3 Number of students in secondary grades 1­6 (thousands)d 202.5 125.1 Population ages 13­18 (thousands) 928.1 1,014.3 Higher Survey-based GER estimate (%) QUID -- 1.3 Census-based GER estimate (%) 0.5 1.9 Total enrollment 3,389 12,757 Population ages 19­23 (thousands) 604.4 660.6 -- Not available. Note: DHS, Demographic and Health Surveys; QUID, Questionnaire unifié sur les indicateurs de développe- ment (also known as Core Welfare Indicators Questionnaire, CWIQ). a.The survey-based estimates refer to 1992; the census-based estimates refer to 1991, the year of the popula- tion census. b.The survey-based estimates refer to 2001; the census-based estimates refer to 2000. c. In 1991­92 primary schooling lasted eight years. Only pupils in grades 1­6 are included here, to ensure comparability with the GER estimate for 2000­01, when the cycle was only six years. d.To enhance the comparability of the data for 1991­92 and 2000­01, the data for 1991­92 refer to enroll- ments in general secondary school, the three-year postprimary vocational training centers (CERAI, SF, and CERAR), and grades 7 and 8 of the primary cycle. Source: The survey-based estimates of the GER rely on data collected in the 1992 and 2000 Rwanda Demo- graphic and Health Surveys and those collected in the 2001 QUID.The census-based estimates rely on enroll- ment data collected by the Ministry of Education through its annual census of schools (see appendix table A2-1) and on population data from the 1991 population census and the United Nations population projections for 2000 as reported in the World Bank's Statistical Information Management and Analysis (SIMA) database. on alternative sources of data for each of the three levels of education.6 For primary schooling, the estimates for both periods are highly consistent, with GERs of about 73­74 percent for the earlier period and 107­108 percent for the later period. Primary schooling lasted eight years in 1991­92; to ensure comparability with the estimates for 2000­01, the calculations have been adjusted so that the GER refers to six years of schooling. According to these results, the educa- tion system is now reaching a larger share of school-age children than it did in the early 1990s. 6. Besides being good practice, the use of alternative data sources to verify the estimates is particularly important in a country such as Rwanda where the routine work of collecting school statistics has suffered severe disruption in recent years and is only now being gradually revived. 36 WORLD BANK COUNTRY STUDY Later chapters of this report, however, show that the system's progress in quantitative terms has probably not been matched by a concomitant gain on the qualitative front. For secondary education, the estimates show greater consistency across data sources for 1991­92 than for 2000­01. Taken as a whole, the general picture appears to suggest a probable decline in coverage over the period. This interpretation needs to be placed in the context of the structural changes in the school system between the two dates. In 1991­92 the primary cycle lasted eight years, and significant numbers entered the three-year postprimary vocational schools (CERAI, SF, and CERAR).7 Here, computations of the GER for 1990­91 include the numerator enrollments in grades 7 and 8 of the primary cycle and those in the postprimary vocational schools. The two extra years of primary schooling in the eight-year cycle and the vocational courses are obviously not comparable to the three-year tronc commun (lower secondary) cycle in the current educational structure, but the treatment here is arguably valid for comparing changes in the course of the 1990s in access to schooling beyond the first six years of the primary cycle. Although coverage may not yet be quite as extensive as in the early 1990s, the current system is perhaps more equitable in that it offers all children who make it to the secondary cycle a chance for continued progression up the educational ladder--rather than foreclosing this prospect by channeling a large proportion of primary school graduates into a vocational path that in fact has been an educational cul-de-sac. In higher education there are two estimates of the gross enrollment ratio for 2000­01: 1.3 percent according to survey results, and 1.9 percent calculated from enrollment and population data. One explanation for the smaller size of the survey-based estimate is that the data were col- lected just before the most recent explosion in enrollments--from 10,058 in 1999­2000 to 12,757 in 2000­01 and to 16,668 in 2001­02. These data suggest that coverage of higher edu- cation as measured by the GER has increased nearly fourfold since 1990­91. Cross-Country Comparison Table 2-2 presents averages for low-income countries worldwide and in Sub-Saharan Africa, as well as data for selected African countries.8 For higher education, the table includes enrollments per 100,000 population as a complementary measure of coverage. Rwanda's GER for primary education is substantially higher than the averages for low-income countries in Sub-Saharan Africa and in the world, whereas its ratio for secondary education is significantly lower and that for higher education is comparable to the corresponding averages.9 These patterns suggest that access to primary schooling relative to the other levels is probably wider in Rwanda than in other low-income settings. It is to be emphasized that this conclusion does not in itself offer a sufficient basis for policy development. For example, if the ratio for pri- mary schooling is high because of widespread grade repetition, the education system may contin- ue to suffer from poor rates of access to and retention within this level of schooling. In secondary education the ratio may be modest in international comparisons, but the pattern needs to be further evaluated by looking at the lower and upper secondary levels separately. In higher educa- tion the comparability of the country's gross enrollment ratio with GERs for other low-income countries does not necessarily imply that public policy has been successful. What is more impor- tant is whether the output of graduates matches the labor market's capacity to absorb the gradu- ates in productive employment. The next section documents patterns of student flow; chapter 8 examines the education­labor market issue. 7. Enrollments in the three-year vocational schools (described in note 2) were between 25,000 and 26,000 in 1990­91, compared with between 36,000 and 39,000 in the general secondary schools (see appendix table A2-1). 8. Despite its flaws, the GER serves our purpose here, as the data for this indicator are readily available for a large number of countries. 9. In higher education the data for 2001­02 suggest that the country has now more or less caught up with the Sub-Saharan African GER average. The number of students per 100,000 population is also at the level that one would expect for a country at Rwanda's level of economic development (see figure 8-1 in chapter 8). EDUCATION IN RWANDA 37 TABLE 2-2: COVERAGE OF THE EDUCATION SYSTEM IN RWANDA AND SELECTED COUNTRIES, LATE 1990S­2000 Per capita Higher education Gross enrollment ratio (%) GNP, 2000 enrollment per Country (1995 US$) Primary Secondary Higher 100,000 population Rwanda, circa 2000­01a 241 107 13 1.9 200 Selected African countries, circa 1997 Mozambique 115 60 7 0.5 41 Ethiopia 146 43 12 0.8 Tanzania 193 67 6 0.6 43 Niger 208 29 7 -- Chad 216 58 10 0.6 54 Madagascar 246 97 16 2.0 194 Burkina Faso 275 39 0.9 90 Kenya 328 85 24 -- -- Uganda 355 74 12 1.9 154 Benin 411 78 18 3.1 208 Mauritania 495 79 16 3.8 374 Country group averagesb Low-income Sub-Saharan African countries 411 75 20 1.9 207 All low-income countries 456 84 33 5.2 575 -- Not available. Note: GNP, gross national product. a.The ratios for primary and secondary education are the averages of the estimates shown in table 2-1. For higher education, the ratio in the next-to-last column is the estimate based on enrollment counts and population projections, which best captures the magnitudes of the recent increases in enrollments. For the last column, the indicator is calculated using total enrollments in higher education for 2001­02 and the population of 8.7 million projected by the United Nations Population Division, as reported in the World Bank's Statistical Information Management and Analysis (SIMA) database. b. Data on gross enrollment ratios refer to the averages for 27, 23, and 17 low-income Sub-Saharan African countries and 42, 37, and 30 low-income countries at the primary, secondary, and higher levels, respectively. Data on higher education enrollments per 100,000 population are for 27 Sub-Saharan African countries and for 40 countries worldwide with per capita GNP below US$755 in 1996. Source: For data on enrollment ratios, UNESCO Website, ; for data on per capita GNP, World Development Indicators (various years). From Cross-Sectional Indicators of Coverage to Student Flow Patterns Because schooling usually requires a multiyear time commitment to yield tangible returns, it is important to go beyond the aggregate indicators of enrollments documented so far. The school- ing careers of children need to be examined, relying on such measures as the entry rate to a given cycle of schooling, the share of entrants who reach the end of the cycle, frequency of grade repe- tition, and transition rates between cycles. 38 WORLD BANK COUNTRY STUDY FIGURE 2-4: RELATION BETWEEN EDUCATIONAL ATTAINMENT AND PROBABILITY OF BEING LITERATE, RWANDA, 2001 1.0 Urban men 0.8 Rural women 0.6 0.4 Probability Rural men and urban women 0.2 0.0 1 2 3 4 5 6 Grade attained Note: Refers to the subsample of the population age 15 and above. Source: Authors' estimates based on the 2001 Questionnaire unifié sur les indicateurs de développement (QUID). At the primary level, where the curriculum is typically designed to impart basic literacy and numeracy skills, children who never enroll are unlikely to acquire these skills, and those who enroll but drop out prematurely are unlikely to become permanently literate and numerate. In Rwanda data from a 2001 household survey show a clear correlation between adults' educational attainment and literacy: the probability of being literate is nearly 100 percent for those who have attained at least five years of primary schooling but less than 50 percent for those attaining only three years (figure 2-4). For a given level of attainment, the probabilities for women are lower than for men. These results suggest that it is particularly important for girls to survive to the end of the primary cycle to achieve permanent literacy in adulthood.10 At postprimary levels of education, access and survival to the end of the cycle remain impor- tant, not least because both are prerequisites for entry to subsequent levels. The curricula likewise offer training in an integrated set of skills, so that premature exit would imply a failed investment, to the extent that the departing student would not have mastered what was being taught. The grade repetition indicator mentioned above is important for two reasons. First, grade repetition is likely to reduce the probability that a student persists to the end of the cycle, in part because it signals to parents that their child is not progressing academically--which would natu- rally predispose them to withdraw the child from school. Second, a repeater costs twice as much per grade attained as a student who has not repeated the grade, and yet there is no strong evi- dence that grade repetition necessarily improves learning outcomes. High rates of grade repeti- tion are therefore uneconomic, in that substantial resources are tied up by students for whom the extra year spent repeating the same grade may at best yield only modest gains in learning. Student Flow Patterns in Primary Schooling This section examines the overall pattern of entry rates to first grade, survival rates to subsequent grades in the primary cycle, and the prevalence of repetition at each grade. (Socioeconomic dis- parities in these parameters are presented in chapter 4.) A later section looks at the transition between the primary and secondary cycles. 10. The results from another survey, the 2000 Rwanda Multiple Indicator Cluster Survey (MICS), show a similar pattern of rising probability of being literate as educational attainment increases, although the gaps between men and women in urban and rural areas are wider than those shown in figure 2-4. The results for Rwanda are comparable to those based on data from the MICS in other African countries, including Burundi, Guinea-Bissau, Niger, Senegal, Sierra Leone, and Togo. EDUCATION IN RWANDA 39 TABLE 2-3: PERCENTAGE ENTRY RATE TO GRADE 1, RWANDA, 1991­92 AND 2000­01 Source of data 1991­92 2000­01 School statistics and 1991 population census 88 -- Household surveys DHS 85 88 QUID -- 87 -- Not available. Note: DHS, Demographic and Health Surveys; QUID, Questionnaire unifié sur les indicateurs de développe- ment. For 1991­92, the estimate in the first row is the number of new entrants reported in the school statistics divided by the population of 7-year-olds enumerated in the census. In the other cells, which are based on house- hold survey data, the estimates correspond to the share of children ages 10­13 who report ever enrolling in school; the age range was chosen to avoid underestimation associated with possible late entry into first grade. Source: Authors' estimates based on data from Ministry of Education school statistics, which are collected through an annual census of schools; the population census of 1991; the 1992 and 2000 DHS; and the 2001 QUID. Rate of Entry to Grade 1 Before presenting the results, a comment on the estimation methodology is in order. One method consists of dividing the number of new entrants to first grade by the cohort correspon- ding to the official age for entering this grade (age 7 in Rwanda).11 For both of the years consid- ered, 1991­92 and 2001­01, data on the numerator were available from school statistics. For the denominator, however, actual data by single years of age were available at the time of analysis for the population census of 1991 but not yet for the population census of 2002. The calculation for 2000­01 would have had to rely on population projections, but because of the large movements of people after 1994, such projections, particularly for single age groups, are probably not suffi- ciently reliable for our purpose. Accordingly, no estimate of the entry rate for 2000­01 based on this method is shown here. A second approach to estimating the entry rate can be used when survey data are available. It involves computing the share of children who report never enrolling in school and subtracting the result from 100 percent.12 The data needed to implement this approach are available for Rwanda for both 1991­92 and 2000­01. For the latter year, two sources are available, both of which are used here. Although survey data represent only a sample of the population, their advantage over school and census statistics is that both the numerator and the denominator refer to the same individuals, thus ensuring internal consistency in the data. The results appear in table 2-3. The estimates for both 1991­92 and 2000­01 are highly comparable across data sources, a feature that enhances our confidence in the estimates. The cur- rent entry rate, averaging 88 percent, is remarkably high for a country with Rwanda's per capita GDP. Still, it is important to note that the rate is, at best, only marginally better than the average of about 85 percent at the start of the 1990s. The results imply that gains in the entry rate to first grade can probably be ruled out as a major source of the large increase in the GER, from 73 11. Because some children may enter earlier or later than the official age, the calculation should ideally be adjusted by dividing the number of entrants by the population corresponding to the average age of the entering class. This adjustment is unlikely to alter the estimates substantially unless significant differences exist in the size of the single-age cohorts around age 7. 12. The calculation refers to children in the 10­13 age bracket. Setting the lower bound at 10 years allows for the possibility that some never-enrolled children may eventually enroll. The upper bound is intended to strike a good balance between two conflicting considerations: (a) ensuring reasonably sized cells for the calculations and (b) minimizing the influence of behavior patterns associated with older cohorts. 40 WORLD BANK COUNTRY STUDY percent in 1991­92 to more than 100 percent in 2000­01. Inducing the remaining children to enroll thus continues to be an important policy challenge in expanding primary school coverage in Rwanda.13 Survival Rates in the Primary Cycle Survival rates are, again, estimated for 1991­92 and 2000­01. Strictly speaking, survival rates should refer to the schooling careers of true cohorts as the members make their way through the school system. In the best of circumstances, such tracking is difficult because of the data require- ments and the complex calculations. In settings such as Rwanda, where the dynamic of student flow has been complicated by the reintegration of children who reenrolled in the years following the genocide, the data needs and the computational algorithm are even more demanding. Beyond these problems, perhaps the more serious drawback is that the results, relating as they do to the experience of children who entered school at least six years ago, may not be sufficiently up-to- date to inform policy development. An alternative approach is to construct pseudocohort survival profiles based on cross-sectional data for two adjacent school years. This method involves computing grade-to-grade transition rates that are then linked to obtain the pattern for the entire primary cycle. For example, if the transition rate between grades 1 and 2 is 85 percent, and that between grades 3 and 4 is 75 per- cent, the survival rate from grade 1 to grade 3 would be computed as 64 percent (= 0.85 × 0.75). By extension to subsequent grades, the process eventually yields the survival rate to the final grade in the cycle. The result is a pseudoprofile because it blends the experience of the different cohorts currently in school. The stability of the profile depends on the extent to which cohorts are similar in their pattern of grade-to-grade progression. Pseudocohort survival rates can be computed using school statistics as well as survey data, and we use both sources for the calculations on Rwanda. Using school statistics, the grade-to- grade transition rates are computed from data on nonrepeaters in adjacent grades in the two years.14 For example, if in year X there were 20,000 nonrepeaters in grade 1, and in year X + 1 there were 18,000 nonrepeaters in grade 2, the transition rate between grades 1 and 2 would be 90 percent (= 18,000/20,000).15 Using survey data, the calculation is feasible if for each child who attended school last year information exists on the following variables: (a) the highest grade attained by the child and (b) whether he or she is still enrolled in the current school year. These variables can be used to compute the grade-specific dropout rates. By subtracting from 100 per- cent, we obtain what would be, in the absence of differences in repetition across grades, the grade-to-grade survival rates. Where these differences are significant, as they are in Rwanda, an adjustment is required to avoid double counting; this is done by multiplying the rates obtained thus far by the ratio between the shares of nonrepeaters in the relevant pairs of grades.16 Once computed, these adjusted grade-specific rates can be linked together as before to obtain the pseudocohort survival rate to the end of the cycle. 13. Chapter 4 examines the characteristics of the unschooled population and presents evidence on the correlates of ever attending school. 14. To avoid double counting, only nonrepeaters are used in the calculation. 15. The method assumes regularity in schooling behavior; that is, children enter the system only in grade 1 and make the transition between school years in one of three categories: as repeaters, dropouts, or gradu- ates to the next grade. Although this assumption is probably valid for the calculations for 1991­92, it is unlikely to hold for 2000­01, given the large population movements following the genocide in 1994 and the reintegration of returning children whose schooling was interrupted by the war. 16. Suppose, for example, that the dropout rate between grades 1 and 2 is 5 percent and the share of nonrepeaters is 80 percent in grade 1 and 75 percent in grade 2. The survival rate between grades 1 and 2, adjusted for differences in grade repetition in the two grades, would then be 89 percent [= (100 - 5) × (75/80)]. EDUCATION IN RWANDA 41 TABLE 2-4: PERCENTAGE SURVIVAL RATES IN PRIMARY SCHOOLING, RWANDA, 1991­92 AND 2000­01 Primary grade 1991­92 2000­01a 1 100 100 2 85 100 3 74 96 4 64 91 5 53 77 6 44 73 7 35 n.a. 8 29 n.a. n.a. Not applicable because the primary cycle was shortened from eight to six years in 1992. a. Estimates have been adjusted for differences in grade repetition between contiguous grades. See the text for further explanation. Source: For 1991­92, authors' estimates based on data from Ministry of Education school statistics; for 2000­01, authors' estimates based on data from two household surveys: the 2001 Questionnaire unifié sur les indicateurs de développement (QUID) and the 2001 Rwanda Multiple Indicator Cluster Survey (MICS). Applying the foregoing methods, we obtain the estimates for 1991­92 and 2000­01 that appear in table 2-4.17 They show a significant improvement in cohort survival over the course of a turbulent decade: whereas in 1991­92, 44 pupils in a cohort of 100 reached grade 6, in the later year 73 did--an astonishing gain of 66 percent. The increase is indeed remarkable, given the context. If sustained, the improvement augurs well for the spread of literacy in Rwanda; as we saw above, adults with at least a fourth grade education have a more than 80 percent chance of being literate. Yet there is concern that the gain in survival rate may not be sustained, a possibility examined in greater detail below in light of the patterns of grade repetition. The information on entry rates and survival rates can be combined to obtain grade-specific enrollment rates for 1991­92 and 2000­01 (figure 2-5).18 These rates show the share of the school-age population enrolled in each grade, providing a finer-grained picture of coverage than is possible with the gross enrollment ratio, which is averaged over the entire cycle of six years. Given the current pattern of entry and survival rates, we can expect that about 37 percent of the present generation of school-age children would not achieve complete primary schooling: about a third of this group would never enter school, about a fifth would drop out between grades 1 and 4, and about 44 percent would drop out after grade 4. These results suggest that attracting chil- dren to school and keeping them past fourth grade remain priorities in the effort to produce a future generation of literate and numerate Rwandan citizens. 17. The calculations for 1991­92 rely on school statistics because survey data were unavailable for that year. Only survey data are used for 2000­01, for the reasons explained in the text. 18. To illustrate, the enrollment rate in grade 3 would be the entry rate multiplied by the survival rate from grade 1 to grade 3. 42 WORLD BANK COUNTRY STUDY FIGURE 2-5: GRADE-SPECIFIC ENROLLMENT RATES IN PRIMARY SCHOOLING, RWANDA, 1991­92 AND 2000­01 100 2000-01 a 80 (%) 60 rate 1991-92 40 Enrollment 20 0 1 2 3 4 5 6 7 8 Gradeb a. See the text discussion for definition and computation method. b.The primary cycle was eight years in 1991­92 and six years in 2000­01. Source: Computed from the data in tables 2-3 and 2-4. Grade Repetition Estimates for grade repetition for 1990­91 and 2000­01 appear in table 2-5. For 1990­91, the table shows only one set of estimates, which is based on Ministry of Education school statistics, the only available source for that year. For 2000­01, there are two sets of estimates based on two independent sources: the ministry's school statistics and the 2001 Multiple Indicator Cluster Survey.19 The comparability between the two sets of estimates increases our confidence in the reliability of the underlying data.20 The results reveal an astonishing threefold rise in repetition rates between 1990­91 and 2000­01. In both years, grade repetition occurs most frequently in grade 1, but its prevalence rises noticeably in the upper grades, reaching 38 percent in grade 5. By contrast, the pattern in 1990­91 shows a declining share of pupils repeating as they progress up the educational ladder through the last grade in the eight-year cycle. The underlying changes that led to the significant rise in grade repetition are unclear, but what is clear is that the increase is one reason why Rwanda's gross enrollment ratio rose so quick- ly in the past decade--from 73 percent in 1991­92 to 109 percent in 2000­01.21 There are, a priori, three possible sources for the increase in the gross enrollment ratio: an increase in the entry rate to first grade, an increase in the survival rate to the end of the cycle, and an increase in grade repetition. Between 1991­92 and 2000­01, the entry rate to first grade did not improve. Of the other two factors, grade repetition contributed an estimated 57 percent of the increase in 19. Other household surveys implemented around 2000, such as the 2000 QUID and the 2001 House- hold Living Conditions Survey (in French, Enquête intégrale sur les conditions de vie des ménages au Rwanda, EICV), do not include questions that would permit estimates of the repetition rate. 20. Even though the rates for 1990­91 cannot be verified against survey data, they are consistent with other indicators of coverage, including the gross enrollment ratio, entry rates to first grade, and survival rates to the end of the cycle, and are therefore fairly reliable for the purpose here. 21. Note that these ratios, computed by dividing enrollments by the relevant population, are reasonably consistent with those implied by the student flow indicators reported so far. The implied gross enrollment ratio (IGER) is given by the relationship IGER = AGSER × (1 + ARR), where AGSER is the average grade- specific enrollment rate across grades in the cycle and ARR is the average repetition rate. The IGER is esti- mated to be 69 percent for 1991­92 and 104 percent for 2000­01. EDUCATION IN RWANDA 43 TABLE 2-5: PERCENTAGE REPETITION RATES IN PRIMARY SCHOOLING, RWANDA, 1990­91 AND 2000­01 2000­01 rate as 2000­01 multiple of 1990­91 Primary grade 1990­91 School statistics Survey data (MICS) ratea 1 16.2 41.5 40.4 2.6 2 13.2 28.8 29.2 2.2 3 11.4 30.3 27.9 2.7 4 9.0 34.4 22.6 3.8 5 8.2 37.6 31.1 4.6 6 8.0 29.2 25.8 3.7 7 6.8 n.a. n.a. n.a. 8 9.7 n.a. n.a. n.a Cycle averageb Grades 1­6 11.0 33.6 29.5 3.1 Grades 1­8 10.3 n.a. n.a. n.a. n.a. Not applicable because the primary cycle was shortened from eight to six years in 1992. Note: MICS, Multiple Indicator Cluster Survey. a.To maintain consistency in the data source, both the numerator and denominator refer to rates based on Ministry of Education school statistics. b. Figures are unweighted averages. Source: For 1990­91, authors' estimates based on Ministry of Education school statistics for school years 1989­90 and 1990­91; for 2000­01, authors' estimates based on Ministry of Education school statistics for 1999­2000 and 2000­01 and on the 2001 Rwanda MICS. the gross enrollment ratio during the period, while the gain in the survival rate to the end of the cycle--from 44 percent in 1991­92 to 73 percent in 2000­01--contributed the remaining 43 percent.22 Thus, although educational coverage undeniably expanded rapidly over the course of the 1990s, much of the growth reflects increased recycling within the system. The high rates of grade repetition raise concerns about the sustainability of the gains in cohort survival rates. Regression analysis of cross-country data suggests that grade repetition cor- relates negatively with cohort survival rates.23 For a country at Rwanda's level of per capita gross national product (GNP), a repetition rate of 33.6 percent would be consistent with a cohort 22. The calculation is as follows. During the period, the entry rate to grade 1 stagnated at around 88 percent, while the survival rate rose from 44 to 73 percent and the repetition rate rose from 11 to 34 per- cent. Suppose the repetition rate in 2000­01 had remained unchanged at 11 percent. The implied gross enrollment ratio would have been 84 percent [= (88 + 88 × 0.73)/2) × (1 + 0.11)]. Thus, of the 35 percent- age point increase (= 104 - 69) in the implied gross enrollment ratio between 1991­92 and 2000­01, 15 percentage points (= 84 - 69), or 43 percent of the total, can be attributed to the increase in the cohort survival rate and 20 percentage points (= 104 - 84), or 57 percent of the total, can be attributed to the increase in grade repetition. 23. On the basis of data for 57 low-income countries, and controlling for the entry rate to grade 1 (EG1), for per capita GNP in constant 1993 dollars, and for public spending per pupil as a multiple of per capita GNP (PSPP), the relation between the cohort survival rate (CSR) and the repetition rate (RR) is estimated as follows: CSR = 38.3 + 0.05 × EG1 - 0.88 × RR + 5.53 × ln(per capita GNP) + 41.8 × PSPP. The coefficient estimates for RR and ln(per capita GNP) are statistically significant at the 1 and 5 percent confidence levels, respectively. 44 WORLD BANK COUNTRY STUDY survival rate of around 47, controlling for the entry rate to grade 1 and the level of public spend- ing per primary school pupil. These results imply that although Rwanda's cohort survival rate (73 percent) exceeds the regression-predicted level, the equilibrium rate may well be lower if the current high rates of grade repetition persist. A decline would hardly be surprising, as parents' willingness to send a child to school and keep the child there may wane in the face of signals, con- veyed by frequent grade repetition, that the child is in fact failing to keep up with the schoolwork. Efficiency of Student Flow What do these student flow patterns imply about the system's internal efficiency? One way to answer this question is to compare the resources that the system actually consumes to produce its annual output of primary school graduates with the resources that it would have spent in the absence of grade repetition and dropout. The calculations and results appear in table 2-6. Consider, for example, the data for 1991­92. Given the patterns of cohort survival and grade repetition in that year, and calibrating the calculations to a starting cohort of 1,000 pupils, the resources spent in grade 1 would amount to 1,193 pupil-years [= 1,000/(1 - 0.162)]. In grade 2, only 849 of the initial 1,000 pupils remain, implying that the resources spent on them would be 978 pupil-years [= 849/(1 - 0.132)]. Continuing in this way for all six grades and adding up the TABLE 2-6: SUMMARY INDICES OF STUDENT FLOW EFFICIENCY IN PRIMARY SCHOOLING, RWANDA, 1990­91 AND 2000­01 1990­91 2000­01 Number Repetition Pupil-years Number Repetition Pupil-years Primary grade remaininga rate (%) investedb remaininga rate (%) investedb 1 1,000 16.2 1,193 1,000 41.5 1,709 2 849 13.2 978 1,000 28.8 1,404 3 743 11.4 839 964 30.3 1,384 4 640 9.0 704 910 34.4 1,387 5 534 8.2 582 775 37.6 1,242 6 439 8.0 477 728 29.2 1,028 Cumulative pupil-years 4,205 4,772 5,376 8,154 (2,634)c (4,366)c Index of student flow efficiencyd Dropout relatede 0.63 0.81 Repetition relatede 0.88 0.66 Overallf 0.55 0.54 Blanks denote not applicable. a. From initial cohort of 1,000. b. Refers to the pupil-years needed to educate new entrants as well as repeaters in the corresponding grade. See the text for an explanation of this computation. c. Refers to the pupil-years invested only in the pupils who reach grade 6 without repeating a grade. d.The index is defined as the ratio between the cumulative pupil-years invested in a system with no student flow problems and the actual cumulative pupil-years invested given the current pattern of dropout and grade repetition.An education system with no dropout and no grade repetition would have an index of 1.0. See the text for further explanation of this computation. e.The numerator of the dropout-related index includes only the pupil-years associated with dropping out.The numerator of the repetition-related index includes only the pupil-years associated with grade repetition. f.Takes account of the pupil-years associated with both dropout and grade repetition Source: Authors' estimates based on the cohort survival and repetition rates reported in tables 2-4 and 2-5. EDUCATION IN RWANDA 45 pupil-years spent in each grade, a cumulative total of 4,772 pupil-years (= 1,193 + 978 + 839 + 704 + 582 + 477) for the whole cycle is derived. This global amount can be compared with the total of only 2,634 pupil-years (= 439 × 6) needed to produce the 439 first-graders in the cohort who reach the last grade without repeating a grade. Thus, the system operates at only 0.55 times (= 2,634/4,772) the efficiency (1.0) of a system in which no one drops out or repeats a grade. When the wastage associated with dropout alone is considered, the implication is that a total of 4,205 pupil-years (= 1,000 + 849 + 743 + 640 + 534 + 439) was spent to produce the 439 graduates. Thus, the dropout-related efficiency index is 0.63 (= 2,634/4,205). Alternatively, counting only the wastage associated with grade repetition implies a total investment of 4,772 pupil-years, when only 4,205 would have sufficed; the repetition-related efficiency index is 0.88 (= 4,205/4,772).24 Comparison of the results for 1990­91 and 2000­01 suggests that the overall efficiency of student flow has not changed much over the decade. Consistent with the rise in the cohort sur- vival rate, the index that takes into account only the impact of dropout was better in 2000­01 than in 1990­91 In contrast, the index associated with the impact of grade repetition alone has deteriorated, reflecting the rise in grade repetition over the decade. If grade repetition persisted at the current level and the cohort survival rate fell to, say, the regression-predicted level of 47 per- cent, the system's student flow efficiency index would drop to 0.42, compared with the current value of 0.54. At 0.42, Rwanda's index would then be among the lowest among low-income countries in Africa. Arresting this potential decline in student flow warrants a high-priority place on the policy development agenda. Cross-Country Comparison Figure 2-6 shows data on entry rates to grade 1 and survival rates to grade 5 for Rwanda, as well as repetition rates for 10 other African countries for which data for around 2000 are available. (Because the primary cycle in some countries is only five years, the figure shows survival rates for five years of schooling rather than the six years in the Rwandan system.) Rwanda ranks among the countries with high entry rates to first grade and high survival to grade 5, but these positive fea- tures are marred by the fact that it has the highest rate of grade repetition. In systems where grade repetition is widespread--as it is in Mozambique, Madagascar, Cameroon, and Benin in the sample--the cohort survival rate has typically been lower, falling, for example, to only 40 percent in Madagascar. To the extent that patterns of student flow in these systems reflect steady-state conditions, the risk of a decline in Rwanda's currently high cohort survival rate cannot be ruled out or ignored. To complete the discussion, we compute the summary index of student flow for the 10 other African countries, using the same method as for Rwanda in the previous section.25 The results, plotted against the countries' per capita GDP in 1999, are shown in figure 2-7. Rwanda is in the middle of the pack. Although it does better than, say, Madagascar and Mozambique, significant room for improvement remains. 24. In passing, we note that the data in table 2-6 can be checked for their consistency with the gross enrollment ratios (GERs) reported in table 2-1, which are estimated using the classical approach of dividing the total number of pupils by the population ages 7­12. These ratios are 73 percent in 1991­92 and 107 percent in 2000­01. Consider first the data for 1990­91 in table 2-6. If all children entered grade 1, the implied GER would be 89.5 percent (= 4,772/6,000); but with a grade 1 entry rate of between 85 and 88 percent (see table 2-3), the ratio is in fact around 69 percent. This calculation yields a result that is compara- ble to the actual ratio of 73 percent for 1991­92. For 2000­01, the data in tables 2-3 and 2-6 imply a GER of 120 percent [= (8,154/6,000) × 0.88], which is much higher than the actual ratio of 107 percent. To evaluate the source of the discrepancy, consider another estimate based only on the survival rate to the ter- minal grade (0.728, from table 2-6), the overall repetition rate (34 percent, from table 2-5); and the entry rate to grade 1 (88 percent, from table 2-3). These three pieces of information imply a GER of 102 percent {= [(100 + 72.8)/2] × 0.88 × 1.34}. The result is much closer to the actual GER of 107 percent, which suggests that the grade-to-grade survival rates in table 2-6 are probably overestimated. 25. To ensure comparability across countries, the indices are for five years of primary schooling. 46 WORLD BANK COUNTRY STUDY FIGURE 2-6: PRIMARY SCHOOL STUDENT FLOW INDICATORS IN RWANDA AND IN SELECTED SUB-SAHARAN AFRICAN COUNTRIES, CIRCA 2000 Average, 21a Mozambique, 30 Mauritania, 17 Tanzania, 4 88 Rwanda, 34 77 Madagascar, 31 Cameroon, 27 Benin, 26 Togo, 24 Guinea, 27 Burkina Faso, 15 Niger, 13 100 80 60 40 20 0 0 20 40 60 80 100 Grade 1 entry rate (%) % of first-graders reaching grade 5 a. The number following each country name is the primary school repetition rate, in percent.The average is for the countries shown in the figure, excluding Rwanda. Source: For Rwanda, tables 2-3 and 2-4; for data on survival rates for the other countries, UNESCO (2001); for entry and repetition rates for the other countries,World Bank education sector country status reports. FIGURE 2-7: EFFICIENCY OF STUDENT FLOW IN RWANDA AND IN SELECTED SUB-SAHARAN AFRICAN COUNTRIES, CIRCA 2000 1.0 a Tanzania efficiency 0.8 flow Niger Burkina Faso Mauritania Guinea 0.6 student Benin Rwanda Togo of Cameroon Mozambique Index 0.4 Madagascar 200 300 400 500 600 Per capita GDP in 1999 (log scale) Note: GDP, gross domestic product. a. The index has a value of 1.0 in a system with no dropout or grade repetition. See the text for details on its computation. Source: Authors' calculations based on data in figure 2-6. EDUCATION IN RWANDA 47 TABLE 2-7: SECONDARY SCHOOL TRANSITION AND SURVIVAL RATES, RWANDA, 1991­92 AND 2000­01 Indicator 1991­92 2000­01 Transition rates between cycles of schooling (percent)a From primary to tronc commun cycle -- 60 From tronc commun to upper secondary cycle 100 Percentage of entrants reaching end of each subcycleb Tronc commun 69 Upper secondary 79 Secondary 1­6 60 57 Blanks denote not applicable. -- Not available. a. Refers to the transition rate between the last grade of one cycle and the first grade of the next cycle.The rates were not computed for 1991­92 because of data constraints associated with structural changes in the system. b. In 1991­92 the secondary cycle lasted six years, with no selection during the cycle.The system was restruc- tured in the postgenocide era into two subcycles, each lasting three years. Students follow a common first cycle (tronc commun) before proceeding to the specialized streams in the upper secondary cycle. Source: Based on Ministry of Education school statistics for school years 1991­92, 1999­2000, and 2000­01. Student Flow Patterns in Secondary Schooling We turn now to an examination of children's schooling careers in secondary education, including the pattern of transition between the primary and secondary cycles and survival and repetition rates within secondary schooling. Transition Rates and Cohort Survival Rates As in many education systems, formal selection for secondary schooling in Rwanda is based on a student's performance in the national examination at the end of the primary cycle. At present, an estimated 60 percent of primary school pupils who reach grade 6 eventually make the transition to secondary schooling (whether public or private), even though some of them do so only after repeated attempts (table 2-7).26 This rate of transition exceeds the average of 50 percent for Fran- cophone Africa in the 1990s and is comparable to the average of 60 percent in Anglophone Africa; it is, however, inferior to the rates in low-income countries in Asia and Latin America, which average 78 and 71 percent, respectively (Mingat and Suchaut 2000). At the secondary level, selection takes place at the end of the tronc commun cycle. As table 2-7 suggests, nearly all the students who currently reach the end of that cycle continue to the upper secondary cycle. For comparison, in Burkina Faso and Madagascar--countries that are similar to Rwanda in their levels of per capita GDP and their rates of transition between the primary and secondary cycles--the shares of lower secondary school completers who proceed to the upper secondary cycle are only 60 and 46 percent, respectively. Table 2-7 also documents survival rates in secondary schooling. In the tronc commun cycle 69 percent of the entrants reach the end of the cycle, and in the upper secondary cycle the share is 79 percent. Of those who start secondary school, we can thus expect 57 percent to complete 26. Although data from the 2000 Multiple Indicator Cluster Survey (MICS) can also be used to estimate the transition rate, small cell sizes make the results unreliable. We therefore report here only the results based on the Ministry of Education school statistics. 48 WORLD BANK COUNTRY STUDY TABLE 2-8: PERCENTAGE REPETITION RATES IN SECONDARY SCHOOLING, RWANDA, 1990­91 AND 2000­01 2000­01 Secondary grade 1990­91 Overall Public schoolsa Private schools Repeaters as percentage of students enrolled 1 2.8 17.4 15.9 19.0 2 10.4 13.8 15.0 12.1 3 10.2 8.5 8.3 8.8 4 8.5 11.5 10.6 13.0 5 9.5 11.0 12.5 9.2 6 4.6 4.7 3.6 6.2 All grades, 1­6 7.5 12.4 12.2 12.8 Unweighted averages Tronc commun cycle 13.2 13.1 13.3 Upper secondary cycle 9.1 8.9 9.4 All grades, 1­6 7.7 11.1 11.0 11.4 Blanks denote not applicable. a. Includes state and libre subsidié schools. Source: Ministry of Education school statistics for school years 1990­91 and 2000­01. the entire six years of the secondary cycle. The rate today is largely unchanged from that in the early 1990s.27 Repetition Rates The average repetition rate for all six grades of secondary schooling rose from 7.7 to 11.1 percent in the course of the 1990s (table 2-8). The pattern of grade repetition has also changed some- what: although almost no students repeated the first year of secondary education in 1990­91, the repetition rate is now 17.4 percent in this grade, the highest rate in the entire secondary cycle. In 2000­01 grade repetition was as high in the public as in the private sector, and the pattern across grades was comparable between the two sectors. As is consistent with the lack of rigorous selec- tion between the tronc commun and upper secondary cycles (and probably for higher education as well), grade repetition in the last grade of each cycle was not dramatically higher than in the other grades. The pattern contrasts with that in Madagascar, for example, where the repetition rate in the final grade of lower secondary school was three times that for the other grades in the cycle. Efficiency of Student Flow Table 2-9 presents indices of student flow efficiency, computed using the same method that was applied to the data for primary schooling. As before, the more efficient is the flow of students through the system, the closer is the index to unity. The results show that the flow was somewhat more efficient in 2000­01 than in 1990­91, reflecting the gains in survival rates to the end of the cycle. 27. In education systems with stable and slowly changing student flow processes, the student flow pro- files for the primary and secondary cycles can be combined to yield a complete profile covering all 12 years in the two cycles. In Rwanda these conditions do not hold. The survival rates presented here are broadly consistent with historical rates of transition between the primary and secondary cycles and repetition rates in secondary schooling. EDUCATION IN RWANDA49 TABLE 2-9: SUMMARY INDEX OF STUDENT FLOW EFFICIENCY IN SECONDARY SCHOOLING, RWANDA, 1990­91 AND 2000­01 Cycle 1990­91a 2000­01 Tronc commun 0.71 Upper secondary 0.80 All secondary, grades 1­6 0.74 0.77 Blanks denote not applicable. a. In 1990­91 the secondary cycle lasted six years, with no selection during the cycle, so no separate index was computed for the tronc commun and upper secondary cycles.An education system with no dropout and no grade repeti- tion would have an index of 1.0. See table 2-6 and related text for details on how the index is calculated. Source: Based on Ministry of Education school statistics for school years 1991­91, 1999­2000, and 2000­01 and the 2001 Questionnaire unifié sur les indicateurs de développement (QUID). Policy Perspectives on Management of Student Flow In the context of the government's poverty reduction strategy, the student flow patterns docu- mented above have at least two important implications for education sector policies. 1. Education policy should ensure that all children attain at least a complete primary educa- tion so that all are equipped with the basic skills of literary and numeracy needed to make a decent living in adulthood. 2. Beyond primary schooling, the education system's production of skilled workers to sup- port economic growth also warrants attention. In the years following the 1994 genocide, the priority was to replenish the depleted stock of human capital as rapidly as possible. As the stock recovers to its pregenocide levels, expansion will increasingly depend on the labor market's capacity to absorb highly educated workers in productive employment (see chapter 8). Although a full treatment of these issues, particularly those pertaining to the labor market, is beyond the scope of this chapter, the student flow data presented in the foregoing sections pro- vide a basis for distilling some implications for managing the quantitative expansion of the educa- tion system. Primary Education Entry rates in primary education have historically been high, at nearly 90 percent. One issue is therefore to identify the last 10 percent or so of the population that is still excluded and to target interventions accordingly. (See chapter 4 for a more detailed discussion of patterns of disparities in schooling.) Survival rates to the end of the cycle warrant even closer attention. As argued earli- er in the chapter, the current rate of 73 percent is respectable compared with rates in other low- income countries but is probably unsustainable in light of the extremely high rates of grade repe- tition. An immediate concern, accordingly, is to lower the repetition rate. A medium-term target of, say, 10 percent at most is feasible in the light of other countries' experience but would proba- bly require measures to rationalize class promotion policies and improve classroom pedagogical practices. 50 WORLD BANK COUNTRY STUDY TABLE 2-10: DISTRIBUTION OF PRIMARY SCHOOLS AND NEW FIRST-GRADERS BY HIGHEST GRADE OF INSTRUCTION OFFERED BY THE SCHOOL, RWANDA, 2000­01 | % of schools % of new first-graders Highest grade of Libre Libre Cumulative instruction offered State subsidié Overalla State subsidié Overalla share 1 0.4 0.8 0.7 0.1 0.4 0.3 0.3 2 2.0 1.4 1.6 0.7 0.7 0.7 1.0 3 2.9 2.3 2.6 1.5 1.2 1.4 2.4 4 3.1 2.6 2.8 2.8 1.5 1.8 4.2 5 8.3 5.0 6.0 5.4 3.3 4.1 8.3 6 83.3 87.9 86.3 89.5 92.9 91.7 100.0 All schoolsb 100.0 100.0 100.0 100.0 100.0 100.0 (551) (1,446) (2,028) (89,134) (229,353) (320,952) Blank denotes not applicable. a. Includes data for private schools. b. Figures in parentheses refer to the number of schools or pupils on which the percentages were computed. Source: Ministry of Education census of primary schools, 1999­2000. One approach is to define two or three subcycles within primary schooling and to apply auto- matic promotion within each subcycle and performance-based promotion between subcycles. An argument for this arrangement is that skills covered by the curriculum overlap between contigu- ous grades, particularly in the beginning years, when pupils are simply getting used to school and are learning the rudiments of reading, writing, and arithmetic. Automatic promotion would avoid unnecessary compartmentalization of the learning process and create space for children to acquire foundational skills at their own pace. A complementary--and arguably indispensable--tool for managing grade repetition is to make sure that teachers are equipped with techniques for forma- tive evaluation and receive relevant professional support from head teachers, inspectors, and oth- ers, so that pupils' learning gaps can be systematically identified and remedied before the children come up for evaluation for promotion to the next subcycle.28 Grade repetition aside, supply-side constraints could also impede progress toward achieving universal survival to the end of the primary cycle. One such constraint is the possibility that some schools may lack the staff or the facilities to offer the complete cycle of schooling. In such set- tings, students leave school prematurely not because they want to, but because they have nowhere else to go in the system. Because supply-side constraints are by definition amenable to policy intervention, it is useful to assess the scale of the problem and the impact that removing these constraints might have on survival rates. Table 2-10 contains relevant information on this point. It shows the distribution of schools and of new first-graders according to the highest grade of instruction in which pupils are current- ly enrolled. Schools with no pupils enrolled in the sixth grade make up about 13 percent of all schools and account for about 8 percent of all first-graders in the system. These schools fall into two categories: either they are new and are still in the process of completing the grades offered, 28. As part of the agreement under the HIPC Initiative, countries such as Benin, Burkina Faso, and Mozambique have committed to lowering their primary school repetition rates. Automatic promotion and support for teachers in managing grade repetition are among the interventions that these countries are put- ting in place to achieve this objective. EDUCATION IN RWANDA 51 or they are constrained by lack of facilities or teachers from offering the complete cycle. The available data do not allow us to distinguish between the two types, but for the sake of illustra- tion, suppose that all of the schools are incomplete because of supply-side constraints and that transfers from incomplete to complete schools occur infrequently. Under these assumptions, if all schools offered the full primary cycle of instruction, the cohort survival rate would rise from its current level of 73 percent to 80 percent (= 73/(1 - 0.083). Under more realistic assumptions, the impact would be smaller. In Rwanda, therefore, the incompleteness of the course of instruc- tion offered can probably be ruled out as a major reason why primary school pupils quit school before finishing the cycle. Beyond Primary Schooling The immediate problem in both secondary and higher education is to manage the mounting pressure for expansion as ever larger cohorts of students in the preceding cycles complete their schooling. One helpful way to think about the problem is to view the tronc commun cycle as a continuation of primary education and the upper secondary cycle as preparation for higher educa- tion. This view receives support from the argument that primary school leavers are still somewhat immature for gainful employment and that continuation in school for a few more years would help round out their preparation for adult life. It explains why governments in some low-income countries have set their sights on achieving universal basic education (defined as the primary through the lower secondary cycle) as a long-term social objective. Beyond the tronc commun cycle, economic considerations become more important in devel- opment of policies affecting student flow. Although it is true that a modernizing economy requires highly trained workers to function well, experience shows that producing such workers faster than the labor market can employ them in relevant jobs typically does not accelerate eco- nomic growth but instead creates educated unemployment or underemployment and, as a conse- quence, social frustration and unrest among the affected individuals. What do these concepts imply for the management of student flow beyond primary educa- tion in Rwanda? Without entering into a full-blown evaluation here that takes into account fiscal implications as well, we consider some simple projections of student numbers to help clarify one aspect of the problem. Table 2-11 shows the current distribution of enrollment by grade in the primary and secondary cycles. The structure of enrollment is very steep: the numbers of students in the lower grades is substantially greater than those in the upper grades. The pressure for TABLE 2-11: GRADE-SPECIFIC ENROLLMENTS IN PRIMARY AND SECONDARY EDUCATION, RWANDA, 2000­01 Primary cycle Secondary cyclea Year in cycle Total Nonrepeaters Total Nonrepeaters 1 494,614 270,266 38,643 31,929 2 324,804 234,273 30,683 26,462 3 238,897 167,044 20,927 19,152 4 189,865 123,549 19,608 17,349 5 145,478 88,938 17,634 15,695 6 81,914 58,923 13,668 13,019 Total 1,475,572 942,993 141,163 123,606 a. Consists of two subcycles, tronc commun and upper secondary school, each lasting three years. Source: Ministry of Education school statistics for 2000­01. 52 WORLD BANK COUNTRY STUDY expansion is thus likely to mount rapidly in the coming years.29 In the primary cycle, the cohort survival rate is currently estimated at 73 percent. If this rate were maintained, the number of new students entering grade 6 in 2006 would be around 197,000 pupils, or more than six times the current cohort of entrants to the first year of secondary schooling. Even if the survival rate were to fall to 50 percent (as is likely if repetition rates remain at their current high levels), the project- ed number of pupils would still be large--about 135,000 new pupils, or more than four times the number of entrants to the first grade of the tronc commun cycle. How many sixth-graders in 2006 will be able to continue to the tronc commun cycle depends on how quickly it would be feasible (both physically and fiscally) to expand places in the second- ary cycle. Between 1995 and 2000, the entire system, including both the public and private sec- tors, expanded at the rate of 18,200 places a year, or 3,000 places annually per grade in the cycle. If the system continued to expand at this rate and repetition rates remained unchanged, 48,000 new entrants to the first year of the tronc commun cycle could be accommodated in 2007.30 In that year the transition rate between the primary and tronc commun cycles would thus be about 24 percent if the survival rate in the primary cycle were 73 percent, and 35 percent if the survival rate were 50 percent. Even if it were feasible to double the pace of expansion of secondary school places to 6,000 places per grade annually, the transition rate between the primary and tronc com- mun cycles would still be lower than at present: 31 percent if the survival rate in the primary cycle were 73 percent, and 46 percent if it were only 50 percent. Although admittedly rough, these projections suggest that in the foreseeable future the share of primary school completers continuing on to secondary school will very likely decline even as the absolute size of enrollments in secondary schools grows substantially.31 The selection process would accordingly need to become even more efficient than it is now in identifying students who have good potential to benefit from continuing past the primary cycle. The data in table 2-11 also have implications for the development of upper secondary educa- tion. In 2000­01 the system admitted nearly 32,000 new entrants to the first grade in the sec- ondary cycle. If the survival rate in the tronc commun cycle remains unchanged at 69 percent, the projected number of candidates for upper secondary schooling will exceed 22,000 in 2003­04, or about 1.27 times the size of the cohort that entered the upper secondary cycle in 2000­01. Currently, almost all students who finish the tronc commun cycle eventually go on to upper sec- ondary schooling, and 79 percent reach the end of the cycle. Continuation of these patterns implies that by 2007, a total of 18,000 upper secondary graduates a year (about 40 percent more than the current cohort size of about 13,000) will be seeking either employment or entry to higher education. We can expect these projections to increase in future years as the number of places in lower secondary school is expanded to accommodate the rising tide of primary school completers. For example, suppose that the number of new entrants to lower secondary school grows to 48,000 students a year by the end of the decade, on the assumption that about 3,000 places are added annually. If the student flow patterns within secondary schooling remained as at present, with limited selection between the tronc commun and upper secondary cycles, by 2015 there would be 29. Note that the current size of the entering cohort in first grade is of the same order of magnitude as the projected population of 6-year-olds in the coming decades, implying that the scale of the problem is likely to persist in the future. 30. This assumes that repetition rates in the first year of the cycle remain at the present 17.4 percent. The number of places for new entrants to the first year of the cycle would thus be the sum of the current number of new places (32,929) and the number of new places created at the assumed rate [= 3,000 × (1.0 - 0.174) × 6], or a total of 47,797 places. 31. This decline obviously does not invalidate the long-term goal of achieving universal basic education through lower secondary schooling, but physical constraints, quite apart from the fiscal ones, make that goal infeasible in the immediate future. EDUCATION IN RWANDA 53 at least 27,000 upper secondary school leavers each year seeking either a job or a place in higher education. At issue, then, is whether the labor market will be able to absorb this large an annual output of workers with at least an upper secondary education. Although it is perilous to project labor market developments in Rwanda's evolving postgenocide context, the scale of the increase in graduate output warrants a closer look at the options for managing student flow. Strengthening mechanisms to tighten selection into upper secondary schooling, particularly in the publicly fund- ed sector, would seem to be a logical place to start. Conclusion This chapter has documented the coverage of the education system and its evolution over time. The absolute number of students at all levels of education has recovered from the severe disrup- tion caused by the genocide and now exceeds pregenocide levels. The private sector share of enrollments remained unchanged during the 1990s: less than 1 percent in primary education, more than 40 percent at the secondary level, and nearly 40 percent in higher education. At the primary level, the gross enrollment ratio is now significantly greater than at the start of the 1990s and is comparable to the average for other low-income Sub-Saharan African countries. The ratio for higher education has also tripled since the early 1990s, although it remains smaller than the average for comparable countries. In secondary education, where the trend is difficult to establish because of structural changes in the system during the 1990s, the ratio in 2000­01 was some- what lower than the average for low-income countries in Sub-Saharan Africa. Nearly 90 percent of Rwandan children currently start school--about the same share as in the early 1990s--and 73 percent eventually make it to grade 6, up from about 44 percent. There is obviously still scope for improvement in both access and survival. But even maintaining the cohort survival rate at 73 percent poses a significant challenge, given the extremely high rates of grade repetition that characterize the system. At nearly 34 percent, Rwanda's repetition rate in primary education is one of the highest in the world, and cross-country experience suggests that its persistence is likely to depress the survival rate, perhaps to as low as 50 percent. Reduction of grade repetition therefore appears to warrant priority attention as part of the country's strategy for achieving universal primary education. As primary schooling improves and expands, the pressure to increase the supply of places at postprimary levels will inevitably mount. Simple projections based on conservative assumptions suggest that the number of potential candidates for entry into lower secondary school at mid- decade (that is, around 2005) could easily be four times the size of the current cohort of entrants to that cycle, and the number of potential candidates for entry into upper secondary school could be at least 30 percent larger than now. By the middle of the next decade (around 2015), given current patterns of student flow in the secondary cycle, the number of upper secondary school graduates seeking either employment or entry to higher education could easily be twice the cur- rent output from the system. In light of these projections, the task of managing student flow throughout the secondary cycle presents two challenges in the next 5 to 15 years. The first is to maintain the current rela- tively high levels of within-cycle survival rates in both the tronc commun and upper secondary cycles. The second is to strengthen the existing selection mechanisms between the primary and tronc commun cycles and between the tronc commun and upper secondary cycles. To the extent that the pace of expansion of places in the tronc commun cycle is impeded by physical and resource constraints, a heavier burden would be placed on the selection process to ensure that high-potential candidates are not lost to the system. For the upper secondary cycle, the rapidly increasing pool of lower secondary school leavers suggests that tighter selection at the entry point might be needed to manage downstream pressure to expand higher education. This task will only become more urgent if the labor market develops signs of a weakening capacity to absorb highly educated workers into productive employment. CHAPTER 3 EDUCATION FINANCE A ggregate public spending on education in Rwanda has been rising steadily since 1996. The prospects for continued expansion are good, given the new climate of support for human development, both internationally and within the country. But even more impor- tant than the aggregate amount of resources is the way they are used. In this chapter we examine the volume of national spending on education, including spending by households; the composi- tion of public spending on education by level; and international perspectives on the pattern of spending in Rwanda. The results provide a factual basis for discussing future directions for man- aging resource allocation to ensure that sectoral priorities are adequately funded. The most important of these priorities is to enable all children to complete a full course of primary school- ing under conditions that promote student learning. National Spending on Education In Rwanda, as in most other countries, the government is the main source of funding for educa- tion. Contributions by households are also significant, however, particularly in secondary educa- tion, where privately financed schools currently cater to about 43 percent of students. Spending by the Government Total spending on education reached a high of 5.5 percent of GDP in 2001, but much of the increase in the postgenocide years reflects rapid increases in capital spending (table 3-1). Nonetheless, aggregate current spending in relation to GDP was 65 percent higher in 2001 than in 1996. Across levels of education, the increased spending has mostly benefited higher educa- tion, whose share of expenditures rose from just under 15 percent in 1996 to more than 37 per- cent by 2001. Secondary education also gained, but its allocation of current spending rose only modestly, from about 15 percent in 1996 to nearly 18 percent in 2001. The combined result of these trends in higher and secondary education has been a dramatic fall in the share of primary education, from 70 percent in 1996 to just over 45 percent in 2001. Despite the large increase 55 56 WORLD BANK COUNTRY STUDY TABLE 3-1: LEVEL AND DISTRIBUTION OF PUBLIC SPENDING ON EDUCATION, RWANDA, 1982­2001 Total spending Current spending on education Share of current spending Capital on primary as % of GDP by level of education (%)a spending as education as Year Total Current % of total Primaryb Secondary Higher % of GDP 1982 3.0 62.7 24.0 13.2 1.9 1983 3.2 12.9 1984 3.3 3.2 2.8 12.6 1985 3.1 2.9 5.7 59.5 27.2 13.3 1.7 1986 3.5 3.2 6.2 60.0 27.6 12.4 1.9 1987 3.5 3.2 7.6 61.1 26.8 12.1 2.0 1988 3.1 59.0 25.5 15.4 1.8 1989 3.0 1990­95 1996 3.2 2.0 37.4 70.1 15.2 14.7 1.4 1997 3.4 2.0 40.9 64.6 16.0 19.4 1.3 1998 3.1 2.2 28.9 49.3 15.7 35.0 1.1 1999 4.3 3.4 21.9 47.7 18.5 33.8 1.6 2000 4.0 3.2 19.1 45.2 17.4 37.4 1.5 2001 5.5 3.3 39.8 45.2 17.6 37.3 1.5 Blanks denote not available. a. Some categories of spending are shared across levels of education (e.g., those associated with the National Curriculum Development Centre, the National Examination Council of Rwanda, the Education Ministry's staff and running costs, and the inspectorate). For our purposes here, such spending is apportioned according to the subsector's share of the spending that could be allocated across levels. b. Published data for the 1980s show combined spending on primary education (which lasted eight years during this period) and on the three-year vocational institutions for primary school leavers that were discon- tinued after 1991­92 (see note 2 in chapter 2).To obtain a time series for spending on primary education corresponding to the current six-year cycle, we prorated the published data according to the share of enroll- ments in the last two years of the eight year cycle and enrollments in the vocational institutions during the 1980s. See appendix table A2-1 for data on enrollments. Source: See appendix table A3-1. since 1996 in aggregate current spending on education relative to GDP, current spending on primary education registered a tepid increase, reaching only 1.5 percent of GDP by 2001. The concentration of spending on higher education after 1996 is consistent with the govern- ment's efforts to rebuild the country's depleted stock of highly qualified citizens following the 1994 genocide. Yet if achieving universal primary schooling is an objective, the current bias in the allocation of current spending appears lopsided. Low-income countries that have managed to universalize primary school completion typically allocate a 50 percent share of public spending on education to the primary level (Mingat, Rakotomalala, and Tan 2002). Admittedly, increased spending alone would not be sufficient to produce results. Although figure 3-1 suggests a generally EDUCATION IN RWANDA 57 FIGURE 3-1: RELATION BETWEEN PRIMARY SCHOOL COMPLETION RATES AND PUBLIC SPENDING ON PRIMARY EDUCATION, LOW-INCOME COUNTRIES, CIRCA 2000 o 100 oo a o o 80 o (%) o o o o o rate o o o o o o 60 o o o o o o o o o o Completion 40 o oo o o oo o o Rwanda o o o o o 20 o o o 0 1 2 3 4 Public spending on primary education as % of GDP a.The completion rate is computed as the number of nonrepeaters in grade 6 relative to the corresponding age cohort for that grade. Source: Based on data for 47 low-income countries reported in Mingat, Rakotomalala, and Tan (2002). positive relation across countries between spending on primary education and grade 6 completion rates, the relation is loose, and Rwanda's position in the graph points to substantial scope for improvement in survival even at the current level of spending. Increasing allocations for primary schooling is a legitimate objective, but there is also a need to improve the system's current per- formance. Subsequent chapters of this volume elaborate on some options in this regard. In the discussion so far, we have focused on public spending on education managed by the Ministry of Education. The ministry is indeed responsible for the bulk of such expenditure, but in recent years a nonnegligible volume of education-related spending has flowed through other ministries and government agencies, notably for vocational and nonformal training and for finan- cial assistance, mostly to secondary students orphaned by the genocide (table 3-2). Including these spending items raises current spending on education in Rwanda to 3.8 percent of GDP in 2001, with 87.2 percent accounted for by spending channeled through the Ministry of Educa- tion.1 This level of aggregate current spending on education is comparable to the average of about 4.0 percent in low-income countries. Spending by Households Rwandan households complement the government's investment in education to a large measure. Data from the 1999­2000 Household Living Conditions Survey (also known as the Enquête intégrale sur les conditions de vie des ménages au Rwanda, EICV), a nationwide survey of some 6,420 households in urban and rural areas, suggest that almost all students, irrespective of level of schooling and no matter whether they attend public or private schools, incur some education- related spending (table 3-3).2 The surprising pattern is that the share of respondents with 1. Because the additional spending outside the ministry is for postprimary services, the results in figure 3-1 remain unchanged. 2. The survey was conducted between October 1999 and December 2000 in urban areas and between July 2000 and July 2001 in rural areas. 58 WORLD BANK COUNTRY STUDY TABLE 3-2: CURRENT PUBLIC SPENDING ON EDUCATION THROUGH THE MINISTRY OF EDUCATION AND OTHER MINISTRIES OR GOVERNMENT AGENCIES, RWANDA, 1999­2001 (Millions of current FRw, unless otherwise indicated) 1999 2000 2001 Ministry of Educationa 21,738 22,347 25,341 Other ministriesb Education and training servicesc 347 287 324 Financial assistance to students 513 3,205 3,400 Total education-related current spending 22,598 25,839 29,065 As % of GDP 3.5 3.7 3.8 Share spent through the Ministry of Education (%) 96.2 86.5 87.2 Share spent through other ministries and government agencies (%) 3.8 13.5 12.8 a. Excludes spending on the university hospital. b. Includes the Ministries of Finance;Youth, Sports and Culture; Gender; and Local Administration and Social Affairs (MINALOC), as well as the Genocide Fund, which became operational in 2000 when the first bursaries were awarded. c. Includes spending on the Institut supérieur des finances publiques (ISFP, now the Institute of Finance and Banking) and on various vocational and nonformal training courses. Source: See appendix table A3-2. nonzero spending is smaller in higher education than in primary or secondary school. In absolute terms, however, families spend more at higher levels of education: from US$5 per child per year in public primary schools to US$318 per year in higher education. Summing across all levels and types of education, household spending on education in 2000 amounted to about 1.5 percent of GDP, or about 41 percent as much as public current spending on education in that year. This level of private spending is relatively high and is comparable to that in Madagascar, another coun- try where the private sector accounts for just under half of the enrollments at the secondary level. As for the composition of spending, school fees are an important item throughout the sys- tem. In private primary schools and at postprimary levels, they are half or more of the total cost to families, while in public primary schools they account for 23 percent of the total. For children in public schools, obtaining uniforms turns out to be the single most important cost item (accounting for about 45 percent of the total), followed by books and school supplies (24 per- cent). The relatively large share of spending on uniforms is not unusual, but it does raise ques- tions about the benefits of requiring families to incur an expense that is arguably only indirectly related to the pedagogical process. In some low-income countries, including Tanzania and Uganda, fees for primary schooling have been abolished, and a system of capitation grants has been instituted to relieve families of the burden of paying for books and supplies. The expectation that these moves would speed progress toward achieving education for all is borne out by the rapid increase in enrollments in both countries in recent years. Closer examination of the evidence shows, however, that, although there is now nearly universal entry to grade 1, many pupils drop out before finishing the cycle. The implication is that additional measures to compensate for the opportunity cost of schooling may be needed to keep children, particularly those from poor families, in school.3 If the Rwandan government followed the example of Tanzania and Uganda by abolishing fees, reimbursing schools for the lost income, and paying for books and pedagogical supplies, 3. See Patrinos and Ariasingam (1997) and Patrinos (2002) for examples of the many mechanisms used by countries to keep children in school. EDUCATION IN RWANDA 59 TABLE 3-3: HOUSEHOLD SPENDING ON EDUCATION, RWANDA, CIRCA 2000 Primary Secondary Publica Private Publica Private Higherb Total Share of students reporting nonzero 97.9 99.5 97.4 97.2 93.8 spending on education (%) Weighted annual spending 1,807 10,370 38,173 56,644 124,002 per student (FRw)c As % of per capita GDP 2.2 12.7 46.6 69.2 151.5 US$ 5 27 98 145 318 Composition of spending per student (%) School fees 22.7 60.2 54.4 68.4 68.4 PTA contribution and other charges 6.4 0.9 5.6 1.9 0.0 Books and school supplies 24.3 15.6 11.8 9.0 8.1 School uniforms 45.0 20.1 11.6 9.0 0.3 Transport to and from school 0.2 2.3 6.1 4.8 11.6 Other expenses 1.3 1.0 10.5 6.9 11.5 Aggregate household spending on 2,647 114 3,042 3,482 1,223 10,508 education (millions of FRw) Subsector share of aggregate (%) 25.2 1.1 29.0 33.1 11.6 100.0 As % of GDP 1.5 As % of public current spending 40.7 on education Memorandum: Number of students nationwide 1,464,594 10,978 79,699 61,464 9,866 1,626,601 Blanks denote not computed. Note: FRw, Rwandan francs, GDP, gross domestic product; PTA, parent-teacher association. Results are based on samples of 6,284 primary school pupils, 945 students in secondary school, and 72 students in universities and other types of institution. a. Refers to state and libre subsidié schools.Teacher salaries in both types of school are financed by the gov- ernment. State schools are managed directly by the government; libre subsidié schools are managed by churches and other nongovernmental organizations. See chapters 5 and 6 for further details. b.To avoid problems associated with small sample sizes, the public and private sectors are not shown separately. c. For all students, including those reporting zero spending. Source: Authors' calculations based on the 1999­2001 Household Living Conditions Survey. Data on enroll- ments are based on sources cited in appendix tables A2-1 and A5-1. Data on GDP in 2000 are from the World Bank Africa Live Database (May 20, 2002, version).Additional details are provided in appendix table A3-3. public spending on primary education would need to rise by 0.18 percent of GDP [= (0.227 + 0.243) × 0.252 × 1.5]. This would bring total current public spending for primary schooling to about 1.68 percent of GDP, an increase of about 12 percent over recent years. Such an increase would seem to be the minimum needed to remove some of the immediate impediments toward achieving education for all. A Closer Look at Public Spending on Education In this section we discuss in greater detail the functional distribution of public spending on edu- cation. We first examine the broad categories of spending reported in the budget. We then build the spending allocations from the bottom up by level of education, as implied by the number and distribution of teaching and nonteaching staff in schools and by the pay structure. 60 WORLD BANK COUNTRY STUDY TABLE 3-4: DISTRIBUTION OF CURRENT PUBLIC SPENDING ON EDUCATION, RWANDA, 1999­2001 (%, unless otherwise indicated) 1999 2000 2001 Salaries of teachers and other school personnela 59.0 55.0 53.4 Primary 41.0 37.0 34.5 Secondary 10.1 8.8 9.1 Higher 8.0 9.2 9.9 Student bursariesb 14.3 24.3 24.0 Secondary 2.3 11.9 11.6 Local higher education 3.4 5.2 4.9 Higher education abroad 8.6 7.2 7.5 Food for students 6.0 8.2 8.7 Secondary 1.2 3.6 3.6 Higher education 4.8 4.6 5.0 System administration 7.2 4.0 7.2 Other operating expensesc 13.4 8.4 6.7 All categories 100.0 100.0 100.0 Memorandum: Total current spendinga,b In millions of current FRw 21,862 25,212 27,520 As % of GDP 3.4 3.6 3.6 Note: FRw, Rwandan francs; GDP, gross domestic product. a. Excludes spending on teacher salary arrears, the university hospital, the Institut supérieur des finances publiques, and various vocational and nonformal training activities offered by other ministries.These exclusions explain the difference between the totals in this table and those shown in table 3-2. b. Includes bursaries administered through the Ministère de l'administration local et des affaires sociales (MINALOC) and the Genocide Fund. c. Includes spending on maintenance, transport, utilities, and other running costs. Source: See appendix table A3-4. Overall Functional Distribution of Spending The rise in overall current spending on education in recent years has been accompanied by a noticeable shift in spending in favor of financial and other assistance to students in secondary and higher education, at the expense of allocations for teacher and staff pay, particularly at the pri- mary level, and for operating expenses (table 3-4). The increase in allocations for student bur- saries coincides with the activation of financial assistance provided through the Genocide Fund to orphaned students; the bulk of current beneficiaries are attending secondary school. Funding for food for students has also been claiming a rising share of spending, and the government has moved to curb this trend by shifting to lump-sum cash allocations instead of basing the alloca- tions on physical quantities of food per student. The new arrangement will help tighten control over this item, but justifying its continued inclusion in the budget is difficult, given that the ben- eficiaries are postprimary students, who are already privileged. Spending Allocations, from the Bottom Up To construct allocations from the bottom up, we merged information on the distribution of per- sonnel and the pay structure to piece together spending on wages and then completed the picture by adding information on spending on nonsalary items as reported in budget documents. The results show spending allocations by level of education and provide details that would otherwise EDUCATION IN RWANDA 61 TABLE 3-5: NUMBERS OF TEACHERS AND SCHOOL-LEVEL ADMINISTRATIVE STAFF IN GOVERNMENT-FINANCED PRIMARY AND SECONDARY SCHOOLS, RWANDA, 1999 Teachers With teaching With other Administrative duties duties staffa Total Primary 24,982 40 1,329 26,351 Secondaryb 3,257 245 964 4,466 Tronc commun cycle 1,907 -- -- -- Upper secondary cycle 1,350 -- -- -- -- Not available. a. Refers to secretaries and school heads with no teaching duties; does not include personnel who provide dormitory and cafeteria services. b.Teachers who teach classes in both the tronc commun and upper secondary cycles are counted in full-time equivalent units according to the distribution of teaching duties between the two cycles. Source: Based on data from Ministry of Education 1999 census of teachers in state and libre subsidié schools. be obscured. The calculations focus on primary and secondary education, which account for the bulk of the system. For higher education, we did not use the bottom-up approach because the requisite data were unavailable. Instead, we relied on budget documents, which, as it turned out, contained sufficient detail for our purpose. Consider first the distribution of staff working in public (state and libre subsidié) primary and secondary schools (see table 3-5). The data are based on a 1999 census of school personnel con- ducted by the Ministry of Education, which documented each employee's educational qualifications and work responsibilities. From the data, it is possible to reclassify teachers with no teaching duties and to evaluate the full-time equivalent number of teachers in the tronc commun and upper second- ary cycles according to teachers' time utilization. The calculation indicates that nearly 60 percent of the teaching time of the 3,257 secondary teachers on the payroll in 1999 was devoted to classes in the tronc commun cycle. We next consider the distribution of these school-level personnel by salary grade (see table 3-6), again on the basis of the 1999 Ministry of Education census of school personnel. Using informa- tion from the census on each staff member's salary grade and years of experience, we are able to locate each person in the salary scale and thus compute that person's annual salary, including benefits. On average, primary school teachers receive remuneration of about 4.0 times per capita GDP; the corresponding multiples were 5.7 for tronc commun teachers and 6.8 for upper secondary teachers. In both the primary and secondary cycles, administrative staff tend to be better educated and to have more years of experience than teachers as a group, and they are generally paid more. By combining information from tables 3-5 and 3-6, we are able to estimate the total salary bill broken down by level of education and disaggregated between teaching and administrative staff working in schools. The salary bill of the central administrative staff was not estimated but was taken directly from the amounts reported in the budget documents for 1999.4 The reported aggre- gate spending on this item was split across the various levels of education by prorating it according to the distribution of the salary bill for school-level personnel, on the assumption that the distribu- tion reflects the cost of personnel management across levels. We then added the details on nonstaff current spending as reported in the budget documents, being careful to eliminate spending unrelated to education, such as the cost of a university hospital. The results appear in table 3-7. As a check 4. For further details, see the notes to table 3-7. 62 WORLD BANK COUNTRY STUDY e teach y inté- pri- ears.y the the ear 0.0 0.2 0.1 3.1 4.0 7.7 env and teacher staff 58.3 23.5 10.8 9.4 tisanal se of 1,209 594.9 ar hours eight-y e to et heads Administrativ of old six y rural the after tructurs y umbern the ment wing household 1999, the earned e on ar A Secondar Upper 0.0 0.8 0.0 2.6 1.6 5.6 6.8 is 42.8 30.1 22.1 lloof b secondar 1,350 526.2 ot it who AND ding d'enseigne tem, W es courses sys Education ,R eachersT accor w of un ne teachers y oncrT 0.0 1.3 0.0 1.8 6.8 0.7 5.4 5.7 centr 74.3 15.1 1,907 443.2 the to ALARIES comm ocationalv Ministr S paid CERAI, equivalents ed under the e; er GE yb e off wances ee VERA full-time schooling; A allo secondair thr y in y qualification. supplied staff 3.6 8.1 2.0 0.2 0.1 0.4 5.0 yb 30.2 55.4 11.1 last 1,369 388.6 famil e AND ecole y The secondar Administrativ counted ES, .y of e ormation RADE structur ar onl inf G cycle; ys earsy y estimated Primar a Y y Rwanda. e bo as and cycles de fiv salar ell ALAR 3.7 0.0 0.0 0.4 8.1 4.0 orf to w the S 22.5 20.5 17.9 35.0 primar eachersT 308.5 24,982 both ee as on BY in ear tisanal thr teachers ar ogram, of c et pr after 1999), AFFT eight-y . S in ormation CERAR, d teaching rural census d old inf SF CERAR earned oduct. security diploma the pr the was oduced 1999 CHOOL of system) system) S w eachersT (intr ditional CERAI, and social y end ad training (old (ne y diploma diploma to y orf y e domestic duties. duties. the d'enseignement onl Education UBLIC P ears),y lodging at es of F ee secondar girls teacher diploma diploma and salaries y A5-1 O attainment thr y y y ossrg, ee benefits teaching teaching centr orf secondar t of table to upper GDP GDP Ministr postsecondar degr FRw was cent with with obtained, upper (one om secondar secondar secondar ear entr pita CERAR, an transpor per fr ca francs; 5 Educational ES, er ersity including,y o-y cur ptitude ogram orf ppendixa of ISTRIBUTION A C Incomplete woL pper pper U U wT experience per pr data Univ salar of teachers teachers d'a system, D See of of y y familiale; SF on y characteristics staff ual onl onl . old benefits or of earsy Rwandan, tificat the 3-6: ann the ultiple staff benefits. thousands m FRw cycle cer, Based contribution categ total 6 5 4 3 2 1 A section, cycle; eragev eragev In As Includes SF s y and ABLE ya of Under each Number A A Note: a. Includes.b C.c d. Includes.e Source: y T P % Ungraded Other Other in gré; mar state' pa EDUCATION IN RWANDA 63 TABLE 3-7: ESTIMATED CURRENT PUBLIC SPENDING ON EDUCATION BY FUNCTION AND LEVEL, RWANDA, 1999 Secondary education (Millions of current FRw, unless Primary Tronc Upper Higher All otherwise indicated) education commun secondary education levels Salaries and benefits Staff at central levela 1,056 158 129 405 1,748 School-level administrative staff } b 532 404 317 3,239 13,755 Teachers with teaching dutiesb 7,707 845 710 Nonstaff current spending System administrationa 1,140 106 79 177 1,502 School-level operating costsc 346 801 475 -- 1,621 Food for studentsd 0 169 95 1,045 1,308 Student financial aid Local study 0 322 191 733 1,246 Study abroad 0 0 0 1,889 1,889 Othere 6 6 6 6 23 Total current spending 10,788 2,810 2,001 7,494 3,093 Memorandum: Amount reported in budget documents FRw -- -- -- -- 21,862 (As % of estimated total) (94.7) -- Not available. Note: FRW, Rwandan francs. a. The information in the budget documents can be categorized separately as (a) spending on services that benefit primary and secondary education only, such as the school inspectorate, the National Examination Council of Rwanda, and the National Curriculum Development Centre, and (b) spending on overall administra- tion at the central level. A distinction can be made in each of these categories between spending on salaries and on materials inputs (transport, fuel, supplies, utilities, maintenance, and so on).We apportioned the report- ed amount in item (a) by level according to the distribution of enrollments, on the assumption that the cost of these services is driven by the size of enrollments.The total spending on item (b) was apportioned according to the distribution of the salary bill for school-level or institution-level staff, on the assumption that the salary bill is correlated with the cost of personnel management. b. Estimated according to the average annual salaries, including benefits, of administrative staff and teachers in publicly funded primary and secondary schools and the number of such staff, as documented in the Ministry of Education 1999 census of teachers. c. Includes spending on materials, books, and maintenance. For secondary education, the amount also includes spending on equipment.Although some of this spending may be more properly treated as capital expenditure, no adjustment is made here because no information is available regarding the nature of the equipment. d. For a few higher education institutions, the amounts spent on food are not itemized in the budget docu- ments and are therefore estimated separately from the data supplied by individual institutions.The estimates are added to the itemized budget amounts to obtain an aggregate for higher education. e. Refers to spending on the National Commission for UNESCO and on research on science and technology. We apportioned these amounts equally across levels on the assumption that such spending benefits the entire education system. Source: Authors' estimate based on data on the number and distribution by pay category of teachers and admin- istrative staff counted in the Ministry of Education 1999 census of teachers, supplemented by data on the structure of teacher remuneration and data from the 1999 executed budget. See also table 3-4 and appendix table A5-1. 64 WORLD BANK COUNTRY STUDY on the overall calculations, we note that the bottom-up procedure yields aggregate spending of 22,093 million Rwandan francs (FRw), an excess of 5.6 percent over the FRw 21,862 million reported in the budget. The gap is larger than we would have liked in this kind of accounting exercise but may be explained by the government's accumulation of arrears on teacher salaries in the past few years. To facilitate understanding of the results in the previous table, we can rearrange them under three major heads: management overhead, service delivery at the facility level, and student wel- fare. This rearrangement brings a new perspective to the management of public spending on edu- cation by drawing attention to the extent to which spending is (or is not) oriented toward the core business of service delivery. Management overhead can be thought of as a backline service that supports the core business; it includes spending on managerial and supervisory staff at the central ministry and decentralized levels and on the material inputs needed by such staff to dis- charge their functions. The core business of service delivery takes place at the level of individual schools and is realized through spending on teachers, school administrators, and support staff, as well as on pedagogical supplies and other operating inputs. Spending on student welfare can be categorized as a backline item, since it is only indirectly linked to teaching and learning at the level of individual classrooms and schools. To continue with the business analogy, student welfare spending can be viewed as a form of customer rebate because it reduces the cost of education services to consumers (students and their families). The rearranged data (table 3-8 and figure 3-2) provide some insights into priorities in public spending on education. Although there is admittedly no obvious benchmark for judging the dis- tribution of spending, a pattern of allocation that glaringly fails to support the core activity of teaching and learning at the facility level would raise legitimate questions about whether resources are sufficiently focused on service delivery. For the system as a whole, less than two- thirds of total public current spending on education in 1999 went to support service delivery at the facility level, whereas 14 percent was absorbed by management overhead and nearly 20 per- cent went into student welfare. The allocation for management overhead does not appear to be overly burdensome, but the share for student welfare is unusually large compared with spending patterns elsewhere. The size of this item reflects the country's postgenocide recovery efforts aimed at rebuilding lost human capital through heavy investment in higher education and at help- ing students orphaned by the massacre. Closer examination of the spending patterns by level of education raises some concerns. In primary education, management overhead exceeds 20 percent, reflecting relatively elevated spending on material inputs such as transport, utilities, and maintenance. At the school level, 96 percent of the allocation goes for staff salaries, leaving a minuscule 4 percent for books, pedagogi- cal supplies, and other running costs and nothing at all for student welfare. Given the paucity of resources for items other than salaries, it is not surprising that a sizable portion of household spending at this level of education goes for school fees and parent-teacher association dues (both kept for use at the school) and for books and school supplies. A reevaluation of the allocations might be warranted to identify the potential scope for shifting some of the government's spend- ing on overhead toward defraying more of the running costs at the school level. There is current- ly no provision for subsidizing orphaned children in primary schools--an omission that poses an obvious impediment to the schooling of the most vulnerable children. In secondary education, spending allocations between the lower and upper secondary levels are comparable--roughly, a tenth on overhead, three-quarters on service delivery, and the rest on student welfare. Within the service delivery category, administration appears to take up a signifi- cant share of resources. A closer look at this item may be worth while in order to improve utiliza- tion of personnel in schools and avoid, for example, poorly justified assignment of teachers to administrative rather than teaching tasks. Allocations for student welfare consist of spending on school feeding and bursaries for orphaned students. The raison d'être for the general food subsi- dies appears weak, since not all students at this level are poor; at any rate, they are not poorer EDUCATION IN RWANDA 65 TABLE 3-8: FUNCTIONAL DISTRIBUTION OF CURRENT PUBLIC SPENDING ON EDUCATION, RWANDA, 1999 Secondary education Primary Tronc Upper Higher (%, unless otherwise indicated) education commun secondary education All levels Management overhead 20.4 9.6 10.7 7.8 14.3 Staff salaries 9.8 5.6 6.5 -- -- Material and other inputsa 10.6 4.0 4.2 -- -- Service delivery at facility level 79.6 73.0 75.0 43.2 66.4 Salaries of teachers with teaching 71.4 30.1 35.5 -- -- duties Salaries of administrative staffb 4.9 14.4 15.8 -- -- Material inputs and other 3.2 28.5 23.7 -- -- operating costsc Student welfare 0.0 17.5 14.3 48.9 19.4 Food for students 0.0 6.0 4.7 13.9 5.7 Bursaries for local study 0.0 11.5 9.5 9.8 5.4 Bursaries for study abroad 0.0 0.0 0.0 25.2 8.2 Total 100.0 100.0 100.0 100.0 100.0 Memoranda: Total current spending 10,788 2,810 2,001 7,494 23,093 (millions of current FRw) Subsector share of spending (%) 46.7 12.2 8.7 32.5 100.0 -- Not available. Note: FRw, Rwandan francs.The categories of functions represent a regrouping of the expenditure categories in table 3-7. a. Includes transport, utilities, supplies, maintenance, fuel, and other running costs. b. Includes the salaries of teachers without teaching duties, school heads, secretaries, and other administrative staff. c. Includes textbooks, equipment, utilities, transport, maintenance, and other running costs at the facility level, as well as salaries of casual labor. Source: Authors' estimate based on data in table 3-7; see the text discussion for definitions of items. than children in primary school who currently do not get free food.5 The justification for bur- saries is stronger because they are directed at orphans. As documented in chapter 1, one result of the 1994 genocide was to swell the number of orphans currently in the secondary school age range. As this bulge passes out of the system, allocations for secondary-level bursaries can be expected to decline but not to disappear altogether, as the HIV/AIDS epidemic continues to produce orphans. With regard to spending allocations in higher education, there is less detail, but even limited information is helpful in identifying questions about spending priorities. The most striking find- ing is that the core business of service delivery received only 43 percent of the subsector's total 5. As noted earlier, in an effort to control spending on food, the government recently replaced calcula- tion of the subsidies on the basis of the number of students enrolled in a school with a lump-sum cash amount fixed in nominal terms in the base year. As more of the country's schools begin to operate as day rather than boarding schools, the amount spent on food can be expected to decrease. 66 WORLD BANK COUNTRY STUDY FIGURE 3-2: FUNCTIONAL DISTRIBUTION OF CURRENT PUBLIC SPENDING ON EDUCATION, RWANDA, 1999 100 80 60 spending of 40 % 20 0 Primary Tronc Upper Higher All levels education commun secondary education education Student welfare Service delivery Management overhead Source: Based on data in table 3-8. current spending, compared with 49 percent for student welfare in the form of food and bur- saries. Because spending on overhead is not especially high (less than 8 percent), a rebalancing of spending in favor of service delivery would essentially necessitate cutting back on student welfare. The only way to avoid the cutback is to raise the share of higher education in overall spending on education, but this option is untenable in view of the subsector's already high share of total spending (37 percent in 2001, up from 33 percent in 1999). The issues for policymakers are to evaluate the merits of shifting spending within the subsector and to form a strategy for minimiz- ing the potential adverse effects associated with cutbacks in student welfare. These negative effects may include, for example, a bias against students from poor families in the socioeconomic composition of the student body and inadequate incentives for students to specialize in fields of study with high social (as opposed to private) returns. Public Spending per Student In this section we discuss patterns in per student spending and the underlying sources of differ- ences across levels of education. Data from other countries are presented to lend a comparative perspective. Unit Spending by Level of Education The relevant results, shown in table 3-9, pertain to the public sector at all levels. (In higher edu- cation, a distinction is made between local study and study abroad.) An immediately obvious pattern is that unit costs are highly skewed in favor of the postprimary levels--another reminder of the relatively meager allocation of public spending for primary education in Rwanda. Each secondary student costs 8.6 times as much to enroll as a primary school child. In higher educa- tion, the corresponding ratio is 95 times for local studies and 275 times for study abroad. Within secondary education, unit costs in the tronc commun cycle are about four-fifths those in the upper secondary cycle, suggesting that service delivery arrangements (for example, use of specialist teachers, class size, teaching workloads, and instructional time for students) in this subcycle more EDUCATION IN RWANDA 67 TABLE 3-9: PUBLIC SPENDING PER STUDENT BY LEVEL OF EDUCATION, PUBLIC SECTOR, RWANDA, 1999 Secondary education Higher education Primary Tronc Upper Local Study education commun secondary Overall study abroad Spending per public sector student In FRwa 7,604 60,273 73,628 65,223 723,899 2,094,422 As multiple of per capita GDP 0.10 0.78 0.95 0.84 9.34 27.03 As multiple of spending on 1.0 7.9 9.7 8.6 95.2 275.4 primary education Memoranda: Total public current spending 10,788 2,810 2,001 4,811 5,605 1,889 (million FRw) Overall number of students 1,428,708 79,454 45,670 125,124 9,866 902b % of students in public sector 99.3 55.7 57.1 56.2 78.5 100.0 (state or libre subsidié ) schools Note: FRw, Rwandan francs; GDP, gross domestic product. a. Excludes spending on bursaries for students attending private secondary schools.The amount is estimated by prorating spending on bursaries according to the distribution of enrollments between public (state and libre subsidié) schools and private schools. Note that although private school students are probably less likely than public school students to receive a bursary, the amount of each award is higher for students attending private schools. b. Includes only those studying abroad on government scholarships. Source: Authors' calculations based on public spending data in table 3-7 and enrollment data in appendix table A2-1; for number of students abroad, Ministry of Education data. closely resemble those in upper secondary education than in primary schooling. The relatively high costs in the tronc commun cycle will limit the government's ability to accommodate the rising demand for places in this cycle as increasing numbers of children complete primary school. Sources of Differences in Unit Spending across Levels Why are the postprimary levels of education so much costlier than primary education? To answer this question, we decompose overall spending per student into its component parts, beginning with the following expression: TS i US = i (3-1) P where US is overall unit spending, TSi is total spending on component i, and P is the number of students in the public sector. Unit spending on the salaries of teachers and nonteaching staff (USS) can be expressed as a function of two items: the average salary cost of these staff, and the corresponding pupil-to-staff ratios: USS = TSS = ASS NS = ASS 1 (P NS) = ASS (3-2) P P PSR where TSS is total spending on staff salaries, ASS is average salary per staff member, NS is number of staff, and PSR is the pupil-to-staff ratio. With the use of equation 3-2, differences in unit 68 WORLD BANK COUNTRY STUDY TABLE 3-10: LEVEL AND COMPOSITION OF PER PUPIL PUBLIC SPENDING IN PUBLIC PRIMARY AND SECONDARY SCHOOLS, RWANDA, 1999 Primary education Secondary education as multiple (FRw, unless of primary educationa otherwise Both Tronc Upper specified) levels commun secondary Overall spending per student 7,604 8.6 7.9 9.7 Cost of teachers per student 5,433 4.1 3.5 5.0 Average annual teacher salary 308,522 1.5 1.4 1.7 Ratio of students to teachers (56.8) 2.6 2.4 2.9 Cost of school-level administrative staff 375 27.3 per student Average annual salary of administrative staff 388,594 1.5 Ratio of students to administrative staff (1,036) 17.8 School-level material inputs per student 244 74.4 Student welfare spending per student 0 n.a. Management overhead per student 1,552 4.4 n.a. Not applicable (no spending on student welfare at the primary level). Note: FRw, Rwandan francs. Public sector includes state and libre subsidié schools. a. Blanks denote that the calculation was not made; differences across the two levels are small and reflect assumptions in apportioning spending across the two cycles of secondary schooling. Source: Authors' calculations based on data in tables 3-6 and 3-7. See also appendix table A3-5. spending across levels or types of schooling can be compared in terms of underlying differences in level of salary per staff and the pupil-to-staff ratio, which may be interpreted as a proxy for the intensity of use of staff. Keeping in mind the above expressions, consider now the results in table 3-10 (which, because of data limitations, pertain only to primary and secondary education). The overall cost per pupil in primary education (FRw 7,604) is divided into FRw 5,433 for teacher salaries, FRw 375 for the cost of school administration, FRw 244 for material inputs at the school level, and FRw 1,552 for management overhead at the systemwide level. Using equation 3-2, we compute the cost of teacher salaries per pupil simply by dividing the average annual salaries of teachers (FRw 308,522) by the pupil-teacher ratio (56.8) and the cost of school-level administration by dividing the average annual salaries of school administrative staff by the ratio of pupils to such staff. Similar calculations yield the corresponding costs per secondary student, overall and sepa- rately for the tronc commun and upper secondary cycles. To facilitate the discussion below, the results for secondary education are expressed in the table as a multiple of the corresponding indi- cator in primary education. Unit spending in secondary education is much higher than in primary education--8.6 times overall, and 7.9 and 9.7 times in the lower and upper secondary cycles, respectively. Part of the reason for the differences lies in teacher salaries, which are the single most important cost item, accounting for more than 70 percent of total spending. The per student cost of teachers in 1999 was 4.1 times as high in secondary schools as in primary schools, reflecting higher teacher salaries and smaller student-teacher ratios in the secondary cycle. A closer look at these variables shows that the latter factor accounted for more of the cost difference. Similarly, higher numbers of staff in relation to students, rather than higher average salaries, were the main reason why the per student cost of school-level administrative staff in the secondary cycle in 1999 was 27 times the EDUCATION IN RWANDA 69 TABLE 3-11: CURRENT PUBLIC SPENDING PER STUDENT, RWANDA, 1999, AND SELECTED COUNTRY GROUPS, 1990S As multiple of spending As multiple of per capita GDP on primary education Local Local Number of higher higher countries Primary Secondary education Secondary education Rwanda, 1999a 1 0.10 0.50 7.3 5.1 75.2 (0.10) (0.84) (9.3) (8.6) (95.2) Regional average, 1990sb Francophone Africa 15 0.15 0.49 5.6 3.3 37.3 Anglophone Africa 9 0.10 0.66 6.3 6.6 63.0 Latin America 10 0.07 0.11 0.7 1.6 10.0 Asia 8 0.08 0.19 0.9 2.4 11.3 Middle East and 6 0.11 0.30 0.9 2.7 8.2 North Africa a. Because the cross-country data refer to spending distributed across all enrollments, public and private, we adjusted the data for Rwanda accordingly.The first row refers to average current spending for all students at each level, regardless of whether they are in public or private schools and regardless of type of institution. In the second row (in parentheses), the numbers for primary and secondary education refer to average current spending only for students in public schools. For higher education, the figures in parentheses refer to per stu- dent spending in traditional public institutions. b.Averages based only on data for countries with per capita gross domestic product (GDP) below US$1,000 in 1993. Source: For Rwanda, table 3-9; for country groups, Mingat and Suchaut (2000). corresponding cost in the primary cycle. Fortunately, school-level administration was not a major expenditure item in 1999, and its impact on overall costs was limited. Secondary schools received more funding per student for material inputs--about 74 times that at the primary level--but this item, too, accounted for a small share of overall spending. Management overhead took up about 20 percent of overall spending per primary pupil in 1999. Spending per student on this item in secondary education was higher than for primary education, by a multiple of about 4.4. Cross-Country Comparison of Unit Spending Table 3-11 presents comparisons with other countries on patterns of unit spending. Public spend- ing averaged across all students in the system, both public and private, as is done in the table, can be interpreted as reflecting the intensity of public spending for each level of education and is suitable mainly for comparing the structure of spending across levels. Public spending per public school student, shown for Rwanda only, is more appropriately seen as a measure of the costliness of service delivery in the public sector.6 Average unit spending in primary education in Rwanda is comparable to the averages for low- and middle-income countries in Anglophone Africa and in the Middle East, somewhat high- er than the averages for countries in Latin America and Asia, and quite a bit lower than those for Francophone Africa. In secondary education Rwanda's figure is higher than the averages for all regions except Anglophone Africa. It is in higher education, however, that Rwanda's level of 6. Because of differences in data definition, comparisons of countries' spending per student should be handled with the usual caution--particularly in higher education, where enrollments in private institutions may not be fully accounted for. Distance education should be treated differently from country to country. 70 WORLD BANK COUNTRY STUDY TABLE 3-12: PUBLIC SECTOR TEACHER SALARIES AND PUPIL-TEACHER RATIOS, RWANDA, 1999,AND SELECTED COUNTRY GROUPS, 1990S Average salary of teachers as multiple Country or region of per capita GDP Pupil-teacher ratio Rwanda, 1999a 4.0 56.8 Regional average, 1990sb Francophone Africa 6.3 53.2 Anglophone Africa 3.6 38.7 Latin America 2.3 31.0 Asia 2.5 37.9 Middle East and North Africa 3.3 25.6 All countries in table 3.9 40.0 a. Data for secondary education are the weighted averages of lower and upper secondary school.All data refer to the public sector. b. Data refer to countries with per capita gross domestic product (GDP) below US$1,000 in 1993. Pupil- teacher ratios pertain to systemwide averages. Source: For Rwanda, appendix table A3-5; for data on country groups, Mingat and Suchaut (2000). spending per student shows up as a true outlier on the high side: it is 75 times the level of spend- ing per primary school child, compared with the corresponding multiples of 63, on average, in Anglophone Africa, 37 in Francophone Africa, and about 11 or less in the other regions. Had higher education study abroad been included in the calculation, the ratio would have been even higher. Overall, the results suggest a structure of spending in Rwanda that is heavily skewed in favor of postprimary levels, especially higher education--a pattern that inevitably translates into significant inequities in the incidence of public spending. Implicit Tradeoffs in the Input Mix in Primary Education For any given level of spending per student, education services can be delivered using different input mixes--for example, less costly teachers and more favorable staffing ratios, or vice versa, or more spending on salaries and less spending on nonsalary inputs. In table 3-12 we use cross- country data on teacher salaries and pupil-teacher ratios in primary education (the level for which comparative data are relatively plentiful) to deduce the implicit tradeoffs in primary schooling in Rwanda. The discussion provides a basis for a dialogue on options for improvement. Recall from equation 3-2 that the per pupil cost of teachers--the single most important com- ponent of overall costs--depends on the level of teacher salaries and the pupil-teacher ratio. In general, the pattern in these variables across countries is what one would expect: the less costly are the teachers, the lower pupil-teacher ratios tend to be. Rwanda's teacher salaries exceed the average for Anglophone Africa by 11 percent, but its pupil-teacher ratio is higher by nearly 50 percent. Its figures for these variables are 21 and 122 percent higher, respectively, than the corre- sponding averages for the Middle East and North Africa.7 Because overall spending per student in Rwanda is comparable to the averages for the selected regions (around 0.10 times per capita GDP), we can deduce from the much higher pupil-teacher ratios in Rwanda that an implicit tradeoff is being made in favor of nonsalary components. As documented above, the tradeoff 7. In Rwanda the exceptionally high pupil-teacher ratio translates into short school days--about 2.5 to 4 hours long--for children in the first three grades of the cycle because teachers are typically assigned to teach two shifts. (See chapter 5 for further details.) EDUCATION IN RWANDA 71 appears to emphasize system-level management overhead rather than services and material inputs at the school level. This is a regrettable result because research and experience consistently point to the critical role of books and pedagogical supplies in enhancing student learning.8 In light of these deductions from the comparative data, one option for improving conditions at the classroom level would be to redirect some of the spending on management overhead toward achievement of smaller pupil-teacher ratios and larger allocations for books and material inputs. If, for example, spending on management overhead were cut by half and the resources thus freed were used to hire more teachers, it would be possible to lower the pupil-teacher ratio to about 50. Although this reduction is certainly an improvement, a ratio of 50 is still high, and books and pedagogical inputs are still undersupplied. Other ways of improving the learning environment clearly warrant consideration. If overall spending on primary education remained unchanged, the only real option within the subsector would be to tighten controls on teacher salaries, either by changing the criteria for new hires or by slowing the pace of salary increases for those already on board. Such policies obviously need careful evaluation, but it should be noted that in low-income countries which have achieved uni- versal primary school completion, teacher salaries typically average about 3.5 times per capita GDP, and these countries generally maintain a pupil-teacher ratio of around 40 (Mingat, Rako- tomalala, and Tan 2002). If in Rwanda average salaries relative to per capita GDP were made comparable to the level in these other countries, all else remaining unchanged, it would be possi- ble to lower Rwanda's pupil-teacher ratio from 57 to 50. If, in addition to the salary change, spending on management overhead were cut in half and the resources thus released were used to hire more teachers, the ratio would drop to 44. As a further context for discussing the management of teacher costs, consider the data in table 3-13 on the pay of teachers and other workers. With the exception of those with the lowest academic credentials, teachers earn a significant premium over other similarly qualified workers. In the lowest pay category, teachers start their careers with a pay advantage of about 8 percent, but by midcareer the gap has just about evaporated. The bias in favor of the better qualified is consistent with the government's policy of attracting the best people into the teaching profession. If it were not for budget constraints, this policy would improve staff quality, and the system could still maintain reasonable pupil-teacher ratios and provide adequate supplies of pedagogical materi- als. The reality, however, is that budget constraints are tight. If lowering the pupil-teacher ratio is an objective, the option of keeping a lid on teacher costs cannot be entirely ruled out. Policy Implications Rwanda's public spending on education has been rising steadily in the postgenocide years, reach- ing a historical high of 5.5 percent of GDP in 2001. Because much of this increase has been for capital spending, allocations to support the day-to-day operations of the education system rose to only 3.3 percent of GDP in 2001--about the same level as in the late 1980s. The amount is aug- mented by education-related spending through channels other than the Ministry of Education, including, in particular, the Ministry of Youth, Sports and Culture, which finances vocational training, and the Genocide Fund, which provides financial assistance to orphaned students. Adding the extra spending brings total current spending for 2001 to about 3.8 percent of GDP. This level is comparable to the 4.0 percent of GDP that low-income countries typically spend on education, but given the extraordinary burden of orphans in the population, one could argue that Rwanda's allocation remains on the low side. 8. The squeeze on public funding for books and pedagogical materials effectively transfers the burden of financing these inputs to pupils and their families, as is confirmed by the pattern of household spending documented earlier in this chapter. Because of the prevalence of poverty in Rwanda, it is likely that families spend less than what is in fact needed for effective learning. 72 WORLD BANK COUNTRY STUDY TABLE 3-13: TAKE-HOME PAY OF PUBLIC SECTOR TEACHERS AND OTHER WORKERS IN RELATION TO PER CAPITA GDP, RWANDA, 2001 Starting pay as multiple Pay at midcareer as multiple of per capita GDPb of per capita GDPc Teacher Other Other pay Educational Teachers workers Teachers workers categorya attainment (1) (2) (1)/(2) (3) (4) (3)/(4) } } 6 Incomplete upper secondary 2.2 2.4 2.4 1.08 3.6 0.97 4 3 to 5 years postprimary 3.0 4.6 education (D3, D4, or D5) 3 Upper secondary diploma, 4.1 3.5 1.17 6.5 5.2 1.25 new system (D6, D7) 2 2-year postsecondary 5.6 4.9 1.14 8.4 7.1 1.18 (e.g., baccalauréat) diploma 1 University degree 8.3 6.8 1.22 11.1 9.5 1.17 Note: Take-home pay refers to salaries plus allowances for travel and housing. See appendix table A5-1 for additional details. a. Pay category 5 is for teachers with a teacher-training diploma, a qualification obtained after completing teacher training at the lower secondary level. It is excluded here for lack of a comparable group in the general labor force. b.The starting pay for teachers is taken from the salary scale shown in appendix table A5-1. For other work- ers, it is simulated at ages 20, 20, 22, and 24, respectively, for the qualifications shown, based on the Mincerian earnings function reported in appendix table A3-6. c. Pay at midcareer for teachers is estimated at the midpoint between entry pay and top pay on the pay scale for teachers (appendix table A5-1). For other workers, it is the pay simulated at age 45 based on the Mincerian earnings function reported in appendix table A3-6. Source: Authors' calculations based on the data in appendix table A5-1 and on simulations based on the Mincerian earnings function reported in appendix table A 3-6. A case can therefore be made for continued increases in spending on education, particularly current spending. Yet in a context in which the government already spends more than a quarter of its resources on education, further increases are likely to come only slowly--certainly more slowly than the pace of increase in the late 1990s. The ineluctable implication is that future progress in educational development will increasingly require more effective allocation of spend- ing within the sector. In a country still recovering from the devastation of the 1994 genocide, it may appear heart- less to even talk of tradeoffs in spending allocations within a sector such as education. The pat- tern of spending in recent years, however, shows that implicit tradeoffs have in fact occurred. The government placed substantial priority on rebuilding the country's intelligentsia in the years fol- lowing the genocide, as reflected in the very rapid changes in the distribution of spending. By 2001, higher education claimed an astonishing 37 percent of total current spending, up from between 12 and 15 percent in the 1980s. Correspondingly, the shares of both primary and sec- ondary education fell significantly below their levels in the 1980s. Public spending on primary education averaged about 1.5 percent of GDP during 1999­2001, compared with an average of about 1.9 percent in the late 1980s. The bias in funding against primary and secondary schooling effectively shifts part of the burden of education finance to households. Nearly 90 percent of all household spending on edu- cation goes for these levels. Parents pay school fees and parent-teacher association dues to help EDUCATION IN RWANDA 73 defray schools' operational costs, and they are responsible for books and school supplies. These arrangements in primary schooling place Rwanda behind countries such as Tanzania and Uganda, where school fees have been abolished and textbooks are provided free of charge as part of an overall strategy for ensuring that all children are able to complete primary schooling and have access to basic learning materials. In secondary education, fees in the public and private sectors in Rwanda are comparably high, blurring the distinction between the two sectors. As expected, the country's current pattern of intrasector allocations shows up in a highly skewed structure of pub- lic spending per student by level. A student in higher education is 75 times as heavily subsidized as a child attending primary school--a result that makes Rwanda's structure of education finance one of the least equitable in the developing world. Although investment in postprimary levels of education will continue to be an important policy aim, ways must also be found to rebalance the allocations to support primary education, particularly if the government is to achieve education for all by 2015 as part of its strategy for educational development and poverty reduction. As more students complete primary schooling, the pressures to expand secondary education--in particular, initially, the tronc commun cycle--will also mount, and resources would have to be found for expanding that level, as well. One poten- tial source of savings is in the untargeted subsidies for secondary and higher education, such as spending on food. To give an idea of the scale of the opportunity cost, eliminating this item from higher education alone could mean a boost of nearly 10 percent in the current budget for pri- mary schooling or could eliminate the burden of paying for textbooks and school supplies for 17,000 children. A closer look at spending on bursaries in higher education might also be warranted to ensure that they are indeed targeted to students from poor families and for fields of study where public subsidization is justified (that is, where society benefits more than the individuals concerned). In secondary education, bursaries currently benefit orphaned children and should obviously continue as long as such students are present in the system. To stretch the resources available for expanding primary and lower secondary education, ways of managing the cost of service delivery at these levels also need to be examined. The options in primary education include reallocating spending from management overhead to pedagogical materials and ensuring that teacher salary costs are kept at sustainable levels. In lower secondary education, policies to reduce unit costs, particularly through better management of staff utiliza- tion, will be critical to efforts to expand enrollments in a fiscally sustainable manner. Conclusion Education finance is at a crossroads in Rwanda today. Financing for the sector recovered quickly in the postgenocide years, but the pace of future increases is likely to slow because of the many competing claims on the public purse. In this context, a combination of policies will probably be needed to achieve an efficient and equitable allocation of spending. The policies chosen should be those that will help create, at all levels, but especially at the base of the education pyramid, effec- tive learning environments with adequate overall funding and a good mix of school inputs. Although the appropriate balance among the various policy options is difficult to determine on the basis of financial considerations alone, the information presented in this chapter highlights potentially fruitful directions that can be explored as part of the process of policy development. It is especially important to increase funding for primary education, both by redirecting resources to that level and through better management of costs and allocations within the subsector. CHAPTER 4 SOCIOECONOMIC DISPARITIES IN EDUCATION C hapters 2 and 3 examined the performance of Rwanda's educational system in the aggre- gate, making no distinction between different groups. This chapter focuses on disparities in education in terms of enrollment, student flow, and the share of public spending appropri- ated by different groups in the country. In particular, it emphasizes differences across provinces, genders, urban and rural localities, and socioeconomic groups. Because so many Rwandan children are orphans, the chapter also compares educational access by orphans and nonorphans. The good news is that social disparities at the base of the educational pyramid, at least as reflected in enrollment patterns, are narrower in Rwanda than in other low-income countries. The fact that even orphans are reasonably well represented among primary school children sug- gests that the country's safety net for vulnerable children is at least ensuring that most of these children have access to primary schooling. Yet the risk of nonparticipation remains elevated among the most vulnerable children, in particular those who have lost both parents and those who live away from their parents, probably as workers or as street children with no adult supervi- sion. Broader efforts are also needed to ensure that all children not only start first grade but also reach the end of the cycle. At postprimary levels, the disparities widen substantially, and rural children and those from poor families lag especially far behind. Differences between girls and boys in educational access and survival emerge only in higher education. Thanks to the government's foresight in setting up the Genocide Fund, orphans are relatively well represented in secondary education, the level that the fund currently targets. Although similar financial assistance to other disadvantaged youths is likely to improve their participation rates, these children's underrepresentation in postprimary education may stem from poor academic progress, as well. To the extent that this is true--and the case is especially convincing for explaining girls' lag behind boys in entering higher education-- interventions to improve learning outcomes would seem particularly important. Aside from redressing socioeconomic disparities in enrollment, there is also scope for improving the incidence of public spending on education. In Rwanda the bias in favor of higher education is 75 76 WORLD BANK COUNTRY STUDY currently so strong that the 10 percent best educated in a cohort appropriates nearly three-quarters of the cumulative public spending on education received by the cohort. As a result, the Rwandan system is one of the least structurally equitable in Sub-Saharan Africa today. Overview of Participation Rates Table 4-1 documents the disparities in school participation.1 At the primary level, the overall gross enrollment ratio (GER) rose from 74 percent in 1992 to 108 percent in 2000. The dispari- ty between the lagging and leading provinces has, however, not become any narrower.2 In 1992 gross enrollment ratios ranged from a low of 61 percent in Kibungo to 92 percent in Kigali Ville--a gap of 31 percentage points. In 2000 the gap was still 31 percentage points, with the ratios ranging from 97 percent in Butare to 128 percent in Kigali Ville. At the secondary level, the overall gross enrollment ratio fell from 20 percent in 1992 to 11 percent in 2000. The dis- parity between lagging and leading provinces widened from 29 percentage points (the difference between 9 percent in Byumba and 38 percent in Kigali Ville) in 1992 to 42 percentage points (6 percent in Gisenyi and 48 percent in Kigali Ville) in 2000. At the tertiary level, the overall gross enrollment ratio was only 1.3 percent in 2000. Kibuye and Ruhengeri Provinces had very few students in higher education, whereas the gross enrollment ratio was above the national aver- age in Gitarama (1.7 percent) and Kigali Ville (2.5 percent). The advantage of certain provinces at the secondary and higher education levels may be partly attributable to students' moving to the localities where these levels of education are concentrated. The survey, however, included all members of a household as long as they had been present at some time in the last six months of the survey period, and so the disparities in participation rates cannot be wholly attributed to dif- ferences in the supply of services across provinces. Differences in gross enrollment ratios across provinces reflect to some extent the disparities between urban and rural localities. These disparities may be seen by using as an index the ratio of the rural GER to the urban GER. At the primary level, in 1992 the ratio between GERs was about 0.79; in 2000 it was 0.85. In this sense, some progress has been made in narrowing the disparity between rural and urban localities.3 At the secondary level, however, the index of rural GER to urban GER fell from 0.49 in 1992 to 0.18 in 2000, implying a significant widening of the gap between rural and urban participation rates.4 At the tertiary level, where the data pertain only to 2000, the rural gross enrollment ratio was only 6 percent that of the urban localities. Gender disparities are not particularly large, except at the tertiary level. At the primary level, GERs for girls and boys were practically the same in both 1992 and 2000. At the secondary level, there has in fact been a narrowing of the disparity between boys and girls; although the GER for girls was only 0.89 times that for boys in 1992, the ratio was about the same for both groups in 2000. Given the similarity of participation at the primary and secondary levels, it is somewhat surprising to observe a large gender gap in enrollments in higher education. In 2000 the GER for women in high- er education was only 0.44 times that for men. In a later section we explore possible reasons for the sudden emergence of a gap at this level of education and the implications for policy development. 1. To maintain internal consistency within the table, the data are based on a single source for each year. The ratios for circa 2000 differ slightly from those reported in chapter 2, for reasons explained in the rele- vant table footnotes 2. As described earlier, the gross enrollment ratio is calculated by dividing the total number of students enrolled at a given level of education by the population in the official age range for that level. An alternative indicator of school participation is the net enrollment ratio, which is computed in the same way except that the numerator includes only students in the official age range for the corresponding level of education. We use the gross enrollment ratio here to provide an overview of the system's coverage. 3. In absolute terms, however, the disparity at the primary level has not narrowed by any substantial amount. In 1992 the gross enrollment ratio in urban localities was 19 percentage points higher than in rural localities. In 2000 the gap was just less than 19 percentage points. 4. In absolute terms, the disparity between rural and urban localities at the secondary level increased from 18 percentage points in 1992 to 24 percentage points in 2000. EDUCATION IN RWANDA 77 TABLE 4-1: GROSS ENROLLMENT RATIOS BY PROVINCE, LOCALITY, GENDER,AND INCOME GROUP, RWANDA, 1992­2000 Primary education Secondary education Higher education, 1992 Circa 2000 1992 Circa 2000 circa 2000 Rwanda 73.9 108.3 20.4 10.8 1.3 By province Butare 69.6 96.9 24.2 9.7 0.8 Byumba 82.9 106.0 8.7 7.6 0.4 Cyangugu 67.6 103.7 22.1 9.6 0.5 Gikongoro 78.2 108.5 25.9 6.5 0.3 Gisenyi 77.2 113.1 15.7 5.9 0.1 Gitarama 67.7 118.9 20.3 9.1 1.7 Kibungo 60.9 110.9 19.6 15.8 0.6 Kibuye 71.2 111.4 10.0 9.6 0.0 Kigali Rural 67.7 106.5 19.6 12.9 0.3 Ruhengeri 85.3 117.6 18.9 6.9 0.0 Umutara -- 124.0 -- 21.5 0.7 Kigali Ville 92.1 128.3 37.7 47.8 2.5 By type of locality Urban 90.6 124.0 36.0 29.9 8.3 Rural 71.3 105.3 17.6 5.5 0.5 Index (urban = 1.00)a 0.79 0.85 0.49 0.18 0.06 By gender Boys 74.3 109.5 21.7 10.8 1.8 Girls 73.5 107.1 19.3 10.8 0.8 Index (boys = 1.00)a 0.99 0.98 0.89 1.00 0.44 By income groupb Richest 20% 89.3 127.6 39.1 29.4 6.1 Middle 40% 76.6 109.1 18.0 7.9 0.4 Poorest 40% 63.8 99.9 12.1 3.0 0.2 Index (richest 20% = 1.00)a 0.71 0.78 0.31 0.10 0.03 -- Not available. a.The indices in the locality and gender blocks are computed as the ratio between the gross enrollment ratio of the less favored group and that of the most favored group. In the income block, the index is the ratio between the poorest and richest groups. b.The income groups are defined using principal components analysis of sample households' (1) ownership of or access to such assets or facilities as piped water, flush toilet, electricity, radio, bicycle, motorcycle, car, and refrigerator and (2) number of persons per room.The asset list is comparable between 1992 and 2000. Source: For 1992­2000, authors' estimates based on the 1992 Rwanda Demographic and Health Survey (DHS); for primary and secondary education in 2000, 2000 Rwanda DHS; for data by province and data for higher education, 2001 Questionnaire unifié sur les indicateurs de développement (QUID) survey. 78 WORLD BANK COUNTRY STUDY FIGURE 4-1: GROSS ENROLLMENT RATIOS IN LOWER AND UPPER SECONDARY EDUCATION BY INCOME QUINTILE, RWANDA, 2000 35 (%) Lower 30 secondary ratio 25 Upper 20 secondary 15 enrollment 10 5 Gross 0 1 2 3 4 5 Poorest Richest quintile quintile Source: Based on results reported in Kline (2002), using data from the 1999­2001 Household Living Condi- tions Survey. A more striking disparity is that between income groups. At the primary level, the disparity-- as measured by the index of the GER of the poorest 40 percent to that of the richest 20 percent--has narrowed over time: in 1992 the index was 0.71, but in 2000 it had risen to 0.78.5 At the secondary level, the GER fell for all income groups, but the decline has been steeper for the lower income groups than for the richest 20 percent. As a result, the GER among the poorest 40 percent relative to that of the richest 20 percent fell from 0.31 in 1992 to only 0.10 in 2000.6 Figure 4-1 shows that the disparity in GERs at the secondary level is in fact largest between the top two and the bottom three quintiles. The pattern is, moreover, similar for the two subcycles within secondary education. At the tertiary level, the disparity between income groups is evident. In particular, the gross enrollment ratio of the poorest 40 percent was only 0.03 times that of the richest 20 percent. Coverage among the middle 40 percent was better but still lagged far behind that of the richest 20 percent. The pattern observed in secondary education, where participation is dominated by the top two quintiles, thus appears to have deteriorated, implying that access to higher education is even more narrowly confined to the most privileged segments of Rwandan society. That the pattern of enrollment in higher education is biased in favor of the better-off is cor- roborated by the data in table 4-2, which compares the occupational backgrounds of the parents of students enrolled at the Université nationale du Rwanda (UNR) and the Kigali Institute of Educa- tion (KIE) with the background of the general adult population ages 35­65.7 Among students with at least one living parent, the data show overrepresentation of students whose parents work as salaried workers (probably in the modern sector, where most salaried jobs are concentrated). 5. In absolute terms, however, the gap in gross enrollment ratios at the secondary level between the poorest 40 percent and the richest 20 percent remained relatively unchanged at around 26 percentage points in 1992 and 28 percentage points in 2000. 6. In 1992 the absolute gap in gross enrollment ratios between the richest 20 percent and the poorest 40 percent was 27 percentage points. The gross enrollment ratio has fallen by similar percentage points across income groups, and so the absolute gap between the top 20 percent and the poorest 40 percent remained relatively unchanged in 2000. 7. The UNR is the largest public higher education institution, accounting for about 56 percent of stu- dents in the public sector in 2000­01. The corresponding share at the KIE was 11 percent. Because of time constraints, similar data could not be collected from the other institutions in the system. The age range of 35­65 was chosen to correspond to the likely age range of the parents of university students. EDUCATION IN RWANDA 79 TABLE 4-2: PERCENTAGE SHARE OF SALARIED WORKERS AMONG PARENTS OF STUDENTS AT TWO PUBLIC HIGHER EDUCATION INSTITUTIONS AND IN THE POPULATION OF RWANDA, 2000 Share of Share of men women ages Fathers of students ages 35­65 in Mothers of students 35­65 in the Occupation UNR KIE the population UNR KIE population Salaried workers 32.4 26.5 16.4 11.6 16.0 5.4 Othersa 67.6 73.5 83.6 88.4 84.0 94.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 Memorandum: Sample size 148 310 2,544 172 338 3,295 Note: UNR, Université national du Rwanda; KIE, Kigali Institute of Education. Students whose fathers are dead are excluded from the "Fathers of students" column; students whose mothers are dead are excluded from the "Mothers of students" column.The share of orphans is somewhat lower among university students than among secondary school students. See table 1-2 in chapter 1 for details on the incidence of orphanhood by level of education. a. Includes farmers, traders, businesspersons, unpaid family workers, and those with unknown occupations. Source: For the UNR and the KIE, random sampling of student files to which the authors of this study were given access in the context of this study; for the occupational distribution of the general population, 1999­2001 Household Living Conditions Survey. Although 16 percent of Rwandan men ages 35­65 held a salaried job in the modern sector in 2000, the corresponding shares among the fathers of UNR and KIE students were, respectively, 2.0 and 1.6 times as large. A similar picture emerges for mothers of the students at the two institu- tions: the share of mothers who worked as salaried employees in the modern sector was 12 percent for UNR and 16 percent for KIE students, compared with the share of 5.4 percent among women ages 35­65 in the population as a whole. Educational Participation Rates among Orphans One of the most devastating legacies of Rwanda's 1994 genocide is the extremely high prevalence of orphanhood. Given the size of the orphan population (see chapter 1), it is important to docu- ment the extent to which orphanhood has compromised the schooling of children. Primary Education As expected, the chances of being in school are higher among children with both parents alive than among those who have lost at least one parent--almost 76 percent, compared with 72 per- cent (table 4-3). The data suggest that the gap between orphans and nonorphans is driven largely by the shortfall in participation for orphaned girls. The participation rate for orphaned boys was less than 1 percentage point lower than for boys with both parents alive, but the corresponding shortfall was more than 5 percentage points for girls. Closer examination of the data suggests that children who had lost their mothers, even if they had not lost their fathers, were most at risk of not attending school. Their participation rate was 71 percent for boys and 64 percent for girls. For children who had lost only their fathers, the rates were 77 percent for boys and 72 percent for girls. Even when both parents were alive, children who lived away from their parents were less likely to attend school than their peers. Their lower participation rate is consistent with the possibility 80 WORLD BANK COUNTRY STUDY TABLE 4-3: PERCENTAGE OF CHILDREN AGES 7­12 ENROLLED IN PRIMARY SCHOOL BY ORPHANHOOD STATUS, RWANDA, 1998­99 Status of child Boys Girls Both sexes Both parents alive 75.3 75.6 75.5 Living with both parents 77 77 77 Living with only one parent or neither 69 73 71 At least one parent dead 74.6 70.4 72.4 Mother dead; father alive 71 64 67 Father dead; mother alive 77 72 74 Both parents dead 70 68 69 Whole sample 75.0 73.5 74.3 Children not living with biological parents 69.8 68.1 68.8 Memorandum: Sample size 1,466 1,548 3,014 Source: Authors' estimates based on the 2000 Rwanda Multiple Indicator Cluster Survey (MICS). See appendix table A4-1 for additional information. that some of them worked at a job and could not find the time to attend school or were simply surviving as street children with little or no adult encouragement to attend school. The overall picture that emerges is that although most children in Rwanda today attend pri- mary school, there remain pockets of the population who do not. Orphans, especially those who have lost their mothers and those living apart from their biological parents, account for most of the out-of-school population. Secondary and Higher Education Table 4-4 provides additional evidence that orphaned children are more vulnerable to nonpartic- ipation than nonorphans. To see this, we compare the share of orphans in the population with the shares of orphans in primary, secondary, and higher education. The data on orphans' shares of enrollments in primary and secondary education are based on a census of the children enrolled in schools associated with the Conseil protestant du Rwanda (CPR), an umbrella organization of schools run by various churches that in 2000­01 served about 20 percent of primary school children and some 16 percent of secondary school students nationwide. For higher education, the data are based on a sample of student records maintained by two of the country's public institutions. Although the data are patchy, they reveal some interesting patterns. First, they confirm that primary school participation rates tend to be lower among orphans: the share in the population of children ages 7­12 who had lost at least one parent was 38 percent in 2000, but the correspon- ding share among children enrolled in CPR primary schools was 31 percent. At the secondary level, the pattern is reversed, with the share of orphans rising to 41 percent instead of falling, as might have been expected. The reason is that many of these children have been specifically targeted for assistance under Rwanda's Genocide Fund. At the tertiary level, and assuming that orphanhood rates among youths are similar to the rates for children ages 7­12, orphans appear to be underrepresented in the two institutions sampled. Overall, orphans accounted for 32 percent of total enrollment in the UNR and 27 percent in the KIE. These proportions are lower than the share of orphans among students in secondary school. EDUCATION IN RWANDA 81 TABLE 4-4: PREVALENCE OF ORPHANHOOD AMONG PRIMARY, SECONDARY,AND HIGHER EDUCATION STUDENTS, RWANDA, CIRCA 2000 % with one parent dead % with Total, one both %with one Mother Father parent parents or both Sample Population group dead dead dead dead parents dead size Children ages 7­12, 2000a 4.6 25.8 30.4 7.1 37.6 3,426 Schools associated with the Conseil protestant du Rwanda, 2001b Primary pupils -- -- 22.2 9.1 31.3 497b Secondary students -- -- 25.2 15.5 40.7 68b Higher education students, 2001c Université Nationale du Rwanda (UNR) 3.9 15.9 19.4 13.0 32.4 207 Kigali Institute of Education (KIE) 4.5 11.6 16.1 10.6 26.6 398 --Not available. a. Based on the 2000 Rwanda Multiple Indicator Cluster Survey (MICS). b.The Conseil protestant du Rwanda is an umbrella organization consisting of schools run by various church organizations. In 2001 it took a special census of the 497 primary and 68 secondary schools belonging to it. One purpose was to ascertain the extent of orphanhood among the 292,258 primary and 26,580 secondary students enrolled in these schools.The results were made available in aggregate form to the authors of this study. c. Based on information supplied to the authors by officials at the two institutions in the context of this study for a random sample of 207 students at the UNR and 398 students at the KIE. Source: Authors' estimates based on sources given in the notes above. The lower participation rates among orphans are consistent with the increased opportunity cost of education as students progress up the educational ladder. Closer examination of the data for higher education reveals an interesting difference between youths who have lost one parent and those who have lost both. Although participation rates among the latter are lower, the drop-off from the corresponding shares in secondary education appears to be less steep among the double orphans. This pattern is consistent with differences in the availability of student aid that gives more support to double orphans. Although the more detailed data necessary for exploring this issue in greater detail are lacking, it seems safe to con- clude that although current policies are already helping orphans gain access to higher education, they could probably be fine-tuned to ensure that disadvantaged youths who have lost "only" one parent instead of both are not overlooked. Disparities in Student Flow Patterns Apart from the disparities in enrollment and participation rates documented above, it is useful to examine differences in student flow across socioeconomic groups. This information is relevant for policy development because it helps identify the locus of the observed disparities in enrollment. Because of limitations in the data, we examine student flow only in primary schooling. 82 WORLD BANK COUNTRY STUDY TABLE 4-5: SELECTED STUDENT FLOW INDICATORS BY GENDER, LOCALITY,AND INCOME GROUP, RWANDA, CIRCA 2000 % of last year's % of cohort ever % of first grade entrants sixth-graders still entering grade 1a surviving to grade 6b enrolled in current yearc Group % Index % Index % Index By gender Boys 86 1.00 74 1.00 68 1.00 Girls 88 1.02 70 0.95 73 1.07 By locality Urban 93 1.00 91 1.00 97 1.00 Rural 87 0.93 70 0.77 64 0.66 By incomec Richest 20% 95 1.00 93 1.00 89 1.00 Middle 40% 89 0.93 68 0.74 65 0.73 Poorest 40% 83 0.87 64 0.69 52 0.58 a. Computed as the average of the percentages of children ages 10­13 who have ever enrolled in school. b. Computed by multiplying the grade-to-grade progression rates after taking into account grade-specific repetition rates. c. Defined in the same way as in table 4-1. See note b in that table. Source: Authors' calculations based on the 1999­2001 Household Living Conditions Survey. Differences by Gender, Locality, and Income Table 4-5 shows three indicators of student flow: the percentage of an age cohort ever entering grade 1, the percentage of entrants who reach the end of the primary cycle, and the share of the previous year's sixth-graders who were still enrolled in the current school year, either to repeat the grade or to continue to secondary school. Consistent with the data presented earlier, the differences between boys and girls are modest: entry rates are comparable, and girls are at a slight advantage in reaching the end of the cycle and remaining enrolled after grade 6. The differences between urban and rural populations widen progressively during a child's schooling career: the entry rate to grade 1 among rural children is 93 percent as high as that for urban children, but the survival rate to grade 6 is only 77 percent as high, and the continuation rate at the end of grade 6 is only 66 percent as high. Differences between children in rural and urban localities reflect, to some extent, differences among income groups. Only 83 percent of children in the poorest 40 percent of households enter grade 1, but 95 percent of those in the richest 20 percent do. The gap widens as children move to higher grades. Only 64 percent of children in the poorest 40 percent of households survive to grade 6, compared with 93 percent among those in the richest 20 percent. Only 52 percent of children in the poorest 40 percent of households remain enrolled after attending grade 6 the previ- ous year, but the share is 89 percent among children in the richest 20 percent. Figure 4-2 clearly illustrates that while children from the poorer households lag somewhat behind richer children in starting grade 1, the gap becomes much wider by the end of the primary cycle because of differ- ential rates of dropping out within the cycle. Further selection at the end of the primary cycle reinforces the disparities, producing the gaps in secondary school participation documented earli- er in the chapter. EDUCATION IN RWANDA 83 FIGURE 4-2: PRIMARY SCHOOL ENROLLMENT RATES IN A COHORT OF CHILDREN ACROSS INCOME GROUPS, RWANDA, 2000 100 (%) Middle 40% 75 enrolled Top 20% 50 cohort of Bottom 40% 25 Percentage 0 Grade 1 Grade 6 Note: The term cohort here refers to a composite population whose schooling career reflects the pattern of promotion, grade repetition, and dropout over two school years in a cross-sectional sample of children. Source: Based on data in table 4-5. Regression Estimates of the Correlates of Student Flow Patterns Given the importance of locality, income, and orphanhood status in determining children's schooling careers, it is useful to examine the relative influence of these factors using regression analysis.8 For this purpose, we define four individual-level indicators of student flow among chil- dren ages 7­12: (a) whether a child who was not yet in school the previous year has entered grade 1 in the current school year; and (among those already enrolled last year) whether the child, in the current year, has (b) advanced a grade, (c) remained in the same grade, or (d) dropped out.9 The regression estimates are reported in appendix table A4-3, and the results based on them are shown in table 4-6. These results are expressed in terms of the probability of a child's being in one of the four schooling statuses defined above. For example, among children ages 7­12 who were not in school in 1999, the probability of entering grade 1 in 2000 is 0.52 for the sample as a whole. It is 0.13 points lower for children who have lost both parents relative to children with both parents alive, after controlling for gender of the child, urban-rural residence, and income group. The difference for this estimate is statistically significant at the 5 percent confidence level or better. None of the other estimates is statistically different from zero, suggesting that the sin- gle most important predictor of nonentry to grade 1 is whether the child has lost both parents. Double orphans are handicapped right at the outset of their schooling career, in that many of them do not even get started. For those who were already enrolled in the previous year, the probability of advancing a grade is greater by 0.12 points among children whose mothers have died than for children with 8. The advantage of this approach is that it allows us to evaluate the impact of a single factor while con- trolling for differences in other dimensions. 9. Use of these new indicators of student flow is necessitated by the reliance in this analysis on individual- level data, whereas the flow indicators computed earlier are estimates for population cohorts based on cross- sectional patterns. See also appendix table A4-2 for tabular information on the differences between orphans and nonorphans in the indicators of student flow considered here. 84 WORLD BANK COUNTRY STUDY TABLE 4-6: REGRESSION-PREDICTED INDICATORS OF SCHOOL PROGRESSION DURING TWO CONSECUTIVE YEARS IN A COHORT OF CHILDREN WHO WERE AGES 7­12 IN 1998, BY POPULATION GROUP, RWANDA, 1998­2000 Probability of being in the indicated schooling status in 1999­2000b Among those already enrolled in 1998­99 Reference Population Sample New entrant Advanced Repeating Dropped group group meana to grade 1 a grade same grade out Whole sample 0.52 0.64 0.33 0.03 Girls Boys 0.49 0.05 0.00 0.00 0.00 Both parents Both parents alive dead 0.08 -0.13* 0.05 -0.09* 0.04* Mother dead; father alive 0.05 -0.13 0.12* -0.12* -0.01 Father dead; mother alive 0.27 -0.01 0.03 -0.02 -0.01 Living with only one parent or neitherc 0.14 -0.07 0.03 -0.03 0.00 Urban resident Rural resident 0.81 -0.03 -0.08* 0.07* 0.01 Poorest 40% Top 20% 0.21 0.02 0.06* -0.04 -0.02* in income Middle 40% 0.42 0.03 0.01 0.01 -0.01 * Statistically significant at the 5 percent confidence level or better. a. Refers to the sample means for those not yet enrolled in 1998­99 among children who were ages 7­12 in 1998.The means are almost identical in the population already enrolled in that year and are therefore not shown separately. b.The figures for the "Whole sample" row refer to the overall probability of being in the indicated status.The other figures refer to the difference in probability between the indicated population group and the correspon- ding reference group. c.Among those with both parents alive. Source: Based on regression estimates in appendix table A4-3. both parents still alive. For the other categories of already-enrolled children, the probability of progressing to the next grade also exceeds that in the reference group, but the difference is not statistically significant. For children who do not advance a grade, the alternatives are to repeat the grade or to drop out. Relative to nonorphans, children who have lost both parents are less likely to repeat and more likely to drop out, whereas among those who have lost their mothers but not their fathers, the probability of repeating is lower (a corollary of the fact that they are more likely to advance a grade) and that of dropping out is about the same. For the remaining children, the estimates are not statistically significant relative to nonorphans. Taken as a whole, the results are consistent with the expectation that when given a chance to go to school, children in highly vul- nerable circumstances--especially those who have lost their mothers or both parents--are likely to be more motivated to succeed than other children. They overcome the enormous odds against them and manage to draw even with or surpass nonorphans in their progression between grades. EDUCATION IN RWANDA 85 But the prospects are unforgiving for the most vulnerable children--the double orphans. Those in this group who fail to advance are more likely to leave the system than to return for a second try the following year. Controlling for orphanhood status, the regression results confirm the patterns found in the cross-tabulations presented earlier. Boys and girls are equally likely to advance a grade, repeat, or drop out. By contrast, locality does make a statistically significant difference, holding back rural children by 0.08 points in terms of the probability of advancing a grade and elevating their chances of repetition by 0.07 points. Household income also makes a difference, but only between children from the richest 20 percent of households and the rest of the population.10 Compared with those in the bottom 40 percent, the chances for children from the most privi- leged homes of advancing a grade are higher by 0.06 points, and the probability of dropping out is lower by 0.02 points. In relation to the overall probabilities of advancement, repetition, and dropping out shown in the table, these gaps are quite large. Distribution of Public Spending on Education Disparities in enrollment and student flow have implications for the distribution of public spend- ing on education because only those who attend school benefit. We examine the issue from two complementary perspectives. The first captures the incidence of spending associated with the system's structural biases; the second measures the incidence of spending associated with cross- sectional differences in school participation across income groups. Spending Pattern Associated with Structural Biases For a given cohort of children, it is possible to determine the number of children who do not go to school at all and the number who will terminate their schooling at each level of education.11 Children with no schooling receive no benefit from public spending on education, whereas those who attain higher education accumulate both the spending associated with higher education itself and that associated with the levels they have already completed.12 The distribution of spending on education in a cohort thus depends on the structure of enrollments, as well as the distribution of spending across levels of education. The degree of inequity in the distribution can be portrayed using a Lorenz curve, which shows the cumulative shares of the cohort by educational attainment on one axis and the shares of resources benefiting each group on the other axis. The degree of 10. The income effect is likely to be confounded by our inability in the regression analysis to control for systematic differences in the characteristics of the adoptive families of orphaned children and those of other families. 11. To illustrate, suppose the gross enrollment ratio for a country is 80 percent in primary education, 30 percent in secondary education, and 10 percent in higher education. In a cohort of children passing through the system, the distribution of eventual school attainment would be as follows: 10 percent would have attained higher education; 20 percent (= 30 - 10), secondary education; and 50 percent (= 80 - 30), pri- mary education. The remaining 20 percent (= 100 - 80) would have had no schooling. More refined meth- ods can be used to estimate the desired distribution (for example, using grade-by-grade survival patterns), but sensitivity tests suggest that calculations based on gross enrollment ratios produce reasonably robust estimates for our purpose here. For countries such as Rwanda, where the gross enrollment ratio for primary education exceeds 100 percent, the calculations cap the GER at 100 percent. 12. Suppose, for example, that the distribution of current spending is such that primary education receives 55 percent, secondary education receives 30 percent, and higher education receives the remaining 15 percent. The cumulative spending would be 100 units (= 55 + 30 + 15) for those attaining higher educa- tion, 85 units (= 55 + 30) for those attaining secondary education, 55 units for those attaining primary education, and zero for those who did not go to school. The distribution of cumulative spending in a cohort passing through the education system would thus be 23 percent [= 55/(55 + 85 + 100)] for those attaining primary education, 35 percent [= 85/(55 + 85 + 100)] for those attaining secondary education, and 42 percent [= 100/(55 + 85 + 100)] for those attaining higher education. 86 WORLD BANK COUNTRY STUDY FIGURE 4-3: SHARE OF CUMULATIVE PUBLIC SPENDING ON EDUCATION BENEFITING THE 10 PERCENT BEST EDUCATED IN A GENERATION, RWANDA AND OTHER SUB- SAHARAN AFRICAN COUNTRIES, CIRCA 1998 Niger Chad Rwanda Mali Madagascar Mozambique Sierra Leone Ethiopia Senegal Malawi Mauritania Comoros Burkina Faso Zambia Lesotho Swaziland Eritrea Gambia, The Cote dÕIvoire Guinea Zimbabwe Mauritius Congo, Rep. Morocco Namibia Benin Togo South Africa 0 25 50 75 100 Percentage share of cumulative spending Note: Data for Rwanda refer to 2001. Source: BREDA,World Bank, and UIS (2002), supplemented by estimates for Rwanda prepared by the authors using the same computational methods. See the text for additional information. inequity can be quantified by calculating the corresponding Gini coefficient.13 For our purpose, we use an even simpler summary statistic: the share of the resources benefiting the 10 percent best educated in a cohort.14 The results for Rwanda appear in figure 4-3, along with estimates for a large number of low- income countries for which appropriate data from around 1998 are available. Although cross- country comparisons should be drawn with caution because of differences in data quality and 13. If we continue to use the same numerical examples as in the preceding two notes, the distribution of the educational attainment of the cohort and the corresponding shares of cumulative public spending on education for the cohort would be as follows: 20 percent with no schooling and with zero share of the cumulative resources; 50 percent attaining primary schooling and receiving 23 percent of the resources; 20 percent attaining secondary education and receiving 35 percent of the resources; and 10 percent attaining higher education and receiving 42 percent of the resources. The Gini coefficient associated with this distri- bution is 0.57 on a scale that ranges by definition from 0 (each person obtains an exactly proportionate share of resources) to 1.0 (one person gets all the resources). See appendix figure A4-1 for a graphic repre- sentation of this hypothetical example. 14. In cases in which the share of the cohort attaining higher education is less than 10 percent, the best educated 10 percent would include some who attained secondary education. The share of resources benefit- ing this group would simply be the sum of the share of those who attained higher education and the prorat- ed share of those who attained secondary education. In cases in which the share of the cohort attaining higher education exceeded 10 percent, the share of resources benefiting the 10 percent best educated would simply be the prorated share of the resources benefiting the group that attained higher education. EDUCATION IN RWANDA 87 coverage and in the structure of the systems, the results support two conclusions. First, in all countries the best educated always receive more than their population share of public spending on education; this pattern is to be expected, given that governments typically subsidize all levels of education. Second, very large differences exist across countries in the degree of bias in favor of the better educated. The share of cumulative resources benefiting the 10 percent best educated in a cohort ranges from about 40 percent in South Africa to about 80 percent in Niger. Rwanda is located near the top end of the spectrum, just behind Niger and Chad among the 28 countries included in the figure. Its position is consistent with what one would expect in a system in which higher education enrolls less than 2 percent of the population in the relevant age group and claims nearly 40 percent of the government's total current spending on education. To anticipate a point made later in this section, note that Rwanda's position is only slightly worse than that of Madagascar, a country with comparable per capita GDP. The implication is that the enrollment structure and public subsidization in these countries are highly comparable in their biases in favor of the best educated in a cohort.15 Spending Pattern Associated with Social Selectivity in Education A second perspective on the distribution of public spending on education is the extent to which it favors the wealthier groups in society. Benefit-incidence analysis is a common tool for document- ing this aspect of inequity in spending patterns. The approach involves using cross-sectional data on school participation patterns by income group and information on public spending per stu- dent to calculate the share of resources benefiting households in each income group.16 As we have seen earlier, wealthier groups in Rwanda become increasingly overrepresented as the level of education rises--a pattern found in practically all countries. Benefit-incidence analysis typically shows that although public spending on primary education may benefit lower-income groups more than wealthier groups, the pattern is often not sustained in secondary education and is usually completely reversed in higher education. The aggregate impact, taking into account all levels of education, usually shows the wealthier groups receiving more than their proportionate share of public spending on education. The real issue is one of degree. Table 4-7 summarizes the results of benefit-incidence analysis for Rwanda in 2000, along with those for nine other countries in Sub-Saharan Africa for which similar calculations have been made. The analysis shows that the poorest 20 percent of households in Rwanda accounted for 15 percent of public spending on primary and secondary education, while the richest 20 percent of the population received 28 percent. Judging by the relative sizes of these shares, Rwanda stands somewhere in the middle among the Sub-Saharan African countries shown in the table. In other words, the country's public sector spending on education is neither as equitable across income groups as it could be, nor is it as inequitable as in some other countries. A comparison between the results for Rwanda and those for Madagascar is particularly inter- esting in light of the findings highlighted earlier regarding the similarity of the structural biases in both systems in favor of the best educated in a cohort. Yet the data in table 4-7 indicate that Rwanda has a more equitable distribution of educational spending across income groups than does Madagascar: in Rwanda the richest quintile appropriates a share nearly twice as large as the share received by the poorest quintile, but in Madagascar the richest receive more than five times as much as the poorest. The more favorable results in Rwanda are consistent with the country's smaller gaps in educational access across income groups. For example, among first-graders from the poorest 40 percent of households, 69 percent reach the end of the primary cycle in Rwanda, 15. In 1998 the gross enrollment ratio in Madagascar was 107 percent in primary education, 16 percent in secondary education, and 2 percent in higher education; the corresponding ratios in Rwanda in 2000 were 107, 12, and 1.4 percent. The distribution of current spending on education in Madagascar in 1998 was as follows: 51 percent for primary education, 33 percent for secondary education, and 16 percent for higher education. The corresponding shares in Rwanda in 2001 were 45, 18, and 37 percent. 16. For further details on the method of calculation, see Demery (2000). 88 WORLD BANK COUNTRY STUDY TABLE 4-7: SHARE OF PUBLIC SPENDING ON EDUCATION BENEFITING THE POOREST AND RICHEST POPULATION QUINTILES, RWANDA, 2000,AND SELECTED SUB-SAHA- RAN AFRICAN COUNTRIES, 1990S Richest quintile's share Year of Poorest Richest as multiple of poorest Country survey quintile (%) quintile (%) quintile's share South Africa 1993 21.1 23.4 1.1 Kenya 1992­93 16.7 20.7 1.2 Ghana 1992 16.4 20.8 1.3 Malawi 1994­95 16.0 25.0 1.6 Rwandaa 2000 15.0 28.0 1.9 Uganda 1992 13.0 32.0 2.5 Côte d'Ivoire 1995 13.5 34.8 2.6 Tanzania 1993 13.0 38.0 2.9 Guineaa 1994 8.5 26.9 3.2 Madagascara 1997 7.0 36.0 5.1 a. Data refer to the distribution of spending on primary and secondary education only. Source: For Rwanda, Kline (2002), based on the 2000 Household Living Conditions Survey; for Madagascar, Madagascar (2000); for other countries,World Bank EdStats database based on various household surveys on living standards. compared with only 9 percent in Madagascar.17 These results imply that in Rwanda it would appear especially important to reduce the education system's structural biases favoring the best educated, even as efforts are made to close the socioeconomic gaps in educational access. Policy Implications In developing appropriate policies to address socioeconomic disparities in education, answers are needed to at least three questions: Who are the disadvantaged? Where in the education system might the government intervene to narrow the disparities? And what would be the most effective interventions? The data presented in this chapter provide some insights on these questions, espe- cially with regard to the first two. Narrowing the Disparities in Primary Education The gaps across socioeconomic groups in Rwanda are not as wide as those observed in other low- income countries. Girls are as likely as boys to be in school; rural children are 85 percent as likely to attend as their urban counterparts; and children in the poorest 40 percent of households are 78 percent as likely to attend as those in the top 20 percent of households. Except among double orphans, entry rates to grade 1 are generally high, so the pattern of underrepresentation stems mainly from differences in survival rates to the end of the cycle. The populations falling farthest behind in this regard include double orphans, children in rural areas, and children from the poor- est 40 percent of households. Double orphans--arguably the country's most vulnerable children--are easily identifiable, and systematic efforts can and must be made to improve their life chances through better access to primary education. Providing financial assistance to such chil- dren is one possibility, but the Genocide Fund--the program enacted by law in 1998 to assist 17. For the pattern in Rwanda, see table 4-5; for that in Madagascar, see World Bank (2002a). EDUCATION IN RWANDA 89 children orphaned by the genocide--currently benefits only students in secondary and higher education. The government also provides assistance to support the education of vulnerable chil- dren through the Ministère de l'administration local et des affaires sociales (MINALOC), but the amount of funding is relatively modest--about 20 percent of the resources available through the Genocide Fund. Financial assistance may also be what is needed to boost survival rates among children in the other lagging groups, particularly those from poor families. In rural areas, raising survival rates may require additional interventions on the supply side, such as ensuring that all primary schools offer the full six grades of instruction in the primary cycle, reducing the distances that children have to travel to reach their schools, and, most important, improving teaching and learning to reduce grade repetition and hence dropout.18 It was not possible to evaluate the relative impact of these promising options in the context of this study, but they clearly need to be explored, through pilot projects and other means, as the next step in policy development. Narrowing the Disparities in Secondary and Higher Education As we have seen, socioeconomic disparities begin to widen substantially at the postprimary levels. For girls, the disadvantage in access emerges only in higher education, but for children in rural areas and from poor households, the barriers to access are already obvious in the first cycle of secondary education. The disparities across provinces also begin to appear in secondary educa- tion, with lags for Gisenyi and Gikongoro, in particular. Orphans are somewhat less likely than nonorphans to participate in higher education but are at least as likely as nonorphans to enroll in secondary education--the level that currently benefits the most from the Genocide Fund. Target- ing financial aid to assist the most vulnerable youths thus appears to be working. Additional inter- ventions would be needed to improve educational access in rural areas, in certain provinces, and among the poorest segments of Rwandan society. Where in the education system might interventions be made to narrow the observed dispari- ties in postprimary education, and what interventions would be most effective? Although a full answer is beyond the scope of this study, it is instructive to consider the specific situation of females' lagging participation in higher education. Table 4-8 shows the scores of males and females in the national examinations administered at the end of the primary, tronc commun (lower secondary), and upper secondary cycles and the scores of applicants to public institutions of higher education. A startling pattern is that girls con- sistently perform worse than boys, particularly in the primary and tronc commun cycles. Even though girls continue to be as well represented as boys up to the upper secondary cycle, their persistently inferior scores take their toll in the competition for the highly coveted places in public higher education. Girls currently account for only about a quarter of the students in higher edu- cational institutions in the public sector, even though they make up half the students in the final year of upper secondary education. These results imply that policies to improve women's repre- sentation in higher education must do more than simply reserve a quota for them. Understanding why girls' academic performance lags behind that of boys, beginning in the primary grades, is a critical first step toward developing sound policies to improve in a meaningful way their access at the top end of the educational ladder. Although data that might illuminate the reasons for the lagging participation among other population groups are lacking, poor academic performance is a likely root cause of their prob- lems, as well. Accordingly, enhancing learning outcomes among the affected populations--rural children and those from poor families--is a task that cannot be ignored in developing a strategy to diminish socioeconomic disparities in postprimary education. 18. About 14 percent of Rwanda's primary schools currently do not offer all six grades of instruction in the primary cycle (see table 2-10). About half of all rural households are more than 30 minutes away from a primary school (see table 5-4 in chapter 5). 90 WORLD BANK COUNTRY STUDY TABLE 4-8: GENDER DIFFERENCES IN PERFORMANCE ON NATIONAL END-OF-CYCLE EXAMINATIONS, RWANDA, CIRCA 2002 Female Male Primary through secondarya Primary (% exceeding cutoff mark for promotion to next level) 17.8 28.5 Tronc commun (% exceeding cutoff mark for promotion to next level) 29.3 55.6 Upper secondary (% exceeding pass mark for graduation) 62.8 76.0 Higher education GPA of applicants to public institutionsb 3.4 3.6 GPA, by institution named by applicant as first choice for placement Université nationale du Rwanda (UNR) 3.5 3.7 Kigali Institute of Science,Technology and Management (KIST) 3.2 3.6 Kigali Institute of Education (KIE) 3.2 3.5 Kigali Health Institute (KHI) 3.6 3.9 Institut supérieur d'agronomie et d'élevage (ISAE) 2.9 3.8 a. Data refer to the results of those who took the examination in 2002. b. GPA refers to the grade point average achieved by the students on six national examinations taken at the end of the upper secondary cycle.The scores refer to intake to higher education in 2001­02. Source: For the data on primary and secondary education, National Examination Council of Rwanda; see also appendix table A4-4. Data on higher education were supplied to the authors by officials at individual institutions. Reducing Overall Inequities in Spending Allocations The socioeconomic disparities in enrollment across income groups inevitably show up in the results of benefit-incidence analysis. In Rwanda the richest 20 percent of households garners nearly twice as large a share of public spending on education as the poorest 20 percent. Although there is obvi- ously still scope for increasing the benefit to the poor, cross-country comparison suggests that Rwanda does not have the most inequitable distribution of spending across income groups among the countries sampled. Far more inequitable is the country's overall structure of spending: primary education receives only 45 percent of the government's current spending on education, whereas higher education gets nearly 40 percent. Redirecting spending in favor of primary education would go a long way toward reducing the current bias in the system in favor of those who manage to climb to the top of the educational ladder. To the extent that the task involves supply-side changes in spending policy, it should be possible for the government to act quickly and effectively. Appro- priate policies on student finance and cost management in higher education are particularly critical in this regard, and these are discussed in greater detail in chapter 7. Conclusion Educational access in Rwanda is relatively equitable in primary education, but disparities across population groups widen dramatically at postprimary levels. Orphans are among the most vulner- able population groups, and the country has admirably institutionalized financial assistance to them through the Genocide Fund. More can be done, however, to reach pockets of children still at risk of lagging behind, particularly double orphans, children in rural areas, and children from the poorest families. Aside from these priorities, success in improving overall equity in the system will require correction of the current imbalance in public spending that so obviously favors higher education at the expense of primary education. CHAPTER 5 SERVICE DELIVERY IN PRIMARY EDUCATION I n Rwanda, as in most other low-income countries, primary education enrolls the largest num- ber of students and absorbs a major share of public spending on education. Because of its central role in basic human capital formation, the performance of the primary sector attracts keen attention from policymakers, as well as the public. This chapter examines the availability of services in primary education, how well existing resources are deployed to provide services, how efficiently the resources are transformed into learning outcomes, the scope for improvement, and the implications for policy development. Although data constraints limit exploration of these topics, the available information presented in this chapter nonetheless helps shed light on the nature of some of the emerging challenges. Overview of the Supply of Services We begin by describing a few salient features of the country's network of primary schools. In this chapter, the school is the unit of analysis.1 Supply Infrastructure Rwanda currently has slightly more than 2,000 primary schools, serving some 1.5 million chil- dren (see table 5-1). The private sector accounts for only 1.5 percent of the schools and 0.7 per- cent of the enrollments. The public sector consists of two types of school: state and libre subsidié. State schools are funded and managed directly by the government; libre subsidié schools are fund- ed by the government but are run by nongovernmental organizations under two main umbrella groups: the Secrétariat nationale de l'enseignement catholique (SNEC) and the Conseil protes- tant du Rwanda (CPR). More than 71 percent of all primary schools fall into the libre subsidié category, and about two-thirds of these are under SNEC management. 1. Unless otherwise indicated, the averages presented in the tables in this chapter refer to unweighted rather than weighted values. 91 92 WORLD BANK COUNTRY STUDY TABLE 5-1: PERCENTAGE DISTRIBUTION OF PRIMARY SCHOOLS AND PUPILS BY SECTOR, RWANDA, 2000 Public sectora State Libre subsidié Private sector Overallb Schools 27.1 71.4 1.5 100 (2,093) Pupils 28.0 69.4 0.7 100 (1,475,572) a. The public sector consists of schools whose staff salaries and other operational costs are paid by the gov- ernment. State schools are run directly by the government; libre subsidié schools are mostly managed by religious organizations.The percentage distribution is based on 2,043 schools for which information on sector affiliation is available. b. Numbers in parentheses refer to total numbers of schools and pupils. Source: Ministry of Education 1999­2000 census of primary schools. FIGURE 5-1: INSTITUTIONAL COMPOSITION OF PRIMARY SCHOOLS BY PROVINCE, RWANDA, 2002 Source: Ministry of Education 1999­2000 census of primary schools supplemented by data from the Conseil protestant du Rwanda (CPR) and the Sécretariat nationale de l'enseignement catholique (SNEC). There is wide diversity across provinces in the institutional composition of the supply infra- structure (see figure 5-1). Almost all of the country's private schools are located in Kigali Ville; Cyangugu and Kibuye have no private schools. As a result, private schools account for nearly 20 percent of the schools in Kigali Ville, compared with the national average of less than 1.5 percent. Provinces also vary in the relative shares of state and libre subsidié schools. In Gikongoro state EDUCATION IN RWANDA 93 TABLE 5-2: CHARACTERISTICS OF STATE, LIBRE SUBSIDIÉ, AND PRIVATE PRIMARY SCHOOLS, RWANDA, 2000 Public sector State Libre subsidié Private All schools Average number of pupils per school 720 678 342 684 Schools with classrooms in poor condition (%) 47.9 54.8 19.9 52.3 Schools with teachers teaching two shifts (%)a Grade 1 94.2 96.0 24.1 94.5 Grade 2 92.8 94.9 24.1 93.4 Grade 3 84.4 88.1 21.4 86.2 Pupil-teacher ratiob 59.7 56.9 33.3 57.3 (56.0) (55.3) (32.8) (55.2) Schools with nonteaching staff (%)c 44.5 41.0 16.1 41.6 Ratio of pupils to nonteaching staffd 725.7 710.5 708.6 714.8 a. Double shifts are not used after grade 3. b. Figures in parentheses refer to weighted averages. c. Counting only school heads without teaching duties, and school secretaries. d. Refers to unweighted ratios in schools with administrative staff. Source: For data on the first four indicators, authors' calculations based on Ministry of Education 1999­2000 census of primary schools; for data on the last two indicators, authors' estimates based on Ministry of Education 1999 census of teachers. schools account for less than 2 percent of the schools, but in Umutara their share exceeds 65 percent. Within the libre subsidié sector, the share of schools under SNEC management ranges from 81 percent in Butare to 39 percent in Umutara. Selected Characteristics of Schools and Teachers Table 5-2 summarizes some features of the supply of primary schooling in Rwanda today. The country has very few private primary schools, and they are strikingly different from the public schools. Average private school enrollment is only about half that of public schools. The private schools operate in better premises; they generally run single shifts in all grades (unlike public schools, where the first three grades generally operate on double shifts); and they have much smaller ratios of pupils to teachers. In the public sector, state and libre subsidié schools are highly comparable, although state schools tend to be slightly bigger (720 pupils, on average) than libre subsidié schools (678 pupils). Almost all first grade classes, whether state or libre subsidié, are taught by teachers with double-shift duties, but by third grade the share of double shifts declines to about 84 percent in the state sector and 88 percent in the libre subsidié sector. By administrative arrangement, there is no double shifting beyond third grade.2 Because of the prevalence of double shifting, Rwanda's average pupil-teacher ratio is currently one of the highest among African countries (see figure 5-2). Libre subsidié schools are better off than their state counterparts in this regard, but the advantage is hardly dramatic. A smaller share of libre subsidié schools than of state schools has nonteaching staff (school heads without teaching duties, and secretarial staff): 41 percent, compared with 44 2. Under double shifting, children attend school either in the morning or in the afternoon, typically for about 2.5 to 4.0 hours, depending on the shift. In some schools, children change shifts on alternate days; in others they do so in alternate weeks. 94 WORLD BANK COUNTRY STUDY FIGURE 5-2: PUPIL-TEACHER RATIOS IN PUBLIC PRIMARY SCHOOLS IN RWANDA AND OTHER SUB-SAHARAN AFRICAN COUNTRIES, CIRCA 1999 Source: For Rwanda, table 5-2; for the other countries,World Bank (2002b). percent of state schools. Among private schools, 16 percent have staff who perform purely administrative tasks. In schools with administrative staff, the staffing ratios appear to be relatively comparable across the three types of school. Table 5-3 presents information on the profiles of teachers in the various types of school. In private schools about half the teachers are women; this proportion is comparable to that in state schools and is slightly less than the share in libre subsidié schools. Teachers in private schools are the most seasoned, with an average of 8.5 years of experience, compared with about 8 years in the other two sectors. They are also the best educated: 77 percent have some sort of upper sec- ondary education, and 47 percent have a D6 or D7 diploma, the best credential at this level of education. (The share of teachers with a D6 or D7 qualification is 33 percent in state schools and 36 percent in libre subsidié schools.) A significantly higher share of private school teachers has received preservice teacher training: 67 percent, compared with 50 and 54 percent in state and libre subsidié schools, respectively. Accessibility and Services: Client Feedback To round out the discussion on the supply of services, we turn to the perspective of the service users. The 2001 Questionnaire unifié sur les indicateurs de développement (QUID) survey of about 5,800 households provides some clues regarding the accessibility of primary schools and pupils' perceptions of the services they receive (see table 5-4). At the primary level, schools should ideally be within walking distance of pupils' homes, given the usually young ages of the children. According to QUID data, only about half the sam- pled households live within 30 minutes of a primary school, but in urban areas the share rises to more than 80 percent. Across provinces, accessibility is best in Kigali Ville, Ruhengeri, and Butare, where between 58 and 74 percent of the households live within 30 minutes of a primary EDUCATION IN RWANDA 95 TABLE 5-3: CHARACTERISTICS OF PRIMARY SCHOOL TEACHERS, RWANDA, 1999 Public sector Libre Private All Teacher credentiala State subsidié schools schoolsb Proportion of women (%) 50.3 54.6 50.0 54.0 Average years of experience 8.1 8.0 8.5 8.1 Distribution by educational attainment (%) Primary CA 1.4 1.3 1.4 1.3 Lower secondary 27.5 23.8 21.6 24.9 General (1­3 years) ES 1, 2, 3 9.1 7.3 5.4 7.8 Vocational diploma holder CERAI, etc. 15.3 12.6 13.5 13.4 Teacher training diploma holder EAP, EMA, etc. 3.1 3.9 2.7 3.7 Upper secondary 70.7 74.6 77.1 73.4 Incomplete ES 4, 5, 6 21.0 20.5 17.6 20.5 Diploma holder (3­5 years)c D3, D4, D5 16.8 18.3 12.2 17.9 Diploma holder (6­7 years)c D6, D7 32.9 35.8 47.3 35.0 Total 100 100 100 100 Number of teachers 6,390 15,653 74 24,982 Share with preservice teacher training (%) 49.6 54.4 66.5 53.2 Blanks denote not applicable. a. See appendix table A5-1 for more details on teachers' educational qualifications. b. Includes teachers for whom data on the type of school where they teach are not available. c. Diploma holders with 3­5 years of secondary schooling refers to those in qualification category D3­D5; those with 6­7 years refers to category D6­D7.Teachers in the first category are those who received their diplomas before the reforms in 1982 that introduced the six-year secondary cycle still in effect today. Source: Authors' calculations based on Ministry of Education 1999­2000 census of schools. school, and is worst in Gitarama, Kibungo, and Cyangugu, where the corresponding share is around 45 percent. These results suggest that the accessibility of primary schools, particularly in the lagging provinces, needs to be improved. We return to this issue in more detail later in this chapter, in the context of the discussion of economies of scale in service provision. Almost half (49 percent) of the pupils surveyed about their perceptions of their schooling reported no problems. Among those who reported a problem, the overwhelming majority identi- fied lack of books and school supplies as the main source of their frustrations. Surprisingly, the share of satisfied pupils is much greater in rural than in urban areas: 51 percent, compared with 32 percent. Despite the advantages of going to school in Kigali Ville, only 37 percent of the pupils from this area report no problems with their schooling. A possible explanation for these patterns is that respondents have different implicit standards, reflecting differences in their expo- sure to and knowledge about alternatives to the schooling they actually receive. It might then be expected that urban children and those in Kigali Ville would be more demanding in their expec- tations and less satisfied with the services they receive. 96 WORLD BANK COUNTRY STUDY TABLE 5-4: ACCESSIBILITY OF PRIMARY SCHOOLS AND PROBLEMS WITH SCHOOLING REPORTED BY CURRENTLY ENROLLED PUPILS, RWANDA, 2001 Households Type of problem within 30 minutes Pupils reporting encountered by those reporting of a primary school no problem a problem (%) Index Index Lack of Poor Buildings (urban and (urban) and books and quality of Lack of in bad % Butare = 100) % Butare =100) supplies teaching teachers condition Rwanda 53 49 91 3 3 7 Type of locality Urban 83 100 32 100 92 3 2 7 Rural 51 62 51 161 83 3 4 6 Province Butare 58 100 52 100 96 5 4 6 Byumba 48 82 59 113 97 1 0 0 Cyangugu 46 79 37 71 87 2 2 9 Gikongoro 52 89 54 104 90 5 2 4 Gisenyi 53 92 53 101 89 2 1 11 Gitarama 45 77 41 78 98 1 1 1 Kibungo 45 77 54 103 91 0 0 4 Kibuye 51 88 43 82 89 12 3 18 Kigali Rural 46 78 57 109 97 1 3 2 KigaliVille 74 127 37 71 93 2 1 4 Ruhengeri 66 114 55 107 95 6 5 10 Umutara 47 81 54 103 77 2 7 10 Blanks denote not applicable. Source: 2001 Questionnaire unifié sur les indicateurs de développement (QUID) household survey. It is surprising that so few pupils report problems other than lack of books and supplies, given the overall characteristics of the supply infrastructure. Here again, a possible explanation is that although pupils can report objectively from personal experience on the availability of books and supplies, they have few benchmarks for judging the other types of problems about which they were asked. For example, few of those surveyed are likely to know much about teaching quality, and respondents would hardly be in a position to identify that as a problem. For many of the pupils, the school they currently attend is probably the only one they have ever attended, so that they have no basis for comparison. Teacher Allocation across Schools In this section we examine in greater detail various aspects of service delivery, beginning with the allocation of teachers across schools. Because teachers embody the bulk of resources that schools, particularly at the lower levels, have to deliver services, teacher allocation across schools is an important management issue. EDUCATION IN RWANDA 97 FIGURE 5-3: RELATION BETWEEN NUMBER OF PUPILS AND NUMBER OF TEACHERS, PUBLIC PRIMARY SCHOOLS, RWANDA, 2000 Note: Each circle represents a school. Source: Ministry of Education 1999­2000 census of primary schools. Overview of Allocation Patterns in the Public Sector Figure 5-3 shows the pattern of teacher allocation based on data for nearly 2,000 of the country's primary schools. (Private schools are not included because of missing data.) Both the state and the libre subsidié panels show the characteristic positive relation between school enrollments and the number of teachers allocated to a school. This pattern is to be expected, but closer examination suggests that there is a wide variation across schools in allocation of teachers. For example, a school enrolling 800 pupils, in either sector, may have as few as 10 teachers or as many as 20. The patterns revealed by the figure raise questions about the application of rules affecting teacher allocation and the extent to which disparities across schools stem from provincial differences or from disparities within provinces. To examine these questions, we turn next to regression analysis of the data.3 Regression Analysis of Teacher Allocation across Schools Results on teacher allocation are presented in table 5-5. The coefficient on the number of pupils is estimated at 0.016 for both state and libre subsidié schools. This implies that, on average, a school with 100 more pupils than the average size would have 1.6 extra teachers; thus, a new teacher is added for every 62.5 new pupils enrolled. Consider next the R2 values shown in the table. The R2 statistic is a summary indicator that captures the explanatory power of the regression equations. It ranges in value from 0, which implies that the allocation of teachers across schools is independent of the number of pupils each school enrolls, to 1.0, which implies that the variation is completely explained by differences in student enrollments across schools. Again, the results for both samples of schools are comparable, with R2 values of 0.79 for the state schools, 0.77 for the libre subsidié sample, and 0.78 for the combined samples. In international comparisons, the latter value places Rwanda in the middle range of R2 values for a sample of 12 African countries for which similar regression analysis has been completed (see figure 5-4). Some of these countries, such as Guinea, achieve better results, perhaps providing a target to aim for; other countries, such as Togo, do much worse. Because both the coefficient estimate on number of pupils enrolled and the R2 values in table 5-5 are comparable in the regressions for the state and libre subsidié samples, we can conclude that teacher allocation rules are actually highly comparable in the two sectors. 3. A simple way to understand the approach is to think of drawing a best-fit line through the observa- tions plotted in the two panels of figure 5-3. 98 WORLD BANK COUNTRY STUDY TABLE 5-5: REGRESSION ESTIMATES OF THE RELATION BETWEEN NUMBERS OF TEACHERS AND PUPILS, PUBLIC PRIMARY SCHOOLS, RWANDA, 2000 Both types State schools Libre subsidié schools of public school Regressor Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Number of pupils 0.016 45.26 0.016 69.55 0.016 83.21 Constant 1.34 4.66 1.58 9.16 1.52 10.27 Number of observations 548 1,440 1,988 R2 0.79 0.77 0.78 Memorandum: R2 in regression models with provincial dummiesa 0.82 0.81 0.81 Note: All coefficient estimates for number of pupils are statistically significant at the 1 percent level. a. See appendix table A5-3 for full regression results. Source: Authors' estimates based on Ministry of Education 1999­2000 census of primary schools. FIGURE 5-4: R2 VALUES OF REGRESSIONS RELATING NUMBER OF TEACHERS TO NUMBER OF PUPILS ACROSS SCHOOLS, RWANDA AND SELECTED SUB-SAHARAN AFRICAN COUNTRIES, CIRCA 2000 Note: The closer the R2 value is to 1.0, the tighter is the relation between school size and the number of teachers allocated to the school. Source: Regression analysis of data from each country's annual census of schools. EDUCATION IN RWANDA 99 TABLE 5-6: REGRESSION ESTIMATES OF THE RELATION BETWEEN NUMBERS OF TEACHERS AND PUPILS IN PUBLIC PRIMARY SCHOOLS BY PROVINCE, RWANDA, 2000 Regression results and related information Simulations for a school Coefficient with 700 pupils estimate on Number of Number of Pupil- number of pupilsa Constant schools (N) R2 teachers teacher ratio Rwanda 0.016 1.52 1,988 0.78 12.7 55.0 Butare 0.016 1.88 174 0.74 13.2 53.0 Byumba 0.015 2.00 162 0.80 12.3 56.9 Cyangugu 0.018 0.43 156 0.89 13.0 53.7 Gikongoro 0.017 2.21 155 0.64 13.9 50.4 Gisenyi 0.011 3.74 219 0.62 11.7 59.7 Gitarama 0.016 1.27 262 0.86 12.5 55.8 Kibungo 0.015 1.36 144 0.73 11.9 58.7 Kibuye 0.015 0.62 186 0.82 11.3 62.2 Kigali Rural 0.017 0.80 190 0.85 12.4 56.7 KigaliVille 0.019 4.21 42 0.90 17.2 40.6 Ruhengeri 0.017 1.49 210 0.76 13.3 52.6 Umutara 0.015 2.62 94 0.77 13.1 53.4 Note: Except in Gisenyi, Gitarama, and Kibuye, the R2 value of the regression for state schools is typically com- parable to that for the corresponding sample of libre subsidié schools. In those three provinces the R2 values for state school samples were 0.52, 0.58, and 0.97, respectively, compared with the corresponding values of 0.64, 0.93, and 0.80 for the libre subsidié samples. a. All coefficients in this column are statistically significant at the 1 percent confidence level. Source: Authors' estimates based on Ministry of Education 1999­2000 census of primary schools. We then repeated the regressions with provincial dummy variables included in the specifica- tion. The R2 values of the new regressions are reported in the last row of table 5-5; full regres- sion results can be found in appendix table A5-3. The values are only slightly higher than those associated with the regressions without the provincial dummy variables: in the regressions using the combined samples of state and libre sub- sidié schools, the R2 statistic rises only slightly, from 0.78 to 0.81. The implication is that differ- ences across provinces in teacher allocation contribute only marginally to the observed disparities in teacher allocation across Rwandan primary schools. A corollary is that the disparities largely result from inconsistent application of teacher allocation policies within each province. Patterns of Teacher Allocation within Provinces Table 5-6 presents additional results based on province-specific analysis. Separate regressions were run for each province, using the combined samples of state and libre subsidié schools.4 As a refer- ence, the results for the whole country are shown in the top row of the table. Consider first the coefficient estimate on the number of pupils. The estimate ranges from a low of 0.011 for Gisenyi to a high of 0.019 for Kigali Ville. These results mean that the overall 4. Combining the two samples helps avoid problems with small sample size in the regression analysis. Because the state and libre subsidié schools are highly similar in their patterns of teacher allocation, this approach poses few problems. 100 WORLD BANK COUNTRY STUDY pattern in the relation between enrollments and teacher allocation is such that for schools in Gisenyi, a new teacher is added, on average, for every 90.1 new pupils enrolled, whereas in Kigali Ville the corresponding figure is 52.6 pupils. Substantial variation thus exists across provinces in the sensitivity of teacher allocation to changes in the size of enrollments. We turn next to the values of the R2 statistic for each regression. This statistic captures the tightness of the relation between enrollments and teacher allocation (that is, how closely the schools in the sample cluster around the regression line relating the two variables). The overall value for the countrywide regression is 0.78, but the province-based regressions yield values rang- ing from only 0.62 in Gisenyi and 0.64 in Gikongoro to 0.89 in Cyangugu and 0.90 in Kigali Ville. In addition to differences across provinces in the slope of the relation between enrollments and teacher allocation, differences thus also exist in the consistency of teacher allocation. Gisenyi stands out among the provinces because teacher allocation there appears to be driven to a larger extent than elsewhere by factors other than enrollments, and the allocations are, moreover, less sensitive to enrollment growth than in other provinces. For better understanding of the implications of the regression results, consider the last two columns in the table, which show simulations of teacher allocations based on the regression esti- mates. The countrywide regression implies that a school enrolling 700 pupils (close to the aver- age size of primary schools in the country) would, on average, receive an allocation of 12.7 teachers, which implies a corresponding ratio of 55.0 pupils per teacher. The same calculation based on the regressions for each province shows teacher allocations ranging from 11.7 in Gisenyi to 17.2 in Kigali Ville and pupil-teacher ratios ranging from around 60 in Kibuye, Gisenyi, and Kibungo to 40.6 in Kigali Ville. Only in Kigali Ville, then, does the level of teacher allocation appear adequate to permit schools to assign teachers to only a single shift of pupils throughout all six grades of the cycle. In the other provinces the compromise solution is to assign two shifts per teacher in the first three grades and single shifts in the upper grades. To summarize, the complement of teachers in Rwanda's public schools is relatively meager, as suggested by the fact that the country's average pupil-teacher ratio of 57 is among the highest in the developing world. Assigning teachers double-shift duty in the early grades helps reduce the inevitable overcrowding in the classroom implied by the high pupil-teacher ratio, but the arrange- ment also means that pupils receive relatively few hours of instruction in the school day. In other countries with high pupil-teacher ratios, the compromise is brokered by tolerating overcrowding in the classroom. Neither compromise appears to work, however, as suggested by the generally high rates of grade repetition where pupil-teacher ratios are high.5 A second result pertains to the pattern of allocation across schools. In Rwanda the allocation appears to be reasonably consistent in relation to school enrollments, but scope for improvement remains, particularly in such provinces as Gisenyi and Gikongoro and, to a lesser extent, in Butare, Kibungo, Ruhengeri, and Umutara. In a growing system such as Rwanda's, teacher recruitment is likely to increase as well, both to serve larger enrollments and as part of an effort to lower pupil- teacher ratios. At this stage, consistency in the allocation of teachers across schools can be improved relatively painlessly by explicitly and unfailingly assigning new teachers to schools with high pupil-teacher ratios. Missing this opportunity would make the task harder in the future, as the pace of teacher recruitment slows. Improvement would then increasingly necessitate the reassign- ment of incumbent teachers across schools--a difficult task in the best of circumstances. Economies of Scale in Service Delivery We now consider a related but different issue: how the cost of service delivery varies across schools of different sizes. The issue is important because of its policy implications for the size and spatial distribution of schools. 5. For example, pupil-teacher ratios in such countries such as Benin, Cameroon, Madagascar, and Mozambique range between 54 and 65 pupils per teacher, and the repetition rate is between 24 and 26 percent. EDUCATION IN RWANDA 101 FIGURE 5-5: RELATION BETWEEN PERSONNEL COST PER PUPIL AND SCHOOL SIZE, PUBLIC SCHOOLS, RWANDA, 1999 Note: FRw, Rwandan francs. Each circle represents a school. Source: Based on Ministry of Education 1999­2000 census of primary schools merged with data from the 1999 census of teachers and data on the salary scale. Relation between Size of Enrollments and Unit Costs Data for the analysis were compiled from two datasets with the requisite information at the school level: the 1999 census of teachers, and the 1999­2000 census of schools. Using three variables common to the two datasets (province, district, and school name), data for 1,774 public schools were successfully merged, yielding a sufficiently large number of observations for the statistical analysis. The teacher census included information on the educational qualifications of teachers and other school personnel, their years of experience, their current duties, and other personal and professional information. By combining this information with the data on salary levels and structure shown in appendix table A5-1, it was possible to estimate the aggregate cost of personnel for each school and then to compute the cost of personnel per pupil in each of the 1,774 schools. Although personnel costs are only part of the total cost of service delivery, they represent the bulk of spending and therefore provide a good proxy for overall spending. The simple scatter plots of enrollments and unit costs in figure 5-5 reveal more diversity across schools than would be expected only on the basis of the distribution of teachers documented earlier. It implies that differences in the educational qualifications and seniority of teachers and other school personnel tend to widen rather than narrow the gap in funding across schools, through the systematic link between these variables and the salary bill. Regression Analysis of the Relation between Unit Costs and School Size The results presented in table 5-7 confirm the pattern in figure 5-5. The relation between cost per pupil and school size is weaker than that between number of teachers and school size. The R2 value falls to 0.70 for the combined sample of state and libre subsidié schools, from 0.78 in the corresponding regression reported earlier that had number of teachers instead of unit costs as the dependent variable. A more important result shown in the table relates to economies of scale in service delivery. The estimated coefficient on the number of pupils implies that a 1 percentage point increase in personnel costs per pupil allows enrollments to expand by about 1.04­1.05 percent. The implica- tion is that unit costs fall as schools increase the size of their enrollments, all else being the same. 102 WORLD BANK COUNTRY STUDY TABLE 5-7: REGRESSION ESTIMATES OF THE RELATION BETWEEN TOTAL PERSONNEL COSTS AND ENROLLMENTS ACROSS SCHOOLS, RWANDA, 2000 State schools Libre subsidié schools All public schools Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic Number of pupils 0.96 33.36** 0.96 53.79** 0.96 60.33** Intercept 8.86 50.58** 8.92 78.15** 8.92 82.81** R2 0.71 0.70 0.70 Number of observations 508 1,266 1,774 Memoranda: Economies of scalea 1.04 1.05 1.05 Average number of pupils per school 720 678 684 ** Statistical significance at the 1 percent confidence level. Note: In all the regression models, both the dependent and independent variables are expressed on a log scale. a. Refers to the percentage increase in the number of pupils enrolled, given a 1 percent increase in spending on personnel, calculated as the inverse of the coefficient estimate on the independent variable (number of pupils). Source: Based on Ministry of Education 1999­2000 census of schools and on data on the salary scale of teach- ing personnel. FIGURE 5-6: ECONOMIES OF SCALE IN PUBLIC PRIMARY EDUCATION, RWANDA, 2000 Note: FRw, Rwandan francs. Source: Simulated from regressions in table 5-7. Yet the magnitude of these economies is modest, as the simulations in figure 5-6, which are based on the regression results, suggest. Taking the simulations for public schools in general (that is, using the combined state and libre subsidié samples), we note that the predicted per pupil cost of school personnel is only 2 percent higher in a school enrolling 400 pupils than in a school with 700 pupils (FRw 5,781, compared with FRw 5,641). As enrollment falls to 100 pupils, the pre- dicted unit cost rises to FRw 6,145--an increase of 6 percent, which is still a modest difference. EDUCATION IN RWANDA 103 TABLE 5-8: PERCENTAGE DISTRIBUTION OF PRIMARY SCHOOLS AND COST PER PUPIL BY SIZE OF ENROLLMENT, RWANDA, 2000 Size distribution of schools (%) Cost per pupil in the public sector Public sector Size of enrollment (thousand FRw)a State Libre subsidié Private sector <150 6.1 3.2 1.8 29.0 150­300 5.9 6.5 7.2 35.5 300­450 5.7 13.4 17.8 6.5 450­600 5.7 19.3 20.9 12.9 600­750 5.6 21.1 18.5 6.5 750­900 5.5 11.4 11.9 0.0 >900 5.5 25.1 21.8 9.7 Note: Columns sum to 100 percent. Percentages are based on data for 2,043 of the 2,093 primary schools for which relevant data are available. a. Costs are simulated from the regression estimates in table 5-7 for schools whose enrollments correspond to the middle of each size bracket except that, for the first category, cost corresponds to a school enrolling 100 pupils, and, for the last category, cost corresponds to a school enrolling 1,000 pupils. Source: Authors' estimate based on Ministry of Education 1999­2000 census of primary schools. Policy Implications for the Size Distribution of Primary Schools The government as service provider has the option of delivering school services through a net- work of relatively few schools, each enrolling a large number of pupils, or, alternatively, through a larger number of relatively small schools. In the Rwandan context, it would appear that the latter option would make more sense, given the very limited economies of scale in service delivery doc- umented above. Large schools are not much cheaper to run, but they are necessarily farther from the average pupil's home. The 2001 QUID household survey indicates that children in about half of the survey households take more than 30 minutes to reach the nearest primary school. Deliv- ering services through smaller schools located closer to the pupil's home would help reduce the physical barriers to schools and lessen the opportunity cost of school attendance. Yet the current network of public schools tends to emphasize size over proximity. Average school enrollment is about 720 pupils in the state sector and 678 pupils in the libre subsidié sec- tor, for a combined (weighted) average of 684 pupils per school. By contrast, private schools enroll only 342 pupils each, on average. Table 5-8 provides another perspective on the differences in the size distribution of schools in the public and private sectors. Among the state schools, less than a quarter enroll fewer than 450 pupils each--about the same share as that of schools enrolling more than 900 pupils. In the libre subsidié sector the emphasis on size is less pronounced but remains significant. The real contrast is with schools in the private sector, where about two-thirds of the schools enroll fewer than 300 pupils each. Student Learning So far, we have addressed aspects of service delivery that affect resource flows to each school. We now move beyond resource allocation to examine the relation between funding and learning outcomes. How strong is the relation between these variables? How greatly do schools differ in their ability to transform funding into student learning? What does the pattern imply for develop- ment of policies to improve the effectiveness of service delivery? 104 WORLD BANK COUNTRY STUDY TABLE 5-9: PRIMARY SCHOOL LEAVING EXAMINATION RESULTS BY TYPE OF SCHOOL, RWANDA, 1999 Type of school Mean score (%) Standard deviation Number of schools State 44.6 8.6 398 Libre subsidié 44.5 8.3 1,135 Private 46.5 13.1 14 All schools 44.5 8.4 1,547 Source: National Examination Council of Rwanda. Evaluating these issues is not easy in the best of circumstances, but in Rwanda the challenges are magnified by data constraints. The country has only recently joined the UNESCO-sponsored Monitoring Learning Achievement Study, and the data and the findings are not yet available. In the absence of other data sources, we use the results on the national examinations administered at the end of the primary cycle as a proxy for student learning. Although this measure is admittedly flawed, it has the advantage that the data pertain to the entire country. The analysis should ideally be performed using individual-level information, but the practical difficulties of preparing the requisite data proved insurmountable; a usable dataset would require for each pupil other information besides examination results, including, in particular, socioeco- nomic background. Given the constraints, we perform the analysis using schools as the unit of observation. We merged the examination data for 1999 with those from the 1999 teacher census and the 1999­2000 school census.6 The resulting dataset pertains to 1,362 of 1,533 schools in the public sector that presented candidates in 1999. For each school, information is available on the average score of the school's grade 6 candidates and on such characteristics of the school as overall spending per pupil on personnel, the educational attainment of teachers, the pupil-teacher ratio, and the condition of buildings. Although the resulting dataset is unique in Rwanda and is very useful for our purpose, its limitations should be kept in mind when interpreting the results. Overview of Examination Results By way of background, table 5-9 shows the mean examination results across the three types of school. The data for state and libre subsidié schools are highly comparable, with mean scores of around 45 and standard deviations of 8.6 and 8.3, respectively. The main difference is with the private sector, where both the average score and the standard deviation are significantly higher. Private school pupils tend to achieve high scores, but the gap between the top and bottom per- formers is much wider. Figure 5-7 illustrates the relation between pass rates on the grade 6 national examinations and the per pupil cost of teachers and other school personnel. Among both state and libre subsidié schools, the relation is weak in two senses: (a) on average, schools with more spending per pupil do not necessarily achieve higher pass rates, and (b) for schools with a given level of spending-- say, FRw 6,000 per pupil--the pass rates vary widely, from a low of 20 percent to a high of 60 percent. These findings are not unique to Rwanda; in other countries where similar analyses have been conducted, such as Mauritania and Madagascar, the patterns are similar. How is the finding that the level of school funding relates only weakly to pass rates on the examination to be interpreted? An attempt to answer this question must begin with the recognition 6. We should ideally have matched the data from the two censuses to the examination results for 2000 to improve the time consistency of the analysis. Because of the practical constraints, we relied instead on the examination results for 1999. Our assumption is that any change in the characteristics of individual schools in the course of a single school year would have been modest. EDUCATION IN RWANDA 105 FIGURE 5-7: RELATION BETWEEN EXAMINATION RESULTS AND PER PUPIL COST OF PERSONNEL, PUBLIC SCHOOLS, RWANDA, 1999 Note: FRw, Rwandan francs. Each circle represents a school. Source: Based on school-level data examination results merged with estimates of personnel costs from the 1999 census of teachers. that the translation of resources into learning outcomes is affected (at the least) by (a) how the resources are used to create learning environments at the classroom level and (b) what incentives are in place to motivate the behavior of teachers and pupils. Use of resources. With regard to how resources are used, the mix of school inputs that character- ize classroom conditions is particularly relevant. These inputs include the availability of teachers (as proxied by the pupil-teacher ratio), teachers' qualifications, the supply of textbooks and learning materials, and so on. Clearly, the higher the level of school funding, the more of each of these inputs can be purchased, and the better the learning environment will be. But in fact, all education systems operate in resource-constrained environments. The issue is to find the right balance among the various inputs so that at each level of funding, the resources are used to create as effective a learn- ing environment as possible. Tradeoffs in the choice of inputs are inescapable, and poor input mixes can give rise to differences across schools in the translation of resources into learning outcomes. Incentives. It is a truism that in the absence of appropriate motivation and support, even the best equipped teachers would fail to produce good learning outcomes among their students. There are various ways of communicating to teachers a set of incentives for good performance, and not all of them need involve financial rewards. In some countries simply clarifying expecta- tions and the scope of a teacher's responsibility has been a first step toward more effective class- room teaching. In others, supportive supervision practices designed to help teachers better man- age classroom processes have yielded positive results. These issues pose difficult challenges for managing the performance of service delivery. Although the data at our disposal do not permit a detailed analysis, some progress in understand- ing the nature of the challenges can be made using simple regression analysis to examine the cor- relates of performance on the grade 6 national examinations. Regression Analysis of the Correlates of Grade 6 Examination Results The nature of the data (which, again, relate to school-level observations) limits us to an assess- ment of the relation between examination results and certain physical attributes of the learning environment. These attributes include the pupil-teacher ratio, teachers' educational qualifications and average years of experience, and the quality of the facilities as proxied by their reported con- dition. Because teacher qualifications and experience can be translated into the salary bill, we can 106 WORLD BANK COUNTRY STUDY TABLE 5-10: CORRELATES OF SCHOOL-LEVEL PASS RATES ON THE PRIMARY SCHOOL LEAVING EXAMINATION, RWANDA, 1999 Sample mean of regression Estimated marginal Regression variable Teacher credentiala variableb impactc Dependent variable: examination 44.5 pass rate (%) Teacher qualification Reference group: upper secondary D6, D7 pédagogique 32.9 diploma with preservice teacher training Primary CA 1.3 ­0.06 Lower secondary General ES 1, 2, 3 8.0 ­0.07** Vocational CERAI, etc. 13.7 ­0.07** Teacher training EAP, EMA, etc. 4.1 ­0.06 Upper secondary Incomplete ES 4, 5, 6 11.8 ­0.11** Diploma (3­5 years) D3, D4, D5 18.3 ­0.03 Diploma (6­7 years) without D6, D7 non- 9.9 0.00 preservice teacher training pédagogique Spending on personnel per pupil (FRw) 5,047 0.31** Blanks denote not applicable. ** Statistically significant at the 1 percent level. Note: FRw, Rwandan francs. See appendix table A5-4 for the full regression results and full list of the variables included in the regression analysis reported here. a. Because of small cell sizes, no distinction is made between those with and without preservice teacher training. b. For the teacher qualification variable, the sample means correspond to the percentage distribution of the sample across qualification groups. c. For the teacher qualification variable, the marginal effect refers to the percentage point change in the examination pass rate in response to a 1 percentage point increase in the share of teachers with the corre- sponding qualification at the expense of a 1 percentage point decline in the share of teachers with the refer- ence qualification (upper secondary school diploma with teacher training). For the spending variable, the mar- ginal effect refers to the percentage point change in the pass rate in response to an increase of FRw 1,000 in per pupil spending on personnel above the sample mean of FRw 5,047. Source: Authors' estimates based on school-level data on examination results for 1999, merged with data from Ministry of Education 1999­2000 census of primary schools and 1999 census of teachers. replace those variables with the estimated per pupil cost of teacher salaries in the specification of the regression equation. The full regression results appear in appendix table A5-4; the main highlights are summarized in table 5-10. Because teachers are the single most important input in the schooling process, absorbing the largest share of spending on education, the highlights in fact pertain mainly to the relation between teacher qualifications and the examination pass rate.7 Using as the reference 7. The variable is defined for this exercise in terms of the share of teachers with the credentials shown in table 5-10. EDUCATION IN RWANDA 107 group teachers with an upper secondary diploma and preservice training (32.9 percent of the whole sample), the regression analysis allows us to compare the performance of teachers with other qualification levels against the performance of the reference group, keeping the other (measured) attributes of the schooling environment constant. For example, in the last column of table 5-10, which shows estimated marginal impact, the value of -0.06 appears in the row for primary education. This means that a 1 percentage point increase in the share of teachers with a primary school certificate at the expense of teachers with the reference qualification (upper sec- ondary diploma with preservice teacher training) would reduce the pass rate by 0.06 percentage points. Several implications emerge from the results. The marginal impact is negative and statistically significant for teachers with general lower secondary education (ES 1, 2, or 3), vocational lower secondary education (for example, CERAI), or an incomplete upper secondary education (ES 4, 5, or 6). In other words, teachers with these qualifications (designated as "unqualified" teachers in the terminology used in Rwanda) are inferior to those with an upper secondary diploma and with preservice teacher training. These negative results are consistent with perceptions among Rwandan policymakers. Yet the magnitude of these teachers' shortfall in performance is modest. For example, the maximum marginal effect for those with ES 4, 5, or 6 credentials is only -0.11. Thus, if the 11.8 percent of sample teachers in this "incomplete upper secondary education" group were replaced entirely with teachers with an upper secondary diploma and preservice train- ing (D6 or D7 pédagogique), the pass rate would rise from the current sample average of 44.5 percent to only 45.7 percent. Taking the more drastic measure of replacing all three groups of "unqualified" teachers (the 33.5 percent who have ES 1, 2, or 3, CERAI or equivalent, or ES 4, 5, or 6 credentials) would boost the pass rate to only 47.2 percent. The modest gain in the pass rate must be balanced against the implied cost of raising teacher qualifications. In the salary structure prevailing in 2001, the annual starting pay of an unqualified teacher ranges from FRw 7,245 a month, plus benefits of FRw 4,000, for those with a general lower secondary education, to FRw 12,462, again with FRw 4,000 in benefits. By contrast, teachers with an upper secondary diploma start at FRw 23,649, plus 6,500 in benefits. The pay gaps are large, and it is reasonable to raise the issue of tradeoffs. In particular, given the current learning environments in most Rwandan schools, would a deployment of resources to hire the top-qualified teachers be the best use of scarce resources? Would the money go farther if it were instead used to increase the supply of textbooks and provide better support for incumbent teach- ers through in-service teacher training and more effective supervisory services? A further detail in the regression results pertains to the impact of the level of spending per pupil on the examination pass rate. The marginal impact on this variable is positive, indicating that, in general, the higher is the level of funding for schooling, the better is its performance as meas- ured by the pass rate. But, consistent with the story thus far, the effect is modest; an increase in spending of FRw 1,000 (that is, by about 20 percent above the sample mean) would boost the pass rate by only 0.31 percentage points, from the current average of 44.5 percent to 44.8 percent. Differences in Performance across Provinces The same regression analysis also allows us to characterize examination performance across provinces (see table 5-11; see also model 2 in the results reported in appendix table A5-4). Butare has the best results, and we use it as the reference province. The other provinces are listed in order of their distance from the average performance of Butare. Controlling for teacher qualifi- cations, the average pass rate of schools in Cyangugu is about 9.2 percentage points below the average for Butare, whereas the pass rate for schools in Byumba is, on average, 19.1 percentage points below Butare's. The provincial ranking is relatively stable whether the regression controls for teacher qualifi- cation or for the level of per pupil spending. Three groups may be identified among the provinces. (a) Cyangugu, Gisenyi, Kigali Rural, Kibungo, and Ruhengeri are a little behind 108 WORLD BANK COUNTRY STUDY TABLE 5-11: REGRESSION ESTIMATES OF PROVINCIAL DIFFERENCES IN EXAMINATION PASS RATES, CONTROLLING FOR DIFFERENCES IN PER PUPIL SPENDING ON PERSONNEL, RWANDA, 1999 Examination pass rate relative to Butare Controlling for teacher Controlling for per pupil Provincial dummy variables qualifications spending on personnel Reference province: Butare Cyangugu ­9.2 ­7.6 Gisenyi ­10.8 ­8.9 Kigali Rural ­10.8 ­9.3 Kibungo ­10.9 ­10.7 Ruhengeri ­11.2 ­8.8 Umutara ­12.8 ­14.5 Kibuye ­15.0 ­12.9 Gikongoro ­15.1 ­14.3 Gitarama ­15.3 ­13.3 KigaliVille ­17.8 ­15.4 Byumba ­19.1 ­17.0 Blanks denote not applicable. Note: See appendix table A5-4 for regression specification and full regression results. All coefficients on the provincial dummy variables are statistically significant at the 1 percent confidence level. Source: Authors' estimates based on school-level data on examination results for 1999 merged with data from Ministry of Education 1999­2000 census of primary schools and 1999 census of teachers. Butare; (b) Kigali Ville and Byumba are far behind; and (c) Gitarama, Gikongoro, Kibuye, and Umutara fall into an intermediate ranking between these extremes. These results tell us that dif- ferences in the characteristics of schools that are explicitly accounted for in the regression analysis-- teacher qualifications, pupil-teacher ratio, condition of the facilities, and per pupil spending on staff--are only part of the reason why schools differ in their ability to produce high-scoring stu- dents. The fact that the provincial dummy variables are all statistically significant suggests that other unobserved differences across provinces are also at work. Furthermore, the size of the inter- provincial gaps implies that these unobserved factors are quite important, making it worthwhile to explore differences in school and classroom management practices that could account for the gaps in performance. Butare's position makes it particularly interesting in this regard. Policy Implications and Conclusion Primary school services in Rwanda are delivered largely through the network of state and libre subsidié schools that make up the public sector. Libre subsidié schools are financed largely by the government through payment of teacher salaries but are by and large run and managed by the Secrétariat nationale de l'enseignement catholique (SNEC) or the Conseil protestant du Rwanda (CPR). Beyond documenting these structural dimensions of the system, the analysis in this chap- ter touched on several issues with important policy implications for the management of service delivery. These include (a) the size and spatial distribution of schools, (b) consistency in teacher allocation across schools, (c) the balance of input mix to support effective learning environments, EDUCATION IN RWANDA 109 and (d) the effectiveness of classroom processes in transforming school funding into learning outcomes. Although data constraints prevent a definitive analysis of these problems, the findings nonetheless provide food for thought in the context of policy development. Public schools are relatively large in Rwanda, enrolling, on average, about 700 pupils each. Yet the available evidence suggests that there are few economies of scale beyond enrollments of about 400 students. Transforming the current network of schools into a system of smaller but more numerous schools would not increase the cost of service delivery but would help bring schools closer to pupils' homes, thus improving access. Rural children would benefit the most, as only half the households in rural areas currently live within 30 minutes of the nearest school. With regard to the consistency of teacher allocation across schools, the findings suggest that, overall, the results are reasonably good relative to performance in other countries. Yet room for improvement remains, particularly in such provinces as Gisenyi and Kibungo where the supply of teachers is both relatively meager and poorly deployed. The prospects for improvement are good provided that the government takes advantage of new recruitment to rebalance the allocation of teachers in favor of schools and provinces that are currently underendowed. The mix of school inputs has both pedagogical and financial dimensions. In an ideal, well- funded learning environment, pupils have sufficient instructional time during the school day, are taught in small classes by teachers with good qualifications, and receive an ample supply of books and other learning resources. In resource-constrained environments, tough choices and tradeoffs have to be made. The challenge facing policymakers is to manage these choices so as to relieve the practical (as opposed to the abstract) constraints on performance. Some of the results presented in this chapter raise important questions in this regard. The overall supply of teachers in Rwanda is generally low in relation to enrollments, as reflected in the country's relatively high pupil-teacher ratio, which currently averages about 57 pupils per teacher in the public sector. Reducing the ratio would appear appropriate in light of experience in other countries. In a context of hard budget constraints, however, this would require tradeoffs against other inputs that might also be deemed important. In particular, the level of teacher qualifications comes to mind. Hiring well-qualified teachers--for example those with upper secondary educa- tion diplomas--has its benefits for the learning process, and the current profile of primary school teachers shows the country making significant headway in this direction. Yet the tenuous link between teacher qualifications and performance--measured imperfectly here using examination results--raises legitimate questions about the wisdom of putting the bulk of resources into teacher qualification at the expense of lowering the pupil-teacher ratio and thereby increasing the length of the school day for pupils in the first three grades of school. In addition, the analysis highlights the need to look beyond physical inputs. Resources mat- ter, but because schooling is a social process in which participants' behavior affects the effective- ness with which resources are transformed into outcomes, it is critical to ensure that the "soft- ware" of managing service delivery is put in place and is used. Teachers are the main agents on the front lines. It is essential that they are equipped for the job through proper training and that they receive the support and incentives they need to perform their jobs well. CHAPTER 6 SERVICE DELIVERY IN SECONDARY EDUCATION S econdary education in Rwanda is a relatively small subsector at present, enrolling about 141,000 students in 2001, compared with nearly 1.5 million pupils at the primary level. Yet, as chapter 2 suggests, the pressure to expand secondary school places is likely to mount in coming years as more and more children complete their primary schooling.1 Managing the sub- sector's expansion in a fiscally sustainable manner is therefore a critically important policy issue. As a contribution to the discussion, this chapter documents the subsector's key features by con- solidating information on the institutional composition of secondary schools and the overall char- acteristics of their services. It also examines the deployment of teachers and resources across schools, economies of scale in service provision, and the correlates of performance in the national examination at the end of the tronc commun cycle. As in chapter 5, data limitations have con- strained the analysis in important ways and have reduced the precision of the findings. Although the results should obviously be understood in their proper perspective, we believe they remain sufficiently robust to help guide policy development for the subsector.2 1. To illustrate, if survival rates at the primary cycle were to stabilize at even 50 percent--much lower than the current level of about 73 percent, which may not persist because of high repetition rates--the probability of a primary school leaver making the transition to the secondary cycle would be less than 40 percent, in contrast to the present rate of 60 percent, even if the secondary school system (public and pri- vate) enrolled twice as many students as it does now. 2. Data problems in secondary education were particularly serious. Information from the Ministry of Education annual school census for the latest available school year was incompletely computerized, and additional effort was needed to fill in the gaps. This task proved a challenge because schools were sometimes known by more than one name and different names were used in different datasets. Thanks to help from various sources, and time-consuming data-cleaning efforts, most of these problems were eventually resolved to our satisfaction. 111 112 WORLD BANK COUNTRY STUDY FIGURE 6-1: NUMBER AND INSTITUTIONAL COMPOSITION OF SECONDARY SCHOOLS BY PROVINCE, RWANDA, 2000­01 4Ð6% Source: For number of schools and their distribution, Ministry of Education 2000­01 census of secondary schools; for population shares, Rwanda (2002b). Overview of the Supply of Services In this section we provide a brief sketch of some key features of the supply of secondary schooling in Rwanda by documenting the institutional composition of the system, selected characteristics of the schools themselves, and the curricula offered by the schools. The picture that emerges is one of a highly fragmented system that probably needs to be streamlined for sustainable expansion. Institutional Infrastructure As at the primary level, the government finances and runs the state secondary schools. It pays the salaries of teachers in the libre subsidié schools but leaves day-to-day management of these schools in the hands of churches and other nongovernmental organizations. The government neither finances nor runs the private schools. Of the nearly 400 secondary schools in Rwanda, 19 percent are state schools, 30 percent are libre subsidié schools, and the remaining 51 percent are private schools (see figure 6-1). These percentages contrast sharply with the figures for the primary level, where the private sector share is tiny. The percentages vary widely across provinces: the private sector share, for example, ranges from more than 80 percent in Kigali Ville to just over 20 percent in Gikongoro; the state schools' share ranges from a low of about 6 percent in Butare to more than 40 percent in Kibungo. It is interesting to compare the supply of schools and their composition across provinces with the corresponding shares of the country's population. For example, about 10­12 percent of the population lives in Gitarama and Gisenyi, but Gitarama has more schools (about 50, compared with Gisenyi's 43), and a larger share of the total is private. The pattern is suggestive of possibly more favorable social conditions for secondary schooling in Gitarama, including stronger house- hold ability and willingness to pay for schooling and probably more effective local initiative in starting schools. Ruhengeri, Kigali Rural, and Byumba also account for about 10­12 percent of EDUCATION IN RWANDA 113 TABLE 6-1: NUMBER AND PERCENTAGE DISTRIBUTION OF TYPES OF SECONDARY SCHOOL, RWANDA, 2001­02 Number of schools by type % privately Network State Libre subsidié Private All typesa financed State 72 0 0 72 (19.0) 0.0 SNECb 0 61 28 89 (23.5) 31.5 CPRb 0 31 37 68 (18.0) 54.4 Other 0 20 129 149 (39.4) 86.6 All networksa 72 (19.0) 112 (29.6) 194 (51.3) 378 (100.0) 51.3 a. Figures in parentheses show total as percentage of all schools. b. Refers to the main church-related networks, the Secrétariat nationale de l'enseignement catholique (SNEC) and the Conseil protestant du Rwanda (CPR). Source: Compiled from lists of schools supplied by the Ministry of Education, the SNEC, and the CPR. the population, but they have only about 30 schools each, and whereas the private share of schools in Ruhengeri and Kigali Rural is about the same as in Gitarama, it is much smaller in Byumba. The three provinces with the smallest shares of population are Kibuye, Gikongoro, and Umu- tara. The first two have about twice as many schools as Umutara--in part, apparently, because of greater government involvement. The lower degree of government involvement in Umutara may be associated with the province's unique features. Most of the area is part of a national park, and many of the inhabitants are returnees from Uganda and have gone through an English medium school. Because the teachers needed to staff such schools are still rare in the Rwandan system, stronger government involvement in the province has simply been impossible to arrange in the short run. The data in table 6-1 show the distribution of secondary schools according to the manage- ment network to which they belong and their funding status. In Rwanda the main management networks are (a) the state, (b) the Catholic Church, through its Secrétariat nationale de l'enseignement catholique (SNEC), (c) a loose association of schools run by Protestant churches voluntarily subscribing to the Conseil protestant du Rwanda (CPR), and (d) independent or smaller associations of private or community schools. Schools belonging to the state, the SNEC, and the CPR each account for about a fifth to a quarter of the schools; the unstructured private sector claims the remaining two-fifths. State schools are by definition funded by the government; but unlike the situation at the primary level, sizable shares of SNEC schools (32 percent) and CPR schools (54 percent) are run entirely with private funding. Faced with limited public funding and a growing demand for sec- ondary schooling, these networks have met the demand in the only practicable way: by tapping into private sources of funding. In the independent schools sector, the surprising finding is that about 20 such schools have gained entry into the libre subsidié category, which entitles them to receive government funding for teacher salaries. Although it is unclear how non­public sector schools qualify for such funding, being in the libre subsidié group is so advantageous that schools can be expected to seek this prize assiduously. As indicated in chapter 5, significant pressures are building up on Rwanda's secondary school system, given the growing numbers of primary school leavers likely to seek entry to secondary school. Because of continuing budget constraints, the government is unlikely to be able to meet the increased demand by itself, and private funding will continue to be critical. Yet the current 114 WORLD BANK COUNTRY STUDY TABLE 6-2: SELECTED CHARACTERISTICS OF STATE, LIBRE SUBSIDIÉ,AND PRIVATE SECONDARY SCHOOLS, RWANDA, 1999­2001 Public sector Libre Private State subsidié schools All schools Number of schools, 2000­01 72 110 191 374a Share of students, 2000­01 (% of total number, 141,163) 19.6 34.6 43.8 100.0 Average enrollment per school, 2000­01 384 467 323 378 Share receiving food from Ministry of Education, 1999­2000 (%) No allocation 0.0 0.0 100.0 51.3 Allocation based on number of boarders only 58.3 46.8 0.0 25.0 Allocation based on numbers of boarders and day students 41.7 53.2 0.0 23.7 Total 100.0 100.0 0.0 100.0 a.Total includes one school for which information regarding the category is missing. Source: For data on number of schools, average enrollments per school, and distribution of enrollments, elec- tronic data files from Ministry of Education 2000­01 census of secondary schools; for data on the share of students in the tronc commun cycle, Ministry of Education statistical abstract, 2000­01; for data on schools receiving food, administrative records of the Department of Secondary Education, Ministry of Education. financing arrangements in the sector are such that the burden of mobilizing private funding falls unevenly across schools, with private schools bearing a disproportionate share of the load. In considering the future prospects for secondary education, the issue of school funding probably warrants additional attention. In particular, policymakers may need to rationalize funding arrangements to distribute the available public subsidies more equitably among schools so that public funds are used to leverage private contributions in all schools, thus putting the expansion of the subsector on a more fiscally sustainable footing. Some Characteristics of the Supply of Services According to table 6-2, the distribution of enrollments across state, libre subsidié, and private schools generally matches the distribution of the schools across these categories, as shown in figure 6-1. Libre subsidié schools tend to be larger than the average size of secondary schools nationwide, and private schools tend to be smaller, but the differences are not large, and the data reinforce the impression of a system composed mostly of relatively small schools. Rwanda's publicly financed secondary schools traditionally offer boarding services. Recently, because of the growing pressure to expand places, they have also been catering to day students. Unfortunately, no data are available on the number of boarders and day students, and we rely on the Ministry of Education's records of allocation of food.3 All secondary schools currently offer boarding services, and it appears that nearly 60 percent of the state schools cater exclusively to 3. Until 2001, the Ministry of Education allocated food to schools according to numbers of boarding and day students. This practice was replaced in 2002 by a new system whereby schools receive a financial allocation to cover the cost of feeding the students. EDUCATION IN RWANDA 115 TABLE 6-3: PERCENTAGE DISTRIBUTION OF SCHOOLS BY LEVEL AND NUMBER OF INSTRUCTIONAL STREAMS OFFERED, RWANDA, 2000­01 Public sector Libre Level State subsidié Private All schools Tronc commun only 51.4 24.5 34.6 34.9 Tronc commun combined with upper secondary cyclea 1 instructional stream 29.2 40.0 38.7 37.3 2 instructional streams 8.3 24.5 22.5 20.4 3 instructional streams 0.0 3.6 2.6 2.4 Upper secondary cycle onlya 1 instructional stream 8.3 7.3 1.6 4.6 2 instructional streams 2.8 0.0 0.0 0.5 Total 100.0 100.0 100.0 100.0 Memorandum: Number of schools 72 110 191 373 Note: Students in upper secondary school are enrolled in one of four streams: general, vocational, technical, or normal (teacher training). a. See appendix table A6-1 for details on the specific combinations of instructional programs offered. Source: Based on data from Ministry of Education 2000­01 census of secondary schools. boarders, compared with almost 50 percent among libre subsidié schools. Although boarding schools may be justified in an elite school system, they are a fiscally unviable model for expanding the system, given the large diversion of state funds to feed students. Finding ways to phase out boarding schools would appear to be a priority item on the agenda for policy development. Instructional Programs Secondary schools in Rwanda offer the gamut of instruction, from lower secondary schooling to specialized vocational and technical courses at the upper secondary level (table 6-3). Slightly more than a third of the schools provide instruction only in the tronc commun (lower secondary) cycle; nearly 60 percent offer both the tronc commun and upper secondary cycles; and 5 percent serve upper secondary students exclusively. The distribution of instructional programs shows dis- tinct differences across types of school: more than half of the state schools specialize in the tronc commun cycle, compared with only a quarter of the libre subsidié schools and about a third of the private schools. Among schools serving both lower and upper secondary students, the spread of course offerings is distinctly wider in the libre subsidié and private sectors; many of the schools provide instruction in two or even three streams at the upper secondary level. The diversity may be a sign of the schools' responsiveness to market demand (given that both public and private schools collect nontrivial levels of fees). Yet because the schools are typically small, with fewer than 500 students each, on average, the implied fragmentation of course offerings may mean that schools are failing to take advantage of scale economies in service delivery. We return to this issue later in the chapter, when evidence is presented on the existence and nature of scale economies in secondary education. 116 WORLD BANK COUNTRY STUDY TABLE 6-4: PERCENTAGE DISTRIBUTION OF SECONDARY STUDENTS BY CYCLE AND STREAM AND BY SCHOOL TYPE, RWANDA, 2000­01 Cycle and stream State Libre subsidié Private All schools Tronc commun cycle 67.1 59.7 64.2 63.1 Upper secondary cycle 32.9 40.3 35.8 36.9 Distribution in upper secondary cycle General 36.1 46.6 16.7 32.0 Normala 9.4 25.0 34.8 26.4 Vocational 40.2 26.3 40.8 34.9 Technical 14.3 2.1 7.7 6.6 Memorandum: Total number of students 27,667 51,277 61,782 140,726 a. Primary school teacher training stream. Source: Authors' calculations based on Ministry of Education electronic files from the 2000­01 census of secondary schools. Data on the distribution of enrollments appear in table 6-4. Consistent with the information presented above, about two-thirds of all secondary students are in the tronc commun cycle, with the slightly above-average share in the state schools counterbalancing the below-average share in the libre subsidié sector. The upper secondary level is divided into general, normal (teacher train- ing), vocational, and technical streams. Overall, less than 7 percent of the students pursue the technical stream, about a third of the total is in the general stream, another third is in the voca- tional stream, and a quarter is in the normal stream. There are substantial differences around these system averages in the distribution across types of school. In state schools there is a stronger focus on the vocational and technical streams than in the other two types of school; together, these streams cater to some 55 percent of state school students. In the libre subsidié category the emphasis is on the general stream, and in the private sector it is on the normal and vocational streams, which together serve more than three-quarters of private school students. Table 6-5 shows the number of schools offering the indicated fields of instruction in each stream, as well as the number of students enrolled in each field. The shaded cells highlight fields in which average enrollment per school in the public sector is fewer than 100 students.4 Six state and 18 libre subsidié schools offered the math-physique curriculum in 2000­01. Each state school enrolled, on average, 118 students in this field; the average for libre subsidié schools was only 86 students. One state and five libre subsidié schools offered the action sociale and secrétariat streams. Enrollments per school in each of these fields were much lower in the state than in the libre sub- sidié schools. Looking at the table as a whole, the general pattern in both the state and libre sub- sidié schools is that the technical streams tend to have small enrollments per field. The question at this point is whether the mix of services provided by the public sector (state and libre subsidié) schools is consistent with economic considerations. In particular, the small enrollments per field and institution merit further scrutiny because of their cost implications. 4. Because private schools in Rwanda rely heavily, if not completely, on fees to fund their operations, we assume that they are recovering their costs even at the low enrollments in some of fields shown in table 6-5. Public schools are shielded from similar market forces and are therefore likely to be less cost-conscious. Because the same labels may mask differences in course content, private schools may well be offering differ- ent curricula--adapted to keep costs down when enrollments are low--than those offered in the public sector. EDUCATION IN RWANDA 117 TABLE 6-5: NUMBER OF SCHOOLS OFFERING UPPER SECONDARY PROGRAMS AND AVERAGE ENROLLMENTS PER SCHOOL BY FIELD, RWANDA, 2000­01 Number of schools offering field Average number of students Libre Libre State subsidié Private State subsidié Private All schools offering a program at the upper secondary level 37 27 66 General Lettres 2 19 10 128 111 41 Science humaines 13 27 20 90 108 78 Math-physique 6 18 7 118 86 67 Bio-chimie 9 28 17 129 110 75 Normal (primary school teacher 4 22 62 215 235 124 training) Vocational Agricole 5 5 5 159 113 66 Foresterie 1 0 0 182 Vétérinaire 5 1 1 137 129 42 Hygiène 0 1 0 79 Laborantins 2 0 1 122 67 Sciences infirmières 4 7 11 263 115 209 Action sociale 1 5 2 54 137 85 Hôtellerie et tourisme 0 0 2 202 Droit 0 2 9 129 100 Secrétariat 1 5 7 33 109 78 Commerce et comptabilité 5 18 38 123 132 110 Informatique 0 0 2 40 Technical Electricité 3 2 5 116 33 44 Electro-mécanique 1 0 1 93 20 Electronique 1 0 1 35 274 Mécanique-générale 2 0 1 79 20 Mécanique automobile 3 0 6 59 89 Construction 3 1 3 85 26 36 Travaux publics et construction 1 1 0 76 159 Plomberie-soudure 0 1 0 20 Menuiserie 1 0 1 57 11 Métal-bois 0 1 1 166 132 Métal-électricité 0 0 2 99 Engins lourds 1 0 0 65 Note: Blank cells denote not applicable. Shaded cells highlight fields in which average enrollments are relatively small. Source: Based on data from Ministry of Education 2000­01 census of secondary schools. In fields where only one school is currently offering a stream, the courses may be new, and the small enrollments may thus consist of single cohorts of students. If so, the menu of course offer- ings appears to be expanding even though the demand is sometimes already being met by exist- ing schools or may be only nascent (e.g., electronique compared with electro-mécanique). 118 WORLD BANK COUNTRY STUDY A more fundamental issue concerns the role of the state in the delivery of vocational and technical education and training. Because these streams are intended to prepare students for the world of work, the production of graduates should ideally be as responsive as possible to labor market signals. Experience from most developing countries suggests, however, that public sector institutions tend to perform poorly in this regard. The weak link reflects in part the inherently feeble incentives for adapting service delivery where a school does not depend for its survival on its ability to attract clients on the basis of its record for successful job placement for its graduates. In Madagascar, for example, most graduates of technical secondary schools end up pursuing courses at the university, often in fields unrelated to their upper secondary school courses, thus wasting the state's investment in their (expensive) training. In light of all this, Rwanda's policymakers might want to give careful consideration to the strategy for expanding vocational and technical education. In doing so, the issue of government financing has to be looked at separately from that of direct government involvement in service delivery. An important argument for government financing is that it serves to advance equity goals, given that students attending vocational and technical streams tend to come from more disadvantaged backgrounds. How strong this argument is in Rwanda's current context is difficult to tell because we have no data on the profiles of students in vocational and technical education. The argument for direct government involvement in service delivery is much weaker, particularly given the well-documented evidence from many low-income countries pointing to the public sector's typical slowness in restructuring course offerings to match fluid developments in the labor market, or even its inability to do so (see, for example, Johanson and Adams forthcoming). Rwanda might find it worthwhile to examine the measures such countries as Côte d'Ivoire, Morocco, and Zambia have taken to achieve their goals in the education sector by combining government finance with nongovernmental delivery of services. Teacher Qualifications, Utilization, and Deployment Teaching staff is the single most costly input in the internal operations of schools. In this section we examine the distribution of teachers by qualification, their teaching workload across school types and subcycles, and the pattern and consistency of their deployment among schools. Although most of the available data pertain to the public sector, information on private schools is included where possible. Teacher Characteristics and Classroom Learning Conditions Women make up less than one-fifth of the teachers in state and private secondary schools and less than one-quarter of those in libre subsidié schools (table 6-6). At least in the public sector, most of the teachers for whom data are available are relatively young, averaging about 5.5 years of teaching experience in state schools and 6.6 years in libre subsidié schools. Throughout the sys- tem, almost all teachers have at least an upper secondary diploma, but there are striking differ- ences in the distribution among school types of those who have only upper secondary diplomas; they account for about 64 percent of teachers in state schools, 60 percent of those in libre sub- sidié schools, and only 48 percent of those in private schools. This distribution is consistent with the bias in state schools toward the tronc commun cycle, where teachers can be expected to have lower qualifications than those in the upper secondary cycle. The majority of public sector teach- ers report having had preservice teacher training; it is not known what share of private school teachers has this qualification. As noted above, a large number of Rwanda's secondary schools offer, under one roof, instruction at both lower and upper secondary levels. In such schools, some teachers teach classes at both levels. According to the data in table 6-7, in the state schools just under a quarter of the teachers fall into this category, whereas the corresponding share is nearly 40 percent in the libre subsidié sector. Teachers who teach both cycles appear to have a slightly heavier teaching load-- about 19 hours per week, compared with the overall average of 18 hours. In contrast to the EDUCATION IN RWANDA 119 TABLE 6-6: CHARACTERISTICS OF TEACHERS BY TYPE OF SECONDARY SCHOOL, RWANDA, 1999­2000 Type of school Teacher credentiala State Libre subsidié Private % women 16.6 24.1 -- Average years of experience 5.5 6.6 -- Teachers' educational attainment (%) Upper secondary Incomplete ES 4, 5, 6 0.8 1.1 -- Diploma (3­5 years) D3, D4, D5 0.8 0.8 1.8 Diploma (6­7 years) D6, D7 64.0 58.8 48.4 Postsecondary 2-year diploma Baccalauréat or 2- 18.7 22.5 33.7 year diploma University degree License, BA, BSc, 12.3 14.9 13.6 master's, PhD Other Includes unknown 3.4 1.8 2.5 % with preservice teacher training 66.9 62.4 -- -- Not available. a. See appendix table A5-1 for a detailed description of the various credentials. Source: For state and libre subsidié schools, Ministry of Education 1999 census of teachers merged with the 1999­2001 census of schools; for private schools, Ministry of Education statistical abstract, 1999­2000. relatively small differences in teaching loads, student-teacher ratios vary significantly across levels of instruction and type of school, implying probably large differences in average class size. We can deduce the magnitude of the differences using the following tautological relation: SH S = TH T CS where SH is the weekly instructional hours received by students, S is number of students, CS is class size, TH is the weekly teaching workload, and T is number of teachers. The left-hand side of the equation represents aggregate instructional hours received by the student body; the right- hand side gives the aggregate hours of teaching offered by the teachers. By rearranging, we have the following relation: CS = SH S =SH STR TH T TH where STR is the student-teacher ratio, derived simply as S/T. Applying the foregoing equation to the data in the "tronc commun only" column in table 6-7, and assuming that students' instructional hours are the same across school types, we can deduce that class sizes in state schools are, on average, about 41 percent larger than in libre subsidié schools [= (26.6/19.3) × (17.8/17.4)]. Thus, if class sizes average about 40 students per section in the libre subsidié schools, they would be as high as 56 students per section in the state schools. 120 WORLD BANK COUNTRY STUDY TABLE 6-7: DISTRIBUTION OF TEACHERS BY LEVEL OF CLASSES TAUGHT,AND TEACHING LOADS AND STUDENT-TEACHER RATIOS, PUBLIC SECONDARY SCHOOLS, RWANDA, 1999­2000 Level of classesa Both tronc Tronc commun Upper commun and upper secondary only secondary only All levels Distribution of teachers by level of classes taught (%) State schools 50.5 23.6 25.9 100 (1,017)b Libre subsidié schools 33.7 38.0 28.3 100 (2,234)b Teachers' average teaching workload per week (hours) State schools 17.4 19.2 18.7 18.1 Libre subsidié schools 17.8 19.0 17.3 18.1 Student-teacher ratioc State schools 26.6 21.3 15.5 23.3 Libre subsidié schools 19.3 22.0 16.6 21.2 a.All calculations in the table count teachers in full-time equivalent units according to their time allocation in each cycle. b. Figures in parentheses indicate the total number of teachers with valid data among the 3,257 enumerated in the 1999 census. c. Refers to school-level averages. For comparison, the student-teacher ratio in private schools averages 24.2. Source: For distribution of teachers by level of classes taught and teaching workload, Ministry of Education 1999 census of teachers, merged with data from the 1999­2000 census of secondary schools; for data on stu- dent-teacher ratios, electronic data files from Ministry of Education 1999­2000 annual census of secondary schools. Using the same approach, we can make a similar comparison of class sizes in the lower and upper secondary cycles. In state schools, class sizes in the tronc commun cycle are, on average, about 84 percent larger than in the upper secondary cycle [= (26.6/15.5) × (18.7/17.4)], whereas in libre subsidié schools the corresponding gap is much smaller, 13 percent [= [19.3/16.6) × (17.3/17.8). Table 6-8 shows the educational qualifications of teachers in state and libre subsidié schools according to the level of classes they teach. Among teachers with only tronc commun classes, almost all have at least an upper secondary school diploma. The share that is university educated is slightly smaller in state schools--just under 10 percent, compared with more than 12 percent in libre subsidié schools. An upper secondary school diploma may be less than ideal for teaching tronc commun classes, but it is nevertheless minimally adequate because the teachers have at least three more years of schooling than their charges. By contrast, the situation in the upper second- ary cycle is alarming: about half the teachers in state schools and more than half of those in the libre subsidié schools have only an upper secondary school diploma themselves. These findings suggest that a clear priority in policy development is to find ways to improve the educational qualifications of upper secondary school teachers. As it may take years to train a sufficient number of teachers to at least two years past the secondary school diploma level, it is important to look at both short-term and long-term measures. One short-term option is to replace incumbent upper secondary school teachers who are underqualified (that is, lack at least a two-year postsecondary diploma) with teachers currently teaching tronc commun classes who are EDUCATION IN RWANDA 121 TABLE 6-8: DISTRIBUTION OF STATE AND LIBRE SUBSIDIÉ SECONDARY SCHOOL TEACHERS BY EDUCATIONAL ATTAINMENT AND LEVEL OF CLASSES TAUGHT, RWANDA, 1999­2000 Level of classes Both tronc Type of school Tronc commun and Upper and teacher qualification commun only upper secondary secondary only Teachers in state schools Number 514 240 263 Share of total (%) Upper secondary diploma 69.0 64.2 49.3 2-year postsecondary diploma 16.9 16.0 25.1 University degree 9.5 12.3 17.9 Other 4.7 2.8 7.6 Teachers in libre subsidié schools Number 752 850 632 Share of total (%) Upper secondary diploma 64.8 57.0 54.4 2-year postsecondary diploma 19.8 23.6 24.2 University degree 11.7 16.7 16.2 Other 3.7 2.7 5.2 Source: Ministry of Education 1999 census of teachers, merged with data from the 1999 census of secondary schools, in which 3,257 teachers with teaching duties were enumerated. at least minimally qualified. Among the teachers who teach classes in either the lower or upper secondary cycle, in 1999­2000 a total of 473 upper secondary teachers lacked the minimum qualifications for teaching in the upper secondary cycle, and 371 tronc commun teachers had those minimum qualifications. Ignoring for the moment the possibility that a complete swap between these two groups of teachers might not work because of mismatches of fields of special- ization, location of work, and other practical constraints, nearly 80 percent of the underqualified teachers currently teaching in the upper secondary cycle could potentially be replaced. The share could conceivably go higher if the swap were extended to teachers who teach both cycles. A legitimate question is how much, on balance, the system would actually gain from the swap, since it would help solve a serious problem in the upper secondary cycle at the possible expense of depressing teaching quality in the lower secondary cycle. We return to this issue later, in a discussion of the relation between teachers' educational attainments and student learning. For the moment, suffice it to say that although the groundwork for long-term solutions needs to be laid through careful expansion of postsecondary education to supply upper secondary teachers with the desired qualifications, the immediate option of rationalizing the deployment of incum- bent teachers also merits consideration. Teacher Deployment across Schools Figure 6-2 hints at substantial variation in the number of teachers across schools of similar size--an unsurprising pattern, given that schools offer different levels and types of instruction. To see how much of the variation persists after controlling for these differences, we turn to regression analy- sis, the results of which are reported in table 6-9. As before, the available data pertain only to schools in the public sector. 122 WORLD BANK COUNTRY STUDY TABLE 6-9: REGRESSION ESTIMATES OF THE RELATION BETWEEN NUMBERS OF TEACHERS AND STUDENTS, PUBLIC SECONDARY SCHOOLS, RWANDA, 1999­2000 Upper secondary cycle Tronc commun cycle Model 1 Model 2 Libre All public Libre All public Libre All public State subsidié schools State subsidié schools State subsidié State Number of 0.040 0.036 0.038 0.048 0.047 0.048 0.035 0.040 0.039 students (8.59)** (8.33)** (12.20)** (9.71)** (13.77)** (16.31)** (7.77)** (10.32)** (13.46)** Number of 2.651 1.423 1.794 streamsa (2.82)** (2.46)* (3.55)** Constant 0.831 2.992 1.983 1.548 1.839 1.676 ­1.561 ­0.025 ­0.556 (0.74) (2.88)** (2.61)** (1.41) (2.25)* (2.46)* (1.08) (0.02) (0.61) Number of observations 53 97 150 30 83 113 30 83 113 R2 0.75 0.50 0.61 0.75 0.52 0.61 0.84 0.55 0.66 Blanks denote not applicable. * Statistically significant at the 5 percent level. ** Statistically significant at the 1 percent level. Note: Figures in parentheses are robust t-statistics. Because some teachers teach in more than one cycle, the number of teachers is counted in full-time equivalent units according to the distribution of each teacher's workload across the two cycles. See appendix table A6-3 for additional results for the combined sample of schools offering either the secondary cycle or both cycles. a.The variable applies only to the upper secondary cycle, where students follow the curriculum in one of four streams: general, normal, vocational, or technical. Source: Based on data from Ministry of Education 1999­2000 census of secondary schools merged with data from the 1999 census of teachers. FIGURE 6-2: RELATION BETWEEN NUMBERS OF STUDENTS AND TEACHERS, STATE AND LIBRE SUBSIDIÉ SCHOOLS, RWANDA, 2000 Note: Each circle represents a school. Source: Based on Ministry of Education 2000­01 census of secondary schools. EDUCATION IN RWANDA 123 FIGURE 6-3: R2 VALUES OF REGRESSIONS RELATING NUMBERS OF TEACHERS AND STUDENTS ACROSS SCHOOLS BY LEVEL OF EDUCATION AND SCHOOL TYPE, RWANDA, CIRCA 2000 Source: For primary education, table 5-5 in chapter 5; for tronc commun and upper secondary education, table 6-9. The regression analysis is carried out separately for the tronc commun and upper secondary cycles. For the purpose of this exercise, teachers who teach classes in both the lower and upper secondary cycles are counted in full-time equivalent units according to the distribution of their teaching workload between the two cycles. For upper secondary education, we ran two sets of regressions, one without controlling for the number of streams of instruction offered and a sec- ond set with that variable included. The coefficients on the number of students are statistically significant and positive on all the regressions, indicating that teacher allocation is generally linked to enrollments, and their values span a narrow range, from 0.035 in the regression for the upper secondary cycle in state schools to 0.40 in the regression for the tronc commun cycle in state schools.5 These values imply that teachers are supplied at the average rate of one teacher per increase in enrollment of 25­29 stu- dents. (At the primary level, the rate is 1 teacher per additional 63 pupils.) Consider next the R2 values of the regressions. Those for the state school regressions are consistently larger than the corresponding values for the libre subsidié schools. This result signifies that although both types of school follow comparable teacher allocation rules, the application of such rules at the level of individual schools is much less consistent in the libre subsidié sector. Figure 6-3 presents the results graphically and includes those for primary education from chapter 5 to provide an overall perspective. Interestingly, the inconsistency in teacher allocation across libre subsidié schools is less of a problem at the primary than at the secondary level. Even so, the fact that R2 values are nowhere more than around 0.85 indicates that there remains room for improvement throughout the system. 5. The model 2 results are used because they control for field of study. 124 WORLD BANK COUNTRY STUDY FIGURE 6-4: RELATION BETWEEN NUMBER OF STUDENTS AND COST PER STUDENT, TRONC COMMUN AND UPPER SECONDARY CYCLES, RWANDA, 1999 Note: FRw, Rwandan francs. Each circle represents a school. Source: Based on Ministry of Education 1999­2000 survey of secondary schools merged with the 1999 census of teachers. Economies of Scale in Service Delivery Beyond ensuring consistency in teacher allocation, another challenge in the management of the secondary education sector is to take advantage of economies in service delivery where these exist. Such economies occur when a school's total cost rises less rapidly than the increase in stu- dent enrollments. Certain fixed costs such as remuneration of specialized teachers and of adminis- trative and support staff at the school level are spread over the rising enrollments.6 As a result, the cost per student of service delivery can be expected to fall as enrollments increase. Overall Pattern Figure 6-4 shows a simple scatter plot of the relation between unit costs and size of enrollments across schools. The scatter is visually even more dispersed than that in figure 6-2 relating number of teachers and enrollments. As at the primary level, the pattern implies that in addition to teacher numbers, differences in the availability of administrative personnel and in the educational profiles of the teachers and the administrative staff tend to exacerbate rather than to narrow dif- ferences in the allocation of resources across schools. Regression Estimates of the Relation between Unit Costs and Enrollments The results shown in table 6-10 suggest that given the way the system currently operates, signifi- cant scale economies in service delivery exist in the tronc commun cycle: at the sample mean; total personnel costs rise by 6.2 percentage points for a 10 percentage point rise in enrollments; and conversely, a 10 percentage point increase in spending on personnel would allow enrollments to 6. For our purpose here, we include the cost of teachers, as well as administrative personnel. Again using data from the 1999 census of teachers, we apportion the cost of teachers to the tronc commun or upper secondary level according to the distribution of each teacher's teaching workload across the two levels, and we apportion the cost of administrative staff according to the share of enrollments in each cycle. EDUCATION IN RWANDA 125 TABLE 6-10: RELATION BETWEEN TOTAL COST OF PERSONNEL AND ENROLLMENTS IN PUBLIC SECONDARY SCHOOLS, RWANDA, 1999­2000 Tronc commun cycle Upper secondary cycle State Libre All public State Libre All public schools subsidié schools schools subsidié schools Number of pupils (log) 0.60 0.60 0.62 0.96 0.93 0.95 (6.06)** (7.43)** (9.64)** (8.74)** (10.71)** (14.33)** Dummy variable for ­0.45 0.07 ­0.20 0.117 0.257 0.123 single-cycle school (4.03)** ­0.74 (2.90)** ­0.62 ­0.78 ­0.71 Constant 12.34 12.26 12.13 10.31 10.66 10.50 (21.74)** (27.29)** (33.30)** (17.27)** (23.14)** (29.54)** Number of schools 54 98 152 31 84 115 R2 0.65 0.37 0.48 0.81 0.69 0.73 Memoranda: Economies of scalea 1.67 1.67 1.61 1.04 1.08 1.05 Average enrollment per school 290 301 296 260 249 253 * Statistically significant at the 5 percent level. * * Statistically significant at the 1 percent level. Note: See appendix table A6-3 for regression results on the relation between numbers of teachers and second- ary students in both cycles across schools. In all the regressions, the dependent variable and the number of students are expressed on a log scale. Num- bers in parentheses are robust t-statistics. Where teachers teach more than one cycle, the cost attributed to each cycle is prorated in proportion to the time allocation of the concerned teachers across the two cycles. Because some teachers teach in more than one cycle, the cost of teachers is computed by apportioning it according to the distribution of each teacher's work- load across the two cycles. a. Refers to the percentage increase in the number of students enrolled for a 1 percent increase in spending on personnel, calculated as the inverse of the coefficient estimate on the independent variable (number of stu- dents). Source: Based on Ministry of Education 1999­2000 census of secondary schools merged with the 1999 census of teachers, as well as data on the salary scale of school personnel. grow by 16.1 percent.7 The magnitude of the economies is comparable across state and libre subsidié schools. In schools that combine instruction in the lower and upper secondary cycles under one roof, the personnel costs tend to be higher, particularly in the tronc commun cycle in state schools, as suggested by the statistically significant and negative coefficient on the dummy variable denoting a single-cycle school.8 The regression results in table 6-10 suggest that economies of scale are also present in the upper secondary cycle. Their magnitude, however, is modest: an increase of 10 percent in 7. The analysis should ideally examine the possible existence of economies of scope in addition to that of economies of scale. We did not implement the latter analysis because the samples involved are small and such an analysis would be inappropriate, given the rough quality of the available data. 8. One explanation is that administrative staffing coefficients are higher for the upper secondary than for the lower secondary cycle. Because these costs are, by construction, spread between the two cycles according to the distribution of enrollments between the cycles, total personnel costs in the tronc commun cycle would tend to be higher in schools with combined cycles. 126 WORLD BANK COUNTRY STUDY FIGURE 6-5: SIMULATIONS OF ECONOMIES OF SCALE IN SECONDARY EDUCATION, TRONC COMMUN AND UPPER SECONDARY CYCLES, RWANDA, 1999­2000 Note: FRw, Rwandan francs. Source: Simulated from the regressions in table 6-10. spending on school personnel would allow enrollments to rise by only about 10.5 percent, on average. As at the lower secondary level, the pattern of scale economies is comparable across types of school. One possible explanation for the near absence of scale economies can be found in the wide menu of course offerings in the upper secondary cycle that was documented earlier in this chapter. Typically, schools expand their enrollments by diversifying their curriculum offerings rather than by increasing the scale of existing offerings, and so they do not benefit from economies of scale in service delivery. In addition, the average enrollment in both the tronc commun and upper secondary cycles is relatively small--no more than about 300 students, on average, per cycle at each school. These small sizes stand in sharp contrast to the current average of about 700 students per primary school. The regression results are represented visually in figure 6-5. For reasons noted above, economies of scale are present in the tronc commun cycle but are largely absent in the upper sec- ondary cycle. For the tronc commun cycle, these economies are particularly apparent as the size of enrollments increases to about 400 students. Taking Advantage of Scale Economies in Service Delivery In chapter 5 the absence of scale economies in primary education led us to suggest that smaller schools may be no more costly to operate than larger schools and would have the advantage of bringing schools nearer to pupils' homes, thus facilitating educational access. In the tronc com- mun cycle, the opposite problem exists: most of the schools offering this cycle enroll too few students and are therefore uneconomic to operate. Because students at the tronc commune level are older and better able to travel longer distances to school, the argument of accessibility is weaker for this level. Table 6-11 shows estimated unit costs at different enrollment sizes and the current size distri- bution of secondary schools in Rwanda. Scale economies are more obviously present in the tronc commun than the upper secondary cycle. Unit costs at enrollments of more than 400 tronc com- mun students are less than two-thirds those for enrollments of between 100 and 200 students and about four-fifths those for between 200 and 300 students. Yet more than a fifth of the stu- dents are currently enrolled in schools serving between 100 and 200 tronc commun students, and more than a third are in schools with enrollments of between 200 and 300. If all the small EDUCATION IN RWANDA 127 TABLE 6-11: SIZE DISTRIBUTION OF ENROLLMENTS IN PUBLIC SECONDARY SCHOOLS AND SIMULATED COST PER STUDENT, RWANDA, 1999­2000 Tronc commun cycle Upper secondary cycle Average cost Size distribution Average cost Size distribution per student of enrollments per student of enrollments Size of enrollment (thousand FRw)a (%) (thousand FRw)a (%) <100 29.7 18.6 100­200b 28.2 21.1 28.0 25.4 200­300 23.2 35.5 27.3 22.0 300­400 20.5 23.5 26.9 19.5 400+ 18.6 19.9 26.4 14.4 Number of schools offering the cyclec 166 118 Blanks denote not applicable. Note: FRw, Rwandan francs. a. Unit cost in the 400+ categories is simulated, for the tronc commun level, for an enrollment of 450. Unit costs in the <100 and 400+ categories are simulated, for the upper secondary level, for enrollments of 50 and 500 students, respectively.These are, roughly, the average sizes of enrollments per institution in the size category. For the other size categories, the costs are simulated for schools in the middle of the size bracket. b. For the tronc commun cycle, the range is <200. c. Includes schools offering only one cycle or both the tronc commun and upper secondary cycles. Source: Authors' simulations based on the regressions in table 6-10. schools enrolled at least 400 tronc commun students each, the average unit cost of delivering services would fall by about 20 percent--arguably a nontrivial decline, considering the overall scarcity of resources and the fact that the costs of personnel who deliver lower secondary services are currently 3.5 times the corresponding costs at the primary level. The foregoing results have obvious implications for management of the size distribution of secondary schools as more and more children complete their primary schooling and the second- ary sector expands. In a growing system such as Rwanda's, taking advantage of scale economies means avoiding a proliferation of schools catering to small cohorts of students. This result can be achieved in many ways, depending on conditions on the ground. In some localities it might be appropriate to combine enrollments in nearby schools that currently offer lower and upper sec- ondary instruction to small enrollments in both cycles. In other settings it might be simply a matter of raising enrollments to accommodate the growing clientele for lower secondary school- ing. In communities where no school currently delivers lower secondary schooling, it might be appropriate to start entirely new schools if the demand is sufficient to allow the school to fill rapidly at least 400 places in the tronc commun cycle. Unlike lower secondary education (which most educators increasingly view as an extension of primary schooling), the expansion of places in the upper secondary cycle needs to proceed more slowly, if only because labor market considerations become increasingly relevant. This issue is explored in more detail in chapter 7, but for the present discussion, the important implication is that enrollments at the upper secondary level are likely to remain relatively small for the foreseeable future. Although, given the results presented above, small enrollments do not appear uneconomic, the fragmentation of course offerings is a problem and may be an important reason for the absence of scale economies. Policies for managing the expansion of this level of schooling, particularly in the public sector, might thus seek specifically to avoid a proliferation of course offerings while at the same time taking advantage of opportunities for consolidation where appropriate. 128 WORLD BANK COUNTRY STUDY Examination Results and Their Correlates A final aspect of service delivery relevant to policy development is the relation between the resources that schools receive and the output in terms of student learning. As at the primary level, the analysis is severely handicapped by lack of appropriate student assessment data. Here again we use the results on the national examination administered at the end of the tronc com- mun cycle as a proxy measure for student learning in lower secondary schooling, and we perform the analysis using schools, rather than individual students, as the unit of observation. This approach suffers from well-known flaws, but it is the best that can be implemented under the circumstances. Although a national examination is also administered at the end of the upper sec- ondary cycle, we do not attempt to analyze the results, in part because of the small number of schools involved and the large number of fields of specialization at that level. Results on the National Tronc Commun Examination The results of the tronc commun examination are expressed in terms of the pass rate and the average score, which can range from zero to a maximum value of X. Because places in the upper secondary cycle are limited, the pass rate is adjusted from year to year to let through the appro- priate number of students. This practice makes it difficult to track trends in the performance of schools on the examination but does not affect the comparison across schools at a single point in time. The results for the 278 schools that fielded tronc commun candidates for the 1999 national examinations appear in table 6-12. A clear dichotomy is seen in the performance of students attending public and private schools. In the public sector, students in libre subsidié schools achieved a mean pass rate of 72 percent, compared with 75 percent in state schools; these high percentages contrast with an average of only 47 percent in private schools. Although the average for libre subsidié schools is lower than that for the state schools, no school in the libre subsidié group had a pass rate below 12.5 percent, whereas the weakest state school had a pass rate of only 1.6 percent--worse than the pass rate of 2.1 percent in the weakest private school. In each school category, at least one school achieved a pass rate of 100 percent. The ranking by school type changes somewhat when performance is measured using average scores instead of pass rates: both state and libre subsidié schools outperform private schools, but TABLE 6-12: RESULTS ON THE NATIONAL EXAMINATION AT THE END OF THE TRONC COMMUN CYCLE, RWANDA, 1999­2000 Pass rate (%)b Average scorec Number Type of school of schoolsa Mean Min. Max. Mean Min. Max. State 44 75.2 1.6 100.0 3.1 1.8 4.6 Libre subsidié 99 72.0 12.5 100.0 3.4 1.7 5.2 Private 135 47.2 2.1 100.0 2.3 0.9 6.1 All schools 278 60.4 1.6 100.0 2.8 0.9 6.1 a.The sample includes all secondary schools that fielded candidates for the 1999 national examination at the end of the tronc commun cycle.The pass rate is based on the number of schools; the average score is based on one fewer school each in the categories for libre subsidié and private schools. b. Refers to the percentage meeting the requirement for entry to the upper secondary cycle. c.The scoring system uses a scale that translates letter grades into numerical scores, with A scored at 11 points, E scored at 0, and letter grades in between scored progressively lower by 1 point. Source: Computed from unpublished school-level data supplied by the National Examination Council of Rwanda. EDUCATION IN RWANDA 129 FIGURE 6-6: RELATION BETWEEN PERSONNEL COSTS PER STUDENT AND RESULTS OF NATIONAL EXAMINATION AT END OF THE TRONC COMMUN CYCLE, PUBLIC SECONDARY SCHOOLS, RWANDA, 1999­2000 Note: FRw, Rwandan francs. Each circle represents a school. Source: Based on data supplied by the National Examination Council of Rwanda, merged with Ministry of Edu- cation 1999 census of teachers. state schools now rank behind libre subsidié schools. The widest range of scores--from a low of 0.9 to a high of 6.1--is for the private schools, implying that the sector had the worst as well as the best schools in the system. In the public sector the worst school was equally likely to be a state or a libre subsidié school, but the top performer was in the libre subsidié group. Regression Analysis of the Correlates of Examination Performance Data for the analysis of performance on examinations were prepared by merging school-level data files supplied by the National Examination Council of Rwanda. Files containing information on each school were based on the census of secondary schools in 1999­2000, and those containing information from the 1999 census of teachers were aggregated to the school level. This proce- dure produced a dataset of 137 public sector (state and libre subsidié) schools. Private schools were eliminated because information was lacking on unit costs; some public sector schools were also dropped because of missing data. Figure 6-6 provides an overview of the relation between performance on the examination and the cost of personnel per student. As in primary education, the patterns shows substantial disparity in two respects: (a) consistent with what has been documented earlier, spending per student varied widely, from FRw 10,000 to as much as FRw 40,000, and (b) for schools with similar levels of funding for personnel, performance on the national examinations was highly dis- persed. For schools with per student spending on personnel of about FRw 20,000, the pass rates ranged from about 20 to 100 percent. The full results of the regression findings appear in appendix table A6-4; highlights are pre- sented in table 6-13. A key finding is that the coefficient on spending per student is not statisti- cally significant, confirming what was hinted at in figure 6-6. One interpretation is that school performance on the national examinations depended more on how resources were used than on the amount of resources schools received. To look into the issue more closely, we also regressed examination results against some aspects of the learning environment. Because of data limitations, we were able to consider only two such aspects: the student-teacher ratio and teachers' educa- tional attainment. Table 6-13 contains findings on the relation of these variables to student per- formance on the national examinations. 130 WORLD BANK COUNTRY STUDY TABLE 6-13: CORRELATES OF SCHOOL-LEVEL PERFORMANCE ON THE NATIONAL EXAMINATION AT THE END OF THE TRONC COMMUN CYCLE, RWANDA, 1999­2000 Dependent variable and regression estimates Pass rate (%) (sample Sample of Score mean = 73.0) mean of (sample regression Marginal mean = 3.3), Regression variable variablea Coefficient impactb coefficient Student-teacher ratio Reference group: <18 23.6 18­27 48.5 -0.47 -6.9 -0.28* >27 27.9 ­0.93* -13.7 -0.22 Teacher qualifications Reference group: university degree (license, BA, BSc, master's, PhD) 8.4 Postsecondary (baccalauréat or 2-year university diploma) 14.9 -0.04 -0.55 -0.004 Upper secondary (D6, D7) 75.0 -0.03 -0.39 -0.009 Other 1.7 0.07 1.06 0.015 % teachers with preservice teacher training 67.9 0.01 0.13 -0.004 R2 0.20 0.31 Number of observations 136 137 Blanks denote not applicable. * Underlying coefficient estimate for the variable is statistically significant at the 5 percent confidence level. Note: See appendix table A6-4 for the full regression results and the full list of variables. a. Corresponds to percentages in the indicated category. b.The regression model follows a log-logit specification: ln[y/(1 - y)] = bX. The marginal effect on the stu- dent-teacher ratio refers to the percentage point change in the pass rate in response to switching the indicat- ed ratio for the ratio in the reference group.The marginal effect on the teacher qualifications variable refers to the percentage point change in the pass rate in response to a 1 percentage point increase in the share of teachers with the corresponding qualification at the expense of a 1 percentage point decline in the share of teachers with the reference qualification (a university degree). Source: Authors' estimates based on school-level data on examination results for 1999 merged with data from Ministry of Education 1999­2000 census of primary schools and 1999 census of teachers. The negative coefficient estimates for the relation between pass rates and the student-teacher ratio imply that, relative to students attending schools averaging fewer than 18 students per teacher, those in schools with higher ratios tended to fare worse on the examination. The detri- mental impact was statistically significant if the student-teacher ratio exceeded 27. For students in such schools, the pass rate was, on average, 13.7 percentage points lower than the sample pass rate of 73 percent. The negative relation persisted when the pass rate was replaced by examina- tion score as the dependent variable, but the regression coefficient was statistically significant only for the dummy variable corresponding to a student-teacher ratio in the 18­27 range. These results suggest that high student-teacher ratios probably compromised student learning as meas- ured by examination results, but the sample size and the quality of the data were such that it was not possible to determine with any precision the exact thresholds above which the negative effects begin to take hold. EDUCATION IN RWANDA 131 We consider next the findings on the relation between teacher qualifications and examination results. Students taught by teachers without a university degree did worse, as the negative coeffi- cient estimates suggest, but none of the coefficients is statistically significant. The result remained the same regardless of which dependent variable was used. These results do not mean that teacher attainment is never an important factor in determining students' academic performance. Rather, they imply that in the current context of Rwanda's secondary schools, other factors, including, perhaps, overcrowding in some schools and lack of adequate learning materials and other facilitat- ing conditions, are probably an even more binding constraint on performance. Policy Implications Taken together, the findings in this chapter highlight the following issues for policy discussions: Developing separate expansion strategies for the tronc commun and upper secondary cycles Managing the cost of expansion Ensuring adequate conditions for effective learning in both cycles of the subsector. Developing Separate Expansion Strategies for the Two Secondary Cycles In Rwanda the tronc commun cycle offers a standard curriculum to all students, while the upper secondary cycle consists of four streams within each of which additional fields of specialization are offered. The system's current design is thus consistent with the view that the lower secondary cycle is part of basic education--in other words, an extension of primary schooling to complete a child's preparation for adulthood. By contrast, the mission of upper secondary education is to prepare students for eventual jobs at the relatively skill intensive end of the labor market, with some students reaching these jobs only after going on to higher education. Acceptance of this distinction between the two cycles implies that the expansion of each would need to be managed very differently. Enrollment growth in the lower cycle would have to keep pace with the rising number of primary school leavers seeking to complete additional years of schooling, whereas growth in the upper cycle would need to be calibrated to the demand for educated labor with significant educational attainment. As in most modernizing agricultural economies, the demand for such labor in Rwanda is likely to expand only moderately fast at best, so it is important to ensure that the numbers admitted to the upper secondary cycle are kept consistent with market conditions. The volume of graduates produced will be a less relevant crite- rion of success than production of graduates who leave upper secondary school with the right skills to succeed in the next phase of their lives. Those going on to higher education need a solid foundation for further studies. Those who will take jobs need skills that will meet employers' demand for skilled labor and enable the graduates to perform their work well. Managing the Cost of Expansion Two important characteristics of the secondary school system make both cycles costly to operate at present. First, almost all secondary schools in the public sector (state and libre subsidié schools) provide boarding facilities. Although some schools have begun to enroll day students, the fiscal burden of expansion is likely to be untenable unless most, if not all, of the future increase in enrollments, particularly at the tronc commun level, is accommodated in day schools. Second, many schools currently enroll too few students to benefit from scale economies in service delivery. At the tronc commun level, more than half of the public sector schools enroll fewer than 300 students each, whereas economies of scale materialize mostly at enrollments beyond at least 400 students. Rationalizing the size distribution of schools in an expanding sys- tem is relatively straightforward because it rarely necessitates collapsing existing schools into sin- gle institutions. What is required is a plan for expansion that (a) raises enrollments in the tronc commun cycle in existing schools where enrollments are currently too small and (b) ensures that new schools are opened only in areas with sufficiently large catchments of potential tronc commun students. Adding on tronc commun classes to existing primary schools is a possibility, but that 132 WORLD BANK COUNTRY STUDY arrangement calls for careful evaluation, since most primary schools are already very large and the benefits of bringing the two cycles under one roof are unclear. Indeed, if the experience from the current practice in secondary education of combining the tronc commun and upper secondary cycles in one school is any guide, the result may well be to encourage an uneconomic size distri- bution of enrollments at both levels. At the upper secondary level, too, school enrollments tend to be small, and the fact that schools seek to supply a wide menu of course offerings only exacerbates the problem. Possible approaches to managing costs in such a system include: (a) reducing duplication in course offer- ings across schools, (b) consolidating related fields of specialization where feasible, and (c) accommodating future growth in enrollments in existing schools rather than building new ones. A related cost-saving measure is to consider shortening the duration of some courses. Currently, all courses last three years, but in some fields, especially in the vocational stream, on-the-job learning may well be more effective than classroom instruction in helping students succeed in their future jobs. Ensuring Adequate Classroom Conditions for Effective Learning Our findings point to at least three constraints that may be compromising the classroom learning environment. Class sizes, particularly in state schools, may be too large. A recent policy of increasing the weekly teaching workload to a minimum of 25 hours (from about 17­19 hours in 1999) may have helped ease the problem. But given the small enrollments across schools and the fragmenta- tion in course offerings, it is unclear how well the policy is being implemented and how effective it has in fact been in ensuring better allocation of teachers' time. A related constraint has to do with the consistency of teacher allocation across schools. Although the number of teachers allocated is positively related to the size of enrollments, the link is not especially close, implying that teachers are more available in some schools than in others. The unequal availability of teachers is a worse problem among libre subsidié schools than among state schools, which suggests that tighter management of teacher allocation across those schools may warrant special attention. The issue of teacher qualifications cannot be ignored. In the tronc commun cycle, most teach- ers satisfy the minimum qualification--at least an upper secondary school diploma--and some have even earned a university degree. By contrast, about 50 percent of the teachers in the upper secondary cycle, in both state and libre subsidié schools, are probably underqualified, having only an upper secondary diploma themselves. Any strategy for improving learning conditions must seek to raise the educational profile of teachers in the upper secondary cycle. The obvious long- term solution is to set clear standards for teacher recruitment, but to solve the problem more quickly, policymakers might consider rationalizing the current allocation of the most qualified teachers (those with postsecondary qualifications) between the tronc commun and upper second- ary cycles. Conclusion Secondary education in Rwanda is in many ways at a crossroads today. The rising tide of primary school leavers is already creating pressures to increase places in secondary schools. As a result, enrollments in both the public and private sectors have grown at an extraordinary pace, averaging some 20 percent a year since 1996. Secondary education is slowly being transformed from a sys- tem serving a small, elite clientele into one that will increasingly be expected to cater to the mass- es. The challenge is to ensure that appropriate policies are put in place so that the system can expand along an efficient and equitable path. In order to assess where and how the government might usefully intervene, this chapter has documented some key characteristics of the secondary education sector and the current arrange- ments for service delivery. The findings highlight three areas for attention by policymakers: EDUCATION IN RWANDA 133 development of separate strategies for the future growth of the lower and upper cycles; management of the cost of expansion by taking advantage of scale economies in service delivery and minimizing fragmentation of course offerings; and management of teacher deployment and teaching assign- ments to promote environments conducive to learning. To be sure, the range of problems in the education sector that policymakers face is broader than this. But these three issues cannot be ignored. They will need to be tackled because of their impact on costs and therefore on the success of the government's efforts to extend educational opportunities in secondary education in a fiscally sustainable manner while ensuring that services are delivered equitably and with maximum value for money. CHAPTER 7 HIGHER EDUCATION H igher education in Rwanda expanded rapidly in the postgenocide years. Since 1997, three new public institutions have been created, the number of government-sponsored students has risen nearly 250 percent, and the public budget for the subsector has grown by a massive 340 percent, to a current level of almost FRw 12 billion (nearly US$27 million). The subsector currently serves a small population but absorbs almost 40 percent of the country's current spending on education. Given the obvious imbalance, it is no surprise that the future development of higher education has been the subject of recent policy documents tabled for dis- cussion by the Cabinet.1 In this chapter our purpose is to consolidate the available quantitative information on higher education so as to illustrate in a concrete manner the nature of some of the key challenges and to enrich the factual basis for policy discussion. Because the system for data collection is still being rebuilt following the genocide, much of the information presented below was gathered directly from the institutions themselves for this study. Although care was taken to ensure consistency and reliability in the data at the time they were gathered, some information may need to be updated in light of ongoing improvements in data collection. Overall, however, the potential flaws in the data are not so serious or so pervasive as to alter the thrust of the policy issues to which they call attention. 1. Specifically, Rwanda's donor partners have sponsored the preparation of three policy documents: "Draft Policy Framework 2000," which examines the changing nature of higher education in general and the major challenges facing the country; "Proposed Higher Education and Continuing Education Bil/The Higher and Continuing Education Act 2002," which sets forth new arrangements covering all aspects of planning, management, governance, and funding of higher education; and "Organization of Research, Sci- ence, Technology, Higher, and Continuing Education," which outlines an organizational structure to ensure that the Ministry of Education can effectively fulfill its proposed roles and functions in higher education. 135 136 WORLD BANK COUNTRY STUDY The treatment here does not pretend to be exhaustive. It focuses on the following areas: the structure of the system and the institutional composition of enrollments; the distribution of enrollments by level and field of study; government sponsorship for study abroad; staffing pat- terns at the institutional level; the cost of service delivery and of student finance; and the pattern of student flow and output of graduates. The analysis highlights several areas that warrant attention if the sector's high costs are to be brought under control: 1. The current system of blanket subsidies under which practically all students who gain entry into public institutions automatically receive a bursary and access to subsidized welfare services. The more than fourfold increase in enrollments in the public sector in the past decade has made this system increasingly unsustainable. 2. Government sponsorship of study abroad, which accounted for a quarter of current spending on higher education. The cost of such studies is high, and local options are increasingly available, particularly for the undergraduate courses that most recipients of scholarships for study abroad are in fact pursuing. 3. Cost management in public institutions. The generous staffing ratios in the public institu- tions, in comparison with private schools, are one reason why the cost of service delivery is so high in the public sector and why so little funding appears to be left for nonpersonnel inputs to support effective teaching. 4. Management of student flow. Issues that should be addressed include not only patterns of dropout and repetition but also the overall size of the intake of students and the distribu- tion of students by field of study. Structure of the System Higher education in Rwanda today is made up of a diverse mixture of public and private institu- tions. In this section we examine the growth of the sector since colonial times and document its current institutional composition and the distribution of enrollments by field of study. Historical Context Table 7-1 shows the number of public and private institutions and the size of their enrollments, beginning in the 1960s. The first institution of higher education, the diploma-granting Grand séminaire de Nyakibanda (GSN), was established in 1936 by the Roman Catholic Church specifi- cally to train men for the priesthood.2 For most of the next three decades, the seminary was the only institution of higher education in Rwanda, and students who were not training for the priesthood pursued their higher education in the present-day Democratic Republic of Congo or in Belgium. The situation changed in 1963 with the founding of the Université nationale du Rwanda (UNR; in English, National University of Rwanda, NUR) under a joint agreement between the Rwandan government and the Dominican Fathers from Québec Province, Canada.3 In 1966 the Institut pédagogique national (IPN; in English, National Institute of Education) was established with support from the United Nations Development Programme (UNDP) and the United Nations Educational, Scientific, and Cultural Organization (UNESCO); its mission was to train secondary school teachers and conduct research in education. The private Faculté de théolo- gie de Butare (FTB) was established in 1969. The 1970s saw no change in the number of public institutions, but in the private sector the Institut Africain et Mauricien de statistiques et d'économie appliquée (IAMSEA) was established 2. See Mazimpaka and Daniel (2000) and Linguyeneza (2001). The latter gives an account of the pio- neering work of the Roman Catholic Church at all levels of education. 3. Although the UNR is a public institution, until 1974 it was run by the Dominican Fathers (Linguyeneza 2001). At that time the university consisted of the Faculty of Medicine, the Faculty of Eco- nomics and Social Sciences, and the Teacher Training College (Ecole normale supérieure). See NUR (2002a) for additional historical information. EDUCATION IN RWANDA 137 TABLE 7-1: NUMBER OF INSTITUTIONS OF HIGHER EDUCATION, OVERALL ENROLLMENTS, AND SHARE OF STUDENTS IN PRIVATE SECTOR, RWANDA, 1960S­PRESENT Number of institutionsa Enrollmentsb Period Public Private Total (thousands) % private 1960s Pre-1963 0 1 -- -- Post-1963c 2 2 -- -- 1970s 2 3 1 -- 1980sd 3 6 1­3 8­28 1990s Pregenocidee 3 7 3­5 35 Postgenocide 6 6 5­11 6­28 2000­present 6 6­8 12­17 30­38 -- Not available. a. Refers to the number of institutions at the end of each period. Four institutions for which information on date of founding is unavailable are excluded: in the public sector, the Ecole supérieure militaire and Centre de formation des adjoints techniques de la statistique, and in the private sector, the Grande séminaire de Kabagyi and Centre d'enseignment supérieur de Kigali. Enrollments in these institutions are, however, likely to be very small. b. Ranges indicate the counts at the start and end of the period. c.The Université nationale du Rwanda was established in 1963­64.The Institut pédagogique national was founded in 1966­67 and merged with the UNR in 1981­82. d. In addition to the UNR, institutions operating in this period were the Institut supérieur de finances publiques (ISFP), founded in 1986­87, and the Institut supérieur d'agronomie et d'élevage (ISAE), founded in 1988­89. e.The genocide took place in April 1994. Source: Authors' compilation based on data in appendix table A7-1. in 1976.4 In the following decade the IPN was incorporated into the UNR, in 1981, and two new public institutions were established: the Institut supérieur de finances publiques (ISFP), in 1986­87, and the Institut supérieur d'agronomie et d'élevage (ISAE), in 1988­89. In the private sector, the number of institutions doubled, from three to six, with the addition of the Université adventiste d'Afrique centrale (UAAC) in 1984, the Ecole supérieure de gestion et d'information (ESGI) in 1985, and the Institut supérieur catholique de pédagogie appliquée de Nkumba (ISCPA) in 1986. The biggest change in the structure of higher education took place in the 1990s, a decade marked by explosive growth in the public sector and large-scale replacement of existing private institutions by new ones. The number of public institutions grew from three to six with the founding of the Kigali Health Institute (KHI) in 1996, the Kigali Institute of Science, Technology and Management (KIST) in 1997, and the Kigali Institute of Education (KIE) in 1999.5 In the 4. New faculties were established at the UNR during this decade: the Faculties of Law and of Applied Science in 1973, and the Faculty of Agronomy in 1979. 5. Apart from the increase in the number of public institutions, significant changes also occurred at the UNR. The School of Modern Languages (Ecole pratique des langues modernes, EPLM), which was closed in 1973, was reopened in 1994; the School of Information Sciences and Technology (Ecole des sciences et tech- niques de l'information, ESTI) was established in 1996 (Kereni 2002; NUR 2002b); and the Faculties of Science and of Applied Science merged in 1998 to form the Faculty of Science and Technology (Bonfils 2002). 138 WORLD BANK COUNTRY STUDY private sector the Institut supérieur de pédagogie de Gitwe (ISPG) was established in 1993, only to close the following year. After the 1994 genocide, three more private institutions--the IAMSEA, the ESGI, and the ISPCA--went out of existence. The FTB and the UAAC reopened in 1996 and the ISPG in 1997. Two new private sector institutions were established, the Univer- sité libre de Kigali (ULK) in 1996, and the Université laïque de Kigali (UNILAK) in 1997. Over- all, the number of private institutions dropped from seven before the genocide to six afterward. The number of students and their distribution between public and private schools tracked the changes in the number of institutions. In the 1960s the system as a whole enrolled perhaps 100 students; the UNR, for example, began with only about 50 students. In the 1970s total enroll- ment stood at about 1,000, but by the end of the 1980s, it had risen to 3,000. Growth during the 1980s was fueled mainly by the establishment of private institutions, and the share of students in the private sector grew during the decade from just 8 percent to 28 percent. Further expansion of the system in the early 1990s saw public and private enrollment rising to about 5,000. The private sector's share rose to about 35 percent just before the 1994 genocide. In the postgenocide 1990s, total enrollments more than doubled, reaching 11,000 by the end of the decade. The private sector share of enrollments fell in the years immediately following the genocide, when several institutions closed, but began recovering as new institutions were established. In the early 2000s enrollments continued to grow throughout the system, reaching nearly 17,000 by 2001­02. The private sector grew even faster; its share of enrollments in that year was a historical high of 38 percent. Institutional Composition of Enrollments Table 7-2 shows that in 2000­01, the most recent year for which complete data are available, enrollment in higher education was concentrated in a small number of institutions. In the public sector the UNR accounted for 56 percent of total enrollments, followed by the KIST, with 18 percent, and the KIE, with 11 percent. The remaining 15 percent of public sector students was distributed among three other institutions. The degree of concentration was even greater in the private sector: the ULK enrolled 81 percent of the students, while the UAAC accounted for 11 percent, the UNILAK for 7 percent, and the ISPG for only 1 percent. To the extent that trends in the UNR apply to the system as a whole, women's access to higher education appears to have improved with time (table 7-2). In the mid-1980s women accounted for less than 18 percent of total enrollment in the UNR. By around 2000, their share had risen to more than 23 percent, and their share in the 10 main public and private institutions taken together was about 33 percent.6 The average, however, masks significant differences in the two sectors: although women accounted for nearly half of enrollments in the private institutions, their share was only 26 percent in the public sector. In the private institutions the shares are rather consistent, in the 49 to 56 percent range, but they vary significantly in the public sector, from a low of 18 percent at the ISAE to 49 percent at the KHI. These differences may be attrib- utable to the stiff competition for places in the heavily subsidized public sector, to the tendency of girls to perform less well than boys on the national examination taken at the end of the secondary cycle, and to the curricula of the institutions; the KHI offers the traditionally female-dominated course in nursing.7 The results of the national examination serve as a basis for selection into the public institu- tions. The private institutions are less selective (presumably, given a minimum level of competence) 6. The rise in women's share in total enrollment mainly stems from growth of enrollments in private institutions. Their share of enrollment in the public institutions appears to have leveled off since the mid- 1990s (table A7-2). 7. See table 4-8 in chapter 4 for data on the differences in examination results between boys and girls on the national examination. EDUCATION IN RWANDA 139 TABLE 7-2: INSTITUTIONAL DISTRIBUTION OF STUDENTS IN HIGHER EDUCATION, SHARE OF FEMALE STUDENTS,AND SCORES OF ENTRANTS ON NATIONAL EXAMINATION AT END OF SECONDARY CYCLE, RWANDA, 2000­01 Average score on national examination of % of students identifying students in % female institution as first Sector and name of institution subsectora Mid-1980s Circa 2000 choiceb Publicc Université nationale du Rwanda (UNR) 55.5 17.7 23.4d 3.6 Kigali Institute of Science,Technology and Management (KIST) 18.2 -- 28.8d 3.5 Kigali Institute of Education (KIE) 11.0 -- 29.3d 3.4 Kigali Health Institute (KHI) 7.9 -- 48.7d 3.7 Institut supérieur d'agronomie et d'élevage (ISAE) 6.0 -- 18.4d 3.5 Institut supérieur des finances publiques (ISFP)e 1.4 -- 34.8d -- All public institutions 100.0 -- 26.4d -- (8,729) Privatef Université libre de Kigali (ULK) 80.7 -- 48.8 -- Université adventiste d'Afrique centrale (UAAC) 11.0 -- 52.2 -- Université laïque de Kigali (UNILAK) 6.9 -- 56.2 -- Institut supérieur de pédagogie de Gitwe (ISPG)g 1.3 -- 49.4 -- All private institutions 100.0 -- 49.2d -- (4,029) All public and private institutions (12,757) -- 33.4d -- -- Not available. a. Figures in parentheses refer to the total number of students. b. Students' score on a 10-point scale on the national examination administered at the end of the upper sec- ondary cycle. c. Enrollments exclude part-time students, for whom little systematic data exist. d.Weighted percentage. e. Now the Institute of Finance and Banking (IFB); in French, Ecole des hautes études des finances et d'économie appliqué. f.The following private institutions, all with probably very limited enrollments not exceeding 100 to 200 stu- dents each, are excluded here for lack of enrollment data: Grand séminaire de Nyakibanda, Faculté de théologie de Butare, Grand séminaire de Kabgayi, and Centre d'enseignement supérieur de Kigali. g. Reestablished in December 2002. Source: For distribution of enrollments, appendix table A7-2; for data on examination scores, personal commu- nication from the National Examination Council of Rwanda 140 WORLD BANK COUNTRY STUDY and cater to those who are able and willing to pay the fees. The available data on the quality of student intake pertain only to the public institutions. The variation is modest, but it suggests that the brightest students tend to rank the UNR and the KHI as their top choices. Distribution of Enrollments by Field of Study The distribution of students by field of study in 2000­01 is shown in table 7-3. Overall, the pat- tern is heavily biased toward the humanities and social sciences, which enroll nearly three-quarters of the students in the system.8 Within these fields, there is a clustering of enrollments in econom- ics and management (whose share is 44 percent) and in literature and the humanities (27 percent). By contrast, only 28 percent of the students were enrolled in the applied and natural sciences, and just over half of these were pursuing courses in science and technology. The distribution of enrollments shows the expected differences between public and private institutions, with the latter focusing almost entirely on the humanities and social sciences. In absolute enrollments, the private institutions have more than three times the number of law students as do the public institutions, whereas enrollment sizes are roughly comparable in literature and the humanities and in economics and management. Trends in Enrollments by Field at the UNR The available data permit a closer look at how the distribution of enrollments at the UNR has evolved from the time when the university was practically the only significant institution to a context marked by a revitalized private sector and an expanded public sector. In general, the evi- dence points to substantial changes suggestive of an institution in the process of adapting to its new surroundings: experimentation with the creation of new faculties, reconstitution of existing faculties and schools, reorganization of course offerings, and so on (see appendix table A7-4). These changes are perhaps best seen as part of the university's ongoing efforts to achieve a good balance in supplying graduates to meet the county's felt need for highly qualified labor (NUR 2002a; Rwamasirabo 2002). Table 7-4 documents the evolution since the 1980s. One change is the large share of stu- dents enrolled in prespecialized courses in the postgenocide era. Fully a quarter of all students are now enrolled in such courses, reflecting the new policy of requiring students to achieve bilingual- ism as part of their university studies. Setting aside these students, the pattern of enrollments shows dramatic shifts between the broad fields of the humanities and social sciences and the applied and natural sciences. In the 1980s the distribution appeared reasonably balanced, with 58 percent of the students enrolled in the humanities and social sciences and 42 percent in the applied and natural sciences. In the postgenocide 1990s, enrollment shifted heavily in favor of the humanities and social sciences, raising the share of that field to 74 percent, a gain of 16 percent- age points over the level in the 1980s. The concentration in humanities and social sciences has declined only marginally in the early 2000s, to 72 percent. Given the country's interest in linking into the global economy through increased capabilities in the hard sciences, it might be appropri- ate to evaluate the current patterns of enrollments across disciplines. In particular, in view of the growth of the private sector, the question is whether the UNR is indeed offering courses that would not otherwise be available. Major shifts have also occurred within each of the two broad fields discussed above. The reori- entation toward courses with an explicit labor market focus, to the detriment of more traditionally academic ones, is especially clear from the data for the humanities and social sciences. In the 1980s 8. The humanities and social sciences include fields of study classified under codes 01­38 and 76­86 in the International Standard Classification of Education (ISCED); applied and natural sciences refer to ISCED 42­72 (see UNESCO 1997). For a breakdown of enrollment by field in each of the six public and four main private institutions, see table A7-3. EDUCATION IN RWANDA 141 , and 5.4 4.4 5.2 0.0 0.0 private 54.6 75.3 47.0 36.3 and % y (ISAE), ersité NSTITUTIONSI Univ (FTB). vage e types % 9.0 8.3 5.4 echnologT, 19.2 12.4 31.9 13.9 d'éle TIONA 100.0 Butar et Both distribution (UNILAK), de Science DUC E of onomie Kigali de théologie d'agr 994 918 599 IGHER students 2,123 1,371 3,537 1,543 1,633 Institute de H Number 11,085 12,718 of laïque ETA Kigali supérieur Faculté b ersité RIV (IFB). P and % 1.3 1.7 1.2 0.0 28.8 25.7 41.3 (UNR), Univ 100.0 Institut AND distribution Banking (ISPG), institutions (ULK), e Rwanda (KHI), and UBLIC du 0 0 Kigali Gitw P 54 68 48 de IN Private 1,160 1,033 1,661 4,024 4,024 students Institute de Finance e Number of of ogie nationale libr Health pédag ersité ersité TUDENTS a Institute de S Kigali F % 4.8 8.5 13.6 26.6 13.3 20.9 12.3 Univ the Univ O 100.0 w (KIE), distribution no supérieur institutions lic institutions: (ISFP), institutions: 963 338 940 870 599 Education Institut ISTRIBUTION Pub students 1,876 1,475 7,061 1,633 8,694 wing of wing C), D Number of ollof ollof AA . publiques (U AND 2000­01, the A7-3 Institute the A in in y table AND pplicablea finances Kigali centrale courses UMBER W des N R humanities management not courses ollments ollments (KIST), enr ppendixa and and technolog enr study e d'Afrique 7-3: e education and denote See of sciences specialized total supérieur in Includes ABLE w especialized erall entiste T Field Literatur La Economics eacherT Science Health Agricultur otalT Blanks a. Includes.b Source: Pr Ov Management Institut adv 142 WORLD BANK COUNTRY STUDY TABLE 7-4: TRENDS IN ENROLLMENTS AND DISTRIBUTION BY FIELD, UNIVERSITÉ NATIONALE DU RWANDA, SELECTED PERIODS Average annual enrollmentsa % distribution of enrollmentsb Postgenocide Early Postgenocide Early Field of study 1980s 1990s 2000s 1980s 1990s 2000s Humanities and social sciences 843 2,467 2,927 58.3 73.9 71.7 Law 126 573 366 14.9 23.2 12.5 Arts and human sciences 257 342 465 30.4 13.9 15.9 Education 153 310 727 18.2 12.6 24.8 Economics, social science, and management 308 1,180 1,329 36.5 47.8 45.4 Journalism and communications 0 62 40 0.0 2.5 1.3 Applied and natural sciences 602 870 1,153 41.7 26.1 28.3 Medicine and health sciences 143 440 430 23.8 50.6 37.3 Science and technology 333 305 623 55.4 35.1 54.0 Agronomy 126 125 100 20.9 14.3 8.7 Total in specialized fields 1,445 3,337 4,079 100.0 100.0 100.0 Prespecialized coursesc 0 730 1,302 0.0 18.0 24.2 Overall total 1,445 4,067 5,381 Blanks denote not applicable. a.The three periods refer, respectively, to 1982­86, 1994­99, and 2000­02. b. Expressed as a percentage of total enrollments.The percentages of the two broad fields--humanities and social sciences, and applied and natural sciences--sum to 100 percent; the percentages for each individual field sum to 100 percent within each of these broad fields. c. Refers to students in common core courses, language trainng, and other preparatory courses. Source: See appendix table A7-4 students studying arts and human sciences accounted for 30 percent of total enrollments in the humanities and social sciences, but their share fell to less than 16 percent in the early 2000s. By contrast, the proportion enrolled in economics and management rose from 37 to 45 percent. The share of enrollments in education fell in the postgenocide 1990s, perhaps reflecting a period of consolidation and reorientation in course content, but by the early 2000s the field had more than recovered, claiming nearly a quarter of total enrollments in the humanities and social sciences group. The years following the genocide saw an explosion of enrollments in law, to nearly a quar- ter of all students in the humanities and social sciences group, but the increase was short lived. By the early 2000s, law students made up just over 12 percent of the total in the category, less than in the early 1980s. This seems to be one field in which the university has been reducing the absolute number of enrollments as similar offerings have become available in the private sector. Within the applied and natural sciences, there is also evidence of major changes in the pattern of enrollment. Fewer students are studying agronomy; they made up less than 9 percent of all students in the hard sciences in the early 2000s, down from 21 percent in the early 1980s. The share of medicine and health sciences shot up in the postgenocide 1990s, accounting for half of science enrollments during the period, but by the early 2000s it had fallen to just over 37 percent of the total as the number of students stabilized at less than 450. Enrollments in science and technology picked up following a period of relative stagnation during the postgenocide 1990s. The lull is similar to that in education enrollments, suggesting a preexpansion period marked by consolidation and reorganization of course offerings in light of the evolution of the sector. EDUCATION IN RWANDA 143 TABLE 7-5: HIGHER EDUCATION STUDENTS ON OVERSEAS GOVERNMENT SCHOLARSHIPS, RWANDA, 1967­2002 Students abroad as % of total Number of students foreign and domestic Period or year abroada % female enrollmentsb 1967­69 220 -- 39.8 1970­79 504 -- 43.6 1980­86 690 8.9 33.1 1999­2000 902 31.0 11.3 2000­01 877 28.3 9.5 2001­02 646 26.9 6.0 -- Not available. a. Refers to the average annual number studying abroad for the periods shown. Data for 1999­2002 exclude students on government scholarships studying at the Rwandan branch of the Université adventiste d'Afrique centrale (UAAC).The number of such students was 79 in 1999­2000, 38 in 2000­01, and 20 in 2001­02. b. Percentage of all students studying abroad and in public institutions in Rwanda. Source: See appendix table A7-5. Study Abroad on Government Scholarships Rwanda has a long tradition of sending students to other countries on government scholarships. In the past, some types of training were simply not available locally. The situation has changed with the explosive growth of higher education in recent years, yet students continue to be sent abroad. As chapter 3 documented, each student on an overseas government scholarship costs the state nearly three times as much as one studying at a domestic public institution. In recent years the government has had to allocate about a quarter of its total current public spending to higher education to finance overseas study. Number of Students Involved Table 7-5 summarizes the trends in government-financed study abroad since the late 1960s.9 The number of overseas scholars rose steadily from 220 in the late 1960s to 902 by 1999­2000 before falling to 646 in 2001­02.10 The increase in the early years, coupled with slow growth in domestic enrollments, meant that an average of 33 to 44 percent of all higher education students were studying abroad in the two decades leading up to 1986. The share dropped steadily after 1999, reaching a floor of only 6.0 percent by 2001­02. Data on women's participation in overseas studies, which are available beginning in the 1980s, show significant improvement with time. In the 1980s women accounted for less than 10 percent of students studying on overseas government scholarships, but during the period 1999­2002, their share averaged just under 30 percent. This proportion is similar to the share of women in domestic public institutions. 9. The information on Rwandan students studying abroad does not include students financed privately, but their numbers are likely to be very small. 10. The Université adventiste d'Afrique centrale (UAAC), an international institution, has a branch in Rwanda. The 20 students studying on government scholarships at the Rwandan branch of the UAAC are excluded from the count in table 7-5. 144 WORLD BANK COUNTRY STUDY Distribution of Overseas Scholars by Field and Level of Study Table 7-6 shows the distribution of overseas scholars in 1984­85 and in 2000­01. Excluding those for whom the field of study is unknown or undefined, 76 percent of the students on an overseas scholarship were pursuing courses in the applied and natural sciences in the earlier year. By 2000­01, the share had declined to 57 percent, a drop of nearly 20 percentage points. The retreat from the applied and natural sciences mirrors a similar shift in the domestic system. Still, the share of overseas scholars in these disciplines is high in comparison with the corresponding shares of 42 percent in the public sector and 10 percent for all domestic students. In both the humanities and social sciences and the hard sciences, the trend has been toward increasing concentration in the pattern of enrollments. Nearly three-quarters of the overseas scholars in the humanities and social sciences were pursuing courses in economics, management, and business in 2000­01, compared with about 46 percent in the mid-1980s. In the applied and natural sciences the scholars have clustered around science and technology, boosting the share of this domain from 61 percent in the mid-1980s to nearly 78 percent in 2000­01. These shifts are again consistent with the trend in the domestic system, where increasing shares of students are enrolled in courses with an explicit orientation toward future jobs. Table 7-6 also offers information (for 2000­01 only) on the level of the courses pursued by overseas scholars. A very high share (87 percent) was enrolled at the undergraduate level or lower. Given the high cost of education abroad, it is important to determine the extent to which these students are pursuing training that is indeed unavailable in Rwanda. Since 43 percent of them were in the humanities and social sciences, and 88 percent of these were studying at the undergraduate level or lower, there is a high likelihood that a large proportion of these students could have been channeled to domestic institutions at a much lower cost to the government. Distribution of Enrollments by Host Country Table 7-7 shows the number of Rwandan students on overseas government scholarships by host country. In 1984­85, 45 percent of these students were hosted by developed countries (members of the Organisation for Economic Co-operation and Development, OECD); 43 percent were in non-OECD countries outside Sub-Saharan Africa; and only 12 percent were studying in Sub-Saharan African countries. By the late 1990s and early 2000s, the pattern had changed dramatically: the OECD share of students had fallen to about 14 percent, whereas the share of non-OECD countries outside Sub-Saharan Africa had risen to 49 percent and that of Sub-Saharan Africa countries, to 37 percent. These changes have been accompanied by an increase in the number of host countries in Sub-Saharan Africa, in contrast to decreases in the other two country groups. Beyond the broad trends, the changes taking place at the country level are also noteworthy. In 1984­85 India received no Rwandan government-sponsored students, but by 1999­2002, it had the single largest contingent of such students, between 35 and 45 percent of the total. South Africa and Uganda also became important host countries during the same period: neither had any government-sponsored Rwandan scholars in 1984­85, but by 1999­2002 South Africa was host to 7­15 percent of these students and Uganda, to 9­16 percent. All of the main OECD host countries except Canada have seen a decline in their shares of government-sponsored Rwandan students. Among the non-OECD countries, the change in Russia's position is especially dramatic: its share fell from 29 percent in 1984­85 to a mere 2 percent in 2001­02. Differences in the cost of education abroad and in the cost-sharing arrangements with the host countries appear to be among the key factors driving these changes. Around 2002, students studying in Algeria, China, India, Poland, and Russia were on special bilateral programs under which the host country paid the fees and offered a small grant, while the Rwandan government topped up the grant and paid for the airfare. (See appendix table A7-7 for details.) As for the other potential hosts, it is much cheaper for the government to send students to neighboring African countries than to OECD countries. For example, annual academic fees alone currently range from US$11,000 to US$14,000 in OECD countries but are less than US$1,200 in South Africa. In addition, welfare grants for students studying in OECD countries range from EDUCATION IN RWANDA 145 OF a (%) d (1986b). EVEL L 9.0 8.7 each 73.2 17.8 43.3 13.6 77.8 56.7 13.3 field 100.0 rof AND Distribution yb Rwanda IELDF elsv centages 31 16 16 e 39 le 251 343 350 450 122 BY 159 per 984­85,1 100.0 e All th C). orf AA study cent; (U data of orf elv 3 PhD 21 71 41 91 4 31 54 42 per le 119 13.0 100 CHOLARSHIPS 2000­01 centrale S yb Masters and to " wn. Education; of c d'Afrique y students eary VERNMENT of ramgo 82 34 63 53 79 220 291 311 382 770 84.2 O 3­4 pr sciences--sum entiste Ministr, G "Others/unkno undergraduate adv systems. Number natural b ersité supérieur VERSEAS ear 0 1 6 8 0 1 and excludes 10 11 14 26 2.8 Univ O 2-y diploma the ON pplieda Anglophone in sciences; of a and ee l'enseignement (%) d social degr branch de s TUDENTS field 11.4 46.3 42.3 23.8 17.0 60.5 22.5 76.2 11.7 and S 100.0 Distribution yb sciences, local ection AN fields. bachelor' pplieda the Dir social a at the AND oad or orf and W br R 1984­85 and omfr F of systems. studying O elsv these le 71 systems 69 63 81 83 149 288 107 476 087 of sciences unication students, all Number each social scholarships comm fields--humanities Francophone ISTRIBUTION within in Francophone and oad D in business (%) br . . o cent and studies tificate tw ernmentvog personal AND sciences 2000­01,Y y etc,y per study tificate cer humanities on social sciences of the 100 cer pplicablea of orf 2000­01, helier UMBER social elv to and le TUD orf N S ts, pharmacy oindustr wn not bac licence students and management, natural technolog yb sum a a ar,e and agr,y to 38 data and centages to subtotals 7-6: study and field denote per of orF of onom w fields Sum ABLE La Economics, Literatur Medicine Science Agr Others/unkno Distribution The Blanks a. Leading.b Leading.c d. Includes.e Source: T Field Humanities Subtotal Applied Subtotal All individual 146 WORLD BANK COUNTRY STUDY e b (%) shar 37.0 14.1 48.9 100.0 1984­85, students Regional of C). AA (U of OUNTRIES C 0 1 1 1 1 1 1 1 1 1 1 1 1 1 countries 15­17 12­14 8­9 2­3 6­10 2­6 31­32 entralec OST Number H host a d'Afrique HEIR 1999­2002 T 0 21 94 61 68 96 51 33 11 7 9 4 17 04 53 5 15 244 326 231 666 entiste AND 2001­02 adv Students 0 10 87 99 22 82 51 6 22 21 9 13 23 9 22 ersité of 334 138 114 674 373 159 2000­01 Univ CHOLARSHIPSS the of Number 0 8 65 369 141 155 151 02 82 52 6 40 81 41 461 42 52 0 9 calculated. 403 981 not 1999­2000 e branch ar VERNMENT O G e (%) Rwandan shar countries the 12.6 44.9 42.5 at 100.0 VERSEAS O students Regional of individual 2001­02. of studying ON to es of host shar 1 1 0 0 9 11 11 1 1 1 1 1 1 5 1 1 0 1 13 10 35 ed, 1984­85 TUDENTSS Number olvv scholarships 1999­2000 countries in om AN fr of AND W 89 32 0 0 25 41 77 38 48 1 0 umbersn ernmentvog R 318 07 33 51 301 42 25 27 207 708 ollments Number