Report No. 26038-RW Rwanda Education in Rwanda Rebalancing Resources to Accelerate Post-Conflict Development and Poverty Reduction June 3, 2003 Human Development Department (AFTH3) Africa Region Document of the World Bank 2 CURRENCYEQUIVALENTS (Exchange Rate Effective January, 2002) CurrencyUnit = RwandanFrancs (RFW) RFW 1.00 - US$0.0022 US$1 - RFW 456.81 Fiscal Year: January 1''- December 31" Vice President: CallistoMadavo CountryDirector: E m n u e l Mbi SectorManager: Laura Frigenti Taskmanager: Jee-Peng Tan RWANDA EDUCATION RWANDA IN REBALANCINGRESOURCESTOACCELERATE POST-CONFLICT DEVELOPMENT POVERTY AND REDUCTION Table of Contents Abbreviations and Acronyms vii Foreword. ............................................................................................................................................................. ............................................................................................................................ ix Acknowledgements .............................................................................................................................................. xi ExecutiveSummary ... XIII RksumeAnalytique .......................................................................................................................................... .......................................................................................................................................... ........................................................................................................................... xxiii The demographic context .................................................................................................................... Chapter 1: Country Context 1 Macroeconomic conditions ................................................................................................................. Overall pattern of governmentfinance................................................................................................ .................................................................................................... Conclusion..... ............................................................................................................................... 10 Chapter2: EnrollmentTrends and P .......................................... ...................................................................................... 13 Trendsineducationalcoverage.................. Aggregate enrollments by level of education 13 .......................................... 16 Fromcross-sectionalindictors of coverage .......................................... 20 Student flow patternsinprimary schooling.................................... ....................... Policy perspectiveson student flow management.......................................................................... Student flow patterns insecondaryschooling ............................... Conclusion...................................................................................................................................... ............................................................................ Nationalspendingon education........................................................................................................................... 39 Chapter3: EducationFinance A closer look at public spendingon education..................................................................................................... ............................................................................................................ 44 Policy implications ...... Public spendingper student.... 50 .............................................................................................................. Conclusion........................ 56 ............................................................................................................. ................................................................................... 58 ......................................................................................... 59 59 Educational participation rates amongorphans.................................................................................................... Overview ofparticipation rates.............................. Chapter 4: Socioecono es inEducation 63 Bygender ....................................................... Bylocality ....................................... ....................................................................................... 66 .................................................................................................... 66 Policy implications ................................................ Conclusion.......................................................................... ................................................................................... 72 Chapter5: Overview oftheService of services.............................................. supplyDelivery in Primary Education .. ..................................................................... ..................................................................... 74 ............................................................. 75 75 Teacherallocation across schools.......................................................... .......................................... 80 .................................. 86 Policy implications and conclusion............................................................................................................ 97 Teachersandtheir utilization and deployment across schools................................................................... Overview ofthe supply of services ...................................................................................................................... 110 Examinations results andtheir correlates........................................................................................................... Economiesof scale inservice delivery............................................................................................................... 114 .................................................................... 118 ...................................................................................................... .................................................................... 119 121 ........................................................................................ .................................................................... 122 128 .................................................................... 137 .................................................................... 144 Policy implications ...... .................................................................... 147 Conclusion............................................................ ....................................................................... 149 Chapter 8: Educationandthe Labor Market ........................................................................................... 151 Employment structure, educationalattainment of workers, andthe r Output of graduatesandtheir absorptioninto the workforce ............ Policy implications .................................................................................... 11 Conclusion.......................................................................................................................................................... 162 Appendix ........................................................................................................................................................... 163 References ......................................................................................................................................................... 197 Map IBRD32003 .............................................................................................................................................. 203 Box 1.1Rwanda at a Glance .................................................................................................................................. Boxes 1 Table 1.1: Population size. poverty rates. and selected healthindicators. Rwanda. 1970-2000............................. Tables 3 Table 1.3: Rwanda"s economic performanceincomparativeperspective. 1980-2000......................................... Table 1.2: Orphansandchildren living apartfrom biological parent(s). Rwanda. circa 2000............................... 4 6 Table 1.4: Governmentrevenues. Rwanda. 1981-2001.......................................................................................... 7 8 Table 1.6: Governmentspendingon education. Rwanda. 1981-2001.................................................................... Table 1.5: Governmentspending. Rwanda. 1981-2001......................................................................................... 9 Table 2.2: Coverageof Rwanda's educationsystem incomparativeperspective. late 1990s-2000..................... Table 2.1: Grossenrollment ratios (GERs) bylevel of education. Rwanda 1991-92 and2000-01 .................... 18 20 Table 2.3: Entryrateto grade 1. Rwanda 1991-92 and2000-01 Table 2.4: Survival ratesinprimary schooling. Rwanda 1991-9 Table 2.5: Repetitionratesinprimary schooling. Rwanda 1990& 2000-01 ................. ..................... Table 2.6: Summaryindices of studentflow efficiency inprimary schooling. Rwanda 1990& 2000 ................28 Table 2.7: Secondaryschool transition and survival rates. Rwanda 1991-92 and2000-01 ................................ 31 Table 2.8: Repetition ratesinsecondaryschooling Rwanda 1990-91 and2000-01 ........................................... Table 2.9: Summary index of studentflow efficiency in secondaryschooling. Rwanda. 1990-91 & 2000-01 ..32 32 Table 2.10: Distributionof primary schoolsandnew first graders byhighest grade of instruction offered by the Table 2.11: Grade-specificenrollments inprimaryand secondaryeducation. Rwanda2000-01......................... school. Rwanda2000-01 ..................................................................................................................................... 34 35 Table 3.1: Level anddistribution ofpublic spendingon education. Rwanda. 1982-2001.................................... 40 government agencies. Rwanda. 1999-2001............. ....................................................................................... Table 3.2: Recurrentpublic spendingon educationvia the ministry ofeducation and other ministries or Table 3.3: Householdspendingon education. Rwandacirca 2000 .............................. ............................. 42 Table 3.4: Distribution of recurrent public spendingon education. Rwanda. 1999-2001.................................... 43 44 Table 3.5: Number of teachers and school-level administrative staff ingovernment-financedprimary and secondary schools. Rwanda. 1999....................................................................................... 45 Table 3.7: Estimatedcurrent public spendingon educationby function andlevel. Rwanda. 1999 ..................... Table 3.6: Staff distribution by salary grade and averagesalariesinpublic schools. Rwan 46 Table 3.8: Functional distribution ofpublic current spendingon education. Rwanda. 1999............................... 47 Table 3.9: Public spendingper studentby level of educationinthe public sector. Rwanda. 1999...................... 49 51 Table 3.10: Level andcomposition of per pupilpublic spendinginpublic primary andsecondaryschools. Table 3.11: International comparisonsof current public spendingper student.................................................... Rwanda. 1999....................................................................................................................................................... 53 Table 3.12: Public sector teacher salariesand pupil-teacherratios inRwanda andother countries..................... 54 Table 3.13: Public sector teachertake homepay andthe pay of other workers. Rwanda. 2001.......................... 55 56 Table 4.1: Gross enrollment ratios byprovince. locality. gender. and income groups. Rwanda. 1992-2000....... 60 Table 4.2: Share of salariedworkers amongthe parentsof students intwo public higher educationinstitutions. ............................. 63 Table4.5: Selectedstudent flow indicators by gender. locality. and income groups. Rwanda. circa 2000 ......... 65 ....................................................... ............................ 66 Table4.6: Regression-predictedindicators of schoolprogressionduring two consecutiveyears ina cohort of childrenaged7-12 in 1998. bypopulation groups. Rwanda. 1998-2000............................................................. 68 Table 4.7: Share ofpublic spendingon educationbenefiting the poorestandrichest population quintiles. Table4.8: Gender differences inperformance on national end-of-cycle examinations. Rwanda. circa 2000......72 Rwandaand other countries. 1990sand 2000 ...................................................................................................... 74 75 Table 5.2: Characteristicsof state. libre subsidik. andprivate primary schools. Rwanda. 2000.......................... Table 5.1: Percentagedistribution of primaryschoolsandpupilsby sector. Rwanda. 2000 ............................... Table 5.3: Characteristicsof primaryschoolteachers. Rwanda. 1999................................................................. 77 79 ... 111 Table 5.4: Accessibility ofprimary schools andproblemswith schooling reportedbycurrently enrolled pupils, Rwanda, 2001................................................ Table 5.5: Regressionest schools, Rwanda, 2000... Table 5.6: Regressionest province, Rwanda, 2000. Table 5.7: Regressionest Rwanda, 2000................. ................................. 88 Table 5.8: Percentage dis Table 5.10: Correlatesof school-level passrateson the primary school leaving examination, Rwanda, 1999...90 Table 5.9: Primary school leaving examinationresultsby type of school, Rwanda, 1999................................... 93 Table 5.11:Regressionestimates ofprovincialdifferences inexamination passrates, controllingfor differences 94 Table 6.1: Number andpercentagedistribution oftypes of secondaryschools, Rwanda, 2001-02 ..................... inper pupilspendingonpersonnel,Rwanda, 1999.............................................................................................. 99 Table 6.2: Selectedcharacteristicsof state, libre subsidie', and private secondaryschools, Rwanda, 1999-2001 ......................................................................................................................................................... 100 ercentage distribution of schoolsby level and number of instructional streams offered, Rwanda, ...................................................................................................................................................... 101 tribution of secondarystudentsby cycle and streamacross schooltype, Rwanda, 2000-01 ...... 101 Table 6.5: Number of schools offeringupper secondaryprograms and average enrollmentsper schoolby field, Rwanda, 2000-01 ......................................................... ...................... 103 Table 6.6: Characteristicsof teachersby type of secon ...................... 105 Table 6.7: Distribution ofteachers by type of classesthey teach andtheir teachingload andstudent-teacher ratios inpublic secondaryschools, Rwanda, 1999-2000.................................................................................... 106 Table 6.8: Distribution of state andlibre subsidie' secondaryschoolteachers byeducationalattainment andtype of classestaught, Rwanda, 1999-2000............................................................................................................... 107 Table 6.9: Regressionestimatesof the relation betweennumber of teachers and students acrosspublic secondary schools, Rwanda, 1999-2000.............................................................................................................................. 109 Table 6.10: Relationbetweentotal cost of personnelandenrollments inpublic secondary schools, Rwanda, 1999-2000.. ................................................................................................................................................ 112 Table 6.11:Size distribution of enrollmentsinpublic secondary schools and simulatedcost per student, Table 6.12: Resultson the national examination at the end of the tronc commun cycle, Rwanda, 1999-2000.. 115 Rwanda, 2000..................................................................................................................................................... 114 Table 6.13: Correlates of school-level performanceon the national end-of-tronc commun cycle national examination, Rwanda, 1999............................................................................................................................... 117 Table 7.1:Number of higher educationinstitutions, overall enrollments, andprivate share of students, Rwanda, 1960sto the present............................................................................................................................................ 123 Table 7.2: Institutional distribution ofhighereducationstudents, share of female students, andexamination score of senior six entrants, Rwanda, 2000-01............................................................ Table 7.3: Number and distribution of students inpublic andprivate higher educatio 2000-01.................................... ........................ ............................................. 126 Table 7.4: Trends inenrollment 199Os, andearly 2000s.............. Table 7.5: Rwandese higher edu 002..............129 Table 7.6: Number andpercentagedistribution of Rwandesestudentson overseasgovernment scholarshipsby field and level of studies, 2000-01 ......... ..................................................... 130 Table 7.7: Number of Rwandesestudent hipsandhost countries, 1984-85,and Table 7.8: Number and composition ofhighereducationfaculty by institution, Rwanda, selectedyears .........13 ....................................... ................................................................................... 1 133 Table 7.9: Number of nationals on the Universite'nationale du Rwanda (UNR) faculty on an overseas scholarship, status as of February 2002.............................................................................................................. 134 Table 7.10: Staffing ratios inpublic andprivate higher education institutions, Rwanda, 2000-01.................... 135 Table 7.11: Student-faculty ratios byfield of study at the Universite' nationale du Rwanda, 2000-01..............136 Table 7.12: Cost of service delivery per student inpublic higher education institutions andtheir composition at the Universite'nationale du Rwanda,2000 ........................................................................................................ 139 Table 7.13: Number of studentsand share amongthem receiving a bursary andvarious student services inUNR, KIE, andISAE, Rwanda, 1994tO 2001.............................................................................................................. 142 Table 7.14: Average annualvalue of studentbursaries, cash transfers, and deductionsto cover the cost of student services inpublic higher education, Rwanda, 2000-01.......................................................................... 143 iv Table 7.15: Promotion, repetition, andsurvival ratesinselectedhigher educationinstitutions, Rwanda, circa 2000.... ..... Table 7 ber ............................................................. Table 8.1:Selecteddata on population, lab ............................................................. Table 8.2: Distributionof employmentby sector, Rwanda, 1991and 2000 ...................................................... 153 Table 8.3: Number anddistribution of workers bytype of employment andjob, Rwanda, 1991and2000 ...... 154 Table 8.4: Percentage distribution ofthe employed population by educationalattainment, Rwanda, 1991and 2000.................................................................................................................................................................... 155 Table 8.5: Average salaries and years of schooling ofwage earners inthe formal andinformal sectors, Rwanda, 155 Table 8.6: Rates of return to educationby level and sector o femployment, Rwanda, 2000.............................. 2000.................................................................................................................................................................... 156 Table 8.7: Percentagedistribution of wage earners by educationalattainmentrelative to workers inthe same jobs, Rwanda, 2000a.......................................................................................................................................... Table 8.8 Unemployment rate by educationallevel and age, Rwanda, 1991and2000 ..................................... 158 159 Figures Figure 1.1:Adult HIV/AIDS prevalenceratesandshareof orphansamong7- to 14-year-olds,Rwandaandother EastAfrican countries, circa 2000....................................................... .............................................................. 5 Figure2.1: Enrollment trends inprimary schooling, Rwanda 1971-2001............................................................ 14 Figure 2.2: Enrollments insecondary education................................................................................................. 15 Figure2.3: Enrollment trends inhigher education, Rwanda, 1975-2002............................................................. 16 Figure2.5: Grade-specificenrollment ratesinprimaryschooling, Rwanda 1991-92 and2000-01 .................... Figure 2.4: Relationbetweeneducationalattainment andprobability ofbeingliterate, Rwanda2001 ............... 21 25 Figure2.6: Rwanda's primary schoolstudent flow indicators incomparativeperspective,circa 2000...............29 Figure2.7: Efficiency of studentflow inRwandaincomparativeperspective,circa 2000 ................................. 30 Figure3.1: Relation betweenprim income countries, circa 2000......... rimary education, low- ................................ 41 Figure 3.2: Functional distribution 9.............................. 50 Figure 4.1: Grossenrollment ratios inlower andupper secondaryeducationby income quintiles, Rwanda, 2000 Figure 4.2: Primary school enrollment ratesina cohort of children across income groups, Rwanda, 2000.. ............................. ........................................................... ....................................... 62 Figure 4.3: Share of cumulative public spending on educationbenefiting the 10percentbest educatedina generation,Rwandaandother African countries, circa 1998.... ............................................ 71 Figure5.1 Institutional composition ofprimary schools by pro 2 ......................................... 76 Figure5.2: Pupil-teacherratios inpublic primaryschoolsinRwandaandotherAfrican countries, circa 1999..78 Figure5.3: Relationbetweennumberofpupilsandnumber ofteachers acrossprimaryschools inthe public sector, Rwanda, 2000 ........................................................................................................................................... 82 Figure5.4: Rzvalues ofregressionsrelating numberofteachersto pupilsacrossschools, Rwandaand other African countries, circa 2000 ............................................................................................................................... Figure 5.5: Relationbetweenpersonnelcost per pupiland school size, Rwanda, 1999 ...................................... 84 87 Figure5.6: Economiesof scale inpublic primaryeducation, Rwanda, 2000....................................................... 88 Rwanda, 1999................................ ................. .............................. 91 ........................................ educationand schooltype, Rwanda, circa 2000............................................ .......................................... 110 Figure6.4: Relation betweenumber of studentsandcost per studentinthe ............................................. ommun and upper secondary 111 Figure 6.6: Relation betweencost ofpersonnelper student andend-of-tronc commun cycle national examination results acrosspublic secondary schools, Rwanda, 1999..................................................................................... 116 Figure 7.1: Arrangementsfor student finance inpublic higher educationinRwanda....................................... 141 Figure 8.1:Relation betweenper capitaincome andcoverageinhigher education, low-income countries, circa 1998.................................................................................................................................................................... 160 V Appendix Tables Table A1. 2: Current and capital public spending on education. Rwanda. 1981-2001....................................... Table A1. 1: Government revenue and expenditure. Rwanda. 1980-2001......................................................... 163 164 Table A2. 1: Enrollmentsby level o f education. Rwanda 1967-68 to 2000-01 ................................................ 165 Table A3. 1:Public spending on education by level. Rwanda. 1971-2002........................................................ Table A3. 2: Public spending on education. Rwanda. 1985-2001...................................................................... 166 Table A3. 3: Itemizedhousehold spending on education. Rwanda. circa 2000 ................................................. 167 168 Table A3.4: Recurrent spending on education by level o f education and function. Rwanda. 1999-2001......... 169 Table A3.5: Per student spending inpublic primary and secondary schools. Rwanda. 1999............................ Table A3.6: Earnings function excluding wage earners working for educational institutions. Rwanda. 2001 . 171 170 Table A4. 1: School attendance status o fa cohort o f childrenaged 7-12 in 1998. Rwanda............................... Table A4. 2: School progression rates duringtwo consecutive years among 7- to 12-year-olds by orphanhood 172 status. Rwanda. 1998-2000................................................................................................................................. 173 Table A4.3: Regressionestimates o f school attendance status in a cohort o f children aged 7-12 in 1998. Rwanda 1998-2000............................................................................................................................................ 174 . Table A5. 1: Teacher's educational attainment. qualification. andremuneration. Rwanda. 2001 ..................... Table A5.2: Pupil-teacher ratio byprovince in state and libre subsidie'primary schools. Rwanda. 2000......... 177 176 Table AS.3: Regressionestimates of the relation between numbers of teachers and pupils acrosstypes of public primary schools with provincial dummy variables. Rwanda. 2000.................................................................... 178 Table A5.4: Regressionestimates o fthe correlates o f school-level pass rates on the national primary school leaving examination. public sector schools. Rwanda. 1999............................................................................... 179 Table A6. 1: Distribution o f secondary schools bylevel andtype o f instruction offered. Rwanda2000-01 .....180 Table A6.2: Number o f secondary school teachers by qualification and type o f classes taught. Rwanda. 1999 ................ .................... ...................................................................................................... 181 tion between number o fteachers and number o f students across public secondary schools. Rwanda. 1999-2000.................................................................................................. 181 Table A6.4: Correlates o f school.leve1. end-of-cycle tronc commun national examination results. Rwanda. 1999 ............................................................................................ .......................................................... 182 Table A7. 1: Trends inhigher education enrollments by institution. Rwanda. 1963-2002................................ 183 Table A7. 2: Enrollments inpublic and private higher education institutions by sex. Rwanda. 1984-85 to 2001- Table A7. 3: Distribution o f students by field o f s 02.......................................................................... ...................................................................................... Table A7.4: Enrollments by field inselected institutions. Rwanda. 2000-01 ...................185 184 Table A7. 5: Number o f Rwandese students on overseas government scholarships and number enrolledlocally. o fstudy inthe Universite' Nationale du Rwanda. 1982-2002...................... 186 1967-2002....................................................................................................................... .............................. 187 Table A7. 6: Number o f Rwandesestudents on government overseas scholarships by host c Table A7.7: Academic fee. welfare. and travel costs paidby the Rwandese government for 1999-2002................................................................................................................................. Table A7.8: Number o f faculty by institution andnationality. Rwanda. 1985-2001......................................... on government overseas scholarships. circa 2002.............................................................................................. 189 190 du Rwanda....... Table A7.9: Current and proposedfuture arrangements regarding student bursaries at the Universite'nationale ......................................................................................................................... 191 192 Table A8. 2: Earningfunctions among wage earners by sector of employment. Rwanda. 1999-2001.............. 193 Table A8. 1: Earningfunctions among wage earners. Rwanda. 2000................................................................ Appendix Figures Figure A4. 1: Cumulative shares o f a hypotheticalcohort by educational attainment and o fpublic spending on education benefitingthe cohort .......................................................................................................................... 195 vii Abbreviations and Acronyms AGSER Average Grade-Specific Enrollment Rate ARR Average Repetition Rate C A Certijkat d'Aptitude CERA1 Centresde 1'EnseignementRural et Artisanal Intkgrk CERAR Centresde 1'EnseignementRural et Artisanal de Rwanda CESK Centred'enseignmentsupkrieur de Kigali CFATS Centre deformation des adjoints techniquesde la statistique CPR Conseil Protestant du Rwanda CSR Cohort Survival Rate CWIQ Core Welfare Indicators Questionnaire DHS Demographic and Health Surveys EFA Education For All EICV EnqueteIntkgralesur les Conditions de Viedes Mknages au Rwanda EPLM Ecolepratique des langues modernes ES Ecole Secondaire ESM Ecole supkrieure militaire EST1 Ecole des sciences et techniquesde 1'information FARG Fonds National pour 1'Assistanceam Rescapks du Gknocide et des Massacres au Rwanda FRW RwandanFranc FTB Facultk de thkologie deButare GER GrossEnrollment Ratio GNP GrossNationalProduct GSK Grande skminaire de Kabgayi HIPC Heavily IndebtedPoor Countries HIVIAIDS HumanImmunodeficiency Virus /Acquired Immune Deficiency Syndrome IAMSEA Institut Africain et Mauricien de statistiques et d'kconomie appliquke IDA International Development Association IFB Institute of Finance and Banking IGER ImpliedGross Enrollment Ratio IMF International Monetary Fund IPN Institutpkdagogique national ISAF Institut supkrieur d'agronomie et d'klevage ISCED International StandardClassification o f Education ISCPA Institut supkrieur Catholique depkdagogie appliquke deNkumba ISFP Institut supkrieur desfinances publiques ISPG Institut supkrieur depkdagogie de Gitwe KHI Kigali HealthInstitute KIE Kigali Institute of Education KIST KigaliInstitute of Science & Technology & management MDG Millennium Development Goal MICS MultipleIndicator Cluster Survey MINALOC Ministgre de I'Administration Local et des Affaires Sociales MINEDUC MinistryofEducation NER Net Enrollment Ratio NPV Net PresentValue OECD Organization for Economic Cooperation and Development ... Vlll PRS Poverty Reduction Strategy PRSP Poverty Reduction StrategyPaper PSPP Public SpendingPer Pupil PTA ParentTeacher Association QUID Questionnaire UnijZsur les Indicateurs deDheloppement RPA RwandesePatriotic Army RR Repetition Rate SF SectionFamiliale SIMA World Bank Statistical InformationManagementandAnalysis database SNEC Secretariat Nationale de 1'EnseignementCatholique UAAC Universitk Adventiste d'Afrique Centrale ULK Universitk libre deKigali UNAIDS Joint UnitedNations Programme on HIV/AIDS UNDP UnitedNations Development Programme UNESCO UnitedNations Educational, Scientific andCultural Organization UNICEF UnitedNationsInternationalChildren's EmergencyFund UNILAK Universite lai'que de Kigali UNR Universitk Nationale du Rwanda ix Foreword This Report makes an important contribution to the on-going strategic planning process for the development o f education inRwanda. Itprovides informative analyses on national trends upto 2001andoffers revealingcomparisonswithother countries insimilar situations. The Report has already been used to informthe draft Education Sector Strategic Plan (March 2003) and was discussed with interest during the joint review of the education sector which was held in Kigali in April 2003. Iam sure that the Report will continue to provide an important reference point for the ongoing Sector Wide Approach programme (SWAP) in the educationplanning and management process. By way o f an up-date, it is important to note that there have been several new policy initiatives since the data for the Report were collected in 2001. In particular, the 2002 Education Sector Policy makes clear the government's commitment to provide quality basic education to all Rwandese children and we have proposed steps to make this a reality, including the abolition o f school fees. We have also taken steps to reduce the highrepetition rates noted in the Report and Iam pleased to report that rates in 2002 showed a reduction from 31.8 to 17 percent. We hope that this salutary trend will continue. Due to highpressure from risingnumbers o fprimary school graduates and the necessity o fproviding young people with the basic skills required for life and work, we shall put in place strategies to ensure lower secondary school education is offered to as many children as possible. Much effort i s being made to raise the quality o f education through the provision of trained teachers and pedagogical materials. We have introduced a radical reform o f higher education financing to reduce government subsidies to university students. This will in turn release additional resources for investment in the primary and secondary sectors. Lastly, the teaching o f science and technology especially Information Communication Technology (ICT) shall be given special attentionat all levels o f our education system. All-in-all Iam confident that Rwanda will continue to make good progress towards achieving the Millennium Development Goals in education and that we will be able to provide the country with an educated and skilled work-force to builda prosperous and peaceful future for all. Iwouldliketotakethisopportunitytothankallwhohaveinonewayoranothermadethis study a success. Special gratitude goes to the World Bank team led by Jee-Peng Tan, lead economist and the Rwandese team led by Claver Yisa, Director o f Planning, for their commendable work on the Report. Prof. RomainMurenzi Minister of Education, Science Technology, and Scientific Research POBox 622 Kigali, Rwanda June 2003 xi Acknowledgements T h i s report i s the result of a collaborative effort by the Government of Rwanda and the World Bank to deepen understanding o f the current status o f education in Rwanda. The Government team was ledby Claver Yisa ,Director o f Researchand Planning o f Education in the Ministry of Education (MINEDUC), under the overall leadership of Prof. Romain Murenzi, Honorable Minister of Education, and Eugene Munyakayanza (formerly MINEDUC's Secretary General and now its Minister o f State for Primary and Secondary Education). The World Bank team was led by Jee-Peng Tan under the overall guidance o f the following Africa Region managers: Emmanuel mi,Laura Frigenti and Arvil Van Adams. The Rwandese team initially comprised Faustin Habineza, Joseph Matsiko, Herman Musahara, and Ernest Rutungisha. As the work progressed, additional support was received from various individuals working in the Ministry o f Education, including Aviti Bagabo, Sampson Kagorora, Leonard Manzi, Catherine Mukankuranga, Susy Ndaruhutse, Joseph L e Strat and Patrick Rwabidadi. Many persons outside the Ministry also offered invaluable assistance with data collection and preparation, including Frangois Rwambonera (Conseil Protestant du Rwanda); A. Aloys Guillaume (Sbcretariat national de 1'enseignement catholique); Joseph Murekeraho and his staff at the National Examinations Council; Straton Nsanzabaganwa (Minist&-e de 1'Administration Local et des Affaires Sociales); John Ruzibuka and Frangois Katangulia (OfJice national de la population); Jacques Gashaka and Omar Sarr (EnquCteIntkgrale sur les Conditions de Vie des Mbnages au Rwanda ); and Jean- Baptiste Rulindamanywa, Camille Rutayisire, Evariste Nzabanita and (MinistBre de la Fonction Public et du Travail); andPatrick Jondoh (Unicef). The World Bank team consisted o f GCrard Lassibille, Jean-Bernard Rasera and Kiong Hock Lee. Valuable assistancewas offered to the team by Susan@per inher role as the task team leader for education inRwanda; and Robert Prouty and Quentin Wodon intheir role as peer reviewers. Inaddition to the contributions made by the people mentioned above, very helpful feedback and support were received from the following interlocutors in Rwanda: Narcissse Musabeyezu (Director of Primary Education), Emma Rubagumya (Director o f Secondary Education), Callixte Kayisire (Director o f Higher Education), John Rutayisire (Director of National Curriculum Development Centre), BCatrice Mukabaranga (Kigali Institute of Education), and Silas Lwakabamba (Kigali Institute o f Science and Technology and Management); and from the following World Bank colleagues: Chukwuma Obidegwu, Toni Kayonga, Edward Brown, Guido Rurangwa, Raju Kalidindi, Birger Fredriksen, Alain Mingat, Mamy Rakotomalala and Jaap Bregman. Contributions fkom Desmond Bermingham, Michael Delens and William O'Hara o f the UnitedKingdom's Department for International Development are also gratefully acknowledged. Inthe course ofpreparing the report, visits were made by members ofthe World Bank team to Rwanda and by members o f the Rwandese team 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, and Anna Rutagengwaova, and Marie Jeanne Uwanyarwaya. Kofi Edoh xii translated the Executive Summary into French; and Julia Anderson put the whole document together and managedits processingfor publication. The work on this report has been financed by the World Bank and by the Government o f Norway through its NorwegianEducation Trust Fundto advance educational development in Afi-ica. ... Xlll EDUCATIONINRWANDA: REBALANCINGRESOURCESTO ACCELERATE POST-CONFLICT DEVELOPMENTAND POVERTY REDUCTION ExecutiveSummary Context 1. Rwanda's recent history was marred by genocide in 1994, in which at least 800,000 people or about 10 percent of the population lost their lives. Stability and security have been restored, and the process of recovery has been under way for several years now. Rebuilding the stock of human capital that perished in the massacre is an important part of that process, and the government has worked hard to make up for lost time in broadening access to education and enhancing the quality of services. Indeed, progress in education figures among the core objectives in the government's economic and social development strategy as articulated inits recent poverty reduction strategy paper (PRSP). 2. On the international stage, the education sector has also come into the limelight in recent years. Under the 2000 United Nations Millennium Declaration, the international community has agreed to work toward ensuring that, by 2015, children everywhere-boys and girls alike-will be able to complete a full course of primary schooling. Another goal has been to work toward eliminating gender disparities in primary and secondary education, ideally by 2005. These are two of the eight Millennium Development Goals (MDGs) explicitly stated under the declaration. Following the 2002 Monterrey International Conference on Financing for Development, consensus now also exists that reaching the goals will require action from both rich and poor countries. Rich countries must, among other measures, boost foreign aid to poor countries, whereas poor countries must put in place the right policies and governance structures to ensure effective use of resources to achieve the goals. The principle of coupling resources with results on the groundis also beginningto permeate discussion o f development policies at the country level. The education sector 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. 3. The foregoing context presents clear challenges for managers of the education system in Rwanda. To attract increased resources, both internationally and domestically, the sector must show evidence of good stewardship of the resources it already receives. What is the scope for improvement inthis regard? What policies are required to ensure that the sector develops in an efficient, equitable, and fiscally sustainable direction? What gaps indomestic financing remain that external resources might usefully fill? What outcomes mightbe agreed on to focus effort and create proper accountability structures inthe system? Purpose, audience, and scope 4. The purpose o f this report is to provide a factual basis for discussing the questions posed above. It is based on data up to 2001, the latest year for which it had been possible to gather the relevant statistical information in the course of report preparation. Because the education system has not stood still since 2001,the snapshot picture captured by these data does not, by definition, track recent developments inthe sector. InRwanda, these changes are occurring because of proactive interventions by the government to address some o f the emerging constraints on sector development. Noteworthy are the efforts to reduce grade repletion in primary education: the latest school census returns for 2003 suggest that there have been significant progress in addressing this problem. Similarly, reforms inhigher xiv education finance have been launched in order to reduce the cost o f government-sponsored overseas studies (e.g. by redirecting students to lower-cost host countries such as South Africa and India) as well as that of local studies (e.g. by treating student bursaries as loans rather than outright grants and creating the Student Financing Agency to institutionalize collection o f repayment). 5. As a stock-taking exercise based o n the situation at a given point in time (i.e. up to 2001), the report is best seen as a diagnostic document whose purpose is to contribute toward a shared understanding of the current performance of Rwanda's education system, the constraints on its progress, and the potential trade-offs involved in charting the system's development inthe coming years. Buildinga shared understanding o f the issues is an integral part o f policy development, one that might follow the process of consultation with partners and civil society on which many governments, including Rwanda's, have already embarked indefining their poverty reductionstrategies. Ineducation, the importance ofthe process can hardly be exaggerated: in the end, the success of policies depends on how they are implemented. This in turn depends to an important degree on how well the problems are understood, particularly among those who must bring about the needed reforms-including policymakers, educational planners, teachers, school managers, parents, and students-and onhow much agreement there is onthe proposed solutions. 6. The audience for this report is, therefore, broad. The report is addressed inthe first instance to Rwanda's policymakers in the education sector as well as to education practitioners and researchers. The report should also be of interest to policymakers and analysts in other parts of the government, particularly those charged with managing the country's overall development strategy and aligning public spending accordingly. The government's development partners should also find information and analyses equipping them to engage more actively in discussing and articulating the country's vision for the education sector and design and implementation of policies to actualize that vision. These partners include donor agencies and nongovernmental organizations that support educational development inRwanda, as well as parents, students, teachers, and the public at large. 7. With regard to scope, the report does not pretend to address the full gamut of issues that Rwanda's policymakers might face. The breadth of its coverage is limited to key economic aspects that are particularly relevant in the PRSP context: cost, finance, service delivery, and education outcomes, particularly those aspects that lend themselves to quantification. Although admittedly incomplete, the treatment yields a picture of the broad structural characteristics of the system and the implicit pattern of resource allocation and effectiveness of service delivery. The key findings are summarized below. Educationalprogressto date 8. Despite the untold havoc caused by the 1994 genocide, the Rwandese education system has recovered remarkably well, at least inquantitative terms. Inplottingthe sector's future development, it is thus important to begin by pointing out some of the signal achievements and the base they provide for buildingup the system incoming years. 9. The most impressive aspect of the system's recovery is the rapid Dace of enrollment increase inthe aftermath of the genocide. Only 5 years after the event, 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. Currently at 107 percent, the xv gross enrollment ratio at this level exceeds the corresponding ratio for the average low- income country in Africa today. In secondary education, the number o f students grew at 20 percent a year since 1996, implying that the system is now nearly three times as large as it was in the earlier year. Although the gross enrollment ratio at this level remains below the average for low-income Sub-SaharanAfrica-1 3 percent compared with 20 percent-the gap would have been even wider, had the system stagnated after the genocide. In higher education, enrollments rose even more rapidly: from 3,400 students in 1990-91 to nearly 17,000 by 2001-02, nearly a fourfold increase in a decade. The system's coverage i s now comparable to the average of about 200 students per 100,000 population inlow-income Sub- SaharanAfrica. 10. As the system expanded, it has done so in ways that has nudged it toward a good balance between the public and private sectors. At the base o f the education pyramid, there hasbeen a consistently strong effort by the government to extend the coverage of the public sector. As a result, the share o f enrollments in private schools has remained modestat less than 1percent. At the secondarylevel, enrollments grew as fast inthe public as inthe private sector in the post-genocide years and the share of students attending private schools has remained steady at about 40 percent-lower than the 62 percent inthe 1980s, but still much higher than the 20 percent on average in low-income Sub-Saharan Africa. In highereducation, the private sector grew infits andstartsinthe two decadesbefore 2000, but it is unmistakably catering to a rising share of students over time-about 38 percent in2001- 02, compared with about 8 percent at the start o f the 1980s. The diversity o f postsecondary institutions and the mix o f public-private sector providers i s a strength o f the system, endowing it with the flexibility to meet the growing demand for places at this level o f study. 11. It is important to note that the expansion of the system has been taking place within a structure that provides for a 6-year primary cycle, a 3-year tronc commun (or lower secondary) cycle, a 3-year upper secondary cycle, and a 4-year higher education cycle in most fields. A meritocratic examination system has been put in place to manage student selection among the various cycles o f schooling. The 6-3-3-4 educational structure, along with the examination-based selection mechanism, provides a sensibly configured system for managing the sector's expansion. InRwanda, evidence exists that children who complete 6 years o f primary schooling usually remain permanently literate and numerate as adults. Universalizing primary school completion is thus entirely consistent with what is needed to buildthe human capital base for broadly based economic and social development, and the system i s already structured accordingly. At the same time, the selection arrangements for progression to post-primary levels are also already inplace, giving managers o f the system the administrative levers to calibrate the pace o f expansion according to the availability o f resourcesand the absorptive capacity o fthe market for highlyeducatedlabor. 12. Systemwide assets aside, the Rwandese system also compares favorably with that o f other low-income countries in Africa in terms o f the socioeconomic disparities in educational access, especially at the primary level. A striking fact is that school participation rates are relatively high, even among orphans. The 1994 genocide left Rwanda with one o f the highest orphanhood rates in the world; nearly 40 percent o f the children aged 7-14 years in2000 have lost at least one parent. Yet, the gap inenrollments between orphans and other children i s noticeable only among the most vulnerable children (i.e., those who have lost both parents or those who live apart from their parents). The result points to the existence of relatively well-developed safety nets that have somehow managed to ensure high primary school participation rates even among orphans. In secondary education, orphans are at least xvi as well represented as non-orphans. This remarkable outcome owes much to the government's decisionin 1998 to establishthe tax-funded Genocide Fundas a mechanism for assisting orphans insecondary school. The challengesahead 13. Inthe comingyears, the challenges consist ofsecond-generation problemsthat are already beginning to emerge following the system's successful rebound from the devastation o f the genocide. The focus i s thus shifting from a situation inwhich emergency measures to re-establish the functioning o f the system take priority to one in which the concern is to chart an appropriate and fiscally sustainable course for the sector's long-term development. Below i s a summary o f some o f the main issues at the systemic level with which policymakers are likely to have to grapple in the foreseeable future. Readers are referred to the individual chapters for detailed findings on specific points that are not mentioned below. 14. Managing student flow and graduate output from the system. To achieve the goal o f providing all children with the chance to complete a full course o f primary schooling requires attention to two issues: raising entry rates to grade 1and boosting the likelihood that children continue to the end o f the cycle. InRwanda, entry rates have been historically high at 90 percent, so the challenge here i s to identify and help the last 10 percent o f the population that still shy away from school. With regard to survival to the end o f the cycle, Rwanda's performance is respectable: the current rate i s an estimated 73 percent, which compares well with the rate inother low-income countries and with the 44 percent inRwanda in 1990-91. Yet this rosy result is unlikely to persist given the exceptionally highrate o f grade repetition inthe system-about 34 percent in2000-01 or more than three times the rate a decade earlier. In line with what i s widely believed to be good practice based on cross- country experience, it would make sense for the country to aim for a medium target of, for example, 10 percent and to put in place measures to rationalize policies and practices regarding grade-to-grade promotion, as well as improve learning outcomes to minimize the needfor pupilsto repeat. 15. Beyond primary education, the pressures to expand access are already beginning to mount, as larger cohorts o f children are now completing the lower cycles o f schooling. To develop appropriate policies governing the progression o f students from cycle to cycle, it is helpful to distinguishthe tronc commun cycle from the upper secondary cycle. The former can be thought o f as a continuation o f the primary cycle; thus, to the extent that resources permit, it would be appropriate to aim for universalizing access to this cycle as a medium-term objective. The upper secondary cycle, however, is more aptly viewed as the preparatory phase for higher education. Because graduates from these levels are being groomed at great expense for jobs inthe modem economy, it i s important to ensure that the number o f graduates produced and their skills mix are compatible with the prospective demand for skilled labor. In the immediate post-conflict context, pervasive shortages o f highly educated workers were felt throughout the government and inthe private sector, and these shortages have stimulated a rapid growth in enrollments, particularly in higher education. Now that the stock o f humancapital i s gradually beingreplenished, inpart through in-migration o f the Rwandese diaspora, and signs o f potential graduate unemployment are beginning to emerge, it would be wise to review the situation and adapt admission policies in upper secondary andhigher education accordingly xvii 16. Mobilizing and making effective use of resources for education. Support for educational development has been and continues to be strong. In the past few years, the convergence of international public opinion and commitments and the government's own prioritization o f the sector has helped to boost public spending on education inRwanda to an all-time high o f 5.5 percent of GDP in2001. Yet, closer examination shows that the increase has been dominated by spending o n capital investments. Although such investments are obviously needed to rehabilitate devastated facilities and expand capacity, the whole system would at some pointneed adequate resources to sustain its day-to-day functioning smoothly. Inthis regard, the system has done less well, as current spending on education has remained at 1980 levels-just more than 3 percent o f GDP. It is nonetheless important to note that, even at this level of spending, the sector already claims more than a quarter of the government's total current budget (net of debt interest payment); the share would be even higher with the inclusion o f education-related spending channeled through other government organs. Increased donor funding for the sector, provided flexibly through budget support, could help rebalance spending in favor of recurrent costs. Yet, even under the best of circumstances, there the scope is probably limited for rapid and large increases inthe sector's share on public recurrent spending, whether from domestic or external sources. 17. 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. Spending by households is significant in Rwanda, equivalent to roughly 40 percent of what the government spends. More can be done, however, to shift the burden o f the contribution to post-primary levels of education; progress inthis regard will require a closer look at student finance policies, particularly in higher education. With regard to the government's own expenditures, room exists for improving allocations across levels o f education and categories o f expenditure. Primary education receives only about 45 percent of public recurrent spending on education, whereas higher education gets nearly 40 percent. The heavy focus on higher education, combined with the fact that higher education currently serves only 2 percent of the relevant population, produces a predictably inequitable result. First, Rwanda's unit costs inhigher education are among the highest inthe world today (about 75 times that inprimary education). Second, the best-educated 10percent ina cohort claims more than 70 percent o f the cumulative public spending on education received by the cohort. It is thus no surprise that the Rwandese 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 and the challenge is for policymakers to identify ways to bring about a shift in the desired direction. 18. The report also highlights the potential scope for improving the functional distribution o f public spending on education. It is helpful to conceptualize spending under three rubrics; overhead, service delivery at the facility level, and student welfare. Overall, the system devotes barely two-thirds o f spending to service delivery; it i s natural to wonder ifthe objectives o f the system are being achieved with this pattern of allocation. On closer examination, the main reason for this pattern is that spending in higher education is heavily skewed toward student welfare services and bursaries (for local and foreign studies), which in recent years have accounted for nearly half the government's current spending on the subsector. The distribution is less skewed at the secondary level, but, even then, just the spending on foodstuff for students claims between 5 and 6 percent of recurrent spending. At the primary level, nothing is currently spent on school feeding or other welfare services, but some improvement in allocations can also be made, notably by channeling resources away from management overhead to allow for increased provision o fpedagogical materials. xviii 19. Ensuring that uublic resources for education reach the front lines. In the education sector, the front line may be thought o f as the facility where services reach the intended beneficiaries. The bulk of resources for primary and secondary education reach the fiont line (i.e., the school or classroom) in the form of teachers. In a well-managed system, the number of teachers that a school receives would bear a close relation to the size o f enrollments. InRwanda, the relation is not as tight as one might expect based on experience in other low-income African countries. A primary school serving 500 pupils, for example, might receive as few as five teachers or as many as fifteen. This diversity in allocation implies that resources are inequitably distributed across schools and, therefore, that learning environments are highly disparate. A few public schools are so well endowed that they allow their teachers to handle only one shift o f teaching instead of the usual two required o f most teachers in the public system. Not surprisingly, the better-endowed schools are typically located in Kigali Ville; the worst-endowed ones are often concentrated in such provinces as Byumba, Gisenyi, Kigali-Rural, Kibungo, and Kibuye. Province-level differences explain only a modest part o f the pattern, however. Our analysis suggests that within-province differences are even more pronounced. The corollary is that the process of teacher deployment within each province can probably be better managed to produce greater consistency inthe availability of teachers across schools. 20. Balancing the accessibility o f schools against considerations of scale economies. The spatial distribution o f schools is important because it determines their accessibility to the intended beneficiaries. Yet, the closer that schools are located to students' homes, the more likely they are to serve small catchment areas and, therefore, the higher are their unit costs. H o w then can a goodbalance be struck between accessibility and the cost of service delivery? A s elsewhere, the answer depends on the level of education. At the primary level, children are too young to travel far, so the issue o f accessibility becomes highly pertinent. Yet the current network of primary schools inRwanda tends to emphasize size over proximity: children inhalf the households in a recent survey take more than 30 minutes to reach the nearest primary schools, and the school they attend is likely to be among the three- quarters of the country's public schools that enroll more than 450 students. The emphasis on large schools confers few economies o f scale, however, because unit costs are relatively flat beyond enrollments of about 450 students. In primary education, consideration could therefore be given to extending the network o f schools to situate schools closer to pupils' homes. The result would be to reduce the physical barrier to school participation, as well as reduce the opportunity cost of attending. 21. At the post-primary levels, students are able to travel farther, so physical distance becomes less of a bindingconstraint on enrollments; at the same time, economies o f scale begin to set in because specialized teaching arrangements become increasingly common. The balance between accessibility and scale economies might thus be struck at a different point. In Rwanda, a sizable share o f the schools providing instruction at the tronc commun level tends to be small and operate at highunit costs. Schools with 400 students cost less than two-thirds as muchper student as those enrolling between 100and 200 students and four-fifths as much as schools enrolling between 200 and 300 students. Yet, schools o f the latter sizes enroll more than 20 percent and more than 35 percent, respectively, o f the students at this level o f instruction. Our calculations show that movingto a situation inwhich no school enrolled fewer than 400 students wouldreduce the unit cost of service delivery by 20 percent systemwide, arguably a nontrivial saving in a context of scarce resources. As enrollments expand inthe coming years, it is thus important to accommodate the increase by xix enlarging enrollments in existing schools until they reach an economic size, rather than by buildingmorenew schools for small catchments. 22. In upper secondary education, as in higher education, enrollments are relatively small and are likely to grow slowly as long as the absorptive capacity o f the modern sector labor market remains limited. This tendency implies that it is even more important for schools and institutions to take advantage of scale economies in service delivery; this may require some consolidation o f their course offerings. InRwanda, only 14 percent of students at the upper secondary level attend a school serving 400 or more students; students at each school tend to be spread thinly across fields of specialization. In higher education, evidence also exists o f duplication across public institutions and between the public and private sectors insome fields of study. These results suggest that, inmanaging the expansion of these levels, policymakers might seek to minimize a proliferation of course offerings, particularly where demand is weak and the potential for scale economies in service delivery is limited. 23. Managing; classroom conditions and processes to enhance student learning. A crucial test of an education system's performance is its effectiveness in transforming the resources at its disposal into learning outcomes. Inbothprimary and secondary education, the available evidence suggests room for improvement, as the performance o f schools as measured, albeit imperfectly, by examination scores relates only weakly to the level of resources they receive. The result raises two types of questions: the effectiveness of the mix of school inputs that support service delivery and, more important, the incentives for performance. 24. With regard to input mix, the ideal situation ina well-funded world would be for pupils to receive sufficient institutional time, be taught in small classes by well-qualified teachers, and be given access to an ample supply of books and other learning materials. In resource-constrained environments, tough trade-offs present themselves, and policymakers must find the mix of inputs that works best under the constraints. InRwanda, the inputmix in public primary schools favors teacher qualification at the expense of class size and instructional time. The country's pupil-teacher ratio at 57 pupils per teacher is currently among the highest inthe world. The highratio translates into a situation where most teachers in the first three grades of the cycle are required to teach two shifts of pupils. Correspondingly, the instructional time for pupils in these grades averages only about 500 hours a year, compared with 1,000 hours inthe upper grades and with the target range of 850 to 1,000 hours targeted by other low-income countries based on international best practices. Reducing the pupil-teacher ratio (i.e., hiringmore teachers) would help address the problem, but ina context of tight budgets, this will require adjustment to the input mix. Adjusting the level at which teachers are recruited is one possible option to manage that trade-off. All new primary school teachers are currently hired from among those who have completed upper secondary education. Although recruiting at this level has its advantages intheory (and may be appropriate as a long-run objective), in the current context of Rwandese schools, such teachers are no more effective than those with only lower secondary education. Given the differences inwages between the two types of teachers, it may be worth taking a closer look at the two ways to achieve a more effective mix o f school inputs. 25. In secondary education, the issue of teacher qualification is also highly relevant. Inthe tronc commun cycle, most of the teachers satisfy the minimumqualification (ie., at least an upper secondary school diploma) and some even have a university degree. In contrast, half the teachers in upper secondary schools are probably underqualified, that is, have only an upper secondary school diploma. Any strategy to improve learning outcomes must thus seek to raise the educational profile of teachers inthe upper secondary cycle. The obvious solutioni s to set and apply clear standards for recruitment, but ifthe problem i s to be solved more quickly, policymakers might consider rationalizing the deployment o f the current cadre o f secondary school teachers, reassigning them to the extent feasible between the tronc commun andupper secondary cycles according to their qualifications. 26. As much as choosing the right input mix is important, it is by no means the entire answer to poor performance. Because schooling i s a social process involving multiple actors, outcomes ultimately depend on how people behave. Putting in place the right incentives to align behavior with the goal o f enhancing student learning is thus a key challenge inmanaging for results. The potential interventions inthis regard are as diverse as the context in which people live and work, but the principle o f establishing clear accountabilities and matching them with spending and management authority at all levels i s probably relevant everywhere. Country experience has also highlighted the crucial role o f defining and monitoringtangible indicators o fprogress as a way to increase the incentives for better performance. 27. Minimizing the barriers to education for oruhans and other vulnerable groups. As indicated above, socioeconomic gaps inenrollments are narrower inRwanda than inother low-income countries. Yet, the result should not provide a reason for complacency. More can and should be done to reach the most vulnerable children, which includes double orphans, street children who do not benefit from systematic adult supervision, children living inrural areas, and children from the poorest 40 percent o f households. Double orphans (i.e. children who have lost both parents)--arguably the most vulnerable children-are easily identifiable, and systematic efforts can and must be made to improve their prospects inlife. For the other groups, the assistance needed i s likely to be diverse; exploratory and pilot interventions are probably appropriate to find the best way to serve them. 28. In secondary and higher education, participation rates among children from the top income quintile far outstrip those in the rest o f the population. Girls are as likely as boys to enroll in secondary school, but they lag significantly behind in higher education. Regarding the choice of interventions to narrow the socioeconomic disparities inparticipation rates, the report'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, the intervention probably needs to be combined with efforts to improve learning outcomes. Such efforts are important, because in a meritocratic selection system based on examination results, enlarging the representation o f lagging groups is possible only to the extent that these groups are able to compete for the coveted places inpost-primary education. Conclusion 29. Nearly a decade after the genocide, Rwanda's leaders can look back with pride and satisfaction at the record o f achievements. They have put a devastated system back on its feet: classrooms have been repairedand new ones built to accommodate the growing number o f 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 created specifically to assist orphans; and, inhigher education, a much diversified system has beencreated. 30. Yet, the task ahead remains daunting as the recovery phase gives way to the work of nurturingthe sector's long-term development. 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 i s hoped that the findings of this study can contribute to the discussion by creating a common understanding of. the isses and by drawing attentionto some o fthe emerging challenges. xxiii L'EDUCATION AU RWANDA : RE-EQUILIBERLESRESOURCESENW E D'ACCELERERLEDEVELOPPEMENTET LA REDUCTIONDELA PAUVRETE DANSUNENVIRONNENEMENTAPRES CONFLIT ResumeAnalytique Contexte 1. L'histoire rkcente du Rwanda a ktk perturbke par un genocide en 1994. Au cours de ce genocide, au moins 800.000 personnes, soit environ 10 pour cent de la population, ont perdu la vie. L a stabilitk et la skcuritk ont 6tk restaurkes et le processus de redressement est en cours depuis plusieurs annkes. L a reconstruction du potentiel de capital humain qui a disparu dans le massacre est une partie importante de ce processus. Par ailleurs, le gouvernement a kgalement fait de gros efforts pour compenser le temps perdu dans l'klargissement de l'accks itl'kducation et le renforcement de la qualitC des services. Eneffet, le progrks en kducation figure parmi les objectifs essentiels de la stratkgie de dkveloppement kconomique et social du gouvernement comme prksentk dans son recent document de stratkgie de rkduction de la pauvretk (DSRP). 2. Sur le plan international, le secteur de l'kducation a kgalement Ctk sous les projecteurs de l'actualitk ces dernikres annkes. Au titre de l a declaration des Nations Unies pourle millknium2000, la communautk internationale a accept6 d'ceuvrer pour s'assurer que d'ici 2015, partout, les garqons comme les filles pourront achever une scolaritk primaire complkte, et tendre vers l'klimination des disparitks entre les genres dans l'enseignement primaire et secondaire, idkalement en 2005 au plus tard. Ce sont 18 deux des huit objectifs de dkveloppement du millknium (ODM) explicitement knoncks dans la dkclaration. D'aprks la confkrence internationale de Monterrey de 2002 sur le financement du dkveloppement, ilse dkgage aussi un consensus que la rkalisation des objectifs nkcessitera l'action conjointe des pays riches et pauvres. Les pays riches doivent accroitre l'aide aux pays pauvres (entre autres mesures) qui, A leur tour, doivent mettre en place des politiques approprikes et des structures de bonne gouvernance pour assurer l'utilisation efficace des ressources afin d'atteindre les objectifs vises. L e principe de lier les ressources aux rksultats sur le terrain, commence ii imprkgner la discussion sur les politiques de dkveloppement au niveau du pays. L e secteur kducatif est l'un des nombreux secteurs qui rkclament des ressources publiques rares et la force de sa revendication dkpend de plus en plus de sa capacitk de fournir des rksultats tangibles. 3. L e contexte antkrieur prksente des dkfis clairs aux responsables du systkme kducatif au Rwanda. Pour attirer des ressources accrues aux niveaux national et international, le secteur doit mettre en kvidence la bonne gestion des ressources dkja disponibles. Quelle est la possibilitk d'amklioration a cet kgard ? Quelles politiques sont nkcessaires pour s'assurer que le secteur se dkveloppe de faqon efficace, kquitable et fiscalement durable? Quelles lacunes de financement au plan national persistent et que les ressources externes pourraient utilement combler? Sur quels rksultats pourrait-on concentrer l'effort en vue de crker les structures approprikes de responsabilitk dans le systkme? xxiv But,audienceet portCe 4. Ce rapport se propose de foumir unensemble de faits qui serviront de base de discussion des questions poskes ci-dessus. I1 est bask sur des donnkes collectkes jusqu'a 2001, demibre annke au cours de laquelle ila ktk possible d'obtenir des informations statistiques approprikes dans le cadre de la prkparation de ce rapport. L e systbme kducatif n'est pas rest6 statique; le tableau que ces donnkes reprksentent ne couvre donc pas les rkcents dkveloppements intervenus dans le secteur. Ces changements ont lieu a cause de la prkvoyance du Gouvernement qui cherche 21 prendre en compte certaines des contraintes kmergeantes dans le dkveloppement du secteur. Les efforts de reduction des effectifs dans les classes du primaire retiennent l'attention: en effet, les rksultats du dernier recensement scolaire de 2003 ont montrk que des progrbs remarquables ont ktk rkalisks dans ce sens. D e la mQme manibre, des rkformes ont ktk lanckes dans les finances de 1'Enseignement Supkrieur, en vue de rkduire les coats des ktudes a l'ktranger supportks par le Gouvernement (des ktudiants ont ktk envoy& dans des pays moins chers comme 1'Afrique du Sud et 1'Inde); les coats des ktudes faites sur place ont ktk kgalement rkduits (les bourses ont kte transformkes en prets, contrairement aux subventions d'antan; une Agence de Financement des Etudiants, visant a institutionnaliserle remboursement des prets, a ktk mise en place). 5. Considirk comme unexercice permettant de faire le point de la situation en un moment donnk (et ceci jusqu'en 2001)' le document serait mieux identifik comme un document diagnostic destine A contribuer ti une comprkhension commune de la performance actuelle du systbme kducatif au Rwanda, les contraintes pour son progrbs, et les choix a faire dans la planification du dkveloppement du systbme au cours des annkes a venir. L'instauration d'un consensus sur les questions constitue une partie intkgrale de l a politique de dkveloppement qui pourrait suivre le processus de consultation avec les partenaires et la sociktk civile dans son ensemble, que beaucoup de gouvernements, y compris celui du Rwanda, ont dkjk initik en dkfinissant leurs stratkgies de rkduction de l a pauvretk. En Cducation, l'importance du processus peut a peine Qtreexagkrke: en fin de compte le succbs des politiques dkpend de la faqon dont elles sont mises en ceuvre et ceci dkpend a son tour, a undegre important,de la bonne comprkhension des problbmes, enparticulier parmi ceux qui doivent provoquer les reformes nkcessaires, y compris les dkcideurs, les planificateurs de l'kducation, les enseignants, directeurs d'kcole, parents et ktudiants et du degrk d'entente sur les solutions proposkes. 6. Ence quiconcerne laportke, lerapport neprktendpas aborder toute la gamme de questions auxquelles les dkcideurs du Rwanda pourraient faire face. L'ktendue de sa couverture est limitke aux aspects kconomiques clks qui sont particulibrement appropriks dans le contexte de DSRP. Ceux-ci concernent le coat, les finances, la prestation de service et les rksultats de I'kducation, en particulier les aspects qui se prQtent a la quantification. Tout en admettant son caractbre incomplet, le document fournit une image des grandes caractkristiques structurales du systkme et du modble implicite d'attribution des ressources et de l'efficacitk de la prestation de service. Les principales dkcouvertes sont rksumkes ci- dessous. Progrcsdel'enseignementjusqu'h cejour 7. En dkpit des ravages indescriptibles provoquks par le genocide de 1994, le systbme kducatifrwandais a rkcupkre remarquablement bien, au moins en termes quantitatifs. En concevant le dkveloppement futur du secteur, ilest donc important de commencer par XXV prksenter certaines des realisations remarquables et la base qu'elles foumissent pour dkvelopper l e systbme dans les prochaines annees. 8. L'aspect le plus impressionnant du redressement du systbme est le rythme rapide de l'augmentation de l'inscription aprbs le genocide. Seulement 5 ans aprhs l'kvenement, le nombre d'enfants inscrits A l'kcole avait d k j i depassk le nombre qui aurait ktk inscrit si l e systbme avait connu une croissance aux taux de progression historiques. Actuellement A 107 pour cent, le taux brut d'inscription dkpasse ce niveau le taux correspondant pour la moyenne des pays faible revenu en Afrique a ce jour. Dans l'enseignement secondaire, le nombre d'ktudiants a augmentk de 20 pour cent par an depuis 1996, impliquant que le systbme est maintenant presque trois fois plus grand que ce qu'il ktait prkckdemment. Tandis que le taux brut d'inscription demeure moyenne pour 1'Afrique subsaharienne des pays a faible a niveau en-dessous de la pour cent - I'kcart aurait Ctk encore plus grand si le systbme avait stagnk aprbs le genocide. revenu -ce13 pour cent au lieu de 20 Dans l'enseignement supkrieur, les inscriptions ont grimpk plus rapidement encore, allant de 3 400 ktudiants en 1990-1991 a presque 17 000 en 2001-2002, une augmentation qui a presque quadruple en l'espace d'une decennie. L a couverture du systbme est maintenant comparable i?i l a moyenne d'environ 200 ktudiants pour une population de 100 000 en Afrique subsaharienne a faiblerevenu. 9. L e systbme s'est dkveloppk tout en maintenant un bon Cquilibre entre les secteurs public et ~rivk.A la base de la pyramide de l'enseignement, uneffort considkrable a et6 constamment fait par le gouvemement en vue d'ktendre la couverture du secteur public. Enconskquence, laplupart des inscriptions dans les kcoles privkes est demeurke infkrieure moins de 1 pour cent. Au niveau secondaire, les inscriptions ont augmentk aussi rapidement dans le secteur public que dans le privk dans les annees d'aprbs gknocide, et la proportion des dtudiants inscrits dans les kcoles privkes est demeurke stable itenviron 40 pour cent, ce qui est plus bas que les 62 pour cent des annkes 80, mais toujours beaucoup plus klevk que les 20 pour cent en moyenne dans les pays faible revenu en Afrique subsaharienne. Dans l'enseignement supkrieur, le secteur privk s'est dkveloppk tant bien que mal pendant les deux dkcennies jusqu'h 2000, mais il rend indubitablement service ti une part croissante d'ktudiants, environ 38 pour cent en 2001-2002 au lieu d'environ 8 pour cent au debut des annkes 80. L a diversite des ktablissements post-secondaires et le mklange de prestataires publics et privks du secteur constituent une force du systbme, le dotant d'une flexibilitk pour satisfaire la demande croissante de places a ce niveau d'ktudes. 10. I1est important de noter que l'expansion du systbme s'est effectuee dans une structure qui prkvoit uncycle primaire de six ans, un cycle de trois ans de tronc commun (ou uncycle secondaire infkrieur), uncycle secondaire supkrieur de trois ans et uncycle de quatre ans d'enseignement supkrieur dans la plupart des domaines. Un systbme meritocratique d'examen a 6te mis en place pour contrdler l a sklection des ktudiants entre les divers cycles d'ktudes. La structure kducative 6-3-3-4, avec le mkcanisme de sklection bask sur le mkrite fournit un systkme raisonnablement structurk pour contrdler l'expansion du secteur. Au Rwanda ily a lieu de remarquer que les enfants qui ont achevk six ans d'instruction primaire savent, presque toujours de faqon permanente, lire et compter comme des adultes. Universaliser l'achbvement de 1'instruction primaire est ainsi entibrement confonne a ce qui est nkcessaire pour dkvelopper l a base du capital humain pour un developpement kconomique et social klargi, et le systbme est deja structurk en conskquence. EnmBme temps, les arrangements de sklection pour la progression aux niveaux post-primaires sont kgalement dkja en place, donnant aux responsables du systbme les leviers administratifs pour adapter le xxvi rythme de l'expansion a la disponibilitk des ressources et h l a capacitk d'absorption d'une main d'ceuvre hautement qualifike. 11. Endehors des ressources du systbme dans son ensemble, le systbme rwandais se compare favorablement a celui des autres pays disparitks sociodconomiques dans l'accbs a a faible revenu en Afrique, en termes de l'enseignement, particulibrement au niveau de l'enseignement primaire. Un fait fiappant est que les taux de participation a l'kcole sont relativement klevks, mQmeparmi les orphelins. L e gknocide de 1994 a laissk le Rwanda avec undes taux d'orphelins lesplus Clevks au monde, avec presque 40 pour cent des enfants Bgks de 7 a 14 ans en 2000 ayant perdu au moins unparent. L'kcart dans les inscriptions entre les orphelins et d'autres enfants n'est remarquable que parmi les enfants les plus vulnkrables (ceux qui ont perdu les deux parents ou ceux qui vivent sCparks de leurs parents par exemple). Cette situation est due a l'existence de filets de skcuritk relativement bien dkveloppks qui ont en quelque sorte permis d'assurer des taux de participation klevks a l'kcole primaire, mQmepanni les orphelins. Dans l'enseignement secondaire, les orphelins sont au moins aussi bien reprksentks que les non-orphelins. Ce rksultat remarquable doit beaucoup la dkcision du gouvernement en 1998 d'ktablir le Fonds de GCnocide, aliment4 par une taxe, comme unmkcanisme pour aider les orphelins dans l'enseignement secondaire. Les dCfis de l'avenir 12. Dans les annkes a venir, les dkfis seront des problbmes de deuxibme gknkration, qui commencent dkjh se poser, suite A l'heureux rebondissement du systbme aprbs l a dkvastation du genocide. L'accent se dkplace ainsi d'une situation ou les mesures urgentes pour rktablir le fonctionnement du systbme ktaient prioritaires, a une situation oG la prkoccupation consiste dkveloppement a long terme du secteur. Un sommaire a donner une direction approprike et fiscalement soutenable au de certaines des questions principales au niveau systkmique auxquelles les dkcideurs devront probablement s'attaquer a l'avenir est prksent ci-dessous. L e lecteur est prik de se reporter aux chapitres individuels pour des rksultats dktaillks sur les points spkcifiques quine sont pas mentionnks ci-dessous. 13. Gestion du flux d'ktudiants et rendement en diD18mks du svstbme. Atteindre le but de donner A tous les enfants la chance d'achever unenseignement primaire complet exige de prQter attention a deux prkoccupations: augmenter les taux d'entrke en premiere annke et renforcer la probabilitk que les enfants continuent jusqu'a la fin du cycle. Au Rwanda, les taux d'entrke ont ktk historiquement klevks et atteignent 90 pour cent. Ainsi le dkfi ici est d'identifier et d'aider les demiers 10 pour cent de la population quirefusent d'aller a l'kcole. En ce qui concerne la survie jusqu'a la fin du cycle, la performance du Rwanda est respectable: le taux actuel, estimk a 73 pour cent, se compare favorablement au taux dans d'autres pays a faible revenu et au de 44 pour cent au Rwanda dans les annkes 1990-1991. Cependant, ilest peu probable que ce bon rksultat persiste; en effet, le taux de redoublement de classe dans le systbme est particulibrement klevk - environ 34 pour cent en 2000 -2001, soit plus de 3 fois le taux de la dkcennie prkckdente. Sur la base de ce qui est gknkralement considkrk comme bonne pratique selon l'expkrience transnationale, ilserait raisonnable que le pays 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 A un autre, et d'amkliorer les rksultats de l'apprentissage afin de rkduire au minimum le besoin de faire doubler les classes aux klbves. xxvii 14. Au-dela de l'enseignement primaire, les pressions en vue l'accbs commencent dkja 21 se faire sentir; en effet, de plus grandes cohortes d'enfants achevent maintenant les cycles infkrieurs d'instruction. Pour dkvelopper des politiques approprikes rkgissant la progression des ktudiants de cycle en cycle, il est utile de distinguer le cycle de tronc commun du cycle secondaire supkrieur. L e premier peut Qtre considkrk comme une continuation du cycle primaire; ainsi, dans la mesure oh les ressources le permettent, ilserait approprik de viser l'accbs universe1 a ce cycle comme objectif a moyen terme. L e cycle secondaire supkrieur quant A lui est considkrk a juste titre comme une phase preparatoire a l'enseignement supkrieur. Les diplbmks de ces niveaux sont formks grands frais pour des emplois dans l'kconomie moderne; ilest donc important de s'assurer que le nombre de diplbmks formks et la diversitk de leurs qualifications sont compatibles avec l a demande prospective de main d'aeuvre qualifike. Dans le contexte post-conflit immkdiat, les pknuries croissantes de travailleurs hautement qualifies ktaient senties dans I'administration et dans le secteur privk, et ces pknuries ont stimulk une croissance rapide des inscriptions, en particulier dans l'enseignement supkrieur. Maintenant que le potentiel de capital humain s'est graduellement reconstituk en partie par l'immigration de la diaspora rwandaise et que des signes de chbmage potentiel des diplbmks commencent A apparaitre, il serait sage de rkexaminer l a situation et d'adapter les politiques d'admission dans l'enseignement secondaire supkrieur et l'enseignement supkrieur en conskquence. 15. Mobilisation et utilisation efficiente des ressources Dour l'enseignement. L'appui au dkveloppement de I'enseignement a ktk et continue d'&e solide. Au cours des dernieres annkes, la convergence de l'opinion publique et des engagements internationaux et la qxioritisation )>du secteur par le gouvernement ont contribuk a faire monter les dkpenses publiques sur l'enseignement au Rwanda au taux historiquement jamais atteint de 5'5 pour cent du PIB en 2001. Cependant, un examen approfondi rkvkle que l'augmentation a ktk dominke par des dkpenses d'investissement en infrastructure. Tandis que de tels investissements sont nkcessaires pour remettre en ktat les installations dktruites et en augmenter l a capacitk, tout le systbme a un certain moment nkcessiterait des ressources adequates pour soutenir son fonctionnement rkgulier. A cet kgard, le systkme a moins rkussi; en effet, les dkpenses actuelles pour l'enseignement sont demeurkes au niveau de celles des annkes 1980, juste au-dessus de 3 pour cent du PIB. I1est nkanmoins important de noter que mQme a ce niveau des dkpenses, le secteur bknkficie dkja de plus du quart dubudget actuel du gouvernement (moins le paiement des intkrets de la dette); la contribution du Gouvernement serait encore plus klevke si on incluait les dkpenses likes A l'enseignement drainkes par d'autres organes du gouvernement. L'augmentation du financement des donateurs du secteur, pour une utilisation flexible a travers l'appui au budget, pourrait aider 21 rkkquilibrer les dkpense en faveur des coats rkcurrents. Cependant, mQmedans le meilleur des cas, les chances pour une augmentation importante et rapide des ressources du secteur dans les dkpenses publiques rkcurrentes sont limitkes, aussi bien dans le cadre des ressources nationales qu'extkrieures. 16. Faire un meilleur usage des ressources existantes et s'en servir pour susciter des contributions privkes doit donc etre une piece maitresse des efforts pour faire avancer les objectifs du secteur. Les depenses des menages sont importantes au Rwanda, kquivalentes a approximativement 40 pour cent des dkpenses du gouvemement. Onpourrait faire davantage cependant, pour faire passer l a charge de la contribution des menages aux niveaux post- primaires de l'enseignement; le progrbs cet kgard exigera un regard plus attentif aux politiques de financement de l'Ctudiant, en particulier dans l'enseignement supkrieur. En ce qui concerne les dkpenses du gouvernement, ilserait nkcessaire d'amkliorer les allocations xxviii aux divers niveaux d'enseignement et aux catkgories de dkpenses. L'enseignement primaire ne reqoit que 45 pour cent des dkpenses rkcurrentes publiques consacrkes a l'kducation, alors que l'enseignement supkrieur en obtient presque 40 pour cent. L a trbs forte concentration sur l'enseignement supkrieur combink avec le fait que l'enseignement supkrieur ne reprksente prksentement que 2 pour cent de la population concemke produit un rksultat inkquitable prkvisible: les coilts unitaires de l'enseignement supkrieur au Rwanda figurent panni les plus Clevks au monde (environ 75 fois ceux de l'enseignement primaire) et les 10 pour cent les mieux formks dans une cohorte absorbent plus de 70 pour cent des dkpenses publiques cumulatives a l'kducation regues par la cohorte. I1n'est donc pas surprenant que le systbme rwandais soit l'un des moins structurellement kquitables en Afiique subsaharienne. L a rkorientation des dkpenses vers l'enseignement primaire est le seul moyen de rectifier la situation. C'est un dkfi pour les dkcideurs d'identifier les moyens de provoquer une rkorientation vers l a direction viske. 17. L e rapport souligne kgalement le champ potentiel d'amklioration de la distribution fonctionnelle des dkpenses publiques sur l'kducation. I1 est utile de conceptualiser les dkpenses sous trois rubriques: frais gknkraux, prestation de service au niveau des kcoles et biendtre de l'ktudiant. D e faqon gknkrale, le systbme consacre deux tiers des dkpenses a a peine l a prestation de service et ilest normal de se demander si les objectifs du systbme sont atteints avec ce modkle d'allocation. A l'examen approfondi, la raison principale de ce modble s'explique par le fait que les dkpenses sur l'enseignement supkrieur sont fortement orientkes vers les services d'assistance sociale de l'ktudiant et les bourses d'ktude (pour des ktudes dans le pays et al'ktranger) qui, ces dernikres annkes, ont reprksentk presque la moitik des dkpenses courantes du gouvemement pour le sous secteur. L a distribution est moins dkvike au niveau secondaire, meme ici les dkpenses pour l'alimentation des klkves reprksentent 5 a 6 pour cent des dkpenses rkcurrentes. Au niveau primaire aucune dkpense n'est actuellement faite pour la restauration A l'kcole ou d'autres services d'assistance sociale, mais on peut kgalement amkliorer les allocations, notamment en diminuant les frais gknkraux de gestion en faveur de l'acquisition d'une plus importante quantitk de materiels didactiques. 18. S'assurer que les ressources uubliaues destinkes l'kducation atteignent les lignes de front. Dans le secteur de l'kducation, la ligne de front peut etre considkrke comme une installation oh les services atteignent les bknkficiaires prkvus. L a majeure partie des ressources de l'enseignement primaire et secondaire atteint l a ligne de front (a savoir l'kcole ou la salle de classe) sous forme d'enseignants. Dans un systkme bien gkrk, le nombre d'enseignants d'une kcole aurait une relation ktroite avec l'importance des inscriptions. Au Rwanda, cette relation n'est pas aussi ktroite que l'on pourrait s'y attendre en se basant sur l'expkrience d'autres pays afiicains a faible revenu. Une kcole primaire accueillant 500 klkves, par exemple, pourrait recevoir aussi peu que 5 enseignants ou autant que 15. Cette diversitk dans l'attribution implique que des ressources sont inkquitablement distribukes aux kcoles et que par conskquent les conditions d'apprentissage sont fortement disparates. Quelques kcoles publiques sont si bien dotkes qu'elles peuvent permettre A leurs enseignants d'assurer seulement un flux d'enseignement au lieu des deux habituellement exigks de la plupart des enseignants dans le systkme public. Comme o npourrait s'y attendre, les kcoles les mieux dotkes sont typiquement situkes dans Kigali Ville; les mal loties sont souvent concentrkes dans des provinces telles que Byumba, Gisenyi,Kigali-Rural, Kibungo et Kibuye. Les diffkrences de niveau au sein des provinces n'expliquent cependant qu'une petite partie dumodble. Notre analyse suggbre que les diffkrences l'intkrieur d'une province sont mCme plus prononckes. L e corollaire est que le processus du dkploiement des enseignants dans xxix chaque provincepeut probablement Qtremieux gCrC pour produire une plus grande uniformit6 dans la disponibilitk des enseignants au sein des Ccoles. 19. Eauilibrer l'accessibilitk des Ccoles avec les considbrations d'Cconomies d'Cchelle. L'implantation spatiale des Ccoles est importante parce qu'elle d6termine leur accessibilitk aux bCnCficiaires cibles. Cependant, plus les &coles sont situCes a proximitd des habitations des Clbves, plus ilest probable qu'elles servent de petits secteurs de recrutement scolaire, et donc plus ClevCs sont les coats unitaires. Comment trouver alors unbon 6quilibre entre 1'accessibilitC et le coat de la prestation de service ? Comme ailleurs, l a rkponse dCpend du niveau de l'enseignement. Au niveau primaire, les enfants sont trop jeunes pour faire de grands dkplacements; la question de 1'accessibilitC devient alors fortement pertinente. Pourtant le reseau actuel des Ccoles primaires au Rwanda tend a privilCgier la taille a la proximitk: les enfants dans la moitiC des menages, selon une enquete rCcente, mettent plus de 30 minutes pour atteindre les Ccoles les plus proches; et ces Ccoles sont susceptibles d'Qtre parmi les trois quarts des Ccoles publiques du pays qui accueillent plus de 450 Clbves. L'accent mis sur les Ccoles de grande taille permet peu d'6conomies d'6chelle cependant, puisque les c o b unitaires sont relativement uniformes au-deli de l'inscription d'environ 450 Clbves. Dans l'enseignement primaire, une attention pourrait donc QtreaccordCe l'extension durCseau des Ccoles afin de les situer plus prbs des habitations des Clbves. L e resultat devrait Qtre de rkduire l a barribre physique 21 la participation scolaire ainsi que de rdduire le coat d'opportunitk de la fikquentation. 20. Aux niveaux post-primaires, les Clbves peuvent voyager plus loin, et la distance physique devient moins une contrainte qui limite les inscriptions; par la mQme occasion, les Cconomies d'kchelle commencent a apparaitre parce que les installations pour les enseignements spCcialis6s deviennent de plus en plus communs. L'Cquilibre entre I'accessibilitC et les Cconomies d'kchelle pourrait ainsi Qtre diffkremment rCalis6. a Au petite taille qui fonctionnent a des coats unitaires dlevCs. Les dcoles de 400 Clbves coatent Rwanda, une part considkrable des Ccoles du cycle dutronc communtend Qtredes Ccoles de deux tiers moins cher par ClBve que celles qui accueillent 100 200 Clbves, et elles coQtentles quatre cinquibmes de celles qui accueillent 200 a 300 Clbves. Pourtant les Ccoles de ces deux dernibres catbgories accueillent respectivement plus de 20 pour cent et plus de 35 pour cent des 61bves a ce niveau d'enseignement. Nos calculs montrent qu'une situation oh aucune Ccole n'enregistrerait pas moins de 400 Clbves rkduirait le coQt unitaire de la prestation de service de 20 pour cent au niveau de l'ensemble du systbme, probablement une Cconomie non nkgligeable dans un contexte de ressources rares. Ainsi, comme les inscriptions augmenteront dans les annCes 21 venir, il est important d'harmoniser l a croissance en augmentant les inscriptions des Ccoles existantes jusqu'a ce qu'elles atteignent une taille Cconomiquement acceptable, plut6t qu'en construisant plus de nouvelles Ccoles pour de petits secteurs de recrutement. 21. Dans l'enseignement secondaire supCrieur comme dans l'enseignement supkrieur, les inscriptions sont relativement faibles et sont susceptibles de se dCvelopper lentement aussi longtemps que la capacite d'absorption du march6 du travail du secteur moderne demeure limitke. Cette tendance implique qu'il est mQme plus important pour les Ccoles et institutions de tirer profit des Cconomies d'bchelle dans la prestation des services et ceci peut exiger une certaine consolidation de leurs offres de cours. Au Rwanda, seuls 14 pourcent des Clbves du secondaire supCrieur vont A une Ccole de 400 Ctudiants ouplus, et les Ctudiants de chaque Ccole ont tendance a se disperser dans plusieurs domaines de sphcialisation. Dans l'enseignement supkrieur, ily a Cgalement une Cvidence de duplication entre les institutions publiques d'une part, et entre les secteurs public et privk d'autre part, dans certains domaines d'ktude. Ces rksultats suggbrent qu'en contrdlant l'expansion de ces niveaux, les dkcideurs pourraient chercher a rkduire au minimumla prolifkration des offres de cours, particulibrement la OG la demande est faible et le potentiel pour des kconomies d'kchelle dans la prestation de service limitk. 22. Gkrer les conditions et les processus de classe en vue d'amkliorer l'amrentissage de l'ktudiant. Un test crucial de performance du systbme kducatif est son efficacitk dans la transformation des ressources mises A sa disposition en rksultats d'apprentissage. Dans l'enseignement primaire et secondaire, ilapparait kvident que des possibilitks d'amkliorer la performance des kcoles existent; en effet, la performance des kcoles, mesurke (mQme de fagon imparfaite) par les rksultats aux examens, n'a qu'un rapport trbs faible avec le niveau des ressources qu'elles regoivent. Ceci soulbve deux types de question: l'efficacitk de la combinaison des ressources qui soutiennent l a prestation des services; et de fagon trbs importante, les incitations a la performance. 23. En ce qui conceme la combinaison des ressources, la situation idkale, sans contrainte financibre, serait que les klbves disposent d'un temps institutionnel suffisant, soient enseignks dans de petites classes par des enseignants compktents et aient accbs a un approvisionnement suffisant en livres et autres matkriels didactiques. Mais dans un environnement de ressources limitkes, des choix difficiles doivent se faire et les dkcideurs doivent trouver les combinaisons de ressources qui fonctionnent le mieux sous ces contraintes. Au Rwanda, la combinaison des ressources dans les kcoles primaires publiques favorise la qualification des enseignants aux dkpens de la taille de la classe et du temps d'instruction. L e ratio klbve-enseignant dupays se situe actuellement parmi les plus klevks au monde avec 57 klbves par enseignant. Suite a ce ratio klevk, ilest exigk de l a plupart des enseignants des trois premiers niveaux du cycle d'enseigner deux vagues d'klbves. En conskquence, le temps d'instruction des klbves dans ces niveaux s'klbve en moyenne a environ cinq cents (500) heures par an, au lieu de mille (1000) heures dans les niveaux supkrieurs et de la fourchette de huit cent cinquante (850) A mille (1000) heures que d'autres pays A faible revenu visent prksentement en se basant sur les meilleures pratiques intemationales. Rkduire le ratio klbve-enseignant (ce qui nkcessiterait de recruter plus d'enseignants) aiderait 5 rksoudre le problbme, mais dans uncontexte de budgets serrks, ceci exigera un ajustement de l a combinaison des ressources. L'ajustement du niveau de recrutement des enseignants est une option possible pour gkrer ce choix. Actuellement tous les nouveaux enseignants sont recrutks parmi ceux qui ont achevk l'enseignement secondaire supkrieur. Tandis que le recrutement a ce niveau a ses avantages en principe (et peut Qtre approprik comme objectif a long terme), la rkalitk est que de tels enseignants ne sont pas plus efficaces que ceux qui n'ont achevk que l'enseignement secondaire infkrieur, dans le contexte actuel des kcoles rwandaises. Etant donnk les diffkrences de salaires entre les deux types d'enseignants, ilpeut Qtre intkressant d'analyser attentivement les deux manibres de rkaliser une combinaison plus efficace des ressources scolaires. 24. Dans l'enseignement secondaire, la question de la qualification des enseignants est aussi hautement pertinente. Dans le cycle du tronc commun, l a plupart des enseignants satisfont a la qualification minimum (ce qui veut dire qu'ils ont au moins un dipl6me secondaire supkrieur) et certains ont mQme un dipl6me d'universitk. Enrevanche, la moitik des enseignants dans les kcoles secondaires supkrieures sont probablement sous qualifiks, n'ayant eux-mbmes qu'un dipldme secondaire supkrieur. Toute stratkgie pour amkliorer les rksultats de l'apprentissage doit donc chercher arelever le profil kducationnel xxxi des enseignants du cycle secondaire supkrieur. L a solution kvidente est de fixer et appliquer des nonnes claires pour le recrutement; mais si le problbme doit Qtre rksolu plus rapidement, les dkcideurs pourraient chercher a rationaliser le dkploiement du cadre actuel des enseignants de cycle secondaire en les rkaffectant (dans la mesure du possible) entre le tronc commun et le cycle secondaire supkrieur selon leur qualification. 25. Autant le choix de la bonne combinaison des ressources est important, ilne constitue en aucun cas toute l'explication d'une la faible performance. Puisque la scolarisation est un processus social qui implique des acteurs multiples, les rksultats dkpendent finalement du comportement des gens. Mettre en place les bonnes incitations pour aligner les comportements avec l'objectif de relever l'apprentissage de 1'618ve constitue ainsi undkfi c16 dans la gestion ciblant les rksultats. Les interventions potentielles a cetkgard sont aussi diverses que le contexte dans lequel les gens vivent et travaillent, mais le principe d'ktablir des responsabilitks claires et de les associer avec l'autoritk de dkpense et de gestion a tous les niveaux est probablement approprik partout. L'expkrience du pays a kgalement soulignk l'importance de la dkfinition et du contr6le d'indicateurs tangibles de progr8s comme moyen d'augmenter les incitations pourune meilleure performance. 26. Rkduire au minimum les barribres a l'kducation pour les orphelins et autres groupes vulnkrables. Comme prkckdemment indiquk, les dcarts socio-kconomiques dans les inscriptions sont plus rkduits au Rwanda que dans d'autres pays a faible revenu. Ce rksultat ne devrait cependant pas pousser ti la complaisance. On peut et on devrait faire davantage pour atteindre les enfants les plus vulnkrables, y compris les orphelins de pbre et de mbre, les enfants de la rue qui ne bknkficient pas de la surveillance systkmatique des adultes, les enfants vivant dans les zones rurales et ceux des 40 pour cent des menages les plus pauvres. Les doubles orphelins -manifestement les plus vulnkrables - sont facilement identifiables et des efforts systkmatiques peuvent et doivent Qtrefaits pour amkliorer leurs perspectives dans la vie. Pour les autres groupes, l'aide nkcessaire devrait Qtrediversifike et des interventions de nature exploratoire et pilote sont probablement approprikes pour trouver la meilleure manibre de les servir. 27. Dans 1'enseignement secondaire supkrieur, les taux de participation des enfants du quintile supkrieur dkpassent de loin ceux du reste de la population. Les filles ont les mQmes chances que les garqons de s'inscrire dans l'enseignement secondaire, mais elles trainent sensiblement derribre les garqons dans l'enseignement supkrieur. En ce qui conceme le choix des interventions pour rkduire les disparitks socio-kconomiques des taux de participation, les conclusions du rapport suggbrent que l'aide financibre pourrait aider les groupes a l a traine de l a mQme manibre qu'elle aide les orphelins dans l'enseignement secondaire dans le cadre du Fonds de Gknocide; cependant, l'intervention devrait probablement Qtre associke a des efforts d'amklioration des rksultats de l'apprentissage. D e tels efforts sont importants car, dans un systbme de sklection baske sur le mkrite, klargir la reprksentation des groupes a la traine n'est possible que si ces groupes sont capables de rivaliser pour les places convoitkes dans l'enseignement post-primaire. Conclusion 28. Presque une dkcennie aprbs le genocide, les dirigeants du Rwanda peuvent Qtre fiers et satisfaits de tout ce qu'ils ont purkaliser. 11s ont remis un systbme dkvastk sur pied: des salles de classe ont ktk rkparkes et de nouvelles ont ktk construites pour faire face au nombre croissant d'klbves; les enseignants qui ont hi la violence et sont rentrks ont ktk xxxii rkintkgrks dans l'enseignement; les arrikrks de salaire des enseignants ont 6tC rCgularisks; le Fonds du Gknocide a 6tC crC6 spkcifiquement pour aider les orphelins; et dans l'enseignement supkrieur, unsysteme diversifik a Ctk mis en place. 31. Cependant, la tiiche restante demeure impressionnante au moment oh la phase du redressement cede la place au travail de consolidation et de dkveloppement a plus long terme du secteur. Les soucis d'efficacitd, d'kquitk et de stabilitC fiscale inkvitablement deviendront de plus en plus rkels au moment ou le pays cherche a faireprogresser le systkme Cducatif dans un environnement de ressources limitkes. I1 faut espkrer que les rksultats de cette Ctude peuvent contribuer la discussion en crkant une comprkhension commune des problkmes et en attirant l'attention sur certains des dkfis kmergeants. Chapter 1: Country Context 1.1 The Rwandese government's recent poverty reduction strategy paper (PRSP) envisions a key role for education in its efforts to improve the population's social and economic well-being (Rwanda 2002a). The emphasis on education is consistent with a growing international consensus-encapsulated in the United Nation's 2000 Millennium Declaration and the accompanying operational definition o f the Millennium Development Goals (MDGs)-that progress in education cannot be left out of the fight against poverty. This confluence o f public opinion i s already leading to support for increased funding for education, both domestically and internationally. 'Yet, any increaseinresources i s contingent on the effective use o f existing resources. It also carries with it the implicit expectation that more and better results would follow, results that must be reflected in a tangible progress toward full coverage o f the primary school-age population, improved student learning, and a better fit between the volume and mix o f skills produced by the school system and the demandso fthe labor market. 1.2 The foregoing developments present exciting opportunities for educational development inRwanda, as well as pose daunting challenges. To set the stage for examining the nature o f some o f these opportunities and challenges, this chapter presents pertinent information on the overall country context. It documents the characteristics o fthe population, highlightinginparticular the prevalence o f orphanhood among the school-age children and youth. It also presents information on the country's macroeconomic conditions in historical and comparative perspective and on the overall situation relating to public finance. For readers unfamiliar with Rwanda, box 1.1 provides additional summary background informationon the country. Box 1.1 Rwanda at a Glance Geography: Known poetically as the "Land of a Thousand Hills," Rwanda i s a landlocked nation bordered to the north by Uganda, to the South by Burundi, to the east by Tanzania, and to the west by the Democratic Republic of the Congo. The country enjoys a mild tropical climate with two rainy seasons, February to April and November to January. With Uganda and the Democratic Republic of Congo, it is home to the only 650 mountain gorillas left inthe world today. Its surface area of 26,338 square kilometers (10,169 square miles) and about 290 inhabitants per square kilometer makethe country one of most densely settledinAfrica. The pressure of peopleon scarce land poses a constantthreat to social harmony andthe physical environment. People: Rwanda had a population of 8.16 million according to preliminary results from the 2002 census. The country's official languages are Kinyanvanda, French, and English, while local dialects include Igikiga, Bufumbwa, Rutwa, andHutu. The likelihood of increased external funding becoming available for this purpose is suggestedby the launching of the Educationfor All @FA) Fast Track Initiativeon June 12, 2002 by the World Bank on behalfof a new EFA Partnership that includedvarious bilateraldonors active in education, the EuropeanUnion, UNESCO, UNICEF, and the multilateral developmentbanks.The trendtowardincreaseddomestic spendingon educationis suggestedby the commitmentmadeby countries benefitingfrom the heavilyindebtedpoor countries (HIPC) to allocate 39 percent of the debt relief monies for investmentineducation(WorldBank2001). 2 Box 1.1Rwandaat aGlance (contd.) Economy: With a per capita GDP of only $2422in 2000, Rwandais one of the poorest countries inthe world today. Its agriculture-dominated economy has changed little over time, consisting mainly of small and increasingly fragmentedfams producing to meet subsistenceneeds. Coffee andtea continue to bethe country's principal exports. The manufacturing sector accounts for 20 percent of GDP and is dominated by firms supplying import substitutes for internal consumption. In the 1960s and 1970s, prudent financial policies, coupled with generous external aid and favorable terms of trade, fostered sustained growth. During the 198Os, when the price of coffee plummeted, growth was erratic. GDP declined sharply duringthe 4 years of civil war that culminated in the genocide of 1994. Since then, GDP has been growing steadily again 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. The first 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 beenthrough one episode of violence after another. To stem the widespread massacre of Tutsis during 1990to 1993, the Arusha Peace Agreement was signed on August 4, 1993, but the deathof PresidentHabyiarimana sparkeda genocide ofunprecedentedswiftnessthat, by one conservativeestimate, left up to 800,000 Tutsis and moderateHutus dead at the hands of Hutu militia. The RwandesePatriotic Army (RPA) launcheda counterstrikeandcapturedKigali, the capital city, on July 4, 1994. More than 2 million refugees fled to Tanzania and the Democratic Republic of Congo. 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 the Democratic Republic of Congo, more than 600,000 refugeesmoved 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 i s estimatedthat fewer than 100,000 Rwandese, thought to be remnants of the defeatedHutu militia, remain outsidethe country. Relationswith the World Bank: Rwanda became a member of the World Bank on September 30, 1963 and receivedits first World Bank credit in 1970for a roadshighway project. Sincethen, Rwandahasreceivedpolicy advice, technical assistance, andproject andprogram financing to support the country's economic development. The World Bank Group's International Development Association (IDA) and the International Monetary Fund (IMF)haveagreedto support acomprehensivedebt reduction packagefor Rwandaunder the EnhancedHeavily Indebted Poor Countries (HIPC) Initiative. This debt relief 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 NPV of Rwanda's debt outstandingat end-1999. Sources: Basedon the following online resourcesaccessedduringMay 2002: CentralIntelligence Agency, "The World Factbook Rwanda," ; Republic of Rwanda, "History ofA People," (online) ; SIL Intemational, "Languages ofRwanda," (online), http://www.ethnologue.com/show-country.asp?name=Rwanda; US Department o f State, Bureau o f Afiican Affairs, November 2001, "Background Note: Rwanda," ; World Bank, "Rwanda to ReceiveUS810 million in Debt ServiceRelief: The World Bank and IMF Support Debt Relief for Rwanda under the Enhanced HIPC Initiative," New Release No. 2001/192/S, ; share of population living in poverty for 1993, under-fivemortality rate and total fertility rate for 1978 and 1991, populationand share of school-agepopulationfor 2000, estimatedpopulationgrowth for 1991-2000,and projectedpopulationgrowth for 2000-20 based on United Nations PopulationDivision estimates and projections as reportedin the World Bank's SIMA database and in World Bank2002; total fertility rate for 2000 from Rwanda(Feb. 2000) DHSreport. 4 1.4 The incidence o f orphanhood inRwanda. As inmany countries affected by the HIV/AIDS epidemic, orphanhood i s a characteristic feature o f Rwanda's demographic makeup. In Rwanda, however, the problem has been greatly exacerbated by the 1994 genocide. What then i s the magnitude o f the problem? The relevant data appear intable 1.2, based on the 2000 UNICEF-sponsored Multi-Indicator Cluster Survey (MICS) that collected information onmore than 9,000 children below 15 years inthe sampled households. The data show that an average of 28.5 percent o f these childrenhadlost at least one parent, usually the father. Among 10- to 14-year-oldsY the share was much higher, exceeding 40 percent. Children who had lost both parents comprised 5 percent o f the sample, but the share among the older childrenwas again significantly higher. Table 1.2: Orphans and childrenlivingapartfrombiologicalparent(s), Rwanda, circa2000 %with oneparentdead -1 I %with %withone both or both %living Sample or Mother Father parents away from popula- dead dead Total 1 parents dead dead parent(s) tion size rIulti-IndicatorCluster Survey,2000" 0- to 3-year-olds 0.7 8.8 9.9 2.7 2,541 4- to 6-year-olds 3.1 18.7 29.58 i: 25.1 11.7 1,862 7- to 9-year-olds 4.0 24.1 28.1 34.4 14.3 1,633 10-to 12-year-olds 5.2 27.4 32.6 40.5 17.1 1,793 13-to 14-year-olds 5.8 28.6 34.4 10.3 44.7 20.4 1,251 Overall 0- to 14-year-olds 3.4 20.1 23.5 5.O 28.5 12.0 9,080 Cstimatedpopulationnationwide ~ I 1,000s) I :: 0- to 6-year-olds 28 208 235 1I 26 261 105 1,594 7- to 12-year-olds 59 332 482 202 1,284 13-to 14-vear-olds 27 134 391 162 210 96 471 a. Percentages are weightedby provincial sampling weights;. As the incidenceof orphanhoodamongboys andgirls is highly comparable, separate figures by gender are not shownhere. Source: Authors' calculationsare basedon data fkomthe 2000 MICS survey. 1.5 Cross-countrv comparisons on the incidence of orphanhood. How do Rwanda's orphanhoodrates compare with those inother countries? Figure 1.1 shows relevant cross-country data for around 2000. Among 7- to 14-year-oldsY the rate was estimated at nearly 40 percent in Rwanda, more than 1.5 times as high as the rates in Uganda and Zimbabwe; more than 2.5 times those inMozambique, South Africa, and Malawi; and nearly 4 times those in Tanzania and Kenya.3 As in these other African countries, the HIV/AIDS epidemic has created many orphans, but the much higher rate in Rwanda leaves no doubt as to the devastating legacy o f the 1994 genocide. Becauseestimatedorphanhoodrates insuch countries as Uganda and Mozambique already reach 20 percent or nearly so, it is unlikely that Rwanda's orphanhoodrate would be as low as the 21percent mentionedinthe previous footnote. 5 Figure 1.1: Adult HIV/AIDS prevalence rates and share of orphans among7- to 1Cyear-olds,Rwanda and other EastAfrican countries, circa2000 p !ze 30 20 10 0 a. Orphans here refer to children ages 7-14 who have lost at least one parent; data refer to 1992 for Malawi, 1996 for Tanzania, 1997 for Mozambique, 1998 for Kenya, South African and Zambia, 1999 for Zimbabwe, and 2000 for Rwanda and Uganda. Source: HIVlAIDS prevalence rates from UNAIDS at 30 5 B m .-C .c 2 0 Y a 20 20 ~ / 1 Source: see appendix table A2.1. 2.5 Inhigher education, enrollments have grown rapidly as well (see figure 2.3). From a base o f some 1,200 students in 1980-81 injust one public institution (Universiti Nationale du Rwanda) and a few private institutions (Institut Africain et Mauricien de Statistiques et d'Economie AppliquCe, Grand SCminaire de Nyakibanda, FacultC de ThCologie de ButarC, Grand SCminaire de Kabgayi, and Centre d'Enseignement SupCrieur de Kigali), the system's enrollment had by 2001-02 expanded to nearly 17,000 students spread across six public and nine private institutions." Enrollments grew at about 10.6 percent a year between 1980-81 and 1990-91, but growth in the post-genocide years quickened to an average rate o f about 29.0 percent a year between 1996-97 and 2001-02. Throughout the past three decades, the private sector has expanded faster than the public sector; this i s seen inthe increasing share o f enrollments inthe private sector. They went from 9.5 percent in 1980-81 to 22.6 percent by 1989-90 and rose even more in the post-genocide period, particularly in the past few years, reaching nearly 38 percent by 2001-02. "Thepublic institutionscurrently include UniversitPNationaleduRwanda,KigaliInstituteof ScienceandTechnology, Kigali Institute o f Education, Kigali Health Institute, Institut SupPrieur d 'Agronomie et d'Elevage, Institut Supkrieur des Finances Publiques. The private institutions include: UniversitP libre de Kigali, UniversitP Lai'que de Kigali, Institut SupPrieur de Pidagogie de Gitwe, Universitk Adventiste d 'ApiqueCentrale, Grand SPminaire de Nyakibanda, Faculti de ThPologie de ButarP, Grand SPminaire de Kabgayi, and Centre d'Enseignement SupPrieur de Kigali. Some private institutions that existed in earlier years have now closed or perhaps merged with other institutions, including the Institut Apicain et Mauricien de Statistiques et d 'Economie AppliquPe, Institut Supkrieur Catholique de PPdagogie Appliquie de Nkumba, and Ecole Supirieure de Gestion et d'lnformation. 16 Figure 2.3: Enrollmenttrends in highereducation,Rwanda, 1975-2002 20,000 40 17,500 35 15,000 30 5 8 0 --!8 8 12,500 25 2.2 E $! g 10,000 20 E `; *d 7,500 15 =I8E b 0 5,000 10 % G 8 2,500 5 0 0 W se 7 co 2 (D m 5 IC $ z .? 003 0 m ? zui 0) 8 0 cv Note: breaks in the series indicate years coinciding with "annees blanches" (i.e. when teaching was cancelled) or when data were unavailable. Source: see appendix table AZ.1. Trendsineducational coverage 2.6 To what do the expanding enrollments correspond in terms o f the education system's coverage o f the school-age population? Typical measures o f coverage are the gross and net enrollment ratios. The gross enrollment ratio (GER) i s defined as the ratio between all students enrolled ina given cycle o f education and the population inthe 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 o f schooling. In Rwanda, the age range i s 7-12 years for primary schooling, and 13-18 years for secondary schooling. For higher education, the standard practice is to define the age range as the 5-year bracket following the last year o f the range for secondary schooling (i.e., 19-23 years in the case of Rwanda). At this level o f education, the inconsistency between the ages of the enrolled and reference populations tends to be wider than inthe previous levels o f schooling, reflecting the cumulative effects o f late entry and grade repetition earlier in the system, as well as the likelihood that many secondary school graduates may work for a few years before enrolling inhigher education. Insteado f the gross enrollment ratio, some analysts thus prefer the number o f students per 100,000 o f the population as an alternative measure o f coverage in highereducation. The cross-country comparisonsbelow alsopresent data onthis statistic. 2.7 Comuarina coverape in 1991-92 and 2000-01. Below we shall focus mainly on the gross enrollment ratio to simplify the presentation and facilitate cross-country comparisons.'2 Table 2.1 shows estimates for 1991-92 and 2000-01 based on alternative Education systems where children begintheir schooling on time and few repeat a grade can expect to show that the gross and net enrollment ratios are practically the same. When the ratios do diverge, both late entry and grade repetitionplay a role, although it is impossible to ascertain their relative contribution merely based on the size of the divergence between 17 sources o f data for each o f the three levels o f ed~cation.'~ For primary schooling, the estimates for bothperiods are highlyconsistent, with a GER o f about 7 3 to 74 percent for the earlier period and between 107 and 108 percent for the later period. Note that, even though primary schooling lasted 8 years in 1991-92, the calculations here have been adjusted so that the GER refers to 6 years o f schooling, thus making them comparable to the calculations for 2000-01. According to the results, the education system is now reachinga larger share o f the school-age children than it didinthe early 1990s.Later sections o f 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. the two ratios. The further up the educational ladder, the more problematic is the net enrollment ratio as a measure o f coveragebecause o f the growing inconsistency inthe age range o f the numerator and denominator populations stemming from the cumulative effects o f late entry and grade repetition inearlier cycles o f schooling. If,for example, 20 percent o f the students ineach enteringclass in secondary school are 1 year older than the official entry age for this cycle (because they started their primary schooling late or repeateda grade) and there is no grade repetition in secondary school, the net enrollmentratio would suggest a level o fcoveragethat is some 20 percent lower than it really is, simply becausethe over- agedstudents havebeen excluded from the calculation. l3 Besides being a good practice, the use of alternative data sources to verify the estimates i s particularly important in a country such as Rwanda where the routine work of collecting school statistics has suffered severe disruption in recent years andis only nowbeing gradually restored. 18 Table 2.1: Gross enrollmentratios (GERs)by levelofeducation, Rwanda1991-92 and 2000-01 I Level ofeducationandtype of estimate 1991-92a 2000-0lb Survey-basedGERestimate(%)' DHS 73.9 106.6 _-_____------___________________________------. QUID .___________________ 108.3 Census-basedGER estimate (%) 72.5 107.3 No. ofpupils in grades 1-6(1.000~)~ 962.3 1,428.7 Population ages 7-12 (1,000s) 1,327.2 1,332.0 Survey-basedGER estimate(%) DHS 20.4 17.0 ............................................... QUID 10.8 Census-basedGERestimate(%) 21.8 12.3 No. of students in see. grades 1-6 (I,ooos)e 202.5 125.1 Population ages 13-18(1,000s) 928.I 1,014.3 ~~~~ Survey-basedGER estimate (%) QUID 1.3 _______-_---______----------------------------. Census-basedGER estimate(%) 1.9 Totalenrollments 12,757 Population ages 19-23 (1,000s) 660.6 Note: Blanks denote not available. a. The survey-basedestimate refers to 1992,whereas the census-basedestimatesrefers to 1991, the year ofthe populationcensus. b.Thesurvey-basedestimaterefersto 2001, whereasthe census-basedmethodrefersto 2000. c. See noteon the sourcefor the meaningofthe two acronyms. d. In1991-92 primary schooling lasted8 years, but only pupils ingrades 1-6 are includedhereto ensure comparabilitywith the GERestimate for 2000-01 when the cycle lastedonly 6 years. e. To enhancethe comparability o fthe data for 1991-92 and2000-01, the data for 1991refer to enrollmentsingeneral secondary, as well as those inthe 3-year post-primary vocationaltraining centers (ie.,CERAI, SF, andCERAR) andingrades 7 and 8 ofthe primary cycle. Source: The survey-based estimates of the GER rely on data collected in the 1992 and 2000 Rwanda Demographic and Health Surveys (DHS),andthose collectedinthe 2001 Quesfionnaire Unifibsur les Indicateurs de Diveloppement (QUID). The survey is alsoknownby its Anglophone acronym, CWIQ, which stands for the Core Welfare Indicators Questionnaire. The census-based estimates rely on enrollment data collected by the Ministry of Educationthrough its annual census of schools (see appendix table A2.1) and on population data from the 1991population censusandthe UnitedNation's population projections for 2000 as reportedinthe Bank's SIMA database. 2.8 For secondary education, the estimates show greater consistency across data sources for 1991-92 than they do for 2000-01. Taken as a whole, the general picture appears to suggest a probable decline in coverage over the period. This interpretation needs to be placed inthe context o f the structural changes inthe school system between the two dates. In 1991-92 the primary cycle lasted 8 years and significant numbers entered the 3-year post- primary vocational schools (i.e., Centres de 1'Enseignement Rurale et Artisanale Intkgrk Rurale et Artisanale deRwanda) counterpart^).'^ Computations o f the GER for 1990-9 1here (CERAI) and their girls-only (Section Familiale) and boys-only (Centres de 1'Enseignement include the numerator enrollments ingrades 7 and 8 of the primary cycle, as well as those in l4Enrollments in the 3-year vocational schools totaled between 25,000 and 26,000 in 199&91, compared with between 36,000 and 39,000 inthe general secondary schools (see appendix table A2.1). 19 the post-primary vocational schools. The two extra years o f primary schooling inthe 8-year cycle and the vocational courses are obviously not comparable to the tvonc commun 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 6 years o f the primary cycle. Inconclusion, although coverage may not yet be quite as extensive as inthe early 199Os, the current system is perhaps more equitable, in that it offers all children who make it to the second cycle a chance for continued progression up the educational ladder rather than foreclosing this prospect by channeling a large proportion o f primary school graduatesinto what hasbeeninfact an educational "cul-de-sac." 2.9 In higher education, there is one estimate of the gross enrollment ratio for 1991-92 and two for around 2000-01-1.3 percent according to survey results, and 1.9 percent according to enrollment and population data. One explanation why the survey-based estimate i s smaller i s that the data were collected just before the most recent explosion in enrollments, when they climbed from 10,058 in 1999-2000, to 12,757 the next year, and then to 16,668 the year after that. These data suggest that coverage o f higher education as measuredby the GER has increasednearly fourfold since 1990-91. 2.10 Cross-country comparisons. The relevant data for the comparisons appear in table 2.2 below." Included inthe table are the averages for low-income countries worldwide and inSub-SaharanAfkica, as well as the data for selectedAfrican countries with aper capita GNP in2000 ranging from $115 (Mozambique) to $495 (Mauritania). For higher education, the table includes enrollments per 100,000 o f the population as a complementary measure o f coverage. 2.11 Rwanda's GER for primary education is substantially higher than the averages for low-income countries inSub-SaharanAfrica and inthe whole world, whereas the ratio for secondary i s significantly lower and that o f higher education i s comparable to the corresponding averages.16 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 i s to be emphasized that this conclusion does not in itself offer a sufficient basis for policy development. For example, if the ratio for primary schooling i s high because o f widespread grade repetition, the education system may continue to suffer from poor rates o f 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 lower and upper secondary separately. Inhigher education, the comparability o f the country's gross enrollment ration with those o f other low-income countries does not necessarily imply that public policy has been successful. What is more important i s that the output o f graduates matches the labor market's capacity to absorb them into productive employment. The next section documents the pattern o f student flow, while chapter 8 examines the education-labor market issue. l5Despiteits flaws. the GERservesourpurposehere, as thedatafor this indicator arereadily availablefor alarge numbero f countries. l6Inhigher education, the data for 2001-02 suggest that the country has now more or less caught upwith the Sub-Saharan average for GER; the number of students per 100,000populationis also at the level that one would expect for a country at Rwanda's level of economic development (see figure 8.1 inchapter 8). 20 Table 2.2: Coverage of Rwanda's educationsystem in comparativeperspective,late 1990s-2000 1 'er capitaGNP, Gross enrollment ratios (%) Higher education 2000 enrollmentsper (in 1995 $) Primary Secondary Higher 100,000 of {he population ~ Lwanda, circa 2000-01a 241 107 13 1.9 191 ;electedAfrican 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 zountry group averagesb Low income Sub-Saharan countries 411 75 I 2o All low-income countries 456 84 33 5.2 575 lore: Blanks denote"not available." a. The ratios for primary andsecondaryeducationarethe averages ofthe various estimatesshown inthe previoustable. Forhigher education, it is the estimatebasedon enrollment counts andpopulation projections, as the former bestcaptures themagnitudesofthe recentincreases inenrollments.For the lastcolumn, the indicator is calculatedusingtotal higher educationenrollmentsfor 2001-02 andthe population of 8.7 millionprojectedby the United Nations Population Division, as reportedinthe World Bank`s SIMA database. b. Dataonthe gross enrollmentsratios referto the averagesfor 27,23, and 17 low-income Sub-Saharan countries, respectively, at the primary, secondary, andhigher levels. Forthe remaining two country groupings, the correspondingnumber ofobservations include 35,3 1, and23 Sub-Saharan countries, and42,37, and30 low-income countries.Dataon higher educationenrollmentsper 100,000 ofthe populationrefersto 27 Sub-Saharan countries and40 countriesworldwide with per capitaGNPbelow $755 in 1996. Source: For data on enrollment ratios, see the UNESCO web site at For data on the per capita GDP, see World Bank WorldDevelopmentIndicators. From cross-sectional indictors of coverage to student flow patterns 2.12 Because schooling usually requires a multiyear time commitment to yield tangible returns, it i s important to go beyond the aggregate indicators o f enrollments documented so far. Instead the schooling careers o f children need examination, relying on such measures as the entry rate to a given cycle o f schooling, share o f entrants who reach the end o f the cycle, frequency o f grade repetition, and transition rate between cycles of education. 2.13 At the primary levelwhere the curriculumis typically designed to impart basic literacy and numeracy skills, children who never enroll are unlikely to acquire these skills, whereas those who enroll, but drop out prematurely are unlikely to become permanently literate and numerate. In Rwanda, data from a household survey in 2001 show a clear correlation between adults' educational attainment and literacy (see figure 2.4): the probability o f being literate i s nearly 100 percent for those who have attained at least 5 years o f primary schooling, whereas it i s less than 50 percent for those attaining only 3 years. For any given level o f attainment, the probabilities for women are lower than men's; these results 21 suggest that it is particularly important for girls to survive to the end o f the primary cycle to achieve permanent literacy inadulthood. l7 Figure 2.4: Relationbetweeneducationalattainment and probability of beingliterate, Rwanda 2001 1.o 0.8 0.6 0.4 0.2 0.o 1 2 3 4 5 6 Grade attained Nofe: Refers to the subsample of population ages 15 and above. Source: Authors' estimates based on the 2001 Questionnaire Unifi sur les Zndicateurs de Diveloppement (QUID). 2.14 At post-primary levels o feducation, access and survival to the end ofthe cycle remain important, not least because both events are a prerequisite for entry to subsequent levels. The curricula likewise offer training on an integrated set of skills, so that premature exit would also imply a failed investment, to the extent that the departing student would not have mastered what was being taught. At all levels of education, therefore, but especially in primary education, rates of entry, survival, and transition are critically relevant when assessing the education system's coverage. 2.15 Grade repetition is the fourth student-flow indicator mentioned above. It is important for two reasons. The first is that it is likely to reduce the probability that a student persists to the end o f the cycle, inpart because grade repetition signals parents that their child is not progressing academically, an outcome that naturally would predispose them to withdrawing the child from school. The second reason is that, although a repeater costs twice as much per grade attained as a student who has not repeated the grade, no strong evidence exists that grade repetition necessarily improves learning outcomes. High rates o f grade repetition are, therefore, uneconomic, inthat 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. Studentflow patternsinprimary schooling 2.16 The overall pattern of entry rates to first grade, survival rates to subsequent grades inthe primary cycle, as well as the prevalence of repetition at each grade are examined l7Theresultsbasedonanothersurvey, the 2000 RwandaMultipleIndicator Cluster Survey (MICS), show asimilar 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 infigure 2.4. The results for Rwanda are comparable to those based on data from MICS inother Africancountries, including Niger, Togo, Guinea-Bissau, Sierra Leone, Senegal, and Burundi. 22 below." A subsequent section shall consider the transition between the primary and secondary cycles. 2.17 Rate o f entry to grade 1. Before presenting the results, a comment on the estimation methods i s in order. One method consists o f dividing the number o f new entrants to first grade by the cohort corresponding to the official age for entering this grade (i.e., 7 years inRwanda)." For bothyears, data on the numerator are available from school statistics. For the denominator, however, actual data are available only for the earlier year when the last population census was conducted. For 2000-01, the calculation would have had to rely on population projections. Because o f the large movements o f people after 1994, such projections, particularly for single age groups, are probably not sufficiently reliable for our purpose here no estimate o f entry rate for 2000-01 based on this method o f estimation i s shown here. 2.18 A second approach to estimating the entry rate can be used when survey data are available. It involves computing the share o f children who report never enrolling in school and subtracting the result from 100percent to arrive at the desired estimate.20The data needed to implement this approach are available for Rwanda for both 1991-92 and 2000-01; for the latter year, two sources are available, both o f which are used here. Although survey data represent only a sample o f the population, their advantage over school and census statistics is that both the numerator and denominator refer to the same individuals, thus assuring internal consistency inthe data. 2.19 The results appear intable 2.3. The estimates for both 1991-92 and 2000-01 are highly comparable across data sources, a feature that enhances our confidence in the estimates. The current entry rate-averaging 88 percent-is remarkably highfor a country at Rwanda's per capita GDP level. Still, it is important to note that the rate is, at best, only marginally better than the average o f about 86 percent at the start o f the 1990s. The results implythat gains inthe entry rate to first grade canprobably be ruled out as a major source of the large increase inthe GER from 73 percent 1991-92 to more than 100percent in2000-0 1. Attracting the last 12 percent o f children to enroll, thus, continues to be an important policy challenge inexpandingprimary school coverage inRwanda." '*Socioeconomic disparities intheseparameters will bepresentedina later chapter. Because some children may enter earlier or later than the official age, the calculation should ideally be adjusted by dividing the number o f entrants by the populationcorresponding to the average age o f the entering class. This adjustment is unlikely to alter the estimates substantially unless significant differences exist in the size o f the single-age cohorts around age 7. 2oThe calculation refers to children inthe 10 to 13 age bracket. Settingthe lower bound at 10years allows for the possibility that some never-enrolled children may still eventually enroll; setting 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 influenceo fbehavior patterns associatedwith older cohorts. 2'Chapter 4 examines the characteristics of the unschooled population as well as presents evidence on the correlates of ever attending school. 23 Table 2.3: Entry rate to grade 1, Rwanda 1991-92 and 2000-01 Sourceofdata 1991-92 2000-01 I I School statistics & 1991population census 88 Household surveys D H S 85 88 QUID 87 Note: Blanks denotenot available. a. For 1991-92, the estimate inthe first row is the number ofnew entrants reportedin the school statisticsdivided by the population of 7-year-olds enumeratedinthe census. Inthe restofthe table, which is basedonhouseholdsurvey data, the estimatescorrespondto the share of children ages 10-13 who report ever enrolling inschool; the age range was chosen to avoid underestimationassociated with possible late entry into first grade. Source: Authors' estimates based on data from MMEDUC's school statistics, which are collected through its annual census of schools, the population census of 1991, the 1992 and 2000 Demographic and Health Surveys (DHS), and the 2001 Questionnaire Un@Psur les IndicateursdeDPveIoppement (QULD). 2.20 Survival rates in the urimarv cycle. To continue with the timeframe used above, these rates are estimated for 1991-92 and 2001. As before, an explanation about estimation methods and data sources i s warranted at the outset. Survival rates should, strictly speaking, refer to the schooling careers o f true cohorts as the members make their way through the school system over time. In the best o f circumstances, such tracking is demanding in terms o f both data needs and ease o f calculation. In settings such as Rwanda, where the dynamic o f student flow hasbeen complicated by the reintegration of children who have re-enrolled in the years following the genocide, the data needs and computational algorithm are even more demanding. Beyond these problems, perhaps the more serious drawback is that the results relate to the experience o f children who entered school at least 6 years ago andmay therefore not be sufficiently up-to-date for policy development. 2.21 An alternative approach is to construct pseudocohort survival profiles based on cross-sectional data for two adjacent school years. The method involves computing grade- to-grade transition rates which are then linked together to obtain the pattern for the entire primary cycle. For example, if the transition rate between grades 1and 2 is 85 percent, and that between grades 3 and 4 i s 75 percent, the survival rate from grade 1to grade 3 would be computed as 64 percent ( = O M x 0.75). By extension to subsequent grades, the process eventually yields the survival rate to the final grade inthe cycle. The result is a pseudoprofile because it blends together the experience of the different cohorts currently in school. The stability of the profile thus depends on the extent to which cohorts differ in the pattern of grade-to-gradeprogression. 2.22 Pseudocohort survival rates can be computed using school statistics as well as survey data. We use both these sources for the calculations on Rwanda. With school statistics, the grade-to-grade transition rates are computed from data on nonrepeaters in adjacent grades inthe 2 years.22For example, ifinyear X there were 20,000 nonrepeatersin grade 1, and in year X+l there were 18,000 nonrepeaters in grade 2, the transition rate between grades 1 and 2 would be 90 percent (=18,000/20,000).23 With survey data, the 22Note that inorder to avoid double counting, only nonrepeatersare usedinthe calculation. 23The method assumes regularity in schooling behavior, that is, children enter the system only in grade 1 and make the transition between school years inone o f three categories: as repeaters, dropouts, or graduates to the next grade. Although this assumption is probably valid for the calculations for 1991-92, it is unlikely to hold for 2000-01, given the large 24 calculation is feasible iffor each child who attended school last year information exists on the following variables: (a) the highest grade attained by the child; and (b) whether or not he or she is still enrolled the current school year. These variables can be usedto compute the grade- specific dropout rates. By subtracting from 100 percent, we obtain what would be, in the absence o f differences in grade repetition across grades, the desired grade-to-grade survival rates. Where these differences are significant, as they are in Rwanda, an adjustment i s required to avoid double counting; this i s done by multiplyingthe rates obtained thus far by the ratio between the shares o f nonrepeaters inthe relevant pairs o f grades?4 Once computed, these adjusted grade-specific rates can be linkedtogether as before to obtain the pseudocohort survival rate to the end o f the cycle. 2.23 Applying the foregoing methods, we obtain the estimates for 1991-92 and 2001 that appear intable 2.4.25They show a significant improvement incohort survival over the course of a turbulent decade:, although 44 pupils in a cohort o f 100 reached grade 6 in 1991-92, 73 currently do-an astonishing gain o f 66 percent. The increase is indeed remarkable, given the context. If sustained, the improvement augurs well for the spread o f literacy inRwanda, given the findingpresented earlier that adults with at least a fourth grade education have a more than 80 percent chance o f being literate. Yet there i s concern that the gain in survival rate may not be sustained-a possibility that we examine below in greater detail inlight o fthe patterns o f grade repetition. Gradea 1991-92 2001b P1 100 100 P2 85 100 P3 74 96 P4 64 91 P5 53 77 P6 44 73 P7 35 - P8 29 - Nore: -denotes"not applicable." a. The primary cycle was shortenedfrom 8 to 6 years in 1992. b. Estimateshave been adjustedfor differences ingrade repetition betweencontiguous grades. See text for further explanation. Source: for 1991-92, authors' estimates based on data from MWEDUC's school statistics; for 2001, authors' estimates based on data from two household surveys, the 2001 Questionnaire Unijii sur les Indicateurs de Ddveloppement(QUD),andthe 2001RwandaMultiple Indicator Cluster Survey(MICS). 2.24 The information on entry rates and survival rates can be combined to obtain grade-specific enrollment rates in 1991-92 and 2000-01 (figure 2.5).26 These rates show the population movement following the genocide in 1994 and the reintegration of returning children whose schooling was interrupted by the war. z4 Suppose, for example, that the dropout rate between grades 1 and 2 is 5 percent and that the share ofnonrepeaters is 80 percent ingrade 1 and 75 percent ingrade 2. The survival rate betweengrades 1and 2, adjusted for differences ingrade in the two grades, would thenbe 89 percent (=(loo-5) x (75/80)). 7.sThe calculations for 1991-92 relied on school statistics only because survey data were unavailable for that year. Only survey data are used for 2001 for the reasons explained above. 26 To illustrate, the enrollment rate ingrade 3 wouldbe the entry rate multipliedby the survival rate fromgrade 1 to grade 3. 25 share of the school-age population enrolled at each grade, thus providing a finer grained picture of coverage than is possible with the gross enrollment ratio which is averaged over the entire cycle o f 6 years. On the current pattem o f entry and survival rates, we can expect that about 37 percent of the present generation of school-age children would not achieve a complete primary schooling: about a third o f these because of non-entry, about a fifth because o f dropping out between grades 1 and 4, and the rest (about 44 percent) because of dropping out after grade 4. These results suggest that attracting children to school and keeping them past fourth grade remain priorities in the effort to produce a future generation ofliterate and numerate Rwandese citizens. Figure 2.5: Grade-specificenrollment rates inprimaryschooling, Rwanda 1991-92 and 2000-01 100 80 60 40 20 0 1 2 3 4 5 6 7 8 Gradeb a. See text discussion for definition and computation method. b. The primary cycle lasted 8 years in 1991/2 and 6 years in 2000/1. Source: computed from the data in tables 2.3 and 2.4. 2.25 Grade repetition. Estimates for 1990 and 2000 appear in table 2.5. For 1990 the table shows only one set of estimates, based on MINEDUC's school statistics, the only available source for that year. For 2000 two sets of estimates appear in the table, based o n two independent sources-MINEDUC's school statistics and the 2001 Multiple Indicator Cluster Survey.27 The comparability between the two sets of estimates increases our confidence in the reliability of the underlying data?* The results reveal an astonishing threefold rise inrepetition rates between 1990 and 2000. Inboth these years, grade repetition occurs most frequently in the first grade, but its prevalence has risen noticeably in the upper grades, reaching 38 percent in grade 5, for example. Incontrast, the pattem in 1990 shows a declining share of the pupils repeating as they progressed up the educational ladder up until the last grade inthe 8-year cycle. *'Other householdsurveys implementedaround2000, suchas the 2000 QUIDandthe 2001 EICV, do not include questions that wouldpennit estimates of the repetitionrate. Even though the rates for 1990 cannot be verified against survey data, they are consistent with other indicators of coverage,includingthe gross enrollmentrate, entryrates to first grade, and survival rates to the end of the cycle, and are, therefore, fairly reliable for the purposehere. 26 Table 2.5: Repetitionratesinprimary schooling, Rwanda 1990 & 2000-01 I (Percent) 2000-0 1 Rate in 2000-01 as a Grade 1990 multiple of rate in School statistics Survey data (MICS) 1990b P1 16.2 41.5 40.4 2.6 P2 13.2 28.8 29.2 2.2 P3 11.4 30.3 27.9 2.7 P4 9.0 34.4 22.6 3.8 P5 8.2 37.6 31.1 4.6 P6 8.0 29.2 25.8 3.7 P7 6.8 - - - P8 9.7 - - - Cycle averagea Pl-P6 11.0 33.6 29.5 3.1 Pl-P8 10.3 Note: -denotes"not applicable." a. Figuresreferto unweightedaverages. b. To maintain consistency in data source, both the numerator and denominator refer to rates based on MINEDUC's school statistics. Source: For 1990, authors' estimates are based on MlNEDUC's school statistics for school years 1989-90 and 1990-91and for 2000, authors' estimates are based on MlNEDUC's school statistics for 1999-2000 and 2000-01 and on the 2001 Rwanda Multiple IndicatorCluster Survey(MICS). 2.26 The underlying changes that led to the significant rise in grade repetition are still unclear. What is clear is that the increase is one reason why Rwanda's gross enrollment ratio rose so quickly in the past decade-fiom 73 percent 1991-92 to 109 percent in 2000- 01.29There are, a priori, three sources for the increase in the gross enrollment ratio: an increase inthe entry rate to first grade, an increase inthe survival rate to the end of the cycle, and an increase in grade repetition. Between 1991-92 and 2000-01, the entry rate to first grade did not improve and can therefore be excluded as a factor behind the increase in educational coverage. Of the other two factors, grade repetition contributed an estimated 57 percent of the increase inthe gross enrollment ratio duringthe period, whereas the gain inthe survival rate to the end of the cycle-fiom 44 percent in 1991-92 to 73 percent in2000-01- contributed the remaining 43 percent3' Thus, although educational coverage had undeniably expanded rapidly over the course o f the 199Os, much of it reflects increased recycling within the system. 2.27 The high rates of grade repetition also raise concerns about the sustainability of the gains in cohort survival rates. Regression analysis of cross-country data suggests that 29Note that these ratios, computed by dividing enrollments by the relevant population, are reasonably consistent with those impliedby the student flow indicatorsreported so far. To be specific, the impliedgross enrollmentratio (IGER) is given by the following relationship: IGER= AGSER x (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 estimated to be 69 percent for 1991-92, and 104percent in2000-01. 30The calculationproceeds as follows, basedon the fact that during the period the entry rate to grade 1 stagnatedat around 88 percent, whereas the survival rate rose from 44 to 73 percent and the rate o f repetition rose from 11 to 34 percent. 84 percent (=(88 + 88x 0.73)/2) x (l+O.ll)). Supposethe repetition in2000 hadremained unchangedat 11percent, the impliedgross enrollment ratio would have been Thus, o f the 35 percentage point increase (=104-69) in the implied gross enrollment ratio between 1991-92 and 2000-01, 15 percentage points (=84 - 69), or 43 percent o f the total can be attributed to the increase incohort survival rate, and 20 percentage points (=lo4 - 84) or 57 percent o f the total to the increase ingrade repetition. 27 grade repetition indeedcorrelatesnegatively with cohort survival rates.31It suggeststhat for a country at Rwanda's level o f per capita GNP, a repetition rate of 33.6 percent would be consistent with a cohort survival rate o f around 47, controlling for the entry rate to grade 1 and the level of public spending per pupil in primary school. These results imply that, although Rwanda's cohort survival rate (73 percent) exceeds the regression-predicted level, the equilibriumrate may well be lower ifthe current highrates o f grade repetitionpersist. A decline would hardly be surprising, because parents' willingness to send and keep a child at school may wane inthe face o f signals conveyed by frequent grade repetition that the child i s infact failing to keep upwiththe schoolwork. 2.28 Summarv index o f student flow efficiencv. What do the foregoing patterns o f student flow imply for 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 against the resources that it would have spent in the absence of grade repetitionand dropping out. 2.29 The calculations and results appear in table 2.6. Consider, for example, the data for 1991-92. Given the pattern o f cohort survival and grade repetition in that year and calibrating the calculations to a starting cohort o f 1,000 pupils, the resources spent ingrade 1 would amount to 1,193 pupil-years (=1000/(1-0.162)). In grade 2, only 849 o f the initial 1,000 pupils remain, implying that the resources spent on them would be 978 pupil-years (=849/( 1-0.132). Continuing likewise for all six grades and adding up the pupil-years spent in each grade yield results in a cumulative total of 4,772 pupil-years (=1,193+978+839+704+582+477) for the whole cycle. This global amount compares against the total o f only 2,634 pupil years (=439x6) 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 o f a system where no one drops out or repeats a grade. Considering the wastage associated with dropping out alone implies that a total of 4,205 pupilyears (=1,000+849+743+640+534+439) were spentto produce the 439 graduates; thus the dropout-related efficiency index i s 0.63 (=2,634/4,205). Alternatively, counting only the wastage associatedwith grade repetition implies a total investment of 4,772 pupil-years, whenonly 4,205 would have sufficed; the repetition-related efficiency index is therefore 0.88 (=4,205/4,772).32 31Based on data for fifty-seven low-income countries, the relation between the cohort survival rate (CSR) and the repetition rate (RR), controlling for the entry rate to grade 1 (EGl), the per capita GNP inconstant 1993 dollars, andpublic spending per pupilas a multipleo f the per capita GNP (PSPP), is estimated as follows: CSR = 38.3 + 0.05 x EG1- 0.88 x RR + 5.53 x In (per capita GNP) + 41.8 x PSPP. The coefficient estimates for RR and In (per capita GNP) are statistically significant at the 1percent and 5 percent confidence levels respectively. 32Inpassing, we note that the data intable 2.6 canbe checked for their consistency withthe gross enrollment ratios (GER) reportedintable 2.1 which are estimated usingthe classical approacho fdividing total number of pupils by the population aged 7 to 12: 73 percent in 1991/92 and 107 percent in 2000/01. Consider first the data for 1990 in table 2.6. If all children enteredgrade 1, the implied GER wouldbe 89.5 percent (4,77216,000); but with a grade 1 entry rate o fbetween 85 and 88 percent (table 2.3), the ratio is infact around 69 percent. This calculation yields a result that is comparable to the actual ratio o f 73 percent for 1991/92. For 2000, the data in tables 2.3 and 2.6 imply a GER o f 120 percent (=(8154/6,000)*0.88), which is much higher than the actual ratio o f 107 percent. To evaluate the source of the discrepancy, consider anotherestimatebasedon only the survival rate to the terminal grade (Le. 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 o f information imply a GER of 102 percent (=((100+72.8)/2)*.88*1.34). The result is much close to the actual GERof 107percent, which suggestthat the grade-to-gradesurvival rates intable 2.6 areprobably over-estimated. 28 Table 2.6: Summary indicesof student flow efficiencyinprimary schooling, Rwanda1990& 2000 I 1990 II 2000 :manrareI Number left Icohort of I rupii-years from an initial Repetition rate Pupil-years 1,0001 (%I invested" cohort of 1,OOC ("/I invested" ~~ P1 1,000 16.2 1,192 1,000 41.5 1,709 P2 849 13.2 978 1000 28.8 1,404 P3 743 11.4 839 964 30.3 1,384 P4 640 9.0 704 910 34.4 1,387 P5 534 8.2 582 775 37.6 1,242 P6 439 8.0 477 I I I 728 29.2 1,028 Zumulative pupil-years 4,205 5,376 (2,634)b - 4,772 (4.366)b - 8,154 0.63 0.81 Repetition-relatedd 0.88 0.66 Overall' 0.55 0.54 t Note:-denotesnotapplicable. a. Refersto the pupil-yearsneededto educatenewentrants as wellas repeatersinthe correspondinggrade. See text for explanationof its computation. b. Refersto the pupil-yearsinvestedonly inthepupils who reachgrade6. c. The index is definedas the ratio between the cumulativepupil-year investedina system with no student-flow problems, and the actual cumulative pupil-yearsinvestedgiventhe current pattemofdroppingout andgraderepetition. See text for additionalexplanationonitscomputation. d. The numerator of the dropout-relatedindex includes only the student-years associated with droppingout; whereas the numerator of the repetition- relatedindexincludesonlythestudent-years associatedwith graderepetition. f. Theindextakes accountofthestudent yearsassociatedwithbothdroppingoutandgraderepetition Source: authors' estimatesbasedon the cohort survival andrepetitionratesreportedintables 2.4 and2.5. 2.30 Comparing the results for 1990 and 2000 suggests that the overall efficiency of student flow has not changed much over the decade. Consistent with the rise inthe cohort survival rate, the index taking into account only the impact of dropping out was better in 2000 than it was in 1990. Incontrast, the index associated with the impact of grade repetition alone has deteriorated, reflecting the rise in grade repetition over the decade. If grade repetitionpersisted at the current level and the cohort survival rate fell to, say, the regression- predicted level o f 47 percent, 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 inAfrica. Arresting this potential decline instudent flow thus warrants priority attention on the policy development agenda. 2.31 Cross-country comparisons. How does the pattern of student flow in Rwanda compare with that inother countries? Figure 2.6 shows data on entry rates to grade 1, survival rates to grade 5 as well as repetition rates for ten other African countries for which data around 2000 are available. Note that because the primary cycle insome countries lasts only 5 years, the figure shows survival rates for 5 years of schooling instead of the 6 years in the Rwandan system. The comparative data places Rwanda among countries with high entry rates to first grade as well as survival to grade 5. These positive features are marred, however, by the fact that Rwanda also has the highest rate of grade repetition. Insystems where grade repetition is widespread-as it is inMozambique, Madagascar, Cameroon, and Benin inthe sample-the cohort survival rate has typically been lower, falling for example to only 40 percent inMadagascar. To the extent that the pattem of student flow inthese systems reflects steady state conditions, the risk o f a decline in Rwanda's currently high rate o f cohort survival rate can neither be ruled out nor ignored. 29 Figure 2.6: Rwanda's primary school student flow indicatorsincomparative perspective,circa2000 . . . . . . . . . . . . . . . . . 1 Average,21 * . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mozambique, 30 . . . . . . . . . . . . . . . . . . . . 11I Mauritania, 17 Tanzania,4 88 I. . . . . . . . . . . . . . . . . . . . Madagascar,31 Rwanda,34 7 1. ....... .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Cameroon,27 . . . . . . . . . . . . . . . . . . .1] Benin,26 p v Togo,24 -y ...................1.I Guinea, 27 Burkina Faso, 15 I , , 1. . . . . . . . . I , , Niger, 13 I 100 80 60 40 20 0 0 20 40 60 80 100 Grade 1entry rate ("A) %of first graders reaching grade 5 *Refers to the average primary school repetition rate (in percent) for the countries shown in the graph except Rwanda. The country-specific repetition rate is shownbythe numberfollowing each country name. Source: Tables 2.3 & 2.4 for Rwanda; UNESCO2001 for data on survival rates for the other countries and various World Bank education country status reports for entry andrepetitionrates for the other countries. 2.32 To complete the discussion, we compute the summary index of student flow for the ten Afiican countries, using the same method as inthe previous section.33The results, plotted against the countries' per capita GDP in 1999, appear in figure 2.7. Given Rwanda's current student-flow indicators, the country is located inthe middle of the pack. Although it does better than such countries as Madagascar and Mozambique, there remains significant room for improvement. 33To ensure comparability across countries, the indices refer to 5 years o f primary schooling irrespective of the duration of theprimary cycle ineachcountry. 30 Figure 2.7: Efficiencyof student flow inRwanda incomparativeperspective, circa 2000 Tanzania BurkinaFaso Mauritania Guinea Rwanda Benin Togo Cameroon Mozambique .4 - Madagascar Per capita GDP in 1999 (log scale) * The indexhas a value o f 1.0 in a systemwith no droppingout or graderepetition. Seetext for detailson its computation. Source: authors' calculationsbasedon datain figure 2.6. Student flow patterns insecondary schooling 2.33 We turn now to examining children's schooling careers in secondary education, including the pattern o f transition between the primary and secondary cycles and survival andrepetition rateswithin secondary schooling. 2.34 Intercycle rates o f transition and cohort survival rates. As inmany education systems, formal selection for secondary schooling in Rwanda i s based on a student's performance in the national examination at the end o f the primary cycle. At present, an estimated 60 percent o f primary school pupils who reach grade 6 eventually make the transition to secondary schooling (whether public or private), even though some of them did so only after a few attempts (table 2.7).34 This rate o f transition exceeds the average o f 50 percent for Francophone Africa inthe 1990s and is comparable to the average o f 60 percent inAnglophone 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 2001). 2.35 At the secondary level, selection takes place at the end of the tronc commun cycle. As the table suggests, nearly all the students who currently reach the end of that cycle continue on to the upper secondary cycle. For comparison, in Burkina Faso and Madagascar-countries at similar levels o f per capita GDP and whose rates o f transition between primary and secondarycycles are also comparable to Rwanda's-the share o f lower secondary school completers who proceedto the upper secondary cycle is, respectively, only 60 and46 percent. 34Although data from the 2000 MICS can also be used to estimate the transition rate, small cell sizes make the results unreliable. We therefore report hereonlythe results basedon MINEDUC's school statistics. 31 Table 2.7: Secondary schooltransition and survival rates, Rwanda 1991-92 and 2000-01 Indicator 1991-92 2000-01 Transition rates between cycles of schooling (%)a Fromprimaryto tronc commun cycle 60 Fromtronccommunto the upper sec cycle 100 % ofentrants reaching end of each subcycle? Tronc commun 69 Upper secondary 79 S1to S6 60 57 Note: Blanks denote not available; -denotes not applicable. a. Refers to the transition rate betweenthe last grade of one cycle to the first grade in the next cycle. The rates were not computedfor 1991-92 becauseofdataconstraints associatedwith structural changes inthe system. b. In 1991-92, the secondary cycle lasted6 years with no selectionduring the cycle. The system was restructuredinthe post-genocide era into two subcycles each lasting 3 years; students follow a common first cycle (rronc commun) before proceedingto the specialized streams inthe upper secondary cycle. Source: Basedon MINEDUC'sschool statisticsfor school years 1991-92,1999-2000, and2000-01. 2.36 The foregoing table also documents survival rates in secondary schooling. In the tronc commun cycle, 69 percent o f the entrants reachthe end o f the cycle, whereas inthe upper secondary cycle, the share i s 79 percent. O f those who start secondary school, we can thus expect 57 percentto complete the entire 6 years o fthe secondary cycle. The rate today i s therefore largely unchangedfrom the rate inthe early 1990s.35 2.37 Repetition rates. The average rate for all six grades o f secondary schooling rose from 7.7 to 11.1 percent in the course o f the 1990s (table 2.8). The pattern o f grade repetition has also changed somewhat: although almost no student repeated the first year o f secondary education in 1990-91, the repetition rate is now 17.4 percent in this grade, the highest inthe two cycles combined. In2000-01, grade repetition was as high inthe public as inthe private sector and the pattern across grades was comparable between the two sectors. Consistent with the lack o f heavy selection between the tronc commun and upper secondary cycles (and probably also into higher education), grade repetition inthe last grade o f each o f the cycles 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 o f lower secondary school was three times that inthe other gradesinthe cycle. 35Ineducationsystemswithstableandslowlychangingstudent-flowprocesses,thestudentflowprofiles fortheprimary and secondary cycles can be combined to yield a complete profile covering all 12 years in the two cycles. In Rwanda, however, these conditions do not hold, so the method cannot be applied. 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. 32 Table 2.8: Repetitionratesinsecondary schooling, Rwanda 1990-91 and 2000-01 2000-01 ?ublic schoolsa Private school: Repeaters as % of students enrolled s1 2.8 17.4 15.9 19.0 s2 10.4 13.8 15.0 12.1 s3 10.2 8.5 8.3 8.8 s4 8.5 11.5 10.6 13.0 s5 9.5 11.0 12.5 9.2 S6 4.6 4.7 3.6 6.2 All grades 7.5 12.4 12.2 12.8 Unweighted averages Tronccommun cycle 13.1 13.3 Upper secondary cycle 8.9 9.4 Overall S1-S6 7.7 11.1 11.0 11.4 Note:-denotes "not applicable." a. Includesstate andlibre subsidie`schools. Source: MINEDUC's schoolstatisticsfor schoolyears 1990-91 and2000-01 2.38 Summary indices o f student flow efficiencv. To complete the discussion on student flow, table 2.9 below presents the indices o f student flow efficiency that are computed using the same method that was applied to the data for primary schooling. Recall that the more efficient the flow o f students through the system, the closer i s the index to unity. The results show that the flow was somewhat more efficient in200041 than it was 1990-91, reflecting the gains insurvival rates to the end o f the cycle. Table 2.9: Summary index of student flow efficiency insecondary schooling, Rwanda, 1990-91 & 2000-01 Cycle 1990-91a 2000-01 Tronc commun Upper secondary S1to S6 0.74 0.77 Note: -denotes not applicable. a. In 1990-91 the secondary cycle lasted 6 years with no selection during the cycle, so no separate index was computed for the tronc commun and upper secondary cycles. Note that an education system with no dropping out and no grade repetition would have an index of 1.0. See table 2.6 and relatedtext for details on how the index is calculated. Source: Based on MINEDUC's school statisticsfor school years 1991-91, 1999-2000, and 2000-01, andthe 2001 Questionnaire Unijipsur les Indicateurs de De`veloppement(QUID). Policy perspectiveson student flow management 2.39 In the context of the government's poverty reduction strategy (PRS), the student flow patterns documented above have at least two important implications for education sector policies. The first is to ensure that all children attain at least a complete primary education, so that all are equipped with the basic skills of literary and numeracy to make a decent living in adulthood. Beyond primary schooling, the production o f skilled 33 workers to support economic growth also warrants attention. Inthe years following the 1994 genocide, the priority was to re-establish the depleted human capital stock as rapidly as possible. As the stock recovers to its pre-genocide levels, expansion will increasingly depend on the labor market's capacity to absorb highly educated workers in productive em~loyment.~~Although a full treatment o f these issues, particularly those pertaining to the labor market, is beyond the scope o f this chapter, the student flow data presented in the foregoing sections nonetheless provide a basis for distilling some implications for managing the education system's quantitative expansion. 2.40 In urimaw education, entry rates have been historically high at nearly 90 percent, so the issue would be to identify the last 10 ercent or so o f the population that are still excluded and to target interventions a~cordingly.'~Survival rates to the end o f the cycle warrant even closer attention. As argued earlier inthe chapter, the current rate is respectably high (73 percent) compared with the rates in other low-income countries, but is probably unsustainable inlight o f the extremely highrates o f grade repetition. An immediate concem therefore i s to lower the repetition rate. A mediumterm target of, say, 10 percent at most is feasible based on other countries' experience, but would probably require measures to rationalize classpromotionpolicies as well as to improveclassroom pedagogical practices. 2.41 One possible approach begins by defining two or three subcycles within primary schooling, and applying automatic promotionwithin each subcycle andperformance- basedpromotion between subcycles. The arrangementreceives support from the idea that the curriculum covers skills that overlap between contiguous grades, particularly inthe beginning years when pupils are simply getting used to school and leaming the rudiments o f reading, writing, and arithmetic. As such, automatic promotion would avoid unnecessary compartmentalization o f the leaming process and create space for children to acquire the foundational skills at their own pace. A complementary-and arguably indispensable-tool for managing grade repetition is to equip teachers with the techniques for formative evaluation, along with the relevant professional support from head teachers, inspectors, and others, so that pupils' leaming gaps canbe systematically identified and remediedbefore they come up for evaluation for promotionto the next s~bcycle.~* 2.42 Grade repetition aside, supply-side constraints could also impede progress toward achieving universal survival to the end o f the primary cycle. One such constraint is the possibility that some schools may lack the staff or facilities to offer the complete cycle o f schooling. In such settings, pupils leave school prematurely not because they want to, but because they have nowhere else to go inthe system. Because supply-side constraints are by definition amenable to policy intervention, it i s useful to assess the scale o f the problem and the impact that removingthem mighthave on survival rates. 2.43 Table 2.10 contains some relevant information in this regard. It shows the distribution of schools and new first graders according to the highest grade o f instruction in which pupils are currently enrolled. Schools without pupils enrolled inthe sixth grade make up about 13 percent of all schools and account for about 8 percent of all first graders inthe system. These schools fall into two categories: either they are new and therefore still inthe process o f completing the grades offered, or they are constrained by lack o f facilities or 36Chapter 8 will further explore the education-labor market link. 37Chapter4 will addressingreater detail the patterno f disparities inschooling. As part o f 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 to manage grade repetition are among the interventionsthey are putting inplace to achieve that objective. 34 teachers from offering the complete cycle. The available data do not allow us to distinguish between these types o f schools, but for the sake o f illustration, suppose that all o f them are incomplete because o f supply-side constraints and that transfers from incomplete to complete schools occur infrequently. Under these assumptions, the impact o f ensuring that all schools offered the full primary cycle o f instruction would be to raise the cohort survival rate from its current level of 73 percent to 80 percent (= 7341- 0.083). Under more realistic assumptions, the impact would be smaller. InRwanda, therefore, the incompleteness o f instruction offered could probably be ruled out as a major reason why primary school pupils quit school before finishingthe cycle. Table 2.10: Distributionofprimaryschools and new firstgradersby highestgradeofinstructionoffered by the school, Rwanda2000-01 Highestgradeof Percentageof schools Percentageofnew first graders instructionoffered 1 Libre Cumulative State 1 subsidie` Overall' Overallb share 0.4 0.8 0.7 0.1 0.4 0.3 0.3 2.0 1.4 1.6 0.7 0.7 0.7 1.o 2.9 2.3 2.6 1.5 1.2 1.4 2.4 3.1 2.6 2.8 2.8 1.5 1.8 4.2 8.3 5.0 6.0 5.4 3.3 4.1 8.3 83.3 87.9 86.3 89.5 I 92.9 I 91.7 100.0 All schoolsa 100.0 100.0 100.0 100.0 100.0 100.0 - (551) (1,446) (2,028) (89,134) (229,353) (320,952) Note: -denotes "not applicable." a. Figuresinparentheses refer to the numberofschoolsor pupils on which the percentageshavebeencomputed. b. Includesdata for private schools. Source: MINEDUC's 1999-2000 census ofprimaryschools. 2.44 Beyond primarv schooling, the immediate problem in both secondary and higher education is to manage the mounting pressure for expansion as ever-larger cohorts o f students inthe preceding cycles complete their schooling. One helpfulway to think about the problem is to view the tronc commun cycle as a continuation o f primary education and the upper secondary cycle as preparation for higher education. This view resonates with 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 off their preparation for adult life. Itexplains why governments insome low-income countries have set their sights on achieving universal basic education (defined as the primary and lower secondary cycles) as a long-term social objective. Beyond the tronc commun cycle, economic considerations become more important inpolicy development regarding student flow. Although it is true that a modernizing economy requires highly trained workers to function well, experience also shows that producing such workers faster that the labor market can employ them inrelevant jobs typically does not accelerate economic growth, but instead creates educated unemployment or underemployment and, therefore, social frustration and unrest among the affected individuals. 2.45 What do these concepts imply for the management o f student flow beyond primary education in Rwanda? Without entering into a full-blown evaluation here that takes into account the fiscal implications as well, consider below some simple projections o f student numbers to help clarify one aspect o f the problem. Table 2.11 shows the current 35 distribution of enrollments by grade in the primary and secondary cycles. The structure of enrollments is very steep in that the number o f students in the lower grades is substantially greater than those in the upper grades; the pressure for expansion is thus likely to mount rapidly in the coming years.39 In the primary cycle, recall that the cohort survival rate is currently estimated at 73 percent. If this rate were maintained, the number o f new students entering grade 6 in2006 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 projected 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 inthe tronc commun cycle. Table 2.11:Grade-specificenrollments inprimaryand secondaryeducation, Rwanda2000-01 Primary cycle Secondary cyclea Year in cycle Total 494,614 324,804 238,897 189,865 123,549 19,608 17,349 145,478 88,938 17,634 15,695 81,914 58,923 13,668 13,019 Total 1,475,572 942,993 141,163 123,606 a. Consistsof two subcycles-tronc commun and upper secondaryschool-each lasting 3 years. Source: MINEDUC's school statistics for 2000-01. 2.46 The number of sixth graders in2006 who would be able to continue on to the tronc commun cycle depends on how fast it would be feasible (both physically and fiscally) to expand places inthe secondary cycle. Between 1995 and 2000, the entire system, including both the public and private sectors, 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 remain unchanged, 48,000 new entrants to the first year of the tronc commun cycle could be accommodated in2007.40Inthat 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 o f secondary school places-to 6,000 places per grade annually-the transition rate between the primary and tronc commun cycles would still be lower than at present: 31 percent if the survival rate inthe primary cycle were 73 percent and 46 percent ifit were only 50 percent. 39Note that the current size o f the entering cohort infirst grade is o fthe same order o f magnitude as the projectedpopulation of 6-year-olds inthe coming decades, implyingthat the scaleo fthe problem is likely to persist into the future. 40This assumes that repetition rates in the first year o f the cycle remain at 17.4 percent as at present. The number o fplaces for new entrants to S1 would thus be the sum o f the current number o f new places (=32,929) and the number o f new places createdat the assumedrate (=3000 x (1.O - 0.174) x 6), for a total o f47,797 places. 36 2.47 Although admittedly rough, the foregoing projections suggest that in the foreseeable future, the share o fprimary school completers continuing on to secondary school would very likely decline even as the absolute size o f enrollments insecondaryschools grows ~ubstantially.~~Correspondingly, the selection process would need to become even more efficient than it i s now inidentifylngstudents with good potential to benefit from continuing past the primary cycle. 2.48 The data inthe foregoing table also carry implications for the development of upper secondary education. In2000-01,the system admitted nearly 32,000 new entrants to the first grade inthe secondary cycle. Ifthe survival rate inthe tronc commun cycle remains unchanged at 69 percent, the projected number o f candidates for upper secondary schooling would exceed 22,000 in 2003-04, or about 1.27 times the size o f the cohort that entered upper secondary cycle in 2000-01. Currently, almost all students who finish the tronc commun cycle eventually go on to upper secondary schooling and 79 percent reach the end o f the cycle. Continuation o f these patterns implies that by 2007, a total o f 18,000 upper secondary graduates a year would be seeking either employment or entry to higher education, about 40 percent larger than the current cohort size o f some 13,000. 2.49 We can expect these projections to increaseinmagnitude inhture years as the number o f places in lower secondary schools is expanded to accommodate the rising tide o f primary school completers. For example, suppose new entrants to lower secondary school grew to 48,000 students a year by the end of the decade, on the assumption that places were expanded at about 3,000 places annually. If the student flow patterns within secondary schooling remained as at presentwith limitedselection between the tronc commun and upper secondary cycles, there would be at least 27,000 upper secondary school leavers annually by 2015 seeking either a job or a place inhigher education. At issue then is whether an annual output o f the projected magnitude would be consistent with the absorptive capacity o f the labor market for workers with at least an upper secondary education. Although it is perilous to project labor market developments in Rwanda's evolving post-genocide context, the scale of the increase in graduate output is such as to warrant a closer look at the options for managing student flow. Strengthening mechanisms to tighten the selection into upper secondary schooling, particularly in the publicly funded sector, would seem to be a logical place to start. Conclusion 2.50 This chapter has docurhented the coverage o f the education system and its evolution over time. The absolute numbero f students at all levels o f education has recovered from the severe disruption causedby the genocide and now exceeds the pre-genocide levels. The private sector share o f enrollments has remainedunchanged duringthe 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 i s now significantly greater than it was at the start o f the 1990s and i s now comparableto the average for other low-income Sub-SaharanAfrican countries. The ratio for higher education has also tripled since the early 1 9 9 0 ~although it remains smaller than the average for other ~ comparable countries. In secondary education, where the trend i s difficult to establish 41This decline obviously does not invalidate the long-term goal of achieving universal basic education (i.e., primary and lower secondary schooling). Physical constraints alone (quite apart from fiscal ones as well) make that goal infeasible in the immediatefuture. 37 because of structural changes in the system during the 199Os, the ratio in 2000-01 was likewise somewhat lower than the average for low-income countries inSub-Saharan Africa. 2.51 With regard to policy development, the detailed look in this chapter at the pattern o f student flow has also clarified some o f the key policy challenges pertaining to the system's quantitative expansion. Nearly 90 percent o f Rwandese children currently start school, about the same share as inthe early 1 9 9 0 and ~ ~ 73 percent eventually make it to grade 6, up from about 44 percent 10years ago. There i s obviously still scope for improvement on both counts. But even maintaining the cohort survival rate at 73 percent would pose a significant challenge, given the extremely highrates o f grade repetition that characterize the system. At nearly 34 percent, Rwanda's repetition rate in primary education i s one o f the highest in the world, and cross-country experience suggests that its persistence i s likely to depress the survival rate, causing it to fall to perhaps as low as 50 percent. Reducing grade repetition would therefore appear to warrant priority attention as part o f the country's strategy to achieve universal primary education. 2.52 As primary schooling improves and expands, the pressure to increase the supply o fplaces at the post-primary levels will inevitably mount. Simple projections based on conservative assumptions suggest that the potential candidates for entry into lower secondary school at mid-decade could easily number four times the size o f the current cohort o f entrants to that cycle. Similarly, by mid-decade the number o f potential candidates for entry into upper secondary school could be at least 30 percent larger than the current number o f entrants to that cycle. By the middle of the next decade, the number o f upper secondary school graduates seeking either employment or entry to higher education-under current patterns o f student flow in the secondary cycle-could easily be twice the current output from the system. 2.53 Inlight ofthese projections, the task ofmanaging student flow throughout the secondary cycle presents two challenges inthe next 5 to 15 years. The first i s to maintainthe current relatively high levels o f within-cycle survival rates in both the tronc commun and upper secondary cycles. The second challenge is to strengthen the existing selection mechanisms-between the primary and tronc commun cycles and betweenthe tronc commun and upper secondary cycles. To the extent that the pace o f expansion o f 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 increasingpool o f 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 can only become more urgent if the labor market develops signs o f a weakening capacity to absorb highly educated workers into productive employment. 39 Chapter 3: EducationFinance 3.1 In Rwanda aggregate public spending on education has been rising steadily since 1996. The prospects for continued expansion are promising, given the new climate o f support for human development, both internationally and domestically. Yet, even more important than the aggregate amount o fresources i s the way they are used. Inthis chapter, we document: the volume o f national spending on education, including spending by households; the composition o f public spending on education by level; and international perspectives on the pattern o f spending in Rwanda. The results provide a factual basis for discussing future directions for managing resource allocation to ensure that key sectoral priorities are adequately funded.Key amongthese is to enableall children to complete a full courseofprimary schooling under conditions that promote student learning. National spending on education 3.2 As in most other countries, the government is the main source of funding for education in Rwanda. Contributions by households are also significant, however, particularly in secondary education where privately financed schools currently cater to about 43 percent o f enrollments at this level. 3.3 Spendinn by the government. Total spending on education reached a highof 5.5 percent ofGDP in2001, butas indicatedinanearlier chapter, muchofthe increase in the post-genocide years reflects rapid increases in capital spending (table 3.1). Nonetheless, aggregate recurrent spending in 2001 relative to GDP was already 65 percent higher than it was in 1996. Across levels o f education, the increase has mostly benefited higher education whose share o f spending rose from just under 15 percent in 1996 to more than 37 percent by 2001. Secondary education also gained, but its allocation of recurrent spending rose only modestly, from about 15 percent in 1996 to nearly 18 percent in2001. The combined result o f these trends inhigher and secondary education i s a dramatic fall in the share o f primary education, from 70 percent to just more than 45 percent in the period. Despite the large increase since 1996 in aggregate recurrent spending on education relative to GDP, spending on primary education thus registered a tepid increase, reaching only 1.5 percent o f GDP by 2001. 3.4 The concentration of spending on higher education after 1996 is consistent with the government's efforts to rebuild the country's depleted stock of highly qualified citizens following the 1994 genocide. Yet, if universalizing primary schooling i s an objective, the current bias in the allocation o f recurrent spending appears lopsided. Low-income countries that have managed to universalize primary school completion typically allocate a 50 percent share o f public spending on education to the primary level (Mingat, Rakotomalala, and Tan 2002). Admittedly, increased spending alone would not be sufficient to produce results: figure 3.1 suggests a generally positive relation across countries between spending on primary education and grade six completion rates, but the relation i s loose, and Rwanda's position in the graph points to substantial scope for improvement even at its current level o f spending. Even though a larger allocation for primary schooling is a legitimate objective, there is, thus, also a need to improve the system's current 40 performance. Subsequent chapters o f the report elaborate on some possible options in this regard. Table 3.1: Level and distributionofpublic spending on education,Rwanda, 1982-2001 Share of current spendin, ~ylevel of Zurrent spendini education (%. , on primary Primaryb Secondary Higher education as % of GDP 62.7 24.0 13.2 1.9 12.9 12.6 59.5 27.2 13.3 1.7 60.0 27.6 12.4 1.9 61.1 26.8 12.1 2.0 59.0 25.5 15.4 1.8 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 (ore: Blanks denote datanot available. I.Some categoriesofspendingare sharedacross levelsofeducation(e.g., those associatedwith theNationalCenter for Curriculum Development,the National Examinations Council, and central ministry staff and running costs, and the inspectorate). For our purpose here, such spending is apportionedaccordingto the subsector's share of the spendingthat was possibleto allocateacross levels. b. Published data for the 1980s show combined spending on primary education (which lasted 8 years during this period) and the Centres de 1'Enseignement Rural et Artisanal Integri- (CERAIs), which were 3-year vocational institutions for primary school leavers; these were discontinued after 1991-92). To obtain atime series for spendingon primaryeducationcorrespondingto a 6-year cycle (which is how longthe cycle now lasts) we prorated the published data according to the share of enrollments in the last 2 years of the 8-year cycle during the 1980s and enrollments in the CERAIs (see appendix table A2.1 for the dataonenrollments). Source: See appendix table A3.1. 41 Figure3.1: Relationbetweenprimary school completionratesandpublic spending on primary education,low-incomecountries,circa2000 100 1 40 D O 20 I 0 1I 2I 3I 4I Public spending on primary education as X of GDP a. The completion rate i s computed as the number o f nonrepeatcrs in grade 6 relative to the corresponding age cohort for that grade. Source: Bared on data for 47 law-income countries reported in Mingat, Rakotomalala, and Tan (2002). 3.5 Inthe foregoing discussion, we have focused on public spending on education managed by the Ministryo f Education. The ministry is indeed responsible for the bulk o f such expenditure, but in recent years, a nonnegligible volume o f education-related spending also flows through other ministries and government agencies, notably to provide vocational and nonformal training and financial assistance to mostly secondary students orphaned by the genocide (table 3.2). Including these items o f spending raises recurrent spending on education in Rwanda to 3.8 percent o f GDP in2001, with 87.2 percent accountedfor by spending through the Ministry of Education?2 This level o f aggregate recurrent spending on education i s comparableto the average o f about 4.0 percent inlow-income countries. 42Note that because the additional spending outside the ministry concerns post-primary services, the results in figure 3.1 remainunchanged. 42 1999 2000 2001 Ministryof Educationa 21,738 22,347 25,341 Other ministriesb Educationandtraining services' 347 287 324 Financial assistanceto students 513 3,205 3,400 Total education-related recurrent spending 22,598 25,839 29,065 ASYoof GDP 3.5 3.7 3.8 %spent via the Ministry ofEducation 96.2 86.5 87.2 %spent via other ministries/govemment agencies 3.8 13.5 12.8 a. Excludesspendingonthe universityhospital. b. Includes the Ministries of Finance; Youth, Sports, andCulture; Gender; andLocalAdministration and SocialAffairs as well as the Genocide Fund(whichbecameoperationalin2000whenthe first bursarieswere awarded). c. Includesspendingonthe Institut Supkrieurdes Finances Publiques,andonvarious vocationalandnonformaltrainingcourses. Source: See appendixtable A3.2. 3.6 Spending bv households. Rwandese households complement the government's investment in education to a large measure. Data from the 1999-2000 Household Living Condition Survey (also known as the Enqugte Intkgrale sur les Conditions de Vie des Mknages au Rwanda or EICV), a nationwide survey of some 6,420 households in urban and rural areas, suggest that almost all students- irrespective o f level of schooling and whether they attend public or private schools- incur some education-related spending.43 The surprising pattern is that the share of the respondents with nonzero spending is smaller inhigher education than it is inprimary or secondary school. Inabsolute terms, however, families spend more as the level of education rises, from $5 annually per child in public primary schools to $318 in higher education. Summing across all levels and types of education, household spending on education in 2000 amounted to about 1.5 percent o f GDP or about 41 percent as much as public recurrent spending on education inthat year. This level of private spending is relatively highand comparable to the level inMadagascar, another country where the private sector also enrolls just under half of the enrollments at the secondary level. 3.7 The composition of spending across levels is also interesting: school fees capture a sizable chunk of spending by households throughout the system, but it is inprivate primary schools and at the post-primary levels that they exceed at least half o f the total. Inpublic primary schools, fees account for 23 percent of the total cost to the families, whereas uniforms are the single most important cost item (accounting for about 45 percent o f the total), followed by books and school supplies (accounting for 24 percent). The relatively large share of spending on uniforms is not unusual, but it does raise questions 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 (e.g., Tanzania and Uganda), fees for primary schooling have been abolished and a system of capitation grants instituted to relieve families of the burden o f paying for books and supplies. The expectation that these moves would 43The surveywas conducted between October 1999 andDecember 2000 inurban areas andbetween July 2000 and July 2001inrural areas. 43 speed progress toward achieving "Education for All" i s borne out by the rapid increase in enrollments in both countries in recent years. Closer examination o f the evidence shows, however, that, although there i s now nearly universal entry to grade 1, pupils continue to drop out before finishing the cycle. The implication is that additional measures to compensate for the opportunity cost o f schooling may be needed to keep children in school once they enroll, particularly children from poor families.44 In Rwanda, if the government followed the example o f Tanzania and Uganda-abolishing fees (and reimbursing schools for the lost income from fees) and paying for books and pedagogical supplies-public spending on primary education would need to rise by 0.18 percent o f GDP (=(0.227+0.243) x 0.252 x 1.5). This would bring total recurrent public spending for primary schooling to about 1.68 percent o f GDP, an increase o f some 12 percent over spending levels inrecent years. Such an increase would seem to be the minimum needed to remove some o f the immediate impedimentstoward achieving education for all. Table 3.3: Householdspendingoneducation, Rwandacirca 2000 Prii uy Secc iary Higher" Total I % of Publicb Private Publicb Private students reportingnonzerospendingon education' 97.9 99.5 97.4 97.2 93.8 Weighted annualspendingper student (FRw)~ 1,807 10,370 38,173 56,644 124,002 - As % of per capita GDP 2.2 12.7 46.6 69.2 151.5 - InUS $ 5 27 98 145 318 - Compositionof spendingper student (%) School fees 22.7 60.2 54.4 68.4 68.4 - PTA contribution and other chargese 6.4 0.9 5.6 1.9 0.0 - Books and school supplies 24.3 15.6 11.8 9.0 8.1 - Schooluniforms 45.0 20.1 11.6 9.0 0.3 - Transportation to andfrom school 0.2 2.3 6.1 4.8 11.6 - Other expenses 1.3 1.o 10.5 6.9 11.5 - Aggregate householdspendingoneducation(mill. FRw) 2,647 114 3,042 3,482 1,223 10,508 Sub-sectorshare of aggregate (%) 25.2 1.1 29.0 33.1 11.6 100.0 As %of GDP - - - - - 1.5 As %of public recurrentspendingon education - - - - - 40.7 Memo. item: Nationwide number of students 1,464,59 1,626,6( 4 10,978 79,699 61,464 9,866 Note: -denotes not computed. a. To avoidsmall sample size problems, separateentries are not shown for the public andprivatesectors. b. Refers to state and libre subsidid schools; teacher salaries inboth types ofschools are financedby the govemment. State schools are manageddirectly by the government, whereas libre subsidie'schools are managedby churches andother nongovemmentalorganizations. See chapters 5 and6 for more details. c. Resultsinthis table are basedon samples of6,284 primary schoolpupils, 945 students insecondaryschool, and 72 students inuniversities andother types of institutions d. Refersto weighted averageacrossall students, includingthose reporting zero spending. e. PTA refersto the Parent-TeacherAssociation. Source: Authors' calculations are based on the 1999-2001Enqugte Inte'grul de la Condition de Vie (EICV). Data on enrollmentsare basedon sources cited in tables A2.1 and 5.1. Data onGDP in2000 are from World BankAfrica RegionLive Database(May 20,2002 version). Additional details are shown inappendix table A3.3. 44 See Patrinos and Ariasingam (1997) and Patrinos (2002) for examples of the many mechanisms used by countries to accomplish this. 44 A closer lookat public spendingon education 3.8 Below we take a closer look at the functional distribution o f public spending on education, first by examining the broad categories o f spending reported inthe budget,and secondbybuildingfrom the bottomupthe spendingallocations by level o f education as implied by the number and distribution o f teaching and nonteaching staff inschools and the pay structure. 3.9 The overall functional distribution of spending. As previously documented, overall recurrent spending on education has been rising inrecent years. As table 3.4 below shows, this increasehas beenaccompaniedby anoticeable shift in spending favoring financial aid and foodstuff for students in secondary and higher education at the expense o f allocations for personnel emoluments, particularly at the primary level, and operating expenses. The increase in allocations for bursaries coincides with the activation o f financial assistance provided through the Genocide Fundto orphaned students; the bulk of the current beneficiaries are those attending secondary school. Duringthis period, funding for foodstuff also claimed a risingshare o f spending. The government has moved to curb this trendby shifting to a system of lump-sumcash allocation instead of basing the allocations on physical quantities of food allowance per student The new arrangementwill helpto tightencontrol over this itemof spending, butjustifying its continued inclusioninthe budget is difficult given that the beneficiaries are the already privilegedstudents inpost-primary education. Table3.4: Distributionof recurrent publicspending on education, Rwanda, 1999-2001 (Percent unless c envise indicated) Type of spending 1999 2000 2001 Salariesof teachers& 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 Studentbursariesb 14.3 24.3 24.0 Secondary 2.3 11.9 11.6 Local higher education 3.4 5.2 4.9 Foreign higher education 8.6 7.2 7.5 Foodstufffor students 6.0 8.2 8.7 Secondary 1.2 3.6 3.6 Higher education 4.8 4.6 5.0 Systemadministration 7.2 4.0 7.2 Otheroperatingexpenses' 13.4 8.4 6.7 100.0 100.0 100.0 Total recurrent spending 21,862 25,212 27,520 Ia. (inmillions ofcurrent FRw) Total recurrent spending as % of GDP 3.4 3.6 3.6 Excludes spendingonteacher salary arrears, the university hospital, t Institut Suutieur des nances Publiaues.and irious vocational and intable 3.2. b. Includesbursaries administered through MINALOCandthe Genocide Fund. c. Includesspending on maintenance,transport, utilities, and other m i n g costs. Source: See appendix table A3.4. 45 3.10 Seeing spending allocations from the bottom up. To obtain this view, we merged informationon the distribution o f personnel and the pay structure to piece together the allocation o f spending on wages and then completed the picture by adding information on spending on nonsalary items as reported inbudget documents. The result reveals spending allocations from the bottom up by level o f education and provides details that would otherwise be obscured. The calculations focus on primary and secondary education, because these levels comprise the bulk o f the system. For higher education, we did not follow the procedure, because the requisite data were unavailable; we relied on budget documents alone, which as it turns out did contain sufficient detail for our purpose here. 3.11 Consider first the distribution o f staff working inpublic (i.e., state and libre subsidie? primary and secondary schools (see table 3.5). The data are based on a 1999 census o f school personnel conducted by the Ministry o f Education, which documented each employee's educational qualification and work responsibilities. From the data, it i s possible to reclassify teachers with no teaching duties and to evaluate the full-time equivalent number o f teachers in the tronc commun and upper secondary cycles, according to the time utilization o f the teachers; thus, o f the 3,257 secondary teachers on the payroll in 1999, the calculation indicates that nearly 60 percent o ftheir teaching time was devotedto classes inthe tronc commun cycle. Table3.5: Numberof teachers and school-leveladministrativestaff ingovernment-financed primary and secondary schools, Rwanda, 1999 Teachers Administrative Withduties teaching With other staff Total duties Primary 24,982 40 1,329 26,351 Secondaryb 3,257 245 964 4,466 Tronc commun cycle 1,907 - - - Uppersecondary cycle 1,350 - - - 3.12 Consider next the distribution o f these school-level personnel by salary grade (see table 3.6), again based on the 1999 Ministry o f Educationcensus o f school personnel. Using information from the census on each staffs salary grade and years o f experience, we are able to locate each person inthe salary scale and thus compute his or her annual salary, including benefits. On average, primary school teachers receive remuneration o f about 4.0 times the per capita GDP, compared with the corresponding multiples o f 5.7 times for tronc commun teachers and 6.8 times for upper secondary teachers. In both the primary and secondary cycles, administrative staff tend to be better educated and have more years o f experience than teachers as a group and they are generally paid more as a result. 46 Table 3.6: Staff distributionby salary grade and average salaries in publicschools, Rwanda, 1999 (Percent unless otherwise indicated) Primary Secondary I Pay category Educational attainment Administra- Administrative tive staff commun I staff sec. I Ungraded CA; ES1,2 & 3; GERAI, SF, CERAR` 22.5 3.6 0.0 0.0 0.0 6 Incompleteupper secondary 20.5 8.1 1.3 0.8 0.2 Lower secondary teacher training diploma 3.7 2.0 0.0 0.0 0.1 4 Upper secondary diploma (old system)d 17.9 30.2 1.8 2.6 3.1 3 Upper secondary diploma (new system)d 35.0 55.4 74.3 42.8 58.3 2 2- year post upper sec. diploma 0.0 0.2 15.1 30.1 23.5 1 University degree 0.0 0.1 6.8 22.1 10.8 Other - 0.4 0.4 0.7 1.6 4.0 Yumber o f staff 24,982 1,369 1,907 1,350 1,209 Average years o f experience 8.1 11.1 5.4 5.6 9.4 Average annual salary, includingbenefitse Inthousands ofcurrent FRw 308.5 388.6 443.2 526.2 594.9 As % o f per capita GDP 4.0 5.0 5.7 6.8 7.7 Vote:-denotes "not applicable." a. Includesonly teachers withteachingduties. b. Teachersteachinginboth cycles are countedinfull-time equivalents accordingto the number o fhoursthey teach ineachcycle. c. Refers, respectively,to Certificat d Yptitude (CA) obtainedat the endofthe old 8-year primarycycle; to Ecole Secondaire(ES) of 1to 3 years; to the completion ofvariousvocational coursesafter the old 8-year primarycycle offered inthe Centre d 'EnseignementRural et Artisanal Intdgrd (CERAI), andthe girls-only SectionFamiliale (SF), andthe boys-only Centresd'EnseignementRural etArtisanal de Rwanda (CERAR). d. Underthe oldsystem, anuppersecondarydiploma is eamedafter 3 to 5 years ofsecondaryschooling, whereas underthenewsystem, itiseamedafter6 to I years. e. Includesbenefitsfor transportationandlodging(introduced in 1999) as well as estimatedfamily allowancespaidto teacherswho are householdheads, and the state's contribution of 5 percent ofsalariesto social security. Source: Basedon data from MINEDUC's 1999census ofteachers andinformationsuppliedby MINEDUC onthe structureofteacherpay andbenefits. See appendix tableA5.1 for additional informationon the salary structureby qualification. 3.13 Puttingtogether information fromthe precedingtwo tables allows us 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 taken directly from the amounts reported inthe budget documents for 1999.4' The reported aggregate 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 o n the assumption that the distribution reflects the cost of personnel management across levels. W e then added the details on nonstaff recurrent spending as reported inthe budget documents, being careful to eliminate spending that is unrelated to education (e.g., the cost o f a university hospital). The results appear in table 3.7. As a check on the overall calculations, we note that the bottom-up view yields an aggregate spending of FRw 22,093 million, compared with FRw 21,862 millionreported inthe budget, an excess o f 5.6 percent. The gap is larger than we would have liked in this kindof accounting exercise, but may be explained by the fact that the government had accumulated arrears on teacher salaries inthe aast few vears. 45 For more details on the way this was done, see the footnotes to table 3.7. 47 Table 3.7: Estimatedcurrent publicspendingon educationby function andlevel, Rwanda, 1999 (Millions I 'current FRw, unless otherwise indicated) Function Primary Secondary education Higher education Tronc Upper education All levels secondary Salaries & benefits Staff at central levela 1,056 158 129 405 1,748 School-level administrative stafp 532 404 317 } } Teachers with teaching dutiesb 7,707 845 710 3,239 13,755 Yonstaffrecurrentspending System administration" 1,140 106 79 177 1,502 School-level operating costs' 346 801 475 - 1,62 1 Foodstufffor studentsd 0 169 95 1,045 1,308 Student financial aid Local studies 0 322 191 733 1,246 Studies abroad 0 0 0 1,889 1,889 %here 6 6 6 6 23 rota1current sDending 10,788 2,810 2,001 7,494 23,093 Memo item: Amount reported inbudget documents 21,862 (As % of estimatedtotal) (94.7) Vote: -denotes datanot available. i.Theinformationinthebudgetdocumentscanbecategorizedseparatelyinto(a) spendingonservicesthatbenefitprimaryandsecondaryeducationonly (such as the school inspectorate, the National Examinations Council, the National Curriculum Development Center) and @) spending on overall administration at the central level. They also permit a distinction in each of these categoriesbetween spending on salaries and that on materials inputs (transportation,fuel, supplies, utilities, maintenance, andso on). We apportionedthe reported amount initem (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 o f school level or institution-level staff, on the assumptionthat it correlates with the cost o f personnel management. b. Estimated accordingto the average annual salaries, including benefits, o f administrative staff and teachers inpublicly funded primary and secondary schools andthe number o fsuch staff as documentedinMINEDUC's 1999census o fteachers. c. Includes spendingon materials, books, and maintenance. For secondaryeducation, the amount also includes spending on equipment, whereas some o f this spendingmay bemore properly treatedas capital spending. No adjustmentis made here, becauseno information is available regarding the nature o f this equipment. d. For a few higher educationinstitutions, the amountspent on foodstuff is not itemizedinthe budgetdocuments, andare thus estimatedseparately from the datasuppliedby individual institutions. The estimatesareaddedto the itemized budgetamounts to obtain an aggregate for higher education. e. Refers to spending on the National Commission for UNESCO and on researchon science and technology.We apportionedthese in equal parts across the various levelsonthe assumptionthat suchspendingbenefitsthe entireeducationsystem. Source: Authors' estimate based on data on the number and distribution by pay category o f teachers and administrative staff counted in MINEDUC's 1999 census ofteachers, supplementedby data on the structureofteacher remunerationanddatafrom the 1999 executedbudget o fthe Rwanda. See also table 3.4 andappendix tableA5.1. 3.14 To facilitate a reading of the results in the previous table, we can rearrange them under three major rubrics: management overhead, service delivery at the facility level, and student welfare. This rearrangement puts a new perspective on the management of public spending o n education by drawing attention to the extent to which spending is in fact oriented to the core business of service delivery. Management overhead can be thought of as a back line service to support 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 discharge their functions. The core business of service delivery takes place at the level of individual schools and is animated by spending on teachers, school administrators, and support staff, as well as on pedagogical supplies and other operating inputs. In addition, spending on student welfare can be categorized as a back line item, given its indirect link to teaching and leaming at the level of individual classrooms and 48 schools. To continue with the business analogy, this item can in fact be viewed as a form of customer rebate, because its effect is to reduce the cost of education services to the consumer (i.e., students and their families). 3.15 The rearranged data appear intable 3.8 and figure 3.2 below. What do they tell us about the priorities in public spending on education? Although there is admittedly no obvious benchmark for judging the distribution o f spending, a pattern o f allocations that glaringly fails to support the core activity of teaching and learning at the facility level would raise legitimate questions on whether resources are sufficiently focused on service delivery. For the system as a whole, less than two- thirds of the total public recurrent spending on education in 1999 went to support service delivery at the facility level, whereas 14percent was absorbed by management overhead and nearly 20 percent went into student welfare. The 14 percent allocation for management overhead does not appear to be overly burdensome, but the share for student welfare is unusually large compared to spending patterns elsewhere. It is a reflection of the country's post-genocide recovery efforts, which aim both to rebuild the lost humancapital through heavy investment inhigher education and help students orphaned by the massacre. 3.16 Yet, closer examination of the spending patterns by level of education does raise some concerns. In primary education, management overhead exceeds 20 percent, reflecting relatively elevated spending on material inputs (such as transportation, utilities, maintenance, and so on), whereas at the school level, 96 percent of the allocation goes for staff salaries, leaving a miniscule 4 percent for books, pedagogical supplies, and other running costs and nothing at all for student welfare. Given the paucity of resources for these items, it is not surprising that a sizable portion of household spending at this level of education goes for school fees and PTA dues (both.kept for use at the school), as well as for books and school supplies. Giventhese patterns, a re-evaluation of the allocations mightb e warranted to identify potential room for shifting some of the government's spending on overhead toward defraying more of the running costs at the school level. Inprimary education, there i s currently no provision to subsidize orphaned children, an omission that poses an obvious impediment to the schooling of the most vulnerable children. 3.17 Insecondary education, spending allocations are comparable between the lower and upper secondary levels-roughly a tenth on overhead, three-quarters on service delivery, and the rest on student welfare. Under service delivery, administration appears to take up a significant share o f resources and may warrant a closer look to improve utilization of personnel in schools to avoid, for example, poorly justified assignment o f teachers to administrative, rather than teaching tasks. With regard to allocations for student welfare, the total is made up of spending on school feeding and bursaries for orphaned students. The raison d'gtre for the general food subsidies appears weak, because not all students at this level are poor--or at any rate, poorer than children inprimary school who currently do not get free food.46The justification for bursaries is stronger, because they are directed at orphans. A s an earlier chapter has shown, one result of the 1994genocide was to swell the number of 46In an effort to control spending on foodstuffs, the government recently replaced calculation of the subsidies based on the number o f students a school enrolled with a lump-sum cash amount, fixed innominal terms inthe basedyear. As more and more o f the country's schools operate as day rather than boarding schools, one should also expect the amount spent on foodstuffs to decrease. 49 orphans currently in the secondary school age range. As this bulge passes out o f the system or goes onto higher education, one should expect allocations for bursaries to decline, though not disappear altogether, because orphans continue to be created by the HIV/AIDSepidemic. Table3.8: Functionaldistribution of publiccurrent spending oneducation, Rwanda, 1999 (Percent unless otherwise ndicated) Functiona Primary Secondar: :ducation Higher education Tronc Upper education All levels commun secondary Managementoverhead 20.4 9.6 10.7 7.8 14.3 Staff salaries 9.8 5.6 6.5 - - Material and other inputsb 10.6 4.0 4.2 - - Servicedelivery at facilitylevel 79.6 73.0 75.0 43.2 66.4 Salaries o fteachers with teaching duties 71.4 30.1 35.5 - - Salaries o f administrative staff: 4.9 14.4 15.8 - - Material inputs & other operating costsd 3.2 28.5 23.7 - - Studentwelfare 0.0 17.5 14.3 48.9 19.4 Foodstufffor students 0.0 6.0 4.7 13.9 5.7 Bursaries for local studies 0.0 11.5 9.5 9.8 5.4 Bursaries for studies abroad 0.0 0.0 0.0 25.2 8.2 Total 100.0 100.0 100.0 100.0 Memo items Total current spending (millions o f current FRw) 10,788 2,810 2,001 7,494 23,093 Subsector share o f spending (%) 46.7 12.2 8.7 32.5 100.0 Note: -denotes data not available. a. The categories represent a regrouping of the expenditure categories in table 3.7; management overhead refers to spending on system administration service delivery refers to spending on the salaries of teachers and administrative staff at the level of individual schools or institutions, as well as 01- operatingexpenses at the facility level;and student welfare refers to spendingon foodstufffor students and student fmancialaid. b. Includesthe cost oftransportation, utilities,supplies, maintenance, fuel, and other runningcosts. c. Includes the salariesofteachers without teaching duties, schoolheads, secretaries,and other administrative staff. d. Includestextbooks, equipment, utilities,transportation, maintenance, and other runningcosts at the facility level, includingsalariesof casuallabor. Source: Authors' estimate based ondata intable 3.7; see text discussionfor item definition. 3.18 With regard to spending allocations in higher education, there is less detail, but even limited information is helpful inraising questions about the priorities in spending. The most striking result is that the core business of service delivery received only 43 percent of the subsector's total recurrent spending, compared with 49 percent for student welfare in the form of foodstuff for students and bursaries. Because spending on overhead is not especially high at less than 8 percent, a rebalancing of spending in favor of service delivery would essentially require cutting back on student welfare. The only possible way to avoid the cutback is to raise the share o f higher education in overall spending on education, but this option is untenable inview of the subsector's already highshare of total spending (which grew from 33 percent in 1999 to 37 percent in2001). The questionthen is for policymakers to evaluate the merits of shifting spending within the subsector and to form a strategy for minimizing the potential adverse effects associated with cutbacks on student welfare; these effects may include, for example, a bias against students from poor families in the socioeconomic composition of the student body and inadequate 50 incentives for students to specialize in fields of study with high social (as opposed to private) returns. Figure3.2: Functionaldistribution of publiccurrentspending on education,Rwanda, 1999 100 0Student welfare Service delivery Management overhead 20 0 Primary Tronc Upper Higher All levels education commun secondary education Source: Based on data intable 3.8. Publicspendingper student 3.19 The foregoing documentation of spending allows us to evaluate and compare public spending on education on a per student basis. Below we elaborate on the patterns that emerge and discuss the underlying sources of differences across levels. Data from other countries are also presented to put the discussion in a comparative perspective. 3.20 Unitwendingacross levels of education. The relevant results appear in table 3.9, expressed in current FRw, as a multiple of the capita GDP, and relative to per capita spending in primary education. They pertain to the public sector at all levels; in higher education, a distinction is made between local and foreign ~tudies.4~ What then are the patterns? The immediately obvious one is that unit costs are highly skewed in favor of the post-primary levels. It is another reminder o f 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 education, the corresponding ratio is 95 times for local studies and 275 times for studies abroad. Within secondary education, unit costs inthe tronc commun cycle are 0.8 times that in the upper secondary cycle, suggesting that service delivery arrangements (e.g., use of specialist teachers, class size, teaching workloads, and instructional time for students) inthis subcycle more closely resemble those inupper secondary education than the arrangements in primary schooling. The relatively high costs in the tronc commun cycle will obviously limit the government's ability to accommodate the rising demand for places in this cycle as more and more primary schoolchildren complete their schooling. `'Torecall, the public sector at the primary and secondary level would include state and libre subsidii schools. 51 Table 3.9: Publicspendingper studentby level of education inthe public sector, Rwanda, 1999 Total public recurrent spending(mill. of FRw) 10,788 2,810 2,001 4,811 5,605 1,889 Overallnumber of students 1,428,708 79,454 45,670 125,124 9,866 902' %inpublic sector schoolsb 99.3 55.7 57.1 56.2 78.5 100.0 a. Excludes spendingon bursaries for studentsattendingprivate secondary schools; the amount is estimatedby proratingspendingon bursaries according to the distribution of enrollments betweenthe public (is.,state andIibre subsidie' schools) andprivate schools. Note that, although private school students are probably less likely than public school studentsto receivea bursary, the amount o feachawardis higher for studentsattending private schools. b.Refersto state or libre subsidii schools. c. Includesonly studentsstudyingabroadongovemmentscholarship. Source: Authors' calculations based on public spendingdata in table 3.7 and enrollment data in table A2.1; data on number of students abroadbased on MINEDUC's unpublishedworking database. 3.21 Sources of differences in unit spending across levels of education. Why are post-primary levels of education so much more costly than primary education? To answer this question, we can decompose the overall spending per student into its component parts, beginning with the following expression: Where US is overall unit spending, TSi is tot&, spending on component anr P I i s the number of public sector students. Furthermore, the unit spending on the salaries of teachers and nonteaching staff can be expressed as a function o f two items, the average salary cost of these staff and the corresponding pupil- to-staff ratios, as follows: 52 uss = - = ASS*l/(P/NS) TSS - ASS NS * = - ASS P P PSR Where USS is unit spending on staff salaries, TSS is total spending on staff salaries, ASS is the average salary per staff, N S is the number of staff, and PSR i s the pupil-to-staff ratio. The last equation makes it possible to compare differences in unit spending across levels or types o f schooling in terms of underlying differences inlevel of salaries per staff and the pupil-to-staff ratio, which may be interpreted as a proxy of the intensity with which staff are used. 3.22 Keeping in mind the above expressions, consider now the results in table 3.11. Note that, because of data limitations, they pertain only to primary and secondary education. The second column inthe table shows the overall cost per pupil inprimary education (FRw 7,604) as well as its composition: 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 above, we compute the cost of teacher salaries per pupilsimply by dividing the average annual salaries of teachers (FRw 308,522) by the pupil-teacher ratio (56.8); likewise, the cost of school-level administration i s the result of dividing the average annual salaries of school administrative staff by the ratio of students to such staff. Similar calculations yield the corresponding costs per secondary student (overall and separately 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 o fthe corresponding indicator inprimary education. 3.23 What then are the sources of the much higher unit spending in secondary education relative to primary education-8.6 times overall and 7.9 and 9.7 times, respectively, inthe lower and upper secondary cycles? A s we have seen earlier, teacher salaries are the single most important cost item (accounting for more than 70 percent o f total spending), so part of the answer lies inthe differences here. The cost o f teachers per student in 1999 was 4.1 times as high in secondary schools as they were inprimary school, reflecting higher teacher salaries and smaller student-teacher ratios inthe secondary cycle. A closer look at these variables in the table shows that the latter factor contributed more to the higher cost o f teachers per student in the secondary cycle. Better staffing ratios rather than higher average salaries were similarly the main reason why the cost o f school-level administrative staffper student inthe secondary cycle in 1999 was 27 times the corresponding cost in the primary cycle. Fortunately, spending on school-level administration in 1999 was not a major item of spending, thus, the impact on overall costs was limited. Secondary schools also received more fundingper student for material inputs-about 74 times that at the primary level. Again, this item accounted for a small share of overall spending, so its impact on the difference in overall costs per student between the two cycles o f schooling was also limited. The last item in the table pertains to management overhead, which took up about 20 percent of the overall spending per primary pupilin 1999 and averaged about FRw 1,552 per pupil. Spending per student on this item in secondary education was even higher-by a multiple of about 4.4 times. 53 Secondary education Overall spending per student and its composition education Primary (as multiple of primary education) Both Tronc Upper (FRw) levels commun secondary Overallspendingper student 7,604 8.6 7.9 9.7 Cost ofteachersper student 5,433 4.1 3.5 5.O 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 per student Average annual salary of administrative staf Studentwelfareper student Managementoverheadper student 4.4 Nores:-denotesnot applicable,becausetherewas no spending on studentwelfareat the primarylevel.Blanksdenote thatthe calculation was notmade, becausedifferencesacrossthe two levelsare small andreflectassumptionsinapportioning spendingacrossthe two cycles ofsecondaryschooling. Public sectorrefersto state and fibre subsidii schools. Source: Authors' calculationsbasedon data intables 3.6 and 3.7. See also appendix table A3.5. 3.24 Comparative perspectives on the pattern of unit spending. How do the foregoing levels and patterns of unit spending inRwanda compare with those inother countries? Table 3.11 contains some relevant data in this regard. To facilitate the comparisons, all values in the table are denominated as multiples of the per capita GDP. Because the cross-country data refer to spending distributed across all enrollments rather than only those in the public sector, we adjusted the data for Rwanda accordingly. For completeness, the table also shows in parentheses the data on spending per student in public schools. Public spending averaged across all students inthe system can be interpreted as reflecting the intensity o f public spending across levels o f education and is suitable mainly for comparing the structure of spending across levels. Public spending per public school student, on the other hand, is more appropriately seen as a measure of the costliness of service delivery through the public sector.48 48 Because of differences in data definition, comparisons across countries' spending per student in the system should be handled with the usual caution-particularly in higher education where enrollments in private institutions may not be fully accountedfor-and distance education treated differently from country to country. 54 Table 3.11: Internationalcomparisonsof current publicspendingper student As amultiple ofper capitaGDP As amultipleof No. of primaryb I countries Local Primary Secondary higher Secondary Local highe education education Rwanda, 1999a 1 0.10 0.50 7.3 (0.10) (0.84) (9.3) (8.6) (95.2) Regionalaverages, 1990sb FrancophoneAfrica 15 0.15 0.49 5.6 Anglophone Africa 9 0.10 0.66 6.3 Latin America 10 0.07 0.11 0.7 Asia 8 0.08 0.19 0.9 11.3 Middle East & NorthAfrica 6 0.11 0.30 0.9 I a. Two sets of figures are shown for Rwanda: The first row refers to current spending averaged across all students at each levr regardless ofwhetherthey are inpublicor privateschools andregardless oftype of institution.The secondrow (inparentheses) refers to currentspendingaveragedacross only pupils inpublic schools. For highereducation, the figuresinparenthesesrefer to per student spendinginthetraditionalpublic institutions. b. Averages basedonly on data for countrieswithpercapitaGNPbelow$1,000 in 1993. 3.25 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 higher than those o f countries inLatinAmerica and Asia, and quite a bit lower than those o f Francophone Africa. Insecondary education, Rwanda's figure surpasses the averages for all regions except Anglophone Africa. It is inhigher education, however, that Rwanda's level o f spending per student shows up as a true outlier on the high side: 75 times as high as the level of spending per primary schoolchild, compared with the corresponding multiple of 63 times o n average in Anglophone Africa, 37 times inFrancophone Africa, and no more than 11times inthe other regions. Had foreign higher education been included inthe calculation, the ratio would have been even higher. Overall, the results suggest a structure o f spending in Rwanda that is heavily skewed in favor o f post-primary levels, but especially higher education) a pattern that inevitably translates into significant inequities in the incidence ofpublic spending. 3.26 Implicit trade-offs inthe input mix inprimarv education. For any given level o f spending per student, education services can be delivered using different inputmixes-for example, less costly teachers combined with more favorable staffing ratios or vice versa and more spending on salaries compared with more spending on nonsalary inputs. In table 3.12 below, we use some cross-country data on teacher salaries and pupil-teacher ratios inprimary education (the level for which comparative data are relatively plentiful) to deduce the implicit trade-offs that characterize primary schooling inRwanda. The discussion also provides a basis for opening up a dialogue on options for improvement. 55 Table 3.12: Publicsector teacher salariesandpupil-teacher ratios inRwanda and other countries Countryhegion Average salary o fteachers (as a multiple o fper capita GNP) Pupil-teacher ratio Rwanda, 1999a 4.0 56.8 Regionalaverages, 1990sb FrancophoneAfrica 6.3 53.2 Anglophone Africa 3.6 38.7 LatinAmerica 2.3 31.0 Asia 2.5 37.9 MiddleEast 3.3 25.6 All countries 3.9 40.0 a. Data for secondaryeducationare the weighted averagesof lower and upper secondaryschool. All datareferto the public sector. b.Datareferto countrieswith per capitaGNPbelow $1,000 in 1993. Pupil-teacherratiospertainto systemwideaverages. Source:Table A3.5 for the data on Rwandaand Mingat and Suchaut 2000 for thecomparativedata 3.27 Recall from equation 3.2 that the cost o f teachersper pupil-the single most important component o f overall costs-depends on the level of teacher salaries and the pupil-teacher ratio. Ingeneral, the pattem inthese variables across countries is what one would expect: the less costly the teachers, the lower pupil-teacher ratios tend to be. Teacher salaries in Rwanda exceed the average for Anglophone Africa by 11 percent, but its pupil-teacher ratio i s higher by nearly 50 percent; compared with the corresponding averages for the Middle East and North Africa, these variables are higherinRwandaby21 and 122percent, re~pectively.~~Becauseoverall spendingper student inRwanda is comparable to the averagesfor these regions (around 0.10 times the per capita GDP), we can deduce from the much higher pupil-teacher ratios in Rwanda that an implicit trade-off was beingmade infavor o f nonsalary components. As we have documented above, the trade-off appears to emphasize system-level managementoverhead rather than services and material inputs at the school level-a regrettable result, because research and experience consistently point to the critical role o fbooks andpedagogical supplies inenhancing student 3.28 What options exist then to improve conditions at the classroom level? Inlightofthe foregoing deduction from the comparative data, one optionwouldbeto redirect some o f the spending on management overhead in favor o f a smaller pupil- teacher ratio and larger allocations for books and material inputs. If, for example, spending on management overhead was cut by half and the resourcesthus freed used to hire more teachers, itwould be possible to lower the pupil-teacher ratio to about 50. The reduction is certainly an improvement, but a ratio o f 50 is still high and the paucity o fbooks and pedagogical inputsremains unresolved. 3.29 Other options to improve the leaming environment would clearly warrant consideration as well. If overall spending on primary education remained unchanged, the only real option within the subsector would be to tighten control on teacher salaries, either by changing the criteria for new hires or slowing the pace of 49InRwanda, the exceptionally highpupil-teacher ratio translatesinto short school days-about 2.5 to 4 hours in duration-for children inthe first three grades o f the cycle, because teachers typically are assignedto teach two shifts (see chapter 5 for more details on this issue). The squeeze on public funding for books and pedagogical materials effectively transfers the burden o f financing these inputs onto pupils and their families, a result confirmed by the pattern of household spending documented earlier inthis chapter. Because o f the prevalence o fpoverty inRwanda, it is likely that families spend less than what is infact needed for effective leaming. 56 salary increases for those already on board. Such policies obviously need careful evaluation, but inexploring their relevance inthe early stages o f policy development, we note that in low-income countries that have achieved universal completion o f primary schooling, teacher salaries typically average about 3.5 times the per capita GDP, making itpossible for these countries generally to maintain a pupil-teacher ratio of around 40 (Mingat, Rakotomalala, and Tan 2002). Ifaverage salaries relative to the per capita GDP in Rwanda were made comparable to the level inthese countries, all things remaining unchanged, it would be possible to lower the pupil-teacher ratio from 57 to 50; if this change were combined with halving the spending on management overhead and using the resources thus released to hire more teachers, the ratio would drop further to 44. 3.30 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 all but 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, butby mid-career, the gap hasjust about evaporated. The bias infavor of the better qualified is consistent with government policy to attract the best people into the teaching profession. In the absence o f budget constraints, the policy would improve staff quality without requiring a trade-off against other inputs-such as reasonable ratios of pupils to teachers and adequate availability of pedagogical materials-that also contribute to teaching effectiveness. The reality, however, is that budget constraints are tight. Ifloweringthe pupil-teacher ratio is an objective, the optionofkeeping a lid on teacher costs cannot thus be ruled out entirely. Table 3.13: Public sector teacher take home pay andthe payof other workers, Rwanda, 2001 (As amultiple c per capita GDP) ":p Starting payb I Payat mid-careerc reacher pay categorf Educational attainment Teachers Other Teachers (1) workers (1)/(2) workers (3)/(4) (2) (3) 6 Incomplete upper sec 2.2 } 2.4 I 2.4 1.08 } 3.6 1 0.97 4 3-5 years post-primary education (Le., D3, D4, or D5) 3.0 4.6 Uppersec diploma (new system) (Le., D6/D7) 4.1 3.5 1.17 6.5 5.2 1.25 2-year post upper sec. dip. (e.g., I Bac.) 5.6 4.9 1.14 8.4 1 University degree 8.3 6.8 1.22 11.1 I II mdhousing. eealso table A5.1 for additional details. a. Pay category5 ;s for teachers with a teacher-trainingdiploma, a qualifickon obtainedafter completing teachingtraining at the lower secondary level. It is excludedherefor lack o fa comparable groupinthe generallabor force. b. The starting pay for teachers is taken from the salary scale shown in table A5.1. For other workers, it is simulatedat ages 20, 20, 22, and 24, respectivelyfor qualifications shown, basedonthe Mincerianeamingsfunction reportedinappendix tableA3.6. c. The payat mid-career for teachers is estimatedat the midpoint betweenthe entry andtop pay on the pay scale for teachem (appendix tableA5.1). For other workers, it is the pay simulated at age 45 basedonthe Mincerianearnings function reportedinappendix tableA3.6. Source:Authors' calculationsbasedon the data intableA5.1 and simulations basedonthe Mincerianeamings functionsreportedintable A 3.6. Policyimplications 3.31 In Rwanda, public spending on education has been rising steadily in the post-genocide years, reaching a historical high of 5.5 percent of GDP in 2001. 57 Because muchof this increasehas beenfor capital spending, allocations to support the day-to-day operations o f the education system rose to only 3.3 percent o f GDP in 2001-about the same level as in the late 1980s. The amount i s augmented by education-related spending that occur through channels other than the Ministry o f Education, including in particular the Ministry o f Youth, Sports and Culture which finance vocational training and the Genocide Fund which provides financial assistance to orphaned students. Adding the extra spending brings the total recurrent spendingto about 3.8 percent of GDP in2001. This level of spending is comparable to the 4.0 percent of GDP that low-income countries typically spendon education, but given the extraordinary burden of orphans in the population, one could argue that Rwanda's allocation remains on the low side. 3.32 A case can therefore be made for continued increases in spending on education, particularly recurrent spending. Yet, in a context where the government already spends more than a quarter o f its resources on education, hrther increasesare likely to come only slowly, certainly more slowly than the pace o fincrease inthe late 1990s. The ineluctable implication is that future progress ineducational development will increasingly require more effective allocations o fspendingwithinthe sector. 3.33 Inacountry stillrecoveringfromthe devastationofthe 1994genocide, it may appear heartless to even talk of trade-offs in spending allocations within a sector such as education. The pattern o f spendinginrecent years, however, shows that implicit trade-offs have in fact occurred. The government has placed substantial priority on rebuildingthe country's intelligentsia inthe years following the genocide, 2001,higher education claimed an astonishing 37 percent of total recurrent spending, and this is reflected in the very rapid increases in the distribution o f spending. By upfrom between 12and 15percent throughout the 1980s.Correspondingly, the shares o f both primary and secondary education fell significantly below their shares in the during 1999-2001,comparedwith anaverageofabout 1.9 percentinthe late 1980s. 1980s. Public spending on primary education averaged about 1.5 percent o f GDP 3.34 The bias in fbnding against primary and secondary schooling effectively shifts part o f the burden o f education finance to households. These levels o f education account for nearly 90 percent o f all spending on education by households. Parents pay school fees and Parent-Teacher Association dues to help defray schools' operational costs; they are also responsible for books and school supplies. These arrangementsinprimary schooling set Rwanda behindcountries such as Uganda and Tanzania, where school fees have been abolished and textbooks are now provided free o f charge as part o f an overall strategy to ensure that all children are able to complete primary schooling and are provided access to basic learning materials. Insecondary education, fees are comparably highinthe public and private sectors inRwanda, blurringthe distinction between the two sectors. As expected, the country's current pattern o f intrasector allocations shows up in a highly skewed structure of public spending per student by level: a student inhigher education is 75 times as heavily subsidized as a child attending primary school, a result that ranks Rwanda as having one o fthe leastequitable structures o f education finance. 3.35 What do the foregoing arrangements in education finance imply for policy development? Investinginpost-primary levels o f education will continue to be important, but ways must also be found to rebalance the allocations to support primary education, particularly if the government i s to achieve Education For All by 58 2015 as part of its strategy for educational development and poverty reduction. As more students complete primary schooling, the pressures to expand secondary education, especially the tvonc commun cycle initially, will also mount, and resources would have to be found for expanding that level as well. One potential source o f savings i s the untargeted subsidies for secondary and higher education, such as spending on foodstuff. To give an idea on the scale o f the opportunity cost, a simple calculation suffices: eliminating this item from higher education alone would boost the recurrent budget for primary schooling by nearly 10 percent, or eliminate the burden of paying for textbooks and school supplies for 17,000 children. In higher education, a closer look at the spending on bursaries might also be warranted to ensure that they are indeed targeted to students from poor families and for fields o f study where public subsidization isjustified (ie., where society benefits more than the individuals concerned). Within 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 to expand primary and lower secondary education, ways to manage the cost o f service delivery at these levels also require attention. The options inprimary education include reallocations of spending from management overhead to pedagogical materials and making sure that teacher salary costs are kept at sustainable levels. In lower secondary education, policies to lower unit costs-particularly through better management of staff utilization-will be criticalto efforts to expand enrollments ina fiscally sustainable manner. Conclusion 3.36 Education finance is at a crossroads in Rwanda today. Financing for the sector has recovered quickly in the post-genocide years, but the pace o f future increases i s likely to slow down in light o f the many competing claims on the public purse. Inthis context, a combination o fpolicies will probablybe neededto achieve an efficient and equitable allocation o f spending, one that would help create effective learning environments-at all levels, but especially at the base of the education pyramid-that are characterizedby adequate overall fundingas well as a good mix of school inputs. Although an appropriate balance among the various policy options would be difficult to determine on the basis o f financial considerations alone, the information presented above highlights potentially fruitful directions that can explored as part o f the process o f policy development. Especially important i s to increase funding for primary education, both by redirecting resources to the subsector as well as through better management o f costs and allocations within the subsector. 59 Chapter 4: Socioeconomic Disparities inEducation 4.1 Chapters 2 and 3 looked at the performance of Rwanda's educational system in the aggregate, making no distinction between different groups. This chapter focuses on disparities in education in terms of enrollment, student flow, and the share o f public spending appropriated by different groups in the country. In particular, it emphasizes differences across provinces, gender, urban-rural localities, and socioeconomic groups. Because orphans are so numerous among Rwandese children, the chapter also compares educational access between orphans and non-orphans. 4.2 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 suggests 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, inparticular those who have lost bothparents and those who live away from their parents, probably as workers or street children with no adult supervision. Effort is needed more broadly to ensure that all children not only start grade one, but also reach the end of the cycle. 4.3 At the post-primary levels, the disparities widen substantially; rural children and those from poor families lag especially far behind. It is noteworthy that differences between girls and boys emerge only in higher education. Thanks to the government's foresight in setting up the Genocide Fund,orphans are relatively well represented insecondary education, the level that the fund currently targets. Although similar financial assistance to other disadvantaged youths is likely to improve participation rates, their underrepresentation inpost-primary 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 wouldseemparticularly important. 4.4 Aside from redressing socioeconomic disparities in enrollment, scope also exists for improving the incidence of public spending on education. In Rwanda, the bias in favor of higher education is currently so strong that the 10percent best educated ina cohort appropriates nearly three- quarters o f the cumulative public spending on education received by the cohort. As a result, the Rwandese system is one of the least structurally equitable inSub-Saharan Africa today. Overview of participationrates 4.5 The existence of disparities in school participation may be seen in table 4.1.51At the primary level, the overall gross enrollment ratio rose from 74 percent in 1992 to 108 percent in2000. The disparity between the lagging and leading provinces has, however, not narrowed.52 In 1992 gross enrollment ratios ranged from a low o f 61 percent inKibungo to 92 percent inKigali Ville-a gap o f 31 percentage points. In 2000 the gap was still 31 percentage points with the ratios ranging from 97 percent in Buhre to 128 percent in Kigali Ville. At the secondary level, the overall gross enrollment ratio fell from 20 percent in 1992 to 11percent in2000, a pattern already noted inchapter 2. With the decline, the disparity between lagging and leading provinces has widened from 29 percentage points 51To maintain intemal consistency within the table, the data are basedon a single source for eachyear. The ratios for circa 2000 differ slightly from those reported inchapter 2 for reasons explained inthe relevant table footnotes. 52To recall, the gross enrollmentratio is calculated by dividing the total number of students enrolled at a given level of education by the population inthe official age range for that level. An alternative indicator o f 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 shall usethe gross enrollmentratio here to provide an overview o f the system's coverage. 60 (ranging from 9 percent in Byumba to 38 percent inKigali Ville) in 1992, to 42 percentage points (6 percent in Gisenyi to 48 percent in Kigali Ville) in 2000. At the tertiary level, the overall gross enrollment ratio was only 1.3 percent in2000. The provinces o f Kibuye and Ruhengeri had very few students in higher education, but the gross enrollment ratio was above the national average o f 1.3 percent inGitarama (1.7 percent) and KigaliVille (2.5 percent). The advantage of certain provinces at the secondary and higher educationlevels may partly be attributed to the fact that students move to the localities where these levels o f education are concentrated; however, because the survey included all members ina household as long as they have been present at some time within the last six months o f the survey, the disparities inparticipation rates cannot be wholly attributed to differences inthe supply o f services across provinces. Table 4.1: Gross enrollment ratiosby province, locality, gender, and incomegroups, Rwanda, 1992-2000 Primaryeducation Secondar 1992 I circa 2000 1992 Rwanda 73.9 108.3 20.4 By province Butare 69.6 96.9 24.2 9.7 0.8 Byumba 82.9 106.0 8.7 7.6 0.4 C Y W w J 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 Kigalirural 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 KigaliVille 92.1 128.3 37.7 47.8 2.5 By locality Urban 90.6 124.0 36.0 Rural 71.3 105.3 17.6 Index(Urban= 1.OO)" 0.79 0.85 0.49 By gender Boys 74.3 109.5 21.7 Girls 73.5 107.1 19.3 Index(Boys= l.OO)a 0.99 0.98 0.89 BYincome group b' 20% richest 89.3 127.6 39.1 29.4 6.1 40% middle 76.6 109.1 18.0 7.9 0.4 40% poorest 63.8 99.9 12.1 3.0 0.2 Index(Richest 20% = 1.OO)a 0.71 0.78 0.31 0.10 0.03 (ole: a dashdenotes datanot available. a. The index in the locality and gender blocks are computedas the ratio betweenthe enrollment ratio of the less-favoredgroup to that of the more-favoredgroup. Inthe incomeblock, it is the ratio betweenthe poorestandrichest groups inthe incomeblock. b. The incomegroups aredefinedusingprincipalcomponentsanalysisofsamplehouseholds' ownershipor accessto suchassets or facilities as pipedwater, flush toilet, electricity, radio, bicycle, motorcycle, car, refrigerator, andnumber of persons per room. The asset list is comparable between 1992and2000. Source: Authors' estimatesbasedonthe 1992RwandaDemographic andHealthSurvey for the data for 1992; 2000 RwandaDemographicand Health Survey for the data on primary and secondary education in 2000; and the 2001 Questionnaire Un@k sur les Indicateurs de Dkveloppement(QUID)survey for the databy province andthe data for higher education.. 4.6 Disparities in gross enrollment ratio across the provinces reflect to some extent the disparities between urban and rural localities. These disparities may be seen using, as an index, the ratio o f the rural to urban gross enrollment ratio. At the primary level, in 1992 the rural gross enrollment ratio was about 79 percent that o f the urban localities. In 2000 it was 85 percent. In this 61 sense, some progress has been made in terms o f narrowing the disparity between rural and urban 10calities.~~ At the secondary level, however, the index of rural to urban gross enrollment ratio fell from 0.49 in 1992 to 0.18 in2000, implyinga significant widening o fthe gap between rural and urban participation rates.54At the tertiary level, where the data pertain only to 2000, the rural gross enrollment ratio was only 6 percent that o fthe urban localities. 4.7 Gender disparities are not particularly large, except at the tertiary level. At the primary level, the gross enrollment ratios for girls and boys are practically the same inboth 1992 and 2000. At the secondary level, there has in fact been a narrowing o f the disparity between boys and girls. Although the gross enrollment ratio for girls was only 0.89 times as highas that o f boys in 1992, the ratio was thus about the same for both groups in2000. Given the comparability o f participation at the primary and secondary levels, it i s somewhat surprising to observe a large gender gap inenrollments inhighereducation. In2000 the gross enrollment ratio for women was only 0.44 times as highas that o f men. Ina later section, we shall explore possible reasons for the sudden emergence o f the gap at this levelo feducationandtheir implications for policy development. 4.8 A more striking disparity is that between income groups. At the primary level, the disparity-as measuredby the index o f the gross enrollment ratio of the poorest 40 percent to that o f richest 20 percent-has narrowed over time: in 1992 the index was 0.71, but in 2000 the index had risen to C1.78.~~ the secondary level, the gross enrollment ratio fell for all income groups but the At decline had been steeper for the lower income groups than for the richest 20 percent. As a result, the gross enrollment ratio among the oorest 40 percent relative to that o fthe richest 20 percent fell from 0.31 in 1992to only 0.10 in2000.P6 Figure4.1 shows that the disparity ingross enrollment ratio at the secondary level i s in fact largest between the top two and the bottom three quintiles. The pattem is, moreover similar betweenthe two subcycleswithin secondary education. 53Inabsoluteterms, however, the disparity has not narrowedby any substantial amount. In1992 the gross enrollment ratio inurban localities was 19percentagepoints higher than inrural localities. In2000, the gap wasjust less than 19percentagepoints. 54Inabsoluteterms, thedisparity betweenrural andurbanlocalities increasedfrom 18percentagepoints in1992to 24percentagepoints in2000. 55Inabsoluteterms, however, the gap ingross enrollment ratio between the poorest 40 percent and the richest 20 percent remained relatively unchanged; around 26 percentagepoints in 1992and 28 percentagepoints in2000. 56In1992the absolute gap ingross enrollment ratio between the richest 20percentandthepoorest 40 percent was 27 percentagepoints in1992. As the gross enrollmentratio has fallen by similar percentagepoints across the income groups,the absolute gap betweenthe top 20 percent andthepoorest 40 percent has remained relatively unchangedin2000. 62 Figure 4.1: Gross enrollment ratiosinlower and upper secondary educationby incomequintiles,Rwanda, 2000 35 sec. 30 25 sec. 20 15 10 5 0 1 2 3 4 5 Poorest Richest quintile quintile Source: Basedon results reportedinKline (2001), using data from the1999-2001EICV survey. 4.9 At the tertiary level, the disparity between income groups is evident. Inparticular, 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 o f the richest 20 percent. The pattern observed in secondary education where participation is dominated by the top two quintiles, thus, appears to have deteriorated. The implication is that access to higher education is even more narrowly confined to the most privileged segments o f Rwandese society. 4.10 That the pattem of enrollment inhigher education is biased infavor of the better off is corroborated by the data in table 4.2, which compare the occupational backgrounds o f the parents of students enrolled at the Universitb Nationale du Rwanda (UNR) and the Kigali Institute of Education (KIE) to that ofthe general adult population aged 35 to 65.57Some students have lost their parents (as the next section will document), but among those with at least one living parent, the data show that students whose parent(s) work as salaried workers (probably inthe modem sector where most salaried jobs are concentrated) are overrepresented. Thus, although 16 percent of Rwandese men aged 35-65 held a salaried job inthe modem sector in2000, the corresponding shares among the fathers o f UNR and KIE students were, respectively, twice and 1.6 times as large. A similar picture holds true among the mothers of the students at the two institutions: the share who worked as salaried employees inthe modern sector was 12percent at UNR, and 16percent at KIE, compared with the share o f 5.4 percent among women aged 35-65 inthe population as a whole. *'The UNRi s the largestpublic higher education institution, accounting for about 56 percent o f students inthe public sector in2000-01. The corresponding share at KIE was 11 percent. Because of time constraints similar data could not be collected from the other institutions in the system. Note that the age range of 35 to 65 was chosen to correspond to the likely age range of the parents of university students. 63 Table4.2: Share ofsalariedworkers amongthe parentsof students intwo publichighereducationinstitutions, Rwanda,2000 Fathersof students Menaged35- Mothers of students Women aged Occupation" Universitk Kigali 65 inthe Universitk 35-65 inthe National du Institute of population National du Kigali Institute of Education population Rwanda Education Rwanda Salariedworkers 32.4 26.5 16.4 11.6 16.0 5.4 Othersb 67.6 73.5 83.6 88.4 84.0 94.6 ITotal 100.0 100.0 100.0 100.0 100.0 100.0 Sample size I 148 I 310 1 2,544 11 172 I 338 I 3,295 a. Excludesstudents whose fathers or mothersare dead. The share of orphans is somewhat lower among university studentsthan among secondary school students. See table 1.3 inchapter 1 for more detailson the incidenceoforphanhoodby level of education. b.Includesfarmers, traders, businesspersons, unpaidfamilyworkers, andothers with unknownoccupations. Source: Figures for the two institutions basedon arandom sampling ofstudent files to which the authors of this study were given access inthe context ofthis study; the occupationaldistributionofthe generalpopulation ofmenandwomen is basedonthe 1999-2001 EICV. Educationalparticipationratesamongorphans 4.11 As indicated in an earlier chapter, one of the most devastating legacies of Rwanda's 1994 genocide i s the extremely high prevalence o f orphanhood. Given the magnitude o f the orphan population, it i s important to document the extent to which orphanhood has compromised the schooling o f children. 4.12 Patterns in urimarv education. The relevant data appear in table 4.3 on the school participation rates among children aged 7 to 12 in 1998-99. As expected, the chances o f being in school are higher among children with bothparents alive than among those who have lost at least one parent-76 percent compared with 72 percent. The data suggest that the gap between orphans and non-orphans i s driven largely by the shortfall inparticipation among girls who have been orphaned. Thus, although the participation rate was less than 1 percentage point lower among orphaned boys than among boys with both parents alive, the corresponding shortfall was more than 5 percentage points among the girls. Closer examination of the data suggests that children who have lost their mothers (whether or not they have lost their fathers) were most at risk o f not attending school. Their participation rate was around 70 percent among the boys and about 66 percent among the girls, comparedwith the corresponding rates o f 77 and 72 percent, respectively, for boys and girls who have lost only their fathers. A further detail from the data is that 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 would be consistent with the possibility that some o fthem worked at ajob and therefore could not find the time to attend school or are simply survivingas 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 primary school, there remain pockets o f the populationwho do not, and most o f the out-of-school population are found among orphans, especially those who have lost their mothers andthose livingapartfrom their biologicalparents. 64 Table 4.3: Percentageofchildrenaged 7-12 enrolledinprimary school by orphanhoodstatus, Rwanda, 1998-99 Status of child Boys Girls Bothsexes Both parentsalive 75.3 75.6 75.5 Livingwith bothparents 77 77 77 Livingwith only one parentor neither 69 73 71 At leastone parent dead 74.6 70.4 72.4 Motherdead; father alive 71 64 67 Fatherdead; mother alive 77 72 74 Bothparentsdead 70 68 69 All childreninsample 75.0 73.5 74.3 Childrennot livingwithbiologicalparent@) 69.8 68.1 68.8 Sample size 1,466 1,548 3,014 Source: Authors' estimatesbasedonthe 2000 RwandaMultiplc idicator Cluster Survey. See appendix table A4.1 for additional information. 4.13 Patterns insecondaryand higher education. Table 4.4 provides additional evidence that orphaned children are more vulnerable to nonparticipation than non-orphans. For this purpose, we compare the share of orphans in the population to the shares of orphans in primary, secondary, and higher education. The data on orphans' shares o f enrollments inprimary and secondary education are basedon a census of the children enrolled inschoolsassociatedwith the Conseil Protestantdu Rwanda (CPR), an umbrella organization o f schools runby various churches that served some 20 percent o f primary schoolchildren and some 16percent o f secondary school students nationwide in2000-01. For highereducation, the data arebasedon a sample o fstudentrecordsmaintained bytwo o fthe country's public institutions. Although patchy, the data nonetheless reveal some interesting patterns. The first confirms the previous finding that primary school participation rates tend to be lower among orphans: although the share of children aged 7-12 who have lost at least one parent was 38 percent in2000, the corresponding share amongchildren enrolled inCPR primary schools was 31percent. 4.14 At the secondary level, however, the pattern is reversed, with the share of orphans rising to 41 percent instead of falling as one might have expected. The anomaly can be explained by the fact that many o f these children have been specifically targeted for assistance under Rwanda's Genocide Fund (Fonds National pour 1'Assistanceaux Rescapbs du Gknocide et des Massacres au Rwanda or FARG). At the tertiary level and assuming orphanhood rates among youths are similar to the orphan rates for children aged 7-12 years, orphans appear to be underrepresentedinthe UNR and KIE. Overall, orphans accountedfor 32 percent of total enrollment inthe UNR and 27 percent inthe KIE. These proportions are also lower than that among students in secondary schools. The lower participation rates among orphans are consistent with the fact that the opportunity cost o f education tends to rise as students progress up the education ladder. Closer examination o f the data 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 appear to be less steep among the double orphans. The pattern would be consistent with differences in the availability o f student aid that gave more support to the double orphans. Although more detailed data are lackingto explore this issue ingreater detail, it seems safe to conclude 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 insteado fbothare not overlooked inthis regard. 65 Table 4.4: Prevalenceof orphanhoodamongprimary, secondary, and higher educationstudents, Rwanda, circa 2000 %with oneparentdead %with %with one Populationgroup both Sample Mother Father parents or both parents deac size I dead ~ Childrenaged7-12,20OOa 4.6 25.8 30.4 7.1 37.6 3,426 Schools associatedwith the Conseil Protestant du Rwanda, 2001b Primarypupils 1 497b Secondary students 6gb Highereducationstudents, 2001` Universite`Nationale du Rwanda 3.9 15.9 19.4 1 13.0 32.4 207 Kigali Instituteof Education 4.5 I 11.6 16.1 10.6 26.6 398 Note: blanksdenotedatanot available. a. Basedonthe 2000RwandaMultiple Indicator ClusterSurvey. b. The ConseilProtestantduRwandais an umbrella organization consistingofschools runby various churchorganizations.In2001, it took a specialcensus of the 497 primary and68 secondaryschools belongingto it; one purpose was to ascertainthe extent of orphanhoodamong the 292,258 primary and26,580 secondary students enrolled inthese schools. The resultswere madeavailable inaggregate form to the authors of this study. c. Basedon information supplied to the authors by officials at the two institutions in the context of this study for a random sample of 207 studentsat the UniversitPNutionaledu Rwanda and398 students at the Kigali Institute ofEducation. Source: Authors' estimatesbasedonsources indicatedinfootnotes above. Disparitiesinstudent flow patterns 4.15 Apart from the disparities inenrollment and participation rates documented above, it is useful to examine differences instudent flow across socioeconomic groups as well. The informationi s relevant for policy development, because it helps to identify the locus of the observed disparities in enrollment. Because o f limitations in the data, we shall examine student flow only in primary schooling. 4.16 Differences by gender. locality, and income. Consider first the data intable 4.5, which shows three indicators of student flow: the percentage of an age cohort ever entering grade 1, the percentage o f the entrants who reach the end o f the primary cycle, and the share o f last year's sixth graders who were still enrolled in the current school year. Consistent with the data presented earlier, the differences between boys and girls are modest: entry rates are comparable, whereas girls are at a slight advantage inreaching the end ofthe cycle and remaining enrolled after reaching grade 6, either to repeat the grade, or to continue on to secondary school. Between urban and rural populations, the differences widen progressively during a child's schooling career: the entry rate to grade 1 among rural children is 93 percent as highas the rate among urban children, but the survival rate to grade six is only77 percent as highand the continuation rate at the end of grade six i s only 66 percent as high. 66 Table 4.5: Selected student flow indicators by gender, locality, and income groups, Rwanda, circa 2000 6 of cohort ever entering %first grade entrants % of last year's 6th graders Group grade la survivingto grade 6b still enrolled this year' % Index % Index % Index By gender Boys 86 1.oo 74 1.oo 68 1.oo Girls 88 1.02 70 0.95 73 1.07 By locality Urban 93 1.oo 91 1.oo 97 1.oo Rural 87 0.93 70 0.77 64 0.66 By income' Richest 20% 95 1.oo 93 1.00 89 1.oo Middle 40% 89 0.93 68 0.74 65 0.73 Poorest 40% 83 0.87 64 0.69 52 0.58 a. Computedas the average ofthe percentagesofchildrenaged 10-13 who haveever enrolled inschool. b.Computedbymultiplyingthe grade-to-gradeprogressionratesaftertaking intoaccount the levelofgrade-specificrepetitionrates. c. Defined inthe same way as intable4.1. See also footnoteb inthat table. Source:Authors' calculation basedon the 1999-2001EICV. 4.17 Differences between children in rural and urban localities reflect, to some extent, differences across income groups. Although only 83 percent of children in the poorest 40 percent of households enter grade 1,95 percent of those inthe richest 20 percent do so. Again, the gap widens as children move to higher grades. Only 64 percent o f children in the poorest 40 percent of households survive to grade 6, compared with 93 percent among those in the richest 20 percent. Furthermore, although only 52 percent of children in the poorest 40 percent of households remain enrolled after attending grade 6 the previous year, the share is 89 percent among children in the richest 20 percent. Figure 4.2 illustrates the disparities in student flow and shows clearly that the gap is biggest between the richest 20 percent of households and the rest of the country. It also shows that, although children from the poorer households lag somewhat behind richer children instarting grade 1,the gap becomes much wider by the end of the primary cycle because o f differential 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 earlier inthe chapter. 67 Figure4.2: Primary schoolenrollmentrates ina cohort of children acrossincome groups, Rwanda, 2000 100I " I Grade 1 Grade 6 Note: the cohort here refers to acompositepopulation whose schoolingcareerreflects the patternof promotion, grade repetition, anddropping over two schoolyears ina cross-sectional sample ofchildren. Source: Basedondata intable 4.3. 4.18 Regression estimates o f the correlates of student flow patterns. Given the importance of locality, income, and orphanhood status in determining children's schooling careers, it is appropriate to examine their relative influence using regression analysis.58 For this purpose, we define four individual-level indicators of student flow among children aged 7-12: whether a child who was not yet inschool the previous year has entered grade 1inthe current school year and, among those already enrolled last year, whether the child has advanced a grade, remained at the same grade, or dropped out inthe current year.59 4.19 The regression estimates are reported in appendix table A4.3 and the results based on them are shown intable 4.6 below. These results are expressed interms of the probability of being in one of the four schooling statuses defined above. For example, among children aged 7-12 who were not in school last year, the probability o f entering grade 1inthe current year was 0.52 for the sample as a whole. It is 0.13 points lower among children who have lost bothparents relative to children with both parents alive, controlling for gender of the child, urban-rural residence, and income group. The bold type on this estimate indicates that the difference is statistically significant at the 5 percent confidence level or better. None of the other estimates is statistically different from zero, suggesting that the single most important predictor of non-entry to grade 1 i s whether or not the child has lost both of its parents. Double orphans are thus handicapped right at the outset of a child's schooling career, inthat many of them do not even get started. 58The advantage of this approach is that it allows us to evaluate the impact o f a single factor, while controlling for differences inother dimensions. 59These new indicatorsof student flow are necessitatedby the fact that the analysis hererelies onindividual 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 andnon-orphans inthe indicators ofstudent flow considered here. 68 Table 4.6: Regression-predicted indicatorsof school progressionduringtwo consecutiveyears ina cohortof childrenaged 7-12 in 1998, by populationgroups, Rwanda, 1998-2000 Probabilityofbeinginthe indicatedschoolingstatus in 1999- 2000 (relativeto reference group, unless otherwiseindicated)b Enteredgrade 1 I Amongthosealready enrolledin 1998-99 (amongthose in1998-99) same grade Droppedout - IWhole sample I - 0.52 I 0.64 I 0.33 0.03 Girls Boys 0.49 0.05 0.00 0.00 0.00 Both parents dead 0.08 -0.13 0.05 -0.09 0.04 Bothparents Motherdead; father alive 0.05 -0.13 0.12 -0.12 -0.01 dive Fatherdead; mother alive 0.27 -0.01 0.03 -0.02 -0.01 Living with only one parent or neither oneC 0.14 -0.07 0.03 -0.03 0.00 Urbanresident Ruralresident 0.81 -0.03 -0.08 0.07 0.01 Poorest40% in Top 20% 0.21 -0.02 income Middle 40% 1 0.42 -0.01 Nofe: a dashdenotes not applicable. Bolded figures refer to estimates for which the underlying regressioncoefficients reported inthe appendix table A4.2 are statistically significant at the 5 percent confidencelevel or better a. Refers to the sample means for those not yet enrolled in 1998-99 among children aged 7-12 years in 1998. The means are almost identical in the population alreadyenrolled inthat year andaretherefore not shownseparately. b. The figures for the "whole sample" line refer to the overall probability of being in the indicated status. The other figures refer to the difference in probabilitybetweenthe indicatedpopulation group and the correspondingreference group. c. Among those with bothparents alive. Source: BasedonregressionestimatesintableA4.3. 4.20 We now consider the other results inthe table for those who were already enrolled last year. The probability o f advancing a grade i s greater by 0.12 points among children whose mothers have died compared with children with both parents still alive; for the other categories o f children the probability o fprogressing to the next grade also exceeds that inthe reference group, but the difference i s not statistically significant. For children who do not advance a grade, the alternatives are either to repeat the grade or drop out. Relative to non-orphans, children who have lost both parents are less likely to repeat andmore likelyto drop out, whereas among those who have lost their mothers (butnot their fathers), the probability of repeating is lower (a corollary of the fact that they are more likely to advance a grade) and that o f dropping out is about the same. For the remaining children, the estimates are not statistically significant relative to the non-orphans. Taken as a whole, the results are consistent with the expectation that when given a chance to go to school, children in highly vulnerable circumstances-especially those who have lost their mothers or both parents-are likely to be more motivated to succeedthan other children. They overcome the enormous odds against them and manage to draw even with or indeed surpass the non-orphans in their progression between grades. Still, the prospects are unforgiving for the most vulnerable children-the double orphans: those who fail to advanceare more likely to leave the systemthan returnfor a secondtry the following year. 4.21 Controlling for orphanhood status, the regression results also confirm the pattems found inthe cross-tabulationspresentedearlier. Children o fboth sexes are equally likely to advance a grade, repeat, or drop out. Incontrast, locality does make a statistically significant difference, holding 69 back rural children by 0.08 points interms of the probability of advancing a grade and elevating their chances o f repetition by 0.07 points. Household income also makes a difference, but only between children from the richest 20 percent of the households and the rest of the population.60 Compared with those in the bottom 40 percent, the chances of advancing a grade are higher by 0.06 points among children from the most privileged homes and those o f dropping out are lower by 0.02 points. Relative to the overall probabilities of advancement, repetition, and dropping out shown inthe table, these gaps are indeed quite large. Distributionof publicspendingon education 4.22 Disparities in enrollment and student flow have implications for the distribution of public spending on education, because only those who attend school benefit. Below we examine the issue from two complementary perspectives. The first captures the incidence of spending associated with the system's structural biases, whereas the second captures the incidence of spending associated with cross-sectional differences inschool participation across income groups. 4.23 The pattern associated with the svstem's 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.61 Children with no schooling receive no benefit from public spending on education, whereas those at the other end who attain higher education accumulate the spending associated with higher education itself as well as that associated withthe previous levels they have already passedthrough. 62 The distribution of spending on education ina cohort thus depends onthe structure of enrollments as well as the distribution of spending across levels of education. The degree of inequity inthe 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 inequity can be quantified by 6oThe income effect is likely to be confounded, however, by our inability inthe regression analysis to control for systematic differences inthe characteristicsoftheadoptive families oforphanedchildrenandthose ofother families. To illustrate, suppose the gross enrollment ratio for a country is 80 percent inprimary education, 30 percent insecondary education, and 10 percent in higher education. In a cohort o f children passing through such a system, the distribution o f eventual school attainment would be as follows: 10percent would have attained higher education, 20 percent (=30-10) would have attained secondary education, 50 percent (=80-30) would have attained primary education, and the remaining 20 percent o f the cohort (=loo-80) would have had no schooling. More refined methods can be used to estimatethe desired distribution (such as 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 it at 100percent. Suppose, for example, that the distribution o f recurrent spending is such that primary education received 55 percent, secondary education received 30 percent, and higher education received the remaining 15 percent. The cumulative spending would be 100 units (=55+30+15) for those attaining higher education, 85 units (=55+30) for those attaining secondary education, 55 units for those attaining primary education, and zero for those who didnot go to school. The distribution o f 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 secondaryeducation, and42percent (=100/(55+85+100) for those attaining higher education. 70 calculating the corresponding Gini-~oefficient.~~For our purpose here, we use an even simpler summary statistic: the share of the resourcesbenefiting the 10percent best educated ina cohort.64 4.24 The results for Rwanda appear in figure 4.3; also shown for comparison are the estimates for a large number of low-income countries for which data from around 1998 are available to make similar calculations. Although cross-country comparisons should be made with caution, because of differences in data quality and coverage and in the structure of the system, the results support two conclusions. The first is that in all countries the best educated always receive more than their population share o f public spending on education; this pattern is to be expected, given that governments typically subsidize all levels of education. The second conclusion is that very large differences exist across countries in the degree o f bias in favor of the better educated. The share of cumulative resources benefiting the 10 percent best educated ina cohort ranges from about 40 percent inSouthAfrica to about 80 percent inNiger. Rwanda is locatednear the top end ofthe spectrum, just behind Niger and Chad among the twenty-eight countries included in the figure. Its position is consistent with what one would expect in a system inwhich higher education enrolls barely 2 percent of the population in the age group and claims nearly 40 percent o f the government's total recurrent spending on education. To anticipate a point to be made later in this section, note that Rwanda's position is only slightly worse than that of Madagascar's, a country with a comparable level of per capita GDP. It means that the structure of enrollments and public subsidization inthese countries are highlycomparable intheir biases infavor ofthe best educated ina cohort.65 63Suppose the numerical examples inthe precedingtwo footnotes pertainto the same country. The distribution o f educational attainment o fthe cohort and their corresponding shares o fcumulative public spendingon education would thus be as follows: 20 percent withno schooling and zero share of the cumulative resources; 50 percent attaining primary schooling and receiving 23 percent o f the resources; 20 percent attaining secondary education and receiving 35 percent o f the resources; and 10 percent attaining higher education and receiving 42 percent o f the resources. The Gini-coefficient associated with this distribution is 0.57 on a scale that ranges by definition from 0 (where eachperson obtains an exactly proportionate share o f resources) to 1.O (where one person gets all the resources). See appendix figure A4.1 for a graphical representationofthis hypothetical example. Incasesinwhichtheshareofthecohortattaininghighereducationislessthan 10percent,thebesteducated10percentwouldinclude some who attained secondary education. The share o f resources benefiting this group would simply be the addition of the share o f those who attained higher education and the proratedshare o fthose who attained secondary education. Incases inwhich the share o f the cohort attaining higher education exceeds 10 percent, the share o f resource benefiting the 10 percent best educated would simply be the proratedshare ofthe resourcesbenefiting the group that attainedhigher education. 6sIn 1998, the gross enrollment ratio inMadagascarwas 107 percent inprimary education, 16percent insecondary education, and 2 percent in higher education; the corresponding percentages in Rwanda in 2000 were 107, 12 and 1.4. The distribution o f recurrent spending on education inMadagascarin 1998 was as follows: 51 percent for primary education, 33 percent for secondary education, and 16percent for higher education. The corresponding share inRwanda in2001was 45,18, and 37 percent. 71 Figure4.3: Share of cumulativepublic spendingon educationbenefitingthe 10percentbest educated in a generation,Rwandaand other African countries,circa 1998 Niger Chad Rwanda M a l i Madagascar Mozambique Sierra Leone Ethiopia Senegal Malawi Mauritania Comoros Burkina Faso Zambia Lesotho Swaziland Eritrea Gambia Cate d'lvoire Guinea Zimbabwe Mauritius Congo Morocco Xamibia South Africa 0 2s so 75 too Percentage share of cumulative spending Note: Data for Rwanda refer to 2001. Source:BREDA, World Bank, and UIS (2002), supplemented by estimates for Rwanda, preparedby the authors using the same computational methods. See text for additional information. 4.25 The pattern associated with social selectivity ineducation. A second perspective o n the distribution of public spending o n education is the extent to which it favors the wealthier groups in society. Benefit-incidence analysis is a common tool for documenting 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 s ending per student to calculate the share of resources benefiting households in each income group.6'As w e have seen earlier, wealthier groups in Rwanda become increasingly overrepresented as the level of education rises-a pattern found inpractically all countries. It is thus no surprise that a typical result of benefit-incidence is that, although public spending on primary education may benefit lower-income groups more than it does 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 o f education usually shows the wealthier groups receiving more than their proportionate share o f public spending on education. The real issue is one of degree. 4.26 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. They show that, although the poorest 20 percent of households inRwanda appropriated 15 percent ofpublic spending o n education, the richest 20 percent of the population received a share of 28 percent. Judgingby the relative sizes of these shares, Rwanda stands somewhere inthe middle among the sample of Sub-Saharan African countries shown in the table. Inother words, the country's public sector spending on education is neither as equitable across income groups as it can be nor is it as inequitable as it is inother countries. 66Formore details onthe method ofcalculation, see Demery 2000. 72 Table 4.7: Shareof public spendingoneducationbenefitingthe poorest and richestpopulationquintiles,Rwanda and other countries,1990s and 2000 Poorest Richest Richest quintile's share Country Year of survey quintile quintile as a multiple of the (%> ("w poorest quintile's share Rwandaa 2000 15 28 1.9 Other African countries South Africa 1993 21.1 23.4 1.1 Kenya 1992-93 16.7 20.7 1.2 Ghana 1992 16.4 20.8 1.3 Malawi 1994-95 16.0 25.0 1.6 Uganda 1992 13.0 32.0 2.5 CGte 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. Datarefer to the distributionofspendingonprimaryandsecondaryeducationonly. Source: Kline 2000 @sedon the 2000 EICV) for Rwanda; Goa. of Madagascar 2000 for Madagascar; and World Bank EdStats database for all other countriesbasedonvarioushouseholdsurveys onlivingstandards. 4.27 The results for Rwanda compared with those for Madagascar are particularly interesting inlight of the finding highlighted earlier regarding the similarity of the structural biases in bothsystems infavor o f the best educated ina cohort. Yet, the results inthe foregoing table indicate that Rwanda has a more equitable distribution of educational spending across income groups than Madagascar: the richest quintile appropriates a share nearly twice as large as the share received by the poorest quintile, compared with more than five times as large in Madagascar. The more favorable results inRwanda are consistent with the country's smaller gaps ineducational access across income groups: among first graders from the poorest 40 percent of households, 69 percent in Rwanda reach the end of the primary cycle, compared with only 9 percent inMadagascar (see table 4.5 above for the pattern inRwanda and World Bank [2002] for the pattern inMadagascar). 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. Policyimplications 4.28 In developing appropriate policies to address socioeconomic disparities in education, answers are needed to at least three generic questions: Who are the disadvantaged? Where in the education system might the government intervene to narrow the disparities? What might be the most effective interventions? The data presented in this chapter provide some insights on these questions, especially with regard to the first two. 4.29 Narrowing; the disparities inprimary education. The gaps across socioeconomic groups inRwandaare not as wide as those observed inother 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 inthe poorest 40 percent o f households are 78 percent as likely to attend as those inthe top 20 percent o f households. Entry rates to grade one are generally high, except among double orphans, so the pattern of underrepresentation stems mainly from differences in survival rates to the end of the cycle. The populations falling furthest behind in this regard include double orphans, children in rural areas, and children from the poorest 40 percent o f households. Double orphans-arguably the 73 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 children is one possibility, but the Genocide Fund-the program enacted by law in 1998 to assist children orphaned by the genocid-urrently benefits only students in secondary and higher education. The government also provides assistance to support the education o f vulnerable children through the Ministsre de I'Administration Local et des AHaires Sociales (Minaloc), but the amount o f funding is relatively modest at about 20 percent of the resources available through the Genocide Fund. 4.30 Financial assistance may also be what is needed to boost survival rates among children inthe other lagging groups, particularly those from poor families. Inruralareas, raising survival rates may require additional interventions on the supply side, including ensuring that all primary schools offer the full six grades of instruction in the primary cycle, reducing distances that children have to travel to reach their schools, and, most important, improving teaching and learning to reduce grade repetition-and hence dropping out-among the children.67 The relative impact of these promising options has not been possible to evaluate inthe context o f this study; it clearly needs to be explored, including throughpilot projects, as the next step inpolicy development. 4.3 1 NarrowinP the disparities in secondarv and higher education. As we have seen, the socioeconomic disparities at these levels begin to widen substantially. For girls, the disadvantage in access emerges only in higher education, but for children in rural areas and in poor households, the barriers to access are already obvious inthe first cycle o f secondary education. The disparities across localities also begin to appear in secondary education; Gisenyi and Gikongoro lag more than other provinces. With regard to orphans, the pattern suggests that such youths are somewhat less likely to participate inhigher education than non-orphans, but are at least as likely as non-orphans to enroll in secondary education, the level o f education that currently benefits the most from the Genocide Fund. Targeting financial aid to assist the most vulnerable youths thus appears to be working. Additional interventions would be needed, however, to improve educational access in rural areas, in selected provinces and among the poorest segments of Rwandese society. 4.32 Where in the education system might interventions be needed to narrow the observed disparities in post-primary education, and of what might the interventions consist? Although a full answer is beyondthe scope of this study to provide, it is instructive to consider the specific situation of females' lagging participation in higher education to illustrate the likely importance of improving academic performance inthe affected population groups. 4.33 Consider table 4.8, which shows the scores of males and females in national examinations administered at the end of the primary, tronc commun @e., lower secondary), and upper secondary cycles and the scores of applicants to the public institutions of higher education. The startling pattern is that girls consistently 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 takes its toll in the competition for the highly coveted places inpublic higher education. Girls currently account for about a quarter of the students in higher educational institutions in the public sector, whereas they account for half the students in the final year of upper secondary education. These results imply that policies to improve women's representation in higher education must do more than simply reserving a quota for them. Understanding why girls' academic performance lags behind boys', beginning inthe primary grades, is a critical first step toward developing sound policies to improve ina meaningful way their access at 67About 14percent o fRwanda's primary schools currently do not offer all six grades o finstruction inthe primary cycle (see table 2.10). About halfof all rural householdslive more than 30 minutes away from a primary school (see table 5.4). 74 the top end o f the educational ladder. Although data are lacking to examine the reasons underlying the lagging participation among other population groups, poor academic performance i s also a likely root cause o f their problems. Accordingly, enhancing leaming outcomes among the affected populations- rural children and those from poor families-sannot be ignored as part o f the strategy to diminish socioeconomic disparities inpost-primary education. Female Male Primary (% exceedingcutoff markfor promotion to the next level) " 17.8 28.5 Tronc commun (% exceedingcutoff mark for promotion to next level)" 29.3 55.6 Upper secondary(%exceedingpassmark for graduation)" 62.8 76.0 Higher education(GPA ofapplicantsto public institutions) 3.4 3.6 UniversitPNationale du Rwanda (LJNR) 3.5 3.7 Kigali Institute of ScienceandTechnology (KIST) 3.2 3.6 Kigali Institute ofEducation (KIE) 3.2 3.5 Kigali Health Institute (KHI) 3.6 3.9 Znstitut Supirieur d 'Agronomie et d'Elevage (ISAE) 2.9 3.8 4.34 Reducing overall inequities in spending allocations. The socioeconomic disparities in enrollment across income groups inevitably show up in the results o f benefit-incidence analysis. In Rwanda, the richest 20 percent o f households gamer nearly twice as large a share o f the public spending on education, as do the poorest 20 percent. Although there i s obviously still scope for increasing the benefit to the poor, cross-country comparison suggests that Rwanda does not have the most inequitable distribution o f spending across income groups. Far more inequitable is the country's overall structure o f spending, where primary education receives only 45 percent o f the government's recurrent spending on education, whereas higher education gets a share o f nearly 40 percent. Redirecting spending infavor o f primary education would go a long way toward reducing the current bias in the system toward those who manage to climb to the top o f 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. Appropriate policies in higher education student finance and cost management are particularly critical in this regard; these are discussed in greater detail in chapter 7. Conclusion 4.35 Educational access in Rwanda is relatively equitable in primary education, but disparities across population group widen dramatically at the post-primary levels. Orphans are among the most vulnerable population groups, and the country has admirably institutionalized financial assistance to them through the Genocide Fund. More can be done, however, to reach pockets o f children still at risk o f lagging behind, particularly double orphans and children from rural areas and the poorest families. These priorities notwithstanding, success in improving overall equity in the system will require correcting the current imbalance inpublic spending that so obviously favors higher education at the expense o fprimary education. 75 Chapter 5: Service Delivery in Primary Education 5.1 InRwandaas inmost other low-income countries,primary educationenrolls the largest number of students andabsorbs a major share ofpublic spending on education. Becauseof its central role in basic human capital formation, the system's performance attracts keen attention from policymakers as well as the public. What is the availability o f services? How well are existing resources deployed to provide services? How efficiently are the resources transformed into learning outcomes? What i s the scope for improvement, and what trade-offs do they imply for policy development? Although data constraints limit the extent to which each o f the questions can be answered, the available information presented in this chapter nonetheless helps shed light on the nature o f some of the emerging challenges. Overview of the supply of services 5.2 We begin by describing a few salient features o f the country's network o f primary schools. Incontrast to earlier chapters, we treat schools as the relevant unito f analysis here.68 5.3 The supply infrastructure. There are currently slightly more than 2,000 primary schools inRwanda serving some 1.5 million children (see table 5.1). The private sector accounts for only 1.5 percent of the schools and 0.7 percent o f the enrollments. The public sector consists o f two types o f schools: state and libre subsidik. State schools are funded and managed directly by the government, whereas libre subsidik are funded bythe government, butmnbynongovernmental organizations under two main umbrella groups-the Secrktariat Nationale de 1'EnseignementCatholique (SNEC) and the Conseil Protestant de Rwanda (CPR). More than 71 percent of the schools fall in the libre subsidik category, and among such schools, about two-thirds are under SNEC management. Public sectora Private sector Overallb State Libre subsidie` Schools 27.1 71.4 1.5 100 (2,093) Pupils 28.0 69.4 0.7 100 (1,475,572) 5.4 There i s wide diversity across provinces in the institutional composition of the supply infrastructure (see figure 5.1). Almost all of the country's private schools are located inKigali Ville, whereas none are found in Cyangugu and Kibuye; as a result, private schools account for nearly 20 percent o f the schools in Kigali Ville, compared with the national average o f less than 1.5 percent. Provinces also vary in the relative shares o f state and libre subsidik schools. In Gikongoro, state schools account for less than 2 percent o f the schools, whereas in Umutara, their share exceeds 65 68Unless otherwise indicated, the averages presented in most of the tables inthis chapter thus refer to unweighted, rather than weighted values. 76 percent. Within the libre subsidik sector, the share o f schools under SNEC management ranges from 81percent inButare to 39 percent inUmutara. Figure5.1 Institutionalcomposition ofprimaryschools by province, Rwanda, 2002 li %private ,27% state ,71 % libre subsidii /34%CPR Rwanda Gikongoro CY~gUgU Kibuye Gitarama Gisenyi Rukngeri Butare Kibungo Byumba Kigali Ville KigaliRural 100 80 60 40 20 0 0 20 40 60 80 100 %distributionofall schools %distributionoflibresubsidik schools Source: MINEDUC's 1999-2000 census of primary schools, supplementedby data from the ConseilProtestant de Rwanda (CPR) andthe Sicretariat Nationale de I'EnseignementCatholique (SNEC). 5.5 Selected characteristics of the schools and teachers. Table 5.2 summarizes some features of the supply of primary schooling in Rwanda today. There are very few private primary schools in Rwanda, and they are strikingly different from the public schools. Their average enrollments are only about halfthat of schools inthe public sector; they operate out of better premises; they generally run single shifts in all grades, unlike public schools where the first three grades generally operate on double shift; and they have muchsmaller ratios ofpupilto teachers. 5.6 Inthe public sector, state and libre subsidik schools are highly comparable, although state schools tend to be slightly bigger, with 720 pupils o n average compared with the average of 678 pupils for libre subsidik schools. Almost all first grade classes, whether in the state or libre subsidik sector, are taught by teachers with double shift duties. By third grade, the share declines to about 84 percent inthe state sector and 88 percent inthe libre subsidik sector. By administrative arrangement, there is no double shifting beyond third grade.69 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 subsidik schools are better endowed than their state counterparts in this regard, but the advantage is hardly dramatic. In terms of nonteaching staff (ie., school heads without teaching 69Under double shifting, children attendschool either inthe morning or afternoon, typically for about 2.5 to 4.0 hours, dependingon the shift. Insome schools, children change shift on alternate days, whereas inother schools they do so on alternate weeks. 77 duties and secretarial staff) libre subsidib schools are slightly worse off, with 41 percent of schools thus endowed, compared with 44 percent in the state sector; the share of private schools with staff performing purely administrative tasks fall to 16 percent. In schools with administrative staff, the staffing ratios appear relatively comparable across all three types of schools. Table 5.2: Characteristicsof state, libresubsidii, and private primary schools, Rwanda, 2000 Public sector Private All schools State I Libresubsidie` I Average number o f pupils per school 720 678 342 684 Percent with classrooms inpoor condition 47.9 54.8 19.9 52.3 Percent with teachers teaching 2 shifts o f pupilsa 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 56.9 II (i;::)II 57.3 (55.2) Percent with nonteaching staff` I 44.5 I 41.0 16.1 41.6 Pupilinonteaching staff ratiod 725.7 710.5 708.6 714.8 a. The systemdoesnot operateon doubleshift beyondgrade 3. b. Figuresinparenthesesreferto the weightedaverages. c. Counting only schoolheadswithout teachingduties, as well as school secretaries. d. Refersto unweightedratios inschools with administrative staff. Source: For data on the first four indicators, authors' calculationsbased on MINEDUC's 1999-2000census ofprimary schools; for data on the lasttwo indicators, authors' estimatesbasedon MINEDUC's 1999census ofteachers. 78 Figure5.2: Pupil-teacher ratiosinpublicprimaryschools inRwandaandother Africancountries,circa 1999 Chad 72 Ethiopia Mali Rwanda Burundi Senegal Cameroon Mozambique Madagascar Zambia Guinea Burkina Cdte d'lvoire Tanzania Nigeria Niger Ghana Kenya 0 20 40 60 80 Ratio Source: For Rwanda, table 5.2; for the other countries, WorldBank 2002. 5.7 Consider now the profile of teachers in the various types o f schools (see table 5.3). In private schools, about half the teachers are women-comparable to the share in state schools and slightly less than the share inlibre subsidik schools. Teachers inprivate schools are the most seasoned, with an average o f 8.5 years of experience compared with an average of about 8 years inthe other two sectors. They are also the best educated: 77 percent have some sort of upper secondary education, and 47 percent have the best credential at this level of education (i.e., D6 or D7). Incontrast, the share of teachers with a D6 or D7 qualification is 33 percent in state schools, and 36 percent in Zibre subsidik schools. A significantly higher share of private school teachers have received pre-service teacher training: 67 percent compared with 50 and 54 percent, respectively, instate and Zibre subsidik schools. 79 Table 5.3: Characteristicsof primary school teachers, Rwanda, 1999 Public sector Teacher credential` Private All State Libre schools schoolsb subsidie` 'ercent women - 50.3 54.6 50.0 54.0 iverage years of experience - 8.1 8.0 8.5 8.1 listribution by educational attainment (%)" Primarv CA 1.4 1.3 1.4 1.3 Lower secondary 27.5 23.8 21.6 24.9 General (1-3 years) ES 1,2, 3 9.1 7.3 5.4 7.8 Vocational diploma holder CERA1etc. 15.3 12.6 13.5 13.4 Teacher training diploma holder EAP, EMA, etc. 3.1 3.9 2.7 3.7 Utmer secondary 70.7 74.6 77.1 73.4 Incomplete ES4,5,6 21.0 20.5 17.6 20.5 Diplomaholder (3-5 years)b D3, D4, D5 16.8 18.3 12.2 17.9 Diplomaholder (6-7 years)b 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) 'ercent with pre-service teacher training - 49.6 54.4 66.5 53.2 Note: -denotes "not applicable." a. See appendix table A5.1 for more details onteachers' educationalqualifications. b. Diploma holderswith 3-5 years of secondary schoolingreferto those inqualificationcategory D3-D5, whereas those with 6-7 years refer to those in categoryD6-D7. Teachers inthe former category arethose who receivedtheir diplomas beforethe systemwas reformed during 1982-92. c. See appendix table A5.1 for more detailedinformationon the variouscredentials. d. Includesteachers with missingdataonthetype ofschoolwhere they teach. Source: Authors' calculationsbasedonMINEDUC's 1999-2000census ofschools. 5.8 Accessibility of schools and client feedback. To round out the discussion on the supply of services, we now consider the perspective of the service users. The 2001 Questionnaire UniJiksur les Indicateurs de Dheloppement (QUID) survey of some 5,800 households provides some clues regarding the accessibility o f 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 to pupils' homes, given the usually young ages o f the children. According to QUIDdata, only about half the sampled households live within 30 minutes of a primary school, but inurban areas, the share rises to more than 80 percent. Across provinces, accessibility is best inKigali Ville, Ruhengeri, and Butare-between 58 and 74 percent of the households live within 30 minutes of a primary school-and worst inGitarama, Kibungo, and Cyangugu, where the corresponding share is around 45 percent. These results provide a first hint of a possible need to improve the accessibility o f primary schools, particularly inthe lagging provinces. W e return to this issue inmore detail later inthis chapter, inthe context of evidence on the economies of scale inservice provision. 5.9 With regard to pupils' perceptions, 49 percent o f the surveyed pupils reported no problems with their schooling and, of those who reported a problem, the overwhelming majority identified lack of books and school supplies as the main source o f their frustrations. Surprisingly, the share of satisfied pupils is much greater in rural than in urban areas-5 1percent compared with 31 percent. Similarly, 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. One possible explanation for these patterns is that respondents use different implicit standards as benchmarks, reflecting differences in 80 their contact with and knowledge about alternatives to the schooling they actually receive. As such, we would expect urban children and those inKigaliVille to be more demanding intheir expectations and thus, to be less satisfied with the services they infact receive. 5.10 With regard to the types of problems encountered, 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 the respondents are likely to know much about teaching quality and would thus hardly be in a position to identify it as a source of problem, particularly because for many o f the pupils, the school they currently attend is likely the only school they have ever attended. Table 5.4: Accessibility of primary schools and problems with schoolingreported by currently enrolledpupils, Rwanda, 2001 Householdswithin 30 minutesof aprimary Pupilsreportingno Type ofproblemencounteredbythose reportinga school problem problem(%) Locality I Index Index Lack of Buildings ir Percent Percent (Urban & books& 'oor quality Lack of bad I Butare =loo) Butare=lo0 supplies )fteaching teachers condition I RWANDA 53 - 49 91 3 3 7 Urban 83 100 32 100 92 3 2 7 Rural 51 62 51 161 83 3 4 6 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 Kigali Ville 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 Note: -denotes "not applicable." Source; 2001 QUIDhouseholdsurvey. Teacher allocation across schools 5.11 Moving beyond the overall characteristics of the supply infrastructure, we now examine in greater detail various aspects of service delivery, beginning with the allocation of teachers across schools. Because teachers embody the bulk o f resources that schools, particularly at the lower levels, receive to deliver services, teacher allocation across schools constitutes an important management issue in this regard. H o w consistent and equitable i s the allocation of teachers across schools? How does the allocation differ across state and libre subsidik schools? 81 5.12 Overview o f uatterns in the state and libre subsidid sectors. Figure 5.3 shows the pattern based on data for nearly 2,000 o f the country's primary schools; data for private schools are missingand therefore not shown. Among both state and libre subsidib schools, the two panels of the figure show the characteristic positive relationbetween school enrollments and the number o f teachers allocated to a school. This pattern is to be expected, but closer examination o f the figures suggests that wide variation across schools nonetheless remains in allocation o f teachers. For example, a school enrolling 800 pupils, whether inthe state or libre subsidid sector, may be endowed with as few as ten teachers, or as many as twenty. The patterns revealed by the graphs raise other questions: how comparable infact are schools inthe two sectors in terms o f the application o f rules affecting teacher allocation? To what extent are the disparities across schools the result o f disparities at the regional level and to what extent the result of disparities within each region? To examine these questions, we turnbelowto regressionanalysisofthe data.70 'OA simpleway to understandthe approachis to think ofdrawinga line with the bestfit linethroughthe observationsplottedinthe two panelsinfigure 5.3. 82 Figure5.3: Relationbetweennumberof pupilsand number of teachers across primaryschools inthe public sector, Rwanda, 2000 State schools Libre subsidik schools 40 40 0 0 0 0 0 O m 30 0 30 K? : c 0 20 0 00 0 0 0 50 300 500 800 1100 1500 1900 50I I I I I I I 300 500 800 1100 1500 1900 Number ofpupils Number o fpupils Source: MINEDUC's 1999-2000census o fprimaryschools. 5.13 Regression analysis of teacher allocation across schools. The relevant results appear in table 5.5. The coefficient on the number o f pupils is estimated at 0.16 for both the state and Zibre subsidik schools, implying that, on average, a school with 100 more pupils than a school of average size wouldhave 1.6 extra teachers; thus, a new teacher i s added for every 62.5 new pupils enrolled. Table 5.5: Regressionestimates of the relationbetweennumber of teachers andpupils across public primary schools, Rwanda, 2000 State schools Number of observations a. Seeappendixtable A5.3 for full regressionresults. Note: All coefficientestimates on"number ofpupils" are statisticallysignificant at the 1percentlevel. Source: Authors' estimatesbasedonMINEDUC's 1999-2000 census ofprimaryschools. 5.14 Consider next the R2values shown inthe table. The R2statistic is a summary indicator that captures the explanatory power o f the regression equations. It ranges in value from 0, which implies that the allocation of teachers across school 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 83 enrollments across schools. Again, the results for both samples o f schools are comparable, with an R2 value of 0.79 for the state schools, 0.77 for the libre subsidik sample, and 0.78 for the combined samples. Ininternational comparisons, the latter value places Rwanda inthe middle range of R2values that characterize teacher allocation in a sample of twelve African countries for which similar regression analysis has been completed (see figure 5.4): some countries, such as Guinea, achieve better results-perhaps providinga target to aim for-whereas other countries, such as Togo, do much worse. 5.15 Returning again to the results in table 5.5, note that because both the coefficient estimate on number of pupils enrolled and the R2values are comparable inthe regressions for the state and libre subsidik samples, we can conclude that teacher allocation rules are actually highly comparable inthe two sectors. What about the pattern of allocation across regions? How important are the gaps in allocation at the regional and subregional levels? To answer these questions, we repeated the regressions with regional dummy variables included inthe specification; the full regression results appear in appendix table A5.3, and to economize on space, only the R2values of the new regressions are reported in table 5.5 above. As can be seen, these values are only slightly higher than those associated with the regressions without the regional dummy variables: in the regressions using the combined samples o f state and libre subsidik schools, the R2statistic rises invalue from 0.78 to only 0.8 1.The implication i s that differences across regions inteacher allocation contribute only marginally to the observed disparities inteacher allocation across Rwandese primary schools. A corollary is that these disparities largely result from inconsistent application of teacher allocation policies within each region. 84 Figure 5.4: R2values of regressionsrelatingnumber ofteachers to pupilsacross schools, Rwandaandother African countries,circa2000 1.o 0.8 0.6 0.4 0.2 Note: The closer the R2 value to 1.0, the tighter is the relation between school size and the number o f teachers allocated to the school. Source: Regression analysis o f data from each country's annual census of schools. 5.16 Intraprovince patterns o f teacher allocation. Given the importance o f within-province variation in teacher allocation, we examine below additional results based on region-specific analysis o f the data (table 5.6). Specifically, separate regressions were runfor each region usingthe combined samples of state and Zibre subsidib schools.71As a reference, the results for the whole country are also shown (top row). "Combiningthetwosampleshelpsavoidsmallsamplesizeproblemsintheregressionanalysis.BecausethestateandZibresubsidit! schools are highly similar interms of the pattemo fteacher allocation, this approachposes few problems. 85 Table 5.6: Regressionestimates ofthe relationbetweennumbersofteachers and pupils acrosspublic schools by province, Rwanda, 2000 Regressionresults andrelated information" Simulations for a school with 700 pupils _ _ Coefficient estimate on Constant Number o f R~ Number of Pupil- number o f pupilsa schools (N) teachers teacher ratio RWANDA 0.016 1.52 55.0 Butare 0.016 1.88 53.0 Byumba 0.015 2.00 56.9 cyangugu 0.018 0.43 53.7 Gikongoro 0.017 2.21 50.4 Gisenyi 0.011 3.74 59.7 Gitarama" 0.016 1.27 55.8 Kibungo 0.015 1.36 58.7 Kibuye 0.015 0.62 62.2 Kigali Rural 0.017 0.80 56.7 Kigali Ville 0.019 4.21 40.6 Ruhengeri 0.017 1.49 52.6 Umutara 0.015 2.62 53.4 a. With the exception three provinces, the due ofthe remi m for state schools is tyuicallv comuarableto that for the corresuonding sample of libre iubsidih schoois. InGisenyi, Gitarama and Kibuye, the R2for state school samples was'0.52, 0.58 and0.97, compared wiG -. . the correspondingvalues of0.64,0.93 and0.80 for the libre subsidihsamples. b.All the coefficients are statistically significant atthe 1%confidencelevel. Source: Authors' estimatesbasedonMINEDUC's 1999-2000census ofprimary schools. 5.17 Consider first the coefficient estimate o n the number of pupils. The estimate ranges from a low of 0.11for Gisenyi to a high of 0.19 for Kigali Ville. These results mean that the overall pattern inthe relationbetween enrollments and teacher allocation is such that, for schools inGisenyi, a new teacher is added, o n average, for every 90.1 new pupils enrolled, whereas in Kigali Ville, the corresponding figure is 52.6 pupils. Substantial variation thus exists across regions inthe sensitivity of teacher allocation to the changes inthe size of enrollments. 5.18 Turn next to the values of the R2statistic for each regression. Recall that this statistic captures the tightness o f the relation between enrollments and teacher allocation (i.e., 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, butthe region-based regressions yield values ranging from only 0.62 and 0.64, respectively, inGisenyi and Gikongoro to 0.89 and 0.90, respectively, inCyangugu and Kigali Ville. In addition to differences across regions in terms of the slope of the relation between enrollments and teacher allocation, differences thus also exist inthe consistency o f teacher allocation. Gisenyi stands out among the regions, because teacher allocation 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 elsewhere inthe country. 5.19 To understand the implications of the regression results better, consider the last two columns inthe table showing simulations of teacher allocations based on the regression estimates. The countrywide regression implies that a school enrolling 700 pupils (which is close to the average size of primary schools inthe country) would on average receive an allocation of 12.7 teachers, which implies 86 a corresponding ratio of 55.0 pupils per teacher. The same calculation based on the regressions for each region shows teacher allocations ranging from 11.7 inGisenyi to 17.2 inKigali Ville and pupil- teacher ratios ranging from around 60 (Kibuye, Gisenyi, and Kibungo) to 40.6 inKigali Ville. Inonly Kigali Ville, thus, does the level o f teacher allocation appear adequate for schools to be able to assign teachers to only a single shift o f pupilsthroughout all six grades o f the cycle. Inthe other regions, the compromise solution is to assign two shifts per teacher in first three grades, whereas in the upper grades, teachers are assignedto single shifts o fpupils. 5.20 In summary, what do the foregoing results, taken as a whole, imply in terms of the pattern o f teacher allocation across public primary schools in Rwanda? Overall, the complement o f teachers is relatively meager, as suggested by the fact that the country's average pupil-teacher ratio o f 57 i s among highest inthe developing world. Assigningteachers double-shift duty inthe early grades helps to reduce inevitable overcrowding inthe classroom implied by the highpupil-teacher ratio, but the arrangementalso meansthat pupilsreceive relatively few hours of instruction inthe school day. In other countries with highpupil-teacher ratios, the compromise is brokeredby tolerating overcrowding inthe classroom. Neither compromise appears to work, however, as suggestedby the generally high rates o f grade repetition where pupil-teacher ratios are high.72 5.21 A second result pertains to the pattern of allocation across schools. In Rwanda, the allocation appears to be reasonably consistent in relation to enrollments across schools, but scope for improvement remains, particularly in such provinces as Gisenyi and Gikongoro and to a lesser extent inButare, Kibungo, Ruhengeri, andUmutara. Inagrowingsystemsuchas Rwanda's-where boththe increase in enrollments and the argument for lowering the pupil-teacher ratios imply that teacher recruitment is likely to grow-the consistency in the allocation o f teachers across schools can be improvedrelatively painlessly at this stage: it only requires that as new teachers are recruited they be explicitly and unfailingly assigned to schools with highpupil-teacher ratios. Missing this opportunity would obviously make the task harder in the future as the pace o f teacher recruitment slows down, becauseimprovements would then increasingly call for the reassignmento fincumbent teachers across school, a difficult task inthe best o fcircumstances. Economies of scale in service delivery 5.22 We now consider arelated, butdifferent issue, namely, how the cost of service delivery varies across schools o f different sizes. The issue i s important, because o f its policy implications for the size and spatial distribution o f schools. 5.23 Overall pattern inthe relation between size o f enrollments and unit costs. First, data for the analysis were compiled from two datasets with the requisite information at the school level, the 1999 census of teachers, and the 1999-2000 census o f schools. Using three variables common to the two datasets-region, district, and school name-1,774 public schools were successfully merged, yielding a sufficiently large number o f observations for the statistical analysis. Inthe teacher census, informationwas available on the educational qualification o fteachersandother school personnel, their years o f experience and current duties as well as other personal and professional information. By combining this informationwith the data on salary levels and structure shown inappendix table A5.1, itwas possibleto estimatethe aggregatecost ofpersonnel for eachschool andthus to compute the cost o f personnel per pupil in each o f the 1,774 schools. Although personnel costs are only a part of the total cost o f service delivery, they represent the bulk o f spending and therefore provide a good proxy for overall spending. '*For example, the pupil-teacher ratios insuchcountries such as Benin, Cameroon, Mozambique, and Madagascarrange between54 and 65 pupilsper teacher andthe repetition raterangesbetween 24 and 26percent. 87 5.24 Figure 5.5 below shows simple scatter plots o f enrollments and unit costs. The pattern reveals more diversity across schools than one would expect based only on the distribution o f teachers documented earlier. It implies that in addition to teacher numbers, differences in the profile o f the teachers and other school personnel-in terms o f educational qualification and seniority-tend to widen rather than narrow the gap infundingacross schools, through the systematic linkbetweenthese variables and the salary bill. Figure 5.5: Relationbetweenpersonnelcost per pupil and schoolsue, Rwanda, 1999 S t a " l S fibre subsidi6schools 10.m- 0 0 0 1Po0 v 1,m 50I , 0 , 250 1 Mo 75b I,& lh 1,Mb 50I 250 500 I 750I 1,ooO I 1250 I 1,500 Nunixrofplpils N& ofpupils 5.25 Regression analvsis o f the relation between unit costs and school size. The relevant results appear in table 5.7. They confirm the visual pattern in the graphs above: that the relation between cost per pupil and school size i s weaker than that between number of teachers and school size: the R2value falls to 0.70 for the combined sample o f state and Zibre subsidik schools from 0.78 in the corresponding regression reported earlier with number o f teachers instead o f unit costs as the dependent variable. 5.26 A more important result in the table relates to the economies o f scale in service delivery. The estimatedcoefficient on the number o f pupils implies that a 1percentage point increase inthe personnelcostsperpupilallows enrollments to expandbyabout between 1.04 and 1.05 percent. The implication i s that unit costs would fall as schools increase the size o f their enrollments, all else being the same. 88 Average number of pupilsper school 720 678 684 5.27 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 (ie., based on the combined state and Zibre subsidie' samples), we note that the predicted per pupilcost of school personnel is only 2 percent higher ina school enrolling 400 pupils relative to that ina school with 700pupils (FRw 5,781 comparedwithFRw5,641). As enrollments fall to 100pupils, the predictedunit cost rises to FRw 6,145, an increase o f 6 percent, which is still a modest increase. Figure5.6: Economiesof scale inpublic primaryeducation, Rwanda,2000 6,500 1 I \ I ,Libre subsidik I 2 State v 2 6,000 1 d .r( 9 a - 1 +I 5,500 u8 5,000 1 I , I , I I I 0 100 200 300 400 500 600 700 800 900 1,000 Numberofpupils Source: Simulatedfromregressions intable 5.7. 5.28 Implications for policies on the size distribution of primarv schools. The government as service provider has the option of delivering school services through a network of relatively few schools, each enrolling a large number of pupils, or alternatively, througha larger number o f relatively 89 small schools. Inthe Rwanda context, it would appear that the latter option would make more sense given the very limited economies o f scale inservice delivery documented above. Large schools are not much cheaper to run, but they are by necessity farther from the average pupil's home. To recall, the 2001QUIDhousehold survey indicates that children inabout half of the survey households take more than 30 minutes to reach the nearest primary school. Delivering services through smaller schools located nearer to pupil's home would help reduce the physical barriers to schools, as well as to reduce the opportunitycost of school attendance. 5.29 Yet, the current network of public schools tends to emphasize size over proximity. The average school enrolls about 720 pupils inthe state sector, and 678 pupils inthe Zibre subsidie! sector, for a combined (weighted) average of 684 pupils per school. In 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 across the public and private sectors. Among the state schools, less than a quarter of the schools enroll fewer than 450 pupils each, about the same share as schools enrolling more than 900 pupils. Inthe Zibre subsidi6 sector, the emphasis on size is less pronounced, butremains significant. The real contrast is with the schools in the private sector, where more than two-thirds of the schools enroll fewer than 300 pupils each. Table 5.8: Percentagedistributionof primaryschoolsand cost per pupilby sue of enrollments, Rwanda,2000 Size of Cost per pupilinthe enrollments public sector Public sector Private schools State Libre subsidie 4 5 0 3.2 1.8 29.0 150-300 6.5 7.2 35.5 300-450 13.4 17.8 6.5 450-600 19.3 20.9 12.9 600-750 21.1 18.5 6.5 750-900 11.4 11.9 0.0 >goo 25.1 21.8 9.7 Vote: Columns sum to 100percent. Percentages are basedon data for 2,043 of the total of 2,093 primary schools `or which relevantdataare available. 1.Costs are simulatedfrom the regressions estimatesintable 5.7 for schoolswhose enrollmentscorrespondto the niddle of each size bracket. For the first categoly, cost corresponds to a school enrolling 100pupils, whereas for helastcategory, to oneenrolling 1,000 pupils. Source: Authors' estimatebasedonMINEDUC's 1999-2000 censusofprimary schools. Student learning 5.30 So far, we have addressed aspects of service delivery that affect the resource flow to each school. W e now move beyond resource allocation to examine the relation between funding and learning outcomes. How strong is the relation between these variables? How different are schools from each other in their ability to transform funding into student learning? What does the pattern implyfor policy development to improvethe effectiveness o f service delivery? 5.3 1 Evaluating these issues is not easy in the best of circumstances. In the Rwandese context, the challenges are magnified by data constraints. The country has only recently started to participate in an international study on student learning (to be verified) and neither the data nor the results are yet available to inform the current analysis. Inthe absence of other data sources, we use the results on the national examinations administered at the end o f the primary cycle as a proxy for student 90 learning. Although admittedly flawed, the advantage is that the data for this variable pertain to the entire country. 5.32 The analysis should ideally be performed using individual-level data, but the practical difficulties o f preparing the requisite data proved insurmountable, particularly because a usable dataset would require for each pupil other information besides examinations results, including especially the pupil's socioeconomic background. Given the constraints, we perform the analysis using schools as the unit of observation. W e merged the examination data for 1999 with those from the 1999 teacher census and the 1999-2000 school census.73 The resulting dataset pertains to 1,362 of 1,533 schools in the public sector that presented candidates in 1999. For each school, there is information on the average score of the school's grade 6 candidates, as well as the characteristics of the school as defined by overall spending per pupil on personnel, the educational attainment of teachers, the pupil-teacher ratio, the conditions of buildings and so on. Although the resulting dataset is unique in Rwanda and very useful for our purpose, its limitations should also be kept inmindwhen interpreting the results. 5.33 Overview o f examination results. By way of background, table 5.9 shows the mean examination results across the three types of schools. The data for state and libre subsidik schools are highly comparable-mean scores of around 45 and standard deviations o f 8.3 and 8.6 respectively. The major difference is with the private sector, where both the average score and standard deviation are significantly higher. Private school pupils tend to achieve highscores, but the gap between the top and bottomperformers i s also much wider. Table 5.9: Primaryschoolleavingexaminationresultsby type ofschool, Rwanda, 1999 Type of school Mean scorea I Std. Dev. I Number of schools State Libresubsidib 1,135 Private 46.5 All schools 44.5 8.4 1,547 a. The scoreis expressedinpercentageterms. Source: NationalExaminations CouncilofRwanda. 5.34 Figure 5.7 below illustrates the relation between pass rates o n the grade 6 national examinations and the per pupil cost of teachers and other school personnel. Each of the 1,362 sample schools is represented by a small circle. Among both state and libre subsidik schools, the relation is weak intwo senses: (a) on average, schools with more spending per pupildo 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. It should be noted that these patterns are not unique to Rwanda. In other countries where similar analyses have been conducted, such as Mauritania and Madagascar, the patternis similar. 73We should ideally have matched the data from the two censuses to the examination results for year 2000 to improve the time consistency of the analysis. Given the practical constraints, we relied instead on the examination results for 1999. Our assumption i s that any changeinthe characteristics of individual schoolswould havebeenmodest inthe courseof a single school year. 91 Figure 5.7: Relationbetweenexaminationresults and spendingon cost ofsalariesper pupilacross schools, Rwanda, 1999 stateschmls 0 70 0 0 0 o c 0 0 0 O O 0 0 0 m 5.35 How can one interpret the finding that the level of school fundingrelates only weakly to pass rate on the examination? An attempt to answer the question must begin with the recognition that at least two stages separate the translation o f resources into learning outcomes: (a) how the resources are used to create learning environments at the classroom level and (b) what incentives are inplace to motivate the behavior o f teachers and pupils. With regardto how resources are used, the mix of school inputs that characterize classroom conditions is particularly relevant: the availability of teachers (as proxied by the pupil-teacher ratio) and their 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; nonetheless all education systems operate in resource-constrained environments, and the issue is to find the right balance among the various inputs, so that at each level o f funding, the resources are used to create as effective a learning environment as possible. Trade-offs in the choice o f inputs is inescapable, and poor input mixes can give rise to differences across schools inthe translation of resources into leaming outcomes. 5.36 The second stage that mediates between resources and learning outcomes concerns the incentives that motivate behavior, particularly that of teachers. 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 his or her students. There are various ways to communicate to teachers a set of incentives for good performance, and not all of them need involve financial rewards. In some countries, simply clarifying expectations and the scope of a teacher's responsibility provided a start toward more effective classroom teaching; in others, helping teachers toward better management of classroom processes throughsupportive supervision practices has yielded positive results. 5.37 These foregoing 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 understanding the nature of the challenges can nonetheless be made using simple regression analysis to examine the correlates ofperformance onthe grade 6 national examinations. 92 5.38 Remession analysis o f the correlates of grade 6 examination results. Recallthat the data relate to school-level observations. The nature of the data limits us to an assessment of the relation between certain physical attributes o f the leaming environment and examination results. These attributes include the pupil-teacher ratio, the composition of teachers according to their educational qualification, the average years of experience o f the teachers, the quality of the facilities (as proxiedby their reported condition). Because teacher qualification and experience can be translated into the salary bill, we can also replace these variables with the estimated cost of teacher salaries per pupil in the specification of the regression equation. 5.39 The full regression results appear inappendix table A5.4, whereas their main highlights are summarized in table 5.10 below. 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 relationbetween this variable and the examination pass rate. The variable is defined for this exercise in terms o f the share of teachers with the credentials shown in the table. Using teachers with an upper secondary diploma and pre-service training (comprising 32.9 percent of the whole sample) as the reference group, the regression analysis allows us to compare the performance of teachers with other levels of qualification against that of the reference group, keeping the other (measured) attributes o f the schooling environment constant. For example, inthe "estimated marginal impact" column, the value of -0.06 i s shown 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 (ie., upper secondary diploma with pre-service teacher training) would reduce the pass rate by 0.06 percentage points. An asterisk against an "estimated marginal impact" in the table means that the estimate is statistically significant at the 1 percent confidence level. 5.40 Several points are noteworthy about the results. The first is that the marginal impact is negative and statistically significant for the following groups of teachers: those with general lower secondary education (ES 1,2, or 3), vocational lower secondary education (CERAI, etc.), and an incomplete upper secondary education (ES 4, 5, or 6). In other words, teachers with these qualifications (designated as "unqualified" teachers inthe terminology used inthe Rwandese context) are inferior to those with an upper secondary diploma with pre-service teacher training. These negative results are consistent with perceptions among Rwandese policymakers. Yet, the magnitude of their shortfall in performance is modest. For example, the maximum marginal effect-for those with an incomplete upper secondary education (i.e., ES 4, 5, or 6), is only -0.1 1. Thus, if the 11.8 percent of sample teachers in this group were replaced entirely with teachers with an upper secondary diploma and pre-service training (Le., D6 or D 7pkdugogique),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 o f "unqualified" teachers (i.e., 33.5 percent of the teachers with ES 1,2, or 3; CERAI [and equivalent]; and ES 4,5, or 6) would boost the pass rate to only 47.2 percent. 93 Table 5.10: Correlates of school-levelpass rates on the primary school leavingexamination, Rwanda, 1999 Regression variable sample mean Estimated marginal (See Appendix table A 5.4 for full list) Teacher credential" of regression variableb impact' )ependent variable: Examinationpass rate (YO) - 44.5 Teacher qualification Reference group: Upper secondary diploma with pre-serviceteacher training D6, D7pkdagogique 32.9 - 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, E M , etc. 4.1 - 0.06 Utmer secondary Incomplete ES4,5, 6 11.8 - 0.11** Diploma (3-5 years)" D3, D4,D5 18.3 - 0.03 Diploma (6-7 years) without pre-service teacher D6, D7 non- training pkdagogique 9.9 0.00 Spendingon personnelper pupil (RFR) - 5,047 0.31 ** Note: -denotes "not applicable." Two stars (**) denote that the underlyingcoefficient estimate for the variable is statistically significant at the 1 percent level. Unmarked estimates indicatethat the underlying coefficients are statistically not significant. See appendix table A5.4 for the full regressionresults. a. No distinctionis madebetweenthose with andwithout pre-serviceteachertraining, becauseofsmall cell sizes. b. For theteacherqualificationvariables, the samplemeanscorrespondto the percentagedistribution ofthe sample across qualificationgroups. c. For the teacher qualification variable, the marginal effect refers to the percentagepoint change inthe examination pass rate inresponse to a 1 percentagepoint increase inthe share of teachers with the correspondingqualification at the expense ofa 1 percentage point decline inthe share of teachers with the referencequalification (Le., upper secondaryschool diploma with teachertraining). For the variable on spending, it refers to the percentagepoint change inthe passrateinresponseto an increaseo fFRw 1,000 inper pupilspending onpersonnelabovethe sample meanof FRw 5,047. Source:Authors' estimatesbasedon school-leveldata on examination results for 1999, mergedwith data f"MJNEDUC's 1999-2000census o f primaryschools andthe 1999 censusofteachers. 5.41 The modest gain in 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). In contrast, teachers with an upper secondary diploma start at FRw 23,649 (plus 6,500 inbenefits). The pay gaps are large, and it is reasonable to raise the issue of trade-offs here. In particular, given the current conditions of the 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 the money were instead used to increase the supply of textbooks, provide better support for incumbent teachers throughin-service teacher training, and bringabout more effective supervisory services? 5.42 A further detail in the regression results reported above pertains to the impact o f the level of spending per pupil o n the examination pass rate. The marginal impact on this variable is positive, indicating that, in general, the higher the level o f funding for schooling, the better is its performance. But, consistent with the story thus far, the effect is modest, because an increase of spending by FRw 1,000 (i.e., by about 20 percent above the sample mean) would boost the pass rate by only 0.3 1percentage points, raisingitfrom the current average of44.5 percent to 44.8 percent. 94 5.43 Differences inuerformance across provinces. The same regression analysis also allows us to characterize the examination performance across provinces. A summary of the results appears in table 5.11 below (see 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 inorder of their distance fiom the average performance o f Butare. Controlling for the qualification o f teachers, 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 inByumba is, o n average, 19.1 percentage points below. Table 5.11: Regressionestimatesof provincialdifferencesinexaminationpass rates, controllingfor differencesin per pupilspendingon personnel,Rwanda, 1999 ~ ~~ ~ ~ ~ ~~ Examinationpassraterelativeto Butare Regiondummy variables Controllingfor teacher Controllingfor per pupil qualification spending onpersonnel Butare(reference) - - Cyangugu -9.2 I -7.6 Gisenyi -10.8 -8.9 KigaliRural -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 Note: -denotes no applicable. See appendix table A 5.4 for regression specification andfullregression results. All coefficients on the regionaldummy variables are statistically significant at the 1%confidence level. Source: Authors' estimates based on school-level data on examination results for 1999 merged with data from MINEDUC's 1999- 2000 census ofprimary schools andthe 1999census ofteachers. 5-44 The provincial ranking is relatively stable whether the regression controls for teacher qualification or, alternatively, for the level of per pupil spending. Three groups may be identified among the provinces: those that fall a little way behind Butare include Cyangugu, Gisenyi, Kigali Rural, Kibungo, and Ruhengeri. Those that fall far behind include Kigali Ville and Byumba. Between these extremes are the provinces with intermediate ranking relative to Butare-Gitarama, Gikongoro, Kibuye, and Umutara. These results tell us that difference in the characteristics of schools that are explicitly accounted for in the regression analysis-teacher qualifications, pupil-teacher ratio, condition o f the facilities, and per pupil spending on staff-are only part of the reason why schools differ in ability to produce high-scoring students; however, the fact that the provincial dummy variables are all statistically significant suggests that other unobserved differences across provinces are also at work. Furthermore, the fact that interprovincial gaps are large implies that these 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 this province a particularly interesting province inthis regard. 95 Policy implications and conclusion 5.45 Primary school services in Rwanda are delivered largely through the network o f state and libre subsidib schools that make up the public sector. Among the latter, schools are financed largely by the government through payment o f teacher salaries, but are by and large runand managed either by the Secrbtariat Nationale de I'Enseignement Catholique (SNEC) or by the Conseil Protestant du Rwanda (CPR). Beyond documenting these structural dimensions o f the system, the analysis in this chapter touched on several issues with important policy implications for the managemento f service delivery. These include (a) the size and spatial distribution o f schools, (b) the consistency in teacher allocation across schools, (c) the balance o f input mix to support effective leaming environments, and (d) the effectiveness o f classroom processes in transforming school funding into leaming outcomes. Although data constraints prevent a definitive analysis of these problems, the findings nonetheless provide food for thought in the context o f ongoing policy development. 5.46 Public schools are relatively large in Rwanda, enrolling on average about 700 pupils each. Yet, the available evidence suggeststhat there are few economies to scalebeyond enrollments o f about 400 students. Transforming the current network o f schools into a system o f smaller, but more numerous schools would not increase the cost o f service delivery, but would help to bring schools closer to pupils' homes, thus improving access. Ruralchildren would benefit the most, as only halfthe householdsinrural areas currently live within 30 minutesof the nearestschool. 5.47 Withregardto the consistency ofteacherallocation 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 o f teachers i s both relatively meager and poorly deployed. The prospects for improvement are good provided the government takes advantage o f new recruitment to rebalancethe allocation of teachers in favor o f schools andregions that are currently under-endowed. 5.48 Concerning the mix o f school inputs, the issue has both pedagogical and financial dimensions. Inan ideal and well-funded learning environment, pupils are given sufficient time to leam during the school day, taught in small classes by teachers with good qualification, and helpedby an ample supply o f books and other learning resources. Inresource-constrained environments, however, tough choices and trade-offs present themselves; the challenge facing policymakers is to managethem inaway that wouldrelievethepractical (as opposedto abstract) constraints onperformance. 5.49 Some o f the results presented in this chapter raise important questions in this regard. The overall supply o f teachers inRwanda is generally low relative to enrollments, as reflected inthe country's relatively high pupil-teacher ratio, currently averaging about 57 pupils per teacher in the public sector. Reducing the ratio would appear appropriate in light o f experiences in other countries. Ina context ofhardbudgetconstraints, however, this wouldrequire trade-offs against other inputsthat mightalso be deemedimportant. Inparticular, the level ofteacher qualificationcomesto mind.Hiring well-qualified teachers, for example those with upper secondary education diplomas, has its benefits for the leaming process, and the current profile o f primary school teachers shows the country making significant headway in this direction. Yet, the tenuous link between teacher qualification and performance-albeit measuredimperfectlyhere usingexamination results-raises legitimate questions about the wisdom ofputtingthe bulko fresources into teacher qualification, at the expense o f lowering the pupil-teacher ratio and thereby increasing the length o f the school day for pupils inthe first three gradeso f schooling. 96 5.50 Inaddition, the analysis inthis chapter highlightsthe needto lookbeyondthe physical inputs. Resources matter, but because schooling is a social process in which participants' behavior affects the effectiveness with which resources are transformed into outcomes, it i s critical to ensure that the "software" of managing service delivery i s put inplace and used. Teachers are the main agents on the front lines, so it i s essential that they are not only equipped for the job by proper training, but are also providedwith the appropriate support and incentives to performtheir jobs well. 97 Chapter 6: Service Delivery in Secondary Education 6.1 Secondary education in Rwanda i s a relatively small subsector at present, enrolling some 141,000 students in2001, compared with nearly 1.5 million pupils at the primary level. Yet, as chapter 2 suggests, the pressure to expand secondary school laces is likely to mount incoming years as more and more children complete their primary s~hooling!~Managing the subsector's expansion in a fiscally sustainable manner i s therefore rapidly emerging as a critically important policy issue. As a contribution to the discussion, this chapter documents the subsector's key features by consolidating information on the institutional composition o f secondary schools and the overall characteristics o f their services. Italso examines the deployment ofteachers and resourcesacross schools, the nature of economies o f scale inservice provision, and the correlates o fperformance inthe national examination at the end o f the tronc commun cycle. As inthe previous chapter, data limitationshave constrained the analysis in important ways and reduced the precision o f the findings. Although the results should obviously be understood intheir proper pers ective, we believe they remain sufficiently robust to help guidepolicy development for the subsector. 7p Overview of the supply of services 6.2 Below we provide a brief sketch o f some key features in the supply o f secondary schooling in Rwanda by documenting the institutional composition o f the system, selected characteristics o f the `schools themselves, and the curricula offered by the schools. The picture that emerges i s one o f a highly fragmented system that probably needs to be streamlined for sustainable expansion. 6.3 The institutional infrastructure. As at the primary level, the government finances and runs the state schools. Itpays the salaries o fteachers inthe Zibre subsidib schools, but leaves their day- to-day management inthe hands of churches and other nongovernmental organizations, and it neither finances nor runs the private schools. 6.4 There are currently nearly 400 secondary schools inRwanda, 19 percent o f which are state schools, 30 percent are Zibre subsidib schools, and the remaining 51 percent are private schools (see figure 6.1). These percentagescontrast sharply with those at the primary level where the private sector i s tiny. The percentages also vary widely across provinces: the private sector's share, for example, ranges from a high o f more than 80 percent in Kigali Ville to just more than 20 percent in Gikongoro; the state schools' share ranges from a low o f about 6 percent in Butare to more than 40 percent inKibungo. 74To illustrate, ifsurvival rates at the primary cycle were to stabilize at even 50percent (which is much smaller than the current level of some 73 percent, which may not persist because o f high repetition rates), the probability that a primary school leaver makes the transition to the secondary cycle would remain at less than 40 percent (compared with the present rate of 60 percent), even if the secondary school system(public andprivate sectors) enrolled twice as many students as it does now. 75Data problems in secondary education were particularly serious. Information ffom MINEDUC's annual school census for the latest available school year was incompletely computerized, thus, requiring additional effort to fill inthe gaps. This taskproved a challenge, because schools were sometimes knownby more than one name, and different names were used indifferent datasets. Thanks to help from various sources and time-consuming data-cleaning efforts, most o f these problems were eventually resolved to our satisfaction. 98 Figure 6.1: Number and institutionalcompositionof secondaryschools by province,Rwanda, 2000-01 Total o f 374 schools Rwanda Gitarama Gisenyi Kigali Ville Ruhengeri Butare Cyangugu Kibungo Provincial population shares : Kigali Rural 10 - 12 % Byumba 1- 8 % Kibuye 4 - 6% Gikongoro Umutara I I I I 60 40 20 0 0 20 40 60 80 100 Number of schools Percentage share o f the schools Source: Number o f schools and their distribution from MINEDUC's2000-01 census o f secondary schools; population shares from Govt. o f Rwanda 2002. 6.5 It is interesting to compare-using both the left and right panels inthe figure above- the supply o f schools and their composition across provinces with comparable shares o f the country's population. For example, about 10-12 percent o f the population lives inGitarama and Gisenyi, but the former province has more schools (about fifty compared with Gisenyi's forty-three schools) and a larger share o f the schools i s private. The pattern is suggestive o f possibly more favorable social conditions for secondary schooling inGitarama, including stronger household ability and willingness to pay for schooling and probably more effective local initiative instarting schools. Ruhengeri, Kigali Rural, and Byumba also account for about 10-12 percent of the population, but they have fewer schools than either Gitarama or Gisenyi-around thirty schools each. The private share of schools in Ruhengeri and Kigali Rural is about the same as inGitarama, but is much smaller inByumba. These patterns again point to possible differences across provinces in the social context and market for secondary schooling. In the three provinces with the smallest shares o f population-Kibuye, Gikongoro, and Umutara-stronger government involvement appearsto be one reasonwhy the former two provinces have about twice as many schools as Umutara. The lower government involvement in Umutara may also be associatedwith Umutara's unique features. Most o f the area i s part of a national park, and many o f 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 Rwandese system, stronger government involvement inthe province has simply beenimpossible to arrange inthe short run. 6.6 Consider next the data intable 6.1, which shows the distribution of secondary schools according to the management network to which they belong and their funding status. InRwanda, the main management networks are (a) the state, (b) the Catholic Church through its Secre`tariat de I'Enseignement Cutholique (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 99 associations o f private or community schools. Schools belonging to the state, SNEC, and CPR each account for between nearly a fifth and nearly a quarter o f the schools; the unstructured private sector claims the remaining 39 percent o fthe schools. Table 6.1: Number and percentagedistribution oftypes of secondary schools, Rwanda, 2001-02 Network Type of school by source of finance `YOprivately State Libre subsidik Private All typesa financed State 72 0 0 72 (19.0) 0.0 SNEC~ 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 72 112 194 (19.0) (29.6) (5 1.3) 378 (100.0) 51.3 a. Figuresinparenthesesshow the correspondingcolumnpercentages. b.Referstotwo ofthemainchurch-relatednetworks, the Secrdtariatde I'EnseignementCatholique (SNEC) andthe Conseil Protestant du Rwanda (CPR). Source:Compiledfromlists of schools suppliedby MINEDUC, SNEC, andCPR. 6.7 State schools are by definition funded by the government; but, unlike the situation at the primary level, sizable shares o f SNEC (32 percent) and CPR schools (54 percent) are runentirely on private funding. In the face o f limited public funding and a growing demand for secondary schooling, these networks have respondedinthe only practicable way to meet the demand, that is, by tapping into private sources o f funding. Among the independent schools sector, the surprisingfinding i s that some twenty o fthem have managedto gain entry into the libre subsidik category, which entitles them to receive government fundingfor teacher salaries. Although it is unclear how nonstate schools qualify for such funding, the prize o f being in the libre subsidik category i s so advantageous that schoolscanbe expectedto seek it assiduously. 6.8 As indicated inthe previous chapter, significant pressuresare buildingup on Rwanda's secondary school system, giventhe growing numbers o fprimary school leavers likely to seek entry to secondary schools. Because o f continuing budget constraints, the government alone i s unlikely to be able to meet the increased demand. Private funding will therefore continue to be critical. Yet, the current financing arrangements in the sector are such that the burden o f mobilizing private funding falls unevenly across schools; private schools bear a disproportionate share of this burden. In considering the future prospects for secondary education, the issue o f school funding probably warrants additional attention. In particular, policymakers may need to rationalize funding arrangements to spread the available public subsidies more equitably across schools so that public funds are used to leverage private contributions in all schools, thus putting the expansion of the subsectoron a more fiscally sustainablefooting. 6.9 Some characteristics o f the supply o f services. Here we pick out some aspects o f the way schools are organized. According to table 6.2, the distribution o f enrollments across state, libre subsidik, and private schools generally tracks the distribution o f the schools across these categories (shown in figure 6.1 above). Libre subsidik schools tend to be bigger and private schools tend to be smaller than the average size o f secondary schools nationwide, but the differences are not large and reinforce the impression o f a system composedmostly o frelatively small schools. 6.10 Rwanda's publicly financed secondary schools traditionally offer boarding services. In recent years, because o f the growing pressure to expand places, they also cater to day students. Unfortunately, no data are available on the number of boarders and day students. Thus, we rely on the 100 Ministry of Education's records of foodstuff allocation to document the situation.76All schools currently offer boarding services, and it appears that nearly 60 percent of the state schools cater exclusively to boarders, compared with nearly 50 percent among the Zibre subsidib schools. Although boarding schools may be justified in an elite system of schooling, they are a fiscally unviable model for expanding the system, giventhe large diversion o f state funds to feed the students. Findingways to phase out boardingschools would appear to be apriority itemon the agendafor policy development. Table 6.2: Selectedcharacteristicsof state, libresubsidit!, and private secondary schools, Rwanda, 1999-2001 Public sector Private All schools State Libre schools subsidie` II Numberofschools, 2000-01 72 110 191 I 374a I Percentageshare of students, 2000-01 19.6 34.6 100.0 43`8 (141,163) Average enrollment per school, 2000-01 384 467 323 378 Percentreceiving foodstuff from MINEDUC, 1999-2000 Noallocation 0.0 0.0 100.0 Allocation basedon number ofboardersonly 58.3 46.8 Allocation basedon number of boarders& day students 41.7 53.2 23.7 Total 100.0 100.0 0.0 100.0 a. Includesone schoolwith missing informationregardingits category; the ' t o fthe table is - - jed ondata fort 373 schools. Source: For data on number of schools, average enrollmentsperschool, and distribution of enrollments, electronic data files from MINEDUC's annual census of secondaryschools for 2000-01; for data on the share of studentsin the tronc commun cycle, MINEDUC's statistical abstract for 2000-01; for data on schools receiving foodstuf, administrative recordso f MINEDUC's SecondaryEducationDepamnent. 6.11 The supplv o f instructional prom-ams. Secondary schools inRwanda offer the gamut o f instruction from lower secondary schooling to specialized vocational and technical courses at the upper secondary level (table 6.3). Slightly more than a thirdof the schoolsprovide instruction only in the tronc commun or lower secondary, cycle; nearly 60 percent offer boththe tronc commun and upper secondary cycles, whereas only 5 percent serve upper secondary students exclusively. The distribution o f instructional programs shows distinct differences across types o f schools: more than halfo fthe state schools specialize in the tronc commun cycle, compared with only a quarter o f the Zibre subsidik schools and about a third among the private schools. Among schools serving both lower and upper secondary students, the spread o f course offerings i s distinctly wider in the Zibre subsidik and private sectors; many o f the schools provide instruction in two or even three streams at the upper secondary level. The diversity may be a sign o f the schools' responsivenessto market demand (given that both public and private schools collect nontrivial levels o f fees). Yet, because most o f the schools are typically quite small, with a total of fewer than 500 students each, on average, the implied fragmentation of course offerings may mean that schools are failing to take advantage o f scale economies in service delivery. We shall return to this issue later in the chapter when evidence i s presentedon the existence andnature o f scale economies insecondary education. 76Up until 2001, the ministry calculated and allocated foodstuff to schools according to number of boarding and day students. The practice was replaced in 2002 by a new system under which schools receive a financial allocation to cover the cost o f feeding the students. 101 Table 6.3: Percentagedistributionof schools by leveland number of instructionalstreams offered, Rwanda, 2000- 01 Public sector Level andnumber of instructional stream State Libre Private All schools subsidie' Tronc commun only 51.4 24.5 34.6 34.9 Tronc commun combined with uppersecondarycycle" 1stream 29.2 40.0 38.7 37.3 2 streams 8.3 24.5 22.5 20.4 3 streams 0.0 3.6 2.6 2.4 Uppersecondarycycle on14 1stream 8.3 7.3 1.6 4.6 2 streams 2.8 0.0 0.0 0.5 Total 100.0 100.0 100.0 100.0 (Number of schools) (72) (110) (191) (373) I Vote: Students in upper secondary schoolare enrolled inone of four pi traininel a/ See appendixtable A6.1 for detailsonthe specific combinationsofthe instructional programs offered. Source: Basedon data from MINEDUC's censusofsecondaryschools for 2000-01. 6.12 In terms of the distribution of enrollments, the relevant data appear in table 6.4. Consistent with the information presented above, the majority o f students-about two-thirds-are in the tronc commun cycle, with the slightly above-average share in the state schools counterbalancing the below-average share inthe libre subsidik sector. At the upper secondary level, students are spread across the general, normal (ie., teaching training), vocational andtechnical streams. Overall, less than 7 percent o f the studentspursuethe technical stream; about a thirdo fthem, the general and vocational streams; and a quarter, the normal stream. Substantial differences exist inthe distribution across types of schools around these system averages: in state schools, there is a much stronger focus on the vocational and technical streams than inthe other two types o f schools; together, these streamscater to some 55 percent o ftheir students. Inthe libre subsidik, the emphasisi s on the general stream, whereas inthe private sector, the focus is on the normal and vocational streams, which together serve more than three-quarters o ftheir students. Table 6.4: Distributionof secondary studentsby cycle and streamacross schooltype, Rwanda,2000-01 Cycle & stream State I Libre subsidie' 1 Private 1All schools %intronc commun cycle 67.1 59.7 64.2 63.1 %inupper secondarycycle 32.9 40.3 35.8 36.9 Distributioninupper sec. cycle (%) General 36.1 46.6 16.7 32.0 Normap 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 a. Refersto the primary schoolteacher training stream. Source: Authors' calculationsbasedon MNEDUC's electronic files from the 2000-01 census ofsecondaryschools. 6.13 For a closer look into the course offerings in upper secondary education, consider the data intable 6.5 below, which shows the number o f schools offering the indicated fields o f instruction 102 in each stream, as well as the number of students enrolled in that field. The shaded cells hi hlight fields where the average enrollment per school inthe public sector is fewer than 100 students?'In the Math-Physique stream, six state and eighteen libre subsidik schools offer the curriculum in2000-01, but each state school enrolled, on average, 118 students, compared with the corresponding average of only 86 students inthe libre subsidik schools. Inthe Action Sociale and Secrktariat streams, there were one state and five libre subsidik schools involved, but enrollments per school in each o f these fields were much lower in the state than inthe libre subsidik schools. Looking at the table as a whole, the general pattern is that small enrollments per field tend to characterize the offerings in the technical streams, not only inthe state sector, but also among libre subsidik schools. 77Because private schools in Rwanda rely heavily, if not completely, on fees to fund their operations, we assume that they are recovering their costs evenat the low enrollments shown inthe table 6.5 insome o f the fields. Public schools are shielded fromsimilar market forces and are therefore likely to be less cost-conscious. Note that because the same labels may mask differences in course content, private schools may well be offering different cumcula than those offered inthe public sector--curricula that are adapted to keep costs down where enrollments are low. 103 Table 6.5: Number of schools offering upper secondary programs and average enrollments per schoolby field, Rwanda, 200&01 Numberof schoolsofferingfield Average number of studentsper schooland field o ;tudy State Libre - I - subsidie' Private Private 111schoolsoffering a program at the upper econdary level 37 27 66 ;enera1 Lettres 2 19 10 128 41 Science Humaines 13 27 20 90 78 Math-Physique 6 18 7 118 67 Bio-Chimie 9 28 17 129 75 v'ormal Primaire 4 22 62 215 124 7ocational Agricole 5 5 5 159 113 66 Foresterie 1 0 0 182 - - Ve'te'rinaire 5 1 1 137 129 42 Hygibne 0 1 0 - - Laborantins 2 0 1 122 67 SciencesInfirmitres 4 7 11 263 209 Action Sociale 1 5 2 85 H6tellerie & Tourisme 0 0 2 - 202 Droit 0 2 9 129 100 Secre'tariat 1 5 7 109 78 Commerce& Comptabilitk 5 18 38 132 110 Informatique 0 0 2 40 rechnical Electricite' 3 2 5 116 44 Electro-mkcanique 1 0 1 93 20 Electronique 1 0 1 274 Me'canique-ge'nkrale 2 0 1 20 Me'caniqueautomobile 3 0 6 89 Construction 3 1 3 36 TravauxPublics et Construction 1 1 0 - Plomberie-Soudure 0 1 0 - I 20 - Menuiserie 1 0 1 11 Me'tal-Bois 0 1 1 - I 166 132 Me'tal-klectricite' 0 0 2 - I I - 99 Enpins Lourds 1 0 0 - Nore: -denotesnot applicable;shadedcellshighlight fields where average enrollmentsare relatively small, Source: Basedondata f?om MINEDUC's 2 0 0 M 1 census ofsecondaryschools. 104 6.14 The question to ask at this point is whether the mix of services provided by the public sector consisting of state and libre subsidik schools is consistent with economic considerations. In particular, the small enrollments per field and institution bear further scrutiny because of their cost implications. Infields where only one school is currently offering the stream, the courses may be new and the small enrollments may thus refer to single cohorts of students. If so, the menu of course offerings appears to be expanding where the demand is sometimes already being met by existing schools and sometimes where it may only be nascent (e.g., Electronique compared with Electro- mkcanique). 6.15 A more fundamental issue concerns the role of the state inthe 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. The experience from most developing countries suggests, however, that public sector institutions tend to perform poorly inthis regard. The weak link reflects inpart, the inherently feeble incentives for adapting service delivery where a school does not depend for its survival o n its ability to attract clients based on its successful job placement for its graduates. In Madagascar, for example, most graduates of technical secondary schools end up pursuing courses at the university, often infields unrelated to their training in upper secondary school, thus wasting the state's investment in their (expensive) training. 6.16 In light of the above discussion, Rwanda's policymakers might give careful consideration to the strategy for expanding vocational and technical education. The issue o f government financing requires separate consideration from that o f direct government involvement in service delivery. With regard to finance, 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 the argument is inRwanda's current context is difficult to tell, because we have no data on the profiles o f students in vocational and technical education. Concerning service delivery, the argument for direct government involvement is much weaker, particularly in the face of well-documented evidence from many low-income countries pointing to the public sector's typical slowness or even inability to restructure course offerings to match fluid developments o n the labor market7' Rwanda might thus find it worthwhile to examine how such countries as Zambia, C6te d'Ivoire, and Morocco have attempted to combine government finance with nongovernmental delivery of service to achieve their goals inthe sector. Teachers and their utilization and deployment across schools 6.17 W e now take a closer look at teachers, the single most costly input for the internal operations o f schools. Inparticular, we examine the distribution of teachers by qualification and their teaching workload across school types and subcycles and document the pattern and consistency of teacher deployment across schools. Although most o f the available data pertain to the public sector, information on private schools is included where possible. 6.18 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 subsidik schools (table 6.6). Most of the teachers, at least inthe public sector, for which data are available, are relatively young, averaging about 5.5 years of teaching experience in state schools and 6.6 years in libre subsidik schools. Throughout the system, almost all teachers have at least an upper secondary diploma. But striking differences across school types exist in the shares of teachers with only this level of educational attainment: about 64 percent among state school teachers, comparedwith See, for example, vocational technical training paper 105 some 60 percent in the libre subsidie' schools and only 48 percent in the private schools. The distribution i s consistent with the bias toward the tronc commun cycle in state schools and the expectation that teachers teaching that cycle would tend to be less qualified than those teaching the upper secondary cycle. The majority of public sector teachers report having had pre-service teacher training; the extent o fthis qualification i s unknownamongprivate school teachers. Table 6.6: Characteristicsof teachers by type of secondaryschool, Rwanda, 1999-2000 rype of school Teacher credentiala State Libre subsid2 Private 'ercent women - 16.6 24.1 9verage years of experience - I 5.5 6.6 reachers' educational attainment (%) Umer secondarv Incomplete ES4,5,6 1.1 Diplomaholder (3-5 years) D3, D4, D5 0.8 1.8 Diplomaholder (6-7 years) D6, D7 58.8 48.4 Post uvver secondary Two-year diploma holder BAC or 2 yr. diploma 22.5 33.7 University degree holder License, BA, BSc., 14.9 13.6 Masters, Ph.D. I Other Includes unknown 1 3.4 1.8 2.5 Percentwith pre-service teacher training - I 66.9 I 62.4 Note: -denotes not auulicable:blankdenotes no data. a. See appendixtable'A5.1 fora detaileddescription ofthe various credentials. Source: For state andlibre subsidii schools, MINEDUC's 1999census o fteachers mergedwith the 1999-2001census o f schools; for privateschools, MINEDUC's publishedstatisticalabstractfor 1999-2000. 6.19 As indicated above, a large number of Rwanda's secondary schools offer, under one roof, instruction at both lower and upper secondary levels. Insuch schools, some o f the teachers teach classes at both levels. Table 6.7 suggests that inthe state schools, just under a quarter o f the teachers fall inthis category, whereas the corresponding share i s nearly 40 percent inthe libre subsidie' sector. Teachers who teach both cycles appear to have a slightly heavier teaching load, about 19 hours per week, compared with the overall average o f 18 hours. Incontrast to the relatively small differences in teaching load, student-teacher ratios vary significantly across levels o f instruction and type o f school, implyingprobably large differences inaverageclass size. 106 Table 6.7: Distributionof teachersby type of classesthey teach andtheir teachingloadand student-teacherratios in public secondary schools, Rwanda, 1999-2000 Type of classes or schoola Tronc Both tronc upper commun commun secondary on1y and upper secondary on1y Distribution of teachers by type of classes taught (%) Instateschools 50.5 23.6 25.9 100(1,017) Inlibre subsidid schools 33.7 38.0 28.3 100(2,234) Teachers' average teaching workload per week (hours) Teachers instate schools 17.4 19.2 18.7 18.1 Teachers inlibre subsidik schools 17.8 19.0 17.3 18.1 Student-teacher ratio` State schools 26.6 21.3 15.5 23.3 Libre subsidid schools 19.3 22.0 16.6 21.2 a. All calculationsinthe table countteachers in full-time equivalentunits accordingto their time allocation ineachcycle. b. Figuresinparentheses indicatethetotalnumberofteachers withvalid dataamongthe 3,257 enumeratedintotal inthe 1999census. c. Refers to school-level averages. Forcomparison, the student-teacherratio inprivate schools averages24.2. Source: MINEDUC's 1999census of teachers, mergedwith data from the 1999-2000 census ofsecondaryschools, for distribution ofteachers by type of classes taught andteachingworkload; electronic data files from MINEDUC's annualcensus of secondaryschools for 1999-2000 for data on student-teacherratios. 6.20 W e can deduce the magnitude o f the differences using the following tautological relation between the student teacher ratio (STR, which is simply the division of the number of students [SI by the number of teachers [TI); the weekly teaching workload of teachers (TH); the weekly instructionalhours received by students; and the class size (CS): The left-hand side gives the aggregate instructional hours received by the student body, whereas the right-hand side gives the aggregate hours o f teaching offered by the teachers. By rearranging, we have the following relation: 6.21 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 instate schools are on average about 41 percent larger than the average inZibre subsidik schools (=[26.6/19.3] x [17.8/17.4]). Thus, ifclass sizes average about 40 students per section inthe Zibre subsidik schools, they would go as high as 56 students per section inthe state 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 they are inthe upper secondary cycle (=[26.6/15.5] x [18.7/17.4]); whereas inlibre subsidiC schools, the corresponding gap is much smaller at 13 percent (=[19.3/16.6] x [17.3/17.8]). 107 6.22 Inaddition, consider the data intable 6.8, which show the educational qualification of teachers instate andlibre subsidid schools according to the type of classes they teach. Among teachers with only tronc commun classes, almost all have at least an upper secondary school diploma, but the share that is university educated i s slightly smaller in state schools-just under 10 percent, compared with nearly 12percent inlibre subsidid schools. Althoughan upper secondary school diploma may be less than ideal for teaching trunc commun classes, it is at least minimally adequate, because the teachers have at least 3 more years of schooling than their charges. In contrast, the situation in the upper secondary cycle is alarming: nearly half the teachers instate schools and more than half o f those inthe libre subsidik schools have only an upper secondary school diploma themselves. The profile o f teachers with duties in both cycles lie mid-way between those teaching either the lower or upper secondary classes. Table 6.8: Distributionofstate andlibresubsidid secondary schoolteachers by educationalattainment andtype of classestaught, Rwanda, 1999-2000 (percent unless otherwise indicated) Type of class taught byteachera Type of school andteacher qualification T~~~~cOmmun commun and Both tronc UPPU only upper secondary secondary only Teachers instate schools (514) (240) (263) Upper secondary diploma 69.0 64.2 49.3 2-year post-secondary diploma 16.9 16.0 25.1 University degree 9.5 12.3 17.9 Other 4.7 2.8 7.6 Teachers inlibre subsidid schools (752) (850) (632) Upper secondary diploma 64.8 57.0 54.4 2- year post-secondary diploma 19.8 23.6 24.2 University degree 11.7 16.7 16.2 Other 3.7 2.7 5.2 I I a. Figuresinparenthesesindicatethe numberof observations on which the percentagesarebased. Source: MINEDUC's 1999 census of teachers, merged with data from the 1999 census of secondary schools, in which 3,257 teachers with teachingduties were enumerated. 6.23 The foregoing findings suggest that a clear priority in policy development is to find ways to improve the educational qualification of upper secondary school teachers. A s it may take years to train a sufficient number of teachers to at least 2 years past the secondary-school diploma level, it is important to distinguish between short-term and long-term measures. One short-term option is to replace the underqualified, incumbent upper-secondary-school teachers &e., those without at least a 2-year post-secondary diploma) with teachers currently teaching tronc commun classes who are at least minimally qualified. Considering only the teachers who teach classes in either the lower or upper secondary cycle, there were in 1999 a total of 473 upper secondary teachers who lacked the minimumqualifications and 371 tronc cummun teachers who hadthe minimumqualifications to teach inthe upper secondary cycle. Ignoringfor the moment the possibility that a complete swap between these two groups of teachers might not work, because of mismatches of fields of specialization, location o f work, and other practical constraints, nearly 80 percent o f the underqualified teachers currently teaching in the upper secondary cycle could potentially be replaced. The share could conceivably go higher ifthe swap were extended to teachers who teach both cycles. 108 6.24 A legitimate question is how much, on balance, the system would actually gain from the swap, given that it would help solve a serious teacher constraint inthe upper secondary cycle, but only at the possible expense o f depressing teaching quality in the lower secondary cycle. W e shall retum to this issue later when some relevant evidence is presented on the relation between teachers' educational attainment and student learning. For the moment, suffice it to say that, although the groundwork for long-term solutions must be laid through careful expansion of post-secondary education to supply upper secondary teachers with the desired qualification, the immediate option of rationalizing the deployment of incumbent teachers also warrants consideration. 6.25 Teacher deployment across schools. W e now examine how teachers are currently distributed across secondary schools; as before, the available data pertain only to schools inthe public sector. Figure 6.2 hints at substantial variation inthe number of teachers across schools of similar size, an unsurprising pattem given that schools offer different levels and types o f instruction. To see how much of the variation persists after controlling for these differences, we tum to regression analysis and report the results below. Figure 6.2: Relation between number of students and teachers across state andlibre subsidie' schools, Rwanda, 2000 (a) State schools 40- O O 6oi (b) Libre subsidid schools 0 0 0 0 m 0 93 8 30- 0 0 0 0 0 rcl 0 40 0 0 0 0 3E 0 0 0 0 00 20- 0 0 O O 0 30- O ,"o 4 0 0 0 0 z O 0 0 0 0 ooo 8 800 0 0 0 0 000 20- 0 0 10- 0 0 , 0 O 0 0 i I i I 100 250 500 750 1,oo I 5 -oo I I I I n Number of students Number of students Source: Basedon MINEDUC's 2000-01census ofsecondary schools. 6.26 W e implement the regression analysis separately for the tronc commun and upper secondary cycles (table 6.9). As indicated earlier, some o f the teachers teach classes inboththe lower and upper secondary cycles. For the purpose o f this exercise, they 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 number of streams of instruction offered and the second set with the variable included. Consider first the coefficient on the number o f students: they are all 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 about 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. These values imply that teachers are supplied at the 109 average rate of one teacher per increase in enrollments o f between 25 and 29 students (which compares with the rate of one teacher per increase o f 63 pupils at the primary level). Table 6.9: Regressionestimatesofthe relationbetweennumber of teachers andstudents across public secondary schools, Rwanda, 1999-2000 Tronc commun cycle" Upper secondarycyclea Model 1 Model 2 State Libre 411public subsidib schools State Libre 411public Libre 411public subsidie` schools State subsidik schools Numberof pupils 0.040 0.036 0.038 0.048 0.047 0.048 0.035 0.040 0.039 (8.59)** (8.33)* * (12.20)** (9.71)** [13.77)** :16.31)** (7.77)** (10.32)** (13.46)** Numberof streamsb - - - - - - 2.651 1.423 1.794 - - - - - - (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 RZ 0.75 0.50 0.61 0.75 0.52 0.61 0.84 0.55 0.66 Note: -denotes not applicable. Figuresinparenthesesrefer to the robustt-statistics.One star (*)denotes statistical significance at the 5% level, andtwo stars (**)at the 1% level. a. Becausesometeachers teach inmorethanone cycle, the number ofteachersis countedin full-time equivalent unitsaccordingto the distributionofeachteacher's workload across the two cycles. See Appendix table A6.2 for additionalresults for the combinedsample ofschools offeringeither or bothsecondarycycles. b. Variable appliesonlyto the upper secondarycycle, where studentsfollow the curriculum inoneof four streams: general, normal, vocational,andtechnical. Source: Basedon data from MINEDUC's 1999-2000 censusofsecondary schools mergedwith data from the 1999census ofteachers. 6.27 Consider next the R2values of the regressions. Those for the state school regressions consistently exceed the corresponding values for the libre subsidid schools. This result signifies that although both types of schools follow comparable teacher allocation rules, the application of such rules at the level of individual schools is much less consistent in the libre subsidid sector. Figure 6.3 presents the results graphically and includes those for primary education from the previous chapter to provide an overall perspective. Interestingly, the inconsistency in teacher allocation across libre subsidib schools is less o f a problem at the primary as it is at the secondary level. Even so, the fact that R2values are nowhere more than around 0.80, indicates that there remains room for improvement throughout the system. 110 Figure 6.3: R2values of regressionsrelating numbers of teachers and students across schools by levelof education and schooltype, Rwanda, circa 2000 Tronc commun Upper secondary Source: For primary education, table 5.5; for tronc commun and upper secondary education, table 6.9. Economies of scale inservice delivery 6.28 Beyond ensuring consistency in teacher allocation, another challenge in the management o f the sector is to take advantage o f economies in service delivery where these exist. Such economies are present when a school's total cost rises less rapidly than the increase in student enrollments, as certain fmed costs, such as specialized teachers or administrative and support staff at the school level, are spread over the rising enrollment^.^^ As a result, the cost per student of service delivery can be expected to fall as enrollments increase. 6.29 Overall uattem. Figure 6.4 shows a simple scatter plot of the relation betweenunit costs and size of enrollments across schools. The scatter is visually even more dispersed than that in figure 6.2 relating number o f teachers and enrollments. As at the primary level, the pattern implies that in addition to teacher numbers, differences inthe availability of administrative personnel, as well as the educational profile o f the teachers and the administrative staff, tend to exacerbate, rather than narrow the allocation of resources across schools. 79For 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 the cost of administrative staff according to the share of enrollments in each cycle. 111 Figure6.4: Relationbetweenumberof students and cost perstudentinthe tronc commun and uppersecondary cycles, Rwanda, 1999 (a) T mcorn" cycle (b)W==wcYce 00 0 0 ajr c. 0 *T 8 0 0 0 0 0 g 0 0 0 0 0 O O 0 0 0 0 0 0 0 43 O 0 0 I O O O O O O O O 0 m 0 0 O 0 ug- 0 0 0 0 0 "1 00 0 1 0 0 0 0 0 00 O 0 0 0 100 I 200 I 300 I 400 I 500 I 600 I 700I 6.30 Regression estimates of the relation between unit costs and enrollments. The results show up in table 6.10.*' They suggest that given the way the system currently operates, significant 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 pointrise inenrollments; conversely, a 10 percentage point increase in spending on personnel would allow enrollments to grow by 16.1 percent. The magnitude of the economies is comparable across state and Zibre subsidib schools. In schools that combine instruction at 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.*' The analysis should ideally examine the possibleexistence of economiesof scope inaddition to that of economies of scale. We didnot implement the analysis, because the samples involved are small and the analysis is inappropriate given the rough quality of the available data. One explanation is that administrative staffing coefficients are higher for the upper secondary than for the lower secondary cycle. Becausethese costs are by construction spread between the two cycles according to the distribution of enrollments between the two cycles, the total personnelcosts inthe tronccorn" cycle would tend to be higher inschools with combinedcycles. 112 Table 6.10: Relationbetweentotalcost ofpersonneland enrollmentsinpublicsecondary schools, Rwanda, 1999- 2000 Tronc commun cycleb w Upper secondary cycleb State Libre All public State Libre All public schools subsidid schools schools subsidid 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)** Dummyvariable for single-cycle school -0.45 0.07 -0.20 0.117 0.257 0.123 (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 Memo. items: Economies of scalea 1.67 1.67 1 .ox 1.05 Average enrollment per school 290 301 249 253 Nore: Inall the regressions, the dependent variable andthe numberofstudentsare bothexpressedon a log scale; robust t-statistics are shown in parentheses; one star (*) indicates statistical significance at the 5% level, and two stars (**)at the 1% level. a. Refersto the percentageincrease inthe number of students enrolled for a 1percent increaseinspending on personnel, calculatedas the inverse of the coefficient estimate onthe independent variable (ie,number ofstudents). b. Where teachersteach morethanone cycle, the cost attributedto eachcycle isproratedinproportion to the time allocationofthe concernedteachers across the two cycles. Source: Basedon MINEDUC's 1999-2000 census ofsecondary schools mergedwiththe 1999census ofteachers, aas well as dataonthe salary scale of school personnel. 6.3 1 The foregoing regression results suggest that economies o f scale are also present at the upper secondary cycle. Their magnitude is very much more modest, however: an increase o f 10 percent inspending on school personnel would allow enrollments to rise by only about 10.5 percent on average. As at the lower secondary level the pattern o f scale economies i s comparable across types o f school. One possible explanation for the near absence o f scale economies can be found in the wide menu o f course offerings in the upper secondary cycle that was documented earlier in this chapter. Schools expand their enrollments mainly by diversifylng their curriculum offerings, rather than by increasing the scale o f existing offerings, thus preventing the schools from benefiting fi-om economies o f scale inservice delivery. 6.32 In addition, the average enrollment in both the tronc commun and upper secondary cycles i s relatively small-no more than about 300 students on average per cycle at each school. These small sizes stand in sharp contrast to the size o f primary schools, which currently average about 700 students per school. 6.33 The regression results are represented visually in figure 6.5. The horizontal axis indicates the size o f enrollments, whereas the vertical axis measures the corresponding unit costs simulated for a school o f the given size based on the regressionestimates. For reasons already alluded to above, economies of scale are present inthe tronc commun cycle, but are largely absent inthe upper secondary cycle. For the tronc commun cycle, these economies are particularly apparent as the size o f enrollments increasesupto about 400 students or so. 113 Figure6.5: Simulationsofeconomies of scale in secondary educationinthe tronc commun and uppersecondary cycles, Rwanda, 2000 (a) Tronc commun cycle 45 - 45 - zfL Schoolswith only uppersec. cycle 35- 8 tronccommun cycle i * Qg25- c.-- 28 Schoolswith combinedcycles ,015- I V 8 V 5 4 I I 0 200 400 600 800 S d0 200 400 600 800 Numberofstudents Numberofstudents Source; Simulatedfiam the regressiauiin table 6.11 6.34 Taking advantage of scale economies in service delivery. Recall from the previous chapter on primary education that the absence of scale economies led us to suggest that smaller schools may be no more costly to operate than larger schools and they would have the advantage of bringingschools nearer to pupils' homes, thus, facilitating educational access. Inthe tronc commun cycle, we have the opposite problem, in that most of the schools offering this cycle enroll too few students and are therefore uneconomic to operate. Because students at this level of schooling are older and therefore more able to travel longer distances to school, the argument of accessibility becomes weaker here. 6.35 To see the importance of taking advantage of scale economies in the delivery of secondary schooling, consider the data intable 6.11, which shows the estimated unit costs at different enrollment sizes and the current size distribution of secondary schools inRwanda. As indicated above, scale economies are more obviously present in the tronc commun than upper secondary cycle. Unit costs at enrollments of more than 400 tronc commun students are less than two-thirds as highas those at enrollments of between 100 and 200 students and about four-fifths as high as the unit costs at enrollments of between 200 and 300 students. Yet, more than a fifth o f the students are currently enrolled in schools catering to between 100 and 200 tronc commun students, and more than a third are inschools catering to enrollments ofbetween200 and 300 students. Ifall the small 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 to deliver lower secondary services are currently 3.5 times as high as the corresponding costs at the primary level. 114 Table 6.11: Sue distributionof enrollmentsin publicsecondary schools and simulatedcost per student,Rwanda, 2000 I I Tronccommun cycle 1 Uppersecondary cycle I I Size o f enrollments Average cost per Average cost peiI Size distribution (1,000s o f (1,000s of o f enrollments (%> < 100 18.6 100-20Ob 25.4 200 -300 23.2 35.5 27.3 22.0 300 -400 20.5 23.5 26.9 19.5 400 + 18.6 19.9 26.4 14.4 No. of schools offeringthe cycle' - 166 - 118 Nore: -denotes not applicable. a. The unit costs inthe first and last categories are simulated for enrollments o f 50 and 450, respectively, for the from commun level andfor enrollments of SO and500students, respectively, which are roughly the averagesizes ofenrollments per institution inthe size. Forthe othersize categories,the costsare simulated for schools inthe middle ofthe size bracket. c. Forthe rronc cycle, the range here is ''< 200." b. includes schools offering only one cycle or boththe fronc commun andupper secondary cycles. Source: Authors' simulations basedon the regressions intable 6.10. 6.36 The foregoing results carry obvious implications for management of the size distribution o f secondary schools as the sector expands. As indicated elsewhere enrollments at this level are likely to expand as more and more children complete their primary schooling. Ina growing system such as Rwanda's, taking advantage of scale economies means avoiding a proliferation o f schools that each cater to small cohorts o f students. Inpractice, the 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 secondary instruction to small enrollments in both cycles. In other settings, it might just be a matter o f enlarging enrollments to accommodatethe growing clientele for lower secondary schooling. Inyet other communities where no school currently delivers lower secondary schooling, it might be appropriate to start entirely new schools, ifthe demandis sufficiently strong to allow the school to fill rapidly at least 400 places inthe tronc commun cycle. 6.37 Unlike lower secondary education (which most educators increasingly view as an extension of primary schooling), the expansion o f places in the upper secondary cycle needs to proceed more slowly, ifonly because labor market considerations become increasingly relevant. This issue will be explored inmore detail in a later chapter, but for the present discussion, the important implication i s that enrollments at this level are likely to remain relatively small for the foreseeable future. Although small enrollments do not appear uneconomic, given the results presented above, the fragmentation of course offerings i s a problem and, as argued above, may be an important reason for the absence of scale economiesas such. Policies for managing the expansion o f this level o f schooling, particularly in the public sector, might thus seek specifically to avoid a proliferation o f course offerings, while at the same time taking advantage o f opportunities for consolidation where appropriate. Examinationsresults and their correlates 6.38 A final aspect of service delivery relevant to policy development is the relationbetween the resourcesthat schools receive and the output in terms o f student learning. As at the primary level, 115 the analysis i s severely handicapped by the lack o f appropriate student assessment data. Here again, we shall use the results onthe national examination administered at the end o f the tronc commun cycle as a proxy measure for student learning in lower secondary schooling and perform the analysis using schools, rather than individual students, as the unit o f observation. The approach suffers -from well- known flaws, but it i s the best that can be implementedunder the circumstances. Although a national examination is also administered at the endo fthe upper secondarycycle, we do not attempt to analyze the results, inpart because of the small numbero f schools involved and the large number of fields of specialization at this levelinthe system. 6.39 Overview o f results on the national tronc commun examination. The results are expressedinterms o f the pass rate and the average score (which can range from zero to a maximum value o f X). Because places in the upper secondary cycle are limited, the pass rate is adjusted from year to year to let through the appropriate number o f passers. The 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 intime. 6.40 The results for the 278 schools that fielded tronc commun candidates for the 1999 national examinations appear intable 6.12. They show a clear dichotomy inthe averageperformance of students attending public and private schools. Inthe public sector, students inlibre subsidik schools achieved a pass rate o f 72 percent, compared with 75 percent instate schools; these highpercentages contrast with the averagepassrate o f only 47 percent inprivate schools. Iti s noteworthy that, although the average for libre subsidik schools i s lower than that for the state schools, no school in the libre subsidik group achieved a pass rate below 12.5 percent, whereas the weakest state school had a pass rate o f only 1.6 percent, an achievement that was even worse than the pass rate o f 2.1 percent inthe weakest private school. Note also that, inall three types o f schools, there was at least one school that achieveda pass rate of 100percent. Table 6.12: Resultsonthe national examination at the end of the tronc commun cycle, Rwanda, 1999-2000 Libre subsidib a. The sample includes all the secondaryschools that fielded candidates for the tronc commun end-of-cycle national examination in 1999. The pass rate is basedonthe numbero f schools shown inthe secondcolumn, whereas the averagescore is basedon one fewer schoolinthe categories for libre subsid2 andprivate schools. b. Referstothe percentagemeetingthe requirementfor entry to the uppersecondarycycle. c. The scoring system uses a scale that translates letter grades into numerical scores, with As scoredat 11points, Es score at 0, and letter gradesinbetweenscoredprogressivelylower by one point. Source: Computedfrom unpublishedschool-level datasuppliedby the NationalExamination Council ofRwanda. 6.41 The ranking by school type changed somewhat when performance is measured using average scores instead o f pass rates: both state and libre subsidik schools outperformed privates schools, but state schools now ranked behind libre subsidik schools. The scores span the widest range across private schools--from a low o f 0.9 to a high o f 6.1-implying that the sector hadthe worst as well as the best school inthe system. Inthe public sector, the worst school was as likely to be a state or libre subsidik school, butthe top performer tended to come from the libre subsidik group. 116 6.42 Regressionanalysis o f the correlates o f examination performance. Data for the analysis were prepared by merging school-level data files supplied by the Rwanda National Examination Council; files containing informationon eachschool were basedon the census o f secondaryschools in 1999-2000, and those containing information from the 1999 censuso f teachers were aggregatedto the school level. The procedure produced a dataset of 137 public sector schools (state and Zibre subsidik schools): private schools were eliminated, because informationwas lacking on unit costs; some public sector schools also dropped out becauseo fmissingdata. 6.43 Figure 6.5 below gives an overview of the relation between performance in the examination and the cost of personnel per student. As in primary education, the patterns shows substantial disparity intwo senses: (a) consistentwith what has been documentedearlier, spendingper student varied widely, going from 10,000 FRw to as much as 40,000 FRw, and (b) among schools with similar levels o f funding for personnel, performance inthe national examinations was also highly dispersed. Among schools where per student spending on personnel was about 20,000 FRw, the pass ratesrangedfrom about 20 percent to 100percent. Figure 6.6: Relationbetweencost ofpersonnelper student and end-of-tronc commun cycle nationalexamination resultsacross public secondary schools, Rwanda, 1999 Pass rates Averagescore 0 100- 0 ' 80 - 60- aE 40 - 0 0 0 I ! I I I 20 , I I 10 20 30 40I 10 30 40 Costperstudent (1,000s ofFRw) costperstudent('OOOFRW) Source: BasedondatasuppliedbyNationalExaminationCouncilofRwanda,mergedto MINEDUC's 1999 census ofteachers 6.44 We turnnow to the regression findings. The full results appear inappendix table A6.4. A key finding is that the coefficient on spendingper student is not statistically significant, confirming what was hinted at in the preceding graphs. One interpretation i s that school performance on the national examinations depended more on how resources were used than on how many resources they received. To look into the issue more closely, we also regressed examination results against some aspects o f the learning environment. Because o f data limitations, we were able to consider only two such aspects, namely, the student-teacher-ratio and the educational attainment o f the teachers. Do these variables relate to student performance on the national examinations? Ifso, what is the nature o f the relationship? Table 6.13 assembles highlights from the full regression results to answer these questions. 117 Table 6.13: Correlates of school-levelperformanceon the nationalend-of-tronc commun cycle nationalexamination, Rwanda, 1999 Dependent variable & regression estimates Score` Regression variables ;ample mean Passrate (%)b I (fulllist appearsintable A6.4) Ifregression sample mean= 73.0) variablea mean=3.3) Coefficient Marginal impactC Coefficient Student-teacherratio Reference group: Lessthan 18 23.6 - - - 18-27 48.5 -0.47 -6.9 -0.28 * >21 27.9 -0.93* -13.7 -0.22 Teacher qualification Reference group: Universitv degree (Le., License, BA, BSc. Masters, Ph.D.) 8.4 - - - Post upper secondary (BAC, or 2 yr. 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 % teacherswith pre-serviceteacher training 67.9 0.01 I 0.13 -0.004 - 0.20 0.3 1 Number o f observations - 136 137 Note: - denotes not applicable. One star (*) denotes that the underlying coeff ent estimate for confidence level; unmarked estimates indicate that the underlyingco&cients are statistically not significant. See app&dG table A6.4 for the full regressionresults. a. Correspondsto percentagesinthe indicatedcategory. b.The regressionmodelfollows a log-logit specification (Le., In[y/(l-y)]=bX)). The marginal effect onthe student-teacher ratio refersto the percentage point change in the pass rate in response to switching the indicated ratio for the ratio in the reference group; the marginal effect on the teacher qualification variable refersto the percentagepoint change in the pass rate inresponse to a 1percentagepoint increase inthe share ofteachers with the correspondingqualificationat the expenseof a 1 percentagepointdecline inthe shareof teachers with referencequalification (Le., university degree). Source: Authors' estimates based on school-level data on examination results for 1999, merged with data from MINEDUC's 1999-2000 census of primary schoolsandthe 1999censusofteachers. 6.45 Consider first the results pertaining to the relation between pass rates and the student- teacher ratio. The negative coefficients estimates on this variable imply that, relative to students attending schools averaging less than 18 students per teacher, those in schools characterized by higher ratios tended to fare worse on the examinations. The detrimental 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 examination score as the dependent variable, but the regression coefficient is statistically significant only for the dummy variable corresponding to a student-teacher ratio in the 18-27 range. These results suggest that high pupil-teacher ratios had probably compromised student learning as measured by examination results. The sample size and quality of the data were such, however, that it had not been impossible to determine with any precision the exact thresholds above which the negative effects begin to take hold. 6.46 Consider next the findings pertaining to the relation between teacher qualification and examination results. Students taught by teachers without a university degree did worse, as the negative coefficient estimates suggested, but none of the coefficients was statistically significant. The result remained the same regardless of which dependent variable was used. These results do not mean that teacher qualification is never an important factor in determining students' academic performance. 118 Rather, they imply that inthe current context of Rwanda's secondary schools, other factors, including possible overcrowding in some schools and lack of adequate learning materials and other facilitating conditions, probably constitute an even more bindingconstraint on performance. Policyimplications 6.47 Taken together, the findings in this chapter highlight the following issues for attention in policy development: (a) treating the trunc cummun and upper secondary cycle separately in developing a strategy for expanding enrollments, (b) managing the cost o f expansion, and (c) ensuring adequate conditions for effective learning inboth cycles o f the subsector. 6.48 Developing separate expansion strategies for the trunc cummun and upper secondary cycles. In Rwanda, the trunc cummun cycle offers a standard curriculum to all students, whereas the upper secondary cycle consists of four streams within each of which additional fields of specialization are offered. The system's current design is therefore consistent with the view that the lower secondary cycle is a part of basic education-in other words, an extension of primary schooling to complete a child's preparation for adulthood. Incontrast, the mission of upper secondary education is to prepare students eventually for jobs at the relatively skills-intensive end of the labor market, with some of the students reaching these jobs only after going on to higher education. 6.49 Accepting this distinction between the two cycles implies that their expansion 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 that in the upper cycle would need to be calibrated to the demand for educated labor with significant years of educational attainment. Like most modernizing agricultural economies, the demand for such labor inRwanda 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 criterion of success than the fact that graduates leave upper secondary school with the right skills to succeed inthe next phase of their lives. For those going on to higher education, these skills must lay a solid foundation for further studies; for those going on to jobs, these skills must satisfy employers' demand for skilled labor and enable graduates to perform theirjobs well. 6.50 Managing the cost of expansion. Two important aspects of the secondary school system make both cycles costly to operate at present. The first is that almost all secondary schools in the public sector+omprising state and libre subsidik schools-provide boarding facilities. Although some of the 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 trunc cummun level, are accommodated inday schools. 6.51 The second aspect is that a large number of schools currently enroll too few students to benefit from scale economies in service delivery. At the trunc cummun 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 inan expanding system is relatively straightforward, because it rarely requires collapsing existing schools into single institutions. Instead, what is requiredis a planfor expansion that (a) enlarges enrollments in the trunc cummun 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 trunc cummun students. Adding on trunc cummun classes to existing primary schools is a possibility, but the arrangement calls for carehl evaluation, given that most primary schools are already very large and the benefits o f combining the two cycles under one roof are unclear. Indeed, ifthe experience from the 119 current practice in secondary education o f combining the tronc commun and upper secondary cycles under one roof is any guide, the result may well be to encourage an uneconomic size distribution o f enrollments at both levels. 6.52 At the upper secondarylevel, enrollments tend to be small across schools as well. The fact that schools also seek to supply a wide menu o f course offerings only exacerbates the problem. Possible approaches to managing costs insuch a system includes the following: (a) reduce duplication incourse offerings across schools, (b) consolidate related fields of specialization where feasible, and (c) accommodate future growth inenrollments inexisting schools rather than by buildingnew ones. A related cost-saving measure is to consider shortening the duration o f some o fthe courses. Currently all courses last 3 years, but in some fields, especially in the vocational stream, learning on-the-job may well be more effective than classroominstruction inhelping students succeedintheir future jobs. 6.53 Ensuringadequateclassroomconditions to sumort student learning. Our findingspoint to at least three constraints that may be compromising the classroom learning environment. The first is that class sizes, particularly in the state schools, may be too large. A recent policy to increase a teacher's weekly teaching workload to a minimum o f 25 hours (as compared to the practice o f some 17 to 19 hours in 1999) may have helpedto ease the problem. But giventhe small enrollments across schools and the fragmentation incourse offerings discussedabove, it i s unclear how well the policy i s being implemented and how effective it has in fact been in ensuring a better allocation of teachers' time. 6.54 A related constraint hasto do with the consistency ofteacher allocation across schools. The number o f teachers allocated relates positively to the size of enrollments, but the relation is not especially tight, implyingthat teachers are more available insome schools than inothers. The unequal availability o f teachers i s a worse problem among Zibre subsidik schools than among state schools, suggesting that tighter managementof teacher allocation across those schools may warrant especially close attention. 6.55 Inaddition, the issue of teacher qualification cannot be ignored. Inthe tronc commun cycle, most o f the teachers satisfy the minimumqualification-i.e., at least an upper secondary school diploma; some o f them have even earned a university degree. In contrast, about 50 percent o f the teachers in the upper secondary cycle-in state and Zibre subsidit! schools-are very probably underqualified, having only an upper secondary school diploma themselves. Any strategy to improve learning conditions must therefore seek to raise the educational profile o f the teachers in the upper secondary cycle. The obvious long-term solution is to set clear standards for teacher recruitment, but if the problem i s to be solved more quickly, policymakers might consider rationalizing the current allocation o f the most qualified teachers (i.e., those with post-secondary qualifications) between the tronc commun andupper secondarycycles. Conclusion 6.56 Secondary education inRwanda i s inmany ways at a crossroads today. The risingtide of primary school leavers i s 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 i s thus slowly being transformed from a system serving a small elite clientele into one that will increasingly be expected to cater to the masses. The challenge is to ensure that appropriate policies are put inplace for the system to expand along an efficient and equitablepath. 120 6.57 Initial conditions are important to establish as a basis for assessing where and how the government might usefully intervene. This chapter has therefore sought to document some key characteristics of the sector and the current arrangements for service delivery. The findings point to three areas for attention inpolicy development: developing separating strategies for the future growth ofthe lower and upper cycles, managing the cost of expansion bytaking advantage o f scale economies in service delivery and minimizing fragmentation of course offerings, and managing teacher deployment and their teaching assignments to promote conducive leamingenvironments. Although the full range o fproblems that policymakers face inthe sector is necessarily broader than these concems, they cannot be ignored because o f their impact on costs and therefore on the success o f 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. 121 Chapter 7: HigherEducation 7.1 Higher education inRwanda expanded rapidly inthe post-genocide years. Since 1997, three new public institutions have been created, the number o f government-sponsored students has risen by nearly 2.5 times, and the public budget for the subsector has grown by a massive 3.4 times, reaching its current level o f almost FRw 12billion (or nearly US$27 million). The subsector currently serves a small population, but absorbs nearly 40 percent of the country's recurrent spending on all education services. Given the obvious imbalance, it is no surprise that the future development o f higher education has been the subject o f recent policy documents tabled for discussion before the Cabinet.82 7.2 In this chapter, our purpose is to consolidate the available quantitative information pertaining to higher education to illustrate in a concrete manner the nature o f some o f the key challenges, as well as to enrich the factual basis for policy discussion. It i s appropriate at the outset to alert the readerthat, becausethe system for data collection i s still beingrebuiltfollowing the genocide, much of the information presentedbelow has been gathered directly fiom the institutions themselves inthe context ofthe presentstudy. Although carehas gone into ensuringconsistency andreliability in the data at the time they were gathered, it i s possible that some o f the information may need to be updated inlight o f ongoing improvements indata collection. Overall, however, the potential flaws in the data are not so serious or so pervasive as to alter the thrust o f the policy issues to which they call attention. The second caveat is that the treatment here does not pretend to be exhaustive, but focuses on the following areas: the structure o fthe system and the institutional composition o f enrollments, the distribution o f enrollments by level and field o f study and the direction o f government sponsorship for overseas studies, staffing patterns at the institutional level, the cost o f service delivery and arrangementsfor student finance, and the pattern o f student flow and output o f graduates. 7.3 The analysis presented below highlights several areas that warrant attention if the sector's high costs are to be brought under control. The first pertains to the current system o f blanket subsidies under which practically all students who gain entry into public institutions automatically receive a bursary and access to subsidized welfare services. Such a system has become increasingly unsustainable with the more than fourfold increase in enrollments in the public sector in the past decade. Government sponsorship o f overseas studies (which accounted for a quarter of the recurrent spending on higher education) is another area that needs to be reassessed. The cost of such studies is high, whereas local options have become increasingly available, particularly for the undergraduate level courses that the large majority o f overseas scholarship-holders are in fact pursuing.Attention to cost management in the public institutions i s a third area for scrutiny. Staffing ratios in the public institutions are generous compared with similar activities inthe private sector and are one reasonwhy the cost o f service delivery i s so highinthe public sector and why so little o f the spending appears to be left for nonpersonnel inputs to support effective teaching. In addition, the findings suggest that policymakers need to pay attention to the management of student flow as well, addressingissues that pertain not only to patterns o f dropping out and repetition, butalso to the overall size of the intake and its distribution by field of study. 82Specifically, Rwanda's donor partners have sponsoredthe preparationo f three policy documents: "Draft Policy Framework 2000," a paper that examines the changingnature ofhigher education ingeneral and the major challenges facing the country; "Proposed Higher Education and Continuing Education Bill/The Higher and Continuing EducationAct 2002," a paper that sets forth new arrangements covering all aspects of planning, management, governance, and funding o f higher education; and "Organization o f Research, Science, Technology, Higher, and Continuing Education," a paper that outlines an organization structure to ensure that the Ministry o f Educationcan effectively fulfill its proposed roles and functions inhigher education. 122 Structure of the system 7.4 Higher education in Rwanda today is made up of a diverse mixture of public and private institutions. Below we examine its growth since colonial times and document its current institutional composition andthe distribution of enrollments by field of study. 7.5 Historical context. Table 7.1 shows the number o f public and private institutions and size o f their enrollments since the 1960s. The first institution o f higher education-the diploma- granting Grand sbminaire de Nyakibanda (GSN)-was established in 1936 by the Roman Catholic Church specifically to train menfor the prie~thood.~~For most o fthe nextthree decades, the seminary was the only institution of higher education in Rwanda and students who were not training for the priesthood pursuedtheir higher education inpresent-dayDemocratic Republic o f Congo or Belgium.84 The situation changed in 1963 with the founding o f the Universitb nationale du Rwanda (National University of Rwanda or UNR) under a joint agreement between the Rwandese govemment and the Dominican Fathers from Quebec Province in Canada.85In 1966 the Institut pbdagogique national (National Institute of Education or IPN) was established with support from the United Nations Development Programme (UNDP) and United Nations Educational, Scientific and Cultural Organization (UNESCO); its mission was to deliver secondary-school teacher training and to conduct researchineducation. Inthe private sector, the Facult6 de thbologie de Buture (FTB) was established in1969. 83See Mazimpaka and Daniel 2000 and Linguyeneza 2001. The latter gives an account o f the pioneering work of the Roman Catholic Churchinthe field ofeducation at all levels. 84Apart from the GSN, two other private sector institutions-the Grande skminaire de Kabgayi (GSK) and the Centre d'enseignment supkrieur de Kigali (GESKFwere established, but the year of establishment o f these institutions are not known. For similar reasons, we also exclude two public sector institutions, Ecole supkrieure militaire (ESM) and Centre deformation des adjoints techniques de la statistique (CFATS), from the discussion here. 85Although a public institution, UNR was, until 1974, runby the DominicanFathers (Linguyeneza 2001). The university then consisted o fthe Faculty o f Medicine, Faculty o f Economicsand Social Sciences, and Teacher Training College (Ecole normalesupkrieure). See also NationalUniversity ofRwanda2002a for additional historical information. 123 Table 7.1: Numberofhighereducationinstitutions,overallenrollments,and privateshare of students, Rwanda, 1960sto the present Period Numberof institutionsa Enrollmend Public Private Total (1,000s) %private Pre-1963 0 1 1960s 1 1 Post-1963' 2 2 2 3 1 1 3 1-3 8-28 Pre-genocidee 3 7 3 - 5 35 1990s Post-genocide 6 6 5-11 6 28 - 2000-present 6 6-8 12- 17 30 38 - 7.6 The 1970s saw no change inthe number o f public institutions, whereas in the private sector, one new institution, the Institut Afi-icain et Mauricien de statistiques et d'kconomieappliquke (IAMSEA), was establishedin 1976.86The following decade witnessed additional development inthe public sector, including the incorporation of the I"into UNR in 1981 and the establishment of the Institut supkrieur de finances publiques (ISFP) in 1986 and the Institut supkrieur d'agronomie et d'klevage (ISAE) in 1988-89. Inthe private sector, the number of institutions doubled from three to six with the addition o f Universitk Adventiste d'Afrique Centrale (UAAC) in 1984, Ecole supkrieure de gestion et d'information (ESGI) in 1985, andInstitut supkrieur Catholique depddagogie appliquke de Nkumba (ISCPA) in 1986. 7.7 The biggest change to the structure of higher education took place in the 1990s, a decade marked by explosive growth in the public sector and large-scale replacement o f existing private institutions with new ones. The number o f public institutions grew from three to six with the founding o f the Kigali HealthInstitute (MI)in 1996, the Kigali Institute o f Science, Technology and Management (KIST) in 1997, and the Kigali Institute o f Education (ME) in 1999.87In the private sector, the Institut supkrieur depkdagogie de Gitwe (ISPG) was established in 1993, only to close the following year. After the genocide in 1994, three more private institutions-the IAMSEA, ESGI, and ISPCA-went out o f existence. The FTB and UAAC reopened in 1996 and the ISPG in 1997. Two new private sector institutions-Universitk libre de Kigali (ULK), Universitk lai'que de Kigali (UN1LAK)-were established in 1996 and 1997, respectively. Overall, the number o f private institutions dropped from sevenbefore the genocide to six afterward. 86However, new faculties were established at UNR during this decade, the Faculties o f Law and o f Applied Science in 1973, and the Faculty ofAgronomy in 1979. *'Apart from the increaseinnumber o fpublic institutions, significant changes also occurred at UNR the School o f Modem Languages, which was closed in 1973, was reopened in 1994; the School o f Information Sciences and Technology (Ecole des sciences et techniques de I'information or ESTI) was established in 1996 (Kereni 2002; National University o f Rwanda 2002b); and the Faculties of Science and Applied Sciencemerged in 1998 to form the Faculty o f Science and Technology (Bonfils 2002). 124 7.8 In terms of enrollments, the number of students and their public-private distribution tracked the foregoing changes in the number of institutions. In the 1960s, the system as a whole enrolled perhaps a 100 or so students; UNR, for example, began with only about 51 students. Inthe 1970s, total enrollment stood at about 1,000, but by the end of the 1980s, the number had risen to 3,000. Because the establishment o f private institutions mainly fueled growth during the 1980s, the share o f students in the private sector rose from just 8 percent at the start o f the decade to nearly 30 percent by the end o f the period. Further expansion of the system in the early 1990s saw enrollment rising to about 5,000 and the private sector's share rising even more, reaching about 35 percent just before the 1994 genocide. In the post-genocide 1 9 9 0 ~enrollments more than doubled, reaching ~ 11,000 by the end of the decade. The private sector share o f enrollments fell inthe years immediately following the genocide, because o f the closure of several institutions, butbegan regaining its share as new institutions were established. In the early 2000s, enrollments continued to grow throughout the system, reaching nearly 17,000 by 2001-02. Because growth had been faster inthe private sector, its share o f enrollments inthat year was at a historical highof 38 percent. 7.9 The institutional composition of enrollments. Table 7.2 shows the relevant data for 2000-01, the most recent year for which we have complete data. They show that enrollment in the system i s in fact concentrated in a small number of institutions. Inthe public sector, UNR accounted for 56 percent o f total enrollment, followed by KIST with 18 percent and KIE with 11 percent, whereas the remaining 15 percent o f the students were distributed across three other public institutions. The degree o f concentration i s even greater in the private sector: among the four main institutions in this sector, ULK alone enrolled 81 percent o f the students, whereas UAAC accounted for 11percent, UNILAK7 percent, andISPGonly 1percent.88 The other four institutions-the GSN, FTB, GSK, and C E S K d a c h with fewer than 200 students, are excluded fromthe table for lack ofdata. 125 Table 7.2: Institutionaldistributionof higher educationstudents, share offemalestudents, and examinationscore of senior six entrants, Rwanda, 2000-01 I %female 1 Average score %share of among senior six Sector andname of institution students in students identifying subsectora Mid-1980s Circa 2000 institution as first 1 Universite'nationale choiceb du Rwanda (UNR) 55.5 17.7 23.4 3.6 Kigali Institute of Science andTechnology &management II 18.2 28.8 3.5 Kigali Institute of Education(KIE) 11.0 29.3 3.4 2 0 Kigali Health Institute (KHI) 7.9 48.7 3.7 Institut supe'rieurd'agronomie et d'e'levage(ISAE) 6.0 18.4 3.5 Institut supe'rieurdesjhancespubligues (ISFP)' 1.4 34.8 All six institutions 1 100.0 (8.7291 Universitk libre de Kigali (ULK) 80.7 Universite'Adventiste d 'Afrigue Centrale (UAAC) 11.0 0 .-2 2> Universitk lai'que de Kigali (UNILAK) 6.9 Institut suptrieur depkdagogie de Gitwe (ISPG)g 1.3 All four institutions 411ten institutions (12,757) {ore; adashdenotes no data areavailable. a. Figuresinparenthesesreferto the total numberofstudents. b.The scorerefersto the mark ona IO-point scale that students obtainedon the nationalexaminationadministeredat the endofthe uppersecondarycycle. c. Enrollmentsexcludepart-time studentsfor which little systematic dataexist. d.The subtotalandoverallfigures referto weightedpercentages. e. The following private institutions, all with probably very limited enrollmentsnot exceeding 100 to 200 students each, are excludedhere for lack o f enrollment data: Grandsdminaire deNyakibanda, Facultd de thdologiede Butare, Grandsdminairede Kabgayi, andCentre d 'enseignement supdrieur de Kigali. f.NowrenamedastheInstituteofFinanceandBanking(IFB) g. Reestablished inDecember2002as Ecole des hautes dfudesdesfinanceset d'konomie appliqude. Source: Appendix table A 7 2 for distribution of enrollments; personal communication from the National Examination Council of Rwanda for data on examination score. 7.10 To complete the description of the overall composition of enrollments, the foregoing table also shows women's share o f total enrollments as well as one indicator o f the quality o f the intake across public institutions. To the extent that trends in UNR apply to the system as a whole, women's access to higher education appears to have improved with time. Inthe mid-l980s, women accounted for less than 18 percent o f total enrollment inUNR, but their share had risen to more than 23 percent by around 2000. Inthe latter year, their share inthe system as whole (i.e., inthe 10 main public and private institutions taken together) was about 33 per~ent.~' The average, however, masks significant differences inthe two sectors: although women accounted for nearly half o f enrollments in the private institutions, their share was only 26 percent in the public sector. A closer look reveals rather consistent shares, in the 49 to 56 percent range, across the private institutions, whereas they vary significantly across the public ones, from a low o f 18 percent at ISAE to 49 percent at KHI.~' One explanation for these differences may be found in the stiff competition for places in the heavily subsidized public sector, coupled with the fact that girls tend to perform less well than boys on the senior six examination, the results of which serve as a basis for selection into the public institution^.^^ 89 Noteworthy is that the rise in women's share in total enrollment has been brought about mainly by their enrollment in private institutions. Their shareo f enrollmentinthe public institutions appears to have leveled off since the mid-1990s (table A7.2). One possibleexplanation for the highshare at the KHImay be becausethe institution offers the traditionally female-dominated course innursing. See table 4.8 for dataonthe differences inexaminationresults betweenboys and girls onthe senior six examination. 126 The private institutions are less selective, presumably as long as the applicants satisfy a minimum level o f competence, and cater to those who are able and willing to pay the fees. Inaddition, interms of the quality of student intake, the available data pertain only to the public institutions. The variation is modest, but it suggests that the brightest students tend to rank UNRand KHIas their top choices. 7.11 Distribution of enrollments bv field of study. The distribution of students by field of study in 2000-01 appears intable 7.3. Overall, the pattern is heavily biased toward the humanities and social sciences, which enroll nearly three-quarters o f the students ~ystemwide.'~Within these fields, there is also a clustering o f enrollments in economics and management (with a share o f 44 percent) and in the area of literature and the humanities (with a share of 27 percent). In contrast, only 28 percent of the students were enrolled inthe applied and natural sciences; among them, only just more than one-half were pursuing courses inscience and technology. The distribution of enrollments shows the expected differences between public and private institutions, with almost exclusive focus among the latter in the humanities and social sciences. It is particularly interesting to note that in terms of absolute enrollments, the private institutions overshadow the public institutions by a factor of more than three in the field of law, whereas the size of enrollments is roughly comparable in literature and the humanities and ineconomics and management. Table7.3: Number anddistributionofstudentsinpublic and private highereducationinstitutions,Rwanda, 2000- 01 I 1 Public institutions" Private institutionsb I Bothtypes Fieldof study No. of ~ students 6 distribution tion Literatureand humanities 963 13.6 19.2 54.6 Law 338 4.8 12.4 75.3 Economics and management 1,876 26.6 31.9 47.0 Teacher education 940 13.3 9.0 5.4 Science and technology 1,475 20.9 13.9 4.4 Health sciences 870 12.3 8.3 5.2 Agriculture 599 8.5 5.4 0.0 Total in specialized courses 7,061 100.0 100.0 36.3 Pre-specializedcourses 1,633 0.0 Overall total 8,694 I - Note: A dash denotesnotapplicable. a. Includesenrollments inthe following institutions: Universite'nationale du Rwanda (UNR), Kigali Instituteof Science and Technology andmanagement (KIST), Kigali Institute of Education(KIE), Kigali Health Institute @HI), Institut sup6rieur d'agronomie et d'devage (ISAE), and Institut supkrieur des financespubliques (ISFP), nowrenamedas the Institute of FinanceandBanking(IFB). b. Includes enrollments in the following institutions: Universitk libre de Kigali (ULK), Universite' lai'que de Kigali (UNILAK), Universite' Adventiste d'Afiique Centrale (UAAC), Institutsupe'rieurdepaagogie de Gitwe (ISPG), Faculte'de the'ologiedeButare (FTB). Source: See appendix table A7.3 7.12 Trends inenrollments by field at UNR. The available data permit a closer look at how the distribution o f enrollments at UNR has evolved as the context has changed, from one inwhich the university was practically the only significant institution to one marked by a revitalized private sector and an expanded public field. Ingeneral, the evidence points to substantial changes that are suggestive of an institution inthe process of adapting to its new surroundings. There is evidence, for example, o f experimenting with the creation of new faculties, reconstitution of existing faculties and schools, and 92 Humanities and social sciences include fields of study classified under codes 01 to 38 and 76 to 86 under the coding system of International Standard Classification o f Education (ISCED); applied and natural sciences refer to ISCED 42 to 72 (see UNESCO 1997). For a breakdownof enrollment by field ineachofthe six public and four mainprivate institutions, see table A7.3. 127 reorganization of course offering^.'^ These changes are perhaps best seen as part of the institution's ongoing search to achieve a good balance insupplying the graduates to meet the county's felt need for highlyqualified labor (National University ofRwanda2002a; Rwamasirabo 2002). 7.13 W e now examine the specific changes that have occurred since the 1980s (table 7.4). The first change i s the large share o f students enrolled inpre-specialized courses inthe post-genocide era. Fully a quarter of the students are now enrolled in such courses, reflecting the new policy of requiring students to achieve bilingualism as part of their university studies. Setting aside these students, the pattern of enrollments shows dramatic shifts between the broad field 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 post-genocide 199Os, enrollment shifted heavily in favor of the former group, raising its share to 74 percent, a gain of 16 percentage points over the level inthe 1980s. The concentration inthe humanities and social sciences has since declined, but only marginally to 72 percent. Given the country's interest in linking into the global economy through increased capabilities in the sciences, it might be appropriate to evaluate the current patterns o f enrollments across disciplines. Inparticular, given the emergence o f the private sector, the question is whether the university is indeed offering courses that would not otherwise be available. Table 7.4: Trends inenrollmentsandtheir distributionby field at the Universiti nationaledu Rwanda, 1980s, 1990s, and early 2000s Average annual enrollmentsa - %distribution of enrollments Fieldof study Post- Post- 1980s genocide Early 1980s genocide Early 1990s 2000s 1990s 2000s Humanities and social sciences 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 Subtotal 843 2,467 2,927 58.3 73.9 71.7 Applied and naturalsciences Medicine and health sciences 143 440 430 23.8 50.6 37.3 Science andtechnology 333 305 623 55.4 35.1 54.0 Agronomy 126 125 100 20.9 14.3 8.7 Subtotal 602 870 1,153 41.7 26.1 28.3 Total inspecialized fields 1,445 3,337 4,079 100.0 100.0 100.0 Pre-specialized coursesb 0 730 1,302 0.0 18.0 24.2 Overall total 1,445 4,067 5.381 Nore: adashdenotes not applicable. a. The threeperiodsrefer, respectively,to 1982-86, 1994-99,and2000-2002; percentagesineachmajorset ofdisciplines sumupto 100percent. b. Refersto students incommon corecourses, languageupgradingor otherpreparatorycourses. c. Expressedas apercentageo ftotal enrollments. Source: See appendix tableA7.4 7.14 Major shifts have also occurred within each of the two broad areas discussed above. The reorientation 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. 93 See appendix table A7.4 for moredetailed evidence inthis regard. 128 Although in the 1980s, students studying arts and human sciences accounted for 30 percent of total enrollment inthe social sciences, their share fell to less than 16 percent inthe post-genocide years; in contrast, the proportion enrolled in economics and management rose from 37 percent to at least 45 percent. In education and law, the trends show an interesting variation. The share of enrollments in education fell in the post-genocide 199Os, perhaps reflecting a period o f consolidation and reorientation in course content, but by the early 2000s, the field had bounced back to claim nearly a quarter of the total enrollments inthe social sciences, somewhat higher than its share inthe 1980s. In law, the years following the genocide saw an explosion o f enrollments, lifting its share of all students in the social sciences to nearly a quarter, but the increase was short lived. By the early 2000s, law students made upjust more than 12 percent of the total in the social sciences, less than their share in the early 1980s. This seems to be one field where the university is reducing enrollments in absolute terms as similar offerings have become available inthe emergingprivate sector. 7.15 Within the applied and natural sciences, there is also evidence o f major changes inthe pattern of enrollment. Fewer and fewer students are studying agronomy, with the result that their share of all students in the hard sciences stood at less than 9 percent in the early 2000s, down from 21 percent in the early 1980s. The enrollment share of medicine and the health sciences shot up in the post-genocide 1 9 9 0 ~ ~ accounting for half the enrollments during the period, but by the early 2000s, it had fallen to just more than 37 percent of the total as the number of students stabilized at just under 450. Inscience and technology, enrollments picked up following a period of relative stagnation during the post-genocide 1990s. The lull is similar to that in education described above, suggesting a pre- expansion period marked by consolidation and reorganization in course offerings in light of the evolving context of the sector. Students on overseas government scholarship 7.16 Rwanda has a long tradition of students studying abroad on government scholarships. Inthe past, students had to be sent abroad because some types of training were simply not available locally. The situation has now changed with the explosive growth of higher education inrecent years, yet students continue to be sent abroad. A s chapter three has documented, each student on an overseas government scholarship costs the state nearly three times as much as a student studying at a domestic public institution. The number o f students involved and the cost of overseas studies are such that the government has inrecent years had to allocate about a quarter of its total recurrent public spending on higher education to finance the students it sponsors overseas. Because of the highcosts involved, it is of interest to review key information on the size of enrollments abroad and their distribution by field, level of study, and host country.94 7.17 Number of students involved. The relevant data appear in table 7.5, summarizing the trends since the late 1960s. The number o f overseas scholars rose steadily from 220 inthe late 1960s, to 902 by the start o f the current decade before falling to 646 by 2001-02.95The increase inthe early years, coupled with slow growthindomestic enrollments, meant that an average of between 33 and 44 percent of all higher education students were studying abroad in the two decades leading up to 1986. Inthe post-genocide era, their share had dropped steadily each year since 1999, reaching a floor of only 6.0 percent by 2001-02. This decline is mainly the result of the explosive growth of higher education within the country. 94The information on Rwandese students studying abroad does not include students financed privately; the number of such students is likely to be very small, however. 95The count excludes the twenty students studying on government scholarship at the localbranch of the Universitk udventiste d'Afrique centrale. 129 Table7.5: Rwandesehigher education studentson overseas governmentscholarships, 1967to 2002 Students abroad as a Period or year Numberof students abroada % female percentage o f total foreign and domestic enrollments 1967-69 220 39.8 1970-79 504 43.6 1980-86 690 8.9 33.1 1999-00 902 31.0 11.3 2000-01 877 28.3 9.5 2001-02 646 26.9 6.0 7.18 The data on women's participation in overseas studies are available for the 1980s and later years. They show a significant improvement with time: inthe 1980s, women accounted for less than 10 percent o f the students studying on overseas government scholarships; inthe period between 1999 and 2002, their share average just under 30 percent. This proportion i s similar to the share o f women indomestic public institutions. 7.19 Distribution o f overseas scholars by field and level study. Consider the data in table 7.6, which show the distribution in 1984-85 and in2000-01. Excluding scholars for whom the field o f study is unknown or undefined, 76 percent of the students on an overseas scholarship were pursuing courses inthe applied and natural sciences. By 2000-01, the share had declined to 57 percent, a drop of nearly 20 percentagepoints. The retreat from the applied andnatural sciencesmirrors a similar shift in the domestic system, but the share of overseas scholars in these disciplines continues to remain relatively high: 57 percent compared with the corresponding shares o f 42 percent inthe public sector and 10percent for all domestic students taken as a whole. 7.20 Inbothbroad fields, the trend has been toward increasing concentration inthe pattern o f enrollments. Nearly three-quarters o f the overseas scholars in the humanities and social sciences were following courses ineconomics, management, or business in 2000-01, 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 o f this domain fiom 61 percent inthe mid-l980s, to nearly 80 percent in 2000-01. These shifts are again consistent with the trend in the domestic system, with increasing shares o f the students enrolled incourseswith an explicit orientation toward future jobs. 130 Table 7.6: Numberand percentagedistribution of Rwandesestudents on overseas governmentscholarshipsby field and levelof studies, 2000-01 I 1984-85 2000-01 ! Field of study No. of Percentage No. of students by level of studies Percentage students, distribution 3-4 year iistributior all levels byfielda 2-year Masters diplomab under- 411levels byfield" graduate' and PhD Humanitiesand socialsciences Law 17 11.4 0 28 3 31 9.0 Economics, Management and Business 69 46.3 10 220 21 251 73.2 Literature, arts and social studies 63 42.3 1 43 17 61 17.8 Subtotal 149 23.8 11 291 41 343 43.3 Applied and naturalsciences Medicine andpharmacy 81 17.0 6 61 13.6 Science andtechnology 288 60.5 8 350 77.8 Agronomy, ago-industry, etc. 107 22.5 0 39 8.7 Subtotal 476 76.2 14 382 450 56.7 Otherstunknown 83 11.7 1 97 122 13.3 All fields 708 100.0 26 770 119 915d 100.0 Percentagedistributionby level of study 2.8 84.2 13.0 100.0 Note: A dash denotes not applicable. a. The percentages ofthe two broadfields, humanities and socialsciences, and appliedandnaturalsciences sum up to 100 percent;the percentages for each individualfield sum upto 100percentwithin each o fthese two broadfields. b. Leadingto a Buchelier certificateinFrancophonesystems. c. Leadingto aLicence certificateinFrancophonesystems or a bachelor's degree inAnglophone systems. d. Includes 38 studentson govemment scholarshipstudying at the localbranch ofthe UniversitCAdventisred Yfique Centrule(UAAC). Source:Basedonpersonalcommunication from the Direction de I'enseignementsupbriar ofthe Ministry ofEducation for data for 2000-01, and on Govemment ofRwanda (1986) for data for 1984-85. 7.21 The foregoing table also offers information (for 2000-01 only, however) onthe level of the courses pursued by overseas scholars (see bottom row o f the table). A very high 87 percent o f them were enrolled at the undergraduatelevel or lower. Giventhe highcost o f education abroad, it i s important to determine the extent to which these students are pursuing training that i s indeed unavailable in Rwanda. Given that 43 percent of the students were studying in the humanities and social sciences and o f these 88 percent were studying at the undergraduate level or lower, there i s a highlikelihood that a large proportion of these students could have been channeled to the domestic institutions at amuch lower cost to the government. 7.22 Distribution o f enrollments by host country. Table 7.7 shows the number of Rwandese students on overseas government scholarships by host country. In 1984-85, 45 percent of them were hosted by OECD countries, 43 percent by non-OECD countries outside Sub-SaharanAfrica, and only 12 percent by countries in Sub-Saharan Africa. By the late 1990s and early 2000s, however, the pattern had changeddramatically: the OECD share o f students hadfallen to about 14percent, whereas the share of non-OECD countries outside Sub-SaharanAfrica had risen to 49 percent and that of the countries inSub-SaharanAfrica had risen to 37 percent. These changeshave been accompaniedby an increaseinthe number o f host countries in Sub-SaharanAfrica, incontrast to an opposite trend inthe other two country groupings. 131 ..i Table 7.7: Numberof Rwandesestudents on overseasgovernmentscholarshipsand host countries, 1984-85, and 1999-2002 H_~ Number of Regional Regional ~~ Number of: share of share of RegiodCountry Students countries students (%) No. of students students :ountries (%)" 1984-85 1999-00 2000-01 !001-02b 1999-2002 Dem. Rep. of 23 0 0 0 25 8 10 21 South Africa 0 65 87 94 Uganda 0 141 138 61 41 9 155 99 68 12- 14 89 11 12.6 369 334 244 15-17 37.0 Belgium 77 20 22 15 Germany 70 6 6 7 France 48 25 15 11 n Canada 38 28 28 33 ~ USA 33 18 12 9 Great Britain 1 40 22 17 Other 51 5 14 9 4 2-3 Subtotal 318 11 44.9 151 114 96 8 - 9 I 14.1 Algeria 42 24 31 40 25 25 32 35 0 403 373 231 USSIURussia 207 0 22 15 Other 27 10 9 9 5 2-6 1 Subtotal 301 13 42.5 461 467 326 6-10 1 48.9 irandtotal 708 35 100.0 981 915 666 31-32 I 100.0 lore: Blankcells inthe column for n u :r ofhost E itriesdenotea count ofone lankcells i heother columnsare left emptyto avoidclutter. a. Computedaccordingto the averageenrollments from 1999-2000to 2001-02. b. Includesstudents on overseas govemmentscholarship studying at the Rwandabranchofthe UAAC. Source: See appendix table A7.6. 7.23 Beyond the broad trends, the changes taking place at the country level are also noteworthy. In 1984-85, India hosted no Rwandese government-sponsored students abroad, but by 1999-2002, itwas hostingbetween35 to 45 percent o fall such students, making itthe country with the single largest contingent o f Rwandese students studying on government scholarship in a foreign country. South Africa and Uganda also became important host countries in the same period: fiom hosting not one Rwandese scholar on government sponsorship in 1984-85 to hosting, respectively, between 7 and 15 percent andbetween 9 and 16 percent o f the total number o f such students by 1999- 2002. Among the main OECD host countries, all except Canada have seen a decline intheir share o f government-sponsored Rwandese students. Among the non-OECD countries, the change in Russia's position is especially dramatic: its share fell fiom 29 percent in 1984-85 to a mere 2 percent in2001- 02. 7.24 Differences inthe cost o f education abroad and the cost-sharing arrangementswith the host countries appear to be among the key factors driving the changes noted above. Circa 2002, students studying in Algeria, China, India, Poland, and Russia were on special bilateral programs under which the host country paidthe fees and offered a small grant, while the Rwandesegovernment 132 topped up the grant and paid for the airfare.96 Among the remainingpotential countries, the cost to the Rwandese government o f sending students is much less burdensome inneighboring African countries than in OECD countries. Annual academic fees alone in the latter countries currently range from $11,000 to $14,000, compared with less than $1,200 in South Africa, for example. In addition, the welfare grants for students studying in the OECD countries range anywhere from $8,400 to $14,400 depending o n level o f study, whereas the corresponding range is $3,400 to $4,200 for students sent to South Africa. These cost differences as well as availability of high quality programs in non-OECD countries inrecent years, including inSub-Saharan Africa, explain inlarge part the significant changes inthe destination ofstudents going abroad ongovernment scholarships. Staffingpatternsinpublic andprivatehigher educationinstitutions 7.25 W e turn now to documenting additional aspects of the domestic higher education system, includingthe composition of the faculty by nationality and academic qualifications, as well as the pattern of staffing ratios across institutions. The issue of faculty remuneration is also relevant to policy development inthe sector, but is not addressed here, because the available information is highly fragmented and not easily consolidated inthe time frame o f the present study.97 7.26 Faculty composition by employment status, nationality. and aualifications. The relevant data for four public institutions and the two private institutions with the largest enrollments appear in table 7.8 showing the number of full-time and part-time faculty, the share of expatriates among the full-time staff, and the academic composition of the faculty. At UNR,the number of staff grew by 78 percent between the mid-1980s and 2000-01-fueled by the more-than-threefold increase in enrollments; the number of expatriate staff appears to have grown as rapidly, as suggested by the fact that their share of the full-time faculty has remainedrelatively stable at around 21 to 23 percent during the period. At the other public institutions, the data pertain only to 2000-01,and they show substantial variation inthe dependency on foreign staff, ranging ftom highs of 40 percent at KIE and around 30 percent at KIST and ISAE, to 14 percent at KHI. Inthe private sector, the share of expatriate faculty was 15 percent at ULK, whereas UAAC uses no foreign teachers at all. The generally lower shares in the private sector are to be expected given that foreign teachers are costly and therefore probably unaffordable inlarge numbers based on the fee income that private institutions are able to generate. 96See table A7.7 for additional details. 97The government recently established standardratesby qualification for all foreign faculty, butthe structure andcompositiono f the pay for nationals remains institution specific. The lack o f comparability in faculty remunerationacross public institutions is perceived to be aproblem bysome andis the subject o fongoingpolicy development. 133 Table 7.8: Number and composition of highereducationfaculty by institution,Rwanda, selectedyears I Full-time faculty \lo. ofpart- Qualification of full-time time & faculty visiting Sector, name o f institutionand year or period faculty per full-time faculty Mid-1980s 218 23.2 0.7 40.2' Universite'nationale du Rwanda (UNR) 2000-01 38gd 20.8e 0.6 34.6' 24.0' E.- o Kigali Institute of Science & Technology & management(KIST) 2000-01 162 30.2 0.1 Kigali Institute of Education (KIE) 2000-01 85 40.0 45.3 15.1 Kigali Health Institute(KHI) 2000-01 63 14.3 55.3 Institut supe'rieur d 'agronomie et d'e'levage(ISAE) 2000-01 31 32.3 0.5 6.5 54.8 +- Universite'libre de Kigali (ULK) 2000-01 55 14.5 1.3 Ez Universite'Adventiste d 'Ajiique Centrale (UAAC) 2ooo~01 12 0 1.8 Note: A dashdenotesdata unavailable. a. Distribution for UNR in2000reflects the distribution among nationalstaff only. b. Includesthose with a licence degree or a bachelor ofarts or a bachelor ofscience degree. c. Includesmedical doctors. d. Includes the 22 lecturers under the UnitedNations faculty volunteer programandthe 35 teachers in the Ecolepratiquedes lungues modernes (EPLM), a language trainingprogramfor first-year students to achieve bilingualism inEnglish andFrench. e. Assumesthat the 35 EPLMteachers are nationals. f. Excludes medicaldoctors to renderthepercentagecomparable tothosefor the other institutions. Source: Rwanda 1986 for data on UNR in the mid-1980s and based on a small survey of teachers and their qualification conducted by the Ministry o f Education's Direction de I'enseignement supbrieur. 7.27 Dependency on expatriate faculty i s not the only difference across institutions. The data visitingor part-time staff for every ten full-time staff in2000-01,a ratio comparable to that inthe mid- also reveal a wide range interms o f the use o f part-time and visiting faculty. At UNR,there were six 1980s. At ISAE the ratio in2000-01 was slightly lower than that at UNR, but at KIST, the ratio was drastically lower, at only one visitor per ten full-time staff. In contrast to the situation at the public institutions, those inthe private sector relymuchmore heavily onpart-time staff, so much so that part- timers outnumber the full-time faculty, by 30 percent at ULKand by 80 percent at UAAC. As with the use o f expatriate staff, cost considerations probably lie behind the pattem. Because many o f the part- timers are actually full-time staff at the public institutions, such staff do not necessarily lower the quality o f teaching even though continuity in teaching arrangements may not be fully assured given the nature o ftemporary contracts. 7.28 Inaddition, consider the datainthe foregoing table onthe qualificationofthe full-time faculty. Because o f the difficulty o f establishingprecise categories o f qualification, we have simply grouped the staff under two main rubrics, those with doctorates and those with less than a master's or maitrise degree?' For UNR, two figures appear for 2000-01: the first includes medical doctors inthe count o f staff with a doctorate to render them comparable with the data for the mid-1980s; the second 98 Part of the difficulty is that the same qualificationlabel may refer to different levels of training, depending on when and where a personobtainedthe training. To illustrate, the licence degree inthe Frenchsystemis currently obtained after a 3-year course after a personhas sat for andpassedthe BucculuurPut examination(which is administeredat the endof the secondarycycle). Inearlier years, however, the licence degree is awarded after a 4-year course of study. In the Belgium system, obtaining the licence degree may require4 or 5 years ofstudy. The availabledatado not allowoneto distinguishamongthe various situations. 134 excludes the medical doctors to render the data comparable with those for the other institutions listed inthe table. At UNR,the share of highly qualified staff has diminished since the mid-1980s, but the decline is surprisingly small, given the heavy losses inhuman capital duringthe genocide. Inabsolute terms, only a quarter o f the faculty at the university currently hold doctorates, whereas nearly half of them have only a licence or bachelor's degree (i.e., less than a maitrise or master's degree). The situation at K I S T at the top end o f the qualification range is comparable to UNR's, whereas it i s much worse at ISAE. Among the public institutions, only KIE has an impressively high stock o f doctorate- holders among its staff?' Inthe private sector, faculty qualifications are generally not muchbetter than at UNR: for example, 40 percent of ULK's full-time faculty lack a postgraduate degree, whereas at UAAC, the share is 25 percent. The overall picture, considering public and private institutions, is one inwhich many ofthe staffinhighereducation arebarely aheadofthe studentsthey teach. 7.29 Given the high proportion o f inadequately qualified staff, faculty upgrading is an obvious priority in staff development. At UNR, one-third o f national full-time faculty in 2002 was pursuing additional training via overseas government scholarships (table 7.9). loo Among them, 55 percent were reading for their master's degree and 29 percent for their doctorate degree. These faculty members accounted for 14 percent of all students studying on overseas government scholarship, but made upmore than two-thirds of those pursuing studies at the post-graduate level. Table 7.9: Number of nationals on the Universite`nationale du Rwanda (UNR) faculty on an overseas scholarship, status as of February 2002 bumber ofUNRfaculty on overseasscholarships II 92 %working toward amaster's degree 55.3 %workingtoward a doctorate 28.7 III II As apercentageoffull-time nationals onthe faculty 33.7 Total scholarshipholdersabroadin2001-02a IIIII 666 II %scholarshipholderswho are UNRfaculty 13.8 %UNRfaculty amongthosepursuingpost-graduatedegreesb 64.7 7.30 Overview of staffing ratios inpublic and private institutions. W e tumnow to examining the pattems of staff utilization across institutions, using student-to-staff ratios as an indicator for this purpose. The relevant data appear in table 7.10 for the main public institutions, as well as the two private institutions with the largest number of students. Consider first the ratio of students to full-time faculty (column A in the table). The ratios are clearly much more favorable at the public than private institutions, but the pattern i s to be expected, given differences between the two sectors inthe reliance on visiting or part-time faculty. H o w would the ratios change if such faculty were also taken into account? W e have little information to make an exact calculation here, but we can perform some simple simulations based o n plausible assumptions about the average teaching load o f the visiting 99Inthe absenceofmoredetaileddata, it is unclear whether the highshare of doctorates is equallyprevalentamongKIE's nationaland expatriatestaff. loo atotherpublicinstitutionsarealsobeingupgradedthroughoverseas govemmentscholarships:twelvefromKHIandfour Faculty from ISAE. Similar informationis unavailablefor KIST and IUE. Among the privateinstitutions, UAAC currently has hvo faculty members studying abroad, one jointly financed by the Roman Catholic Church and the Government of Rwanda and the other sponsoredby Canada. 135 staff. In the table, the ratios in column (B) assume that the visiting or part-time faculty supply, on average, one-quarter of the regular load of a full-time faculty, whereas incolumn (C) they supply one- half of the regular load. The simulation results show that the differences instaffing ratios between the public and private sectors diminish, but remain large. Under the assumption that each visiting and part-time faculty supply about half the load o f full-time staff, the ratio ranges from lows of 9 and 10, respectively, at KIST and UNR, to a highof 14 at ISAE, incontrast to the ratios of 36 at ULK and 20 at UAAC. Table 7.10: Staffing ratios in public and privatehigher educationinstitutions,Rwanda, 2000-01 Number of staff and students Staffing ratios Sector andinstitution" faculty visiting non-teaching facultv staff staff UNR 1 389' 1 222 701d 4,840 12 11 10 9 2 IKLST 162 19 189 1,592 10 10 9 8 .e 0 KIE 85 148 959 1 1 / - I - I 6 KHI 63 43 690 ISAE 31 15 76 526 I I I 30 3,250 6 445 a. See table 7.8 above for the fullnameofthe institutions referredto hereby their acronyms. b. For the ratio in column (A), the denominatorincludesonly full-time faculty; incolumns (B)and (C) it includespart-time and visiting faculty as well, assuming that, on average, the latter categoryof staffrenders the equivalent o f onequarter and one-half, respectively, o f the official teaching load of a full-time staff. c. Includes35 teachers inthe &ole pratique des languesmodemesand22 UnitedNations volunteer lecturers. d. Excludesthe228 administrativestaffemployedatthe university hospital. Source: Appendix table A7.1, A 7.3, and A7.8 for data on enrollments and number o f faculty; personal communication 60m each institution on the numberofnonteachingstaff 7.3 1 With regard to the nonteaching staff, the magnitude of the differences between public and private institutions is even greater. The ratio of students to nonteaching staff lies below 10:1 at all the public institutions, except at KHI, where it rises to 16. Incontrast, the ratio is 108 at ULK and 74 at UAAC. The large differences in staffing ratios-for both teaching and nonteaching faculty- inevitably imply much costlier services in the public than private institutions. It may thus be appropriate to explore the potential for tighter cost management throughbetter staff utilization at the public institutions. 7.32 Student-facultv ratios bv field o f study at UNR. A s a refinement on the estimates presented above, we turn now to examining the situation at UNR.The available data allow us to adjust the ratios by removing the number of faculty who are on overseas government scholarships from the denominator and including the actual hours of teaching supplied by visiting and part-time faculty'o'; they also allow us to compute the ratios by field of study. The results should provide a more accurate picture of staffing ratios at the university (table 7.11). For the institution as a whole, the ratio o f students to full-time faculty on active duty was 16.3, but this falls to 11.3 ifvisiting and part-time staff Recall that UNR has about a quarter of its full-time staff on overseas government scholarship. Even though some of them may be away only part of the time and may thus continue to teach some classes, we exclude them completely to generateratios that are likely to be over- rather than underestimated.For the conversiono f visiting and part-time faculty into full-time equivalents, see the relevant footnotes intable 7.1 1. 136 were also included as units of full-time equivalent staff. The latter estimate is comparable to the simulated ratio o f 11reported earlier under the general assumption that visiting faculty supply one-half o f the teaching load o f a full-time staff. In the discussion below, we shall focus only on the ratios taking into account both the full-time and visiting/part-time faculty. Table 7.11: Student-facultyratiosby field of study at the UniversitPnationaledu Rwanda,2000-01 Full-time-faculty Part-time & visiting faculty Student faculty ratio I I I Full-time Fieldof study No. of No. on Total hours Number in Full-time students No. on &full- Total overseas ofteaching full-time faculty scholar- active services equivalent on active time ships duty provided unitsa dutyb equivalent faculty Law 338 26 7 19 1,599 8.0 12.5 Arts andhuman sciences 371 45 17 28 2,445 12.2 9.2 Econ., social science& management 1,142 61 15 46 6,858 34.3 14.2 Education 562 43 8 35 1,545 7.7 13.2 Joumalism& communications 30 7 0 7 315 1.6 3.5 Agronomy 79 31 17 14 1,386 6.9 3.8 Medicine 419 51 8 43 2,790 14.0 7.4 Science& technology 569 63 17 I 46 I 3.730 I 18.7 8.8 Galinspecializedfields 3,510 327 II 89 I 238 I 20,668 I 103.3 10.3 Ecolepratique des languesmodemes 1,330 35 2 33 5,585' 27.9 21.8 UnitedNations volunteer lecturers 22 0 22 0 0 Otherd 5 1 4 0 0 Overall total 4,840 389 I 92 I 297 I 20,668 I 131.3 1 16.3 11.3 Note: A dash denotesnotapplicable, andn.d. denotesnodata. a. Assumes aconversionrateof200 hoursannually per full-time staff. The officialloadfor faculty by rank is as follows: 180hours for assistants,210 hours for char& des cows andchar& des cowsassociks,and240hours forprofesseurs associks andprofesseurs titulaires. b. That is, excludingthoseonoverseasscholarships. c. Estimatedfrom the reportedexpenditure of FRw 29.6 million in 2000 on visiting faculty teaching in the Ecolepratique des langues modernesand on the assumptionthat the visiting faculty are remuneratedat $15 perhour(is.,FRw 5,300), the ratetypicallyappliedto holdersofmaster's degrees. d. These arethe teachers inthe Ecoledesanfkpublique et denutrifion.Ithadfaculty in2000-01, butnoenrollments. Source: Table A7.4 for enrollment by field of study, table A7.8 for number of faculty, and personal communicationfrom officials at UNR for the hours of teachingprovidedbypart-timeandvisitingstaff. 7.33 The average ratio for the institution as a whole masks wide differences across fields o f study. For the 27 percent o f the university's students who were enrolled for language training inthe Ecolepratique des langues modernes (EPLM), the ratio was nearly 21.8, whereas it was 10.3 for the rest who were enrolled in specialized programs. Within the latter programs, the ratios in the humanities and social sciences tend to exceed those inthe applied and natural sciences, but there are important exceptions. The ratio for joumalism and communications was lower at 3.5 than any o f the specialties in the applied and natural sciences, whereas that for arts and human sciences was only slightly higher at 9.2 than the corresponding ratios of 7.4 for medicine and 8.8 for science and technology. lo2 7.34 The foregoing estimates by field of study provide even clearer evidence o f the highly favorable staffing ratios in the public sector. Consider the student-faculty ratios for law and for economics, social science, and management, two fields catering to the bulk of the students at ULK and lo*The low ratios for the three fields mentionedhere maybedrivenby temporarydips inenrollmentsfor the year to whichthe data relate. Evenifthe higherenrollmentsfrom adjacentyears were usedinstead, the ratioswouldremainmodest. 137 UAAC, the two mainprivate institutions inthe co~ntry."~ The ratio for law at UNR was 12.5 and that for economics, social science ,and managementwas 14.2. The average ratios for the ULK and UAAC were much higher, even under the conservative assumption that part-time and visiting staff supplied, on average, half the teaching loado f full-time faculty. Underthis assumption, the ratio would be 36 at ULK and 20 at UAAC. The available evidence, at least as they pertain to the humanities and social sciences, thus, points to potential scope for improving staff utilization inthe public sector. Cost of service delivery and student finance 7.35 Higher education i s costly in Rwanda, in part because o f the staffing patterns documented above. As an earlier chapter has shown, the cost to the government per student in the public system is much higherinRwanda than inother developing countries, averaging more than nine times the per capita GDP in 1999, compared with an average o f 6.3 times among countries in Anglophone Africa, the region with the highestlevel o f costs among low-income countries. Below we document differences inthe cost o f service delivery across institutions inthe public sector and within UNRby field of study. We also examine the arrangements for student finance and contrast them to fees charged for private highereducation. 7.36 Cost of service delivery in public institutions. The term i s used here to refer to the runningcost o fprovidingteaching and other student services. The bursaries received by students, net o f the portion deducted to help defray the cost o f student services, are an additional burden on the public purse, but they are not part o f the direct operating costs o f the institutions and are therefore considered separatelyinthe next section. 7.37 The relevant data on costs per student at five of the country's six public institutions appear intable 7.12.'04 Consider first the pattern within UNR.Unitcosts are lowest for students inthe preparatory language training program offered through the Ecole pratique des langues modernes, but at FRw 545,000 per student equivalent to 6.6 times the percapita GDP), they nonethelessappear high for a nonspecialized course.' Across the specialized programs, the costs vary from FRw 669,000 per 6 student at the low end for economics, social sciences, and managementto FRw 1,385,000 at the high endfor science andtechnology. Thesepatterns are generally (although not completely) consistent with the staffing ratios consideredpreviously, in that costs tend to be higher in fields with more favorable student-faculty ratios.lo6 7.38 As expected, a wide variation also exists in the cost o f service delivery across institutions: itranges from a low o fFRw 757,000 at I S M (where, to recall, the student-faculty ratio is These two institutions together accounted for 91 percent o f all students in the private sector in2000-01, and less than 2 percent o f their students were enrolled inapplied and natural scienceprograms. The ULK catered exclusively to programs inthe humanities and social sciences, and UAAC had only 68 students inthe applied and natural sciences ina total student body o f 440 (see appendix table A 7.3). As in other aspects o f higher education documented so far, the fmancial data presentedhere have been carefully culled from reports supplied to the authors by each institution in the context o f this study. To illustrate the work underlying the cost estimates intable 7.12, consider those for UNR. The data refer to actual spending in 1999-2000 and 2000-01, weighted respectively by 9 and 3 months o f the year to coincide with the spano f the academic year. Furthermore, care was taken to exclude spending at the university hospital and the laboratory associated withit, because they are largely separate operations from the teaching function o f the university (apart from the fact that they might provide medical and pharmacy students with the site for their practical training). The different components o f spending per student-on faculty-level academic services and administration, university-level administration, and student services-reflect actual reported spending. The latter two categories o f spending are, for obvious reasons, uniform across faculties, but the first category is not. lo'For comparison, the cost o fpublic upper secondary education (excluding bursaries) was only 0.85 times the per capita GDP in 1999, implyingthat the operating costs for the language training program at UNR are more than eight times as highas those o fpublic upper secondaryeducation. Aside from staffing ratios, other factors, including the cost o f materials and supplies and the need for support staff such as laboratoly assistants, also influence costs. 138 the highest among the public institutions) to a high o f FRw 1,360,000 at KIST. The per student cost at KIST is comparable, however, to that for courses in science and technology at UNR; this comparability i s not surprising, given the similarity in the corresponding staffing ratios. Operating costs at KIE are also on the very highside: at FRw 1,066,000 per student, it is 23 percent higherthan the corresponding cost in the field o f education at UNR. KIE's higher cost i s undoubtedly associated with its more favorable student-faculty ratio (estimatedat no higher than 11to 1,comparedwith about 13.2 to 1 at UNR), but its relatively large shares o f expatriate staff (40 percent in 2000-01) and o f highly qualified faculty (45 percent holding doctorate degrees) may also be contributing factors.lo7 With regard to KHI, operating costs are smaller than those for medicine at UNR-FRw 702,000 per student compared with FRw 933,000. KHI has a slightly higher student-faculty ratio, which may explainits cost advantage, butthe mainreasonprobably has to do with its focus on less costly training programs such as nursing,physiotherapy, andradiography, rather than the trainingo fmedical doctors. lo' See tables 7.10, 7.11, andtheir explanatory notes for the relevantdata anddetails on their estimation. 139 Table 7.12: Cost of service delivery per student inpublic highereducationinstitutionsand their compositionat the Universitknationale du Rwanda, 2000 I Overall cost of servicesper nta %distribution of overall costsb'c Field of specialization University- In 1,000s of 4s amultiple Faculty-level level Student current FRw' ofthe per .cademicservices capita GDP' Lk administration administra- services tion Yniversite'nationale du Rwanda (UNR) Law 8.2 Arts andhuman sciences 11.7 Economics, social sciences, & management 8.1 Education 10.6 Journalismandcommunication 10.0 Medicine 11.3 Science andtechnology 16.8 Ecolepratique des langues modernes 6.6 Overdl for UNR~ 9.5 42 35 23 Share of spendingon personnel (0.89) (0.28) (0.24) - - 16.5 13.0 CigaliHealth Instituteg 8.5 9.2 Note: Dashdenotesnot available; two periodsdenotenot shownto avoidclutter. a. The costs include the salarieso ffull-time faculty, part-time andvisitingfaculty, andadministrative andservicepersonnel, as well as the cost ofmaterials, supplies, andotheroperatingoutlays, includingthat associatedwithproviding student services (ix.,student feeding, housing, andhealth care). Spendingon studentbursaries,net ofthe amounttaken at the sourcebythe institutionto cover registration fees, andfor studentservices (Le., food, lodging, andmedical care) constitutes atransferto studentsandis thereforeexcluded!?om the calculation here. ForUNR, the costs excludethose associatedwith the universityhospital andthe hospital laboratory. b. Faculty-level costs are specific to each faculty, reflectingthe numberand composition ofthe staffand the amount of nonpersonneloutlays. University-level administrationincludes expenditureon personnelandmaterialsincurredby offices with university-wide functions (e.g., rector's office, departmentso f finance, audit, personnel, printing, library, computer center, etc.). Student services include the cost of personneland materials for student feeding, housing, and health care. Note that the per student cost of university-level administration andstudent services is the same across all faculties, about FRw 276,000andFRw 182,000 respectivelyin2000. c. Figuresinparenthesesrefer to the shareso ftotal spendingon personnelunderthe correspondingcategow. d. Includesdata for agronomyandthe Ecole de santCpublique et denutrition whose dataare insufficiently reliable to show separatelyinthe table. e. The first ratio refersto the share on personnelwhenthe denominatorincludesspendingonall three categoriesshown inthe lastthree columns; the second ratio refersto the share ofpersonnelwhen the denominatorexcludesspendingonstudent welfare services. f. Totheextenttheunderlyingdatacontains some doublecountingofstudentbursariesandthe costofstudentservices,the figuresshownheremaybe somewhatoverestimated. g. Reflectsanestimate for 2000 basedondata for 2001. Source: Authors' estimationbasedonexpendituredatasuppliedinthe context ofthis study by officials at each institution. 7.39 In addition, the available data for UNR permit us to examine the distribution o f the operating costs across three broad categories of spending: faculty-level academic services and administration; university-level administration; and student services (feeding, housing, and student health), as well as the share of spending on personnel, overall and within each category."' The per lo*The shareofspendingonpersonnelwithin each category is estimatedby first computingthe total spendingonpersonnelbasedon (a) data on the number of staffby salary grade and seniority and the structure of staff remunerationby pay grade and seniority, @) the number and cost of Visiting and part-time staff, and (c) the number, seniority, and qualificationof nonteachingstaff and their pay scale. The result from this bottom-up approach is then checked for consistency with the volume of aggregate spending of the university. 140 student cost of university-level administration and student services is the same throughout the university-about FRw 275,500 for the former and FRw 181,600 for the latter. As a result of this uniformity,the variation inthe faculty-specific shares of spendingby category is driven exclusively by differences inthe absolute level of academic and administration costs at the faculty level and not at all by faculty-level differences inspending priorities. For this reason, the table shows only the distribution of spending for the university as a whole. 7.40 Nearly a quarter o f U N R ' s operating expenses is devoted to student welfare services, whereas more than a third goes for systemwide administrative overhead, leaving just 42 percent for actual teaching and related support services at the faculty level. Although no clear benchmark indicates that the current balance is suboptimal, an allocation of 42 percent for what is, after all, a core function of the university does appear low, suggesting a possible area for additional exploration inthe process of policy development. Furthermore, most o f the resources that do reach the faculties are taken up by personnel costs-averaging 89 percent for the institution as a whole and falling no lower than 85 percent in any faculty. It is possible that some of the nonpersonnel costs under system administrative overhead (which accounts for 72 percent of spending under this rubric) include materials bought in bulk by the university for distribution to the f a c u l t i e ~ . 'Although the available ~ ~ data do not permit this aspect of spending to be traced, the share of personnel in the combined spending at the faculty-level and university-wide administration works out to be about 61 percent, suggesting that the balance between staff and material resources to support teaching might in fact be better than appears at first sight. Whether or not it is optimal is beyond the scope o f the present report to confirm; we therefore simply flag ithere as yet another issue for additional analysis. 7.41 Student finance. A short explanation is appropriate at the outset concerning the arrangements for student finance in higher education. As in many francophone systems of higher education, full-time students in all public institutions generally receive a bursary.'" Up untilJanuary 2002, the arrangement was for deductions to be taken at the source to cover the cost of food, housing, and health care services, and the remainder is given to students as a cash transfer. '''Students also pay a registration fee, but the amount is paid directly rather than deducted from the bursary. The registration fee is small and typically covers a minuscule fraction of the cost of the academic services provided by the institutions. A s it turned out, the size o f the deductions from the bursaries was also modest relative to the costs of student welfare services. Eachpublic institution thus receivedadditional budget allocations to make up the difference (figure 7.1). The arrangements implied that beyond the bursary, students also receive in-kind benefits in the form of subsidized services. In the context of preparing the 2002 budget, the government agreed to a plan under which the operating costs of student welfare services was transferred to the students rather than to each establishment. This change was adopted January 2002, effectively making explicit the full value of the welfare subsidies provided to students. logNonpersonnelcosts make up an evenlarger share o f total spending for student welfare services, reflecting the fact that much of the spending i s for foodstuff for the studentcafeteria. 'loAlthough the bursary is given inthe form o f repayable loans, no mechanism has ever been put inplace to collect repayment on the loans. Students in fact perceive their bursaries as nonrepayable grants. The ministry has proposed establishment o f a Student Financing Agency that will monitor student loans and collect repayment, and it i s anticipated that the agency will be operationalby early 2003. See appendix table A7.9 for details on the amounts o f deduction for UNR students who are fed and housedon campus and those who eat and live offcampus. The arrangements at other public institutions are comparable to those at UNR. 141 Figure7.1: Arrangements for studentfinance inpublic highereducationin Rwanda (Statusprevailing up untilJanuary2002) Total amountofbursary Total subsidy =cash transfer + cost o f student welfare services 4 Amount deductedat source as contributiontoward student welfare services Cashtransfer to student Cost ofstudentwelfareservices Source:Authors' consmction. 7.42 Keeping the foregoing context inmind, we now turn to examine the available data on students receiving a government bursary at three public institutions-UNR, KIE, and ISAE (table 7.13). The share of such students at UNRhas beenrelatively constant at almost 99 percent throughout the post-genocide period, with the exception o f 2001-02 when it fell to 92 percent. At KIE, their share has been rising, from 72 percent in 1998-99 to 97 percent in 2000-01. Incontrast, the trend i s inthe opposite direction at ISAE, where the share o f students receiving a bursary has fallen from 96 percent in 1999-00 to 90 percent in 2000-01. Although precise data are not available for the other public institutions-KIST and KHI-comparably highshareso ftheir students also receive bursaries. Overall, therefore, securing a place inpublic higher education literally guarantees that a student will receive a bursary.' l2 '''For completeness, the data for UNRintable 7.13 also show the shares o fstudents who take their meals at the university andwho live instudenthousing. Those on a meal planhave beenrising from 60 percent in 1997-98 to 80 percent in2001-02,whereas the share living oncampus has fluctuated between 37 to 50 percent. 142 Table 7.13: Number of students and share amongthem receivinga bursary and various student services inUNR, KIE, andISAE, Rwanda, 1994tO 2001 Znstitut supe'rieur Universite' nationale du Rwanda (UNR) Kigali Institute of Education (KIE) i'agronomie et d'e'levage 1 1 1 Year E) Number of %receiving 9'0 on a YOin Number of YOreceiving Number of % receiving students a bursary meal plan student housing students a bursary students a bursary 1994-95 3,261 98.5 49.8 1995-96 3,948 98.8 41.1 1996-97 4,178 98.7 38.8 1997-98 4,548 98.6 60.4 40.1 1998-99 anne'eblanche (classes cancelled) 299 72.2 164 96.3 1999-00 4,535 98.4 77.2 40.2 597 95.8 314 89.5 2000-01 4,840 98.4 76.3 45.1 959 97.3 526 89.2 2001-02 5,922 92.4 79.9 36.9 Note: Blanks denotedatanot available. Source: Authors' compilation basedondata suppliedby each institutioninthe context ofthe presentstudy. 7.43 Inthe past few years, the bursary amounted to FRw 11,000 per monthfor a period of 10 months per academic year. Of this amount, deductions were made at source for health services, meals, and housing, and the remainder was transferred to the student in cash. The amounts deducted differ across institutions and across students according to the meal plan they choose and whether or not they live in student housing. At UNR, for example, the deductions amounted to FRw 500 health services, FRw 5,400 for students on a meal plan, and FRw 650 for those who live in student housing. For students boarding on campus, the cash transfer was thus FRw 4,450 per month, and for those fending for themselves, it was FRw 10,500 per month. Because students paid an annual registration fee o f FRw 7,300, the pocket money left from the bursary ranged from FRw 37,200 to FRw 97,700 a year, depending on whether or not the student ate and lived on campus. 7.44 What did the foregoing arrangements imply inaggregate terms? Some estimates inthis regard appear in table 7.14 for four public institutions. The data are not precise, especially those for KIST and KIE (where the reporting system is still being streamlined), butthey are sufficiently robust for our purpose here. The average amount ofbursary per student is computed simply by weighting the value of a full bursary by the share of students who receive the benefit. The average cash transfer to students is computed from the financial accounts supplied by each institution, divided by the number o f students receiving a bursary. These pieces of information allow us to derive the amount deducted on average for student services. At UNR, the deductions covered 25 percent of the cost of all student services and about 28 percent of the cost of foodstuff alone. Because the cost o f student services amounted, on average, to FRw 180,000 per year, the value of welfare subsidies received by UNR students, in cash and in kind, amounted to a total o f FRw 236,700 annually (=180,000 + 101,200 - 44,500)."3 The amount was generous, as it was nearly equivalent to 80 percent of the average pay, including benefits, of a primary school teacher (see table 3.6 for the relevant data o n teacher pay). 'I3Figure2.1 provides a visual aidfor these calculations. 143 Table7.14: Average annualvalue ofstudent bursaries,cashtransfers, anddeductionsto cover the cost ofstudent services in publichighereducation, Rwanda,2000-01 1 Amount o f bursary per student (FRw) Institution Relative to the Relative to Total" Cashtransfer cost o f all the cost o f to students InFRwb student welfare foodstuff services (%) (%I Universite' nationaIe du Rwanda (UNR) I 101,200 I 56,739 44,461 24.5 27.8 Kigali Institute of Science & Technology & Management(KIST) 1I 104,500 1I 60,214 44,286 51.0 Kigali Institute o f Education(KIE) 106,700 89,728' 16,972 59.8 Kigali HealthInstitute 110,000 56,440 53,560 41.9 7.45 As indicated above, the government decided duringthe 2002 budget preparation that all direct costs o f student welfare would be transferred to the students, rather than having each o f the institutions provide indirect subsidies for student accommodation and food. The plan was implementedinJanuary2002, andall government-financed studentsbeganreceiving abursary ofFRw 25,000 per month. As inthe past, the bursary i s being offered for 10 months per academic year, for a total o f FRw 250,000 per year. Underthe new system, the annual amount o f the subsidies i s expected to shrink gradually to FRw 200,000 in2003 and FRw 150,000 in2004. Implementation o f this part o f the plan, however, has been delayed, so that during 2003, students continued to receive FRw 25,000 per month. Ifthe value o f the bursaries fell to FRw 150,000 a year and ifthe cost o f services remained at about FRw 180,000 a year, students would have to top up their bursaries with out-of-pocket contributions o f FRw 30,000 if students are expected to pay the full cost o fthese services. This would represent a break from the tradition of complete subsidization, but it is only a small step in the right direction. At UNR, for example, this change would affect only the 23 percent o f the operating cost associated with student services; the remaining 77 percent would still require wholesale government subsidization unless additional cost-sharing policies are put in place. It should also be noted that, although higher education institutions should no longer be providing in-kind subsidies to students under the new plan (and thus no longer receive a budget allocation for food), they may still be providing in-kind subsidies to the extent that they still pay for the salaries o f staff providing student welfare. 7.46 Fees and student finance in the private sector. To provide a comparison with the arrangements inthe public sector, table 7.15 summarizes the patchy information that was possible to compile inthe context o f this study. Inhigher education institutions, the majority o f students pay only 144 a registration fee, currently amounting to FRw 7,300 a year.'14 In the private sector, students pay annual fees ranging from a total o f around FRw 160,000 to FRw 210,000 at ULK, depending on the level of study, and to FRw 225,000 at the ISPG. These rates are comparable to the charges for privately financed students who enroll at the public institutions, and they appear to be consistent with what private institutions would need to collect to be financially viable, assuming that fees constituted their primary source o f i n ~ 0 m e . l 'It~ i s interesting to note, nonetheless, that at some o f the private institutions, a significant share o f students defray their costs through a government bursary awarded through Fonds d'aide aux rescapbs du gknocide (FARG), a fund specially set up by the government to help those orphanedby the genocide. Table 7.15 Fees inprivatehigher educationinstitutionsand percentageoftheir students on governmentbursaries, Rwanda, 1998-2002 Itema I ULK I UAAC UNILAK 1 ISPG Annual registrationfees (1,000s FRw) 10 10 5 Annual tuition fees (1,000s FRw) Bachelier degree 120 201 120-130 200 Licence degree 150 190 I Other annual fees (1,000s FRw) 30 -50b 3/0 of students on government bursar9 1998-99 18 1999-00 31 2000-01 13 2001-02 36 Note: See table 7.2 for the fullnames o fthe institutions listed here by their acronyms. A dash denotes that no dataare available. a. The dataon fees referto circa 2001. b.Amount reflects the approximate annualizedrangesoffees pais de mkmoire) ofFRw 100,000 charged to final-year students. c. Amount reflects the charges for internship training. d. Almost allthe bursaries are givenbythe Fonds d'aide aux rescapCsdugknocide (FARG). Source: Compiled by the authors basedon informationsupplied by officials at the institution inthe context ofthe present study. Studentflow efficiency and outputof graduates 7.47 Higher education institutions produce many outcomes, includingresearchand scientific knowledge, but a primary function i s to produce graduates. How smoothly do students make the transition from year-to-year through the system, and how many are being produced each year and in what fields of specialization? Answers to these questions are documented below, again based on data supplied by each o f the institutions inthe context o f the present study. Inrecent years, some ofthepublic institutions have beguncateringto self-financed studentswho incur other costs inaddition to the registration fee that all students pay. For example, at UNR,there were 67 such students in 2001-02, and they paid FRw 200,000 in annual tuition fees. At KIST, the amounts charged for self-financed students include the following: FRw 198,000 a year for tuition, FRw 1,000 inapplicationfees, FRw 4,000 inregistration fees, FRw 5,000 for examinationfees, FRw 1,000 for identity cardcosts, and a refundable FRw 30,000 as deposit (or caution) money. Lunch expenses, averaging FRw 88,000 a year, are also levied on students '"whoa participate inthe institution's mealplan. As simple check, we note that the cost per student at UNR amounts to about FRw 784,000, and its student-faculty ratio is about 10.3. At the ULK,for example, the student-faculty ratio (after taking into account the probable teaching loads o f visiting andpart-time faculty) is about 36. If the structure of costs between teaching and other activities i s comparable between the public and private sectors, these datawould imply a cost per student o f about FRw 224,000 inthe private sector. 145 7.48 Promotion, repetition. and survival rates. The relevant data appear in table 7.16 for selectedpublic and private institutions and by field o f specialization within UNR.'l6Becausethe data for the private institutions are somewhat patchy, we shall cover them only briefly, noting only that the pattern o f student flow i s not worse than the corresponding pattern in the public institutions and, in some cases, they are in fact much better. With regard to the public institutions, the data are more complete and allow us to compare across fields within UNR and across several o f the institutions. At UNR, the promotion rates among students intheir first year ina specialized field range from a very low 44 percent inscience and technology to a high o f only 74 percent ineconomics and management andineducation. The promotionrates at KHIandISFP are muchbetter, butat ISAE, it falls to only 55 percent. Students who fail to be promoted either drop out or repeat the year. Among first-year law students, for example, 70 percent o fthem get promoted, 10percent repeat the first year, thus implying that 20 percent drop out. Across all fields, some dropping out does occur among first-year cohorts. The attrition rate appears to be particularly high for students in science and technology at UNR (35 percent) and among ISAEstudents (30 percent). 7.49 Once past the first year, the rates of progression from year to year improve, with much better promotion rates and lower repetition rates across all fields at UNR and across the other public institutions as well; dropping out correspondingly falls to reasonablylow levels. These salutary trends are perhaps the result o f a selection process that has winnowed out the weaker students in the first year. The situation continues to improve among final year students at KHI and ISAE, but it deteriorates dramatically at UNR across all fields. The rates o f promotion (i.e., o f graduation) fall to only 11percent inlaw and medicine and rise no higher than about 50 percent ineducation and inthe arts and human sciences. What i s the reason for this startling lapse? Although part o f the answer may be attributable to practices and policies peculiar to UNR, a more sinister view suggests that prevailing labor market conditions and the current bursary scheme combine to give students an incentive to remain at the university as long as possible. The prospects o f salaried employment are not bright, with an unemployment rate in 2000 o f some 35 percent among degree holders aged 25-29, whereas the value o f the bursary and in-kindwelfare services that repeatersapparently continue to receive i s quite attractive, amounting to 80 percent o f a primary school teacher's pay (with benefit^)."^ Under these circumstances, it makes sense for students to repeat their final year. Although this explanation o f the highrepetitionratesamong final year students at UNRis only conjecture at this point, the fact remains that it does point to a serious problem in student flow management that warrants attention as part o f policy development inthe sector. Becausethe promotionrates inthe second to the penultimate year tend to be relatively flat, they are shown inthe table as averages rather than for each year inthe cycle. The number of years involved varies across field of study, ranging from 4 years (excludingthe year spent in the language training program at the Ecole pratique des langues modemes) to 5 years for agronomy and 6 years for medicine.The promotionrate for studentsinthe languageprogramis reasonablyhighat 84.7 percent. ' I 7 For a moredetaileddiscussionon the labormarket, see the nextchapter. 146 Table 7.15: Promotion,repetition,andsurvivalratesinselectedhighereducationinstitutions,Rwanda,circa2000 I Iby year of Promotion rates I Index of . . study inthe cycle (%)b by year of siudy inthc ycle (%)' I ReDetitionrates Institution and field of studya Survival student I"year penultimate Final 'inal yea] rate (%)' flow efficiencyd JNR Law 70 78 11 10 16 89 58 38 Economics & management 74 85 40 6 4 60 53 69 Education 74 90 51 8 3 49 60 75 Arts and humansciences 73 87 49 13 3 51 58 74 Agronomy 90 30 6 70 76 65 Medicine 66 89 11 8 7 89 50 46 Science & technology 44 77 * 21 12 * * * (HI 84 84 89 5 10 9 81 85 SAE 55 85 94 15 6 0 55 73 SFP 90 100 7 0 97 95 JLKe 78 11 JNILAK 69 88 8 0 66 77 SPG 90 98 95 5 0 5 90 94 Note: An asterisk indicates that the rate was not computedbecause the underlyingdata pertain to only twelve students in physics and chemistry; a dash denotes nodataare available. a. See table7.8 above for the fullname ofthe institutionslistedherebytheir acronyms. b. The first year refers to the first year of study inthe selected field, not the first year ofenrollment at the university, duringwhich some studentsmay attend the Ecole pratique des langues modemes to achieve bilingualism in French and English. The duration of study differs across fields, lasting 5 years in agronomy, 6 years inmedicine, and4 years inall other fields showninthis table. The rateinthe middlecolumnofthe blockrefers to the averageofthe rates from the secondto the penultimateyears inthe cycle(whichshows arelativelystable pattern). c. Refers to the percentage of first-year students in the indicated field of specialization who eventually obtain their diploma or degree, with or without repeatingayear. d. The index has a maximumvalue of 100correspondingto a situation inwhich all students complete their studies without repeatingor dropping out. The index here is computedfrom data shown inthe precedingcolumnson promotionand repetitionrates andby weighting dropouts inproportionto the number ofyears they completebeforeleavingthe system. e. The ratereflects the averageacrossall years ofstudy. Source:Computedbyauthors basedonrawdatasuppliedby eachinstitutioninthe contextofthis study. 7.50 As a summary measure of the efficiency of student flow, the foregoing table shows two additional indicators inthe last two columns: the percentage o f first-year students who graduate and an index o f student flow efficiency, which measures the wastage in the system relative to a hypothetical situation in which students neither drop out nor repeat (in which case the index would be 100). The results suggest that much room for improvement exists, especially at UNR, where the graduation rate reaches no higher than 76 percent (for agronomy) and sinks to a low of around 50 percent (for medicine, and economics and management). The impact of the highrepetition rate among final year students shows up in the dismal efficiency indices for law and medicine. In the remaining public institutions for which data are available, performance i s excellent at ISFP where the graduation rate is 97 percent and the efficiency index is 95 percent. The KHI boasts reasonably good results too, but ISAE's performance is no better than UNR's. 7.5 1 The output of eraduates. The number of graduates, grouped under two broad rubrics- social sciences and management, and sciences and technology-appear in table 7.17. UNR is now producing more than three times the number o f graduates that it did in the 1 9 8 0 ~but its share of ~ graduates in 2001-02 (26 percent), nonetheless, seems modest in relation to its share of enrollments (36 percent). For the public sector as a whole, the volume o f output exceeds the private sector's by nearly 40 percent and the distribution of skills mix is roughlybalanced between the arts and sciences, whereas it is skewed almost entirely toward the arts in the private sector. Combining the output of all 147 institutions, the system is currently producing about 1,700 higher education graduates in total, two- thirds of them in the social sciences and management. The output can be expected to increase in coming years, given recent increases in enrollments. To illustrate, if enrollments remained at the estimated total of some 17,000 in2001-02 and if survival rates were conservatively assumed to be 50 percent (which is comparable to the average rate across fields of study at UNR), the output o f graduates would exceed 2,100 annually (assuming that courses last an average o f 4 years). If survival rates were higher, say 75 percent (which is possible because survival rates at other public institutions, and probably at private ones as well are better than they are at UNR),the output of graduates would approach 3,200 annually. The issue raised by these simulations i s whether the Rwandese labor market can effectively absorb this volume of output. Unless the pace of job creation keeps apace with the recent explosive growth inhigher education enrollments, graduate unemployment could easily develop into a major social problem incoming years. `I Table 7.16: Numberofgraduatesby institution and broad fields of specialization,Rwanda, 1980sand circa 2000 Sector, institution, and perioda Social Sciences I& Manaeemen! Technology Sciences All fields 90 53 142 IKIST 325 124 449 2001 34 98 132 Projectedfor 2002' 90 90 180 a' /KIE KHI 2001 0 110 110 1999-00 0 71 71 2000-01 48 0 48 Subtotal 497 ~ 493 990 1999-00 605 0 605 I e, UAAC 2000-01 66 0 66 4- (d .; UNILAK 2000-01 38 0 38 ISPG 2000-01 0 10 10 ISubtotal 709 10 719 lverall total 1,206 503 1,709 Note: A dashdenotesnot available or not applicable. a. See table 7.8 for the fullname ofthe institutions listedhereby their acronyms. b. Data for the 1980srefer to the average annualnumberof studentsgraduatingwith alicence degree. For 2001-02, they refer to the numberof final year students inthe secondcycle, which leadsto the licence degree. c. The number of graduates inthe two domainsis an approximation basedon the distributionofenrollmentsbetweenthem in 2000-01. d. Includesgraduates ofthe first andsecond cycles, which leadto, respectively,the baccalaurht andlicencedegrees. Source:Rwanda 1981to 1986 for data for UNR inthe 1980s; data for recentyears suppliedby eachinstitution to the authors inthecontextofthepresentstudy. Policyimplications 7.52 Rwanda's system of higher education has expanded and diversified rapidly inthe post- genocide period. Once dominated by UNR, it now comprises three new public institutions-IUST, KIE and KHI-with combined enrollments in2001-02 roughly equal to UNR's, as well as a handful o f new private institutions, the largest of which (ULK)enrolled about as many students as UNRinthat year. The system's expansion has been fueled by a strong demand for higher education, which intum has been stimulated by widespread scarcity of qualified labor in government and inthe modem sector ofthe economy inthe aftermath o f the genocide. 148 7.53 Inthe past few years, the system has developed inways that have strengthened it in several respects. The increased diversity has created healthy competition across institutions, as each t i e s to adapt its course offerings to match labor market needs. The changes are especially evident at UNR,where enrollments have shifted dramatically since the 1980stoward fields with an explicit labor market orientation. The highlypragmatic approach, inboth the public and private institutions, of using expatiate teachers and visiting and part-time staff to deliver courses has enabled the system to rise quickly to the challenge o f rebuilding the country's depleted human capital. At the same time, it has also built into the system the flexibility and suppleness that it will need to remain responsive in an evolving environment. 7.54 Yet, not all is well in the sector. The single most important issue concerns the unsustainability of current arrangements for higher education finance. At present, the subsector absorbs nearly 40 percent of total public recurrent spending o n education, a level of spending that can continue only to the detriment of efforts to develop primary and secondary education and at the cost of seriously compromising the country's broader poverty reduction agenda. Three factors combine to create the crisis in financing: the arrangements for student finance for domestic studies, govemment financing of overseas studies, and the highcost of service delivery inthe public institutions. 7.55 With regard to student finance for domestic studies, almost all students in Rwandese public institutions, including repeaters, receive a full government bursary and are further aided inthe form o f subsidized welfare services; the total value of the cash and in-kindsubsidies currently amount to about FRw 237,000 per student per year, roughly 80 percent of the pay of the average primary school teacher. These arrangements have persisted even as the number of students in the system has . risen nearly 2.5 times since the genocide. That the current system of student subsidization is unsustainable is now quickly becoming evident, and the government has proposed addressing the problem through a gradual reduction in the value o f the subsidies. Yet, the proposed reform appears timidrelative to the scale o fthe problem. The real challenge is infact to move away from the present blanket system of subsidization to one in which subsidies are offered much more selectively-for example, to students pursuing particular fields or from particularly disadvantaged backgrounds-and thus made more sustainable. 7.56 Withregardto overseas government scholarships, the item currently accounts for about a quarter of the government's recurrent spending o n higher education. The trends have beenmoving in the right direction in recent years: fewer students are being sent abroad and those sent are being increasingly directed to such lower-cost countries as India (where the host government has waived tuition fees) or Uganda and South Africa (where their physical proximity has helped to lower costs), instead o f to OECD counties, as inthe past. The fact remains, however, that, on average, an overseas scholar still costs nearly three times as much as a student studying on government expense domestically. The proportion enrolled overseas at the undergraduate or lower level is currently very high at 87 percent, and nearly half pursue degrees inthe humanities and social sciences. Inthe past when Rwandese higher education was still nascent, there might have been a stronger rationale for sending students abroad, even those at the undergraduate level. Intoday's greatly changed situation, it might make sense to channel students to domestic institutions whenever feasible, rather than to institutions abroad. The move would reduce the burden on the public purse, while supporting the development ofthe domestic system. 7.57 With regard to the highcost of service deliverv, the problem pervades the entire public sector. Unit costs inRwanda's public sector are now among the highest across low-income counties. One o f the mainreasons are the highly favorable staffing ratios bothfor teaching and nonteaching staff inallpublic institutions, with the result that most of the available resources are eaten upbypersonnel expenses with little left to finance other complementary inputs. Student-faculty ratios in the public 149 sector are only one-third to one-half as high as those for comparable fields in the private sector. Improving staff utilization inthe public system thus poses a key challenge inpolicy development. A potential interventioninthis regard i s to rationalize course offerings-not only within eachinstitution, butalso acrossthe public sector as awhole-to avoid duplication and take advantage of economies o f scale in service delivery. More broadly, it is also appropriate to explore the scope for reducing duplication incourse offerings with the emergingprivate sector, particularly infields inwhich private institutions have demonstratedtheir capacity to deliver training. The main advantage o f allowing the largely self-financing private institutions to expand and cater to more o f the students would be to reducethe financial burden on the public sector. 7.58 In addition, the analysis presented in the chapter also points to the need for tighter management o f student flow inthe public system. Inthe newer institutions, such as KHIand possibly also KIST and KIE, the first batches o f students appear to have moved through the system expeditiously, without too much dropping out and repetition. Incontrast, the pattern o f student flow at UNR leaves much to be desired: repetition inthe final year of study is high across all faculties, but they reachwhat mightbe describedas astronomical levels ofnearly 90 percent inlaw andmedicine. A more general issue in managing student flow pertains to the volume and skills mix o f the system's output o f graduates. The current output o f 1,700 graduates annually could rise to 3,200 a year under plausible assumptions about the pattern o f student flow. Although the system's capacity to ramp up production so dramatically is indeedremarkable, success must ultimately be judged by how well the output matches the availability o fjobs. As the country moves from the emergency o f replacing lost human capital to the task o f supplying a steady stream o f graduates to fill positions created by a modernizing economy, tighter management o f intake to the system, particularly in the highly subsidizedpublic sector, and o f students' transitions through the system would increasingly be needed to achieve a good balancebetweenthe supply o fand demand for highlyeducatedlabor. Conclusion 7.59 Higher education faces many development challenges in Rwanda. Following its explosive growth in the post-genocide years, the time is now ripe to take stock and explore possible directions for its future development. The most important task facing policymakers i s to get the system onto a development paththat i s more fiscally sustainable than the one it has been traveling on so far. The data and analysis presented in this chapter suggest that this will require not only changes to the system o f student finance, but also improvements in staff utilization within the public institutions, reform o f government sponsorship o f overseas studies, care in expanding the public sector to avoid crowding out the private sector, and better management o f student flow, particularly at the Universitk nationale du Rwanda, the country's largest public institution. Although these are not the only issues that matter inhigher education, they cannot be ignored inthe searchfor a more viable way forward to correct what has now become a serious imbalance in the allocation o f public spending on education, one that i s clearly at odds with the country's expressedcommitment to poverty reduction. 151 Chapter 8: Educationand the Labor Market 8.1 The previous chapters examined Rwanda's education system interms o f its financing, coverage, and internal operations. In this chapter, we look at the system's external efficiency. An important concern here is how well the education system i s responding to the labor market's demand for educated workers."* The issue is complicated and inevitably dynamic in nature. In Rwanda, its complexity has been accentuatedby the country's recent history and ensuing disruption o f economic activity and massive movements o f population, both within and across the country's borders. Inthe initial phase o f recovery inthe mid-l990s, acute shortages o f educated workers were felt everywhere inboththe public andprivate sectors.Manyqualified Rwandeseinthe diasporahave sincereturnedto the country, while the education system has also expanded. The situation i s thus shifting from one o f meeting emergency shortages o f qualified workers to one inwhich a steady flow o f school leavers is entering the labor force insearcho f suitablejobs. The issue then i s whether school leavers are landing jobs for which their education prepared them. Is the investment in producing qualified workers yieldingthe expectedreturns? 8.2 Although the sort o f data needed for a comprehensiveassessmento f these issues-such as tracer studies to track the employment experience o f recent school leavers-are unavailable, some insights can nonetheless be gained by reviewing cross-sectional data on the overall structure of employment, educational profile of workers, returns to education by level, and indirect evidence on the school-to-work transition experience o f recent entrants to the labor market. These aspects o f the linkbetweeneducation andthe labormarketarethe subjectofthischapter. Employment structure,educationalattainmentof workers, andthe returnsto education 8.3 Below we present data for 1991 and 2000 on the structure o f employment and the educational profile o fworkers; we then examine the current pattern o freturns to educationbylevel. 8.4 Structure o f emplovment. Consider first table 8.1, which presents data on the overall size o f the labor force and selected indicators on labor force participation, unemployment, and dependency. Between 1991 and 2000, the labor force (i.e., those working and actively seeking ajob) fell by nearly 2.5 percent and the population with ajob, whether paid or unpaid, fell by more than 3.4 percent (equivalent to a loss o f 122,000 jobs), whereas the total population grew by 11.5 percent and the population ages 10 and above grew by 19.7 percent. As a result o f these trends, the labor force participation rate (Le., the labor force relative to the population ages 10 and above) fell fiom 76.4 percent in 1991 to 62.2 percent in2000, whereas the unemployment rate climbed fiom 0.6 percent to 1.6 percent and the dependency ratio &e., the nonworking population relative to the working population) rose fiom 1.O to 1.3. "* Besides supplying educated labor for the economy, investments in education also produce other social benefits, such as better health- seeking behaviors and outcomes, lower fertility, and greater participation in community life. Inthis chapter, however, we shall be concemed only with the link betweeneducation and the labor market. 152 Table 8.1: Selecteddata on population,labor force, employment, and relatedindicators,Rwanda, 1991and 2000 1991 2000 Total population(1,000s) 7,157.6 7,979.9 Populationaged 10andabove (1,000s) 4,674.1 5,595.1 Populationinthe labor force (1,000~)~ 3,569.4 3,482.0 Employedpopulation(1,000~)~ 3,547.0 3,425.2 lNonworkingpopulation(1,000s)~ I 3,610.6 I 4,554.7 I Dependencyratiod I I I - 1.3 I 1.o I ILaboriparticipationrate e/.>' 76.4 62.2 Unemploymentrate e/.)' 0.6 1.6 8.5 Continued dominance of jobs in agriculture. Trends inthe distribution o f employment by sector appear intable 8.2, based on data for 1991 from the population census and data for 2000 fiom the Household Living Conditions Survey (also knownas the Enqukte Intkgrale sur les Conditions de Vie des Mknages au Rwanda (EICV).'19 Eventhough agriculture employed nearly 190,000 fewer workers in 2000 than in 1991, the sector continues to dominate employment in Rwanda, accounting for nearly 89 percent o f the jobs in 2000. Jobs in industry and public works also shrank in number between 1991 and 2000-by roughly one-third inboth cases-resulting ina further reductionof their already tiny shares o f employment in 1991. Incontrast to these trends, the number o fjobs grew inthe transport and communications sector and especially in the commerce and services sector. The latter employed 120,000 more workers in2000 than in 1991, raising its share of employment from 6 percent to almost 10percent. ' I 9The EICV is a nationwidesurvey of 6,420 urbanandruralhouseholdsconductedbetween 1999and2001. 153 Table 8.2: Distributionof employmentby sector, Rwanda, 1991and 2000 1991 Sector No. of No. of workers % workers % (1,000s) (1,000s) Agriculture 3,224.4 90.9 3,034.6 88.6 Industry 49.7 1.4 33.1 1.o Energy 3.5 0.1 3.7 0.1 Public works 31.9 0.9 20.4 0.6 Transportandcommunications 14.2 0.4 23.0 0.7 Commerceandservices 195.1 5.5 306.2 8.9 Other 28.4 0.8 4.2 0.1 All sectors 3.547.2 100.0 3.425.2 100.0 Source: 1991Population Census;authors' estimates basedonthe 1999-2001HouseholdLivingConditionsSurvey. 8.6 An increasing: share of unpaid familv workers and continued importance of self- employment. Table 8.3 shows the number and distribution o f workers by type o f employment and type ofjob. Compared with 1991,in2000, there were some 180,000 more salariedjobs, about 11,000 more employers, and an estimated 8,000 more apprentices. Correspondingly, the share o f salaried jobs in total employment rose from 7.6 percent in 1991 to 11.0 percent in 2000; employers' share rose from 0.1 percent to 0.4 percent; and apprentices' share grew from a negligible share to 0.2 percent during the same period. These shifts are nowhere near the increase o f 417,000 unpaid family workers between 1991 and 2000, an increase that has swelled the share o f such workers intotal employment from 25.5 to 38.6 percent. Moving in the opposite direction is self-employment, which dropped sharply by nearly 639,000 during the period. Despite the large drop, self-employment continues to provide nearly 50 percent o fthejobs inthe Rwandeseeconomy. 8.7 Interms ofthe distribution bytype ofemployment, the declinebetween 1991and2000 inagricultural jobs has been noted above. Similar job losses also occurred elsewhere: 11,000 fewer production jobs and 2,000 fewer managerial jobs inthe same period. Incontrast, administrative jobs swelled by 2,000, professional and technical jobs by 7,000, service provider jobs by 39,000, and commerce and sales jobs by 55,000. Despite the increase, these jobs, taken together, still account for less than 9 percent o ftotal employment in2000. 154 Table 8.3: Numberand distributionof workers by type ofemployment andjob, Rwanda, 1991and2000 1991 2000 Type o f employment andjob go. of workers No. of workers (1,000s) % (1,000s) % By type ofemployment Salariedjobs Public sector 54.1 1.6 Quasi-public sector 21.3 0.6 Private formal sector 59.3 1.7 Private informal sector 242.8 7.1 Subtotal 269.6 7.6 377.5 11.0 Apprentices 0.0 0.0 8.2 0.2 Employers 3.5 0.1 14,8 0.4 Self-employed 2,341.2 66.0 1,702.5 49.7 Unpaidfamily workers 904.5 25.5 1,32 1.5 38.6 Other 28.4 0.8 0.7 0.0 All twes o femployment 3,547.2 100.0 3.425.2 100.0 By type ofjob Professionalandtechnical 46.1 1.3 52.6 1.5 Managerial 3.5 0.1 1.3 0.0 Administrative andrelated workers 21.3 0.6 23.4 0.7 Commerce and sales 35.5 1.o 90.0 2.6 Services 81.6 2.3 120.3 3.5 Agricultural and related workers 3,224.4 90.9 3,031.6 88.5 Production andrelatedworkers 110.0 3.1 99.5 2.9 Other 24.8 0.7 6.5 0.2 All types o fjobs 3,547.2 100 3,425.2 100.0 Note: Blanksdenotethat dataare unavailable. Source: 1991Population Census;authors' estimatesbasedonthe 1999-2001HouseholdLiving ConditionsSurvey. 8.8 Trends inworkers' educational twofile. Despite the disruption to the education system caused by the 1994 genocide, Rwanda's labor force had a better educational profile in 2000 than in 1991. The number of workers with no schooling fell by 21 percent, whereas those with between 1to 3 years o fprimary schooling fell by 8 percent; incontrast, the number of workers with at least 4 years of primary schooling swelled by 22 percent. These shifts show up inthe trends documented intable 8.4 regarding the educational composition o f the workforce in 1991and 2000. Although 40 percent o f the workers hadno schooling in 1991, the share was down to 33 percent in2000. Workers with at least 4 years o f primary schooling made up 47 percent o f the workforce in2000, in contrast to the 37 percent in the earlier year. The improvement extends all the way to the top end of the educational ladder: workers with at least some secondary education accounted for 4.1 percent o f the workforce in 2000, comparedwith 2.6 percent in 1991. 155 Table 8.4: Percentagedistribution ofthe employedpopulationby educationalattainment, Rwanda, 1991and 2000 Educationalattainment 1991 2000 Noschooling 40.1 32.8 Primaryeducation Primary 1-3 years 20.9 19.9 Primary4 years andmore 32.6 40.2 Subtotal 53.5 60.1 Post-primary vocational andtechnical education 2.2 3.0 General secondaryeducation 3.7 Higher education } 2.6 0.4 Other 1.5 0.0 Total population 100.0 100.0 Sources: 1991Population Census;authors' estimatesbasedonthe 1999-2001Household Living ConditionsSurvey. 8.9 The rising educational attainment o f Rwanda's labor force may be the result of the improvement incohort survival rates inprimary schooling documented in an earlier chapter. A more likely explanation-particularly for the observed improvements at the high end o f the educational ladder-is that as peace was reestablished in the country following the genocide, employment conditions and other factors had improved sufficiently to attract significant numbers o f educated Rwandese living abroad to take upjobs inthe country. As we shall see below, the pattern of returns to education is consistent with this explanation. 8.10 The structure o f earnings and the returns to education. The 1999-2001 Household Living Conditions Survey provides the most up-to-date data on earnings, but the information pertains only to wage earners. According to the survey, the wage differentials are relatively wide across workers in different sectors o f the economy (table 8.5). The average worker in the informal sector earns only about one-fifth as much as his or her counterpart inthe formal sector. Significant gaps in educational attainment are part of the reason: workers inthe informal sectorhave, on average, only 3.5 years o f schooling compared with more than 9 years among formal sector workers. Table 8.5: Averagesalaries andyears of schooling ofwage earnersinthe formal and informalsectors, Rwanda, 2000 Sector Average annualsalary (1,000s of FRw) Years of schooling Formal sector Public 575.3 10.4 Quasi-public 1,095.4 9.2 Private 706.6 8.2 Average 711.0 9.3 Informal sector 142.9 3.5 Overall average 356.7 5.7 Source: Authors' estimatesbasedonthe 1999-2001HouseholdLiving ConditionsSurvey. 8.11 Giventhe significant disparities inearnings and education, one can expect the private returns to education, particularly at the top end o f the educational ladder, to be high. Table 8.6 confirms the expectation, with an extra year o f schooling above the sample mean, yielding a return o f about 11 percent for the average wage earner. The returns differ across the sector o f employment; however, they are highest for employment in the public and quasi-public sector (about 14 percent a 156 year), followed by employment in the private sector (about 13 percent a year), and lowest in the private informal sector (about 7 percent). Table 8.6: Ratesofreturnto educationby leveland sector ofemployment,Rwanda,2000 Level of education All sectors Public and quasi- public sector Formal Informal iverage rate ofreturna 11.2 14.6 12.9 6.5 iate of returnby levelb Primary 1-3 years ns. n.s. ns. n.s. 4-6 years 8.1 ns. ns. 8.2 All 3.0 ns. ns. ns. Secondaryvocational andtechnical 9.5 n.s. 4.7 9.6 Generalsecondary 20.5 ns. 16.3 20.2 Higher education 24.7 27.9 20.0 Note; A dash denotes that the estimate is unavailable for lack of sufficient observations in the dataset; n.s. denotes that the estimate is statistically insignificant. a. Correspondsto the coefficient onyears ofschooling inthe Mincerianeamings functionsreportedinappendix tables A8.1 andA8.2. b. Calculatedas [(eb /e" )- l]/[m,-m,] where bj andb, are, respectively,the regressioncoefficients ofthe educationaldummy variablesj andi reportedin appendixtables A8.1 andA 8.2 andm, andm, arethe correspondingaverage years ofschooling. The average durationof schooling i s 2.3 years for those with 1-3 years of primary schooling, 5.4 years for those with 4-6 years of primary education, 12 years for those with secondary vocational andtechnicalschoolingandgeneral secondary, and 16years for those with higher education. Source: Authors' estimatesbasedonthe 1999-2001HouseholdLivingConditionsSurvey. 8.12 Consider now the pattern of returns across levels o f education. For the average wage earner, the private returns to some primary schooling relative to no schooling are about 3 percent a year; those to post-primary vocational and technical education relative to primary schooling, are about 10 percent a year; those to general secondary education relative to primary education are about 21 percent; those to higher education relative to secondary schooling are about 25 percent a year. The much higher returns to general secondary education compared with those for post-primary vocational and technical education runs counter to the commonly held belief that vocational and technical education would confer significant labor market advantages to students who follow this curriculum. Although deficiencies inthe content o fcourses offered inthe past may explainthe modest retums, the possibility that such courses are inherently less responsive than general education to labor market needs cannot be dismissed. This possibility is, moreover, consistent with the findings in most developing country settings regarding the labor market performance o f public sector vocational and technical education and training (see, for example, a comprehensive summary o f these findings in Johanson and Adams 2003 forthcoming). The fact that such training tends to cost more than general education only reinforces the need for a careful and thorough assessment before significant public resourcesare committedto expanding the subsector. 8.13 Noteworthy too are the exceptionally high returns to higher education, an investment that typically entails a total of about 15 years o f study. Highrates o f return to this level of education are common inlow-income countries. Inmany of these countries, the modern sector is usually small and dominated by public sector employment, and wages at the high end tend to be "sticky," in that imbalances in the supply and demand for highly educated workers are accommodated more often through quantity than through price adjustments. These forces are reinforced inRwanda by the needto set sufficiently high wages to attract highly educated Rwandese and others to fill positions in government and elsewhere in the aftermath o f the genocide. That the tight labor market conditions 157 have also pulled upwages for general education graduatesi s suggestedby the highreturns to this level o f education. 8.14 For completeness, the table also shows the rates o f return to education by level and sector o f employment. The average retum to a year of education is highest in the public sector, followed by the formal private sector, and lowest inthe informal sector-a pattern consistent with the "stickiness" o f wages inthe modern sector. With regard to the rest o f the table, some o f the estimates are associated with relatively small cell sizes, so it i s prudent to avoid making too much o f the differences in returns across sectors. Suffice it to say that the overall structure o f returns remains unchanged: higher education fetches the highest private returns, followed by general secondary education, whereas vocational and technical education and 4 to 6 years o f primary education fetch private returns lessthanhalf as high. Outputof graduates andtheir absorptionintothe workforce 8.15 The foregoing evidence on the pattern o freturns to education provides one piece o f the puzzle regarding the link between education and the labor market. Because they pertain to earnings differentials for those who already hold ajob, the information may not capture the experience of more recent entrants to the labor market, particularly in a context where the scarcity o f qualified labor i s probably evolving from one o f general scarcity to one o f shortages in specific areas. From a policy perspective, we would want to know whether, at the current rate o fproduction o f school leavers, there are too many or too few graduates relative to the availability o fjobs and whether or not graduates are landing jobs in their field of specialization and earning a reasonable return on their education. Although a full assessment would require data not available at this writing, we can nonetheless glean some insights from evidence based on the 1999-2001 Household Living Conditions Survey on the incidence o f over- andundereducation inthe workforce, as well as unemployment ratesby educational levels among new entrants to the labor market. Because possible overproduction o f degree holders poses a special concern-in part because o f the high cost o f higher education and the difficulties o f managing graduates' job expectations-the discussionbelow also provides a cross-country perspective onthe size o fhighereducation inRwanda. 8.16 Incidence o f over- and undereducation. Following the literature (see for example Sicherman 1991and Verdugo and Verdugo 1989), we can define three categories o fworkers: (a) those who are adequately qualified, (b) those who are overeducated, and (c) those who are undereducated. The "adequately qualified" are those whose educational attainment lies within one standarddeviation o f the meanyears of schooling o f wage earners inthe same occupation specified at the two-digit level. The "overeducated" refers to wage earners whose years of schooling exceed the mean of those inthe same occupation by more than one standard deviation, whereas the "undereducated" refers to those whose years o f schooling are smaller than the mean in their occupational group by more than one standarddeviation. 8.17 Table 8.7 shows the overall pattern o f over- and undereducation in Rwanda in 2000. According to the foregoing definition, 60 percent of the wage eamers with higher education are injobs for which they are overeducated, compared with 13 percent among general secondary school graduates, and 37 percent among graduates of vocational and technical education. Because jobs become more skill-intensive with time, particularly ina context o f technological progress, we would expect some overeducation to exist in most occupations. However, the extent o f overeducation, particularly among those with higher education and those with post-primary vocational and technical education andtraining, appear somewhat excessive inRwanda. 158 Table8.7: Percentagedistributionofwage earners by educationalattainment relativeto workers inthe samejobs, Rwanda,2000a Primary Post-primary Higher All levels of vocational & General years technical secondary education education Adequately educated 66.1 73.2 46.4 79.1 40.2 69.4 Overeducated 27.9 18.8 37.2 12.8 59.9 16.3 Undereducated 35.1 12.1 6.0 8.0 16.4 8.1 0.0 14.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 a. A wage earner is classified have whenhis "adequately educated" whenhisher years of schooling fall within one standarddeviation of the sample mean years of schooling for workers inthe same occupation(specifiedat the level twodigit); as "overeducated' or "undereducated" when the years o fschoolingextendsbeyondthe samplemeanby at least one standarddeviation inthe correspondingdirections. Source: authors' calculationsbasedon the1999-2001HouseholdLiving ConditionsSurvey. 8.18 Two cautionary remarks on the foregoing results are inorder at this point. The first is that the results pertain to the overall conditions inthe country and does not rule out the possibility that in some occupations workers with these qualifications might be adequately educated or even undereducated. The second caveat is that in some occupations the bulk o f the workers may be underqualified if they were evaluated relative to an objective definition o f the minimumqualification required to perform the job adequately. For example, a large number of upper secondary school teachers have themselves no more than upper secondary education, and are therefore clearly underqualified for thejob. Insuch situations, our calculations mightconsider a degree holder teaching uppersecondaryclasses to be overqualified relative to other secondary school teacherswheninfact he or she i s appropriately qualified and the other teachers underqualified. Although such instances o f misclassification cannot be ignored in our calculations, they are unlikely to be so pervasive as to eliminate the likelihood that too many degree-holders and graduates o f post-primary vocational and technical education are in fact currently injobs for which their training exceeds the skills generally requiredto performthem. 8.19 Unemployment amone; new entrants to the labor market. As a complement to the foregoing picture o f labor market dynamics, consider now the pattern o f unemployment among workers with different educational attainment. Because the majority o f workers entering the labor market for the first time tends to be concentratedincertain age groups, we can capture school leavers' likely experience in the transition from school to work by disaggregating the incidence of unemployment by age group. The results appear intable 8.8 based on survey data for 2000; data for 1991basedon aggregatedcensusresults are also included inthe table for comparison. 8.20 Across all age groups, unemployment in both 1991 and 2000 is more widespread among those with secondary and higher education than among the lesser qualified, but the gap is substantially wider in 2000 than it was in 1991. The jump in unemployment rates among secondary and higher education graduates i s consistent with the loss o f modern sector jobs between 1991 and 2000 that was documentedinan earlier section o f this chapter. Focusingnow on recent entrantsto the labor market (whose unemployment rates are denotedinthe shadedcells inthe table), the unfavorable situation of secondary and higher education graduates come into even sharper relief. In 2000, the unemployment rate was 14 percent among general secondary school leavers in the 20-24 age group, and an astonishing 35 percent among degree holders in the 25-19 age group. In contrast, unemployment rates among new labor market entrants with no schooling or only primary education stood, respectively, at less than one percent andjust more than two percent. 159 Table 8.8 Unemployment rate by educational level and age, Rwanda, 1991and 2000 1991 2o0oa Educational attainment All ages 30-34 >34 All ages Noschooling 0.1 0.2 0.2 0.5 Primary 0.2 1.5 0.8 1.5 Post-primary vocational & technical 1.4 1.3 1.3 3.1 Generalsecondary } 6 4.4 9.3 3.0 Higher 2.0 5.6 19.0 All groups 0.3 2.3 I 2.4 I 2.6 I 2.4 1.4 0.7 1.6 Note: a dash indicatesthat the numberof observations inthe dataset is too small to compute reliablerates ofunemployment. a. The shadedcells refer to the age groups where first-time entrants from eacheducationcategory are likely to be concentrated. Source: Population Census of 1991andauthors' estimatesbasedon the 1999-2001HouseholdLiving Conditions Survey. 8.21 A cross-country perspective on the size of higher education inRwanda. Inview of the substantial rates of unemployment among recent degree-holding entrants to the labor market, a reasonable question i s whether or not higher education in Rwanda, taken as a whole, i s overly developed. One would expect that the more developed an economy the stronger would be the demand for highly educated workers, and therefore the larger would be the appropriate size of higher education. This expectation i s generally borne out in figure 8.2 which shows the relation across low- income countries between a country's per capita GDP inPPP international dollars, and the number of students inhigher education relative to the size of the population. In 1997-98, Rwanda was well-below the regression-predicted level of enrollments inhigher education given its level of per capita GDP. By 2001-02, enrollments inhigher education, particularly inthe private sector, had expanded so rapidly that they are now very close to the level one would in a country at Rwanda's level o f per capita income. This result, coupled with the evidence presented earlier on graduate unemployment, suggests that the catch-up phase of expanding highereducation is largely over, and that the expansion o f higher education, particularly the publicly financed sector, would increasingly need to be calibrated to the growth ofthe economy and creation o fjobs requiringhighlyeducatedlabor. 160 Figure8.1: Relationbetweenper capitaincomeand coveragein higher education,low-incomecountries,circa 1998 0 o * * 0 I- I 0 / 400 - / O Rwanda in2001-02 200 - Rwanda in 1997-98 ,-- o Other African countries * Non-African countries 50 - 300, , 500 1,000 2300 5,000 ! Per capita income (1998 PPP dollars; log scale) Source: authors' construction basedon data on per capita income, higher education enrollments, andpopulation sue from the World Bank's StatisticalInformation Management and Analysis (SIMA) database (version of December 16, 2002). Policyimplications 8.22 Inthe aftermath ofthe genocide, acute shortagesofqualifiedlabor were felt throughout the Rwandeseeconomy. These shortageswere met duringthe latter halfo f the 1990sby increasing the output o f degree-holders and other graduates, both in local institutions and through scholarships for studies abroad, as well as by attracting qualified Rwandese inthe diaspora to return home to fill the vacancies. Although highly educated workers with at least a general secondary education remain a small share o f Rwanda's workforce today, the share i s significantly larger today than it was in 1991. Enrollments inhigher education relative to the population is now also at par with that in other low- income countries at Rwanda's level o f economic development, thanks to the very rapid expansion o f the sector after 1997. 8.23 Although shortages o f certain types o f qualified labor will undoubtedly continue to be felt inparts o f the economy, the labor market at the highendi s beginningto show signs o f saturation: a significant share o f degree-holders are now in jobs for which they appear to be overqualified; unemployment rates are highamong probable first-timejob seekers holding a university degree or a general secondary school certificate. These patterns suggest that the demand for educated labor i s being transformed from one of mass shortages everywhere to one that is and probably will be increasingly driven by the pace and direction o f job creation in the economy. In other words, the challenge of filling the gaps left by the genocide-related loss o f educated labor i s giving way to the same generic task faced by policymakers inall low-income countries, namely that of ensuringa good match between the volume and skills mix of graduates leaving the education system and the economy's capacityto absorbthem into productive employment. 161 8.24 What then are the implications for policy development inthis regard? Although a full treatment would require more extensive data and analysis than were available or possible in this chapter, the results presented here provide some suggestions for shaping future policy. An important contextual consideration is the nature of the Rwandese economy: agriculture dominates employment, accounting for nearly 90 percent of the labor force in 2000 and is likely to remain as such for the foreseeable future. A majority o f Rwandese currently work in self-employment or as unpaid family labor. As the economy develops, the pattern o f employment is likely to evolve along the trajectory followed by most low-income economies: a gradual shift inproduction away from agriculture toward industryand servicesand a concomitant shift away from unpaid family labor toward paidemployment inthemodernsector. 8.25 Given the context and its likely evolution, investments in education can make an effective contribution in two directions. The first is to ensure that all children receive at least the minimum amount of schooling required to remain literate and numerate throughout their lives. The second i s to manage the expansion o f post-primary education inresponse to both the increased social demand created as increasing numbers o f children complete primary schooling and the absorptive capacity o fthe labor market for highly educatedworkers. 8.26 With regardto primary education, the suggesteddirection for policy development rests on the critical role that this level of schooling plays ina country's economic and social transformation. Evenintraditionalagriculture, this modicumo f schooling hasbeen shown inmost developing country contexts to boost substantially farmers' productivity (e.g., Lockheed, Jamison, and Lau 1980 and Foster and Rosenzweig 1996), and it is this superior productivity that facilitates the shift of employment away from agriculture. Ensuring that all children become literate and numerate adults also generates societal benefits, including better health, greater and more effective engagement in community life, and so on. These widely recognized benefits have indeed led countries worldwide to include universal primary school completion as one o fthe eight MillenniumDevelopment Goals inthe September 2000 United Nations Millennium Declaration. As an earlier chapter has demonstrated, Rwanda i s making good progress toward achieving universal primary education inthe post-genocide era. The challenges inthe future will be to persist inthis direction by addressingremaining pockets o f nonenrollment and reducing the country's excessively highrates o f grade repetition. 8.27 With regard to post-primary education, the pressures for expansion will be felt most immediately inlower secondary education. Accommodating this pressureto the extent allowed by the availability o f resources would be consistent with the increasingly widespread view among development policymakers and educators inside Rwanda and elsewhere that lower secondary schooling i s best viewed as an extension o f the primary cycle. The reasoning is that primary school leavers are often still too young to enter the labor force and the extra few years o f lower secondary schooling would reinforce their basic skills andbetter preparethem for adult life. 162 8.28 These social considerations become less persuasive in upper secondary and higher education. At these levels, cost considerations and the capacity o f the economy to absorb graduates into productive employment intheir field of training become increasingly relevant. As we have seen above, signs o f saturation are already emerging inRwanda, particularly for higher education. Keeping enrollments under careful management is thus an important challenge for the future. The menu o f policy instruments include: (a) administrative measuressuch as setting tighter selection criteria for the intake o f students, (b) other supply-side interventions such as reforming curriculum and developing new courses or fields o f study inresponse to labor market demand, (c) mobilizing market signals to guide students' choice of studies, including cost-sharing arrangements to encourage students to view their studies more as an investment than an entitlement, and better dissemination o f labor market information, and (d) encouraging greater participation by the private sector to satisfy the demand for upper secondary and higher education. Although the Rwandese government is already taking appropriate action inmay o f these domains, it remains important to monitor the policies' success in ensuringa good match between the education system's output o f graduates and their absorption into the labor market. Conclusion 8.29 As in most other low-income countries, the education sector in Rwanda receives priority attention inthe country's poverty reduction strategy. Yet, ina context o f limitedresources, the sector's claim on the public purse needs to be justified by evidence o f its contribution toward improving the welfare of the population. In this regard, the link between education and the labor market warrants close consideration. At the highend o fthe educational ladder, investmentsper student are costly and are intended to equip students with specialized skills to perform modern sector jobs. Because these jobs are createdby economic growth, rather than by simply producing more graduates inthe hope they will land the desiredjobs, it is important to heed the signs of saturation that have recently emerged and to manage expansion accordingly. At the low end o f the educational ladder, investments are less costly and are intended to equip students with the general purpose skills o f basic literacy and numeracy. Such skills are effective intraditional agriculture and informal sector work and generatenonmarket social benefits as well. These considerations imply placingthe highest priority on ensuring that all children receive at least a complete primary education and, eventually, a complete lower secondary education as resources and implementation capacity allow. InRwanda, these broad policy directions inmanaging the education-labor market link needto be strengthened, particularly by ensuring that basic education i s prioritized inthe allocation o f public spending. Because the situation is not static, it is important to monitor and evaluate it at regular intervals to ensure that policies are adjusted to maintain a good balance between what the education system supplies and what the labor market canabsorb productively. 163 Appendix Table Al. 1: Government revenueand expenditure,Rwanda, 1980-2001 (Billions of rrent FRw) Govemment revenue Govemment spending GDP at Current Total Current Interest payment Year market revenue, govern- spending Capital Total prices excluding Grants ment net of )nexternal spendingovem-men. grants revenue interest domestic spending payment 1980 108.0 10.2 0.2 0.1 0.0 10.5 1981 122.6 10.4 4.5 14.9 13.0 0.2 0.2 9.1 22.6 1982 131.0 11.7 3.6 15.2 13.9 0.3 0.3 11.1 25.7 1983 142.2 11.6 3.9 15.4 15.3 0.2 0.5 12.5 28.6 1984 159.1 14.3 3.6 17.9 14.8 0.3 0.7 8.0 23.9 1985 173.7 17.2 4.0 21.2 15.9 0.4 0.9 11.2 28.4 1986 170.3 19.8 3.7 23.5 18.2 0.4 1.o 12.0 31.6 1987 171.4 19.7 3.4 23.1 21.5 0.6 1.2 14.0 37.3 1988 183.1 17.1 5.7 22.9 21.1 0.7 1.3 11.6 34.6 1989 192.8 19.2 5.3 24.4 21.5 0.7 1.3 12.2 35.7 1990 213.5 21.6 5.9 27.5 27.3 0.6 1.5 12.7 42.1 1991 239.3 25.0 11.1 36.1 34.5 1.o 3.3 16.8 55.6 1992 272.9 27.6 16.7 44.3 40.6 1.4 3.8 20.1 65.8 1993 282.2 25.9 18.1 44.0 37.4 1.6 4.5 21.6 65.2 1994 165.8 6.0 1.5 7.5 14.8 2.5 4.9 4.4 26.6 1995 339.0 23.1 38.4 61.5 34.3 3.8 4.0 27.4 69.5 1996 424.3 39.4 31.4 70.8 49.0 4.0 2.9 39.4 95.3 1997 558.3 58.0 37.8 95.8 57.2 3.8 3.0 45.6 109.6 1998 621.3 66.0 33.0 99.0 69.6 3.4 2.3 42.1 117.4 1999 644.0 63.5 38.4 101.9 80.6 3.5 2.4 40.4 126.9 2000 696.6 68.5 63.8 132.3 82.8 4.5 2.0 42.5 131.8 2001 765.8 83.9 55.6 139.5 99.4 5.0 1.6 58.2 164.2 Note: -denotesdatanot available. Source: WorldBankAfrica RegionLive Database, version ofApril 24,2002. 164 Table Al. 2: Current and capitalpublicspendingon education, Rwanda, 1981-2001 (Millions E Year Current Capital Total 1981 3,285 1982 3,868 1983 4,573 1984 5,119 145 5,263 1985 5,050 308 5,358 1986 5,528 367 5,895 1987 5,557 456 6,012 1988 5,686 1989 5,690 1990 6,239 1991-95 1996 8,626 5,158 13,784 1997 11,360 7,871 19,23 1 1998 13,916 5,650 19,566 1999 21,738 6,103 27,841 2000 22,347 5,267 27,614 2001 25,341 16.769 42.110 de: Blanksdenotedatanotav able. Source: World Bank 1989 and Rwanda 1985 for data up to 1990; Rwanda 1996, 1998, 1999,20OOa, 2000b, and 2001 for data for 1996to 2001. .-f I u * * k E C o m -$ 2 0 0 0 0 o o o o g s 0 0 0 0 0 0 0 0 - 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 166 Table A3. 1: Public spendingon education by level, Rwanda, 1971-2002 (Millions of current FRw) Current spending Capital Spending Total Year Primary & All 'rimary & rimary & Total Sec.a Sec. Higher education Sec. Higher ,11 educatior Sec. Higher Education 1971 463.0 1972 563.2 1973 694.7 1974 930.7 1975 1,063.8 1976 1,231.9 1977 1,397.5 1978 1,455.2 1979 1,962.0 1980 2,361.8 1981 2,865.8 419.3 3,285.1 1982 3,378.1 490.3 3,868.4 261.0 3,639.1 1983 3,981.3 591.3 4,572.6 132.2 4,113.5 1984 4,472.7 645.8 5,118.5 34.8 110.0 144.8 4,507.5 755.8 5,263.3 1985 4,319.1 731.0 5,050.1 198.1 110.0 308.1 4,517.1 841.0 5,358.1 1986 4,768.7 759.0 5,527.7 257.2 110.0 367.2 5,025.8 869.0 5,894.8 1987 4,821.5 735.2 5,556.8 291.1 164.5 455.5 5,112.6 899.7 6,012.3 1988 4,810.1 876.3 5,686.3 297.5 5,107.6 1989 5,690.0 5,690.0 341.4 6,031.4 1990 5,812.2 6,239.1 331.1 6,143.4 1991 1992 1993 1994 1995 1996 7,357 1,311 1,269 8,626 5,158 13,784 1997 9,161 1,822 2,199 11,360 7,871 19,231 1998 9,046 2,183 4,870 13,916 5,650 19,566 1999 14.386 4,020 7.352 21.738 6,103 27.841 2000 8,349 22,347 5,267 27,614 2001 9,44 1 25,341 16,769 42,110 Note: Blanks denotethat dataare not available. a. From 1996onward, the data exclude spendingon secondaryeducation. Source:Rwanda(n.d.) for data for 1971-80; World Bank 1989 for data for 1981-88; Rwanda 1985aand 1985bfor the dataonhigher educationin 1985; Rwanda(budget documents for various years) andRwanda2000 for the data for 1996-2001; seeappendixtableA3.2 for further detailsonthedata for 1996-2001. 167 Table A3.2: Publicspendingon education, Rwanda,1985-2001 fMillions of currentFRw) 1985 1996 1997 1998 1999 2000 2001 Ministry of Education - 13,784 19,231 19,566 27,841 27,614 42,110 Recurrent spending 4,919 8,626 11,360 13,916 21,738 22,347 25,341 Primary & preschool 3,295" 5,505 6,624 6,305 9,7 13 9,895 10,914 Secondary 771 1,194 1,645 2,006 3,767 3,805 4,24 1 Local higher education 606 . 5,000 6,353 6,947 Scholarshipsfor higher educationabroad 49 1,155 I1,985 L 4,475 1,889 1,818 2,05 1 Administration 198 772 1,106 1,130 1,369 476 1,188 Capital spending - 5,158 7,871 5,650 6,103 5,267 16,769 Domestically financed - 581 - 550 503 321 583 Extemally financed - 4,577 - 5,100 5,600 4,946 16,186 Share of capital spending(YO) 37.4 40.9 28.9 21.9 19.1 39.8 Other ministriesor government agencies - - (recurrentspendingonly) 860 3,492 3,724 Institut Supkrieur desFinances Publiquesb 56 30 96 68 81 Vocational training` 15 42 216 142 165 Nonformal educationd 121 74 35 77 78 Subsidies for secondarystudents GenocideFunde 0 0 0 2,500 2,650 MINALOC` - - 513 505 550 Subsidies for highereducationstudents GenocideFunde 0 0 0 200 200 Note: -denotesno data. a. Includesspendingonthe Centres de 1'Enseignement Rural et Artisanal IntCgr6(CERAI) andothersimilar institutions, which offer 3-yearvocational trainingafter the 8-year primary cycle; these were discontinuedafter 1991-92. See also appendixtableA2.1 for additional detailson the structure andduration ofthe various cycles. b. Undervariousconfigurations ofthe MinistryofFinance. c. Under various ministries during 1996-2001, includingthe Ministry ofLabor andSocialAffairs (1996); the Ministry o f Youth, Sports, andTrades (1998); the Ministry of Youth, Culture, andSports (1999); andthe MinistryofYouth, Sports, andCulture (2000 and2001). d. Undervariousministries during 1996-2001, includingthe MinistryofLabor andSocialAffairs (1996); the MinistryofGender, Family, andSocialAffairs (1998); the Ministryo f SocialAffairs 1999); andthe MinistryofLocalAdministration and SocialAffairs (2000 and2001). e. Also known by its acronym, FARG, which stands for Fondsnationalpour l'lsistance aux rescapCsdu ghocide et des massacres auRwanda. f. Refersto theMinisthredeI'AdministrationLocaletdesAffaires Sociales. Source: World Bank 1989, Rwanda(various years & 1985) for the data for 1985; Rwanda(variousyears and 1999) for the data for 1996,1997, and 1998; Rwanda (various years) for the data for 1999,2000, and2001; supplementedby unpublisheddatasuppliedby MINALOCon the subsidies to secondarystudentsin2000 and 2001. 168 Table A3.3: Itemizedhouseholdspendingon education,Rwanda, circa2000 Primarv - Secondary Highera Public 'rivate Public Private Share oftotalenrollments(YO) -- 99.3 0.7 56.6 43.4 YOreportingnon-zerospending on: Overall school-related expenses 97.9 99.5 97.4 97.2 91.7 School fees 65.0 83.5 73.3 81.5 79.2 PTA contribution or other chargesb 40.8 28.0 17.9 13.2 4.2 Books and school supplies 90.4 92.3 87.8 88.8 70.8 Transportation to and from school 0.8 11.5 56.1 50.4 62.5 School uniform 59.7 78.6 75.2 79.3 5.6 Food, board and lodging 0.2 1.6 20.6 13.7 12.5 Clubfees and field trips 8.0 19.2 18.9 17.9 8.3 Miscellaneous school-related expenses 27.1 22.5 34.5 38.4 27.8 Annualspending per student amongthose reportingnon-zero spending(FRw)' Overall school-related expenses 1,847 (0,421 38,958 58,362 136,433 As % o f per capita GDP 2.3 12.7 47.6 71.3 166.6 us$ 4.7 26.7 99.9 149.6 349.8 School fees 689 8,010 28,941 48,076 108,568 PTA contribution or other chargesb 272 290 10,356 6,722 2,098 Books and school supplies 486 1,717 5,080 5,664 15,821 Transportation to and from school 614 3,011 4,361 5,65 1 21,867 School uniform 1,374 2,653 5,959 6,594 4,034 Food, board and lodging 49 54 9,960 10,446 36,888 Clubfees and fieldtrips 66 97 911 758 73,371 Miscellaneous school-related expenses 61 349 3,324 4,362 13,359 Weightedannualspendingper studentaveragedacross all students (FRw) Overall school-related expenses 1,807 10,370 38,173 56,644 124,002 As % o f per capita GDP 2.2 12.7 46.6 69.2 151.5 InUS$ 4.6 26.6 97.9 145.2 317.9 School fees 411 6,242 20,767 38,730 84,776 PTA contribution or other chargesb 117 95 2,130 1,098 48 Books and school supplies 439 1,616 4,514 5,096 10,071 Transportation to and from school 4 235 2,340 2,738 14,391 School uniform 814 2,083 4,418 5,076 402 Food, board and lodging 0 0 2,505 2,083 4,581 Club fees and field trips 6 25 196 141 5,209 Miscellaneous school-related expenses 17 74 1,302 1,683 4,525 a. Higher education includesuniversities and other institutions not classified as primary or secondary schools; as there were o I72 observatiom for higher education,the dataarenot shownseparately for public andprivate institutions. b.PTA refersto uarent-teacherassociations. c. Because individuals incur non-zero spending for different items of spending, the sum of the separate items o f spending exceeds the overall average across all items; for example, at the primary level, the sumofthe items was 3,612 FRw, comparedwith 1,847 FRw for the overall average. The differencesuggests that households incur expenses on some items by economizingon spendingin other categories. The extent that this occurs is largely, but roughly comparable across levels ofeducationandbetweenthe public andprivate sectors. Source: Data on the distribution of enrollments are from MINEDUC's 2000-01 censuses of primary and secondary schools and sources cited in appendix table A2.1; data on spendingon schooling aretabulatedby authors from the 1999-2001EnquLe Intbgrul de la Conditionde Vie (EICV). 169 Table A3. 4: Recurrent spendingoneducationby level of educationand function, Rwanda, 1999-2001 1999 2000 2001 Primary & preschool servicesa 8,963 9,325 9,483 Secondaryeducation Education servicesa 2,205 2,222 2,508 Student feeding 263 908 1,000 Higher education Education servicesa 1,739 2,33 1 2,717 Student feeding 1,045 1,150 1,389 Bursaries For studies inlocal institutions 733 1,114 1,144 For studies abroad 1,889 1,818 2,05 1 Administrationb 1,585 1,011 1,975 Other operatingexpenditures' 2,927 2,128 1,854 Teacher salary arrears 390 340 1,222 TotalMOErecurrent spending 21,739 22,347 25,342 Bursariesadministeredoutside MOEd Secondary 513 I 3,005 3,200 Higher education (local) 0 200 200 Memo item: GDP at factor prices (billions) 644.0 I 696.6 765.8 I supplies. b. Excludesspendingonthe universityhospital, which is listedunderthis budgetheadfor the educationmirv. c. Includesspendingonmaintenance, repair, andotherrunningcosts. d. Refersto bursariesawardedthroughMINALOC andthe GenocideFund. Source: Basedon Rwanda (various years) for the data for 1999,2000, and2001; supplementedby unpublisheddata suppliedby MINALOC onthe subsidies to secondarystudentsin2000 and2001. 170 Table A3.5: Per student spendinginpublic primary and secondary schools, Rwanda, 1999 Itemized spending per student Primary Secondary education education Tronc Upper secondarv Both levels I Overallspending commun 7,604 60,273 73,628 65,223 Teacher salaries 5,433 19,100 27,252 22,122 Average annual teacher salary 308,522 443,211 526,23 1 477,48 1 Ratio of students to teachers 56.8 23.2 19.3 21.6 Salariesof school-level administrative staff 375 10,250 Average annual salary of administrative staff 388,594 594,943 Ratio o f students to administrative staff 1,036 58.2 Material inputs at the school level 244 18,137 Studentwelfarea 0 7,845 Management overhead 1.552 6,869 Memo items: Average annual teacher salaries, incl. benefits (as multiple of per capita GDP) 4.0 5.7 6.8 6.2 No. of students inpublic schoolsb 1,4 18,707 44,256 26,064 70,320 No. of students inprivate schools 10,001 35,198 19,606 54,804 No. ofteachers inpublic sector with teaching duties 24,982 1,907 1,350 3,257 Nofe: Blanks denote items of cost that aresharedbetweenthe tronccommun anduppersecondarycycles, which arenot calculatedseparately. a. Excludesbursariesawardedto students attendingprivate schools; estimatedaccordingto the distributionofstudents betweenthe public and privatesectors. b. Refersto state andZibresubsidii schools. Source: Basedon data on educationspending intables3.6 and3.7; data onenrollments intable A2.1; data on numberofteachem on MINEDUC's 1999census ofteachers. 171 ITable A3.6: Earningsfunction excludingwage earnersworkingfor educationalinstitutions, Rwanda, 2001 Coefficient t-statistic Sample mean Level of education (omitted category: 0 to 5 years of primary education) 6 years ofprimary education 0.179 8.33 0.218 1-2 years ofpost-primary education 0.330 12.73 0.110 3-4 years of post-primary education; D4, D5 0.535 24.26 0.194 6 years oftechnical secondary education; D6, D7 0.883 38.24 0.193 Baccalaurbat; graduate 1.206 35.28 0.049 Licence, ingbniorat, doctorat 1.490 34.91 0.026 Age (in years) 0.016 25.59 37.4 Age2(inyears) -0.000 24.27 2,564.3 Intercept 11.422 393.86 RZ 17127 0.46 Number of observations 0.46 INote: The regressionsexclude the followingworkers: all nondwandese workers, emplo es ofintemation; manizations. 1workerswho with dummy variables for eachenterprise;these variables andtheir coefficient estimateare not shownto economizeon space. Source: Authors' analysisofJune2001survey of enterprises andtheir workers conductedby the Ministhrede laFonctionPublic et du Travail (Minfotra). N l- z 2 2 I I 9909 s s % I - q o o m 0 m " I I 173 Table A4.2: Schoolprogressionratesduring two consecutiveyears among7- to 12-year-olds by orphanhoodstatus, Rwanda, 1998-2000 try rate to grade 1 Schooling status in 1999-2000amongthose 1999-2000among alreadvinschool in 1998-99 Orphanhood status thosenot 1) inschool in 1998-99 Advanced a Repeatingsami Dropped out (percent) grade grade Both parents alive 53.2 62.1 34.2 3.6 Livingwith both parents 55 62 35 3.5 Living with only one parent or neither 48 64 32 4.2 At leastone parent dead 49.8 66.4 30.0 3.6 Mother dead; father alive 42 74 23 3.2 Fatherdead; mother alive 54 65 33 2.7 Both parentsdead 42 69 24 7.0 All children insample 51.8 63.8 32.6 3.6 Children not living with biologicalparent(s) 38.3 66.0 27.4 6.5 ;ample size 776 Source: Authors'estimates basedon the 2000 RwandaMultipleIndicatorClusterSurvey; see tableappendix A4.1 for moredetails. 174 Table A4. 3: Regressionestimatesof school attendancestatus in a cohort of childrenaged 7-12 in 1998, Rwanda, 1998-2000 School School attendance status in 1999-2000 attendance Among status those not Amongthose enrolledin Right-hand-sideregressionvariables in 1998-99 enrolledin 1998-99and 1999-2000 1998-99 Probabilityof Probability Probability being ofentering Probabilityof of dropping enrolleda grade lb repeating' out' Sex (girlis the omittedcategory) BOY 0.062 0.209 -0.0 13 -0.070 (0.73) (1.44) (0.15) (0.30) lrphanhoodstatus (livingwith bothparentsis the omitted :ategory) Bothparentsdead -0.588 -0.529 -0.381* 0.723 (3.77)* * (2.07)* (1.96) (2.03)* Mother dead -0.475 -0.521 -0.616 -0.332 (2.38)* (1.62) (2.38)* (0.54) Father dead -0.112 -0.022 -0.105 -0.383 (1.07) (0.12) (0.95) (1.24) Both parentsalive& livingwith one or neither ofthem -0.289 -0.28 1 -0.145 0.066 (2.22)* (1.30) (1.01) (0.19) ,ocation (living inurbanarea isthe omittedcategory) Livinginrural area -0.812 -0.113 0.342 0.502 (5.95)** (0.45) (2.80)** (1.48) ncomegroup (40% poorestis the omittedcategory) Middle 0.274 0.120 0.003 -0.450 (2.91)** (0.76) (0.03) (1.76) Richest 0.659 0.061 -0.196 -0.799 (5.05)** (0.27) (1SO) (2.26)* :onstant 1.631 0.151 -0.797 -2.890 (9.91)** (0.53) (5.03)* * (6.96)** I 'seudo Rz 0.031 0.009 0.012 Ibservations 3.014 776 2,238 2,238 Note: Robust statistics appear in parentheses, with one star (*) demoting statistical significance at the 5% level and two stars (**) denoting statistical significance at the 1%level. a. Logit model; sample: 7-12 school age children (in 1998-99) not enrolled or enrolled in 1998-99 in primary education; omitted category: children not enrolled in 1998-99. b. Logit model; sample: 7-12 schoolage children (in 1998-99) not enrolled in 1998-99 andentrying or not into 1stgrade in 1999-2000: omitted category: childrennot enrolled inbothschool years 1998-99and 1999-2000. c. Multilogit model; sample: 7-12 school age children (in 1998-99) enrolled in 1998-99; omittedcategory:children upgrading during schoolyear 1999-2000. Sources:Authors' estimatebasedonthe 2000 MICS survey. 175 Table A4.4 Number of candidates, percentagefemales and percentageexceedingcut-off marksfor promotionor pass markinthe end of-cycle examinationsfor primaryand secondary education,Rwanda, 200182 2002 ~ ~ ~ _ _ _ _ _ _ Level of education& indicator 2001 2002 No. ofcandidates 63,931 79,226 % females 48 49 IOverall 26 24 %exceedingcut offmark 1 I for promotionto next cycle Among girls 20 18 %exceeding cut offmark for promotion to next cycle No. o f candidates 17,718 24,852 n %females 47 49 Overall 77 69 3> 3v i2 %exceeding pass mark Among girls 71 63 2 2 % Among boys 81 76 a 1 Overall 17 12 %amongpassersselected for public higher education1Among girls IAmong boys 24 16 Table A5.1: Teacher's educationalattainment, qualification, and remuneration, Rwanda, 2001 Qualification //Structure o f gross monthly remuneration Benefits 3ducational attainment II I Salary range, ntroduced in Abbreviatior Explanation Grade III c a z ry (increased at 3 % II 1996 1999d a year since)c Dkplacement f Logement) 'rimary (8-year cycle) CA Cert$cat d 'Aptitude .ower secondary General E.S. 1, 2, 3 1-3 years general secondary Unqualified 6,250 -7,750 CERA1 Centred'Enseignement Rural et Artisanal Intkgrk SectionFamiliale (for Vocational diploma SF girls) Unqualified 6,750 Centre d'Enseignement CERAR Rural et Artisanal de Rwanda (for boys) Enseignement EAP Apprentissage Pkdagogique 4,000 EMA Ecole Moniteur Auxiliaire Teacher training E M M Ecole Mknager Moyen Instituteur diploma 9,300 -28,800 EMP Ecole Mknager auxiliaire Pkdagogique ENA Ecole Normale Auxiliaire ENTA Ecole Normale Technique Auxiliaire Jpper secondary' Incomplete VI 10,750 - 12,700 Complete (old system) IV 15,400 - 36,100 Complete (new system) 111 20,400 -50,100 6,500 I1 27,000 -62,000 10,000 'ost-secondary I 30,800-66,200 25,000 Doctorat or Ph.n. IDoctorate lore: -denotes not applicable. a. Teachers with an incomplete upper secondaly education are those who have not received the relevant diploma for this level o f schooling. Those with a complete upper secondary education fall into two groups, depending on whether they received their diploma before or after the reforms in 1982, which introduced the 6-year secondary education cycle that is still currently inplace today. b. For teachers inthe "qualified" category, the top andbottom pay in the range are separated by a series o f between 9 and 11 steps in the salary scale. For teachers inthe "unqualified" group, the pay structure contains no stepwise progression with seniority. c. Although the annual increasemay rangebetween 0 and3.5 percent, 80-90 percent o fthe teachers receive an average increaseo f3 % annually. d. Refers to allowance introducedin1999for transportation andlodging. Source: MINEDUC. 177 State Libre subsidik Ratio (Butare=loo) Index Ratio (Butare=loo) Index RWANDA 59.7 - 56.9 - Butare 55.2 100 51.8 100 Byumba 58.3 106 58.6 113 Cyangugu 56.6 103 55.7 108 Gikongoro 50.8 92 51.2 99 Gisenyi 69.7 126 63.1 122 Gitarama 57.4 104 55.8 108 Kibungo 62.9 114 62.0 120 Kibuye 68.4 124 62.4 120 KigaliRural 66.8 121 58.5 113 KigaliVille 46.0 83 40.4 78 Ruhengeri 53.3 97 55.8 108 Umutara 52.3 95 51.5 99 Note; -denotes not applicable. Ratiosreflect unweighted averagesacross schools. 178 Table A5.3: Regressionestimatesof the relationbetweennumbers ofteachers and pupils across types of publicprimary schoolswith provincialdummy variables, Rwanda, 2000 State schools Libre subsidie' schools Bothtypes of schools Regressors Coefficient t-statistic Coefficient t-statistic Coefficient t-statistic \lumber ofpupils 0.02 41.67** 0.02 69.53** 0.02 81.42** 'rovincial dummies ~ Butare (reference)" - - - - - - Byumba -0.46 -0.84 -1.07 2.79** -0.90 2.92** Cyangugu -0.43 -0.5 -0.38 -1.13 -0.27 -0.88 Gikongoro -0.13 -0.07 0.46 -1.39 0.60 -1.94 Gisenyi -2.00 3.29** -2.06 6.31** -2.00 6.95** Gitarama 0.30 -0.47 -0.87 2.85** -0.61 2.23* Kibungo -1.05 -1.78 -1.23 3.29** -1.19 3.80** Kibuye -2.08 2.67** -1.86 5.75** -1.79 6.04** Kigali Rural -0.04 -0.07 -1.03 2.70** -0.66 2.21* Kigali Ville 4.74 5.19** 5.92 10.15** 5.42 11.07** Ruhengeri 1.27 2.20* 0.03 -0.1 0.39 -1.34 Umutara 0.11 -0.19 0.58 -1.04 0.04 -0.1 zonstant 1.98 4.14** 2.34 8.58** 2.24 9.42** \lumber of schools (N) 548 1,440 1,988 c2 0.82 0.81 0.8 1 Note: - denotes not applicable. One star on the robust t-statistics indicates statistical significant at the 5 percent level, whereas two stars indicati statistical significance at the 1percent level. a. As the omittedvariable, Butare is therefore the reference province. Source: Authors' estimatesbasedon data from MINEDUC's 1999-2000census ofprimaryschools. 179 Table AS. 4: Regression estimates of the correlates of school-level pass rates on the nationalprimary school leavingexamination, public sector schools, Rwanda, 1999 Model 1 vlodel2 Regressors amp11 Marginal - mean hfficient -statistic effectb hefficient t statistic `upil-teacher ratio Lessthan 40 (referencedcategory) 6.0 - - - 40-55 48.2 0.006 0.15 0.14 >55 - 45.9 0.052 1.41 1.28 reacher qualification (percent in category) Upper secondarydiploma (D6-D7) with pre- - serviceteacher training (referencecategory) 32.3 - - Primary education 1.3 -0.003 1.07 -0.06 Lower secondary 8.0 -0.003 3.08** -0.07 General(1-3 years) 13.7 -0.003 3.78** -0.07 Vocational 4.1 -0.002 1.78 -0.06 Teacher training 11.8 -0.004 5.26** -0.11 Uuuersecondary 18.3 -0.001 1.5 -0.03 Incomplete 9.9 0.000 0.04 0.00 Diploma holder (D3-D5)a 0.6 -0.006 1.7 -0.14 Diploma holder (D6-D7) with no pre-service 1.3 -0.003 1.07 -0.06 teachertraininga Other -8.0 -0.003 3.08** -0.07 reachers' average years of experience 8.2 0.002 0.84 0.05 - Percentof classroomsin acceptable 48.5 0.000 0.1 0.00 :ondition Per pupilcost of all personnel (FWR, log scale) - 16,681 - - - 0.3052 iegion Butare (referenceregion) 11.3 - - - - - - Byumba 9.8 -0.008 21.49** -19.1 -0.007 22.25* * -17.0 CYWYgu 8.4 -0.004 9.20** -9.2 -0.003 8.79** -7.6 Gikongoro 9.2 -0.006 17.27** -15.1 -0.006 16.7** -14.3 Gisenyi 0.0 -0.004 6.19** -10.8 -0.004 4.83** -8.9 Gitarama 14.8 -0.006 16.20** -15.3 -0.005 17.38** -13.3 Kibungo 6.8 -0.004 10.95** -10.9 -0.004 11.15** -10.7 Kibuye 8.8 -0.006 15.24** -15.0 -0.005 16.28** -12.9 Kigali Rural 12.4 -0.004 12.64** -10.8 -0.004 12.5** -9.3 Kigali Ville 1.5 -0.007 10.54** -17.8 -0.006 9.89** -15.4 Ruhengeri 12.6 -0.005 11.93** -11.2 -0.004 11.22** -8.8 Umutara 4.5 -0.005 8.32** -12.8 -0.006 10.51** -14.5 Jonstant 0.3699 6.71** -0.4645 1.75 \lumber of observations 1,362 1,362 z2 0.29 -.26 `ore: - denotes not applicable. The regressionspecification fc logit model, that is, In[y/(l-y)]=bX, where y is the passrate at the school level, and is a vector of school-ievel characteristics. Robustt-statistics appear in1 table, with one star (*) denoting statistical significance at the 5 percent level and two stam (**) at the 1 percent level. a. Those inthe D3-D5 group receivedtheir diplomas with 3-5 years' secondaryschooling under the old structure ofthe education system, whereas those inthe D6-D7groupreceivedtheirswith 6-7 years ofsecondaryschooling underthe new6-3-3 structurethat was fully inplaceby 1992. The shareofteacherswith pre- serviceteachertraining is sizable only inthe D6-D7group, so the distinction betweenthose with andwithout suchtrainingis madeonly for this group. b.The marginal effect is evaluatedat the samplemeanfor continuousvariables (e.g., years ofteacherexperience), accordingy( I-y)b, where y is thesamplemean and b is the coefficient estimate; for categorical variables (e.g., teacher qualification), it is evaluated relative to the omitted category. The results show the percentagepoint change inthe pass rate inresponseto a unit change inthe correspondingregressor. c. The marginalchange is evaluatedat the samplemean for a change ofFRw 1000inthe per-pupil spendingon schoolpersonnel. Source: Authors' estimatebased on school-level data on examination pass rates for 1998-99 supplied by the National ExaminationCouncil of Rwanda, merged with datafrom MINEDUC's primary schoolcensusdatafor 1999-2000andMINEDUC's 1999 census ofteachers. 180 Table A6.1: Distributionof secondary schools by level andtype of instructionoffered, Rwanda 2000-01 Level andnumber of Combinations of levels and Streamsa I Number of schools bl YPe upper sec. streams I TC All rronc commun alone I x 130 62 ri4 X 1stream X 39 X 32 6 33 26 15 1 1 1 1 All secondary schools 373 a. Exceptfor teacher training, the coursesineachstreamare further divided intovarious fields ofstudy (see table 6.6 for the completelist). Source: Authors' summarybasedonelectronicdata files basedonMINEDUC's 200041 censusof secondary schools. 181 Table A6.2: Number of secondaryschool teachers by qualificationand type of classes taught, Rwanda, 1999 Teaching only Teaching tronc Teaching only tronc commun commun upper classes upper secondary secondary classes classes Teachers with post-upper secondary diploma or university degreea State schools 135 67 105 Libre subsidie'schools 232 334 251 Teachers with upper secondary diplomab State schools 335 163 127 Libre subsidie'schools 488 489 342 All teachers State schools 5 14 240 263 Libre subsidik schools 752 850 632 Public Libre subsidiC Total IXT..-L-- I Y UIIIUGI V I DLUUGIILB -r-h..A--+.. A AA? V . W - t J n nzn W . W J 7 A A A q w.w-rL. (14.72)** (8.67)* * (16.66)** Constant 1.268 4.870 2.866 (1.26) (2.46)* (2.78)** Number of observations 60 102 162 It2 0.85 0.55 0.70 Table A6.4: Correlates of school-level, end-of-cycle ironc commun nationalexamination results, Rwanda, 1999 Pass rate (%) Log-logit xificatio i.e., ln[y/(l Average score Regressionvariables Sample )I=bX)) (linear specification) mean Moc 1 Moc 2 Coefficieni Marginal llarginal effecta 2oefficient effecta hefficient Coefficient Studentkeacher ratio (omitted category =< 18) 18-27 48.53 -0.466 -6.872 -0.283 (1.24) (1.99)* >27 27.94 -0.930 -13.707 -0.217 (2.14)* (1.30) reacher qualification (in YO)(omitted :roup=univ.) Post upper secondary 14.90 -0.038 -0.549 -0.004 (1.57) (0.58) Upper secondary 74.97 -0.027 -0.394 -0.009 (1.33) (1.50 Other 1.70 0.07 1 1.061 0.015 (1.14) (1.07) % teachers with pre-service teacher training 67.92 0.008 0.126 -0.004 (0.61) (0.16) Experienceof teachers (in years) 5.04 -0.031 -0.466 0.003 (0.37) (0.81) Region (omitted region = Butare) Byumba 12.50 -0.265 -4.195 -0.142 -2.456 0.277 0.298 (0.44) (0.24) (1.37) (1.63) CYWPgU 7.35 -0.186 -2.650 0.520 7.371 0.423 0.546 (0.23) (0.86) (1.92)* (2.78)** Gikongoro 8.09 1.203 11.153 1.695 16.383 0.601 0.702 (1.18) (1.62) (3.07)** (3.78)** Gisenyi 13.97 -0.828 -14.155 -0.612 -11.903 -0.319 -0.325 (1.10) (0.90) (1.60) (2.03)** Gitarama 11.76 -0.326 -4.788 0.136 2.179 0.002 0.063 (0.50) (0.21) (0.01) (0.40) Kibungo 7.35 -0.002 1.033 0.465 6.711 -0.040 0.058 (0.00) (0.54) (0.15) (0.24) Kibuye 10.29 -0.824 -14.676 -0.474 -8.931 -0.211 -0.180 (1.32) (0.78) (1.06) (1.01) Kigali Rural 8.09 0.264 3.869 0.306 4.635 0.357 0.378 (0.33) (0.36) (1.23) (1.31) Kigali Ville 2.21 0.620 8.662 0.093 1SO0 0.934 0.866 (1.OO) (0.14) (2.24)** (2.78)** Ruhengeri 5.88 1.090 10.585 1.483 15.346 0.366 0.516 (0.97) (1.38) (1.35) (1.99)* Umutara 1.47 -2.407 -5 1.267 -0.278 -4.991 -0.433 0.104 (1.33) (0.15) (1.09) (0.49) Ln(Cost of personnel per student) 10.01 0.717 0.468 0.252 (1.78) (1.57) Zonstant 3.978 -5.847 3.870 0.586 (2.11)* (1.43) (7.75)* (0.36) Vumber o f observations 136 136 137 137 x2 0.20 0.13 0.31 0.27 Vote: Robustt-statisticsinuarentheses; one star (*)denotes sta icalsigni anceat the 5 cent confid :e level; two I s (**), at t percentlev( a. Marginaleffectsreferto he percentagechangein thepassrate, calculatedat the samplemeans for continuousvariables andwith reference to the omitted categoryfor categoricalvariablesand for an increase ofFRw 1,000 aroundthe samplemeanfor the cost variable. Source: School-leveldataon examinationresults for 1999mergedwith datafrom MINEDUC's 1999-2000censuso fsecondaryschools andthe 1999 census ofteachers. L 184 TableA7. 2: Enrollments inpublic andprivate highereducationinstitutionsby sex, Rwanda, 1984-85to 2001-02 1 Sector andname of institution Sexa 11984-8511986-8: 994-9511995-9611996-97 1999-OC !OOO-0 1 !oOl-o: M 1,299 1,222 2,472 2,942 3,032 3,475 3,705 4,640 Universite'nationale du 1 Rwanda (UNR) F 789 1,006 1,146 1,060 1,135 1,282 ~ T 12::2 12f;l 4,535 4,840 5,922 ~ Kigali Institute of 793 1,169 ScienceandTechnology 356 423 andManagement(KIST) 1.149 1,592 415 689 928 Kigali Institute of 182 270 341 I Education (KIE) 597 959 1,269 88 151 184 377 Kigali Health Institute (KHI) 59 128 199 313 147 279 383 690 Institut supe'rieur 100 145 262 420 i'agronomie et d'e'levage 13 annie 19 52 106 (ISAE) danche I 113 164 314 526 I Institut supkrieur des 51 74 75 59 85 'inancespubliques (ISFP, 16 20 35 39 36 now known as the Institute of Finance and 67 94 110 98 121 Banking (IFB) M 1,299 3,203 5,188 6,445 Subtotal F 273 1,184 1,888 2,283 T 1,572 4,387 -- 7,076 8,728 270 764 1,187 1,658 Universite'libre de Kigali ( U W 136 642 1,125 1,592 406 1,406 2,312 3,250 60 128 117 Universite'lafque de 49 110 163 .- B Kigali (UNILAK) 109 238 280 i- ."5 38 22 Y 33 Y Institut suptrieur de .- 2 pe'dagogie de Gitwe - 1 - 1 - 43 32 39 CI a, (ISPG) 54 &> d 81 72 .C Universite'Adventiste 89 116 191 208 245 d 'Afrique Centrale 60 100 160 237 250 (UAAC) 149 216 351 445 495 1,544 2,005 Subtotal 1,438 2,024 - 2,982 4,029 6,732 8,450 Grandtotal 3,326 4,307 10,058 12,757 Note: Shadedareas indicatedyears when the institutionhadnot yet beenestablished; adashdenotesthat dataare unavailable. a. M, F,andT refers to male, female, andtotal. Source: Rwanda 1981to 1986, 1990, 1994 for data for the 1980s, andpersonalcommunication from each institutionfor the 1990sand2000s. 185 Table A7.3: Distribution ofstudentsby field ofstudy inselectedinstitutions, Rwanda, 2000-01 - ----- blic ir itution! Private institutions" 1. All Fieldo fstudy KIST KHI ISAE ISFP Total % NILAX ISPG JAAC Total No. of - itudents % .-g 371 371 371 `6 Lit.&humanities c Social sciences 0 986 986 Educ. Psychology 0 126 126 126 c 4 Theology 0 48 48 48 Education 562 562 562 .- 30 30 30 -1 Subtotal f Journalism 963 963 13.6 174 1,160 --- 28.8 2,123 19.2 338 338 901 132 1,033 1,371 a Subtotal 5 Law 338 338 4.8 901 132 1,033 25.7 -- 1,371 12.4 Management 686 686 686 .-oEo3f Economics 1,142 1,142 1,363 100 1,463 2,605 m % Accounting 198 198 198 $ Public finance 48 48 Subtotal -- 1,142 686 ---1,876 26.6 1,363 100 198 1,661 --- 41.3 3,537 31.9 Sciences 569 569 569 ag @ Engin.& technology 749 749 749 2SCC8 8 Z Foodscience 157 157 157 Computer science 0 68 68 68 Subtotal -- - --- 569 906 1,475 20.9 68 68 1.7 1,543 13.9 ~ Medicine 419 419 419 Anesthetics 43 43 43 Dentistry 66 66 66 52 52 .-8P 8 Laboratoryscience 52 Physiotherapy 48 48 48 4 Radiology 23 23 23 Mentalhealth 45 45 45 Nursing 174 174 174 Medicaltechnology 0 48 48 48 Subtotal 419 -451 a70 12.3 48 48 --- 1.2 918 8.3 Agronomy 79 79 79 Crop science 1a3 1a3 1a3 Animal husbandry 160 160 160 `63 Agr. engineering 174 174 174 4 Soilscience 3 3 3 Subtotal - 79 - -- - --- 520 599 8.5 599 5.4 Sciences 523 523 54 577 .c 3 .o E $ 8 Arts 417 417 417 I3 Subtotal 940 -- 940 13.3 54 --- 1.3 994 9.0 , p.- 0Commoncore 73 73 73 !3 5 1,330 230 1,530 1,530 & Languagetraining' ` 0 0 Subtotal - 1,330 230 - 73 1.633 - 1,633 oral inspecializedfields 3,510 ,592 940 451 520 48 7,061 100.0 3,250 280 54 100.0 11,085 100.0 verall total ------ - 4,840 ,592 940 681 520 121 8.694 3,250 280 54 -12,718 ote: Blanks denotezero enro ients i le ind ted fie and ir itution lashes motenot applic e. a. The institutions are Universitk nationale du vanda JNR), Kigali Institute Science i d Technology and Management (KIST), Kigali Institute of Education (KIE), Kigali Health Institute (KHI), Institut su@rieur d'agrono ie et d'elevage (ISAE), and Institut su@rieur des financespubliques (ISFP), now known as Instituteo fFinance and Banking(IFB). b. The institutions are Universitb libre de Kigali (ULK), Universitb lai'que de Kigali (UNILAK), Universitb Adventiste d 2fiique Centrale (UAAC), Institut supbrieur de pdagogie de Gime (ISPG), Facultk de thkologiedeButare (FTB). c. Offered to first-year enrollees by the Ecole Pratique des Langues Modemes inUNR and other comparable arrangementsto monolingual students to become bilingual in EnglishandFrench. Source: personalcommunication kom officials at eachinstitution. 186 TableA7.4: Enrollmentsby field of study inthe UniversitiNationaledu Rwanda, 1982-2002 -- Field of studya 1994- 1995- 1996- 1997- ~ 95 96 97 98 139 - 142 111 121 110 121 Law 114 I it: I 117 143 562 670 552 625 ~ -- Literature 213 292 291 375 387 247 345 Literature& humanities 356 I 371 I 559 Educational sciences 199 225 152 238 Ecolenormale supCrieureb 67 168 105 69 Education E c y s o c i a l sciences, & managemenl -1150 1340 1,077 1,259 et techniquesde 114 112 cole de journalisme et communicationd Medicine 127 114 263 296 238 331 Ecole de santkpublique et de nutritione 155 219 158 131 Pharmacy -- Sciences 219 260 160 121 87 204 Applied sciences 81 95 191 245 179 93 Science & technology Ecole supkrieure des techniquesmodernesf 64 64 -- Subtotal 1,572 1,565 3,261 3,813 3,020 3,528 Prespecializedcourses (language upgrading etc.) 1,158 1,020 Overall total 1,572 1,565 3,261 3,813 4,178 4,548 Note: blanks denotezero enrollments. a. The names of the specializedschools are left inthe original Frenchto facilitate Rwandesereaders'recognitionoftheterms b. SchoolofTeacherTraining. c. Schoolof Information Sciences andTechnology. d. SchoolofJoumalismandCommunications. e. SchoolofPublic HealthandNutrition. f. SchoolofModemTechnology. Source: Rwanda 1981to 1986for datafor the 198Os, andpersonalcommunication from officials at the Universitbnationale du Rwanda for datafor lateryears. 187 Table A7.5: Numberof Rwandesestudents on overseasgovernmentscholarships and number enrolled locally, 1967-2002 Numberonoverseas Numberenrolledinlocal Students on overseas cholarships as apercentagt Year scholarships institutions of all enrollmentsa Total ?LOwomen 'ublic sectoi Public& xivate sectors 1967-68 156 n.d. 233 n.d. 40.1 1968-69 233 n.d. 330 n.d. 41.4 1969-70 271 n.d. 446 n.d. 37.8 1970-71 364 n.d. 390 n.d. 48.3 1971-72 508 n.d. 526 n.d. 49.1 1972-73 525 n.d. 489 n.d. 51.8 1973-74 530 n.d. 619 n.d. 46.1 1974-75 498 n.d. 672 n.d. 42.6 1975-76 559 n.d. 657 1,108 46.0 33.5 1976-77 486 n.d. 657 n.d. 42.5 1977-78 610 n.d. 760 n.d. 44.5 1978-79 515 n.d. 976 n.d. 34.5 1979-80 448 n.d. 1,037 n.d. 30.2 1980-81 593 n.d. 1,125 1,243 34.5 32.3 1981-82 533 n.d. 1,211 1,309 30.6 28.9 1982-83 590 8.8 1,317 1,495 30.9 28.3 1983-84 626 7.6 1,367 1,579 31.4 28.4 1984-85 708 7.5 1,572 1,885 31.1 27.3 1985-86 833 8.8 1,565 1,987 34.7 29.5 1 1986-87 945 11.6 1,513 2,029 38.4 31.8 1999-00 981b 31.0 7,673 10,655 11.3 8.4 2000-01 915b 28.3 8,729 12,903 9.5 6.6 2001-02 666b 26.9 10,354 16,668 6.0 3.8 lable; a dashinc ites not applic le. well asprivate domestic enrollments. b. Includes students on scholarship at the Rwandabranchofthe Universitb Adventiste d'Afrique Centrale. See appendixtable A7.6 for their numberineachofthese years. Source: Rwanda 1986a, 1982, and 1986b for the number of studentsabroadfrom, respectively, 1967-68to 1977-78, 1978-79to 1980-81, and 1981-82 to 1986-87; personalcommunication from the MINEDUC's Directionde l'enseignement sup6rieur for the number abroad and percentagewomen from 1999-2000to 2001-02; appendix table A7.1 for the data on domestic higher educationenrollments. 188 Table A7.6: Number of Rwandesestudents on government overseasscholarships by host country, 1984- 85 and 1999-2002 Region & host country 1984-85 1999-2000 2000-2001 2001-2002 Benin 3 0 0 Burkina Faso 0 3 03 3 Burundi 10 31 15 8 C.A.R. 1 0 0 0 Cameroon 1 13 11 13 Cbte d'lvoire 5 0 0 0 Dem. Rep. ofCongo 23 0 0 0 Ethiopia 0 7 6 4 Gabon 0 1 1 0 Ghana 0 3 0 1 Kenya 0 11 12 9 Lesotho 0 2 2 2 Niger 4 1 1 1 Rep.of Congo 2 0 2 1 Senegal 25 8 10 21 SouthAfrica 0 65 87 94 Tanzania 13 3 5 3 Togo 0 1 0 0 Tunisia 2 0 3 3 Uganda 0 141 138 61 Subtotal 89 290 296 224 Austria 21 1 1 0 Belgium 77 20 22 15 Canada 38 28 28 33 France 48 25 15 11 Germany 0 6 6 7 Great Britain 1 40 22 17 Holland 1 0 0 0 Ireland 2 0 0 0 Italy 21 5 4 2 RFA 70 0 0 0 Switzerland 6 8 4 2 USA 33 18 12 9 Subtotal 318 151 114 96 Algeria 42 24 31 40 E m t 0 1 1 0 Israel 0 1 1 1 Libya 4 0 0 0 SaudiArabia 2 0 0 0 Syria 1 0 0 0 Subtotal 49 26 33 41 Bulgaria 6 0 0 0 Czechoslovakia 2 0 0 0 Poland 2 4 4 3 Romania 6 0 0 0 Ukraine 0 1 1 1 USSRRussia 207 0 22 15 Yugoslavia 2 0 0 0 Subtotal 225 5 27 19 China 25 25 32 35 India 0 403 373 231 Jamaica 1 0 0 0 Mexico 1 0 0 0 Philippines 0 2 2 0 - Subtotal 27 430 407 266 .I1countriesexcluding students at 902 877 646 11countriesincluding studentsat 708 981 915 666 a. UAAC refers to the Universiti Adventiste d'Afrique Centrale. The institution has a local campus inRwanda. Source: Rwanda 1981 to 1986 for data for 1984-85 and personal communication from MINEDUC's Direction de l'enseignement supkrieur. 189 Table A7.7: Academicfee, welfare, andtravel costs paidby the Rwandesegovernment for Rwandese students on governmentoverseasscholarships, circa 2002 Air ticket per course of Hostcountry Average annual Annualwelfare grant by level of study (US$) academic fees (US$) study Jndergraduate Master's Ph.D. Outbound Inbound Algeria Paidbyhost country 600 563 997 China Paidbyhost country 1,800 1,800 1,800 1,156 1,082 India Paidby host country 1,800 2,400 2,400 408 408 Russia Paidby host country 1,800 1,800 1,800 709 684 Poland Paidby host country 3,600 3,600 3,601 709 684 Uganda n.d. 750 1,250 - 1,550 - 2,050 1,550 - 2,050 0 0 SouthAfrica 1.091 - 1.121 3,415 3,488 4.247 406 496 Tunisia n.d. 600 600 600 604 905 Burundi n.d. 2,400 0 0 Congo, DR n.d. 2,400 488 488 Ethiopia n.d. 2,400 321 321 BurkinaFaso n.d. 3,600 4,800 6,000 767 947 Cameroon n.d. 3,600 4,800 563 563 Egypt n.d. 3,600 4,800 6,000 469 469 Gabon n.d. 3,600 4,800 6,000 522 947 Ghana n.d. 3,600 4,800 6,000 n.d. n.d. Kenya n.d. 3,600 207 268 Niger n.d. 3,600 4,800 6,000 1,036 872 Senegal n.d. 3,600 4,800 6,000 857 1,155 Tanzania n.d. 3,600 286 305 Lesotho n.d. 4,600 455 455 Ukraine n.d. 6,000 709 684 Israel n.d. 7,200 787 1,083 Canada 11,000- 14,OO1 7,200 -8,400 5,400 - 10,20C 1,000 - 14,00( 888 875 Austria 11,000-14,000 8,400 10,200 14,400 663 662 Belgium 11,000-14,000 8,400 10,200 14,400 663 662 France 11,000-14,000 8,400 10,200 14,400 663 662 Germany 11,000-14,000 8,400 10,200 14,400 709 569 GreatBritain 11,000- 14,000 8,400 8,400 9,000 615 615 Italy 11,000-14,000 8,400 10,200 14,400 663 662 Switzerland 11,000-14,000 8,400 10,200 14,400 663 662 USA 11,000-14,000 n.d. 8,400 9,000 888 876 Note: n.d. denotes tk io dataare available. Inm African countrie: he fees are likelyto be modestandpi iblynot greaternthe level shown for SouthAfr Source: Personalcommunication from the MinistryofEducation, Direction de l'enseignement supkrieur. 190 Table A7.8: Number of faculty by institutionand nationality,Rwanda, 1985-2001 Sectorandnameof Institution Year Full-time faculty Part-time & visiting faculty Nationals Total Nationals Expatriates Total All staff 1985-86 163 1 IExpatriates1I 54 217 121 1986-87 172 47 219 177 1994-95 125 1995-96 160 Universitk nationale du 1996-97 164 Rwanda (UNR) 1997-98 233 1998-99 1999-2000a 2000-01a Kigali Institute of Scienceand 1997-98 Technology and Management 1998-99 (KIST) 1999-2000 2000-01 Kigali Institute of Education 1998-99 W E ) 1999-2000 2000-01 1998-99 Kigali Health Institute (KHI) 1999-00 2000-01 lnstitutsupkrieur 1997-98 d'agronomie et d'devage 1998-99 (ISAE) 1999-2000 2000-01 lnstitut suptrieur desfinances 1986-87 Dubliaues (ISFP)' 2000-01 1996-97 Universiti Iibre de Kigali 1997-98 [ULK) 1998-99 1999-2000 2000-01 UniversittAdventiste 1985-86 I'AfriqueCentrale(UAAC) 1986-87 2001-02 1997-98 Universite'lai'que de Kigali :UNILAK)~ rnstitutsuptrieur de Note: Blanksdenote that dataare unavailable. a. Excludes teachers in the Ecole pratique des langues modemes, a language upgrading program for fust-year students, which was launched in 1996-97. They numbered35 in200061. a. IncludesUnitedNationsVolunteerswho numbered21 in 1999-00and22 in200061. b. Includes35 teachers in the Ecole pratiquedes langues modemes, a programstarted in 1996-97 to enable first-year studentsto achieve bilingualism in Frenchand English.The numberofsuchteachersinearlieryears is unknown. c. Now !mown as the Institute ofFinanceandBanking(IFB). d. The distribution ofteachersby nationality andemploymentstatus is highlyapproximate. e. So-calledpermanentteachers andvisiting faculty have almost the sameteachingloads, about 15 a week. Source:Rwanda 1986 for dataon UNR inthe 1980s;personalcommunicationfrom eachinstitution for all other institutions andyears. 191 Table AI. 9: Current and proposedfuture arrangementsregardingstudent bursariesat the Universite' nationale du Rwanda Type o f student o f bursaryper amount Monthly amount deducted from bursary cash received by amount Of student Meals Housing Health services each student -5 y Studentsfedand E E 8 housed on campus 11,000 5,400 650 500 4,450 uB % 8 Other students 11,000 0 0 500 10,500 35ta 8 4 Studentsfed and housed on campus 25,000 15,000 1,300 500 8,200 Q- & 8 Other students 25,000 0 0 500 24,500 Note: The amountsabove apply for 10months ofthe year only Duringthe other 2 months, students receivenobursary. Source: Personalcommunication from UNR officials inthe context ofthe presentstudy. Table A8. 1: Earning functions amongwage earners, Rwanda, 2000 Regressionvariablesa Coefficient t Coefficient t Coefficient t Zonstant 9.624 33.07 ** 9.747 33.11 ** 9.794 33.36 ** Years o f schooling 0.112 14.73 ** Level o f education attainedb Primary 0.126 1.92 1-3 years o f schooling 0.018 0.22 4-6 years o f schooling 0.207 2.87 ** Secondary vocational & technical 0.578 5.91 ** 0.620 6.25 ** General secondary 0.929 9.88 ** 0.974 10.18 ** Higher education 1.838 14.37 1.888 14.64 ** 4ge 0.076 4.45 ** 0.080 4.60 **** 0.077 4.40 ** 4g9 0.001 2.82 ** 0.001 3.05 ** 0.001 2.87 ** Male/Female 0.295 5.76 ** 0.324 6.29 ** 0.320 6.23 ** Employment sector' Quasi-public 0.083 0.70 0.090 0.79 0.097 0.84 Private formal 0.064 5.61 ** 0.84 0.070 0.90 0.073 5.35 ** 0.93 Private informal 0.450 0.464 5.61 ** 0.442 Economic sectord Mining 0.033 0.12 0.213 0.79 0.135 5.55 ** 5.24 ** 0.48 Manufacturing 0.532 4.54 ** 0.664 5.55 ** 0.626 Energy 0.933 4.60 ** 1.141 5.88 ** 1.112 Construction andpublic works 0.642 6.05 ** 0.807 7.46 ** 0.765 7.04 ** Commerce, hotel and restaurant 0.247 2.02* 0.451 3.72 ** 0.405 3.31 ** Transport and communication 0.870 8.04 ** 0.977 8.79 ** 0.933 8.26 ** Banking and insurance 0.679 5.86 ** 0.715 5.99 ** 0.679 5.64 ** Services 0.294 4.02 ** 0.407 5.66 ** 0.369 5.02 ** Other 0.754 3.71 ** 0.873 3.81 ** 0.834 3.66 ** Provincee Butare 0.165 1.70 0.186 0.808 0.171 1.81 Byumba 0.435 3.97 ** 0.436 4.07 ** 0.435 4.07 ** cyan@@ 0.525 5.34 ** 0.516 5.16 ** 0.507 5.14 ** Gikongoro 0.622 5.94 ** 0.615 6.00 ** 0.614 6.02 ** Gisenyi 0.149 1.24 0.196 1.55 0.194 1.54 Gitarama 0.258 2.52* 0.233 2.29* 0.213 2.09* Kibungo 0.591 5.21 ** 0.606 5.30 ** 0.612 5.35 ** Kibuye 0.657 4.25 ** 0.694 4.63 ** 0.674 4.54 ** Ngali Kigali 0.167 1.85 0.165 1.81 0.159 8.77 ** 1.77 Ruhengeri 0.779 8.42 ** 1.97* 8.97 ** 0.793 Umutara 0.134 1.08 0.106 0.86 0.099 5.30 ** 0.79 Secondary activity (occupation) 0.554 5.28 ** 0.550 5.14 ** 0.560 RZ 0.58 0.58 0.58 1866 1866 Note: One star (*)demotes statisticalsignificance at the 5 percent level andtwo stars (**) at the 1percent level. a. The dependant variable is the annualsalary (including benefits) from the main andsecondaryjobs. The resultsreflect estimates basedon a semi- logarithmic modelspecificationcorrectedfor heteroskedasticity,butnot for selectionbias. b. Relativeto those with noschooling. c. Relativeto wage eamers inthe public sector. d. Relativeto wageearnersinaprimarysectorjob. e. Relative to Kigali city. Source: Authors' estimates basedonthe 1999-2001Household Living Conditions Survey. m o\ m P ' , O O V ) ~ m w w m ~ m m m a ~ m o m m m w w 0 - 0 o m ~ w c r~ m m mw d - ~ l - o m m m : m o - 195 FigureA4.1: Cumulative shares of a hypotheticalcohortby educationalattainment and ofpublic spendingon educationbenefitingthe cohort :f 100 . I -- I 1 I 80 . I I I curve s I h 0 8 60 c, ~~ I rtI I I E rt I 2 40 20 Gini-coefficient= A/(A+ B) 0 0 20 40 60 80 100 Populationshare ("h) Source: Authors' construction,basedonhypotheticaldataon gross enrollment ratios andshareso fpublic spendingon educationindicatedinfootnote 13 inthis chapter. 197 References Ainsworth, Martha andDeonFilmer.2002. "Poverty, AIDS and Children's Schoo1ing:A Targeting Dilemma," Paperpresentedat the annual meetings o f the PopulationAssociation o f America (May 9-11,2002), Atlanta, Georgia. Bonflis, Safari. 2002. "A Word fromthe Dean," (online) http://www.nur.ac.rw/technoll .htm, [AccessedJanuary 9,20031. Cuenin, Serge. 1992, "Bourses et prgts, propositions pour unfinancement mixte de l'enseignement supkrieur au Rwanda", report prepared for UNESCO, mimeograph. Demery, Lionel. 2000. "Benefit Incidence: A Practitioner's Guide." The World Bank, Washington, DC, mimeo. Devarajan, Shantanayan.2001. "Growth i s not enough," World Bank, Washington D.C. (processed). Devarajan, Shantanayan; and Ritva Reinikka. 2002. "Making serviceswork for the poor," Washington D.C. (processed). Foster, A. and M.Rosenzweig 1996. "Technical changesandhuman capital returns and investments: Evidence from the greenrevolution," American Economic Review 86 (4): 931-952. Government o fMadagascar. 2000. Stratkgie nationale de lutte contre lapauvretk Primature. 2000-20015. versionprkliminaire (mimeo), Antananarivo, Madagascar: Govt. o fRwanda & UNICEF. 2001. Enqukte a indicators multiples (MICS2). Kigali: MinistBre desJnances et de laplanification kconomique,Direction de la statistique. Govt. o fRwanda (1981to 1986) Annuaires statistiquede 1'enseignementsupkrieur au Rwanda, MINESUPRES: Kigali. Govt. o fRwanda (1994). "L'assistance d'urgence et lareconstruction dusystbme Cducatifau Rwanda", background document for a seminar on emergency assistancefor the reconstruction o fthe education system inRwanda, 7-8 November, 1994, Kigali. Govt. o fRwanda (1994). "L'assistance d'urgence et lareconstruction du systbme 6ducatif au Rwanda", background document for a seminar on emergency assistancefor the reconstruction o fthe education systeminRwanda, 7-8 November, 1994, Kigali. Govt. o fRwanda (Nov). 1985. Annuaire Statistiquede I'EnseignementSupkrieur 1983-84.Kigali. 198 Govt. o f Rwanda (undated, a). Statistiques de 1'EnseignementAnnke Scolaire 1989-90 and 1990-91, Directionde la Planification. Ministire de 1'Enseignement Primarire et Secondaire, Kigali. Govt. o fRwanda (undated, b). RecensementStatistique Annie Scolaire 2000-01. Ministbre de 1'Education. Direction des Etudes et de la Planification de l'Education, Kigali. Govt. ofRwanda (undated, c), Education pour Tous. Bilana l'an 2000. Ministbre de 1'Education: Directionde laPlanification de I'Education. Govt. ofRwanda (various years) Annuaires statistiques(1981-82-1986-87), MINESUPRES:Kigali. Govt. ofRwanda (various years). Official Gazetteof theRepublic of Rwanda. MinistryofFinanceandEconomic Planning, Kigali. Govt. ofRwanda 1985 (Nov). Annuaire Statistique de I'EnseignementSupkrieur 1983-84.MinistryofHigher Education and Scientific Research.Kigali. Govt. o fRwanda 1985. Rapport Annuel Exercise 1985. MinistryofHigher Education and Scientific Research, Kigali. Govt.ofRwanda 1999(Nov.). Education SectorReview. Objectivesand Expenditure Scenarios2000-03, Ministry o fFinance and Economic Planning, Kigali. Govt. of Rwanda 1999. EnseignementPrimaire Statistiques. Annke Scolaire 1998- 99. Ministbre de 1'Education. Directiondes Etudeset de la Planification, Kigali. Govt. o fRwanda2000 (Jan). Budget Execution Exercisefor 1999. Ministry o f Finance and Economic Planning, Kigali. Govt. ofRwanda2000a. EnseignementPrimaire Statistiques. Annie Scolaire 1999- 2000. Ministbre de 1'Education. Directiondes Etudes et de laPlanification, Kigali. Govt. o fRwanda 2000b Law modifying andcomplementing LawNo. 25/99 o f 31/12/1999 determining the state finances for the 2000 fiscal year. Oflcial Gazetteof the Republic of Rwanda. Kigali: Ministry o fFinance. Govt. ofRwanda 2000c peb). Enqukte dhographique et desantk EDSR-I% Rapport prkliminaire. Kigali: Office national de lapopulation and Calverton, Maryland, USA: Demographic and HealthSurveys, ORC Macro International, Inca Govt. ofRwanda 2001. Law determining the state financesfor the 2001 fiscal year. OfJicial Gazetteof theRepublic of Rwanda. Kigali: Ministryo fFinance. Govt. ofRwanda n.d.Statisitiques de 1'Enseignement1989-90. MinistGre de 1'EnseignementPrimaire et Secondaire, Direction de la Planijkation, Kigali. 199 Govt. ofRwanda. 1986.Annuaire statistique de 1'enseignementsupe'rieur au Rwanda, MinistBre de 1'enseignementet de la recherche scientiJique, Kigali. Govt. ofRwanda. 1982. Statistiques de 1'enseignementsupe'rieur. MinistBre de 1'enseignementet de la recherche scientiJique,Kigali. Govt. o fRwanda. 1986a. "Aspects quantitatgs de l'e'volution de 1'enseignement supe'rieurau Rwanda de I963 ii 1985," Ministire de 1'enseignementet de la recherche scientifique, Kigali. Govt.o fRwanda. 1990.Rationalisation des ressourcesA I'UNRet perspectivesde financement de l'enseignement supdrieur au Rwanda, Rapport d'une commission d'ktude, Ministere de 1'EnseignementSupdrieur et de la Recherche Scientifique, Kigali. Govt. o fRwanda. 1990. Rationalisation des ressourcesA 1'UNR et perspectives de financement de l'enseignement supdrieur au Rwanda, Rapport d'une commission d'ktude, Ministere de l'enseignement supdrieur et de la Cuenin, Serge. 1992, "Bourses et prets, propositions pour unfinancement mixte de l'enseignementsupkrieur au Rwanda", report prepared for UNESCO, mimeograph. Govt. ofRwanda. 1994(April). Recensementge'ne'ralde lapopulation et de 1'habitat au I5 aout 1991. Kigali, Rwanda: Commision nationale de recensement, Ministe're duplan, and Service national de recensement. Govt. of Rwanda. 1996. Law no10/96 of 16/10/1996fixing the Ordinary Budget and the Development Budget for the financial year 1996, Oflcial Gazetteof the Republic of Rwanda. Kigali: Ministry ofFinance. Govt. ofRwanda. 1998. Law amending andcompleting Law No. 01/98 of 31/12/1997 determining the state finances for the 1998 fiscal year, Ofzcial Gazetteof theRepublic of Rwanda. Kigali: Ministry of Finance. Govt. ofRwanda. 1999 (Nov.) Education Sector Review, Objectives and Expenditure Scenarios,2000-2002. Kigali: Ministry o fFinance and Economic Planning (mimeo). Govt. ofRwanda. 2000a (Jan.). Budget executionfor 1999. Kigali: Ministryof Finance and Economic Planning. Govt. ofRwanda. 2000b. Enseignement SecondaireStatistiques. Annkes Scolaires 1998-99, Ministryo fEducation, Planning and ResearchDepartment, Kigali. October. Govt. ofRwanda. 2000c. Enseignement SecondaireStatistiques. Annkes Scolaires 1999-00, Ministryo fEducation. Planning and ResearchDepartment, Kigali. March. Govt. o fRwanda. 2000d. "Proposed highereducation sub-sector policy", mimeograph (March), Ministry o f Education, Kigali. 200 Govt. o fRwanda. 2000d. "Proposedhighereducation sub-sector policy", mimeograph (March), Ministryo fEducation, Kigali. Govt. ofRwanda. 2001. EnseignementSecondaireStatistiques. Annkes Scolaires 2000-01, Ministryof Education. Directiondes Etudes et Planification, Kigali. June. Govt. o f Rwanda. 2002 (Feb). A Profle of Poverty in Rwanda. An Analysis based on the results of the Household Living Condition Survey 1999-2001(English Version) Ministryo fFinance and Economic Planning, National Poverty ReductionProgramme and SatisticsDepartment, Kigali. Govt. of Rwanda. 2002a (June). Poverty Reduction Strategy Paper. Kigali: Ministry ofFinance and Economic Planning. Govt. o fRwanda. 2002b (Feb). A Profle of Poverty in Rwanda. An Analysis based on the results of theHousehold Living Condition Survey 1999-2001. Kigali: MinistryofFinance. International Bureau of Education (IBE), 2001. L e dkveloppement de l'kducation (1990-2000), Rapportnational duRwanda.Geneva: IBE Johanson, Richard K.; and Arvil Van Adams. 2003 (forthcoming). Skills Development inSub SaharanAfrica. Washington D.C.: The World Bank. Kelley, MichaelJ. 1991.Education in aDeclining Economy: The Case of Zambia,1975-1985,Washington, DC: World Bank. Kereni, Ildephonse. 2002. "A Word from the Director," (online) httD://www.nur.ac.rw/eulml.htm, [AccessedJanuary 14,20031. Kigali Institute o f Science, Technology and Management. 2003. "KIST Interim Administration and Organizational Chart," (online) httdlwww.kist.ac.rw1, [AccessedJanuary 14,20031. Kline, Patrick, 2002. "Note on the Distributionof Public Spending inRwanda in 2000: A Benefit-Incidence Analysis for Healthand Education," Mimeograph (June). Linguyeneza, Venuste. 2001. "Rwanda -Jubilee Year of Evangelisation," White Fathers- WhiteSisters,Number357, April-May. Lockheed, M., D.T. Jamison and L.Lau 1980. "Farmer education and farm efficiency: A survey," Economic Development and Cultural Change,29 (1): 37-76. Mazimpaka, Jolly and G.F. Daniel. 2000. "Post-Genocide Reconstructionof Education inRwanda: An Overview," Occasional Paper, Accra-North, Ghana: Association ofAfrican Universities, (online) http:/lwww.aau.org/english/documents/genocide.htm#Introduction, [AccessedJanuary 13,20031. 201 Mingat Alain, RamahatraRakotomalala; and Jee-PengTan 2002. Financing Educationfor All by 2015: Simulationsfor 33 Africa Countries. Africa RegionHumanDevelopment Working Paper Series. Washington D.C.; The World Bank. Mingat, Alain; and Bruno Suchaut. 2000. Les sythmes Cducatifs afiicains. Une analyse Cconomiquecomparative. Brussels, Belgium: DeBoekUniversitC. Mingat, Alain; and Bruno Suchaut. 2000. Les systkmes kducatgs africains. Une analyse kconomique comparative, Brussels, Belgium: De Boek Universitk. Mingat, Alain; RamahatraRakotomalala and Jee-PengTan (2002). Financing Educationfor All by 201.5: Simulationsfor 33 African Countries. Africa RegionHumanDevelopment Working Paper Series, Washington D.C.: The World Bank. NationalUniversityo fRwanda (NUR).2002a. "History," (online) http://www.nur.ac.rw/history.htm, [Accessed January 9,20031. NationalUniversity of Rwanda (NUR). 2002b. "Historical Background," (online) http://www.nur.ac.rw/ejc4.htm, [AccessedJanuary 9,20031. Patrinos, HarryAnthony (2002). "A Review o fDemand-Side Financing Initiatives in Education", (processed), HumanDevelopment Department, LatinAmerica and Caribbean Region, the World Bank, Washington D.C. (August) Patrinos, Harry Anthony; and David LakshmananAriasingam (1997). Decentralization of Education: Demand-Side Financing Directions in Development Series (June), Washington D.C.: The World Bank. Rutazibwa, Gerard. 2002. "A Word from the Dean," (online) htts://www.nur.ac.rw/econom1.htm, [AccessedJanuary 9, 20031. Rwamasirabo, Emile. 2002. "Rector's Message," (online) htb://www.nur.ac.rw/economl .htm, [AccessedJanuary 9,20031. Sicherman, N. 1991. "Overeducation inthe labor market, "Journal ofLabor Economics, 9 (2): 101-122. Task Force on HigherEducation and Society (TFHES). 2000. HigherEducation in Developing Countries. Promise or Peril. Published for the TFHES by the World Bank. Task Force on HigherEducationand Society (TFHES). 2000. Higher Education in Developing Countries: Promise or Peril, Washington, DC: World Bank Transitional National Assembly. 2001."Law Establishing Kigali HealthInstitute has beenVoted,'' July 30, (online) h~://www.rwanda~arliament.gov.rw/30-iuly.htm, [AccessedJanuary 14,20031. 202 UNESCO. 2001. Educationpour tous: rapport de la zoneAfrique Subsaharienne. ADEA working group onEducation Statistics, National Education Statistical InformationSystems (NESIS), Harare, Zimbabwe. UNESCO Regional Office for Education inAfrica (BREDA),World Bank and UNESCO Institute o f Statistics (UIS). 2002. Universal Primary Education goalfor all. Statistical-Document MLVEDAF VIII. Document preparedfor the EightConferenceof Ministers ofEducation ofAfrican Member States (MINEDAF)heldinDar-es-Salaam,TanzaniaduringDecember2-6,2002. UNESCO. 1995.Plan-cadre de reconstructiondu systbme Cducatif 1995-2000, Kigali. UNESCO. 1997. International Standard Classijkation of Education, ISCED, 1997,Paris: UNESCO. UniversityofBuffalo, undated, "The BudgetProcessat UB," (online) http://www.provost.buffalo.edu/policies/BUDPROC.Ddf[AccessedJanuary 23,20031. Verdugo, R. andN.Verdugo 1989. "The impact o f surplus schooling on earnings: some additional findings," Journal of Human Resources,24 (4): 629-643. World Bank (Jan) 1989. Rapport sur le CoGt et Financement de IEnseignement Primaire, Skcondaire et Supkrieure." Report No. 7362-RW. World Bank: Africa Region. World Bank. 1999. Madagascar. An Agendafor Growthand Poverty Reduction. WorldBank Country Study. Washington D.C. World Bank. 2002. Education and Training in Madagascar. Toward a Policy Agendafor Economic Growth and Poverty Reduction. A World Bank Country Study. Washington D.C.: The World Bank.