Report No. 28037-ET Ethiopia Education in Ethiopia Strengthening the Foundation for Sustainable Progress February 28 , 2005 AFTH3 Human Development Department Africa Region Document of the World Bank CURRENCYEQUIVALENTS (ExchangeRate EffectiveAugust 6,2004) CurrencyUnit = Birr Birr 8.45 = US$1 US$1 = 1.46275SDR FISCALYEAR July 8 - July 7 ABBREVIATIONSAND ACRONYMS InstructionalManagementby Parents, Communitiesand ANDM AmharaNationality DemocraticMovement IMPACT Teachers BESO Basic Education System Overhaul LFS Labor Force Survey BRAC Bangladesh Rural Advancement Committee MDGs MillenniumDevelopment Goals DHS Demographic and Health Survey MOE Ministry of Education EFA Education for All MOFED Ministry of Financeand EconomicDevelopment Organization for Economic Cooperationand EMIS Education Management Information System OECD Development EPLF EritreanPeople's LiberationFront OPDO Oromo People'sDemocraticOrganization EPRDF Ethiopia Peoples' RevolutionaryDemocraticFront PRSP Poverty Reduction StrategyPaper ERHS Ethiopia Rural Household Survey PRSC Poverty Reduction Support Credit SustainableDevelopment and PovertyReduction ESACLFS Ethiopia StandAlone Child Labor Force Survey SDPRF' Programme ESDP Education SectorDevelopmentProgram SEPDO SouthEthiopiaPeoples' DemocraticOrganization GDP GrossDomesticProduct TGE TransitionalGovernment of Ethiopia GER Gross EnrollmentRatio TPLF TigrayPeople's Liberation Front Household Income and ConsumptionExpenditure HICES TTC TeacherTrainingCollege Survey HIPC HeavilyIndebtedPoor Countries nr TeacherTrainingInstitute Human ImmunodeficiencyVirudAcquired HIV/AIDS W E T Technicaland Vocational Educationand Training ImmunodeficiencySyndrome United Nations Educational, Scientificand Cultural IDA InternationalDevelopmentAssociation UNESCO-UIS Organization IMF InternationalMonetaryFund WMS WelfareMonitoringSurvey CountryDirector: Ishac Diwan SectorManager: Laura Frigenti -i- Tableof Contents Acknowledgments ...................................................................................................................... xiii ... ExecutiveSummary .................................................................................................................... xv Designingthe OverallPolicyFramework ..................................................................... xxi GettingResourcesto Schools. Communities. andHouseholds ................................ xxxii TakingAdvantageofDecentralizationto GetResults .................................................. xl Conclusion ....................................................................................................................... xlii Chapter 1 Demography.Economy. andGovernmentFinance . ........................................... 1 DemographicandSocialConditions ................................................................................ 4 EconomicConditions.. ......................................................................................................... 8 PatternofGovernmentFinance ..................................................................................... 10 FiscalDecentralization . . ...................................................................................................... 10 Trends inOverallGovernment Revenues andExpenditure ............................................. Aggregate Trends inPublic Spendingon Education........................................................ 11 14 Conclusion ........................................................................................................................ 17 Chapter 2. EnrollmentsandPatternsofStudent Flow ...................................................... 19 Structure ofthe EducationSystem ................................................................................ 19 Overview ofEnrollmentTrends Primary and Secondary Education.................................................................................... ..................................................................................... 21 PostsecondaryEducation.................................................................................................. 22 23 EnrollmentsinGovernmentandNongovernmentEstablishments ............................ 23 EveningClasses.DistanceEducation. andNonformalBasicEducation .................... 25 Evening Classes................................................................................................................ 25 NonformalEducation.,...................................................................................................... Distance Education ........................................................................................................... 26 27 SummaryandCross-country Comparisons ...................................................................... 27 Trends andLevelsinthe GrossEnrollmentRatio PrimaryEducation............................................................................................................ ....................................................... 29 30 31 Higher Education .............................................................................................................. SecondaryEducation ........................................................................................................ 32 StudentFlow inPrimary andSecondaryEducation .................................................... 33 Years of Schooling Completed andAdult Literacy .......................................................... 34 .11. .. Entryto Grade 1................................................................................................................ 35 Completion Rates.............................................................................................................. Survival Rates.......;........................................................................................................... 37 GradeRepetition............................................................................................................... 40 43 Summarizing the Efficiency o f StudentFlow ................................................................... 46 Issuesfor PolicyDevelopment ........................................................................................ Structure ofthe Education SystemandLoci of StudentFlow Regulation....................... 48 48 Indicators for Monitoring Progresstoward the EducationMDGs .................................... 50 Conclusion ........................................................................................................................ 51 Chapter3. EducationExpenditures ..................................................................................... 52 OverallPatternofRecurrentPublicSpendingon Education ..................................... National andRegionalTrends inAggregate Spending.,................................................... 52 52 Trendsinthe Functional Composition of Spending ......................................................... 55 Regional Diversityinthe Functional Composition of Spending ...................................... 57 A Closer Look at RecurrentPublicSpendingon Educationin2001-02 ...................59 .............60 Spending on StaffRemuneration inthe other Subsectors ................................................ Spending on Teachers and Other PersonnelinPrimary and Secondary Schools Consolidated Picture of Spending on Staff andother Recurrent Costs ............................ 63 64 Relation between Spending andPrimary School Completion Rates ................................ 68 PublicSpendingPer Student byLevelandType of Education .................................. NationalandRegionalEstimates ...................................................................................... 69 Decomposition of SpendingPer Student inPrimary and Secondary Education 74 Summaryofthe PattemofPublic Spendingper Student ................................................. ..............69 82 HouseholdSpendingonEducation ................................................................................ 82 Issuesfor PolicyDevelopment ........................................................................................ 85 Conclusion ........................................................................................................................ Chapter 4. DisparitiesinEnrollments.StudentFlow. andBenefitIncidence ..................95 97 Overview of ParticipationRates .. ..................................................................................... 97 DisparitiesinStudent Flow Patternsin PrimaryEducation ..................................... 102 Distributionof PublicSpendingon Education ........................................................... 109 SpendingPattem Associated with Structural Biases ...................................................... 109 SpendingPattem Associated with Social Selectivity inEducation ................................ 111 Issuesfor PolicyDevelopment ...................................................................................... 113 Managing Constraints on the Supply Side ...................................................................... 114 Addressing Constraints onthe Demand Side .................................................................. 115 Conclusion ...................................................................................................................... 119 Chapter 5 . Service Delivery inPrimary and Secondary Education ................................ 120 Overview of the Network of Schools ............................................................................ 120 Selected Characteristics of Schools .............................................................................. 125 Staffing Patterns ............................................................................................................ 128 Administrative Staff ........................................................................................................ 128 Teachers .......................................................................................................................... 129 Teacher Allocation across Government Schools ......................................................... 134 Economies of Scale in Service Delivery ....................................................................... 138 Student Achievement ..................................................................................................... Overview o fPerformance onNationalExaminations .................................................... 142 142 Results o f Two Recent Assessments of StudentAchievement Additional Results Basedon Grade 8 Examination Performance inOromiya...............144 ....................................... 146 Issues for Policy Development ...................................................................................... 150 Conclusion ...................................................................................................................... 160 Chapter 6. Aspects of the Market and NonmarketBenefits of Schooling ...................... 161 Employment Patterns and Workers' Educational Attainment ................................. 161 The Market Returns to Education ............................................................................... 166 The Impact of Education on Poverty and Nonmarket Outcomes ............................. 170 Issues for Policy Development ...................................................................................... 174 Conclusion ...................................................................................................................... 176 Technical Note 1: Preparation of the Data on Current Public Expenditure on Education in Ethiopia................................................................................................................................ 177 GeneralDefinitions ......................................................................................................... 177 Economic Classification ................................................................................................. 178 Detailed Estimates o fthe Composition of Spendingin 1994E.C. (2001-02) ...............178 Functional Classification ................................................................................................ 180 Nonsalary Expenditures Wage Billo f Teaching andNonteaching StaffinGrades 14.5-8. and9-12 ...............180 .................................................................................................. 180 Teaching andNonteaching StaffRemuneration inTTIs. TTCs. andHigher Education Institutions........................................................................................................... 181 Technical Note 2: Estimating Student Flow Profiles ............................................................. 183 The Basic Data................................................................................................................ 183 .iv. Pseudo-Cohort Method................................................................................................... 184 UNESCO Reconstructed-Cohort Method ....................................................................... 186 Computing the Repetition, Promotion, andDropout Rates............................................ 186 The Results Usingthe Classical Algorithm.................................................................... SettingUp the Calculations ............................................................................................ 188 The Resultswith aModified Algorithm......................................................................... 188 189 Composite Cohort Method .............................................................................................. 190 Technical Note 3: Selected Studies on Rates of Return to Education inEthiopia ..............192 An Overview of SevenStudies....................................................................................... 192 Summary of the Estimates ofthe Returns to Education ................................................. 196 References............................................................................................................................. 232 - v - ListofTables Table 1. Low Rates of Entryto Grade 1and Survival to Grade 4, Especially inRuralAreas, despiteUnambiguous Overall Improvement since 1993-94 ............................................. xxiii ... Table 2. HighTeacher Cost Leads to Adverse Tradeoff against Other School InputsinEthiopia .......................................................................................................................................... ... XXVlll Table 3. Scope Exists for Increasingthe Efficiency of Time Use by Teachers beyond Grade 4 ............................................................................................................................................. xxx Table 4. Possible Indicators for Monitoring Progress inPrimary Education inEthiopia .......xxxii * . Table 5. Urban-Rural Disparities inPrimarySchool Participation are Especially Wide in Table 6. Primary Schools are Still Too Inaccessiblefor Many Children inRuralEthiopia....xxxiii Ethiopia ............................................................................................................................ xxxiv Table 7. Incompletenesso f Instructional Program CharacterizesManyRuralPrimary Schools .......................................................................................................................................... XXXVl Table 8. The Direct Cost o fPrimary Schooling CanBe Significant for the Poorest Families ......................................................................................................................................... Table 1.l. ..............................XXXVll..5 Table 1.2. SelectedHealthIndicators, Ethiopia, 1984to 1999-2000 ............................................. SelectedPopulation Characteristics, Ethiopia, 1984to 1999-2000 6 Table 1.3. Percentageo f Childrenbelow 15 Years ofAge Who Have Lost One or BothParents, Ethiopiaand Other African Countries, circa 1999 .................................................................. 7 Table 1.4. Sources o fRecent Economic Growth, Ethiopia, 1992-93 to 1999-00 ......................... 9 Table 1.5. Trends inOverall Government Revenues, Ethiopia, 1980-81 to 2001-02 .................12 Table 1.6. Trends inTotal Government Expenditure, Ethiopia, 1980-81 to 2001-02 ..................15 .................13 Table 1.7. Trends inPublic Spending on Education, Ethiopia, 1980-8 1to 2001-02 Table 2.1. Enrollments by Level inGovernment andNongovernment Institutions, Ethiopia, 1967-68 to 2001-02 ............................................................................................................. Table 2.2. Numbero f StudentsbyType and Level o fEducation, Ethiopia, 2001-02 .................21 .................26 Table 2.4. Three Indicators o fEntryto Grade 1, Ethiopia, 1993-94 to 2000-01 ........................ Table 2.3. Primary Gross Enrollment Ratios (GER), Ethiopia, 1993-94 to 2001-02 30 Table 2.5. Cohort SurvivalRatesUsingDifferent Estimation Methods, Ethiopia, 2000-01 .......36 39 Table 2.6. Composite Cohort SurvivalRates and Intercycle Transition Rates, Ethiopia, 1993-94 andcirca 2000-01 ................................................................................................................. 40 Table 2.7. RepetitionRates and Share o fReadmitted Students among Repeaters, Ethiopia, 1993-93 and2002-03 ........................................................................................................... 44 Table 2.8. Efficiency o f StudentFlow inPrimaryEducation, Ethiopia, 1993-94 and 2001-02 -47 Table 2.9. Enrollments inGrades 8-1 2 inGovernment andNongovernment Schools, Ethiopia, 1999-00 to 2002-03 ............................................................................................................. Table 2.10. PossibleIndicators for MonitoringProgress inPrimary Education inEthiopia ........491 5 Table 3.1. Regional Distributiono f School-Age Population andRecurrent Public Spending on Education, Ethiopia, 1993-94 to 2001-02 ........................................................................... 54 Table 3.2. Recurrent Public Education Spending Trends and Composition by Subsector, Ethiopia, 1993-94 to 2001-02 .............................................................................................. 56 Table 3.3. Regional Recurrent Spendingon Education, andAmount and Share o f Spending on Administration, Ethiopia, 2001-02 ....................................................................................... 58 Table 3.4. Regional Recurrent Spending on Primary, Secondary, andTechnical ........................ 59 .vi. Table 3.5. Distributionof Teachers inGovernment SchoolsbyLevel of Teaching Assigned, Certification. and Average Age. Ethiopia. circa 2002.......................................................... 61 Table 3.6. Average Annual Teacher Remuneration andAggregate Spending on Teachers and Administrative Staff inGovernment Schools. Ethiopia. 2001-02 ....................................... 62 Table 3.7. Number of Academic andAdministrative Staff. Their Average Remuneration and Aggregate Wage BillinTeacher Training andHigher Education. Ethiopia. 2001-02 ........63 Table 3.8. Recurrent Public Spendingon EducationbyFunction andLevel. Ethiopia. 2001-02 65 Table 3.9.Recurrent 2001-02 ................................................................................................................................. Spending on Primaryand SecondaryEducation across Regions. Ethiopia. 67 Table 3.10.Teacher Wages andEducational Suppliesas aPercentageo fRecurrent Spending in Primary and SecondaryEducation across Regions.Ethiopia 2001-02 ............................... Table 3.11 .Public . Recurrent SpendingPer Student by Level and Type o f Education ................68 70 Table 3.12. Public Spending Per Student inGovernment Primaryand Secondary Schools across Table 3.13. Estimates o f Teaching Loads by Level of Education. Ethiopia. 2001-02 .................74 Regions. Ethiopia. 2001-02 .................................................................................................. 76 Table 3.14. Sources o fDifferences inSpendingPer Student inGovernment Primary and SecondaryEducation. Ethiopia. 2001-02 ............................................................................. 78 Table 3.15. SourcesofRegional Differences inSpendingPer Student inGrades 1-4 and Grades 5-8 Across Regions. Ethiopia. 2001-02 ............................................................................... 80 Table 3.16. Sources o fDifferences inGovernment SpendingPer Student across Cycles in Primary EducationinEachRegion. Ethiopia. 2001-02 ....................................................... 81 Table 3.17. Household Spendingon Educationby Level. Ethiopia. circa 2000 ........................... .......................... 83 Table 3.18. Aggregate Household Spendingon Education. Ethiopia. 1999-00 84 Table 3.19. National Spending on Education. Ethiopia. 1999-2000 Table 3.20. Household SpendingPer Child inPrimary School by HouseholdConsumption Table 3.21 .Public ..................................................85 87 Recurrent SpendingPerPrimary Pupil. Teacher Wages. Pupil-Teacher Ratios. and Spending on InputsOther ThanTeachers. Ethiopia and SelectedWorld Regions. circa 2000 ............................................................................................................... 88 Table 3.22. Earningso f Teachers andPrivate Sector Workers by Level of Education. Ethiopia. 1997....................................................................................................................................... 89 Table 3.23. Teacher's Weekly Teaching Hours by Level of Education. Ethiopia(2001-02) and Other Countries (1999) ......................................................................................................... 92 Table 4.1. Gross Enrollment Ratios byRegion. Ethiopia. 1993-94 and2001-02 98 Table 4.2. Gross Enrollment Ratios by Locality. Gender. andWealth. Ethiopia. 1999-2000 ...100 Table 4.3. Female Share o fPrimaryand SecondaryEnrollments across Regions. .................... 101 Table 4.4. Percentageo f ChildrenAges 7-14 Currently EnrolledinPrimary Schoolby .......................................................................... Table 4.5. StudentFlow Indicators inPrimaryEducation by Gender. ....................................... OrphanhoodStatus. Ethiopia. 1999-2000 102 103 Table 4.6. Social Selectivity inPrimaryEducation according to Three Cross-Sectional Indicators o fAccess. Ethiopiaand Sub-SaharanAfrican Countries. circa 2000................................. Table 4.7. Benefit Incidence o fPublic Spendingon Education. Ethiopia. circa 2000 ...............108 112 Table 4.8. Share of Public Spending on Education Benefitingthe Poorest andRichest Population Quintiles. Ethiopia(1999-2000) and Other Countries (1990s) .......................................... 113 Table 4.9. Distributionof ChildrenAges 7-14 and Their School Participation Rates byDistance to Nearest Primary School. Ethiopia. 2000 ......................................................................... 114 .vii. Table 4.10. Percentageo f Children Ages 7 to 14Years Currently Registeredfor School. by 118 Table 5.1. Distributionof Primary and Secondary Studentsand Schools.................................. Mother's Native Language. Ethiopia. 2000........................................................................ 121 Table 5.2. Distributionof Government andNongovernment Schools ....................................... 122 Table 5.3. PercentageDistributionof Government andNongovernment Primary Schools by Date o f Establishment. Ethiopia. 2001-02 ......................................................................... 123 Table 5.4. Distributionof Government Primary SchoolsbyHighest Grade Table 5.5. Prevalenceof Double Shifting andSection andSchool Sizes o f Government .........124 .............................. Table 5.6. Selected Indicators of Conditions inGovernment andNongovernment 128 Table 5.7. Administrative Staff Allocation and StaffingRatios across Primary ........................ ...................126 129 Table 5.8. Distributionof Teachersby Gender and Qualification inGovernment and NongGovernment Primary and Secondary Schools. Ethiopia. 2001-02 ............................ 130 Table 5.9. Pupil-teacher Ratios by Levelof Instruction inGovernment andNongovernment Schools across Regions. Ethiopia. 2001-02 ....................................................................... 132 Table 5.10. PercentageDistributiono f Government andNongovernment Schools byRange of Pupil-teacher ratios. Ethiopia. 2001-02 ............................................................................. 134 Table 5.11.RegressionEstimates of the Relation betweenNumbersofTeachers and Students across Government Schools by Level o fInstruction. Ethiopia. 2001-02 ........................... 136 Table 5.12. Regression Estimateso f the Relation betweenTotal Costs and Enrollments across Government Schools. Ethiopia. 2001-02 ........................................................................... 140 Table 5.13. Pass Rates onNationalExaminations at Grades 8. 10. and 12 byRegionand Sex. Ethiopia. circa 2000 ............................................................................................................ 143 Table 5.14. Grade 8 Examination Results for a Sample of Pupils inGovernment Primary Schools inOromiya Region. Ethiopia. 2001-02 .............................................................................. 147 Table 5.15. Correlates o f Average Grade 8 Examination Scores. Oromiya. 2001-02 ...............149 Table 5.16. Pupil-Teacher andPupil-Section Ratios andTeaching Loads Relative to Pupils' Instructional Hours inUrbanandRural Government Primary Schools. Ethiopia. 1994-95 Table 5.17. Average Weekly Teaching Loads by Teacher Certificationand Grade..................154 and 2002-03 ........................................................................................................................ 155 Table 5.18. Pupil-Teacher andPupil-Section Ratios andTeaching Loads Relative to Pupils' Instructional Hours inUrbanandRuralGovernment Primary Schools. by Region. Ethiopia. 2002-03 ............................................................................................................................... 156 Table 6.1, Trends inthe Economically Active Population. 162 Table 6.2. Labor Force Participation by School-Age Children. Ethiopia. 2001 ......................... Table 6.3. Distributiono fEmployment by Economic Sector. Ethiopia. 1994. 1999and2001 164 .163 Table 6.4. Distributiono fEmployment by Type. Ethiopia. 1994. 1999and 2001..................... 165 Table 6.5. Educational Attainment of Workers Aged 10 andAbove. Ethiopia. 1994. 1999 and ..................................................................................................................................... Table 6.6. RegressionEstimates o fthe Effect o fEducation....................................................... 2001 166 167 Table 6.7. Education's Influence on Fertilizer Use. Ethiopia. 1994 ........................................... 168 Table 6.8. MonthlyMeanEarnings by Level o f Education. Ethiopia. 2001 Table 6.9. Private Rates o fReturns to Education byLevel o f Education. Ethiopia. 2001.........169 .............................. 170 Table 6.10. Incidence of Poverty by Gender andEducation Table 6.11 .RegressionEstimates ...................................................... Table 6.12. FertilityRate andNumberof ChildrenPer Woman................................................ o fthe Impact o f Educationon Poverty. Ethiopia. 1999.......171 172 172 Table 6.13. ParentalEducation and Child Malnutrition. Ethiopia. 1995-98 .............................. 173 - V...l - l l Table 6.14. Unemployment Rateby Educational Attainment and Age, Ethiopia, 2001 ............175 Table 6.15. Social Rates ofReturns to EducationinEthiopia, 2001 .......................................... 176 List of Figures Figure 1. Historically UnprecedentedGrowth inPrimary School Enrollments since 1993-4 xvii ... Figure2. Persistent DeteriorationofPedagogicalConditions inSchools since 1994...............xviii Figure 3. Especially RapidDeterioration ofPedagogicalConditions inRuralSchools since 1994 .............................................................................................................................................. x1x Figure 4. A Conceptual Framework for EducationPolicyDevelopment and SystemManagement .............................................................................................................................................. xxi Figure 5. Recurrent Spendingon Education Has Risen, but Share o fPrimary Education Still Needs to Grow .................................................................................................................... xxv Figure 6. Cross-country Data Show aPredictable Tradeoff between Teacher Cost and the Pupil-Teacher Ratio.. ....................................................................................................... xxviii Figure 7. Allocation of Teachers across Government Primary Schools is HighlyRandomin Ethiopia............................................................................................................................. xxxv Figure8. An Example ofthe Weak LinkbetweenSpendingperPrimaryPupilandExamination Scores..................................................................................................................................... x l Figure 1.1. Adult HIV/AlDS PrevalenceRates and Orphanhoodamong PrimarySchool-Age Children, Ethiopia and SelectedEast African Countries, circa 2000 ..................................... 8 Figure 1.2. Real Gross Domestic Product Per Capita, Ethiopia, 1960-2002 ................................. 9 Figure 1.3. RelationbetweenGovernment RevenuesandCurrent Spending, Ethiopia, 1980-81to 2001-02 ................................................................................................................................. 14 Figure 1.4. RelationbetweenOverallGovernment SpendingandGovernment Spendingon Education, Ethiopia, 1980-81 to 2001-02 ............................................................................ 16 Figure 1.5. Trends inGovernment OverallRecurrent SpendingandEducationRecurrent Spending, Ethiopia, 1993-94 to 2001-02 ............................................................................. 16 Figure 1.6. Public Recurrent SpendingonEducation andPer Capita GDP, EthiopiaandOther Countries, circa 2000 ............................................................................................................ .......................................... 17 Figure 2.2. Enrollments inGrades 1-12, Ethiopia, 1967-2002 .................................................... Figure 2.1. Structure ofthe Ethiopian Education Systemin2003-04 20 22 Figure2.3. Enrollments inRegular ProgramsinHigher Education, Ethiopia, 1985-2002 .........23 Figure2.4. Share of StudentsinNongovernment Establishments, Ethiopia, 1967-68 to 2001-02 ............................................................................................................................................... 24 Figure 2.5. Share o fFee-Paying or Nongovernment-Financed Students inPrimary and SecondaryEducation, Ethiopiaand Other Countries, circa 2000 ......................................... 28 Figure2.6. Share of Privately FinancedTertiary Students, Ethiopia, 2001, andOther Countries, 1999 ....................................................................................................................................... 29 Figure2.7. Primary Gross Enrollment Ratio (GER), Ethiopia andOther Countries, circa 2000.31 Figure2.8. SecondaryGrossEnrollment Ratio (GER), Ethiopia, SelectedAfrican Countries and ................................................................................... Figure 2.9. Relation between GDP Per Capita and Coverage inHigher Education..................... Country Group Averages, circa 2000 32 33 Figure2.10. PercentageLiterate amongAdults Ages 15-49, byHighest GradeAttended, Ethiopia, 2000 ....................................................................................................................... 35 .ix. Figure 2.11 New Entrantsto Grade 1and PercentageIncreasefrom Previous Year, Ethiopia. . 1993-94 to 2000-01 ............................................................................................................. 35 Figure2.12. Pseudo-Cohort Survival RatesinPrimary and SecondaryEducation, Ethiopia, 1993-94 to 1998-99 ............................................................................................................. 38 Figure2.13. Grade-Specific CompletionRates, Ethiopia, 1995-96 andcirca 2000 .................... 42 Figure2.14. RelationbetweenGrade6 Cohort CompletionRates andPer Capita GDP, African Countries, circa 2000 ............................................................................................................ 43 Figure 2.15. Repetition RatesinPrimaryand SecondaryEducation, Ethiopia, 1992-93 to 2001- 02........................................................................................................................................... 45 Figure2.16. Repeatersas aPercentageofEnrollments inPrimaryEducation, EthiopiaandOther ....................................................................................... Figure3.1. TrendsinRecurrent Spendingon SelectedComponents ofEducation,..................... Low-Income Countries, circa 2000 46 57 Figure3.2. RelationbetweenPublic SpendingonPrimaryEducation andCompletion Rates, Ethiopiaand Other Countries, circa 2000............................................................................. 69 Figure3.3. SpendingPer Pupil inPrimary Education, EthiopiaandOther African Countries, circa 2000 .............................................................................................................................. 71 Figure3.4. EstimatedTeachers' WeeklyTeaching Hours inGovernment Primary andSecondary Schools, Ethiopia, 2001-02 .................................................................................................. 77 Figure3.5. Household Spendingas aPercentageofTotalNational SpendingonEducation, Ethiopia, 1999-2000, and Other Countries, circa 1998 ........................................................ 86 Figure 3.6. Relation betweenTeacher Remuneration andPupil-teacher ratio inPrimary Education among Low-Income Countries, circa 2000 88 Figure3.7, Where GraduatesofTeacher Training Work ............................................................. ......................................................... 90 Figure4.1. Female Share ofTotal Enrollments byLevel, Ethiopia, 1967-68 to 2001-02 ........100 Figure4.2. Cohort EntryRateto Grade 1andCompletion RatesinGrades4 and 8 across Regions, Ethiopia, circa 2000 ............................................................................................. 104 Figure4.3. Cohort EntryRateto Grade 1and Completion RatesinGrades 4 and8 across Figure4.4. Cohort EntryRatesto Grade 1inUrbanandRuralAreas ....................................... Regions, Ethiopia, circa 2000 ............................................................................................. 106 Figure 4.5. Socioeconomic ParityinPrimarySchooling inEthiopia......................................... 107 109 Figure4.6. Share ofCumulative Public SpendingonEducation, Benefitingthe 10Percent Best Educated ina Generation, EthiopiaandOther African Countries, circa 1998 ..................111 Figure5.1, PercentageDistributionof Government andNongovernment Primary and Secondary Schools byRegion, Ethiopia, 2001-02 ............................................................................... 121 Figure5.2. Average Size ofEnrollments inGovernment andNongovernment Primary and SecondarySchools by Locality andRegion, Ethiopia, 2001-02 ........................................ 127 Figure5.3. Relationbetween GenderParity inGrade6 Completion andFemale Share of Teachers inPrimary Education, Ethiopia, 2001-02, and Other African Countries, circa 2000 ............................................................................................................................................. 130 Figure5.4. Primary SchoolPupil-teacher Ratios inEthiopia, 2001-02, andOther Countries, ............................................................................................................................ Figure5.5. RelationbetweenNumbersofStudentsandTeachers across Government .............135 circa 2000 133 Figure 5.6. RandomnessinTeacher Deployment Across Schools inEthiopia, 2001-02, and Other African Countries, circa 2000 ................................................................................... 137 Figure5.7. RandomnessinTeacher Deployment across Government Primary Schools by Region, Ethiopia, 2001-02 ................................................................................................. 138 - x - Figure 5.8. Relationbetween Cost Per Student and Size of Enrollments across Government Schools. Ethiopia. 2001-02 ................................................................................................ 139 Figure 5.9. Size Distributionof Government Primary Schools andRecurrent Cost Per Pupilby School Size, Ethiopia, 2001-02 .......................................................................................... 141 Figure5.10. Size DistributionofGovernment SecondarySchools andRecurrent Cost Per Student by Size of School, Ethiopia, 2001-02 ................................................................... 142 Figure 5.11.Relationbetween SpendingPer Pupil andMeanGrade 8 Examination Score across A SampleofPrimarySchools inOromiyaRegion, Ethiopia, 2001-02 ............................. 150 Figure 5.12. Trends inPupil-Teacher andPupil-Section Ratios andinNonsalary Public Spending Per Primaryand SecondaryStudent, Ethiopia, 1990-2001 ............................... 152 Figure 5.13. Pupil-Teacher andPupil-Section Ratios inUrban ................................................ Figure5.14. Relation betweenRelative Teaching Loads inRuralandUrbanSchools ..............153 157 Box Box 1.1.Ethiopia at aGlance ......................................................................................................... 2 TechnicalNoteTables Table T2.1. Numberof StudentsandRepeaters. Girls andBoys. Ethiopia. 1992-93 to 2002-03 Table T2.2. Number ofNonrepeaters. Boys and Girls. Ethiopia. 1993-94 to 2002-03 .............184 ............................................................................................................................................. 185 Table T2.3. Pseudo-Cohort Survival Rates. Ethiopia. 1993-94 to 1999-00 .............................. 186 Table T2.4. Example Calculation ofPromotion. Repetition. andDropout Rates ...................... Table T2.5. Promotion. Repetition. Dropout Rates. Ethiopia. 1993-94 to 2001-02 ..................186 Table T2.6. Number of Nonrepeaters. Boys andGirls. Ethiopia. 1993-94 to 2002-03 .............187 191 Table T3.1. Selected Studies on the Returns to Education inEthiopia ...................................... 193 Table T3.2. Returnsto Education inEthiopia. Computed UsingRegressionCoefficients in .............................................................................. Table T3.3. Estimation ofEarnings ModelBasedon Data ........................................................ Krishnan. Selaissie. and Dercon (1998) 195 196 Table T3.4. Summary of the Rates ofReturnsto Education. Ethiopia ....................................... 197 TechnicalNoteAppendix Figures Figure T2.1. Example Calculation inthe Reconstructed-Cohort Method .................................. 188 Figure T2.2. Computing Survival Rates Usingthe UNESCO Reconstructed-Cohort Methodwith Repetition Algorithm ApplieduntilAll RepeatersinEachGrade Exitthe System ...........189 Figure T2.3, Computing Survival RatesUsingthe UNESCO Reconstructed-Cohort Methodwith ....................................................... Figure T2.4. Example Calculation inthe Composite Cohort Method........................................ Repetition Algorithm AppliedOnly Twice Per Grade 190 190 AppendixTables Table A1.1. GDP and OverallGovernment Revenues. Ethiopia. 1981-2002 ............................ ..............198 Table A1.3. Government Expenditures. Ethiopia. 1981-2002 ................................................... Table A1.2. Evolution o fRevenue-SharingFormula. Ethiopia. 1994-95 to 2001-02 199 200 .xi. Table A1.4. Current and Capital Government SpendingonEducation. Ethiopia. 1980-81 to 2001-02 ............................................................................................................................... 201 Table Al.5. Capital Spending on Education, Ethiopia, 1993-94 to 2001-02 ............................ 202 Table A2.1. Numberof Participants inNonformal EducationbyAge Group andRegion. Ethiopia. 1999-2002 ........................................................................................................... 203 Table A2.2. Percentageof PopulationEver Enrolled inGrade 1at EachAge. Ethiopia. 1996to 2000 ..................................................................................................................................... 204 Table A2.3. Cohort Completion Rates, ....................................................................................... 204 Table A3.1. Recurrent Public Spending on Education byRegion, Ethiopia, 1993-94 to 2001-02 ............................................................................................................................................. 205 Table A3.2. Current Expenditures by Function, Ethiopia, 1993-94 to 2001-02 (1986 E.C. to 1994E.C.) Table A3.3 .Recurrent ........................................................................................................................... Spendingby Subsector andRegion, Ethiopia, 1993-94 to 2001-02., ...205 206 Table A3.4. Eamings Functionfor Teachers inGrades 1-12 inGovernment Schools, Oromiya, 2002-03 ............................................................................................................................... Table A3.5, Teacher Salary Scale inPrimary and SecondaryEducation, Ethiopia Table A3.6. Datafor Decomposition o f SpendingPer Student inGovernment Primary...........210 ...................209 Table A4.1, Percentageo f the Population Ever EnrolledinGrade 1at Each Age..................... 211 212 Table A4.2. Student Flow Indicators inPrimaryEducation across Regions, Ethiopia, circa 2000 ............................................................................................................................................. 213 Table A4.3. ProbitEstimates o f Changes inthe Probabilityo fBeingRegistered for School among Children Ages 7 to 14Years UsingDatafrom the 1999/2000WMS/HICES ........214 Table A4.4. Estimates o f Changes inthe Probabilityo f Currently Attending School, Working in a "Job, " or Doing UnpaidHousework inLast SevenDays, UsingData from the 1999 Labor Force Survey ....................................................................................................................... 217 Table A4.5. Estimateso f Changesinthe Probabilitythat Child has Ever AttendedSchool Based on Data from the 2000 EthiopiaDemographic andHealth Survey .................................... 220 Table A5.1. PercentageDistributionGovernment andNongovernment Primary and Secondary Schools inUrbanandRural Areas by Instructional Program, Ethiopia2001-02 ..............223 Table A5.2. Distribution of Section Sizes byLevelof Instruction,............................................ Table A5.3. Average Section Sizes ofPrimary and SecondaryClassesinGovernment ............223 Table A5.4. RegressionEstimates o fthe RelationbetweenNumberso f Teachers andPupils 225 ..224 Table A5.5. RegressionEstimates o fthe Relation betweenNumbersof Teachers andPupils in 226 Table A5.6. Primary School Construction Costs, Ethiopia, 2002 .............................................. Government Primary Schools byRegion, Ethiopia, 2001-02 ............................................ 226 Table A5.7. Number o f CandidatesWho Took the NationalGrade 8, 10, and 12 Examinations by Region ,Ethiopia, circa 2000 .............................................................................................. 227 Table A5.8. Results from the 2000 EthiopiaNational Baseline Assessment on ........................ 227 Appendix Figures Figure A2.1, Probability o fBeing Literate among Adults with Six Years ofPrimary Schooling. ............................................................................................................................................. 229 FigureA3.1. Trends inAverage Teacher Salary inPrimaryEducation. SelectedWorld Regions. 1975-2000 ........................................................................................................................... 229 - xii - Figure A4.1. Cumulative Distribution of a Hypothetical Cohort by Education Attainment andof the Public Spending Benefiting It....................................................................................... 230 Figure A5.1. Percentage of Government andNongovernment Schools Serving Two or More Shifts of Students by Locality andRegion, Ethiopia, 2001-02 .......................................... 231 Figure A5.2. Relation betweenEnrollments and Number of Teachers inGrades 1-5 andGrades 5-8 across Government Schools, Ethiopia, 2001-02 ......................................................... 231 - XI11- ... Acknowledgments 1. This report owes its origin to an agreement in 2002 between Woiz. Genet Zewdie, Ethiopia's Minister of Education and Arvil Van Adams, then the sector manager for human development operations for Ethiopia at the World Bank, to embark on a stock-taking study o f education in the country. Ethiopia's Second Education Sector Development Program was then drawing to a close and it was felt that such a study would provide timely input for policy dialogue regarding the next phase o f the sector's development. Accordingly, the Ethiopian government and the World Bank organized ajoint effort to prepare the study. The work was led on the Ethiopian side by Setotaw Yiman, Head o f the Federal Ministry o f Education's Programming and PlanningDepartment, under Woiz. Genet's overall direction and support from At0 Derege Terefe and Dr Teshome Yizengaw, the Vice Ministers of Education; and on the World Bank side by Jee-Peng Tan, a Lead Economist in the Africa Region's Human Development Department, under the overall guidance of the following managers inthe Region: Arvil VanAdams, LauraFrigentiandIshac Diwan. 2. The Ethiopian team was structured intwo parts: a Steering Committee to provide overall direction for the study and guidance on policy development; and a Technical Working Group to collect and analyze data andcontribute to report writing. The Steering Committee was chaired by Setotaw Yiman and includedAwash Gebru, Belete Demissie, Kassaw Ali, Mebratu Birman, and Workiye Tegegn; the Technical Working group was led by Damte Demeke and consisted of Adefi-is Belachew, Almaz Beyene, Amare Bizuneh, Getachew Tadesse, Getahun Workneh, Masresha Geleta, MulugetaMekonnen andTeshome Lemma. As the work progressed, additional help came from Aleazar Tilahun as well as Yezid Reshid and his colleagues in the Education Management Information System (EMIS) panel at the Oromiya Bureau o f Education. Many people outside the Ministry o f Education also helped by sharing with the team data and documents that would otherwise have been inaccessible. They include Jutta Franz and Horst Sommer o f the German Technical Cooperation (GTZ); Bemdt Sandhaas o f Institute for Intemational Cooperation o f the German Adult Education Association; Mulat Demeke o f the Addis Ababa University; Alebachew Tiruneh; and Tesfaye Kelemework and Thomas Tilson of the United States Agency for Intemational Development. Various staff at the Ethiopian Central Statistical Authority facilitated access to the household datasetsusedinthis report. 3. The World Bank team consisted of Gerard Lassibille, Julie Schaffner, Andrew Dabalen, and Gong Hock Lee. Getahun Gebru and Gary Theisen, task team leader for education in Ethiopia, provided invaluable support and facilitated the team's interactions with Ethiopian policymakers and donor representatives inthe country; Mesfin Girma answeredour questions on the public finance data and helped closed many o f the gaps; and Mamy Rakotomalala, Lianqin Wang and Dina Abu-Ghaida prepared and documented the work on the Ethiopia Education Policy Simulation Model. The visits betweenAddis Ababa and Washington D.C. which the team members exchanged in the course of this work were arranged with skill and cheerfilness by Eleni Albejo and Southsavy V. Nakhavanit. The report was edited by Deidre Ruffino and was - xiv - processed for publication with help from Cornelia Jesse, Elsie Maka and Southsavy V. Nakhavanit . 4. The report has benefited from feedback from Jeni Klugman and Jamil Salmi, the World Bank's formal reviewers for the report. Very usefbl comments were also received from Ishac Diwan, Laura Frigenti, Getahun Gebru, Dina Abu-Ghraib, Trina Haque, GaryTheisen and other participants at a formal review o f the report at the World Bank. The report was subsequently presented at various meetings in Ethiopia, including two for donor representatives; one for regional bureau heads and some o f their staff along with officials from the federal Ministry o f Education; and one for high-level Ethiopian policy-makers and researchers. Feedback from the participants at these fora has helped to remove factual inaccuracies and improve our understanding o f the report's findings. 5. The report was financed by the Government o f Ethiopia through contributions in staff time, the World Bank and the Government o f Norway through a grant channeled via the Norwegian Education Trust Fundto support educational development inAfrica. - xv - Executive Summary Introduction 1. Ethiopia emerged from civil war in 1991. The government that replaced the communist regime introduced a new constitution in 1994 to formalize the creation of a federal structure of government, and organized the country's first multiparty elections in 1995. These momentous political reforms were accompanied by major changes in economic management policy whose impact hasbeen to help reverse more thantwo decades of persistent decline inthe real per capita gross domestic product. The country's profoundtransformation creates promising conditions for social progress. The obstacles are many, however. Ethiopia is one o f the poorest countries inthe world today; in 2004 incomes barely exceeded $100 per head. An overwhelming 80 percent of the population ekes out a living from a mostly rain-fed agriculture, and an estimated 44 percent of the country's nearly 70 million people survivebelow the poverty line. On the economic front, the country's fortunes are highly vulnerable to extemal economic shocks because of their heavy dependence on just one major export, coffee. Droughts have occurred at the rate of one every three years during the past decade, puttingat risk the lives ofmillions of Ethiopians. 2. Despite these daunting difficulties, the optimism today about the country's future i s palpable. The government has made poverty reduction the centerpiece o f its development strategy and it has continued to advance the reform o f governmental structure, functions, and finances. These contextual factors are highly pertinent for education. Ethiopia's Sustainable Development and Poverty Reduction Program (SDPRP, equivalent to what in some other countries is called a Poverty Reduction Strategy Paper [PRSP]) envisions a key role for the sector, placing on it the following expectations: progress toward universal primary education; improvement in the quality o f services throughout the system; and the production of a trained workforce that is responsive, in quantity and skills mix, to the demand for educated labor inthe country's modernizing economy. At the same time, governmental decentralization is opening the way for regional and local governments (and through them, local communities) to take greater responsibility, financial and otherwise, for managing their own affairs, including the delivery o f social services such as education. 3. For the education sector, these features inthe country's political and economic landscape prompt a host of questions: I s the current education policy framework adequate for fulfilling the expectations placed on the sector? What resources are available to achieve the stated goals? Are these sufficient to meet the requirements? Ifnot, how will the gap between resource availability and need be closed? What changes in the financing of education, inresource allocation across subsectors and schooling inputs, and in the arrangements for service delivery will help ensure that the education system develops over the long runina fiscally viable andpedagogically sound - xvi - manner? This report serves as a contribution to the continuing dialogue on these strategic questions regardingthe goals o f andmeans for education sector development.' Scope of the ReportanditsAudience 4. Ethiopia is a large and complex country and a report such as the present one must be selective infocus. The emphasis is on aspects of the education sector that are deemed especially relevant inthe PRSP context-costs, finance, and service delivery and their impact on schooling outcomes, especially among the poor; and the economic and social benefits o f investments in education. The report offers a sector-wide view on the allocation o f public spending, the structure o f costs, and the size and growth o f the system's coverage across all levels of education; explores key aspects o f service delivery in primary and secondary education, subsectors that constitute the bulk o f the system and that serve large numbers o f the poor; and takes stock of the market and nonmarket benefits of schooling. However, it does not discuss in detail technical and vocational training and education (TVET) and higher education. The exclusion o f TVET was made for a practical reason: the subsector is rapidly evolving and the data required for a proper analysis could not be assembled to add value to what was already known. Higher education was excluded because a parallel study on the subsector was inprocess while this study was being prepared. Because of time and data constraints, the study does not cover nonformal education and early childhood education, nor does it address (except when relevant to the topics covered) such traditional educational issues as curriculum content, teacher training, pedagogicalmethods, language o f instruction, and so on. 5. This report provides a snapshot o f the education sector up to 2001-02 (and, on some dimensions, up to 2002-03). Its intent i s to portray the selected aspects as accurately as possible so as to discover potentially important areas for policy development. Therefore, the report is deliberately diagnostic in orientation. This approach is important because the process o f policy development andimplementation must begin with a broadly shared understandingo f the goals o f sector development, the constraints on the sector, and the tradeoffs that might be required to overcome them. The report is thus aimed at a wide audience. In the first instance, it i s directed toward Ethiopia's policymakers andmanagers inthe education sector at all levels o f government and toward education practitioners and researchers. It seeks to inform decision makers and analysts in other parts of the government as well, especially those with direct responsibility for designing the country's overall development strategy and for aligning public spending with that strategy. As a knowledge asset the study should be o f interest to the government's development partners in education. This audience includes donors who are active in supporting educational development in Ethiopia; nongovemental organizations that are active in providing various education services; teachers and other school personnel, in both the government and nongovernment sectors, who constitute the front-line agents for service delivery; parents and their children, the intended beneficiaries of education services; and the public at large for whom a strong education system is important for the country's overall economic and social well-being. All these partners play an indispensable role, and the report can help provide a factual basis to The report thus complements two recentlycompletedWorld Bad-sponsored studies-the public expenditure review on public spending in the social sectors and the study on higher education-as well as the many studies on education sponsored by the government and its other development partners. - xvii - enrich discussions on the government's vision for sector development and its action plan to put that vision into effect. 6. Below we summarize the report's main findings. To economize on space, the presentation focuses on the country-wide highlights. Readers are invited to read the rest o f the report for regionalperspectives on the key sector development issues. Remarkable Progress since 1993-94 7. Evencasual observers of Ethiopia's education sector will marvelat the dramatic growth in enrollments throughout the system in recent years. Under the new government, aggregate enrollments ingrades 1-12 rose at a steady pace o f about 9 percent a year between 1992-93 and 2001-02; and in grades 1-4, the first cycle o f primary schooling, they grew even faster: 15 percent a year. These trends are a remarkable achievement given the patterns of stagnation, reversals, anduneven growth inthe past (figure 1).By 2002-03, the education system had grown to about 8.6 million studentsinprimary schools, andmore than 0.6 million insecondary schools. Enrollments in technical and vocational training and education grew from fewer than 3,000 students in 1995-96 to an estimated 54,000 students in 2001-02. Inpostsecondary education, enrollments have also ballooned, from around 18,000 in 1990-91 to more than 48,000 in2001- 02. Figure 1. Historically Unprecedented Growth inPrimary School Enrollments since 19934 7.0 n3 6.0 .C( 0 =I .-8 5.0 W 4 sc rcl 2 4.0 0 3.0 jz !I3 2.0 1.o A Note: Dataincludeonly students inregularprograms in government andnon-governmentschools. Source: See figure 2.2. 8. The growth inenrollments shows up ina sustainedincreaseinthe gross enrollment ratio, a common indicator for measuring coverage, at all levels inthe system. Inprimary education, it more thantripledbetween 1993-94 and 2001-02, from 20 to 62 percent. The ratio for secondary education rose from 8 to 12 percent inthe same period, while that for higher education climbed from 0.5 percent to 1.7 percent. Despite these dramatic improvements, Ethiopia's gross - xviii - enrollment ratios are still somewhat smaller than the corresponding averages for Sub-Saharan Africa. EmergingSigns ofDistressinthe System 9. While undeniably impressive, the system's expansion has not been problem-free. The clearest distress signals are the deteriorating conditions in`classrooms throughout most o f the country. Since 1993-94, the pupil-teacher ratio, the pupil-section ratio (Le., the number o f pupilsper section), andthe realspendingper student on nonsalary inputshave steadily worsened (figure 2). In 2001-02, Ethiopia's pupil-teacher ratios of 65:l in government primary schools and 52:l in government secondary schools are among the highest in the world. Inevitably, section sizes at both levels have also risen to extremely high levels, averaging about 75 and 82 students per section, respectively. Figure 2. Persistent Deterioration of Pedagogical Conditions inSchoolssince 1994 90 9 80 8 h I m.E a. in 6 2 5 3 40 4 P 30 3 k 20 2 I `PTR grades 9-12 \PTR, grades 1-8 I 10 1 0 1 I o 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 PTRPupil-teacher ratio; PSRPupil-section ratio. Source: See figure 5.12. EspeciallyDesperateConditionsinRuralPrimary Schools 10. As the system has expanded, conditions have worsened particularly rapidly in rural schools (figure 3). The trends are such that by 2002-03 Ethiopia's school system found itself in the anomalous situation o fhavingpupil-teacher ratios inrural schools that exceedthose inurban schools, by an astonishing average o f 60 percent in grades 1-4 and 50 percent in grades 5-8. Almost everywhere else in the world pupil-teacher ratios are typically smaller in rural areas because of their lower population densities. By contrast, classrooms in rural Ethiopia are as crowded as they are in urban areas. The combination o f high pupil-teacher ratios and section sizes translates into much heavier teaching loads for teachers assigned to rural than to urban schools. InOromiya, for example, the weekly load inrural schools averages 29 hours ingrades 1-4 and 24 hours ingrades 5-8, compared with the corresponding averages o f 22 and 18 hours, - xix - respectively, in urban schools. These disparities imply that, in rural schools, there is almost no scope for specialized teaching; they also make rural teaching jobs even more unattractive than they already are, thus increasing the difficulty of rationalizing teacher deployment across schools. Figure 3. Especially RapidDeterioration of Pedagogical Conditions inRural Schools since 1994 80 UrbanPSR 70 60 50 40 30 20 PTR Pupil-teacher ratio; PSR Pupil-section ratio. Note: Because data are available separately for government and nongovernment schools only after 1997-98, the figure uses data for both sectors in all years. Data after 1997-98 indicate that the difference is small between the govemment-only series and that for both sectors. Source: Based on data supplied by the Ministry of Education on number of pupils, teachers, and sections. A Frameworkfor Consideringthe ChallengesAhead 11. Ethiopia is at a crossroads today. The main question i s this: I s the model that framed the system's expansion since 1994 still valid to guide its future development? Ifthe answer i s inthe negative, as the data increasingly appear to suggest, how then should we adapt the model to the conditions in Ethiopia today and in the years to come? Beyond this, we also need to address some searching questions on ways to improve management o f the system, so that all available resources maketheir fullest contribution to the sector's development. 12. These are large and complex issues with no easy answers. One approach is to begin with a fiamework to conceptualize the way in which policies and actions work together to produce results on the ground. At risk of oversimplification, figure 4 below shows three loci where the behavior of and decisions byvarious actors inthe education sector make an important difference. The top layer corresponds to decision making by the federal and subnational governments, the impact o f which permeates the system as a whole. It i s the place where government sets overall sector goals and mobilizes the resources to reach them; where it defines standards for service delivery, andthe roles of government andnongovernment providers; and where it makes explicit - xx - and agrees upon indicators for tracking overall progress as the basis for system-wide accountability. 13. The middleblock inthe figure reflects the ideathat the impact o fthe broad parameters o f educational policy is mediated, inthe first instance, through activities to bringresources down to the level o f schools, communities, and households. For the government, the figure lists three types o f activities that are particularly pertinent, all having to do with good system administration: ensure that schools receive the resources, human and financial, required to deliver services; ensure that constraints on classroom facilities and the lack o f schools are consistently and routinely addressed; and ensure that vulnerable population groups receive the assistance they need to participate in schooling. The impact o f these activities can be enhanced by creating appropriate information flows and providing support and incentives for problem solving. 14. Finally, the figure shows a third locus-schools and classrooms-where policies are acted on to produce results on the ground. The actors here are school directors and teachers who must work with the resources at hand, material and otherwise, to manage the schooling careers o f their charges and their acquisition o f learning while at school. Once again, appropriate information flows and design o f support and incentive structures can help strengthen performance. 15. The analysis in this report yields findings that are relevant for discussing the potential scope for improvement at all three levels o f decision making described in the figure. They are summarized below using the figure's general structure. To move from diagnostic findings to a plano f action will obviously require considerationo f a practicalcontext. Inthis regard, the rapid move toward greater decentralization inEthiopia is particularly pertinent, and some lessons from international experience on this score are distilled below to enrich the discussion o f the implications for Ethiopia. - xxi - Figure4. A ConceptualFrameworkfor EducationPolicyDevelopmentand SystemManagement Designing the Overall Policy Framework Policy management of: Educationalgoals inthe PRSPcontext Public resourcesfor education Public-privatesector rolesineducation Standards for servicedelivery Role of indicatorsfor monitoringprogressandaccountability Getting Resources to Schools, Communities, and Households Administrative management of: Deploymentofhumanandfinancialresourcesacross schools Placementof improvedclassroomfacilities andnew schools across communities Targetingof assistance to vulnerable groups to reduce socialdisparities Role of information, support, andincentives for problemsolving andimpactevaluation) Transforming Schools' Resources into EducationalOutcomes Pedagogicalmanagement of: Children's schoolingcareers Leamingachievement Role of information, support, andincentivesfor effectiveteaching and leaming Source: Authors' construction. Designing the Overall Policy Framework 16. Ina strategic policy framework, the key options for sector development and their fiscal implications are considered side-by-side rather than in isolation from each other. This approach i s important because in a world o f limited resources, tradeoffs are inevitable and will occur in one of two ways: through proactive choice or by default. When the tradeoffs occur by default, implicit choices are made that are likely to produce unforeseen and potentially adverse consequences. In the context o f decentralization and increased social participation in the - xxii - decision-making process, makingthe tradeoffs inan explicit and strategic fashion i s probably the best way to balance the interests of competing groups inorder to chart a course that benefits the country as a whole. The report's main findings on areas where harddecisions warrant attention are summarized below. Rethink the Goalsfor Coverage and Differentiate Themfor Urban and Rural Areas 17. As Ethiopia's SDPRP states, the education sector is expected to help reduce poverty by universalizing primary education andby producing a workforce capable of fillingjobs requiring skilled labor. What do these goals mean in practice? Does universal primary education mean getting all children to complete eight years of primary schooling? Does producing skilled workers mean turning out as many graduates as the education system can put through the upper levels of the system? 18. Universalize four years of schooling as an immediate and Dractical Driority. The 2004 World Bank Poverty Assessment for Ethiopia calculates that if all adults in the country had at least four years o fprimaryeducation, the share of households living inpoverty would drop by 18 percent. This sizable impact is consistent with the widespread agreement among policymakers and analysts that the core skills that primary education imparts-literacy and numeracy-can boost a child's life chances, even if he or she eventually earns a livelihood in agriculture. Because five years o f primary schooling are generally considered a minimumfor one to become permanently literate and numerate, a large number o f countries made the commitment in the UnitedNations 2000 Declaration on the MillenniumDevelopment Goals (MDGs) to ensure that all children would be able to complete at least five years o fprimary schooling. 19. Ethiopia's education system has a 4-4-2-2 structure which at first sight makes it somewhat awkwardto state a clear-cut goal for coverage: universalizing eight years o f schooling i s probably fiscally unviable even in the medium term, while universalizing only four years seems too modest in light o f international experience. Inpractice, the issue is simple to resolve because coverage is still modest at present: only 60 percent in each age cohort ever enroll in grade 1, and barely 60 percent reach grade 4 (table 1). Under these circumstances, implementation constraints alone would make universalizing eight years of schooling for the whole populationan impracticalgoal inthe short- or even medium-term. 20. AdaDt the goals for coverage to conditions in urban and rural areas. Because initial conditions differ so widely across locality, setting different time-bound targets for urban and rural areas i s a practical necessity. In urban areas the goal o f universalizing eight years o f primary schooling is well within reach (and is also consistent with the stronger demand for educatedlabor insuch areas)-most children already enter grade 1andthe majority who survive past grade 2 eventually reach grade 8. The challenge here i s to minimize the dropout rate between the first two grades where the problem i s currently concentrated. In rural areas, by contrast, less than half the children in each age cohort enter grade 1, and o f those who do, only 55 percent survive to grade 4 and only 19 percent, to grade 8. These patterns imply that universalizing even four years o fprimary education inrural areas will be a demanding goal, and that expansion beyondthis minimumwould need to be made instages. - xxiii - 2001-02 Indicator 1993-94 Overall Urban Rural Cohort entry rate to grade 1ai 0.37 0.61 0.92 0.46 Composite cohort survival rates from grade 1 To grade 4 0.52 0.59 0.77 0.55 To grade 8 0.39 0.36 0.80 0.20 To grade 12 0.15 0.08 - - Intercycle cohort transition rates Grades 4 & 5 b' 0.94 0.88 1.07 0.79 Grades 8 & 9 0.84 0.91 - - Grades 10 & 11 0.86 0.33 - - -Datanotavailable. 21. Allow labor market conditions to guide the pace of expansion beyond wade 4. As the economy matures, the demand for educated labor typically expands, making it increasingly appropriate and feasible to universalize eight or even more years of schooling. At issue i s how fast to expand the education system. If the system grows too fast and produces more graduates qualified beyond the core skills o f basic literacy and numeracy than can be absorbed into the prevailing job market, the investment would be wasted to the extent that graduates fail to find work appropriate to their level o f training and so lose the expected return on their investment. With enough frustrated job seekers on the streets, the situation could easily escalate into serious social unrest. In part to minimize this risk and to institute a merit-based system for allocating scarce places in postprimary education, most governments in low-income countries use standardized examinations to regulate the flow o f students through the system. 22. InEthiopiathe demandfor educatedlaborhasbeengrowing, butonlyslowly. Nearly80 percent o f country's workforce i s still in agriculture today, and more than 90 percent o f the workers are either self-employed or usedas unpaid family labor-features that put a limit on the economy's capacity to absorb large numbers o f educated labor into modem sector jobs. Data from the most recent labor force in 1999 suggest, for example, that among recent graduates, the unemployment rate was about 25 percent among those attaining grades 5-8 or vocationaVtechnica1 education and 44 percent among those attaining general secondary education. Only among university graduates was the unemployment rate at a modest 4 percent, and even this favorable situation may be short-lived unless jobs outside public administration (which have increased greatly in recent years because o f decentralization) become the major source o f employment in the future. These pattems underline the importance o f aligning the expansion paceofpostbasic education to labor market conditions. 23. Improvethe education system's responsiveness to labor market signals. Fostering a closer linkbetweenthe education system and the labor marketwill require not only regular monitoring - xxiv - o f the labor market to generate the relevant data, but also fairly flexible and responsive arrangementsfor service delivery. With regard to the latter, an important option is to enlarge the role o f private service providers at postprimary levels. Because private providers serve clients who pay for most, ifnot all, of the cost of services and who are thus likely to insist on getting a reasonable return on their investment, such providers tend to respond more nimbly to labor market signals in designing their course offerings. This asset is especially important when the objective is to prepare graduates for employment rather than for further studies. 24. Beyond encouraging private providers, the government can also tighten the management of student flow through its own institutions, not least to counteract the tendency toward overexpansion in a highly subsidized sector. In recent years, the pressure on preparatory and higher education has risen sufficiently for the government to limit the number ofplaces ingrade 11to 50,000 in2002-03, compared with some 175,000 in2000-01, and to divert the remaining grade 10 completers to programs inteacher training, and technical and vocational education and training. A selection mechanism also exists ingrade 8 where students take standardizedregional examinations. However, as table 1shows, today there is practically no selection at this point, as nearly 91 percent of the students eventually continue on to grade 9. This fact, coupled with the absence o f a formal selection mechanism earlier than grade 8, may not matter much at present becausenearly two-thirds o f eachcohort of first-graders eventually drop out much earlier. But as basic education expands and survival rates improve, as they should, the need for a selection mechanism earlier than grade 8 is likely to present itself. Prioritize Education, Especially Primary Education, in theAllocation of Public Spending 25. Duringmost of the 1980s and 1 9 9 0 ~ ~ education was not systematically prioritized inthe allocation of public spending. Since 2000, however, Ethiopia's government has been spending more on education, and aggregate public recurrent spending on education has now beenrestored to the 3 percent of GDP that prevailed in 1993-94 (figure 4). Most o f the increasehas gone into administrative overheads and postsecondary levels of education, however, leading to a significant shift in the allocation o f spending. Although this trend has not yet distorted the functional distribution o f spending, the share o f primary education-50 percent of the total in 2001-02-remains much less than the international benchmark o f 67 percent under the Education For All Fast Track Initiative. Thisbenchmark is itselfbasedon the pattern of spending in countries that have made good progress toward universalizing the completion of primary school (Bruns,Mingat, and Rakotomalala 2003).2 26. Continue to allocate,moreto education while ensuringwise use of the resources. Looking to the future, we see that the increased attention to education, both domestically and internationally, imply that the prospects of more resources for education are good. Yet the magnitude of the increase will inevitably be constrained by the competition among sectors for the government's ultimately limited resources. Among low-income countries, very few have managed to sustain levels o f recurrent spending on education beyond 4 or 5 percent o f their GDP. Thus, if Ethiopia faces the same intemal pressures as the governments in these other countries in managing revenue generation and resource allocation across sectors, the amount o f * In most countries, the primarycycle lasts six years. The benchmark under the EducationFor All Fast Track (EFA-FTI) is 50 percent for the subsector.For an eight-yearcycle, this benchmark is increasedproportionatelyto 67 percent, assumingunit costs are constant for the extra two years of schooling. t tXt - xxvi - Set Sustainable National Standardsfor TeacherRecruitment and School Construction 28. Givingmore resources to primary education does not remove the need to make tradeoffs within the subsector itself. Indeed, if all children are to receive at least a primary education, questions arise about the options for scaling up. The distress signals already present inthe system suggest that a simple replication o f the current model o f service delivery is unlikely to work. Indeed, expanding the system in a fiscally viable manner without compromising its ability to function well would probably require significant change in two key dimensions o f service delivery-standards for teacher recruitment and school construction. A related issue pertains to a potential need to simplify the curriculum so as to reduce the use o f specialized teaching and facilities, particularly ingrades 5-8. Inthe context o f decentralization, it is important to consider giving regional governments greater flexibility to adjust national standards in order to achieve their educational goals within the budgetconstraints they face. 29. Choose fiscally sustainable standards that benefit the system as a whole. Currently, the recruitment standards are as follows: teachers in grades 1-4 should have graduated from a teacher training institute (TTI; Le., they should have had ten years o f general schooling plus one year o f teacher training); teachers in grades 5-8 should have graduated from a teacher training college (TTC; i.e., they should have had ten years o f general schooling plus three years o f training); and teachers in grades 9-12 should have received a university degree. Ingrades 1-4, almost all the teachers in government schools have the required certification, but in grades 5-8, only 26 percent do; in grades 9-12, the share is less than 18 percent. The remuneration of teachers averages about 6.8 times the per capita GDP for those teaching ingrades 1-4, 8.0 times for those in grades 5-8, and 11.8 times for those in grades 9-12. With regard to construction standards, they are suchthat the cost can range up to $19,000 per primary school classroom. 30. Standards are obviously important: they help create a network o f schools with common characteristics and are intended to enable schools to function well. But if standards are set beyond what the country can afford (or mobilize through donor assistance), schools may meet the requirements in the areas explicitly specified, but only by cutting back in other areas; that eventually undermines their ability to provide good services. The result is a tradeoff by default. The fact that a poor tradeoff has occurred is reflected inthe continual deterioration in schooling conditions since 1993-94 (shown earlier infigure 2). Other telling signs are: (a) ubiquity o f large andovercrowded schools (more than 20 percent o f the government primary schools enroll more than 900 pupils, and about 25 percent o f government secondary schools enroll more than 2,500 students); and (b) the pervasive shortages o f certified teachers, especially beyond the first cycle o f primary education. In regions such as Amhara and Oromiya, the teacher shortages have in recent years led the authorities to begin using or allowing the use o f paraprofessional teachers (ie., graduates o f 10thor 12thgrade without TTI or TTC certification) to fill the gaps. 31. More broadly, one implication o f overly ambitious standards for teacher certification and construction is that fewer schools can be built and staffed for a given budget than would otherwise be the case. This tradeoff means that many villages in Ethiopia do without a government primary school and many children must travel long distances to get to a school. In some villages, the solution is offered by alternative basic education centers, which provide a simplified and shorter instructional program. These are set up with the help o f nongovernmental organizations using locally recruited teachers (who have no formal certification and are paid - xxvii - much less than certified teachers) and facilities that cost substantially less than government schools to build. The use o f paraprofessional teachers in government schools and the emergence of alternative basic education centers are creative solutions to very difficult circumstances and they make an obvious, if localized, contribution. It is important to note, however, that they cannot substitute for the reform of national recruitment and construction standards that are neededto benefitthe entire systemofgovernment schools. 32. Use recruitment standards that leave room for immoving other aspects of classroom conditions. Recruitment standards that are set without adequate attention to their financial implications can produce unintended consequences. InEthiopia, the use o f costly standards for teacher recruitment has reduced the resources available for other school inputs that are also essential for effective teaching and learning. A comparison o f the pattern o f resource allocation across school inputs in Ethiopia and other countries illustrates this point (table 2). Expressed relative to the per capita GDP, the amount that Ethiopia spends per student in grades 1-8 is comparable to what the average low-income country spends. But the composition of its spending differs strikingly from that in other countries. In the first cycle, Ethiopia's recurrent public spendingper pupil is comparable to that inAsian and Eastern European countries. But Ethiopia achieves the result by combining a much higher cost o f teachers with a substantially less favorable pupil-teacher ratio and a significantly smaller share o f spending on inputs other than teachers. In grades 5-8, Ethiopia's per pupil spending is comparable to the average for Latin American countries. Yet here again, the composition of spending reveals the same tradeoffs in which the higher cost o f teachers is absorbed through a less favorable pupil-teacher ratio and fewer resources for inputs other than teachers. This is true even when compared with other African countries, where teacher remuneration as a multiple o f per capita GDP typically exceeds that in other low-income countries-the composition of spending in Ethiopia reflects a substantial tradeoff against student-teacher ratios and allocations for inputs other than teachers, infavor ofhigherteacher costs. This fairly inevitable tradeoffbetweenteacher remuneration and pupil-teacher ratios is suggestedby the pattern in figure 5, which i s based on the experiences o f a large cross-section o f countries around 2000. With regard to Ethiopia's low level of spending on inputs other than teachers, the impact materializes in the form of a scarcity of pedagogical materials, including textbooks, throughout the system, as well as limited provision for teacher supervision and other support services for effective teaching. - xxviii - Table 2. HighTeacher Cost Leadsto Adverse Tradeoff 1 againstOther SchoolInputsinEthiopia I Public 11 I Pupil- Percentageof spending er Average teacher ratio recurrent spending Zountrylregion pupiJ teacher wagea/ " on inputsother than ~ Ethiopia,2001-02 Grades 1 4 0.10 6.8 II 74.3 7.8 Grades 5-8 0.18 8.0 48.9 10.3 Xegional averages,circa 2000 Francophone Africa (20) 0.13 4.4 51.0 26.8 Anglophone Africa (13) 0.13 4.3 41.6 20.2 South & East Asia (10) 0.09 2.4 36.5 23.4 I Latin America (4) 0.16 4.3 33.6 18.5 Eastern Europe & Central Asia (5) I 0.11 1.2 18.2 31.3 Figure6. Cross-CountryData Show aPredictableTradeoff betweenTeacher Cost and the Pupil-Teacher Ratio 12 10 Ethiopia \ 8 6 4 2 0 I , 0 10 20 30 40 50 60 70 80 90 Pupil-teacher ratio Note: Data for Ethiopiarefer to grades 1-4. Source: See figure 3.6. Diversijj the Optionsfor Managing the Cost of Services 33. Broadly, the options fall into two categories: those that involve the use o f alternative service delivery arrangements, and those that involve lower cost inputs or increased efficiency within the regular government sector. An important approach would be to make the menu o f sensible options as large as possible, and empower and motivate involved decision makers to choose wisely amongthem. - xxix - 34. Support alternative basic education centers as an option for service delivery. InEthiopia such centers provide low-cost schooling covering the first four years of primary schooling. By compressing the program into three years, the services are particularlyresponsive to the needs of the large numbers of overage children currently seekingprimaryschooling. Although the centers have achieved reasonably good outcomes in terms of student learning, the majority operates without government funding as of this writing. As a result, they serve a limitedclientele totaling perhaps 500,000 pupils in 2001-02, or about 8 percent of total enrollments in grades 1-4. If alternative basic education is to play a central role in universalizing primary schooling, the government will need to finance it and make the centers an integral part o f the whole system rather than leave them as ad hoc experiments. Inthe process, it will be important to retain the features that make alternative basic education highly relevant in the first place, particularly in rural contexts: a flexible school day and academic calendar, use o f teachers from the community, simpleclassroom facilities, proximity to children's homes, strong in-service teacher support, and reliance on communityparticipation. 35. Accept flexible standards in order to lower costs in the formal school system. Because spending on teachers makes up the bulk of education costs, any strategy for cost management must consider ways to keep this cost item under control. A straightforward option is to reduce the pay structure, but inpractice it is politically difficult and also largely irrelevant unless labor markets for the educated function poorly and teachers are somehow overpaid. If this were the case, one should find low attrition rates among teachers and hightake-up rates o f teaching jobs. Attrition is not a problem among TTI-certified teachers but it is serious among those with TTC diplomas and university degrees (i-e., those expected to fill teaching jobs in grades 5-8 and grades 9-12). If anything, the indications are that teachers with such qualifications are leaving the profession for more attractive nonteaching jobs; and among fresh graduates trained to be teachers, significant numbers do not even turn up at their posts, particularly inthe more remote regions. 36. A secondandmorepromising optionis to accept greater flexibility inteacher recruitment standards, for example, by allowing new teachers to be recruited at lower levels of formal certification. This does not mean filling teaching posts with the clearly unqualified candidates, nor does it mean that once recruited, the teachersremainstuck intheir careers. Rather, the idea is to lower the floor for entry while encouraging teachersto upgrade their qualification over time as part o f a long-term career plan. As indicated above, recruitment o f paraprofessional teachers is already happening to some extent in some regions. Such teachers may have completed lothor 12th grade without certification from a teacher training institute or college, and they may typically receive a pay that i s perhaps only half as high as that o f a certified teacher. Even so, recent processing o f recruitment in Oromiya, for example, suggests that in some localities there may be as many as 20 qualified applicants for each available position. Recruiting at lower levels of certification i s admittedly not ideal because of possible adverse effects on the quality o f services. Yet it may be the only way to manage tradeoffs within a limited budget that will achieve a balanced andpedagogically effective combination o f school inputs, one characterized notjust by acceptable levels o fteacher certification, but also by reasonable student-teacher ratios andadequateavailability ofbooks and other pedagogical materials. 37. Reduce subiect specialization to make better use o f teacher time. Although the present report did not concern itselfwith the details o f the curriculum content, the impact o f specialized - xxx - teaching after grade 4, possibly exacerbated by constraints on classroom facilities, can be deduced fiom the pattern o f teachers' use o f time: whereas weekly teaching loads in grades 1-4 average 24 hours, they fall to 17 hours for teachers in grades 5-8, and 15 hours, for those in grades 9-12. Comparison with the practice in other countries suggests that there is probably some scope for improving the management o f teacher time in Ethiopia (table 3). Exploring this potential is important because differences inteaching loads make a big difference to the cost o f service delivery. Our analysis suggests that, all else being the same, the impact o f the difference inteaching loadsingrades 1-4 andgrades 5-8 is to raise unit spending inthe secondcycleby an estimated 40 percent. This simulation implies that unless teachers ingrades 5-8 are used as fully as they are in grades 1-4, the fiscal requirements o f a large-scale expansion o f the system after grade 4 will probably be hardto absorb. Teachers' weekly teaching hours Countrylcountry group Lower Upper Primarya' secondaryb' secondary Ethiopia 24.0 17.2 15.0 OECDMean(27 countries) 21.2 19.4 - Minimum 15.2 15.0 - Maximum' 31.6 31.3 - WE1Mean(17 countries) 21.6 20.0 19.1 Minimum 17.6 12.9 12.9 Maximum 29.4 29.4 24.5 38. Include lower cost models among the options for classroom construction. The Ministry o f Education recently commissioned two studies to look into the options for more affordable designs. One study found that construction costs could be reduced to between $4,400 to $5,900 per classroom depending on choice o f materials and design. Costs are naturally higher in the more remote areas, but the second study found that even after taking this factor into account, existing construction costs could be reducedby about a third. For comparison, the calculations in the Education For All Fast Track Initiative use a rate o f $8,000 per classroom, a benchmark based on cross-country experience in Africa. The scope for reducing construction costs in Ethiopia thus appears to be substantial. While a $6,000 classroom may not be as good as a $20,000 classroom, choosing the latter standard effectively means that for a given budget less thana third as manynew classrooms wouldbe built. - xxxi - Focus on Key Indicators to Monitor Progress 39. Indesigning the overall policy framework, it is not enough simply to define goals and make tradeoffs; it is also critical to track progress toward the intended goals. This activity also generates the feedback that is so critical for adjusting policies as they are implemented. As mentioned above, three goals in education are pertinent in the PRSP context: universal primary education, improvement inthe quality of services, andproductionof a skilled workforce. Below we summarize the report's findings regarding monitoring o f the first goal. With regard to the second goal, we note in passing that in 2000 the government completed a national baseline assessment of student achievement in grades 4 and 8 and hasjust repeatedthe exercise in2004. Regular follow-up o f this nature, including possible participation in international student assessments inthe future, is a good way to track progress on the second goal. With regardto the production o f a skilled workforce, it is particularly important to track the transition from school to work among graduates of those courses o f study that are especially expensive for the government to provide (e.g., TVET and higher education). While a systematic and routine tracking of this transition does not appear to be in place at present, the gap can be bridged by exploring collaboration with the Central Statistical Authority to generate the required data through its regular surveys (e.g., by adapting and adding appropriate questions in the survey instruments). 40. Keep the set o f indicators for primary education small and relevant. In the past, most countries have used cross-sectional measures such as the gross and net enrollment ratios to measure progress, in part because these indicators are easy to compute. While appropriate for some aspects o f education planning, they are deeply flawed for tracking the country's progress toward ensuring that all children have access to a full cycle o f primary schooling. This i s particularly true inthe context o f Ethiopiawhere the systemi s still catering to a large backlog o f overage children. Increasingly, countries are instead using the primary school completion rate (typically at grade 5 or 6) as the relevant measure of success. Because this indicator reflects underlyingpatterns of access and continuation in school, two other measures, the entry rate to grade 1 and survival rates to subsequent grades, are also included as part of the package. Furthermore, repetition rates are typically reported as well, for two reasons: First, an education system characterized by high rates o f repetition costs more to operate. And second, high repetition rates point to possible problems in managing student flow, which, if left unattended, are likely to make it harder to improve completion rates. 41. Measure the indicators correctlv. Assuming the foregoing indicators are to be tracked, it i s important to agree on the methods and sources o f data for making the estimates. Tradeoffs are inevitable but important considerations include the following: (a) use o f cohort rather than cross- sectional measures (for the reasons elaborated in the report); (b) reliance on data generated by existing mechanismsto avoidoverburdening the available capacity for data collection ;(c) use o f simple calculations that keep tedious data manipulation to a minimum. Table 4 shows the indicators that are likely to meet these criteria. Cohort entry rates require survey data that include the educational status of children. The fact that the data pertain to a sample rather than to the entire population i s admittedly a flaw, but it cannot be avoided because o f the nature o f the desired statistic. The highfrequency o f household surveys inEthiopia should make it possible to cross-check the estimates, all o f which can be generated at relatively low cost by consistently adding to the survey questionnaires a question for eachhousehold member aged 5 and above on - xxxii - whether or not they have ever been enrolled in school (and if so, in what type of school). The other indicators can be estimated using data generated from the Ministry o f Education's annual school census. While accuracy in the census data, particularly with regard to repetition status, continues to be an issue, these data pertain to the whole population of students; they can be further cross-checkedby including, as is already done in some household surveys in Ethiopia, two additional questions in the survey instruments: (a) Are you currently enrolled, and if so, in what grade? and (b) Were you enrolled last school year, andifso, inwhat grade? Indicator Definition/comment Source of data Cohort entry rate to grade 1 Percentage o fage cohort ever enrolled Nationalhousehold survey I I I I `Oh'* Annual school censuses for data on rate, grades 1-5 Percentage o f entrants who reach grade 5 enrollments and repeaters by grade Composite cohort completion Product ofthe cohort entry and composite Computed rate, grade 5 cohort survival rate to grade 5 Repeaters (definedto include "readmitted" students) ingrades 1-5 in Annual school censuses; verified by Repetition rate' grades year Y dividedby enrollments ingrades surveys as available 1-5 inyear Y-1. Source: Table 2.10, based onauthors' construction; see text for further elaboration of the indicators. Getting Resources to Schools, Communities, and Households 42. This activity corresponds to the second block of management responsibility in the conceptual framework presented earlier. It canbe thought o f as a packageo f interventions whose aim i s to improve services to the groups that currently lag behindin school participation. This is important because the country's overall progress in education will depend on how fast these groups are brought into the system. Pay Special Attention to the Needs of Rural Populations 43. Table 5 summarizes the differences in primary school participation across three commonly defined population groups: girls and boys, children in urban and rural areas, and children from rich and poor households. The indices shown in the table are a composite of various indicators o f school participationexplained inthe report. Among the three sets o f groups compared, the gap is widest between children in urban and rural areas: the index for rural children is, on average, only 31 percent that o f their urban counterparts. More detailed analysis reveals that whereas 90 percent o f urban children ever enroll in grade 1, the share i s only 45 percent among rural children. Once in school, rural children survive to grade 4 at about 72 percent the rate o f urban children, but their survival rate to grade 8 i s only 25 percent as high. Overall, these results suggest that improving primary schooling outcomes in Ethiopia and - xxxiii - reducing socioeconomic disparitiesinthe higher levels will require a concerted effort to address the constraints in rural areas. This does not mean that other population groups should be neglected, but it does meanthat attention to the needs ofrural localities will beparamount. Index of parity in primary school participation Average across all Girls Rural areas Poorest quintile three sets of Countrylregion (Boys = 1.00) (Urban = 1.0) (Richest = 1.0) comparisons Ethiopia 0.77 0.3 1 0.69 0.59 Sub-SaharanAfrica 0.83 0.63 0.50 0.65 44. Managers in the education system can influence progress toward the country's goals in basic education through their management of the supply- and demand-side factors that affect schooling behavior. Below we summarize the report's findings inthis regard. Note that although we discuss supply- and demand-side issues separately to facilitate the presentation, the distinctionis not strict inpractice because of interactive effects. Act on the Supply of Services 45. The report identifies three aspects of the supply o f services that warrant attention: (a) the availability and accessibility o f schools; consistency in teacher deployment across schools; and (c) completeness of the instructional program offered in schools. With regard to teacher deployment, it is important to note that inEthiopia, very little is spent on nonteacher inputs, and the bulk o f public spending on education reaches schools in the form o f teachers. Focussing on teacher deployment i s therefore a good way to ensure that resources are equitably allocated at the facility level. 46. Improve the availability and accessibility of schools in rural areas. According to published tabulations from the 2001 Child Labor Force Survey, about one in seven Ethiopian youths (excluding those who were too young to enroll) cited the lack o f a school as the reason why they were not enrolled. The problem o f distance, especially among rural children, is illustrated in table 6, based on data from the 2000 Welfare Monitoring Survey. Fully a third o f the rural children in the survey lived at least 5 kilometers from the nearest primary school. Multivariate analysis o f the data indicates that when differences across households are controlled for, each additional kilometer o f distance from the nearest primary school reduces school registration rates by two to three percentage points, up to distances o f 12 to 15 kilometers. Ln other words, reducing the distance to school from 10 kilometers to 0 (i.e., to within the neighborhood) could raise registration ratesby 20 to 30 percentagepoints-a substantial increase by anymeasure. - xxxiv - I Distance from home to,nearest primary school Locality 0 1-2km. 3-4km. 5km.+ All Percentage of all sample children Urban 47.4 45.3 6.2 1.1 100 Rural 15.4 28.1 27.8 33.3 100 Total 1,9.6 30.4 24.9 25.1 100 Percentage of children registered for school Urban 85.5 83.1 78.9 71.0 84.0 Rural 43.6 38.8 32.6 19.8 32.4 Total 57.0 47.4 34.1 20.0 41.6 47. It is important to note, however, that distance is far from beingthe only determinant of school registration: among children inrural areas with a school inthe neighborhood, less than 44 percent registered for school; inurbanareas, the percentageis much higher-86 percent. But the fact that participation is still not universal suggests that even in such areas, factors other than distance continue to influence schooling decisions. These results imply that if school attendance i s to be improved, other constraints would need to be addressed. The multivariate analysis suggests that the quality of learning environment also matters: when taken together, the variables associated with this aspect o f supply explain a large fraction o f the differences in enrollment rates across localities that are not explained by differences in the characteristics o f the households. 48. Deploy teachers across schools more consistently in relation to size of enrollments. Improving the quality o f the learning environment requires action on many fronts. One i s to ensure that teachers are available in adequate numbers relative to the number o f students in the school. Because pupil-teacher ratios are so high on average inEthiopia, all schools suffer from chronic shortages o f teachers. The situation is much worse in some schools than in others, however, because o f the inconsistency in teacher deployment across schools. Among government primary schools enrolling about 1,000 pupils, for example, the number of teachers could range fiom less than 5 to more than 30. Across government primary schools, an estimated 28 percent o f the variation in the number o f teachers on staff i s accounted for by factors other thanthe sizeofenrollments; acrossgovernment secondaryschools, the corresponding shareis 30 percent. Figure 6 shows cross-country results that place Ethiopia among African countries having a fairly highly degree o f randomness inthe allocation o f teachers across primary schools. More detailed analysis shows that the relationbetween number o f teachers and number of pupils i s even weaker in grades 1-4. More consistent allocation o f teachers across schools i s clearly a goal that better administrative managemento fthe system can and should strive for. - xxxv - Figure 7. Allocation of Teachers across Government Primary Schools is Highly RandominEthiopia I h 7 50 B 8 40 g 30 20 ci I O 0 I- #% v1 Note: The degree o f randomness is computed as 1 minus the R2value o f regression equations relating number o f teachers assigned to teach the indicated grades to number o f pupils in those grades, with schools as the unit o f observation. The index has a 0-1 range, with higher values indicating greater inconsistency in the deployment o f teachers across schools. Data refer to government schools in all countries. Source: Figure 5.6. 49. Encourage schoolsprom-essivelvto offer complete instructional programs. Another factor on the supply side that affects the quality o f services i s the completeness o f the instructional program offered by schools. Where instructional programs are incomplete, some students drop out sooner than desired or repeat the highest grade offered simply because they have nowhere else to go inthe system. InEthiopia, only a fifth o f government primary schools offer instruction up to grade 8 (and these are generally the oldest schools), while just over 31 percent stop after grade 6, and 34 percent, after grade 4 (table 7). Rural schools are more likely than those inurban areas to offer an incomplete instructional program. The legacy o f the previous 6-4-4 structure of education system (which was replacedby the current 4-4-2-2 structure in 1994) i s evident inthe large share o f schools that stop at grade 6. Most of these schools were created around 1980 and appear not to have evolved to matchthe new structure o f the system. Schools that stop at grade 4 were created, on average, around 1987, and most also appear not to have developed beyond the scope o f their initial instructional program. Schools that offer fewer than the first four grades make up less than 8 percent o f Ethiopia's primary schools. Nearly69 percent o f them are new in the sense that they have not been around long enough to fill out the complete instructional menu. These schools have, on average, the highest pupil-teacher ratios than other schools, suggesting that staffing constraints may be one factor standing inthe way o f their development. Part o f the task of strengtheningthe administrative management o f the systemis to ensure that constraints at the school level, whether in staff or classroom facilities, are addressed so that a complete instructional program can be offered. Obviously, clarifying the number o f years o f schooling that are to be universalized will affect what i s considered a fillcycle o fprimary school instruction. - xxxvi - nstructionalProgramCharacterizes Many Rural Primary Schools Highf fered Grade 3 All Indicator/localitv or lower Grade 4 Grade 5 Grade 6 Grade 7 Grade 8 schools Percentage distribution of government schools"' 7.6 34.0 4.4 31.4 2.5 20.1 100.0 Inurbanareas 4.8 11.9 1.8 16.4 2.7 62.4 100.0 Inruralareas 8.0 37.0 4.7 33.4 2.5 14.3 100.0 Percentage distribution ofnew first- gradersb' 5.4 23.5 3.7 32.5 3.1 31.8 100.0 Average year schools were created 1997 1987 1985 1980 1978 1971 1982 Percentage ofthe schools that are new' 68.8 10.7 11.1 2.2 5.4 3.8 11.0 Average pupil-teacher ratio 84.5 73.6 82.4 79.1 77.8 72.5 76.3 a. Basedon data for 10,919 schools, of which 1,303 are inurbanareas, and9,616 are inruralareas. b. The number of new entrants totaled 1,599,110 for Ethiopiaas awhole, excludingSomali. c. Schoolscreatedno more than nyears ago, where n is the highestgrade of instructionofferedby the school. Source: See table 5.4. Address Demand-side Constraints 50. The demand-side constraints on school participation are many. They include household- level factors such as the affordability of schooling relative to household incomes, the opportunity cost of children's time, and the value of schooling perceived by parents; individual-level factors such as the child's personal characteristics (e.g., gender andorphanhood status); and community- level variables such as language. While many of these variables are impervious to policy interventions, particularly in the short run, documenting their impact helps planners understand the dependence of educational progress on success in solving problems in other sectors, and identifypossible interventions andtarget groups for special attention. 51. Reduce the direct costs of schooling for children from poor families. In Ethiopia schooling is tuition-free but families incur other fees and expenses to enroll their children. Among youths surveyed in 2001, about 13 percent cited affordability as the reason why they were not in school. The analysis shows that the direct costs o f schooling can indeed be significant: households spent an estimated total o f 387.8 million Birr on education services in 1999-2000 (nearly 30 percent o f the government's recurrent spending on education that year), and slightly more thanhalf o f the total was for primary schooling. Across households, the direct costs ofprimaryschooling weighmore heavily on the budgets o f the poorest households than on those oftheir richer neighbors, notjust because the per child outlay is a larger share o fhousehold income, but also because poorer families tend to be larger (table 8). The implication is that in mobilizingprivate household resources for education, it i s important for the government to avoid increasing the already large burden on poor households. Indeed, if private funding for postprimary education is successfully mobilized, the result should be to free up public funds for primary schooling, perhaps even to eliminate the cost that families still face to enroll a child in primary school. In countries such as Kenya, Lesotho, Malawi, and Uganda, concern that costs may be keeping children from school has already led the governments in these countries to implement free primary education in the 1990s as part o f their poverty reduction strategies - xxxvii - (Avenstrup 2004); more recently countries such as Madagascar, Rwanda, and Tanzania, to name a few, havejoined the list. Annual educationoutlay per child inprimary schoola' (Birr, unless otherwiseindicated) Poorest Richest Type of school quartile 2 3 quartile Government schools 21.9 27 36.5 57.2 Nongovernment school 37.4 56.9 64;l 340.9 Weighted average 22.6 28.1 38.2 80.8 As a percentage of consumption expenditure per adult equivalent 3.3 2.6 2.5 2.4 Memorandum: Annual average consumption expenditure per equivalent adult 688 1,095 1,555 3,366 a. Includes expenditure on fees, books, school supplies, and other education-related items. The estimates are based on data for the 8,112 households inthe sample that have children enrolled only inprimary education in the indicated type of school. The estimates are underestimated to the extent that spending on eveningprograms i s excluded. Source: Table 3.20. 52. The impact o f the direct costs o f schooling can be expected to interact with poverty. Analysis o f household data shows that after controlling for supply-side conditions and other household and personal characteristics, moving children from the poorest quintile to the richest quintile would, on average, raise attendancerates by 6 to 10percentagepoints. This is a modest increase, but it i s important to note that the effect o f income i s much stronger among children who live within 2 kilometers of the nearest primary school: moving from the poorest to just the next-poorest quintile would boost attendance rates by nearly 8 percentage points. The result suggests that income matters and that its influence is likely to grow as the direct costs o f attending school decline. The implicationfor policy development is that additional effort beyond making schools more accessible and free of charge may be needed to encourage the poorest families to enroll their children. 53. Offer additional help where neededto reduce the opportunity costs of schooling. When a child's presence in school conflicts with participation in work activities-whether at home, on the family farm, or in the labor market-families are said to incur an opportunity cost in enrolling a child at school. The data suggest that more than half o f rural Ethiopian boys and nearly a third o f rural girls report having worked in a range o f activities that might constitute a "job." Children who work are generally less likely to attend school, yet it appears that not all work is incompatible with schooling. This interpretation i s consistent with the modest size o f the impact o f children's work on schooling reflected in the multivariate analyses: all else being unchanged, a 10percent increase inthe availability o f adult labor inthe household reduces child participation inwork by 1.8 percentage points, and raises school attendanceratesby 0.6 to 0.8 o f 1percentagepoint. Inurbanareas, the estimates suggest that the addition of each young child in the household (which increases the demand for childcare by older siblings, especially sisters) - xxxviii - reduces by 4 percentage points the school participation rate of girls, while increasing their participation inunpaid domestic work by a comparable amount. 54. Help parents and communities better appreciate the value of schooling. In a country where schooling rates are as low as in Ethiopia, a plausible barrier to the schooling of children may simply be the fact that parents themselves have not beento schools andhave no idea what schools can do for their children. Descriptive information from the Welfare Monitoring Surveys reveals, for example, that among rural households who were askedwhy they have not sent their children to the nearest primary school, a higher percentage answered"no experience" or "have no need o f it" than "too expensive" or "too far" or "poor quality." The multivariate analysis shows that a child's chances of being enrolled is influenced by the percentage o f household heads in a locality (excluding the child's own household head) who are literate, even after controlling for such factors as household poverty, distance from school, and whether or not the child's own household head is literate. The fact that the impact of parental literacy diminishes sharply with the inclusion o f community-level literacy rates inthe regression analysis suggests that beliefs about importance of schooling may be formed at the community-level, and that exposing communities to the potential benefits of schooling may be a more useful approach in the short run for enhancing school participation than targeting each school-aged child's parents individually. 55. Remove barriersto school participation posedby language. sender, and orphanhood. The available data also allow an assessment o f the impact of these potentially important demand-side factors. Substantial differences in school participation exist across language groups, particularly inrural areas. Inmultivariate analysis, these gaps are diminishedonly modestly after controlling for the influence o f a full range o fregional andhousehold factors andthat o f community-specific effects. The results lend themselves to the interpretation that it is language per se, and not just associated differences in culture and prosperity, that cause registration rates to differ. The government introduced teaching in the child's mother tongue in its 1994 education policy, an action entirely consistent with the policy's intention to facilitate expansion o f primary school participation throughout the country. 56. With regardto girls andorphans, the patterns are consistentwith what mightbe expected. Gender gaps are primarily a rural phenomenon, and appear to be region-specific, implyingthat efforts to increase school participation by girls will require a region-by-region approach. For orphans, the multivariate analysis shows that after controlling for a variety o f household, community, andregional characteristics, childrenwho have lost one parent are 5 to 6 percentage points less likely to attend school than nonorphans, and that the effect i s twice as great for children who have lost both parents. The fact that orphans are at a disadvantage is evident from the regression results showing such children less likely to enroll in school than the biological children o fthe householdheadwith whom they live. Transforming Resources in Schools into Educational Outcomes 57. The outcomes of interest are the children's schooling careers-as reflected in entry, survival, andrepetition rates-and their learning achievement. Inboth domains, goodresults will require teachers and school directors to manage the pedagogical process effectively. The decisions made by both school and parents are interconnected. While teachers and school - xxxix - directors may have limited influence over entry rates, they are directly responsible for year-end decisions on whether a child will be promoted or asked to repeat a grade. It i s those decisions that intum provide parents the information they need, along with other considerations, to choose whether or not to keep their child enrolled in the next school year. And a child's academic progress influences to a large degree the decisions on promotion and grade repetition. Inlight o f this, and because student leaming is valued for its own sake, successful delivery o f education services must be reflected inprogressive improvements inthe level o f student achievement. Make it a Key Goal to Raise StudentAchievement 58. As indicated earlier, the Ministry o f Education embarked in 2000 on the first national baseline assessment o f grade 4 and grade 8 student achievement andhasjust completed a follow- up assessment in 2004 (data from which are not yet available at this writing). In the 2000 assessment, about 10,500 fourth-graders were tested inReading (inthe language o f instruction), English, Mathematics, and Environmental Science; and some 5,500 eighth-graders were tested in English, Mathematics, Chemistry, and Biology. The average percentage o f correct answers for all the subjects combined was 48 percent inthe grade 4 sample; and 41 percent in the grade 8 sample. Given that the test items were chosen from a range o f key topics in the curriculum for the grade tested as well as that for the previous grade, these scores indicate that a large number o f students were not achieving the curriculum objectives. 59. A key challenge for pedagogicalmanagement is therefore to improve uponthese baseline scores. Finding the right interventions will not be easy. The national baseline assessment evaluated the importance o f such factors as school infrastructure; school organization and management practices; teacher characteristics and practices; and the availability o f instructional materials and equipment. In both grades 4 and 8, the study confirmed the importance o f textbooks and the use o f radio for supplemental instruction, findings that are consistent with those reported inmany similar studies. Ingrade 8, while schools differed inthe share o f teachers who graduated from a teacher training college or higher levels o f training, the study found that this variable had no impact on student achievement. By contrast, the study found that process variables (i.e., those that reflect attitudes and behaviors) were important, including how well teachers prepared their lessons, how much the director focused on school matters, and the extent to which students engaged inhomework. Tighten the Relation between School Inputs and StudentLearning 60. Inan effort to understand the avenues for promoting student learning, the present study supplemented the national baseline assessment with an ad hoc analysis o f the correlates o f grade 8 examination scores inOromiya Region, based on data for 2001-02 specially assembled for 314 o f the region's total o f just over 1,000 primary schools with students in grade 8. Analysis o f these data found that differences across schools in some o f the tangible factors often considered important for school quality-especially teacher certification-accounted for very little o f the differences inperformance across students or schools. Although this finding appears puzzling, a moment's reflection reveals why it might not be so strange. While the tangible aspects o f schools define the infrastructure for teaching and learning, they alone cannot yield good results unless the humanagents put incharge o f them are properly motivated, supported, and held accountable - x l - for well-defined and measurable outcomes. Indeed, it is conceivable that even the best-equipped andstaffed school would fail inapoor management environment. 61. The importance o f good pedagogical management can hardly be exaggerated. Infigure 7 below, the tangible factors were converted into monetary terms and the results used to relate spending per student to examination score across schools. The figure shows virtually no link between the two variables: schools with relatively highlevels o f recurrent spending were just as likely to outperform as lag behind less well-endowed schools. This pattern, which is familiar from similar analyses in other countries, including developed ones, confirms the weak relation between school inputs and learning outcomes. The absence o f a link does not imply that money i s irrelevant; for example, money to buy more books would be a help given their consistently demonstrated impact on learning in many diverse settings. What it does say is that money by itself is insufficient to achieve good results. The national baseline assessment indicates that certain behaviors on the part o f teachers, school directors, and students are particularly important. The real issue from a systemic perspective is to discover the underlyingsupport and incentive structures that motivate teachers and school directors to adopt these and other helpfbl behaviors andpractices. Figure8. An Exampleof the Weak Link betweenSpendingper PrimaryPupiland ExaminationScores 45 i ,. 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 50 100 200 300 400I Average spendingper student (current Birr) Note: Data are for a sample of some 300 primary schools in the Oromiya region, each of which is representedby a circle in the figure. The mean score refers to the average percentageof correct answers of studentsin each school for eight subjects. Source: Figure 5 11 Taking Advantage of Decentralizationto Get Results 62. Ethiopia is currently undergoing a far-reaching decentralization o f the structure o f government. Budgets are allocated to the regions in block grants, and the regions in turn are expected to make block grants to the woreda (lower level) governments. Inthe new govemance landscape, these subnational governments are directly responsible for service delivery inprimary and secondary education (as well as some other subsectors). For them, two o f the issues discussed above are immediately pertinent: (a) the weak discipline inthe allocation o f resources - xli - across schools, so that human and financial resources are ineffectively channeled to the front lines &e., the schools); and (b) the apparent disconnection between resources and learning outcomes. The third aspect of management-making tradeoffs among alternative goals and means-will become increasingly relevant as power, finances, and responsibilities are transferred to the regional authorities. Learnfrom International and Domestic Experience 63. Eventhough the process of decentralization is still at an early stage, Ethiopiamight take note of the lessonsemergingfrom the experience incountries further along inprocess, andmight learn from its own implementation efforts as these mature. Winkler and Gershberg (2003) distill the following lessons,basedmostlyon experiencesinLatinAmerica andEasternEurope: "Efficiency and effectiveness are most likely to improve under decentralization when service providers-schools, local governments, or regional governments-are held accountablefor results. Accountability requires clear delineation of authority and responsibility and transparent andunderstandable information onresults (both educational and financial). Decentralization o f real decision making power to schools or school councils can significantly increaseparental participation inthe school, andhighlevels o fparental and community participation are associatedwith improvedschool performance. Decentralization o f education to sub-national governments does not in and o f itself empower parents and improve school performance. Further decentralization to schools (school councils or school boards) or local communities does empower parents and can improve school performance. For decentralization to schools to be successful, principals must acquire new skills in leadership andmanagement-financial, of teachers, andwith the community. The design of financial transfers to sub-national governments or schools has powerful effects onboth efficiency and equity. Decentralization requires that national and/or regional ministries of education be restructured; failure to restructure ministries i s a serious obstacle to realizing the benefits of decentralization. The decentralization of teacher management is critical to creating accountability and realizingthe potential benefits o f decentralization. National education ministries frequently resist decentralization on the grounds that sub- national governments, communities, and/or schools lack the capacity to manage education. Inpractice, this i s seldom true. Realdecentralization i s a long, evolutionary process." Some o f the foregoing ideas are already being made concrete in the Ethiopian context. For example, in 2002 the Ministry o f Education issued a set o f "Guidelines for Organization of Educational Management, Community Participation and Educational Finance" to clarify roles and responsibilities at all levels o f management in the education system, as well as to give explicit recognition to the importance o f community participation and parental involvement in school-based management. Identifying the most critical elements in these guidelines, putting them into practice, and evaluating the impacts of alternative arrangements are important tasks - xlii - that warrant attention in the years to come. If carried out systematically and routinely, they will provide lessons from Ethiopia's own implementation of decentralization inthe sector. Mobilize and Disseminate Information to EnhanceAccountability for Results. 65. As education is decentralized in Ethiopia, the challenges of ensuring accountability for results will increasingly come to the fore. The task i s difficult becausemany actors are involved and the outcomes are typically multidimensional, but the difficulty can be reduced by creating the right kinds o f information flow. In the first instance, the availability of appropriate information at the level o f schools can help stimulate a community's interest inensuring that its schools get a fair share o f the public resources for education. InUganda, for example, a large- scale public dissemination o f information on grants for schooling gave schools and parents the information they needed to monitor the grants. Over the space o fjust six years, the leakage o f funds fell from 80 to 20 percent, and schools were able to use the resources now reaching them to purchasetextbooks and other pedagogical inputs(World Bank 2004a). 66. O f course, information flow can also be channeled in other ways. Report cards such as those usedinParanaState inBrazil are aninteresting example (Servilla and Winkler 2003). They provide schools with information that compares indicators for the school (e.g., pupil-teacher ratios, examination results, and parent feedback) with reference averages at the national and relevant subnational levels. Report cards for school districts and other higher levels o f aggregation can also provide similar comparative information. The availability o f such information may be expected to improve conditions in schools in two ways: by creating incentives for underendowed schools to seek redress, and by pinpointing specific schools inthe systemthat warrant the attention of those incharge. InEthiopia, as inother countries, the use of information on student achievement for accountability purposes remains at an early stage. However, the country has already established a baseline, and this can be built on as part o f the broader effort to mobilize appropriate information for better management. Conclusion A decade after launching its 1994 New Education and Training Policy, Ethiopia's government can look back withjustifiable pride on the progress achieved. The very rapid growth of primary education reflects a genuine commitment to transforming the country's historically elitist system into one that serves all of Ethiopia's children. The government's recent Sustainable Development andPoverty Reduction Programreaffirmsits goal of universalizing primary educationwhile also envisioning improvements in quality and the expansion of other levels to enhance the skills profile of the workforce. These ambitions present sobering challenges on many fronts. Accomplishing them will likely require more public spending on education than the sector currently receives, as well as increased spending on education by households in postprimary education. Money alone i s unlikely to produce the desired results, however. In primary and secondary education-the levels on which the present study is focused-the system is truly at a crossroad today. The continual deterioration o f pedagogical conditions in schools across the country since 1993-94 sounds an alarm that bears heeding. I t i s a signal that the existing arrangements for service delivery-characterized in particular by the standards for teacher recruitment and classroom construction-may not be suited for expansion on the scale required to attain the goal of universal primary school completion. The work o f finding a new model will - xliii - need to be informed by a hard-nosed, system-wide assessment of potential tradeoffs among the options for moving forward. Better administrative management is also critical for continued progress, so that the available human and financial resources for education are channeled effectively to schools, communities, and households, Transforming these resources into higher levels of school participation and, even more important, into learning outcomes, i s the ultimate measure o f success in service delivery. This task i s difficult in the best of circumstances, but it can be made less so by cultivating a culture of accountability for results and good stewardship o f resources throughout the system. - 1 - Chapter 1. Demography,Economy,and GovernmentFinance 1.1 The fbture holds great promise for Ethiopia. Since the end o f the long andtragic civil war in 1991, the country has enjoyed a period of relative peace, albeit one marred by brief border wars with Eritrea in 1998 and again in2000. As a result, a greater potential now exists to channel more of the country's resources, both material and intellectual, into basic services that directly affect the welfare of the population. Moreover, the reformist government that took power in 1991 introducedmajor policy changes to shift the economy away from a centralized, command system toward a more liberal, market-based one, the impact o f which i s already helping to improve economic condition^.^ Finally, the government set inmotion a profound reform ingovernance to replace the centralized power structure o f the communist era with a federal system in which regional and subregional governments play explicit and substantive roles in the country's decision-making process. 1.2 These domestic developments provide a propitious context for education in Ethiopia. Reinforcing them is the growing international consensus on the goals o f development assistance and the means for achieving them. These goals have been crystallized inthe form o f the eight targets (along with the corresponding indicators and timelines) set forth in the 2000 United Nations Millennium Development Goals (MDGs).~Among them are two that relate directly to education: achieving universalprimary school completion by 2015, and ensuring gender parityin primary and secondary education by 2005. Regarding the means to reach the goals, some 200 countries, including Ethiopia, have endorsed the view espoused at the 2002 International Conference on Financing for Development inMonterrey, Mexico, that action is required by both rich andpoor countries. Rich countries can andmust increase aid flows andopen their markets to exports from poor countries, while poor countries can and must reform domestic policies to reduce poverty and boost progress toward the MDGs. The tone o f the dialogue continues to evolve, however, and a view is emerging among some leaders o fpoor countries that they have an obligation to set their own houses in order regardless o f the halting success of the industrialized countries inmatching rhetoric with a ~ t i o n . ~ 1.3 On balance, the prospects for educational development look promising, but the expectations and demands are also commensurately high. Some o f these demands are explicitly stated in the government's most recent Sustainable Development and Poverty Reduction Key components of the economic reforms included devaluation, gradual removal of trade restrictionsand price controls, and taming inflation, all of which spurred long-neglected private activities. The World Bank is currently preparing a Country EconomicMemorandumfor Ethiopia that will look more closely into the record of and prospects for economic growth (World Bank, forthcoming). The MillenniumDevelopment Goals (MDGs) include eight explicit objectives for reducingpoverty: (1) eradicating extreme poverty and hunger; (2) achieving universal primary education; (3) promoting gender equality and empowering women; (4) reducing child mortality; (5) improving matemal health; (6) combating HIV/AIDS, malaria, and other diseases; (7) ensuring environmental sustainability; and (8) developing a global partnership for development. For details on these goals, see www.developmentgoals.org. The World Bank's OperationsEvaluationDepartmentreportedthat in2000 net transfers to HIPCswere no higher than in 1995, while those to non-HIPCs remainedlower. The most recent InternationalMonetaryFunWorld Bank ImplementationReport on the Highly IndebtedPoor Countries (HIPC) Initiative suggests, however, that aggregatenet flows to HIPCs rose from US$4.0 billion in 1999 to US$7 billion in 2002 (although in 7 of the 23 African HIPCs, these flows actually fell), and concomitant increases in public spending on education and other poverty-reducingactivities have occurred (Hinchliffe 2003). In education, donors took the additionalstep of launchingthe Educationfor All FastTrack Initiative (EFA FTI) in 2002 to mobilize additional donor fundingspecificallyfor primary education, but gaps infinancingremainas of this writing. - 2 - Program (Govt. of Ethiopia 2002a). Among them are: achieving equitable universal primary education, improving the quality o f the education system at all levels, and training sufficient workers with skills relevant to the country's growing economy. The challenge for the managers o f the education system i s how to use existing and additional resources to deliver education serviceseffectively and thereby fblfill these expectations. 1.4 This report is intended as a contribution to ongoing discussions on the way forward, focusing on primary and secondary education6. As such, it i s a diagnostic rather than a prescriptive study. To set the stage for examining some of the key policy challenges in these subsectors, the rest of this chapter highlights the country's demography and its social and economic conditions; it also draws attention to key patterns and trends in government finance, including overall allocations of public spending to education. Box 1.1 below provides readers unfamiliarwithEthiopiaadditionalbackgroundinformationonthe country. Box 1.1. Ethiopia at a Glance Geography: Ethiopia is a landlocked nation bordered to the north by Eritrea and Djibouti, to the South by Kenya, to the east by Somalia, andto the west by the Sudan. With a surface area o f 1,127,127 sq km(435,186 square miles), it is twice the size o f France and slightly less than twice the size o f the State o f Texas. The country has a temperate climate inthe highlands, and a tropical climate inthe lowlands with a rainy season from mid-June to mid-September. It is characterized by extreme geographic diversity, and elevations range from 100 m below sea level (Kobar Sink in the DallolDepression) to 4,620 mabove sea level (Ras Dashen Mountains). People: Ethiopia's population o f 67 million in 2002 consists o f more than 100 distinguishable ethnolinguistic entities. The major groups include Oromo (40 percent), Amhara (30 percent), Tigray (12 percent), Sidama (9 percent), Shankella and Somali (6 percent each), and the Afar (4 percent). Amharic and English are the official languages. About 50 percent o f Ethiopians (mainly Amhara and Tigray) are Ethiopian Orthodox Christians, 40 percent Muslims, and about 2 percent Protestants and Roman Catholics. Almost 43 percent o f the populationis below age 15. Economy: Ethiopia's per capita gross national income o f US$lOO in 2002 is much lower than the corresponding averages o f US$480 for Sub-Saharan Africa and US$420 for low-income countries as a whole. After years o f socialist experiment, Ethiopia began to reform its economy in 1991. It has largely dismantled the extensive system o f price controls, lowered tax rates, and removed restrictions on the private sector. Agriculture nonetheless continues to dominate the Ethiopian economy: it employs 80 percent o f the workforce and accounts for 52 percent o f the gross domestic product (GDP), compared with the manufacturing sector's share o f 7 percent. Although Ethiopia has one of the lowest exports per capita inthe world, the economy is vulnerable to terms o f trade shocks because one product- coffee-accounts for as much as 60 percent o f its exports revenue in recent years. The country also suffers from frequent and severe droughts, averaging one per decade during the past 100 years, and one every three years during the last 10 years. The most recent drought occurred in2002 andput as many as 12 million people at risk o f starvation. T h e impact of these factors i s reflected inthe volatility o f GDP growth: it expanded by 7 percent in2000, grew by only 1.2 percent the following year, and shrank 3.8 percent in2002. History, Politics, and Civil Conflicts: Ethiopia has one o f the oldest civilizations in the world and is unique among African countries in not having been a foreign colony. According to legend, it was founded by Menelik I,the son 01 King Solomon and the Queen o f Sheba, around the 5" century B.C. Christianity was introduced by missionaries fron Egypt and Syria inthe 4" century A.D. However, the country experienced bitter religious conflict and isolation fron the rise o f Islam inthe 7" century, Ethiopianresistance to Portuguese hegemony over the Indian Ocean, and attempts to convert Ethlopia to Roman Catholicism. Under the Emperors Theodore I1(1855-68), Johannes IV (1872-89), anc Menelik I1 (1889-1913), the kingdom began to emerge from its medieval seclusion. When Menelik I1 died, h i s grandson, Lij Iyasu, succeededto the throne but soon lost support because o f his Muslimties. In 1996he was deposed See World Bank (2004b) for a discussionof key policy challengesinhigher education; and Franz(2003) for atreatment of the issues intechnical andvocationaleducationand training. - 3 - by the Christian nobility, and Menelik's daughter, Zewditu was made empress. Her cousin Ras Teferi Makonnen (1892-1975) was made regent and successor to the throne. After the empress' death, he ascendedthe throne in 1930 as Emperor Haile Selassie. His reign was interrupted in 1936 when Italian forces invaded and occupied Ethiopia, and reestablished five years later when Ethiopian forces, aided by the British, defeated the Italians. In 1974 Haile Selassie was overthrown by rebel soldiers who installed a military government. Major MengistuHaile Mariamemerged as their leader and in 1977 became the chairman o f the ruling council known as the Derg (Coordinating Committee o f the Armed Forces, Police, andTerritorial Army; later called the ProvisionalMilitary Administrative Council). In 1977, the Derg launched a campaign o f terror-the Red Terror-that resulted in the jailing, torture, and death o f untold thousands, in an attempt to stem rebellion against its policies promoting a unique nationalism through suppression o f ethnic diversity. These struggles galvanized the predominantly ethnic-political movements into action and the Derg was eventually overthrown in 1991. The new Transitional Government o f Ethiopia (TGE) was established by the Ethiopian Peoples' Revolutionary Democratic Front (EPRDF), a coalition whose key members include the Amhara Nationality Democratic Movement (ANDM), the Oromo People's Democratic Organization (OPDO), the South Ethiopia Peoples' Democratic Organization (SEPDO), and the Tigray People's Liberation Front (TPLF). The year 1991 also saw the Eritrean People's Liberation Front (EPLF) assuming control o f Eritrea and establishing a provisional government. Following a referendum in which Eritreans voted overwhelmingly for independence, the international community recognized Eritrea as a country in 1993. InEthiopia, the TGEadopted anew constitution in1994, whichcreateda federal structure ofgovernment consistingof nine ethnic regional governments and two administrative urban councils (Addis Ababa and Dire Dawa). Ethiopia's first multiparty elections were held in 1995. Representation at the federal level i s based on population (House o f Peoples' Representatives with 545 members) and ethnicity (House o f the Federation, with each ethnic group or nationality represented by at least one member, plus an additional member per million inhabitants). The government is headed by a Prime Minister chosen for a five-year term by the majority party. Regional governments were set up by proclamation No. 711992. They are run by an elected state council (the highest organ o f state authority) and a State Administration (highest organ o f executive power). The state council defines the region's policy and has all legislative, executive, and judicial powers over the affairs o f the region. As o f 2001-02, W h e r decentralization has been set in motion in four regions (Tigray, Amhara, Oromiya, and SNNPR [Southern Nations and Nationalities Peoples Republic]) and two adrmnistrative councils (Addis Ababa and Dire Dawa), with the objective of transferring responsibilities from the regional governments to the woredas and kebeles (the level o f administration below the woreda) in a well-defmed structure that formalizes the h c t i o n a l responsibilities and constitutional powers o f these levels o f government. The central government retains control over such areas as defense, foreign affairs, and economic i Relations with the World Bank Ethiopia became a member o f the World Bank on December 27, 1945, and received its f i s t loan in 1951. Since then Ethiopia has receivedpolicy and technical support, andproject andprogramfinancing. Ethiopia is the International Development Association's (IDA) largest beneficiary in Africa; $1.85 billion was committed by January 7, 2004. The World Bank, the International Monetary Fund(IMF), and bilateral development partners are currently supporting a comprehensive debt reduction program under the enhanced Heavily Indebted Poor Countries (HIPC) Initiative, but the details o f how much debt relief Ethiopia will receive are still under discussion. Sources: Bureauof African Affairs, US.Departmentof State 2001,"Background on: Ethiopia," ~httv:llwww.state.~ovlrioalei~gnl2859.h~~; CentralIntelligenceAgency, "The World Factbook:Ethiopia," ; EmbassyofEthiopia, Washington, D.C., "Ethiopia: History," ; SIL Intemational,"Languages of Ethiopia," ; World Bank, "Country Brief:Ethiopia," ~ht~:llwww.worldbank,ordafriet2.htm>; World Bank, "Ethiopia at aGlance," ; orphanhoodrates for Ethiopia,from Goa. of Ethiopiaand ORC Mam,2001; for Rwanda,World Bank2003a; and for allothercountries, Ainsworth and Filmer 2002. Economic Conditions 1.14 Inthis section we highlight a few key features of economic conditions inEthi0~ia.l~ The country has had fairly rapid growth in the 1 9 9 0 ~but it has not always been that way. The ~ previous two decades were characterized by volatile growth and a major deterioration o f the real per capita gross domestic product (GDP). Infact, only bythe start ofthe 2000s didthe per capita GDP recover to the historicalpeak achieved inthe early 1970s(figure 1.2). l4 For a thorough treatment of this topic, readers are referred to the World Bank's Country Economic Memorandum which is currentlyunder preparation. - 9 - Figure 1.2. Real Gross Domestic Product Per Capita, Ethiopia, 1960-2002 s 2001 Y I 150 I I 4 I , I I 1960 1965 1970 1975 1980 1985 1990 1995 2000 Source: World Bank2004d. 1.15 The growth spurt o f the 1990s boosted real per capita GDP at the rate o f 1.1percent a year between 1992 and 2000. The cessationof civil war has obviously been an important factor, butthe economic reforms that the new government introduced-which, among other results, kept inflation under control and improved management of the budget deficit, exchange rate, ind investment in infrastructure-probably also played a role in the recovery. It i s noteworthy that although agriculture i s the mainstay o f the economy and the focus o f the government's development strategy, it was the industrial and services sectors that grew fastest, with output expanding at, respectively, 8.5 and 9.1 percent a year during 1992-2000 (table 1.4). O f the 5.5 percent per year growth in total GDP over the period, 65.4 percent (or 3.6 percentage points) were contributed with the services sector, compared with agriculture's contribution of 18.2 percent (or 1.O percentagepoints) (Easterly 2002). Table 1.4. Sources of Recent Economic Growth, Ethiopia, 1992-93 to 1999-00 Average share of GI)P Percentage contribution to during 1992-93 Average annual growth aggregate average annual Sector, to 1999-00 rate (percentage p.a.) growth,, Agriculture 48.8 2.1 18.2 Industry 11.0 8.5 16.4 Services 40.1 9.1 65.4 Total GDP 100% 5.5 100.0 1.16 For the purpose of this report, what i s important from the foregoing discussion i s that while the recent rapid growth is very promising, the prospects for sustaining the pace of expansion remain uncertain as the country seeks to move beyond the recovery phase. The fragility and volatility o f Ethiopia's macroeconomic future serve as important reminders that resources are tight and are likely to remain so. Under these conditions, the education sector's claim on public funding will obviously be facilitated by demonstrating good stewardship in the management of resources to achieve tangible results-results that must be measurednot only in - 10- quantitative expansion o f enrollments but also interms of gains in student learning and, beyond that, inthe sector's responsivenessto labor marketneeds. Pattern of Government Finance 1.17 Inlight of Ethiopia's demographic, social, and macroeconomic conditions, we can now examine overall government finance. We begin with a brief description o f the fiscal decentralization that has characterized intergovernmental relations since 1994. We then document the composition of andtrends ingovernment revenues and expenditures, and finishthe discussionwith somebroad observations on aggregatepublic spendingon education.l5 Fiscal Decentralization 1.18 Since the adoption of a federal structure of government in 1994, the nature of intergovernmental fiscal relations has beenundergoing a profound transformation. Inthe present arrangement, there are three tiers of government and each is accorded, constitutionally, the authority and autonomy to manage its expenditures and revenues. The federal government retains the authority to collect most of the taxes, some exclusively. Only it can run budget deficits and borrow internationally. The revenues collected from all sources are shared with the regions through a block grant formula that has three components-population, level of poverty, and development or expenditure needs (see table A1.2 for a summary of the formulae that have been used since 1994). The grants received by regions represent the bulk of regional revenues (averaging nearly 80 percent of total regional budgets excluding external funds); these grants are untied,meaningthat regionalgovernmentshavewide latitude to allocate the funds acrossvarious regional priorities. 1.19 At the regional level, these grants are supplemented by the region's own revenues generatedthrough a number o f taxes that the regions have the authority to collect, primarily on income from regional employees, landuse, and sales within regions. The grant from the federal government and the region's self-generated revenues, along with the regional allocation of foreign funds received by the federal government, add up to the regional budget. Once the regions determine their resource envelope, they are expected to transfer some of the resources to the lower levels of government (woredas).For the regions undertakingfurther decentralization to the woredas,the sharing i s done also as untiedblock grants according to a formula, similar to the national formula with slight regional variations. In the nondecentralizing regions, regional governments transfer resources according to planned expenditures submitted by the woredas to the regional governments.16 1.20 Under the new structure of fiscal relations, each level o f government is responsible for delivering certain services. In education, the federal government is responsible for higher education and some teacher training; regional governments for secondary education, teacher training, and technical and vocational education and training; and woreda governments for A moredetaileddiscussion of education finance is developedinchapter 3. Readersare also referredto the World Bank's 2003 EthiopiaPublic ExpenditureReview for an assessmentofoverall government finance. l6The decentralizingregions andadministrativecouncils are: Tigray, Amhara, Oromiya, and SNNPR andAddis Ababa and Dire Dawa. The planis to implementthe decentralizationprogramover the period2001--02 to 200647 to transfer many responsibilitiesfrom the regionalgovernmentto the woreda andevenkebele level. During this period, the framework for the new decentralizedstructurewill be laid down, andwill includeconstitutionalempowerment of woreda governments, division of functionalresponsibilitiesbetweenworedas andregionalgovernments, andclarification of organizationalstructure andfunctions of offices. - 11 - primary education. This arrangement means that when evaluating government spending on education, it is important to account for expenditures on educationby all the different actors.17A proper accounting i s also neededto provide a baseline for informing discussions about spending priorities, as well as for tracking the impact o f the new structure on the global pattern of spending. Trends inOverallGovernmentRevenuesandExpenditure 1.21 Below we present key indicators o f revenue and expenditure trends since the 1980s and draw attention to the relationbetweenthem. 1.22 Government revenues. Table 1.5 shows selected indicators of consolidated government revenues between 1980-81 and 2001-02 (the horizontal lines in the table mark the transitional period from the Mengistu regime to the present government). The data offer two clear observations. The first is that revenues were volatile in the last two decades. Three phases are observable, consistent with the recent history o f Ethiopia. Inthe first phase, the Derg years (1981-89), total revenues including grants, as a percentage o f gross domestic product (GDP), rose from 18 percent in 1981 to 28 percent in 1989, when they peaked. Inthe second phase, which coincides with the collapse o f the Derg regime and the immediate transition (1990-92), total revenues including grants, as a share o f GDP, fell to their lowest level. Inthe final phase, 1993-2001, which defines the current government's rule, revenues as a percentage of GDP rose again, andreached24 percent by 2002. 1.23 The second observation i s that the economy is grant-dependent, since almost 20 percent o f revenues come from external grants. Nonetheless, revenues excluding external grants have been growing, rising from 11.9 percent of GDP in 1992-93 to 19.1 percent in2001-02, or the equivalent o f 79.4 Birr per person to 194.7 Birr per person (in constant 1994-95 prices)-a 2.4- fold increase over the period. Despite this very positive trend, Ethiopia's level of domestic resource mobilization, which averaged about 17.2 percent o f GDP throughout the period, remains slightly below the corresponding statistic o f 19.0 percent for 33 low-income African countries around 2000, and 20.0 percent for 21 non-African low income countries (Bruns, Mingat, andRakotomalala 2003).18 "Fordetailsonthepreparationofthedataoneducationfinancefor thisreport, seeTechnicalNote 1ontheDataonCurrent Public Expenditureon EducationinEthiopiaat the end ofthis report. '*Countriesinthe citedstudyare thosethat haveyet to achievethe goalofuniversalprimary schoolcompletionas of2002. - 12- Table 1.5. Trends in Overall Government Revenues, EthioDia. 1980-81 to 2001-02 Per capita revenues net of Total E vernment rei lues extern; ;rants Percentage share of In constant Index AS external In constant Index Ethiopian Gregorian 1994-95 (1994- bercentage grants in total 1994-95 (1994- Calendar Calendar Birr 95=100) of GDP revenues Birr 95=100) 1973 1980-81 4,837.8 100 18.2 9.8 124.7 100 1974 1981-82 5,097.4 105 19.0 12.2 127.8 102 1975 1982-83 5,628.7 116 19.4 10.7 137.3 110 1976 1983-84 6,09 1.4 126 21.5 10.0 144.3 116 1977 1984-85 5,343.8 110 21.3 21.4 123.1 99 1978 1985-86 6,206.1 128 22.5 13.6 138.8 111 1979 1986-86 6,608.7 137 21.1 9.9 143.4 115 1980 1987-88 8,102.9 167 25.8 15.5 170.2 137 1981 1988-89 8,779.0 181 28.0 17.0 178.1 143 1982 1989-90 6,404.7 132 19.9 11.3 125.1 100 1983 1990-9 1 4,806.0 99 16.1 14.6 90.7 73 1984 1991-92 3,601.3 74 13.2 19.7 65.7 53 1985 1992-93 4,231.1 87 13.7 12.7 79.4 64 1986 1993-94 5,552.4 115 17.4 20.0 101.1 81 1987 1994-95 7,044.6 146 20.8 16.1 124.7 100 1988 1995-96 7,987.1 165 21.3 13.6 137.2 110 1989 1996-97 8,628.6 178 21.8 16.6 144.3 116 1990 1997-98 8,125.1 168 20.9 13.6 132.5 106 1991 1998-99 8,808.7 182 21.4 16.9 140.3 112 1992 1999-00 9,208.4 190 21.6 15.4 143.2 115 1993 2000-01 11,236.8 232 24.2 20.5 170.8 137 1994 2001-02 13,101.1 271 23.5 18.6 194.7 156 Note: 1982 E.C.to 1984 E.C. correspondto the transitionfrom the Mengisturegime to the new govemment Source: Basedon datain appendix table AI.I 1.24 Government expenditures. The relevant data appear in table 1.6. In the 1990s, total govemment spending as a percentage o f GDP rose steadily and peaked by the end o f the decade, a pattern that is strikingly similar to the trends inthis statistic inthe 1980s. Throughout the 1980s and 1990s, expenditures exceeded revenues, implyinga routine reliance o n borrowingand deficit financing to fund government spending. Total spending includes a substantial share devoted to capital spending, averaging about 34 percent in the decade since 1992-93, which is slightly larger than the average of 30 percent in the 1980s. With regard to interest payments o n public debt, the level i s noticeably higher in the 1990s than in the 1980s. Still, the amounts relative to the GDP remain modest, averaging about 2 percent of GDP in the years since 1996-97 or between 6 to 9 percent of total government spending. The composition of the debt burden has changed, however; the interest paid on external debt rose from 23 percent of total interest payment when the new government took power, to 47 percent in2001-02. - 1 3 - Table 1.6. Trends in Total Government Expenditu lEthiopia, 1980-81 to 2001-02 Btalspendi L 1 Recurren spending Interer Inall pub !debt arcentagc Per share of [n1994- capita As As interest 95 Index As pending ercentagt As 'ercentagc As paidon Ethiopian hegorian :onstant (1994- percentage LS ratiool (1994- of total Nercentagc of total ercentagf external Calendar Zaalendar Birr 95=100) of GDP -evenues" )5=100) Fending of GDP spending of GDP debt 1973 1980-81 5,668.1 100.0 21.3 1.17 100 77.9 16.6 3.7 0.8 22.8 1974 1981-82 6,629.0 117.0 24.7 1.30 114 74.3 18.3 3.0 0.7 21.1 1975 1982-83 8,756.2 154.5 30.2 1.56 146 67.1 20.3 2.4 0.7 22.8 1976 1983-84 8,005.0 141.2 28.3 1.31 130 72.1 20.4 4.4 1.2 18.2 1977 1984-85 7,024.4 123.9 28.0 1.31 111 69.4 19.4 5.8 1.6 12.9 1978 1985-86 7,780.0 137.3 28.3 1.25 119 63.9 18.0 4.8 1.3 20.0 1979 1986-86 8,187.3 144.4 26.1 1.24 122 65.6 17.1 5.5 1.4 31.9 1980 1987-88 9,663.0 170.5 30.8 1.19 139 70.2 21.6 5.2 1.6 38.8 1981 1988-89 10,709.8 188.9 34.2 1.22 149 66.2 22.6 4.3 1.5 33.9 1982 1989-90 9,543.6 168.4 29.7 1.49 128 72.7 21.6 4.3 1.3 20.7 1983 1990-91 7,353.0 129.7 24.6 1.53 95 75.0 18.5 5.4 1.3 19.2 1984 1991-92 5,504.9 97.1 20.2 1.53 69 77.4 15.6 7.3 1.5 19.9 1985 1992-93 6,037.8 106.5 19.6 1.43 78 65.8 12.9 10.2 2.0 22.9 1986 1993-94 7,995.9 141.1 25.0 1.44 100 62.0 15.5 13.5 3.4 15.4 1987 1994-95 8,371.9 147.7 24.7 1.19 101 62.3 15.4 10.0 2.5 28.8 1988 1995-96 9,058.7 159.8 24.1 1.13 107 61.0 14.7 10.1 2.4 33.9 1989 199697 9,46 1.4 166.9 23.9 1.10 108 57.9 13.8 9.3 2.2 30.8 1990 1997-98 9,733.0 171.7 25.0 1.20 109 63.1 15.8 7.4 1.9 37.0 1991 1998-99 12,615.5 222.6 30.6 1.43 138 67.9 20.8 6.4 2.0 38.6 1992 1999-00 14,099.7 248.8 33.0 1.53 150 80.0 26.4 6.5 2.2 35.6 1993 2000-01 13,473.0 237.7 29.1 1.20 140 67.4 19.6 7.0 2.0 46.8 1994 2001-02 16,402.5 289.4 29.4 1.25 167 65.3 19.2 - - - -Datanot available. a. Includinggrants. Source: Basedon data inappendix table Al.3. 1.25 Relation between revenues and expenditures. To round out the foregoing discussion, we plot in figure 1.3 below the relationship between per capita revenues and per capita recurrent spending. The data show that for every additional Bin per capita mobilized, the government i s able to set aside about 0.64 Birr for current spending. The positive relationship between revenues and recurrent spending draws attention to the reality that the more resources that can be raised domestically, the more the government can finance its day-to-day effort to deliver services. This relationship i s especially important for a sector such as education where the availability of recurrent spendingi s so crucial for schools to hnction smoothly. - 14- Figure 1.3. RelationbetweenGovernmentRevenues and Current Spending, Ethiopia, 1980-81 to 2001-02 --- I I A 150 100 10 ° 0 50 100 150 200 250 Percapitarevenues(1994-95 constantBirr) Note:Govemrrentreve~wexcludegrants. soUroe:Basedcndafaf~1980-81t o U x l l - 0 2 i n ~ xtabiesAl.1andA1.3. Aggregate Trends inPublic Spendingon Education 1.26 The preceding section described recent developments in the resource envelope o f the government. It i s now appropriate to ask how much o f the available resources have been allocated to education, and how well Ethiopia compares with other low-income countries in terms o fthe resources it devotes to this important areao f social development. 1.27 Recent patterns in historical perspective. Table 1.7 shows that total public spending on education has been rising since 1992-93, markinga significant break from the past. It averaged 3.6 percent o f GDP in the decade following 1992-93, and 13.8 percent o f total government spending, compared with an average of 2.4 percent of GDP and 8.7 percent of total government spendingfrom 1980-81 to 1988-89 under the Mengisturegime.Public spendingon education as a percentage o f domestic revenue resources rose even higher, averaging 21 percent inthe period from 1992-93 to 2001-02, compared with 12.9 percent between 1980-81 and 1988-89. 1.28 These broad positive trends in education financing have to be seen within the context o f two related facts. First, a disaggregation o f the data between recurrent and capital spending (possible only for the series afier 1992-93) shows that a sizable share o f the total amount has gone into capital investment. Thus, while total public spending on education as a percentage of GDP has been rising in the 1 9 9 0 ~recurrent spending as a percentage o f GDP has remained ~ constant for most o fthe decade. - 1 5 - Table 1.7. Trends in Public SpendingonEducation,Ethiopia, 1980-81 to 2001-02 Total recu rent and capi a1spending Recurrent spe I AS As percentage As percentageof percentage of totalgovernment As of total government i s percentageof :urrent spendingnet Xhiopian Gregorian iercentage government revenues net totalspending of debt interest Zalendar Calendar of GDP , spending of grants on education pay,ment 1973 1980181 2.1 10.0 13.0 - - 1974 1981/82 2.4 9.7 14.4 - - 1975 1982183 2.2 7.3 12.7 - - 1976 1983184 2.5 8.8 12.9 - 1977 1984185 2.4 8.4 14.1 - 1978 1985186 2.4 8.5 12.3 - 1979 1986186 2.4 9.3 12.8 - 1980 1987188 2.5 8.2 11.6 - 1981 1988189 2.9 8.5 12.4 - 1982 1989190 2.8 9.4 15.8 - 1983 1990191 2.5 10.1 18.1 - 1984 1991i92 2.5 12.6 23.9 - 1985 1992193 2.8 14.3 23.4 1986 1993194 4.0 15.8 28.5 3.0 75.6 24.6 1987 1994195 3.5 14.3 20.3 2.7 76.1 20.9 1988 1995196 3.8 15.7 20.6 2.6 68.2 21.0 1989 1996197 3.6 15.3 20.1 2.6 71.7 22.5 1990 1997198 3.5 14.1 19.5 2.6 74.2 18.8 1991 1998199 3.5 11.5 19.9 2.6 73.6 13.8 1992 1999100 3.1 9.5 17.3 2.6 82.0 10.6 1993 2000101 4.1 14.2 21.5 3.O 71.5 16.9 1994 2001102 4.0 13.6 21.0 3.0 75.9 -Datanotavailable. a. Excludes spending on items such as vehicles, equipment, building, and construction materials which were classified in the budget data as recurrent spending; for actualdata, see appendix table Al.5 Source: Computedfrom data in appendix tables Al.1, A1.3.-A1.5. 1.29 The secofid related fact regarding education financing trends is that in the last two decades, increases in government spending did not translate immediately into increases in spending on education. A plot of total government spending relative to GDP and total education spending relative to GDP, between 1980-81 and 2001-02, reveals no systematic relationship (figure 1.4). This suggests that historically education did not automatically become a priority when the government increased its overall spending, which is understandable within the historical context of the country's long civil wars and recurrent famines. Nonetheless, the trend in recent years under the new government is in the right direction, as figure 1.5 shows. The figure first converts all recurrent education spending, for which we have data only since 1993- 94, into constant 1994-95 Birr. It then plots these real expenditures relative to the 1993-94 spending. By indexing spending in every year to the spending in 1993-94, it becomes clear that spending on education rose after 2000, trailing by two years the rise in total recurrent government spending that coincided with the period of the border war with Eritrea. But since the - 16- end of hostilities in 2000, real recurrent government spending has remained higher than the historical levels inthe 199Os, and education spendinghas edgedup as the government turned its attention to its program for poverty reduction, which includes education as a keycomponent. Figure1.4. RelationbetweenOverallGovernment Spending and GovernmentSpendingonEducation, Ethiopia, 1980-81to 2001-02 4.5 4.0 c 8: a $8 Fo 3.5 8 % 3.0 0 0 M$ 0 3 g E 2.5 0a a - 0 ia a Kn E' a 2.0 0 0 15 20 25 30 35 40 Overallspending(as %of GDP) Note:Bothaxesreferto recurrentandcapitalsperading. Sowce: Basedondataintables 1.6and1.7. Figure1.5. Trends inGovernment OverallRecurrent Spendingand EducationRecurrent Spending, Ethiopia, 1993-94 to 2001-02 250 1 200 150 100 50 0 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 Note: The spendinginboth series are expressedin 199495 constantBirr, indexedto spendingin 1993-94, the first year inthe series. Source: Basedon data in tables 1.6 and 1.7. 1.30 Cross-country comparisons. Ethiopia's recurrent spending on education, at 3.0 percent o f GDP in2001-02, i s slightly lower than the average o f 3.2 percent o f GDP among 54 other low- income countries circa 2000, but about the same as the average for African countries in the sample (Bruns,Mingat, and Rakotomalala 2003). It falls far short, however, o f the average o f 3.8 percent among the countries deemed closest to achieving the goal o f universal primary - 17- completion. The comparative data, which are plotted in figure 1.6, suggest two lessons: first, public spendingon education inEthiopiaprobably needs to be increasedbeyond its current level ifthe country is to advance toward the education MillenniumDevelopment Goal; second, the fact that only a few countries have managed to raise spending above 4.0 percent o f GDP means that in Ethiopia as elsewhere it will probably require an extraordinary confluence o f favorable circumstances to sustainclaims onpublic spendingbeyondthis threshold. Figure1.6. Public Recurrent SpendingonEducationand Per CapitaGDP, Ethiopia and Other Countries, circa 2000 8.0 - A 'r: 8 Other developingcountry 0 ' A 2 7.0 - a8 0' 2: 0 0 0 c2gE 5.0 - A +d 4.0 - A A 0 '' 2 e3 A A A A A A A A 3.0 - A EthiohaA 0 & AA A O A 0 0 2.0 - A A A A 0 0 A 0 o % A 0 A 0 1.0 - A 0 d o 560 1,doo 2,doo 3,dOO Per capitaGDP (PPP$) GDP Grossdomesticproduct; PPPPurchasingpowerparity. Source: Bms, Mingat, andRakotomalala2003 for dataon recurrentpublic spendingoneducation; World Bank SlMA database for data Conclusion 1.31 There is wide acknowledgment about the primacy of education in the fight against poverty and its role in promoting social well-being (e.g., Case 2001). For Ethiopia, the end o f internal conflict has created the opportunity for govemment to give even more attention to the development o f the sector. Indeed, a clear political commitment and momentum now exist to transform the historicallyelitist andnarrowly accessiblesysteminto one that offers opportunities to the mass o f the population. The recent decentralization of govemmental structure, finances, and functions will undoubtedly help make this goal a reality. Accordingly, the government's expectation is that the policy o f decentralization will produce results by clarifying responsibilities for service delivery at each level o f govemment; by enabling and empowering lower levels of govemment to take direct responsibility for managing their own affairs; and by makinglocal administrations more accountable for service delivery, particularly insuch areas as education, health, andagricultural extension. 1.32 Matching goals with means i s at the crux o f the challenge facing policymakers in Ethiopia. The country's historical pattern indicates that education has not always beenprioritized inthe allocation of public spending, quiteunderstandable inthe context ofthe country's lengthy civil wars and persistent food insecurity. However, recent developments suggest a positive turn o f events, as real recurrent spending since 2000 has begun to edge upward. This progress has been aided by intemal developments-including an end to hostilities with Eritrea and increased - 18- mobilization o f domestic resources-and intemational consensus to build partnerships across countries to accelerate the accessibility o f education through such initiatives as HIPC, MDGs, EFA, and the increased use o f budget support mechanisms (such as the World Bank's Poverty Reduction Support Credit) to channel donor funds. Since all these initiatives provide additional sources o f increased funding for education, there is a good prospect that recent increases in public spending on education will continue. Yet there i s no escaping the reality that the competition for public resources will be tight, as other sectors such as roads, water, and health can also make a legitimate claim based on their contribution to poverty reduction. Thus, while every effort must be made to raise recurrent spending on education from its present level of 3.0 percent o f GDP, pushing the sector's claim beyond 4.0 or 5.0 percent o f GDP is likely to prove progressively difficult, perhaps even elusive. 1.33 For the managers o f the education sector, the challenge is how best to capitalize on the favorable conditions o f political commitment and the promise o f increased domestic and intemational funding in order to meet the expectations for the sector: achieving universal basic education and expanding the opportunities for postbasic education; equipping students and frontline teaching and administrative staff in such a way as to improve learning outcomes; and producing graduates insufficient numbers andwith the appropriate skills to sustain the growth o f the economy. This report will focus on the issues as they pertain to primary and secondary education and, where appropriate and possible, draw in relevant information on the other levels to provide the big picture. Moving beyond the question o f the global resource envelope, the rest o f the report will examine the policy choices that are available to manage the fiscal viability o f a system that is expected to serve a vastly expanded clientele, as well as draw attention to the possibilities for enhancing the efficiency and effectiveness of service delivery. - 19- Chapter 2. Enrollments and Patterns of Student Flow 2.1 This chapter traces the historical expansion of enrollments across levels of education in Ethiopia, places the system's coverage ina cross-country perspective, and draws attention to key indicators o f student flow-in particular, entry rates to first grade, survival and completion rates, and repetitionrates inprimary and secondary education. The results show that the country has made enormous progress on all quantitative measureso f coverage. Primary school enrollments in 2002-03 were 2.8 times what they were in 1993-94 when the new government came to power, while those in secondary education and regular government higher education alone were, respectively, 1.9 and 2.3 times as large. Interms o f student flow, an estimated60 percent o f each age cohort in 2000-01 were expected eventually to enter grade 1, compared with 37 percent in 1995-96; and about 54 percent of eachenteringcohort of first-graders in2000-01 were expected to reach grade 5, up from 49 percent in 1993-94. While these trends are beginningto look like a clear break from a past inwhich access to evenbasic education was highly restricted, there i s no room for complacency if the country is to achieve the Millennium Development Goal o f universal primary school completion by 2015. Critical decisions loom on the horizon regarding the duration o f schooling to be universalized, the placement of selection points to manage the growing pressures on postprimary education, the choice o f indicators to track progress in the sector, particularly at the lower levels, and the design o f interventions to continue boosting primaryschool completion. Structureof the EducationSystem 2.2 From 1962 to 1994, the education system had a 6-2-4 structure: six years of primary schooling, followed by two years o f junior secondary education, and four years of senior secondary education. National examinations at the end o f each cycle (Le., ingrades 6, 8, and 12) regulated student selection from one cycle to the next. Following the 1994 Education and Training Policy and Strategy, the government created a new structure (figure 2.1). Still inplace today, it consists of an eight-year primary education cycle, which i s itself divided into a basic education cycle covering grades 1-4, and a general primarycycle covering grades 5-8, followed by two years of general secondary education (grades 9-10>, and two years of preparatory secondary education (grades 11-12). National examinations are now administered only at the endof grades 10and 12; regional examinations have replacedthose at the endofgrade 8. 2.3 The changeover to the 4-4-2-2 structure was motivated in part by a concem that the education system existing in 1994 was too similar to the models in developed countries and therefore unsuited to conditions in Ethiopia. It was also perceived to be performing poorly in terms o f access, equity, relevance, and quality. The 1994 education reforms sought to address theseproblems andincludedthe following features among others: 0 Renewedemphasison student learningandproblem-solvingskills; 0 Revitalizationo ftechnical andvocational education and training; 0 The teaching ofprimary studentsintheir mother tongue; and 0 Self-contained classesingrades 1-4 (i.e., one teacher for all the core subjects). - 20 ,. 2.4 The objectives at each level of education were clarified as follows: primary education was to provide functional literacy; general secondary education in grades 9 and 10 was to prepare students for further education, including technical and vocational options; and preparatory secondary education (grades 11-12) was to groom students for studies in higher learning institutions (colleges and universities) or in technical and vocational fields. As figure 2.1 shows, students who finish grade 8 can compete for places in programs o f junior technical and vocational education and training (TVET); those who complete grade 10can vie for places in preparatory general education, teacher training institutes or colleges (TTIs or TTCs), or programs offering medium level TVET; and those who make it through grade 12 can compete for places indiploma- or undergraduate-level coursesinthe universities or inTVET program^.'^ Figure2.1. Structure ofthe EthiopianEducation Systemin2003-04 23+ 22 21 3rd 20 2nd 19 1st 18 12 17 11 16 10 15 9 14 8 13 7 12 6 11 5 10 4 9 3 8 2 7 1 4-6 KG I Kindergarten ITeacher General education TechnicaUvocational training education&training Note: The duration of postsecondarycourses mayvary by field of study. Source: MinistryofEducation, Govt. of Ethiopia. 2.5 While the length of the primary and secondary cycles has been stable since 1994, the duration o f some courses inpostsecondary education has been adjusted over the past few years. Programs in the TTIs and TTCs were recently redesigned to last one and three years, respectively, after grade 10; previously, TTI courses lasted three years after grade 10, and TTC l9 Graduates o f TTIs are officially certified to teach ingrades 1-4, and those o fTTCs, ingrades 5-8. - 21 - courses, one year after grade 12. Diploma-level courses in the universities used to last two to three years, but as o f 2003-04, were shortened to one year; and undergraduate-level courses that used to last three to four years inmost fields, shortenedto three years. As of the 2001-02 school year, entry to preparatory secondary education (grades 11 and 12) was sharply reduced, and repetitionpractically eliminated. Most of the grade 10 graduates are expected to pursue courses inTVET inlinewiththegovernment's decisionto expandthis subsector. Overview of EnrollmentTrends 2.6 The Ethiopian Ministry of Education maintains extensive records of enrollments dating back to 1967-68 for general education, 1985-86 for higher education, and 1995-96 for TVET and teacher training. This section documents the aggregate number of students over time, the share of enrollments innongovernment institutions, and the magnitude o f enrollments in evening classesand innonformalbasic education. 2.7 In 2002-03, the formal system catered to a total of more than nine million students, compared with just over half a million in 1967-68 (table 2.1). Between these years, enrollments increased 16-fold in primary education, and by 26 times in secondary education. Dramatic increases also occurred inthe other subsectors: the number o f studentsinTVET rose from fewer than 3,000 in 1995-96 to more than 54,000 by 2002-03; those in higher education rose fiom around 18,000 in 1990-91 to more than48,000 by 2001-02. Table 2.1. Enrollments by Level in Government and Nongovernment Institutions, - Eth pia, 1967-68 to 2001-02 ~ Technical and Vocational Teacher [igher educatior Primary Secondary Education and Training (Regular School Year Preprimary `Grades 1-8) (Grades 9-12) Training Institutes, wograms only)! 1967-68 - 496,334 26,690 - - - 1970-71 - 728,548 53,220 - - - 1975-76 - 1,226,124 90,09 1 - - - 1980-81 - 2,341,437 216,876 A - - 1985-86 - 2,811,910 292,385 - - 18,457 1990-91 - 2,871,325 453,985 - A 17,895 1995-96 - 3,787,919 402,753 2,738 5,900 17,378 2000-01 109,358 7,274,121 649,221 8,639 6,224 46,812 2001-02 118,986 7,982,760 684,630 38,176 6,080 48,143 2002-03 123,057 8,572,315 626,714 54,026b' 7,002 77,946 2003-04* - 9,343,428 725,059 - A 80,698 -Datanotavailableattimeofreport\i ing. Note: All data inthis table refer to-enrollmen%inregularprogramsonly. a. Includesstudents indiploma-level teacher training colleges, but excludespart-timestudents and those ineveningor summer (Kiremt)courses. b. Dataprobablyincompletebecauseof missinginformation from severalregions. c. Preliminarydata available as of September 2004. Source: Govt. of Ethiopia 1994 E.C., for 197G1993 data; Govt. of Ethiopia 1997, for 1995 data; Govt.of Ethiopia2001, for 2000 data; and Govt. of Ethiopia2002, for 2001data. - 22 - Primary andSecondaryEducation 2.8 The increase in enrollments since 1967-68 did not follow a smooth path. Instead there were o f periods o f expansion followed by reversals that generally coincided with political turbulence inthe country and severe famines (figure 2.2). Enrollments ingrades 1-12 grew at 12 percent a year during 1967-68 to 1976-77, a pace not significantly disturbed by the famine o f 1973-74 andthe overthrow o f Haile Selassie ina 1974 coup ledby Teferi Benti. In1977-78, the year that Mengistu Haile Mariam took power, enrollments at all levels fell by 2 percent, but the setback proved temporary. Enrollments grew by 20 percent the following year, by 28 percent the year afier that, and at the rate o f 13 percent a year until 1982-83 when the beginnings o f a severe famine caused enrollments to increase by only 6 percent from the previous year. In the subsequent two years, the country fell victim to one o f the worst famines in its history, and the number of children in school shrank in absolute terms. Although enrollments did recover by 1987-88 to the prefamine peak, the gains proved to be short-lived as the country plunged into the political turmoil that eventually culminated in the overthrow o f Mengistu's regime in 1991. Enrollments in 1992-93 were 31 percent smaller than they were in 1988-89 and about the same as at the start o f the 1980s. As the figure clearly demonstrates, the large swings in enrollments are due mainly to instability in the enrollments in grades 1-4. Under the new government, aggregate enrollments ingrades 1-12 have been growing at a sustained pace o f about 9 percent a year between 1992-93 and 2001-02-a remarkable achievement indeed given the history o f stagnation anduneven growth inthe past. Figure 2.2. Enrollments inGrades 1-12, Ethiopia, 1967-2002 7.0 1.o 0 Source: Govt. of Ethiopia 1994, for 1967-68 - 1992-93 data; Govt. of Ethiopia 1995, for 1993-94 data; Govt. of Ethiopia 1996, for 1994-95 data; G o a Note: Datainclude only students inregularprogramsingovemment andnongovemmentschools. of Ethiopia 1997, for 1995-96 data; Govt. of Ethiopia 1998, for 1996-97 data; Govt. of Ethiopia January 1999, for 1997-98 data; Govt. of Ethiopia August 1999, for 1998-99 data; Govt. of Ethiopia 2000, for 199940 data; Govt. of Ethiopia2001, for 2000-01 data; and Govt. of Ethiopia 2002, for 2001-02data. - 23 - PostsecondaryEducation 2.9 Enrollments include students in diploma-level programs as well as those in programs leading to undergraduate and postgraduatedegrees. Figure 2.3 shows the trends since 1985-86, the earliest year for which data are available. Because the private sector has historically been relatively undeveloped, the figure pertains only to students in government institutions. In the decade up to 1996-97, enrollments at all three levels of postsecondary studies had been somewhat stagnant, rising by only 13 percent among diploma-level students and by 11percent among undergraduates. The number of postgraduate students jumped nearly threefold over the same period, but the increasestarted from an extremely small base of only 279 students in 1985- 86. AAer 1996-97, enrollments at all levels o fpostsecondary studies climbedrapidly-by 2001- 02 the system was catering to more than 12,000 diploma-level students and more than 22,000 undergraduate students, compared with the corresponding enrollments of fewer than 6,000 and about 10,000 in 1995-96. At the postgraduate level, the historical increase in enrollments continued to be sustained, even though total enrollments remained relatively small in 2001-02, just over 1,300 students. Figure 2.3. EnrollmentsinRegular ProgramsinHigherEducation,Ethiopia, 1985-2002 25.0 * d 20.0 g 8 c y 15.0 -0 ti 2 k O 10.0 P ti iE 5.0 Note: Dataon enrollments inprivate institutionsand eveningclasses are excludedfor lack ofasufficiently longtime series. Source: Govt. of Ethiopia 1994, for 1985-86 - 1992-93 data; Govt. of Ethiopia 1995, for 1993-94 data; Govt. of Ethiopia 1996, for 1994-95 data; Govt. of Ethiopia 1997, for 199596data; Govt. of Ethiopia 1998, for 1996-97 data; Govt. of Ethiopia January 1999, for 1997-98 data; Govt. of Ethiopia August 1999, for 1998-99 data; Govt. of Ethiopia 2000, for 1999-2000 data; Govt. of Ethiopia2001, for 2000-01 data; and Govt. of Ethiopia2002, for 2001-02 data. EnrollmentsinGovernment andNongovernmentEstablishments 2.10 Schools in Ethiopia fall into two broad categories: government and nongovernment. Those in the first group charge little or no fees and they are managed and staffed by employees paid by the federal or regional governments. By contrast, institutions in the nongovernment - 24 - sector rely partly or wholly on nongovernment sources o f financing (e.g., contributions by communities andstudent fees), and they are managedprivately.20 2.11 Figure 2.4 shows trends since 1967-68 in the share of enrollments in nongovernment primary and secondary schools, and in nongovernment postsecondary institutions since 1999- 2000. Although nongovernment schools served three times as many students in grades 1-12 in 2000-01 as they did in 1967-68, their share o f total enrollments fell from 24 percent to only 4 percent over this period. In grades 1-8, the decline started in earnest around the early 1980s (following an explicit policy in 1975-76 to nationalize all schools). It was reversed for about five years between the late 1980s and the early 1 9 9 0 ~a~period of political turbulence in Ethiopia during which total enrollments fell in absolute terms. Since around 1991-93, the decline inthe nongovernment sector's share resumed, reflecting the impact o f the large-scale expansion of government schools underthe new government's first education sector development plan. Figure2.4. Shareof StudentsinNongovernmentEstablishments,Ethiopia,1967-68to 2001-02 35 ,Grades 1-4 30 25 i-' E 8 20 a 15 10 5 0 . . . . . . . . . . . . . . . . . . . . . . . I I , , , , Y , i l l l l l $r- & z m 2 -m ! I & Z PI \o r4 wl m 00 -3 00 m 00 m 10 g Rx` d ` 0 00 0 I 2 2 2 2 2 2 z 2 2 2 2 0 0 r 4 Note: Datafor highereducationavailableonly from 1999-2000 onward; enrollmentsat all levelsexcludestudents inevening, part- time, andsummer courses. Source:Govt. of Ethiopiavarious years, EducationStatisticsAnnual Abstract. 2.12 Ingrades 9-12, the nongovernment sector's share of enrollments had already started to decline in the 1960s. By 1975-76, barely 2 percent o f secondary students attended a nongovernment school, and the modest share has persisted to this day. In higher education, enrollments inthe private sector have grown very rapidly inrecent years: from 5,472 students or 15 percent of total enrollments in regular programs in 1999-2000, to 12,269 or 26 percent of total regular enrollments by 2001-2002. Nongovemment schools in Ethiopia include the following: (a) public schools, found mainly in Addis Ababa Municipality and OromiyaRegion; most of these schools startedoff as private schools but were taken over by the govemmentat various times and subsequentlydenationalized (in Addis Ababa, the schools depend to varying degrees on the govemment to pay the salaries of their teachers, while in Oromiya, public schools are financed by community contributions in the form of cash, materials, and labor); (b) community schools, schools that serve foreign communities and are managed by the relevant communities and financed by student fees; (c) mission schools, run by religious bodies and financed by school fees; and (d) private schools, operatedbyprivateentrepreneursthat rely entirely on studentfees to cover operatingcostsand generateaprofit for the owner. - 25 - 2.13 Taken together, the trends in primary and secondary education point unmistakably to sustained efforts by various governments to expand formal education through direct service delivery in the government sector. In higher education, the explosion in enrollments in private institutions suggests that the latent demand for services has always been strong and only with the easing o f constraints on private providers inrecent years is such demand being met. EveningClasses, DistanceEducation,andNonformalBasicEducation 2.14 Besides the regular daytime programs, a diversity o f other educational options are available. Most students pursuing these options either pay out-of-pocket to receive instruction or rely on nongovernment organizations for subsidized services.21 Below we describe the characteristics and size o f enrollments in evening classes, distance education, and nonformal basic child and adult education. The data imply that fee-paying students in Ethiopia make up a larger share o f the total enrollments than i s impliedby the data for daytime programs only, and the difference is especially large at the tertiary level. EveningClasses 2.15 These classes offer instruction at all levels but are available mainly in urban areas because electric power for lightingis still largely confined to towns and cities. Financed entirely by student fees, the classes may meet in government or nongovernment premises. Those using government facilities are typically staffed by teachers and other personnel with daytime jobs in the same while those meeting in nongovernment facilities presumably draw on more diverse sources for their staff. The curricula for evening courses prepare students for the same end-of-cycle examinations as students inregular programs; however, they are designed to cover the material over a longer period because students inevitably attend fewer hours o f instruction at each class than students inthe regular programs23.In 2001-02, some 162,000 students attended evening classes in grades 1-8, 80,000 in grades 9-12, and nearly 45,000 in higher education (table 2.2). About 37 percent o f the students in the primary grades attended classes in nongovernment facilities, but the share was only 3 percent at the secondary level and 15 percent inhigher education. Exceptions are students inteacher training courses offered inthe summer; such students are typically sponsoredby the regional governments. *'In2002-03, fees from students ranged between Birr 6 to15 per month in the primary grades and Birr 19 per month in the secondary grades. Fees in higher education vary across fields o f study. According to govemment stipulation, the school hosting the classes is entitled to receive 11 percent o f the fees collected from its evening students; the remaining 89 percent is used to pay the teachers. A parent-teacher committee associated with the day programo f the school decides on how this income is used. 23 The curriculum in grades 1-8 is taught at the same pace as inregular daytime classes, but those for grades 9-10 and grades 11- 12 cover the material over a period of three years instead o f two years inthe regular classes. Inhigher education, the duration of studies i s also longer but its precise length varies across fields and by level o f study, On average, students in the evening and summer courses take 1.5 times as long to complete their studies as daytime students. - 26 - Table 2.2. Number of Students by Type and Level of Education,Ethiopia, 2001-02 Primary cycle Secondary cycle Higher Adult Type of student kades 1-4 Grades 5-8 Total (grades 9-12) educationa' education Regular daytime students Total 5,725,954 2,256,806 7,982,760 684,630 48,143 0 Innongovernment establishments 229,265 140,050 369,315 15,667 12,619 0 Nongovernment share ("A) 4 6 5 2 26 n.a. Evening students Total 87,863 73,714 161,577 80,011 44,567 0 Innongovernment establishments 33,806 26,145 59,951 2,695 6,472 0 Nongovernment share (%) 38 35 37 3 15 n.a. Students indistance education 0 0 0 1,080 - 0 Students inKiremt coursesb` -500,000 0 0 0 0 9,119 0 Students innonformal education` 0 -500,000 0 0 -500,000 Grand total (thousands) 6,314 2,331 8,644 766 102 -500 Memorandad/: Percentage o f students inregular daytime classes 91 97 92 89-93 47-57 n.a. Percentage of fee-paying students Percentage of students not financed by governmente` -No data available; n..a.not applicable. a. Includesdiploma-levelteacher training. b. Refersto summer coursesmainly for trainee teachers seekingadiploma-levelcertification. c. Participants innonformaleducationare dividedintotwo groups, those aged7-14 andthose aged 15 and older. The younger group follows acurriculumthat allowstransfers to the formal systemandis thereforeclassifiedhereundertheprimarycycleingrades 1 4 . Participantsinthe older group are groupedunder adulteducationto correspondto the intentandcontent of the courses. The total is a roughestimatebasedondatabetween 1997-98 and2001-02; see appendix table A2.2 for additionaldetails. d. The rangesshown for secondary andhighereducationcorrespondto altemativeassumptionsinaddingup enrollmentsinthe various types of programs.For the first figure in eachrange, enrollments are computedby simple additionacrossall types ofprograms; for the secondfigure, they are added after converting enrollments inevening, Kiremt,anddistancecourses to regular daytimeequivalentsonthe assumptionthat students insuchprogramstake 1.5 times as longto completetheir studiesas regular daytime students.Fornonformalbasiceducation,a similar calculationis notattemptedbecausewhile pupils inthe well- organizedprogramscompletetheir studiesinthree insteadof the four years informalprimaryschools, the arrangementis notuniformacrossthe country, and pupilsinsomeof the less well-organizedprogramsmay infact take longeror even fail ever to reachgradeequivalencywiththe formalprogram. e. Includesfee-paying students. Source: Govt. of Ethiopia2002a; supplementedby IIDDW-ETHvarious issues, for data onnonformaleducation; and Govt.of Ethiopia1994 E.C., for data on distanceeducation. DistanceEducation 2.16 Distance programs at the secondary level have been offered by the Educational Media Agency since 1971, but they reach relatively few people-around 1,000 students in2001-02. In higher education, distance courses have also become popular in recent years, but the numbers enrolled are not known. Kiremt courses offered by universities and colleges are yet another option. Their main clientele were teachers, numberingjust over 9,000 in 2001-02, who attended undersponsorshipbyregional governments to upgrade their formal qualification. - 27 - NonformalEducation 2.17 This option, sometimes referred to as Altemative Basic Education, is especially important in rural communities. Available to both adults and children and financed largely by nongovernment organizations, including international ones (e.g., Action Aid and Save the Children) or local agencies dependent on foreign sources o f funding (e.g., Rift Valley Children and Women Development Association), nonformal education programs may be managed directly by the funding organization or in collaboration with the regional governments. While not all nonformal education programs share a common curriculum across localities, they have a common aim: imparting basic literacy and numeracy. Adults in these programs often benefit from job-related skills training as well. The programs for children (ie., clients below age 15) typically offer three levels o f instruction that, in aggregate, match the first cycle o f primary schooling (grades 1-4). Focusing mainly on the core subjects o f languages (typically the local language and English) and mathematics, the programs are staffed by teachers living in the community who may or may not be paid.24 Classes meet in structures built by the villagers or donated by a local benefactor, and the school calendar and hours o f instruction are deliberately arranged to free children for work on the family farm or at home. At their best, the programs succeed in qualifying their pupils for transfer to the formal schools at the fifth grade after only three years o f instruction (Govt. o f Ethiopia 2002). Performance is not uniformly good, however, because o f finding and other constraints. Indeed, in the most poor1 funded centers, the programs offer what can at best be described only as informal instruction? Inthe late 1990s and early 2000s, nonformal education served an estimated 1 million people, more or less evenly divided between adults andchildrenbelow the age o f 15. Summaryand Cross-CountryComparisons 2.18 The information presented above describes an education system with diverse, if uneven, options. About 90 percent o f primary and secondary students attend regular daytime programs at present; the share inhigher education is much smaller, between 47 and 57 percent. Because the government finances only the regular daytime programs in government schools, it is easy to deduce the share o f students who either pay fees for their schooling or depend on nongovernment organizations to defray the costs. In primary education, the share i s currently 12 percent (counting pupils innonformal education); the pupils are evenly split betweenthose who pay out- of-pocket and those who rely on nongovernment organizations. Insecondary education, the share o f fee-paying students rises to between 9 and 13 percent, depending on how enrollments in nonregular programs are treated. Inhigher education, such students account for between 50 and 56 percent o f total enrollments. At all levels, the share o f students who rely on sources o f financing other than the government is thus much higher than what would have been implied by data for the regular daytime programs only. Also noteworthy is that in secondary and higher education, most o f the fee-paying students attend evening classes taught by teachers from the 24 InOromiya Region, unpaid volunteers earn five points for each year of service in nonformal programs. These points count toward their score on a set of criteria the regional government uses to evaluate applicants for the coveted places in teacher trainingprograms; the programsqualify successful candidates for teachingpositionsingovernmentprimary schools. 25 Inone center the authors of this study visited, instructionwas offered in aroommade availableby a residentbut the class was cancelledfor the day becausethe owner forgot to unlock the door. Inanother center, the teacher was a boy from the village who was attendingeighthgrade in aformal primary schoolsomedistancefromthe village. - 28 - daytime programs. In some ways, persistent constraints on private education in the past have simply given rise to what amounts to a private system within the public sector. 2.19 Figure 2.5 puts into cross-country perspective the extent o f privately financed education inEthiopia at the primary and secondary levels.26At 12percent (with enrollments innonformal basic education included), Ethiopia's share at the primary level exceeds that in most o f the countries at a comparable level o f per capita GDP inthe figure. At the secondary level, the share i s similar to that in a cluster of countries at Ethiopia's level of per capita GDP (Benin, Chad, Malawi, Mali, Mozambique, and Niger), but it i s much smaller than the shares o f about 50 percent inMadagascar, Rwanda, and Tanzania, and 30 percent in Burkina Faso-countries with per capita GDP inthe $500 to $1,200 range (inPPP dollar Figure2.5. Share of Fee-Paying or Nongovernment-FinancedStudents inPrimary and SecondaryEducation, Ethiopia and Other Countries, circa 2000 Rinerylevel secadarylevel A 'I A A A 40 A A A A O A ''1 A A ettdmiab A A A ~ A A 2.20 For higher education, the available data-presented in figure 2.6-pertain mainly to countries at higher levels ofper capita GDP than Ethiopia. What they reveal is the great diversity across both middle-income and OECD countries in the role o f privately financed tertiary education. Among middle-income countries, the share of fee-paying students ranges from 5 percent in the Russian Federation, to 85 percent in Indonesia, and 74 percent in the Philippines; andamongOECD countries, from 1percent inSweden, to 80 percent inKorea, and79 percent in ~~~~ 26 For primary education, the data for countries other than Ethiopiacome from Bruns, Mingat, and Rakotomalala2003, inwhich privately financed students exclude those who attend private schools subsidized by the government. The data for secondary education come from the UNESCO Institute of Statistics (UIS), which unfortunately makes no distinction between students in self-financingprivate schools andthose inprivateschools subsidizedby the government. "For agroup of 18middle-incomenon-OECDcountries participatingin UNESCO's World EducationIndicators(WEI) Programme, the share of students in independentprivate institutionsin 2002 averaged 9.5 percent at the lower secondary level, and 13.0percentat theupper secondarylevel. Ethiopia's overallaverage for secondaryeducationis comparableto these shares. - 29 - Japan2*The corresponding share inEthiopia in 2001-02 was between 50 and 56 percent, which is higher than inmost countries inthe figure. Nevertheless, it could go higher by the standard in such countries as Brazil, Chile, Indonesia, Japan, the Philippines, and Republic o f Korea- countries where basic education has historically received top priority in the allocation of public expenditure on education. Figure 2.6. Share of PrivatelyFinancedTertiary Students,Ethiopia, 2001, and Other Countries, 1999 Ethiopia Rwanda Madagascar Indonesia Philipirazil ines Chile .MalaPeru sia Thairand Ar entina duguay Russian Federation KoreaJapan Rep Portugal UnitedMexico States Poland Italy Spain France Switzerland Ireland Sweden 0 20 40 60 80 100 Percent Note: The middle block includes countriesparticipating in UNESCO's World EducationIndicators(WEI) Programme in which detailed data are collected on education finance; the second block includes OECD countries for which the relevant data are available. Source: For Ethiopia, table 2.2; for Madagascar, World Bank 2002; for Rwanda, World Bank 2003b; for the other countries, UIS/OECD2001. Trends andLevels inthe Gross EnrollmentRatio 2.21 This section documents the coverage o f the system as depicted by the gross enrollment ratio (GER), which i s defined as the ratio between all students enrolled in a given cycle and the population inthe official age range for that cycle.29While flawed as a measure o f coverage, this statistic is easily computed, and data are available for a large number o f countries.30 Other 28 According to UNESCO-UISIOECD (2001) the precise definition is the share o f students in independent private institutions. Because fees are the main source o f income in most o f these institutions, we equate this statistic with the share o f fee-paying students. 29 InEthiopia the age range for primary schooling is 7-14 years and for secondary schooling, 15-18 years. For higher education, the standard practice is to define the age range as the five-year bracket following the last year o f the range for secondary schooling (ages 19-23 inthe case o f Ethiopia). 30 The GER is often supplemented bythe net enrollment ratio (NER), which is computed inthe same way as the GER, except that the numerator includes only students in the official age range for the cycle o f schooling concemed. Both measures are flawed because they refer to the volume of enrollments (calibrated against population size) rather than to the dynamics of student flow. Because o f the way it i s defined, the NER is especially problematic as a measure o f coverage inpostprimary cycles o f education. This i s because the cumulative effects o f late entry to first grade, grade repetition, and interrupted studies inevitably mean that - 3 0 - indicators o f coverage that more closely reflect the dynamics of student flow will bepresentedin a later section o fthe chapter. PrimaryEducation 2.22 Table 2.3 shows trends in the GER computed from two sources o f data: (a) the administrative statistics collected by the Ministry of Education through its annual school census and population projections supplied by the Central Statistics Authority; and (b) household surveys. The survey-based GERs are generally higher, particularly in the earlier years, but the discrepancy i s not large. Looking over time, we see that the rapid increase in enrollments between 1993-94 and 2001-02 i s reflected ina tripling o f the GER, from 20 to 62 percent. This remarkable achievement notwithstanding, Ethiopia's GER remains modest in cross-country comparison. This is evidenced in figure 2.7 where the statistic for Ethiopia has been reestimated for six years o f schooling to match the corresponding statistic for all the other countries shown in the figure. The adjusted GER o f 68 percent puts Ethiopia only slightly ahead o f such countries as the Democratic Republic o f the Congo, Guinea, Sierra Leone, and Sudan. Because the GER provides a measure of the availability o f school places relative to the potential demand for schooling, the result for Ethiopia implies that even in the absence of repeaters who inevitably take up extra space, new classrooms will continue to be needed in the foreseeable future as the country strives to achieve universalprimary school c~mpletion.~' Table 2.3. PrimaryGrossEnrollmentRatios(GER), Ethiopia, 1993-94 to 2001-02 Basedon schoolcensus and Basedon householdsurveys governmentpopulation estimates a/ I GEREstimate I Survey and data collectionperiodb/ I 20.5 1994-95 26.2 - 1995-96 30.1 34.0 WMS, Jan./Feb. 1996 1996-97 34.7 - 1997-98 41.6 45.8 WMS, MarcWApril 1998 ~ 1998-99 45.8 41.3 LFS,May 1999 1999-00 51.0 53.9 WMS, Jan./Feb. 2000 51.6 DHS,Feb./May 2000 2000-0 1 57.4 - 2001-02 61.6 - significant numbers of students must be excluded in computing the NER; this results in systematic underestimationof the true reachofthe educationsystem. 31As will becomeclear in chapter 5, managementof the country's currently high construction costs is an important concern in light ofthe magnitudeof thepotentialdemandfor new classrooms. - 31 - Figure2.7. Primary GrossEnrollmentRatio (GER), Ethiopiaand Other Countries,circa 2000 Indonesia Honduras Uganda Rwanda Nicaragua The Gambia Nigeria Congo, Rep. Angola Chad Guinea-Bissau Senegal Ethiopia Sierra Leone Guinea Sudan Congo, Dem. Rep. C.A.R. 0 20 40 Percent60 80 100 C.A.R., Central African Republic; Congo, Dem.Rep., Democratic Republic o f the Congo; Congo, Republic of Congo. Note: The GER has been standardized for six years of primary schooling for a l l countries in this figure. Source: For Ethiopia, computed from Ministry of Education data on enrollments in grades 1-6 and on the population ages 7-12; for other countries, Mingat, Rakotomalala, and Tan 2002. Secondary Education 2.23 Incontrast to the sharp rise inthe primaryGERbetween 1993-94 and2001-02, the ratio for secondary education rose only marginally duringthis period, from 8 percent to 12 percent (or 13 percent if evening students are included; nearly 14 percent if students in technical and vocational education and training are included). For cross-country comparisons, Ethiopia's statistic again needs to be adjusted because secondary schooling lasts only four years in the country compared with the more common structure of 6-7 years in most other countries. In figure 2.8, the GER for Ethiopia is computed for grades 7-12. At nearly 17 percent, the adjusted GERis smaller thanthe averageof24 percent for Sub-SaharanAfrica, and far below the average of 56 percent indeveloping countries as a whole. Still, Ethiopia's figure is more thanthree times as high as Tanzania's 5 percent, and somewhat greater than the ratios for Guinea, Madagascar, Mali, andRwanda-all countries with higher levels ofper capita GDP in2000. - 32 - Figure 2.8. Secondary Gross EnrollmentRatio (GER), Ethiopia, SelectedAfrican Countries and Country 1 Group Averages, circa 2000 Sub-SaharanAfrica Developingcountries Zimbabwe Ghana Togo Kenya Zambia CBte d'Ivoire Senegal Congo, Dem.Rep. Ethiopia Mali Madagascar Guinea Rwanda Tanzania 0 10 20 30 40 50 60 Percent Congo, Dem.Rep. DemocraticRepublicof the Congo. Note: Country group averagesarepopulationweighted; to improvecross-countrycomparability,the ratio for Ethiopiarefers to grades 7-12. Source: For Ethiopia, computedby authorsfrom Ministry of Educationdataon enrollments andpopulation, ages 13-18; for the other countriesand countrygroups, UNESCO 2002. Higher Education 2.24 At this level of education, the discrepancy between the ages of the enrolled and reference populations tends to be even wider than at the lower levels of schooling.32Thus, in addition to the GER, some analysts also rely on the number of students per 100,000 populationas a measure o f coverage at this level o f education. 2.25 We examine first the GER. It rose from 0.5 percent in 1993-94 to 1.7 percent in 2001- 02, reflecting the sharp rise inenrollments after the mid-1990s. The country's ratio nonetheless remains one o f the lowest worldwide, ranking ahead of only a few low-income countries that include Angola, Chad, Mozambique, and Tanzania. The picture remains unchanged when the other statistic-higher education students per 100,000 inhabitants-is used (figure 2.9). Ethiopia's statistic in 2001-02 i s estimated at between 125 and 150 students per 100,000 inhabitants-clearly below the nearly 200 students per 100,000 predicted from regression analysis for a country at a similar level of gross domestic product per capita. 32The gap between the two populations reflects the effect not only o f late entry and grade repetitionearlier inthe system, but also the likelihood that many secondary school graduates will work for a few years before enrolling in higher education. - 3 3 - Figure 2.9. Relation between GDP Per Capita and Coverage inHigher Education inLow-Income Countries, circa 2000 A A 0 A A A A A A A AA A 100 I 500 I 1,do0 2,dOO 3,000I Per capita GDP (2000 PPP dollars) Note: GDP, gross domestic product; PPP, purchasingpower parity; datafor Ethiopia refer to 2001-02 andincludestudents ineveningandKiremt courses, adjustedto MI-time equivalents,assumingthat such students take 1.5 times as longto completetheir studies as those inregular programs(see table 2.2 for additionaldetails). Source: Ethiopia highereducationenrollment data, EducationManagementInformation System(EMIS) Panelofthe Ministry of Education; all other data, World Bank SIMA database. 2.26 It is important to caution that even though these cross-country comparisons seem to suggest a possible need for catch-up expansion o f higher education, a convincing justification must rest on direct evidence o f market demand. At this level o f education, such evidence needs to confirm that graduates are indeed landing jobs appropriate to their training and are also earninggood returns on their education. Just as important is the fact that even if expansionwere justified on these terms, there is no prima facie case for expansion through the public sector alone. Inlight of the government's fiscal constraints, a continued large role for the private sector, encouraged by the government in recent years, may be the only fiscally viable and sustainable way to expand coverage while also ensuring responsiveness to the demand for highly educated workers. StudentFlow in PrimaryandSecondaryEducation 2.27 Because schooling requires a multiyear commitment in order to yield the hoped-for results, it is important to go beyond the gross enrollment ratio to examine the dynamics o f student flow, as documented by such indicators as rates o f entry, survival, completion, repetition, and transition between cycles. At the primary level, where the curriculum typically aims to impartbasic literacy andnumeracy skills, children who never enroll are unlikely to acquire these skills, and those who enroll but drop out prematurely are also unlikely to become permanently literate and numerate. At the postprimary levels, access and survival to the end o f the cycle remain important, not least because both are prerequisites for entry to subsequent levels. The cumcula likewise offer training inan integrated set o f skills, so that premature exit would imply - 34 - a frustrated investment.To set the stage for our discussion on student flow, we beginbelow with a briefpresentationo fthe link betweenschooling attainment andadult literacy. Years of SchoolingCompletedandAdult Literacy 2.28 Data from a 2000 household survey confirm the expected positive relation between educational attainment and literacy: the probability o f being literate, objectively measured by a demonstrated ability to read a written sentence inthe respondents' own language, i s 100 percent for those with seven or eight ears of primary schooling and falls to 80 percent for those with only four years (figure 2.10).% While these results compare well with those o f other African countries (see appendix figure A2.1), it is to be recalled that they pertain to adults who were in school when the systemwas much smaller (Le., whenit served the country's elite) andclassroom conditions better (e.g., the pupil-teacher ratio averaged 32 in 1993-94, compared with 65 in 2001-02). The view that performance hasprobably deteriorated is consistent with the results of a recent small-scale assessment o f student leaming, which showed that fourth-graders across the country correctly answered fewer than half of the items on which they were tested (Keder et al. 2001; Hassenet al. 2001b). 2.29 The foregoing results have important implications for how we examine the pattem of student flow. Although the country's primary cycle currently lasts eight years, divided into two subcycles o f equal duration, it would appear that completion o fthe first cycle maybe insufficient to ensure that most children achieve permanent literacy (given the probable deterioration in schooling conditions discussed earlier). However, completion of eight years may go beyondwhat i s minimally required. A noteworthy fact i s that Ethiopia i s one o f only 47 among 194 developing countries worldwide in which primary schooling lasts more than six years. As countries develop economically, many do indeed set themselves the goal o f ensuring that all childrencomplete eight or nineyears o f basic education (usually combining what are labeled the primary andlower secondarycycles). Such a goal is undoubtedly also appropriate for Ethiopiain the long run, but in the short term, resource constraints are likely to present difficult tradeoffs between the duration o f basic education and the number o f childrenwho can be accommodated. Assessing the fiscal implications o f the current eight-year cycle compared with, say, a five- or six-year cycle, will therefore be important ininformingfuture policy development. 33 A unique feature of the Ethiopian data is that it contained both a subjective and an objective measure o f literacy. As figure 2.10 shows, the subjectivemeasureyields amoreoptimistic result, reflectingan understandablebiasin all self-evaluation. - 35 - Figure 2.10. PercentageLiterateamongAdults Ages 15-49,by HighestGrade Attended, Ethiopia, 2000 100 - 80 - \Based on demonstrated + 8 60 - ability to read sentencein 40 - I I I I I I I I I 0 1 2 3 4 5 6 I 8 Highestgrade attended Source: Estimated from the 2000 EthiopiaDemographicandHealthSurvey. Entryto Grade 1 2.30 Growth in the number o f children entering grade 1 was the main reason for the spectacular gains inprimary school enrollments since 1993-94. The gains were largest between 1994-95 and 1996-97: each school year saw the number o f new entrants exceed those in the previous year by some 20, 26, and 15 percent, respectively (figure 2.1l), by about 200,000 or children in absolute terms. The number of new entrants has continued to grow since 1996-97, but at the slower paceofjust 4 percent eachyear. Figure 2.11. New Entrantsto Grade 1andPercentageIncreasefromPreviousYear, Ethiopia, 1993-94 to 2000-01 5 1,200 - Note: Includesonlyregularpupils. Source: New entrantsare computedby subtractingrepeaters from total enrollments in Grade 1, data for which are reportedin Govt. of Ethiopiavarious years, EducationStatisticsAnnualAbstracts. 2.31 One way to put the growth in enrollments in perspective i s to relate the number o f new entrants to the 7-year-old population, 7 being the official age o f entry to grade 1. This indicator-called the apparent intake rate-climbed from 45 percent in 1993-94 to nearly 100 percent by 2000-01 (table 2.4). While the increase is undoubtedly spectacular, it by no means - 36 - implies that all children are now entering school. In a system such as Ethiopia's where rapid growth had been occurring since only 1993-94, new entrants to grade 1 typically included children who were entering on time, as well as older children who would have started at an earlier age had there been a school in the locality at the time they were ready to start. In other words, the numerator inthe apparent intake rate contains a cross-sectiono f children from several cohorts. Thus, when it i s related to the population in a single age cohort, the result tends to overstate the share o f childrenina cohort who would ever enter grade 1. Table 2.4. Three Indicators of Entry to Grade 1,Ethiopia, 1993-94 to 2000-01 Percentage of Apparent intake Cohort entry rate Ratio of AIR to children ever Cear rate (AIR, %) a' (CER, %) b/ CER enrolled by age 7 1993-94 45.2 - - - 1994-95 55.8 - - - 1995-96 71.3 37.2 1.9 12.6 1996-97 80.6 - - - 1997-98 90.5 44.6 2.0 17.3 1998-99 92.9 54.5 1.7 21.9 1999-00 52.4 1.8 20.6 I 95.7 99.5 60.5 1.6 - -Datanotavailable. a. Defined as the number of new entrants to grade 1, regardless of age, relative to the populationof 7-year-olds; data exclude evening students. b. Refers to the estimatedshare ina cohort who ever entered school. The rates are estimatedfrom data for sampled youths ages 7-18 inthe Welfare MonitoringSurveys of 1996,1998, and 2000; youths ages 7-14 inthe 1999Labor ForceSurvey; and youths ages 7-17 inthe 2001 Child Labor ForceSurvey.Because the percentage ever enrolled tends to peakby ages 13-14, itprovides a reasonablygood estimateof the share of childrenin a cohort who would ever enroll. For 1995-96 and 1997-98, the peak occurs slightly later, so the data shown in the table correspondto the percentage ever enrolled by ages 15-16 and 14-15, respectively (see tableA2.3). Source: For the apparent intakerate: computedfrom data on enrollments and repeatersreported in Govt. of Ethiopia 1995, 1996, 1997, 1998, 1999,2000, and 2001, and projections of the 7-year-old populationpreparedby the Central Statistical Authority of Ethiopia; for percentage ever enrolled in a cohort of children: computed from the Welfare Monitoring Surveys of 1996, 1998, and 2000, and the 1999Labor Force Survey, and for 200&01, from table 3.12 in Govt.of Ethiopia 2002, which is based on the Child Labor ForceSurvey fielded inMarch2001. 2.32 A better estimate of the cohort rate of entry to grade 1is the share o f children who have ever been to school by a certain age. Typically, the share rises with age and then flattens out or declines slightly after age 13 or so (see table A2.3 for the pattem in Ethiopia). This pattem impliesthat children who have neverbeento school by age 13 are unlikelyever to do so. We can therefore use the share o f children who have ever enteredby this age as an estimate o f the cohort entry rate to grade 1.34 The results, basedon data from several national household surveys, show a steady rise inthe cohort entry rate, from 37.2 percent in 1995-96, to 60.5 percent in2000-01. The gain i s indeed very substantial, particularly because it occurred over a relatively short period; yet the fact remains that entry to grade 1i s still far from universal, and is muchbelow the average of 75 percent among subsaharan African countries.35By comparing the apparent and 34In Ethiopia, the share of children who have ever entered grade 1 peaked at slightly older ages in 1996 than in 2000. This pattem implies that in the earlier years, schools were indeed catering to a backlog of demand among older children that is now being satisfied at younger ages in the expanded system. Because we are interested in cohort entry rates, we used the peak rate in the age range 13-17, without necessarily choosing a specific age within the range. 35The average is for 18 low-income countries. - 37 - cohort entry rates to grade 1, we clearly see that the entrants to grade 1 over the past few years consist o f a large number of older children who were responding to newly available schooling opportunities. As the backlog of such children i s cleared, this multiple cohort effect will inevitably dissipate. Consistent with this expectation, the share o f children who enter grade 1by age 7-the official (and probably desired) age o f entry-has been rising steadily; it nonetheless remains modest, just over 20 percent. SurvivalRates 2.33 Cohort survival rates are conceptually easy to understand:they are the share o f entrants to a given cycle o f education who eventually reach a subsequent grade. Inthe 4-4-2-2 structure o f Ethiopia's education system, the relevant survival rates are between grades 1 and 4, between grades 5 and 8, between grades 9 and grade 10, and between grades 11 and grade12. Survival rates between grades 1 and 8 and between grades 1 and 12 provide additional perspective, while those between grades 1 and 5 have become increasingly relevant in the context o f the Millennium Development Goals. Because Ethiopia's education system is at an early stage o f development, survival rates in the primary grades are o f particular interest, and the discussion below will focus on them. 2.34 Data and estimation methods. To clarif the discussion, it is useful to begin with a brief comment on how survival rates are estimated. 32One method involves the use o f longitudinal data to trace students' schooling career over time. It is not often used, however, because: (a) such data are typically unavailable; and (b) the results are dated because they refer to students who have already completed their schooling up to a given grade, rather than to those currently going through the system. Instead, survival rates are more usually estimated using cross-sectional data. Below we present the results based on three methods o f estimation. All o f them are approximations of the phenomenon we wish to document, but the composite method would appear to provide the best compromise for our needs-it combines ease o f computation, lightness of assumptions, and an ability to generate real-time estimates. 2.35 Estimates for pseudo-cohorts. Assuming stable student flow patterns over time, this method calculates the survival rate between two grades, say, grades 1 and 4, as the ratio o f the number o f nonrepeaters ingrade 4 inyear Y to the number o f new entrants who started in grade 1three years earlier, i.e., inyear Y-3. Since the data pertain to assumed rather thanreal cohorts, the method yields what we might call pseudo-cohort survival rates. The results for cohorts starting their schooling careers in 1993-94 to 1998-99 appear in figure 2.12. Duringthis period, survival rates improved slightly between grades 1 and 4 and between grades 9 and 12, but deteriorated between grades 5 and 8.37These patterns reflect the combined effect of underlying influences andtheir evolution: (a) the perceived benefits o f schooling; (b) the age composition o f the students (because of its implications for the opportunity costs o f attending school); and (c) the quality o f school services. The first factor appears to have counteracted the negative impact associated with the massive expansion of the system since 1993-94 and the unambiguous deterioration in services (as reflected, for example, in the rapidly rising pupil-teacher ratios 36A more detailedexplanationcanbe foundat the endofthis report inthe TechnicalNote 2 on estimatingstudent flow profiles. 37Intheorywe can applythe methodto computepseudo-cohort survivalratesbetweengrades 1and 8, for example, butthe latest year for which we couldcompute estimateswould be 1994-95, seven years earlier than the latest year for which datawere available. -38- duringthe period), as well as that associatedwith the large numbersof overage students among entrants in the early years. These patterns are an indirect confirmation that the demand for schooling during the past decade had indeedbeenreasonably strong. Nonetheless, a survival rate of barely 60 percent in the first cycle of primary schooling remains low. Efforts will obviously be needed to continue boosting the demand for schooling, and this will probably require a combination of interventions to lower the cost of schooling to families andto improve the quality o f services. Figure 2.12. Pseudo-Cohort Survival Rates in Primary and Secondary Education, Ethiopia, 1993-94 to 1998-99 100% I 80% 5 h -.E5 3 60% E m 40% 20% z 0% q In \o p. 0, m 01 01 m pi 01 01 00 2 2f 01 v) 2 0, 2 f f Source Technicalnotetable TZ 3 2.36 Notice that because o f data truncation, the survival rates pertain to children who started their schooling at least three years prior to the last year for which data are available. Inacountry where enrollments have been growing very rapidly, this drawback is unfortunate because it provides a less than up-to-date picture o f the situation. For this reason, we turn below to the results from two other methods o f estimation, which overcome this handicap. 2.37 Estimates for reconstructed and comDosite cohorts. Again, the details o f these methods can be found in the technical note 2. Briefly, the reconstructed cohort method employs grade- wise enrollment data for two consecutive years and repeaterdata for the later o f the two years to compute promotion, repetition, and dropout rates inthe first year. These rates are then applied to a fictive cohort of 1,000 students to simulate their grade-to-grade flow through the system. The composite cohort method uses the same data but does the calculation in two steps, first computing survival rates between adjacent grades by relating the number o f nonrepeaters in grade X inyear Y, to those in grade X-1 the year before; and then computing survival rates to a given grade by multiplyingthe relevant adjacent grade-to-grade survival rates. Effectively, this method provides a simulation o f the survival rate in a hypothetical cohort that exhibits the same grade-to-grade transition patterns as currently observed in a cross-section o f children in the various grades. 2.38 The results basedon these two methods are shown intable 2.5. The reconstructedmethod yields higher survival rates than the composite method, and the gap widens as the number o f - 39 - grades for which they are computed rises. This is not surprising since the former method involves repeating the algorithm o f repetition and promotion in each grade for as many times as it takes for all members o f the cohort to exit the grade, either through promotion to the next grade or through dropping out. The Ministry o f Education makes a modified calculation inwhich the repetition algorithm i s processed only twice per grade (to reflect the fact that under current rules, pupils are allowed to re eat a grade no more than twice); and as the table shows, the result i s to reduce the survival rate?' Insome ways, the calculation can be described as a simulation of survival rates under certain assumptions about repetition anddropout behavior; ifpupils repeated fewer than twice before dropping out, the simulated survival rate would obviously fall further. The bootstrapping procedure is also a simulation; however, the grade-to-grade repetition and dropout rates are not assumed but simply incorporated as an intrinsic feature in the calculation. For this reason, and because the calculations are less tedious than those involved in the reconstructed cohort method, the discussion below and in subsequent chapters relies on survival rates computed using this method. Table 2.5. Cohort SurvivalRatesUsingDifferentEstimationMethods,Ethiopia,2000-01 Method Grades 1to 4, Grades,lto 5 Grades 1to 8 Classical reconstructed cohort method"' 0.64 0.60 0.47 Composite cohort method 0.60 0.53 0.36 Memorandum: Modified reconstructed method withrepetition algorithm repeated only twice per gradeb' 0.63 0.57 0.40 2.39 Table 2.6 compares the composite cohort survival rates for students enrolled in 1993-94, and circa 2000-01. The data show an unambiguous rise in the share o f first-graders who complete the first four years o f primary schooling, but the trends are less clear-cut regarding the share reaching subsequent grades, especially from grade 6 onward. Following a change inpolicy restricting access to grades 11 and 12, the share o f students in 2001-02 reaching grade 12 has fallen to only 8 percent-a sharp decline indeed compared with the corresponding figure o f 15 percent among students enrolled in 1993-94. In cross-country comparison, Ethiopia's cohort survival rate to grade 6 lags behind those o f other countries: 45 percent in 2001-02 compared with an average o f61percent circa 2000 among 20 low-income subsaharan countries for which a composite cohort survival rate could be computed. 38The calculationpreparedbythe MinistryofEducationis reportedinits Annual StatisticalAbstract. - 40 - Table 2.6. Composite Cohort Survival Rates and Intercycle Transition Rates, Ethiopia, 1993-94 and circa 2000-01 ndicator 1993-94 1999-00 2000-01 2001-02 2ohort survival rates from grade 1 To grade 4 0.52 0.60 0.61 0.59 To grade 5 0.49 0.53 0.54 0.5 1 To grade 6 0.48 0.47 0.49 0.45 To grade 8 0.39 0.36 0.40 0.36 To grade 10 0.23 0.20 0.3 1 0.27 To grade 12 0.15 0.19 0.13 ntercycle cohort transition rates grades 4 & 5 -aades 0.94 0.88 0.90 0.88 8 & 9 0.84 0.89 0.98 0.91 grades 10 & 11 0.86 --a/ 0.34 Source: See technical note table T2.5. 2.40 The table also includes transition rates at three points inthe system, corresponding to the system's 4-4-2-2 structure. There is no formal selection mechanism between grades 4 and 5, but students must pass a regional examination in grade 8 in order to reach grade 9, and a national examination ingrade 10 inorder to go on to grade 11.Until 1999-00, about 90 percent of tenth- graders proceeded to the next grade, but under a new policy introduced in 2000-01 to regulate student flow, only about a third o f them do so now. At the other two selection points, the transition rates remain very high at around 90 percent. These patterns suggest that the bulk of selection up to grade 10 is mediated not by any formal selection mechanism, but by a steady loss o f students through dropping out. CompletionRates 2.41 As a concept, completion rates are again easy to understand: they measurethe proportion o f children who attain schooling o f a given grade. Estimatingthem is another matter, however. To maximize sharedunderstandingof this important indicator andto facilitate discussion on how best to monitor performance over time, we elaborate on two methods below, and suggest that the cohort method is probablymore suited to the purpose at hand. '1 2.42 Cross-sectional estimates. The simplest calculation i s to divide the number o f nonrepeaters at the grade of interest, by the population o f the corresponding age for that grade.39 39Note that this calculation tends to underestimate the completion rate since the numerator includes students who may drop out inthe course ofthe school year. The fact that Ethiopia's school census is conducted only one monthafter the start ofthe school year tends to increase the risk o f overestimation. The altemative o f usingdata for the next grade to make the calculation i s also unsatisfactory: it is likelyto underestimate completion rates to the extent that some of the students who complete a given grade may not go on to the next grade. Since the latter calculation does not offer a clear-cut advantage and may be confusingbecause it - 4 1 - Because o f its simplicity and the relative ease o f assembling the required data, this calculation has been made for a large number o f countries in the context o f background documents for the Education For All Fast Track Initiative (see, e.g., Mingat, Rakotomalala, and Tan 2002; and Bruns, Mingat, and Rakotomalala 2003). By this method, Ethiopia's grade 6 completion rate is estimated at 24 percent in 1999-00 (the figure cited inthe foregoing publications) and 34 percent in 2001-02.40 By contrast, the average primary completion rate was 58 percent for 54 low- income countries (excluding Ethiopia) around 2000. 2.43 While the above statistic facilitates comparisons across countries, the estimate for Ethiopia suffers from two problems associated with the rapid growth o f the system in recent years. First, because o f the large number o f overage students currently in the system, the numerator in the completion rate pertains to a population that i s inevitably older (and therefore smaller insize) than the population inthe denominator. Second, the numerator includes students from several age cohorts while the denominator refers to only one age cohort. The effect o f the first problem is to underestimate the completion rate, while that o f the second is to overestimate it. We can minimize the first problem by ensuring consistency inthe ages o f the populations in the numerator and denominator. Making this adjustment using data from the 2000 Ethiopia Welfare Monitoring Survey-the latest available at this writing-yields a grade 6 completion rate o f 42 percent4'-a muchhigher figure thanthe unadjusted rate o f 24 percent based on school census data andpopulationprojections for 1999-2000, the corresponding school year. 2.44 With regardto the second problem, however, there does not appear to be an obvious way to adjust the calculations. This handicap implies that the partially corrected completionrate o f 42 percent ingrade 6 (and 53 percent ingrade 5) reported above is almost certainly an overestimate o f what we wish to measure. Note, too, that because the overestimation arises from a transitory phenomenon that must dissipate as the backlog o f unmet demand for schooling is satisfied, the partially adjusted completion rate can be expected to decline over time inthe absence o f changes inthe underlying cohort entry and survival rates. Obviously we would expect the latter rates to improve, inwhich case the trend inthe completion rate is likely to become unpredictable as the improvement i s offset by the negative transitory effects. 2.45 Cohort estimates. For monitoring trends in completion rates over time, we would need a statistic that i s free from the transitory effects associated with the explosive expansion o f the system in recent years. A simple approach is to multiply the cohort entry rate by the cohort survival rate to each grade in order to obtain the corresponding cohort completion rate. The completion rates thus computed show the proportion in a cohort o f children that would attain uses data for one grade to compute completionratesfor the previousgrade, we will follow standardpracticeinusing the former definition. 40 Ethiopia's grade 6 completionratewas computedby dividing the nonrepeatersinthis grade by the 12-year-oldpopulation, 12 beingthe sum of the official age at entry to grade 1 and the number of years betweengrade 6 and grade 1. A similar calculation for grade 5 yields acompletionrate of34 percent in 1999-2000 and44 percentin2001-02. 4' The adjustments were made using data from the 2000 Welfare Monitoring Survey (which was fielded in JanuaryRebruary 2000 andwhose datathus correspondedto those collectedfor school year 1999-2000). Becausethis survey provided information on enrollment status for individual children by age and grade, it was possibleto evaluate the age of nonrepeatersin each grade and to relate the number to the population of the correspondingage cohort (after smoothing the distribution to correct for the effects of ageheapingin the reporteddata). - 42 - each grade if current pattems of transition between adjacent grades remain unchanged.42Figure 2.13 shows the results for 1995-96 and 2001-02 obtained inthis way, based on the cohort entry and composite survival rates reported earlier.43 For comparison, the partially corrected completion rates basedon cross-sectional data are also shown. Figure 2.13. Grade-Specific Completion Rates, Ethiopia, 1995-96 and circa 2000 100 I 1 4 b Cross-sectional .J b completionrates, 2000 80 - 4 4 808 + 60 a 40 20 0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 Grade attained Source: For the cross-sectional completion rates, authors' analysis of data from the 2000 Ethiopia Welfare Monitoring Survey; for the composite cohort survival rates, computed by multiplying cohort entry rates in table A2.2 and the composite cohort survival rates in technical note table T2.5, with results shown intable A2.3. 2.46 Several features are noteworthy in the figure. First, with regard to the results for around 2000, the cross-sectional completion rates exceed the cohort completion rate at each grade, the difference being especially large in the earlier grades. The pattern i s consistent with our expectation that the multiple cohort effect would be smaller inthe upper grades as older students ineachgrade are likely to leave the systemsooner than their younger peers. Second, the cohort completion rate at each grade shows a clear improvement between 1995-96 and 2001-02; for example, the grade 5 completion rate rose from 19 to 31 percent during this period, while the grade 6 completion rose from 17 to 27 percent. Although rising entry and survival rates both contributed to the improvement, the former is by far the more important factor, havingrisen from around 37 percent to more than 60 percent duringthe period (a gain o f 63 percent); survivalrates to each subsequent grade rose much more modestly by comparison (e.g., the largest gain in survival rate occurred betweengrades 1and2, and the increasewas only 24 percent). 2.47 Figure2.14 putsEthiopia's grade 6 cohort completionrate ina cross-country perspective. While Ethiopia's rate is not the lowest among countries at its levelof development, there is much scope for improvement. It is important to recognize that as new entrants to grade 1 must 42As mentionedabove, the cohort survival rate can be computed usingthe reconstructedcohort methodor the composite cohort method. For reasonsexplainedearlier, we will use the survival rates computed usingthe secondmethodto derivethe completion rates. 43Becauseno survey data are availableto estimatecohort entry rates for 2001-02, we used the cohort entryrates estimated from surveydata for 2000. -43 - inevitably come from more andmore difficult-to-reach populations, this indicator may not rise as rapidly in the coming years as it had in the past. If this expectation materializes, cohort completionrates can rise only as the systemimproves its ability to retain the children it currently attracts to enroll inschool. 100 n 5 80 c 0 B 60 - 4 4 % 3 u f 40 * * 20 4 b t h i o p i a 0 GradeRepetition 2.48 Grade.repetitioni s an important problemto addressbecauseit increasesthe probabilityo f dropout-by signaling parentsthat their child is not progressing academically. It also reducesthe efficiency o f student flow in that a repeater costs at least twice as much per grade attained as a nonrepeaters,while yieldingonly modest gains, ifany, interms o f student learning. 2.49 InEthiopia, the authorities make adistinctionbetweenrepeatersandreadmitted students: the former are students who sit for and fail the year-end examination, and return to the same grade the next school year; the latter are those who drop out during the school year without sitting for the year-end examination and retum to the same grade the next school year. The distinction helps educators isolate two factors associated with grade repetition: inefficiencies in the teaching process itself, and demand-side constraints, which cause students to stop their schooling before the end of the school year and which in turn retard their leaming and make it necessary for them to repeat the grade. While the separation o f the two types of students has its use, we set it aside for our analysis o f student flow for three reasons. First, the inclusion o f readmitted students in the count of repeaters would avoid a serious underestimation o f cohort survival rates (because such students would otherwise be treated as dropouts inthe calculation). Second, it would render Ethiopia's data comparable to those o f other countries. Third, it would make it possible to construct an internally consistent time series for Ethiopia as well, since the - 44 - distinction between repeaters and readmitted students was made only fiom 1993 E.C. (Le., school year 2000-01) onward. 2.50 Trends over time. Table 2.7 shows the grade-specific repetition rates among primary and secondary students in 1992-93 and 2001-02. For completeness, it also shows the share of readmitted students among those who were enrolled in the same grade for two or more years. The large share of such students emphasizes the importance of clarifying the definition of the repetition rate, especially if the indicator i s to be used for monitoring the system's performance over time. Based on the definition usedhere, repetitionrates rose between 1992-93 and 2001- 02, for an overall average increase o f 42 percent for the primary cycle as a whole. The increase in grades 9-10 was much smaller, only 8 percent; in grades 11-12, repetition has practically disappearedas a result o f the newpolicy that was implemented in2000-01. Table 2.7. Repetition Rates and Share of Readmitted Students among Repeaters, Ethiopia, 1993-93 and 2002-03 Percentageof udentswho repeated the grade a' 2001-02 Readmittedstudents as a share of all Index repeaters, 2002-03 Grade 1992-93 Percent (1992-93=1.00) (percent) b' 1 15.4 19.1 1.24 43.1 2 8.7 14.1 1.62 43.3 3 6.9 13.2 1.91 44.4 4 8.4 16.1 1.92 35.5 5 6.2 15.9 2.58 37.6 6 11.0 11.5 1.05 48.4 7 12.5 22.6 1.81 33.1 8 21.3 22.1 1.04 21.0 ~ Grades 1-8 11.6 16.4 1.42 39.8 Grades 9- 10 19.3 20.8 1.08 38.3 Grades 11-12 8.5 0.4 0.05 i 22.8 I I I I Vote: The repetition rate is defined as the number o frepeatersin grade X inyear Y relative to the number o f students in grade X in year Y-I. a. Repeaters include all students who were enrolled in the same grade for two consecutive years, regardless of whether or not they sat for the year-end examination. b. Readmittedstudents are those who drop out during the school year and retum to the same grade the next year. Source:Based on data on enrollments by grade and year reported in Govt. of Ethiopia 1993-94 and 200142. 2.5 1 Figure2.15 puts a historical perspective on the trends inrepetitionrates. The rates inboth subcycles o fprimaryschooling are comparable, and have been rising consistently since 1997-98. Ingrades 9-10, the rates have oscillated up and down inthe 20 to 25 percent range for most of the period, while in grades 11-12, they have beenrising steadily since 1992-93; they reached a high of more than 20 percent by 1999-00 before dropping to near 0 under the newly implementedpolicy, which sharply curtailed repetitioninthese grades. -45 - Figure 2.15. RepetitionRates inPrimary and SecondaryEducation,Ethiopia, 1992-93 to 2001-02 30 /Grades 9-10 25 20 10 I \Grades 1-4 \Grades 5-8 5 - \ o i 2 ma o\ a. r? " $ 3 , * pi ? o\ 9 k9 9 9 9 : s q 5 N j 8 Note: Repetitionrates inthis figure are computedby dividingthe total numberrepeaters(includingreadmittedstudents) ingradeX in year Y by totalenrollments ingradeX inyear Y-1. Source: Basedon data intechnicalnotetableT2.1 2.52 Cross-csuntrv comparisons. Ethiopia ranks in the middle o f a group o f low-income countries for which data are available on the percentage o f repeaters among those enrolled in grades 1-6 (see figure 2.16; note that for this comparison, Ethiopia's data have been adjusted accordingly). The lesson to draw from the figure is that while grade repetition i s not yet a serious problem in Ethiopia, it has been growing in the past few years and could get worse-given the experience o f suchplaces as Rwanda andBangladesh-in the absence o f explicit attentionto this aspect o f student flow. Requiring repeaters to exit the system, as was recently instituted as a policy in upper secondary education, is a possible solution. While the approach is probably appropriate at that level to help manage the huge demand for places in government secondary schools, it detracts entirely from the goal o f achieving universal primary completion. Automatic promotion is also flawed as a solution because it merely shifts to the next grade the underlying problem o f inadequate student learning. More promising are solutions that combine at least three elements: (a) automatically promoting students within subcycles o f two or three years; this would avoid unnecessary compartmentalization o f the learning process and allow lagging students to build on skills already acquired rather than starting each year with curricula covered the previous year; (b) stabilizing teacher assignments in each subcycle so that teachers could work with the same group o f pupils over several years and familiarize themselves with their charges and their learning needs; and (c) strengthening support for teachers, enhancing their skills in formative evaluation o f student learning and introducing other measures to improve teaching effectiveness inthe classroom. - 46 - Figure2.16. Repeatersas a PercentageofEnrollmentsinPrimaryEducation,Ethiopia and Other Low- IncomeCountries,circa 2000 Rwanda C.A.R. Congo, Rep. Bangladesh Guinea-Bissau Angola Chad Guinea Lesotho Cambodia Ethiopia Congo, Dem. Rep. Senegal Nicaragua The Gambia Uganda Sierra Leone Honduras Indonesia Pakistan Bolivia Sudan Nigeria 0 5 10 15 20 25 30 3 5 40 Percent C.A.R. Central African Republic; Congo, Dem Rep., DemocraticRepublic of the Congo; Congo Rep., Republic o f Congo. Note: Sample includescountries with a six-year primary cycle. The rate for Ethiopia refers to grades 1-6 in 2000-01. Source: For Ethiopia, authors' calculationbasedon data on enrollmentsand repeaters (including readmittedstudents) in Govt. o f Ethiopia 2000-01; for all other countries, BIUUS,Mingat, and Rakotomalala2003. Summarizingthe Efficiency of StudentFlow 2.53 An education system that is characterized by grade repetition and dropout uses more resources to produce its graduates than one free from these problems. We can thus use information on these indicators to evaluate the system's student flow efficiency. Table 2.8 assembles the data for 1993-94 and 2001-02 from which an index o f flow efficiency can be calculated. To illustrate the computation, consider the data for 2001-02. Starting with a cohort o f 1,000 pupils, the resources spent in grade 1 amount to 1,235 pupil-years (=1,000/[1-0.1911). In grade 2, only 746 pupils are left, and the resources invested inthem amount to 867 pupil-years (=746/[ 1-0.14 11). Continuing likewise to the remaining grades, the cumulative resources spent on the cohort for eight years o f primary schooling amount to 5,704 pupil-years. Since only 363 eighth-graders are actually produced, only 2,904 pupil-years (=363 x 8) would have beenneeded inthe absence ofrepetition anddropout. Inother words, the system infact operates at only 0.51 (=2,904/5,704) times as efficiently as a system in which all entrants reach grade 8 and do so without repeating a grade. Isolating the contribution o f dropout alone, a total o f 4,749 pupil-years are invested to produce 398 eighth grade graduates, implying a dropout-related efficiency index o f 0.61. Considering only grade repetition, we see that a total o f 5,704 pupil-years were used when4,749 pupil-yearswould have sufficed, which yields arepetition-related efficiency index of 0.83. - 47 - Table 2.8. Efficiency of Student Flow in Primary Education, Ethiopia, 1993-94 and 2001-02 2001-02 Number left Number left from rom an initial an initial cohort of Repetition Pupil-years cohort of Repetition Pupil-years Grades 1,000 rate" invested 1,000 rate" invested 1 1,000 0.139 1,162 1,000 0.191 1,235 2 580 0.075 627 746 0.141 867 539 0.055 570 652 0.132 751 4 518 0.086 567 587 0.161 699 5 489 0.064 522 515 0.159 612 6 483 0.137 559 450 0.115 508 7 413 0.147 484 437 0.226 564 8 385 0.243 508 363 0.221 467 hmulative pupil rears 4,406 4,999 4,749 5,704 ndex of efficiencyb' Dropout-related 0.70 0.61 Repetition-related 0.88 0.83 Overall'' 0.62 0.5 1 (0.71) (0.61) a. Definedas the number ( repeaters ingrade X inyear Y+l relativeto the number of sl lentsingrade X inyear Y. b. The indexis defined as the ratiobetweenthe cumulative pupil-yearsinvestedinasystemwithneither droppingout nor graderepetition, and the actualcumulative pupil-yearsinvestedgiventhepattemof droppingout andrepetition.Thus, the closer is the indexto 1.0, the moreefficient is the studentflow. c. Figuresinparenthesesreferto the indexassociated with studentflow pattems ingrades 1-5. Source: Authors' estimates basedon survival andrepetitionratesreportedintables 2.5 and2.6. 2.54 Comparing the results for 1993-94 and 2001-02, it is clear that the system has become less efficient over time, and that dropping out contributes a greater share o f the efficiency loss than repetition inboth years (as reflected inthe smaller index associated with dropout inboth years). A system operating at an efficiency rate o f 51 percent i s indeed incurring a significant loss o f resources. For comparison, figure 2.17 ranks Ethiopia's performance against those o f education systems inother low-income African countries for which an index could be computed, standardizing across all countries for five years o f schooling. Although Ethiopia i s not currently at the bottom of the group, it could easily slip into that position by allowing repetition rates to rise as high as they have in such countries as Madagascar, or, altematively, improve its position byincreasing survivalrates andlowering grade repetitionto the levels inTanzania (which were, respectively, 74 percent and 4 percent in 1999). - 48 - Figure 2.17. Efficiency of Student Flow in Grades 1-5 inEthiopia and Other African Countries, circa 2000 Ethiopia Benin RWm& Togo Cameroon Mozambique Madagascar 0 0 0.2 0.4 0.6 0 8 1.o Index o f studentflow efficiency No@:'Ihe indexrefersto admt flow inGrades 1-5 inallcounaies shown above. Itrangesh m 0 to 1.0,resptively, lo a riaation whae all radmu dropoutor repeaf lo one inwhich all studmureachGrade5 withoutrepetitioncorresponding, or droppingout. Source:Authors' calculationbasedon data intable 2.6 far Ethiopia,andU N E S M 2001 far dataan nWival mdrepetitionrates for allthe0 t h countries. ~ Issues for Policy Development 2.55 Following the reforms in 1994, Ethiopia's education system now serves a much larger clientele, particularly at the bottom of the pyramid. While the stresses o f rapid growth have undoubtedly produced many new challenges for policy development, the analysis presented in this chapter has identified three issues that warrant close attention: 0 The system's 4-4-2-2 structure andthe loci of student flow regulation; 0 Indicators for monitoringprogresstoward the educationMDG; and 0 Options for raising entry and survivalrates inprimaryeducation. We briefly review the first issues below, deferringthe third issue to chapter 4 where an nalysi o f social disparities inschooling provides a fuller context for considering these options. Structure of the Education System and Loci of Student Flow Regulation 2.56 There is now widespread agreement that literacy and numeracy are core skills that can boost a child's life chances in adulthood. Accordingly, most governments seek to universalize basic education as part o f their strategy for social development and poverty reduction, a goal that was formalized in the 2000 UnitedNations Declaration on the Millennium Development Goals. Inthat Declaration, basic education is definedas five years ofprimaryschooling, but obviously, as countries develop and their economies mature, we would expect it to be extended to eight or nine years o f schooling, typically covered by primary and lower secondary education, and indeed, even to 12 years, as inthe most advanced economies. The appropriate speed for moving toward a longer cycle i s ultimately constrained by labor market considerations. Ifthe education systemproducesmore graduates qualified beyond the core skills o f basic literacy and numeracy thancanbe absorbedinthe prevailingjob market, the investment would bewasted to the extent that graduates fail to find work appropriate to their level o f training and so lose the expected retumsontheir investment. With enough frustratedjob seekers on the streets, the situation could - 49 - easily escalate into a serious social problem. Inpart to minimize this risk andto institute a merit- based system that allocates scarce places in postprimary education, most governments in low- income countries use standardized examinations to regulate the flow o f students beyond basic education. 2.57 InEthiopia, the current 4-4-2-2 structure is not initself a problem. What is ofconcern is that the loci o f selection tend to be loaded toward the high end of the system, the first standardized examinations occurring only at grade 8. The lack o f a selection point earlier in the system may not matter much at present because most first-graders drop out long before they reach grade 8, leaving only slightly more than a third of the entrants who eventually do so. But its absence would undoubtedly increase the difficulty o f managing the pressure on postprimary education as basic education expands. Already the government has found it necessary to radically restrict access to grade 11, diverting grade 10 completers to programs in teacher training, and technical and vocational education andtraining, inorder to relieve the rapidbuildup o f pressure on preparatory and higher education in recent years. In some ways, the new policy effectively amounts to a shrinking of the overall size of upper secondary education, as well as a redirection o f enrollments toward occupationally defined fields. The shaded cells in table 2.9 show the drastic drop in enrollments in grades 11 and 12 as the new policy took effect from 2001-02 on. Table 2.9. EnrollmentsinGrades8-12 in Government and NongovernmentSchools, Ethiopia, 1999-00 to 2002-03 i__ Grade 1999-00 2000-01 , , ,2001-02 2002-03 8 268.4 299.7 352.5 444.3 9 211.3 265.7 312.1 355.0 1 :: 10 156.1 112.5 156.1 193.1 , , Note: Data refer to sti ents in regular programs only; shaded cells draw attention to large reductions in anrrrllmon+m frnm orio4nn ,.-.., ,.-l;,.., +*+.-h+-,. ,,1,,.6:,, Ah ----,.A--. -A.."-c-- 2.58 As the country makes progress in expanding enrollments in primary education, the pressures on the system will inevitably grow. If other countries' experiences are a guide, it will infact be necessary not only to tighten selection between grades 8 and 9 but at an even earlier point as well. Where precisely to do so is entwined with decisions about the duration o f schooling to be universalized. The current 4-4 structure inprimary education does not make for an easy choice: universalizing eight years o f schooling is probably fiscally unviable even in the medium term, while universalizing only four years seems overly modest against the five years defined in the MDG Declaration (and probably also inadequate if pupils are to achieve permanent literacy and numeracy). IfEthiopia chooses to universalize five years o f schooling, it would make sense to introduce, as part o f the policy package, a new selection mechanism at the end o f grade 5 inorder to strengthen the management of student flow throughout the system. - 50 - Indicators for Monitoring Progress toward the Education MDGs 2.59 Inthe past, cross-sectional measures such as the gross and net enrollment ratios have been popular (inpart because they are easy to compute), but they are flawed inways that make them unsuitable for tracking how well a country is progressing toward ensuringthat all children have access to a full cycle of primary schooling.44Increasingly used instead, particularly in the arena o f international development assistance, is the primary school completion rate (typically at grade 5 or 6). Yet, as the analysis inthis chapter has revealed, cohort rather than cross-sectional estimates of this indicator are required ifthe intentionis to track changes over time. InEthiopia's rapidly expanding system, cross-sectional estimates are particularly inadequate because they are highly distorted by the transitory presence of multiplecohorts andby the related mismatches in the age profiles of the denominator and numerator populations relevant to the calculations. Clarity i s also needed with regard to the measurement of grade repetition, another relevant indicator for monitoring performance. While incomplete as an indicator o f system efficiency, high rates of repetition point to possible problems in managing student flow, which, if left unattended, are likelyto make it harder to raise survivalrates. 2.60 Assuming that school completion and repetition are to be tracked, it will be important to agree on the methods and sources o f data for making the estimates. Tradeoffs are inevitable, but important considerations include the following: (a) use o f cohort rather than cross-sectional measures, for the reasons explained above; (b) reliance on data generated by existing mechanisms to avoid overburdening scarce data collection capacity; and (c) use o f simple calculations that keep tedious data manipulation to a minimum. Table 2.10 shows the indicators that are likely to meet these criteria. Cohort entry rates require survey data in which the educational status o f childrenis reported. The fact that the data pertainto a sample rather than to the entire population is admittedly a flaw, but it cannot be avoided because of the nature of the desired statistic. The highfrequency of household surveys inEthiopia should makeit possible to cross-check the estimates, all o f which can be generated at relatively low cost by consistently including in the survey questionnaires a question for each household member aged 5 and above on whether or not they have ever been enrolled in school (and if so, inwhat type o f school). The other indicators can be estimated using data generated from school censuses. While accuracy in the census data, particularly with regard to repetition status, continues to be an issue, these data pertain to the whole populationo f students, and can further be cross-checked by including, as is already done in some household surveys in Ethiopia, two additional questions in the survey instruments: (a) Are you currently enrolled, and if so, inwhat grade? and (b) Were you enrolled last school year, and ifso, inwhat grade? ~~ ~ 44 Other measures have also been identified, including: (a) the proportion of pupils starting grade 1 who reach grade 5; (b) the literacyrate of 15-24-year-olds; (c) ratio ofgirls to boysinprimary, secondary, andtertiary education; and (d) the ratio of literate femalesto males of 15-24-year-olds. See 110 22.8 4.3 4.3 8.6 1.6 0.0 Total 100% 100% 100% 100 % 100 % 100Yo Memoranda: Sample size 11,022 5,939 376 592 437 53 Pupil-teacher ratios Unweightedmean"' 89.0 52.2 57.0 57.3 36.2 21.7 Weighted meanb/ 75.2 48.8 51.5 48.3 35.1 20.0 Mediand 76.6 45.8 49.0 45.8 31.2 20.2 Standard deviation 60.9 38.6 56.7 48.5 30.2 13.0 a. Refers to the average of the pupil-teacher ratios across schools, computed by first calculatingthe ratios for each sample school, and then taking the average of these ratios. b. Refers to the ratio obtained by dividing total number of students by the total number of teachers. For govemment schools, data excludes Somali, Benshangul-Gumuz, and Gambella. c. Refersto the ratio that dividesthe sample intotwo equal groups, one with ratiosexceedingit and the other with ratios smaller than it. Source: Analysis of the Ministry of Education's 2001-02 EthiopiaSchoolCensus. Teacher Allocation across Government Schools 5.22 Inthis section, we take the foregoing analysis a step further by examiningthe pattern of teacher allocation across schools. The presentation i s limitedto the government sector, inpart to simplify the presentation, but, more important, to focus attention on issues in system managementthat the government might address. 5.23 Relationbetween enrollments andteacher allocation across schools. The relevant data are displayed in figure 5.5, which plots the size o f a school against the number o f teachers it has.96 The left panelinthe figure pertains to primaryschools, andthe rightpanel, to secondary schools; in both panels, the horizontal axis shows a school's enrollment while the vertical axis, the numberofteachers assignedto the school.97Across bothprimary and secondary schools, the data suggest that school size relates positively, as it should, to the number o f teachers; yet at the same time, this positive relationship is not particularlyprecise: among secondary schools, for example, a school with, say, 1,000 students, may have as few as 10 teachers or as many as 25. The loose 96For similar graphs showing the data separately for grades 1-4 and grades 5-8, see appendix figure A5.2. 97 Data from Oromiya Region indicate the presence o f a very small number of teachers (about 0.3 percent o f the total) who are paid by the community. These teachers are not counted inthe data for most regions, and are thus excluded from the analysis here. - 135 - relationship between school size and teacher allocation i s a corollary o f the wide disparities in the pupil-teacher ratio already documented above. Figure 5.5. Relation between Numbers of Students and Teachers acrossGovernment Primary and Secondary Schools, Ethiopia, 2001-02 (a)Grades 1-8 @)-Grades9-12 110 0 0 0 70 0 d o 0 0 0 90 f6 0 0 0 0 0 0 s 6 0 50 0 p 70 0 r- ` 0 c .k Pb 2 0 z 5 50 30 30 10 10 0 0 0 0 0 0 1 0 l b 1 0 5dO lob0 1500 2000 30b0 4000 10 500 1000 1500 2000 3000 Numberof students Numberof students Note: For separatefigures for grades 1-4 and grades 5-8, see appendix figure A5.2. Source :Analysisof the Ministry of Education's 2001-02 Ethiopiaschoolcensus. 5.24 Regression analysis o f teacher allocation across schools. We turnto regression analysis to achieve a more systematic assessment o f the relation between school size and number o f teachers. The results o f such analysis based on data for government schools are summarized in table 5.11 and suggest three main observations. First, the coefficient estimates on the number o f students indicate that, on average, a new teacher i s added for every 111 students in grades 9-12, compared with the corresponding figures o f 83 students ingrades 1-4 and 56 pupils ingrades 5- 8.98 Teachers are scarce throughout the education system, but especially in the first cycle o f primary education and in secondary education. These allocation pattems, particularly in secondary education where the curriculum requires specialized teaching, implythat schools must enroll largenumbers o f students inorder to keep the student-teacher ratio incheck. '*These figures are computed by taking the inverse o f the coefficient estimates on "number o f students" in the regression equations reported inthe table. - 136 - Table 5.11. Regression Estimates of the Relation between Numbers of Teachers and Students across Government Schools by Levelof Instruction, Ethiopia, 2001-02 Item Grades 1-4 Grades 5-8 Grades 1-8 Grades 9-12 Number of students 0.012 0.018 0.016 0.009 (91.58)** (132.49)** (164.73)** (26.86)** Constant 15.78 15.417 26.244 9.428 (31.22)** (30.36)** (35.88)* * (3.96)** No. ofobservations 11,022 5,939 11,122 372 R2 0.50 0.79 0.76 0.75 Memorandum: R2for regressionwithout regional dummies 0.44 0.75 0.72 0.70 Note: Regression equation includes regional dummy variables, not shown here to economize on space. Value of t-statistics appear in parentheses; two stars (**) denote statistical significance at the 1 percent confidence level. Results for nongovemment schools are shown inappendix table A5.1. a/ Refers to the number of schools offering the indicated level of instniction; for the purpose of the regression, the number of teachers includes all who teach at the indicated level, whether or not they have classes at the other levels. Source: Analysis o f the Ethiopia2001-02 School Census; see table A5.4 for full regression results. 5.25 The other two observations center around the values o f the R2statistic. This statistic is a measure o f how well the estimated regression equation accounts for the disparities in teacher allocation across schools. Ina system where teachers (or fractions o f their time) are allocated to schools strictly on the basis o f enrollments, the R2 statistic would have a value o f 1.0. Conversely, the further the R2value is from 1.0, the more factors other than school size intervene in the allocation of teachers across schools. Table 5.11 reports the R2values for two sets of regressions corresponding to specifications with and without dummy variables for regions. The fact that the R2variables are highly comparable between the two specifications implies that subregional processes play an important role in the placement o f teachers. This result is consistent with those from similar analyses in other countries. It reflects the reality that while teacher allocation rules may exist, the outcomes are decided at the local level by officials who ultimately implement the rules inresponse to the pressures andpriorities they feel andperceive. 5.26 The third observation is that the R2 values for the regressions for grades 1-4 are significantly smaller than for grades 5-8 or grades 9-12, which implies greater randomness in teacher allocation at that level. One possible explanation may be the fact that some primary schools-58 percent, to be precis-ombine both subcycles o f primary schooling under one roof. In such schools, decisions about teaching assignment between the two subcycles are activated at the school level; it is possible that the staffing requirements in grades 5-8 are prioritized, perhaps to facilitate implementation o f a curriculum that requires specialized teaching. If so, the number of teachers for grades 1-4 would simply be a residual complement, after the requirements ingrades 5-8 have more or less beenmet. 5.27 Cross-countrv and regional commrisons. Whatever the underlying processes, the fact is that staffing ratios are highly disparate across schools in grades 1-4. While the situation is better at the other levels, room for improvement probably exsists throughout the system. This possibility becomes clear in light o f cross-country experience, as illustrated in figure 5.6. Inthe - 137- figure, countries are ranked in order o f the degree o f randomness in teacher allocation across public sector primary schools, as measured by how far the R2 values for each country's regression fall short o f unity. Ethiopia's randomness index for grades 1-8 i s moderately highby African standards, but its index for grades 1-4 exceeds that o f all the countries inthe figure. It is noteworthy that in such countries as Burkina Faso, Madagascar, Mauritania, Mozambique, and Niger (countries where the degree o f randomness i s smaller than in Ethiopia) efforts are already under way to reduce inconsistency inteacher allocation as a priority for sector development. Figure 5.6. RandomnessinTeacher DeploymentAcross SchoolsinEthiopia,2001-02, andOther African Countries, circa2000 I h s 50 85 y1 40 j 30 20 10 0 mts Note: The degree o f randomness i s computed as 1 minus the R2value o f regression equations relating number o f teachers assigned to teach the indicated grades to number o f pupils in those grades, schools being the unit o f observation. The index has a 0-1 range; higher values indicate greater inconsistency in the deployment o f teachers across schools. Data refer to government schools in all countries. Source: Analysis o f each country's school census data. 5.28 Regression analysis also provides a simple means for regional comparisons o f the consistency o f teacher allocation across schools. As before, the method involves estimating for each region the relation between enrollments and number o f teachers; and computing the index o frandomness by subtracting from unity the R2value o f the regression estimates, and expressing the result in percentage terms.99 Figure 5.7 below summarizes the results for grades 1-8 and grades 1-4. Performance is starkly different across regions: among the four largest regions, the index of randomness for grades 1-8 varies from just 8 percent in Tigray, to 41 percent in SNNPR; Amhara and Oromiya are ranked inbetween and have an index o f 22 and 28 percent, respectively. The range is as wide as what is observed internationally; the outcome inTigray is comparable with the best-performing countries, which suggests the practices inthis region might hold lessons for other regions on ways to manage the deployment o f teachers. Among the remaining regions, the networks o f schools are relatively small (e.g., Harari and Dire Dawa have fewer than 45 schools each), and one would expect greater consistency in the relation between school size and number o f teachers. While this expectation is borne out in most regions, the degree o f randomness in teacher allocation is relatively high in Harari, Afar, and Benshangul- Gumuz, and exceptionally SO in Gambella. Finally, the figure shows that in all regions, 99See appendix table A5.4 for the full regression results. - 138 - substantially more randomness exists in teacher allocation in the first four grades o f primary schooling than inwhole cycle. EveninTigray, 26 percent o f the variation inteacher allocation is accounted for by factors other than the size of enrollments. Taken as a whole, the regional results confirm that the management of teacher deployment i s generally weak not only across schools, but also betweenthe two cycles ofprimary schooling within schools. Figure 5.7. RandomnessinTeacher Deployment across Government Primary Schools by Region,-Ethiopia, 2001-02 Grades 1-8 \ Frades 1-4 Benshangnl, Benshangnl-Gnmuz; SNNPR, SouthernNations andNationalitiesPeoplesRepublic. Now The degree of randomnessis computed as 1 minus the R' value of regression equations relatingnumber of teachers assignedto teach the indicated grades to number of pupilsin those grades, schools beingthe unit of observation. The index has a 0-1 range; higher values indicategreater inconsistency in the deploymentof teachers across schools. Source: Analysis o f the Ministry o f Education's 2001-02 school census; see also appendix table A5.5. Economies of Scale in Service Delivery 5.29 Inthis sectionweconsider howcostsrelateto the sizeofenrollments. Beforestarting, we note that in Ethiopia, the school census contains information on the number o f administrative staff and teachers, along with their distribution by salary range, as well as`reported spendingon nonsalary school inputs. We combine this information to compute the cost per student and plot the result against the size o f the school (Figure 5.8).'0° The relation shows the characteristic disparities across schools, which are unsurprisinggiven the large gaps across schools inteacher allocation that we havejust documented. looSince salaries are related to these variables, the costs thus computed capture information not only on the number o f staff, but also their qualification and experience. - 139 - Figure 5.8. Relation between Cost Per Student and Size of Enrollments across Government Schools, Ethiopia, 2001-02 (a) Grades 1-8 500 4 - l o b 800 700 O0 600 1 , , 50b 1ado 15dO 2000 2500 3000I Number o f students (b)Grades 9-12 8B 500 400 V 8B 300 200 100 10 , 100 500 I f 1000 1500 2000 2500 3000 3500 4000 4500 5000I I I I I I 1 I I Numberofstudents Source: Analysis of the Ministryof Education's 2001-02 Ethiopia school census. 5.30 Reaession analysis. As before, we use regression analysis to provide a systematic evaluation o f the relation between size o f enrollments and total spending across schools. The results appear intable 5.12. The fact that the R2values in these regressions are smaller than the corresponding R2values inthe regressions inthe previous section (involving number o f teachers andschool size) implies that the disparities across schools inoverall recurrentcosts are even less well explainedby school size alone. 5.31 Our focus here, however, is on the coefficient estimate o f the school size variable. Inthe regressions, both the dependent variable (total recurrent costs) and the size o f school (i.e., enrollments) are expressed in logarithmic terms. The coefficient estimate on the school size variable i s thus interpreted as the percentage increase in cost for a given percentage increase in enrollments; the inverse o f this coefficient is a measure o f the magnitude o f the economies o f scale in providing services. Because of the way services are currently organized, there are few - 140- economies o f scale in primary education: a 1 percent increase in total spending allows only a proportionate increase in enrollments; by contrast, the economies of scale are substantially higher in secondary education: a 1 percent increase in total resources allows a doubling of enrollments. This result i s consistent with the fact that secondary schools have more administrative staff and specialized teachers whose time can be fully utilized only in larger schools. Table 5.12. Regression Estimates of the Relation between Total Costs and Enrollments across Government Schools,Ethiopia, 2001-02 Item Primary schools (grades 1-8) Secondary schools (grades 9-12) Ln(Numberofpupils) 0.987 0.5 11 (66.91)** (5.99)** Constant 4.131 8.439 (44.49)** (13.53)** Numberofobservations 9,608 264 R2 0.32 0.12 Memorandum: Economies of scale"' 1.01 1.96 5.32 Economies o f scale andthe size distributiono f schools. The foregoing analysis provides a lens through which to view the size distribution in the current network of government schools. Figure 5.9 below shows the distribution o f the primary schools and the unit costs for given school sizes simulated from the regression equations. The absence o f economies o f scale at this level o f education i s reflected inthe relatively flat cost curve. Yet some 23 percent o f the schools enroll more than 900 pupils, and about 12percent, more than 1,250 pupils. The presenceo f these relatively large schools suggests that factors other than recurrent cost considerations have also been influencing school size. InEthiopia, one such factor is the highcost o f school construction. Under the Education Sector Development Program (ESDP), construction was budgeted at $20,000 per furnished and equippedclassroom (allowing for peripheral facilities); to put this cost in perspective, recall that it is equivalent to 25 times the average salary of primary school teachers. A recent study estimates that a four-classroom school could be built for between 150,000 and 200,000 Birr, depending on choice o f materials and design, for a per-classroom construction cost o f between$4,400 and $5,900 (McArthur 2003). Another study that collected costs from several sites basedon interviewswith experts and on tender documents puts the range at between $7,000 and $19,000, dependingon location (Coyle 2003); the study also suggests that alternative buildingdesigns and materials couldpotentially lower these costs by about a third.lo' The prospect of taking advantage of economies of scale in service delivery will depend greatly on how well capital costs are managed. If costs persist at $20,000 per classroom, large schools are inevitable; by the same token, fewer schools would be built; this would compromise the lo'See appendix table A5.6 for details. - 141- accessibility o f schools, particularly in rural areas where most of the expansion o f the system must take place. ,Costpapupil 45 0 N S F. V t A N. schoolsize(numberofstudents) h e : SizedistriihbasedonanalysisofMinistryofFklwxtiion's 2001-02scbwlcensus; wst p~pupilsirmlatedh mregressionequationreported intable 5.12 5.33 For secondary education, the data on recurrent cost and size distribution o f schools are summarized in figure 5.10 below. The cost simulation shows the expected decline in recurrent cost per student as enrollments rise: it falls from 400 Birr per student inschools with fewer than 200 students, to just above 100 Birr in schools with enrollments o f 1,500 to 2,500 students; indeed, beyond 1,500 students, the decline in costs becomes increasingly small as economies o f scale are exhausted. As inprimary education, the size distribution o f secondary schools does not appear to be fully coherent with the pattern o f scale economies. About 15 percent o f the schools enroll too few students and are three to four times as costly per student as the larger schools. At the same time, a sizable share of the schools appear to be much larger than they need be to benefit from economies of scale: more than 40 percent have more than 1,500 students, andnearly 25 percent have more than 2,500. As inprimary education, the high cost o f school construction may be driving the system toward the larger schools. The Education Sector Development Program, for example, budgets construction o f secondary schools at $30,000 per furnished and equipped classroom. Equivalent to about 23 times the average salary o f a secondary school teacher, this cost i s clearly on the high side and creates inevitable pressures toward large schools.'02 lo2Another way to evaluate these costs is to assess how the amortized cost per student relates to the recurrent cost per student. Assuming that classrooms accommodate 180 students each (e.g., two shifts of 90 students each), that they have a useful lifespan o f40 years, and that the interest rate averages 3 percent a year, the amortized capital cost would be 57.7Birr per student, which is about 35 percent as high as the average recurrent cost per student. - 142 - Figure 5.10. Size Distribution of Government Secondary Schools and Recurrent Cost Per Student by Size of School, Ethiopia, 2001-02 25 1 500 F 20 400 8 av1 .u BP B r- 15 300 3c 13 P 200 a lo & 100 0 .- N- School size (number o f students) Source: Size distribution o f schools based on analysis o f the Ministry o f Education's 2001-02 school census; unit cost simulated from regression estimate reported in table 5.12. StudentAchievement 5.34 Inany discussionofservice delivery, learningoutcomes arejust as important to consider as costs. For this reason, we assemble below the available data on student learning and attempt to piece together a picture o f the system's performance in this regard. To forewarn the reader, we must note that the data for Ethiopia are sparse and the little that exist are fragmentary and not readily accessible in formats conducive to detailed analysis. Because Ethiopia does not participate as yet in any o f the growing number o f international or regional studies on student learning, we cannot put the system's performance in a cross-country context, as we did for various aspects o f service delivery discussed in earlier sections. lo3 Nonetheless, the available data capturing experience within the country does help to raise important issues for policy development, even if they cannot provide definitive answers at this point. The sections that follow provide an overview o f national examination results, a summary o f two recent studies on student achievement, and an analysis o f student learning in Oromiya prepared specially for this study. Overviewof Performanceon NationalExaminations 5.35 National examinations at grades 8 and 12 date back to the days o f Emperor Haile Sela~sie."~The grade 8 examinations were regionalized in 1999 but the results continue to be used for selection to grade 9. The examinations at grade 12were replaced in2003 by the College Entrance Examination as the mechanism for selection into postsecondary education. National IO3Examples of regional studies that collect information on student leaming include the following: UNESCO's Monitoring Learning Achievement (MLA) project; the Southem and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ), Programme d'Analayse des Systkmes Educatifs de la CONFEMEN (PASEC), Trends in Mathematicsand Science Study (TIMSS); Progressin InternationalReadingLiteracy Study (PIRLS, etc. . IO4Examinationsat grade 6 were also administered inthe past when the systemhada6-4-4 structure. - 143 - examinations at grade 10 are a recent mechanism, introduced in 2000 to manage access into grade 11. 5.36 Table 5.13 shows pass rates on the various examination^.'^^ Ingrade 8, the passing mark in 1999-2000, the last year for which the examinations were administerednationally, was set at 32.6 percent, and on this basis, 66.9 percent o f the nearly 270,000 who took the exam passed; this implies a pool of some 180,000 students who were qualified to enter grade 9. Ingrades 10 and 12, the passing grade is set at a grade-point average of 2.0. Among the large regions, pass rates for all three examinations are consistently high in Tigray, Amhara, and Oromiya, while they lagbehindnoticeably inSNNPR.'O6Passrates are also highinAfar, butthe regiontypically fields very few candidates. Inthe remaining regions, pass rates are low for Gambella at the grade 8 examination (57 percent as high as the national average); for Addis Ababa at the grade 10 examination (61 percent as high as the national average); and for Harari at the grade 12 examination (32 percent as high as the national average). For the grade 12 examination, pass rates are available separately for boys and girls, and they show that girls are only 63 percent as likely as boys to pass. In all regions, girls lag behind boys, but the gender gap is widest in S W R . Table 5.13. Pass Rates on NationalExaminations at Grades 8,10, and 12 by Region and Sex, Ethiopia, circa 2000 G Grade 8 Grade 10 Females 'ountrylregion 1999-00" 2001-02b/ Overall Males (Males=1.00) 3thiopia 66.9 58.5 54.1 63.1 0.63 3y region" Tigray 0.96 1.12 1.33 83.8 0.73 Afar 1.12 1.29 1.45 91.6 0.94 Amhara 1.20 1.09 1.16 73.4 0.64 Oromiya 1.17 1.16 1.13 71.5 0.67 Somali 0.91 0.76 1.04 65.4 0.96 Benshangul-Gumuz 0.88 1.16 0.98 61.8 0.69 SNNPR 0.69 0.82 0.86 54.1 0.39 Gambella 0.57 0.80 0.80 50.3 0.69 Harari 0.91 0.98 0.32 20.5 0.64 Addis Ababa 0.93 0.61 0.76 48.2 0.55 Dire Dawa 0.78 1.01 0.97 61.2 0.54 vlemorandum: No. taking exam 267,642 156,970 182,903 112.561 1 38.5% ;NNPR, SouthemNationsand ? Note: A breakdownbv sex is avi a. Nationalexaminationsreplacedby regionalexaminationsafter 1999-2000; pass rates refer to the percentageof studentswho scoredat . I or above the passingmark of 32.6 percent. b. Refersto the percentageof studentswho obtained agrade-point averageof 2.0 or higher. c. Indexedto nationalaverage. Source: Personalcommunicationfrom the National ExaminationOrganization. Although trend data are available on pass rates, their interpretation is unclear since the thresholds that determine pass rates vary from year to year, and insome years they also vary across regions. For the years shown inthe table, the passing mark is the same across all regions. See appendix table A5.7 for data on the number of candidates fielded by each region at the various examinations. - 144- Resultsof Two RecentAssessments of StudentAchievement 5.37 Two studies holdparticular interest inthe context o f this study.lo7The first is the Ethiopia National Baseline Assessments o f Grade 4 and Grade 8 Student Achievement completed in2001 by the National Organization o f Examinations; the second is an evaluation o f the costs and performance o f alternative routes to basic education, completed by the Ministry o f Education in 2000. Below we briefly summarize the results, focusing on aspects that are particularly pertinent to the relation betweeninputs or costs and leaming outcomes. 5.38 National baseline assessments o f grade 4 and grade 8 student achievement."* This nationwide study collected information in 2000 by testing more than 10,500 grade 4 students randomly sampled from the relevant sections in256 government schools (which were themselves randomly selected from each region), and more than 5,500 grade 8 students sampled from the relevant sections in 136 government schools; inaddition, questionnaires were administeredto the school directors and the more than 750 grade 4 teachers and nearly 470 grade 8 teachers associated with the sample classes. The tests were carefully designed to gauge grade 4 student achievement in reading (in the language o f instruction), English, mathematics, and environmental science; and grade 8 student achievement inEnglish, mathematics, chemistry and biology, They yield the following results, which provide a baseline for monitoring future trends instudent achievement: the overall percentage of correct answers for the four subjects averaged 48 percent in the grade 4 sample, and 41 percent in the grade 8 sample; in both samples, girls scored somewhat lower than boys. Since the test items were chosen from a range o f key topics in the curriculum for the grade tested as well as that for the previous grade, these scores indicate that a large number o f students were not achieving the curriculum objective^.'^^ 5.39 The studies also evaluated the correlates o f student achievement, focusing on such factors as: school infrastructure; school organization and management practices; teacher characteristics andpractices; and the availability o f instructional materials and equipment. Ingrade 4, the study stressed the positive contribution o f the availability o f textbooks (as well as teacher guides) and frequent use o f the radio for supplementary instruction. However, it showed that introducing instruction inlocal languages appears to have compromised student learning, andthat teaching in the local languages would require better support and training in order to close the performance gap. Especially interesting is the study's finding on the importance o f such process variables as how well teachers prepare their lessons and how much the director focuses on school matters. In grade 8, the study again confirmed the positive influence o f textbooks and instructional radio on student achievement and urged that investments in these areas be prioritized. Similarly, process variables also matter, including whether or not the school director tends to focus on school matters, and the extent to which students engage in homework. An important fact is that the study shows student achievement is not affected by the percentage o f teachers who graduated lo'Both studies were conducted inassociation with the ProjectBasic EducationSystem Overhaul (BESO) funded by the United StatesAgency for IntemationalDevelopment(USAID). logFor details of the study and the results, see Kedir, Zewdie, and Afework (2001a, b and c); see appendixtable A5.8 for further details. logMany of the teachers in the study felt that the curriculum was too difficult. Nonetheless, it is interestingthat in the schools where teachers tend towardthisview, test scores were higher. - 145 - from teacher training colleges or higher levels o f training."' To summarize, the results from the baseline assessment are consistent with a large literature that consistently points to the cost- effectiveness o f textbooks and other learning materials as well as that o f instructionalradio (e.g., ADEA 2003). The findings onprocess variables are also similar to those inother studies; the real issue here is to discover the underlying support and incentive structures that motivate teachers andschool directors to adopt the desirable behaviors. 5.40 Student achievement in and costs o f alternative basic education. As noted in chapter 2, some 500,000 pupils inEthiopia are estimated to be attending altemative basic education centers that provide the equivalent o f the first four grades o f primaryschooling."' In2000, the Ministry o f Education conducted a small study to evaluate whether the altemative routes "provide promise interms ofenhancing accesswhile realizingat least equivalent levels ofquality andefficiency as government schools.''112 The study documented the practices in 11 centers run by various nongovernment organizations (one o f them in collaboration with the local government), and administered achievement tests to pupils in grades 2 and 3 in seven o f the centers as well as to counterpart pupils innearby government school^"^ Because the study's research design did not involve random sampling and the samples o f centers and pupils were very small, its findings must obviously be treated with great care.114Nevertheless, the results raise questions that are highly relevant to discussions about educational access and performance-even if the picture they paint may require additional research to confirm them. The study finds that grade 1dropout rates at the centers are only a third as high as in the nearby government schools and that most pupils persist to the end o f the cycle; some even transfer to government schools for continuation in grade 5. Further, pupils in the centers tested reasonably well relative to their government school counterparts, obtaining overall scores on the four subjects, scores that were on average higher by 18 percent ingrade 2 andonly 7 percent lower ingrade 4.115 5.41 On the cost side, the basic education centers involve much lower capital and recurrent costs. At the time o f the study, the cost o f constructing a standard first-cycle govemment school (four classrooms, a pedagogical center, and a staff room), using standard materials (i.e., hollow blocks, concrete, or stone), ranged from 400,000 Birr in reasonably accessible areas, to three times as much in the more remote sites. By contrast, basic education centers comprising the typical two to four classrooms built with local materials, andwith active community contribution o f labor and some materials, may cost as little as 8,000 to 46,000 Birr; when the construction 'loAs notedelsewhere, ingrades 5-8, only about 26 percentof government teachers are certified as graduates of teacher training colleges (TTCs); in grades 1-4, more than 95 percent of the govemment teachers meet the requirement of being certified as graduates of teacher training institutes (TTIs). Becauseof the overwhelmingshare of TTI-certified teachers in grades 1-4, the impact ofteacher certification at this level was not evaluated. 111The curriculainthe nonformal centersare typically designedto cover in three years the content of what i s taught in four years in the government schools. In some of these centers, preschoolis offered, so the number of years of schoolingmay in fact be equivalentbetweenthe two systems. "2 See Govt. of Ethiopia(Ministry of Education, 2001). The test instruments covered comprehension in four local languages, English, mathematics, and environmentalscience, and were prepared by experts from the National Organization of Examinations and the Regional Curriculum Development and ResearchDepartmentsinthe participatingregions. 'I4The achievementtests were administeredto atotal of 387 grades 2 and 3 pupils inthe centers and to atotal of 449 counterpart pupilsingrades 2 and4 innearbygovernment schools. Gaps are computedfrom table 4.6A inGovt. of Ethiopia(Ministry of Education, 2001) anddiffers slightly from percentages reportedinthe text. - 146- uses hollow blocks, the costs only double. For even the most expensive altemative basic education centers, the construction costs rise to no more than one-sixth the cost of a standard government school. Regardingrecurrent costs, altemative basic education centers typically hire teachers who may be 1Oth or 12thgrade completers without formal teacher training certification, and pay them perhaps no more than half the salaries received by TTI-certified teachers in government schools. By economizing on teacher costs, the centers not only are able to operate with much lower pupil-teacher ratios than those ingovernment schools, but can also spend more on pedagogical inputs, all the while keeping cost per pupil about comparable to that in regular government schools. The mix of inputs is more adapted to rural populations, and possibly more effective pedagogically. 5.42 Thus the economics of altemative basic education centers appear attractive. However, the study did note that because these centers operate on a small scale, they may not be easily replicated to meet the country's vast needs inbasic education. Even so, the centers provide cost and performance benchmarks that are useful in assessing options for expanding educational opportunities inrural areas; and the innovative aspects o f their operations may contain ideas for improving service delivery in government schools. An important fact i s that teachers are recruited locally and paid at a rate consistent with local labor market conditions-although the payment i s modest relative to the pay scale for government teachers, it i s sufficient to motivate and retain the teachers. Just as important, the teachers invariably receive initial training and regular follow-up training, and they benefit from continual support during the school year. Finally, a high level o f community engagement exists: through participation in construction activities, through involvement in decisions about the school calendar and daily schedule o f classes, and through on-site supervision of attendance by teachers and pupils. It may be noted that inone form or another, these beneficialaspectso f the basic education centers inEthio ia are inevidenceinmanyother countries that have soughtto bringservicesto ruralpopulations. P,6 AdditionalResultsBasedon Grade8 ExaminationPerformancein Oromiya 5.43 To complement the foregoing studies, we compiled a special dataset inthe context o f the present study to examine more closely the relation between spending and learning outcomes, as well as the role ofteacher certification. The dataset pertains to a stratified random sample o f 314 schools in Oromiya chosen from the 1,037 schools in the region that enrolled pupils in grade 8. Data from the 2001-02 school census on the basic profile o f these schools were merged with information for the same year on grade 8 regional examination results for individual students in the schools; they were also mergedwith information on the teachers who taught at the schools. As with the data usedinthe studies described earlier, the Oromiya dataset has its flaws. The fact that it was pieced together from existing data rather than created from scratch means that not all the information desired inan evaluation o fthe determinants o f student learning were a~ai1able.l'~ 'I6A few of these are described in the study, including the Bangladesh Rural Advancement Committee (BRAC) schools in Bangladesh; the Escuela Nueva schools in Colombia and Guatemala; Instructional Management by Parents, Communities and Teachers (IMPACT) schools in the Philippinesand Indonesia, Malaysia, Jamaica, Liberia, and Bangladesh; and the community schools inEgypt. Examples of flaws in the data include the following: (a) it was not possible to match individual students to the specific teachers who taught them; (b) there was no information on process variables, such as whether the school had a parent-teacher association, and how frequently teachers and parentsmet; (c) there was no information on how intensivelyand well the teachers usedpossibly important educationalresources, such as radio-assisted instruction; and (d) the fact that learningoutcomes are not measureddirectly but throughgrade 8 examinationresults. - 147 - The results reported below should thus be treated with caution and used simply to stimulate a healthy debate regarding the transformation o f resources into learning outcomes in Ethiopian schools. 5.44 Overview o f examination scores. Table 5.14 summarizes the data on grade 8 examination results, the proxy for learning outcome inthe analysis here. Some o f the sample schools serve as examination centers for evening and private students, but such students were relatively few in number and were excluded from the main analysis."* The overall score averaged only 33 percent for the eight subjects included in the examination, and the score was somewhat higher among boys than girls. More than three-quarters o f those taking exams passed the examination-the passing mark i s typically calibrated to generate a pool o f qualified students for entry to grade 9 that more or less matches the number ofplaces available inthe system. Table 5.14. Grade 8 Examination Resultsfor a Sample of Pupils inGovernment Primary SchoolsinOromiyaRegion,Ethiopia, 2001-02 Evening Regular Result student Private student student All Average scorea' 25.3 26.8 33.0 32.8 Girls 23.9 26.9 31.0 30.9 Boys 26.0 26.6 34.1 33.9 Pass rate 38.3 36.3 83.0 81.9 Girls 26.2 34.7 75.3 74.1 Boys 44.9 37.7 87.4 86.4 Memoranda: No. of observations 551 201 30,393 31,145 Girls 195 95 10,960 11,250 Boys 356 106 19,433 19,895 5.45 Regsession analvsis. The learning environment i s defined by tangible features, as well as by processes that work inthe background. Typically the tangible features constitute what might be called the mix o f school inputs-availability of books, teacher qualification, pupil-teacher ratios, the physical environment, and so on; the background processes refer to classroom management practices, which may be shaped by such factors as the availability o f technical support for problem solving, accountability or incentive structures that motivate behavior, information flows that improve management, and so on. Because the choice o f school inputs has explicit and direct cost implications, policymakers are justifiably interested in the relation between the inputs and learning outcomes. A particularly relevant input in this regard is the certification o f teachers. As earlier sections have noted, the national policy in Ethiopia calls for teachers in grades 5-8 to be graduates of teacher training colleges (TTCs), and for teachers in grades 1-4 to be graduates o f teacher training institutes (TTIs). In practice, however, a large 'IsIn passing, it is noteworthy that evening and private students achieve comparable results; that girls fare worse than boys amongthem; andthat bothgroups lagsignificantly behindthe regularstudents. - 148 - number o f the TTI-certified teachers are assigned to teach classes at grades 5-8, andthe question is whether they are less effective than their better certified colleagues. 5.46 The regression analysis reported in table 5.15 i s an attempt to answer that question. Two specifications are shown: one uses individual students as the unit o f observation, the other uses schools. Recognizing that teacher certification i s not the only factor that influences performance, the regression equation includes other variables that are commonly used in student achievement regressions, including a student's personal characteristics, pupil-teacher ratios, availability o f books, and teacher experience. With regard to the results, the coefficient estimates are generally consistent between the two regressions interms o f magnitude and sign. But only three variables consistently achieve statistical significance in both regressions, namely, age o f the student, gender, and teacher experience. A curious fact i s that teachers with longer experience tend to be less effective, contrary to what one might expect. One possible explanation is that inthe absence o f continual in-service teacher training, the older teachers may not have kept up with changes in the curriculum and are thus not as effective as their younger colleagues. Again contrary to expectations, the results suggest that TTI-certified teachers are, if anything, slightly more effective than TTC-certified teachers; the difference between them is nonetheless very modest as the coefficient implies that raising the share o f TTI-certified teachers by 10 percentage points andreducing the share o f TTC-certified teachers to a similar extent would improve scores by a mere 0.1 percentage points above the sample average o f 33 percent. - 149- Table 5.15. Correlates of Average Grade 8 Examination Scores, Oromiya, 2001-02 Regression with individual Regressionwith schools as units of students i units of obs .vations&' lbservation a' Sample Sample Regression variable average Coefficient t-statistic average Coefficient t-statistic Student characteristics Age (inyears) 14.77 -0.579 ** -23.16 14.88 -0.65 * -2.39 Boy(girl)b/ 0.64 2.487 ** 34.24 0.72 3.971 * 2.02 Oromo as language of instruction (Amharic)b' 0.86 3.584 ** 34.35 0.95 1.700 1.21 Teacher characteristics" Percentagewith 10+1 or 10+2 0.42 -0.106 ** -8.75 0.26 -0.087 -0.56 Percentagewith 12 or 12+1 1.49 -0.026 * -2.53 1.25 -0.009 -0.13 Percentagewith TTI certification 87.23 0.010 ** 3.50 87.5 1 0.029 1.67 Average years of service 12.58 -0.205 ** -18.99 12.17 -0.174 ** -2.95 Other school characteristics Pupil teacher ratio' 53.17 -0.013 ** -6.04 54.68 -0.006 -0.44 Numberofbooks per pupil 2.76 0.097** 4.43 2.55 0.35 1 1.93 School located inurban area 0.42 0.681 ** 7.42 0.41 0.480 0.95 Intercept n.a. 39.407 72.64 n.a. 38.600 7.26 RZ n.a. 0.146 n.a. n.a. 0.140 n.a. Number ofpupils insample 29,905 n.a. n.a. n.a. n.a. n.a. Number o f schools insamplee' 314 n.a. n.a. 272 n.a. n.a. n.a. Not applicable. ***Statisticailysignificantcoefficient at the 5 percent confidence level. Statistically significant coefficient at the 1percent level. a. The dependent variable inboth regressionsrefers to students' average scores on the following eight subjects: Afan Oromo, Amharic, mathematics, English, biology, chemistry, physics, and social science; for studentsnot taking Afan Oromo, the average is computed usingthe scores for seven subjects. The sample average i s 33.0 inthe regression usingstudents as the unitof observation; and 34.6 inthe regression using schools. b. D u m y variable, with the omitted category indicatedinparentheses. c. Teachers refer only to those assignedto teach grades 5-8; omitted category are teachers with teacher trainingcollege certification; teachers with 10+1 or 10+2 refers to those with one or two years of schooling beyond 10" grade; and teachers with 12 or 12+1 refer to those with 0 or one year beyond the 12" grade; teachers withTTI refer to those with teacher training institute certification. d. Ratio refers to pupils and teachers in grades 5-8. e. The sample for the regression using schools as the unit of observation is smaller because it includes only schools whose average grade 8 examinationscore can be computed from the scores of each of the grade 8 students who took the examination. Source: Analysis o f data from arandomsample of some 314primary schoolscollected and coded by the Oromiya Bureau of Education in the context of the present study. 5.47 The scope for better management. While the analysis presented above can be refined ina numbero fways, the results, ifconfirmed, are truly disturbing. They implythat differences across schools in some o f the tangible factors often considered important account for very little o f the differences in perfonnance across students or schools. A moment's reflection reveals why this puzzling result might not be as strange as it first appears. While the tangible aspects o f schools define the infrastructure for teaching and learning, they alone cannot yield good results unless - 150 - the human agents put incharge of them are properly motivated, supported, andheld accountable. Indeed, it is conceivable that even the best-equipped school would fail in a poor management environment. The scope for improvement inthe sample o f Oromiya schools considered above i s illustrated in figure 5.11 below. Inthe figure, the tangible factors for each school are converted into financial terms and expressed as spending per student, and the result is plotted against the average score of students in the school at the grade 8 e~amination.''~As anticipated, the figure shows hardly any relation between the two variables: schools with relatively high levels o f recurrent spending arejust as likely to outperform as lag behind less well endowed schools. This pattem, which is familiar from similar analyses in other countries, confirms the weak relation between school inputs and learning outcomes already documented above. It identifies a clear need for better management, and it suggests that while the choice o f school input mix i s important, progress will probably also require reforms to increase accountability for results throughout the system. Figure5.11. RelationbetweenSpendingPer Pupiland MeanGrade 8 ExaminationScore acrossA Sampleof Primary SchoolsinOromiyaRegion,Ethiopia, 2001-02 45 11 0 0 0 0 0 0 0 0 00 0 e 00 40 0 00 0 0 35 30 0 25 50I 100 I I 200 300I 400I Average spendingper student (current Birr) Note: Each circle in the figure representsa sample school. The mean score refersto the average score (percentagecorrect) of studentsin each sch. the following eight subjects: Afan Oromo, Amharic, mathematics, English, biology, chemistry, physics, and social science; for students not t Afan Oromo, the average is computedusingthe scores for seven subjects. Source: Analysis of Grade 8 examinationscores for students in 314 primary schools merged with information for those schools from the 20( school census for Oromiya,and with additionalinformationon teachersin those schools. Issues for Policy Development 5.48 Ethiopia faces many challenges in managing the delivery o f services in education. The analysis presented in this chapter highlights two issues that warrant special attention. The first concems the structural policies, particularly those pertaining to standards for teacher recruitment 'I9The salaries of the teachers and their teaching assignment are known, which makes it possible to compute the aggregate teacher wage bill for grades 5-8. The other recurrent costs-salaries of administrative staff and spending on items other than salaries-are reportedfor the whole school and are prorated to each subcycle accordingto the size of enrollments.The resulting aggregate spendingis dividedby the enrollmentsingrades 5-8 to obtain the spending per pupil at this level. - 151- and school construction, that help define the tangible characteristics o f the supply o f services. The second has to do with management weaknesses in the system that show up in the poor allocation o f human and financial resources across schools, and apparently also in the schools' disparate effectiveness in using the resources to deliver learning outcomes. Below we discuss policy options pertaining to these issues, drawing on international experience for an added perspective. 5.49 Setting appropriate standards for teacher recruitment and school construction. InEthiopia as inother countries, these standards create a network o f schools with some key commonalities. While they are intended to help schools function well, the opposite effect may infact materialize when budget constraints are ignored in the choice of standards. If standards exceed what the country can afford, schools may meet requirements in the areas explicitly specified, but do so through cutbacks elsewhere that eventually compromise their ability to provide effective services.12' That this has indeed happened in the Ethiopian system can be seen from the following features: (a) a severe shortage o f teachers and classrooms throughout the system leading to both very highpupil-teacher ratios (among the highest inthe world) and section sizes; (b) a sizable share o f schools with very large enrollments; (c) large shares o f teachers ingrades 5-8 and 9-12 who do not meet the certification standards; and (d) very limited spending on school administration and nonsalary pedagogical resources. Figure 5.12 shows that all levels o f the school system have suffered a persistent deteriorationinpedagogicalconditions since 1994. 5.50 Regions such as Amhara, Oromiya, and Tigray have responded to the deterioration by introducing paraprofessional (ie., uncertified) teachers to relieve the teacher shortage, and encouraged the emergence and growth o f lower cost alternative basic education to serve remote rural populations.'21 While such adjustments at the marginwill certainly help a school here and there, an improvement for the whole system will require a systematic review of national recruitment and construction standards. Such a review can be facilitated by simulation models to test the impact o f alternative options. One such model estimates, for example, that inthe absence o f changes inrecruitment standards-and therefore inthe cost of teachers-Ethiopia would need to spend 4.5 percent o f its GDPjust to attain the Millennium Development Goal o f 100 percent primary school completion by 2015.'22 Since the country's recurrent spending on all levels o f education i s only about 3.0 percent o f GDP, that very few low-income countries are able to mobilize more than 4.0 percent o f the GDP for the whole sector, the fiscal unsustainability o f the current standards becomes very clear. As already mentioned, the scope for reducing construction costs is also large; alternative school designs yield an estimated cost reduction o f up to 60 percent over certain construction designs currently inuse.123 I2O Teacher recruitmentandconstructionstandards have importantimplicationsfor curriculum design. This complextopic cannot be addressedfully here, but we note inpassingthat what teachersare expectedto teach must obviouslybe calibratedto what they CQn teach, and this in turn depends on their academic profile. Further, the need for supplementary learning space, such as laboratories,particularlyinthe lower grades, will dependon what is requiredinthe curriculum. 12' Suchteachers are typically paid at about halfthe rateof the regular teachers; inAmhara, the cost of paraprofessionalteachers i s bomeby the regionalgovernment, while in Oromiyait i s borneby the communityinwhich the school is located. 12* See World Bank/MOE (2004a) for adescription o fthe modeland an explanationon its use as a tool for policy evaluation. See McArthur (2003) andCoyle (2003). - 152- Figure5.12. Trends inPupil-Teacher and Pupil-Section Ratios and inNonsalary Public Spending Per Primary and SecondaryStudent,Ethiopia,1990-2001 90 9 8 0 0 Note: The PTRis conputedbydividingtotalemllmntsbythe totaln& ofteachers, andthe PSRbydividingtotalemlmnts bythe total numberofsections. Thesecalculationsyi eldweightedratioswhichdifferfromthe unweightedratiosconputed(as intable 5.10) byaveraging the ratiosacrossschools. Non -salarypublic spendingper student is conputedbydividingthe reportedbudgetmunt for thu itembythetotal numberof pirraryandsecondarystudentsingovemnt schools. Source:basedondatasuppliedbytheMinishy ofEducationone " t s , teachers andsections;andonMOFEDbudgetdataonexpenditure. 5.51 Allowinn for flexibility around the national standards. Flexibility in this regard is particularly relevant in Ethiopia because conditions differ so widely across the country, particularly between urban and rural areas. To appreciate the need for adapted standards, consider first the disparate trends since 1994-95 in the pupil-teacher and pupil-section ratios (PTRs and PSRs) in urban and rural areas (figure 5.13). Rural schools began with an advantage intheseratios in 1994-95, but the advantagehad completely disappearedby2002-03. Indeed, a highly anomalous situation now exists, inwhich the average pupil-teacher ratio inrural schools exceeds that in urban schools (71.7 compared with 51.5, a difference o f nearly 40 percent); and their average pupil-section ratio is likewise greater (74.1 compared with 69.7, a difference of 6 percent). These patterns are unusual because population densities are typically lower in rural areas. They exist largely because in the past, uniformly unaffordable standards for teacher recruitment and school construction had been applied throughout the system regardless of conditions at the local level. - 153 - Figure 5.13. Pupil-Teacher and Pupil-Section RatiosinUrban 80 , andRural PrimarySchools,Ethiopia, 1994-2003 I Urban PSR 70 - Rural PSR 20 c I I I , I I , I m VI \o b 00 m m ;2 3 ; wI 00 I 2 2 2 3 m m m m m m 0 2 2 5 2 2 0 0 ct 0 N 0 N Note: Because data are available separately for govemment and nongovemment schools only after 1997-98, the figure uses data for both sectors in all years. Data after 1997-98 indicate that the difference i s small between the govemment-only series and that for both sectors. Source: Based on data supplied by the Ministry of Education on number of pupils, teachers, and sections. 5.52 The need for flexible standards is reinforced when one considers the unintended outcomes o f past policies. As shown above, applying uniform standards has caused the pupil- teacher ratio to rise and diverge widely between urban and rural schools. We show below that it also causes teaching loads to rise over time, particularly in rural schools. The increase is such that teachers in rural schools now teach as many hours as students spend in class, leaving little scope for the specialized teaching envisaged by the curricula for grade 5 and beyond. These trends make it increasingly difficult both to attract teachers to rural areas and to rationalize the overall deployment o f teachers across schools. 5.53 For policy development, it is important to understand the precise sources of the adverse consequences described above. Consider then the following algebraic tautology, which simply states that the aggregate weekly hours o f instruction that pupils in a school receive (left-hand side) is exactly the weekly hours o f teaching suppliedbythe teachers (right-handside): - 154- NS-SH = T-TW where N S is the number of sections of pupils, SH is the number of instructional hours that pupils receive, and T i s the number of teachers. Given that N S is the number o f pupils divided by the number ofpupilsper section (i-e., PSR) andPTR is the numberofpupils dividedbythe number o f teachers, the above equation can be expressed in terms o f pupil-teacher and pupil-section ratios, teacher workloads (TW), and students' instructional hours, as follows: - = - PTR TW PSR SH 5.54 Based on the foregoing relationship, we can easily relate teachers' teaching loads to pupils' instructional time using the relatively plentiful data on the pupil-teacher and pupil- section ratios. Between 1994-95 and 2002-03, teaching loads relative to pupil hours in urban schools rose from 0.55 to 0.74, an increase o f 33 percent; in rural schools, it rose from 0.77 to 0.97, a gain of 25 percent (table 5.16). Although the percentage rise is smaller inrural schools, a ratio that is close to unity implies that by 2002-03, the majority of teachers inrural schools had most likely ceased teaching as subject specialists, even though they may have done so inthe past. In both 1994-95 and 2002-03, the ratio between teaching load and instructional hours was significantly greater inrural thanurban schools, by 40 percent and 31percent, respectively; these trends imply that just as specialized teaching has becomepracticallyimpossible inrural schools, ithas also becomeincreasingly difficult to practice inurbanschools. Table 5.16. Pupil-Teacher and Pupil-Section Ratios and Teaching Loads Relative to Pupils' Instructional Hours in Urban and Rural Government Primary Schools,Ethiopia, 1994-95 and 2002-03 Ratio between teachers' teaching load Pupil-teacher ratio Pupil-section ratio and pupils' instructional hours" Index (urban Yearlregion Urban Rural Urban Rural Urban Rural =1.00) 1994-95 34.1 31.7 61.7 41.1 0.55 0.77 1.40 2002-03 51.5 71.7 69.7 74.1 0.74 0.97 1.31 5.55 When the ratio between teaching load and instructional hour rises, an interestingquestion i s how the numerator (i.e., TW) relates to the denominator (Le., SH). If, for example, schools in both urban and rural areas follow the curriculum strictly and offer the same weekly number of instructional hours, the higher ratio inrural schools must implythat rural teachers carry a heavier teaching load than their urban counterparts. Data on actual teaching loads for teachers in Oromiya confirm the pattern: whereas in urban schools, teachers average 19.5 hours a week of classroom teaching, the load is 27.2 hours a week inrural schools-an astonishing difference of 39 percent (see table 5.17). Giventhe large gap, i s it any surprise that it i s hardto attract teachers to rural schools and keep them there? In Oromiya, graduates fresh out o f teacher training -155 - colleges typically accept postings to the remotest rural schools, but most seek a transfer to schools nearer a town as soonas they have gained sufficient seniority. Table 5.17. Average Weekly Teaching Loads by Teacher Certification and Grade inWhich Teacher is Teaching, Oromiya, Ethiopia, 2002-03 (inhours) ~~~~ Teacher Certification Locality & grade to which teacher TTC and i s assigned 10+1; 10+2 12; 12+1 TTI equivalent All teachers Oromiya 17.7 21.0 24.2 18.1 23.8 Grades 1-4 18.0 27.4 26.0 22.2 26.0 Grades 5-8 17.5 16.2 21.2 18.0 20.6 Urban schools 17.5 16.6 19.9 17.0 19.5 Grades 1-4 18.1 21.6 21.6 18.1 21.5 Grades 5-8 17.1 15.8 17.9 17.0 17.6 Rural schools 23.0 27.7 27.3 22.3 27.2 Grades 1-4 15.0 28.9 28.6 28.9 28.6 Grades 5-8 27.0 20.3 24.5 21.9 24.3 Note: Refers to govemment teachers only. Source: Analysis of Oromiya teacher census for 2002-03. 5.56 If instructional hours are relatively uniform across schools in urban and rural areas throughout the country as they are in Oromiya, differences in the teaching loads of teachers in urban and rural schools can be inferred from data on pupil-teacher and pupil-section ratios, based on the algebraic relationships discussed earlier. Inparticular, the average teaching load of rural teachers as a multiple o f the average load of urban teachers is simply the ratio between teacher's teaching load and pupils' instructional hours in rural schools relative to the corresponding ratio for urban schools. The results across regions in 2002-03 appear inthe last column of table 5.18 below. The diversity across regions is striking: workloads for ruralteachers are comparable to those for urban teachers inAddis Ababa; they are between 11and 16 percent heavier in Tigray, Benshangul-Gumuz, Harari, and Dire Dawa; and they are as much as 67 percent heavier inGambella. - 156 - Table 5.18. Pupil-Teacher and Pupil-Section Ratios and TeachingLoadsRelativeto Pupils' Instructional HoursinUrban and Rural GovernmentPrimary Schools, by Region,Ethiopia, 2002-03 Zatio betweenteachers'teachingload Pupil-tea ier ratio Pupil-sec .onratio andpupi ' instructioi .1hoursa' Index (urban Region Urban Rural Urban Rural Urban Rural =LO) Ethiopia 51.5 71.7 69.7 74.1 0.74 0.97 1.31 rigray 59.8 59.7 67.1 60.2 0.89 0.99 1.11 Afar 33.9 25.5 56.0 32.1 0.61 0.79 1.31 Amhara 60.3 74.7 74.4 74.5 0.81 1.oo 1.24 Oromiya 55.8 75.2 74.4 75.8 0.75 0.99 1.32 Benshangul-Gumuz 51.8 48.8 72.2 59.8 0.72 0.82 1.14 SNNPR 51.5 73.7 76.4 79.4 0.67 0.93 1.38 Gambella 31.9 41.2 70.3 54.5 0.45 0.76 1.67 ;NNPR Southem Nations and Nationalities Peoples Republic. Vote: Data for Somali excluded because unavaiiable; data for Addis Ababa, Harari, and Dire Dawa excluded because of the largely urban character of these localities. a. Data derived by taking the ratio between the pupil-teacher ratio and the pupil-section ratio, as explained inthe text. Source: Analysis of data from the Ministryof Education's school censuses for the year indicated. 5.57 Consistent with the wide urban-rural gap inteaching load inGambella, the region (which has a total of only 108 schools offering primary school instruction) also has the greatest difficulty among all the regions inrationalizingthe deployment o f its government primaryschool teachers across schools. As indicated earlier in this chapter, we can use the R2statistic from a regression relatingnumbers o f pupils and teachers across schools to compute 1-R2as a measure of the degree o f randomnessinteacher deployment. Figure 5.14 relates this index of randomness to the teaching loads o f rural teachers relative to their urban counterparts. The positive link between these variables is what one might expect: the more disparate the work conditions between urban and rural areas, the more difficult it i s to rationalize teacher deployment across schools. - 157 - Figure 5.14. Relation between Relative Teaching Loads in Rural and Urban Schools and Randomness in Teacher Deployment acrossRegions, Ethiopia, 2002-03 0.75 1.oo 1.25 1.50 1.75 2.00 Ruralteachers' teaching loads relative to those of urbanteachers Note: Eachpoint denotes a region; Somali excluded for lack ofdata; and Harari, Addis Ababa, and Dire Dawaexcluded because of their largelyurbancharacter. Source: Basedon data intables A5.5 and 5.18. 5.58 Assessing the tradeoffs in teacher recruitment options in rural areas. Setting fiscally unsustainable standards typically sets o f f a vicious cycle o f worsening conditions that eventually harms rural schools the most. In 1994-95, the standards did not seem overly demanding, but as enrollments grew andthe demand for teachers rose, the effects o fbudget constraints began to set in, even as the government started to commit more resources to education. Yet the increase was constrained by competing demands on government funds, and the education system adapted by default through increases in pupil-teacher and pupil-section ratios as well as in teaching loads. Teaching posts inrural areas are typically less attractive to begin with than those inurban areas: there is no electricity or running water, no social services, entertainment, or other amenities to provide for a comfortable family life; the opportunities for teachers to eam extra income by teaching in evening classes are also limited, if not entirely absent. But under the circumstances, they have become even more unattractive because teachers in rural schools carry substantially heavier teaching loads than their urban counterparts. Indeed, in terms o f earnings per hour o f classroom teaching, rural teachers currently earn only about 75 percent as much as similarly certified teachers inurban schools. This fact is not lost on teachers, andit exacerbates the already difficult task o frecruiting and retainingteachers inrural schools. 5.59 Relieving the staffing pressures in rural schools will require a recruitment policy adapted to conditions inrural areas. One option involves incentive packages that are attractive enough to hire andretain the desired number o f certified teachers inrural areas. Naturally, the issue o f costs arises. If our estimates are a guide, teacher wages for rural postings would have to rise by as much as 30 percent, on average, just to compensate for the bigger workloads in rural schools, andthey would needto rise even more ifquality-of-life differences were also taken into account. The question is whether the resources can be found to pay for these packages, and whether the - 158 - money thus spent will have been put to the best use in terms o f schooling outcomes. A second option currently implemented throughout the country (albeit inan ad hoc manner), particularly in the centers offering alternative basic education, involves hiringteachers at lower cost from local labor pools who satisfy minimum standards o f certification (e.g., grade 10 or grade 12 graduates rather than TTI graduation). The issue here i s quality: can such teachers teach the curriculum; can they be trained to do so; andmost important, are they effective? Inthe end the choice comes down to these tradeoffs: increasing public spending dramatically to put sufficient numbers o f teachers that meet national standards into rural schools so that these schools can function properly, versus adapting teacher certification and curriculum standards to local conditions in rural areas inorder to hire the desired number o fteachers within the available budget. 5.60 Taking advantage of decentralizationbased on lessons from international experience. As mentioned above, Ethiopia is currently undergoing a profound decentralizationo fthe structure o f government. Budgets are allocated to the regions in block grants, and the regions in turn are expected to make block grants to the woredas governments. In the new governance landscape, these subnational governments are directly responsible for service delivery in primary and secondary education (as well as some other subsectors). As noted in this chapter, two major challenges loom large for them in this regard, namely, improving discipline inthe allocation o f resources across schools, so that human and financial resources are effectively channeled to the front lines (Le., the school); and reducing the disconnect between resources and learning outcomes. Although the process o f decentralization is at an early stage, and roles, responsibilities, and governance frameworks are still being defined and refined, Ethiopia might well note the accumulated experience o f other countries who are further along in education decentralization. Winkler and Gershberg (2003) distill the following lessons based mostly on experiences inLatinAmerica and Eastern Europe: "Efficiency and effectiveness are most likelyto improve under decentralizationwhen service providers-schools, local governments, or regional governments-are held accountablefor results. Accountability requires clear delineation of authority and responsibility and transparent and understandable information onresults (both educational and financial). Decentralizationo f real decisionmakingpower to schools or school councils can significantly increaseparentalparticipation inthe school, andhighlevels o fparental and community participation are associatedwith improved school performance. Decentralizationo f educationto sub-national governments does not inand o f itself empower parents and improve schoolperformance. Further decentralizationto schools (school councils or school boards) or local communities does empower parents and can improve school performance. For decentralization to schools to be successful, principals must acquire new skills in leadership andmanagement-financial, o f teachers, andwith the community. The design of'nancial transfers to sub-national governments or schools has powerful effects on both efficiency and equity. Decentralization requires that national and/or regional ministries of education be restructured; failure to restructure ministries is a serious obstacle to realizing the benefits o f decentralization. - 159 - 0 The decentralizationo f teacher management is critical to creating accountability and realizing the potentialbenefits o f decentralization. 0 National education ministries frequently resist decentralization on the grounds that sub- national governments, communities, and/or schools lack the capacity to manage education. Inpractice, this i s seldom true. 0 Real decentralizationis a long, evolutionary process." 5.61 Bearing in mind Winkler and Gershberg's last point, one sees clearly that decentralization in Ethiopia will take time to produce the results expected o f it. While not all aspects o f other countries' experiences are applicable-or indeed, even useful-the concepts are highly relevant and the lessons from international experience can already be incorporated to cultivate the key characteristics and behavior patterns that characterize high-performance systems: clarity of mission, roles, and responsibilities; effective empowerment o f the relevant actors; greater parent participation in schools; and, above all, an emphasis on accountability for results. The Ethiopian government has already set the process in motion, as reflected in its recently issued "Guidelines for Organization o f Educational Management, Community Participation and Educational Finance" (Govt. o f Ethiopia 2000d). Regional and local governments are moving rapidly to claim and exercise their new powers, for example, by adapting policies on teacher training, recruitment, and remuneration, as well as school construction to local realities. In this evolving context, it will be important for the central government to provide flexible guidelines without stifling local initiatives, to encourage problem solving and learning-by-doing through systematic evaluation o f new initiatives, and to facilitate the exchange o f ideas for improving the system's performance at all levels ofmanagement. 5.62 Mobilizing and using information to enhance accountability for results. Strengthening accountability for results is difficult in schooling because many actors are involved and the outcomes are typically multidimensional. Yet appr.opriate use o f information flows can help improve performance by ensuring that resources are at least reaching schools. InUganda a large- scale public dissemination o f information on grants for schooling gave schools and parents the information they need to monitor the grants. Over the space o f just six years, the leakage o f funds fell from 80 to 20 percent, and schools were able to use the resources now reaching them to purchase textbooks and other pedagogical inputs (World Bank 2004a). Information flow can o f course be channeled in other ways too; report cards such as those used in Parana State in Brazil are an interesting example (Servilla and Winkler 2003). They provide schools with information that compares indicators for the school (e.g., pupil-teacher ratios, examination results, and parent feedback) with reference averages at the national and relevant subnational levels. Report cards for school districts and other higher levels o f aggregation can also provide similar comparative information. The availability o f such information may be expected to improve conditions in schools intwo ways: by creating incentives for underendowed schools to seek redress; and by pinpointing specific schools inthe system that warrant the attention o f those incharge. InEthiopia, as inother countries, the use of information on student achievement for accountability purposes remains at an early stage. The country has already established a baseline, and this can be built on as part of the broader effort to mobilize appropriate information for better management. - 160 - Conclusion 5.63 Ethiopia's education system is poised to continue in the coming years as the country makes progress toward achieving the Millennium Development Goal o f universalizing primary school completion by 2015. If in that year all the children ages 7-10 were enrolled, the system would be catering to nearly 11million pupils injust grades 1-4 alone, up from about 6 million today. Insuch a large system, the challenges o f good management become even more daunting. Ideally, access and learning outcomes must both improve without putting a heavy burden on the public purse. While the solutions to these complex challenges will require a deeper assessment than has been possible here, this chapter has highlighted the need to adapt national standards, particularly for teacher recruitment and school construction, so as to avoid continued and probably inevitable deterioration o f the country's already poor conditions in schools. These conditions are manifested in exceptionally high and rising pupil-teacher ratios, overcrowded classrooms and schools, and a dearth o f instructional materials. At the same time, the process o f decentralization i s a very promising development, but its ability to improve the education system's perfonnance will require, among other things, changes that give parents and communities a real say inhow schools are run, increase accountability at all levels inthe system, and provide appropriate information flows to give concrete meaning to the concept o f accountability. - 161- Chapter 6. Aspects ofthe Market andNonmarketBenefitsof Schooling 6.1 This chapter brings together some key aspects o f the market and nonmarket benefits o f education. It documents the impact o f education on agricultural productivity and on workers' wages inthe modem sector; consolidates available information on the relation between education and poverty, fertility and child health and nutrition; and sheds light on the transition between school andwork among recent labor market entrants so as to provide an added perspective on the implications for policy development. While admittedly not exhaustive, the information presented creates a basis for dialogue on ways to ensure that investments ineducation respond to the economy's requirements for trained labor, while also promoting widespread social well- being. 6.2 Inbrief, the assembled evidence offers clear support for the government's intention to make education a centerpiece o f its strategy for economic and social development. Because o f past neglect, the average Ethiopian adult has fewer than two years o f formal schooling. This weak humancapital base has not only impeded economic productivity among farmers and other workers but also made it difficult to reduce poverty and improve the country's social indicators. Raising the educational profile o f the population is thus critical for the country's future progress. Inpractice, it makes sense to expand the education system instages: universalizing at least four years o f schooling as soon as possible, which serves both economic and social goals, while expanding the other levels more slowly. This approach recognizes that tradeoffs are inevitable in the allocation o f ultimately limited resources for education and that the economy may also face constraints inits capacity to absorb largenumbers o f educated labor, particularly inthe short run. EmploymentPatternsandWorkers' EducationalAttainment 6.3 To set the stage for discussing the extemal efficiency o f education, this section presents data on the structure of the economy and the educational profile o f the w ~ r k f o r c e . 'To ~ ~ summarize, more than 80 percent o f Ethiopia's workforce in2001 is employed inagriculture and most o f the workers are either self-employed or used as unpaid family labor. The latter category remains dominant, even though the share has shifted in favor o f self-employment since 1994. Literacy rates are generally low in Ethiopia: in 2001, seven in ten workers could not read or write. In agriculture, nearly 80 percent o f the workforce is illiterate, while the corresponding share inthe formal sector is about 28 percent. 6.4 m e g a t e pattems o f employment. Table 6.1 shows the size o f the labor market and selected indicators o f labor force participation. During the 1990s and in 2001, the share o f Ethiopians aged 10 years and older who were working or actively looking for ajob has remained relatively stable at between 72 and 75 percent. As is typical o f low-income economies, the rate o f unemployment is modest: in 2001, only 5 percent o f the active labor force reported being jobless, while the share was 3 percent in 1994 and 8 percent in 1999. Because o f differences in definitions and data collection methods, the data for 1994, 1999 and 2001 are not strictly comparable. They nonetheless offer a perspective on trends inlabor market developments over time. - 162- Indicator 1994 1999 2001 Population aged 10and above 36,626.4 36,022.2 36,795.9 Population inthe labor force 26,503.1 27,068.3 27,168.2 Employedpopulation 25,732.2 24,896.6 25,747.6 Labor force participationrate (%)"' 72.4 75.1 73.3 Unemployment rate (%lb' 2.9 8.0 5.2 6.5 Among those in the labor force, a significant share are children of school-going age: according to the 2001 Ethiopia Stand-Alone Child Labor Force Survey (ESACLFS), those ages 10-14 account for some 24 percent of the employed population; more than half o f these young workers are engagedineconomically productive activities (e.g. production of goods and services for sale or e~change).'~~Although the legal minimum age for starting work in Ethiopia is 10 years, nearly 40 percent o f those ages 5-9 already report being engaged in a productive activity; among boys, the share rises to nearly half (table 6.2). The significant rate o f labor market participation by school-age children is consistent with the pattem inother low-income countries. As discussed elsewhere in this report and confirmed by the data in the table, not all work by children is incompatible with schooling, particularly if the activities do not require a substantial time commitment. Thus, among children ages 10-14 who perform housekeeping and other unpaid domestic chores, two-thirds are enrolled for school (which is only slightly smaller than the share o f 70 percent among non-working children); and among those who work inproductive activities (with or without housekeeping or other household chores), the share enrolled for school i s still highat 43 percent. See table A.6.1 for percentages based on the 1999 Ethiopia Labor Force Survey. In general the percentagesare typically smaller than those basedon the 2001 EthiopiaStand-AlongChildLabor Force Survey-a pattemthat is perhapsto be expected given that the lattersurveywas intendedto ensure that children's work activitiesare fully captured. - 163 - Table 6.2. Labor Force Participation by School-Age Children, Ethiopia, 2001 Indicator Ages 5-9 Ages 10-14 Ages 5-14 Population (thousands) 8,625.4 6,533.1 15,158.5 Percent not working 25.4 4.2 16.3 Percent working 74.6 95.8 83.7 Inunpaid domestic activities only 35.4 33.0 34.4 Inproductive activities onlyb' 5.3 6.7 6.0 Inbothproductive and domestic activities 33.9 56.0 43.4 Percent enrolled in school among: Non-workmg chddren 19.3 69.9 24.9 Working children 26.9 51.5 39.0 Inunpaid domestic activities only 31.1 68.2 46.5 Inproductive activities only or inproductive anddomestic activities 23.1 42.7 33.9 Percent working inproductive activities Among boys 48.3 73.2 59.1 Among girls 29.7 51.8 39.2 Among children inurban areas 5.0 22.5 14.2 Among children inrural areas 42.9 69.8 54.1 a. Domestic activities refer to personal services of a domestic natureprovided by unpaid household child members in the householdof their own parents, grandparentsor guardians, or spouse. The services include such chores as preparing and serving meals; making, mending, washing and pressing clothes; shopping; caring for siblings or sick and infirmpersons in the household; cleaning and maintaining dwelling; transporting household members or their goods; and so on. b. Productive activities refer to the production of goods and services for sale or exchange; and production and processing of primary products (e.g. agriculture, hunting, etc.) for own consumptionor saleonthe market; and own accountconstructionor formationof other fixed asset. Source: authors' estimatesbasedon the 2001 ESACLFS. 6.6 Distribution o f emplovment by sector and type o f work. Between 1994 and 2001, the share o f Ethiopian workers in agriculture declined-by some 7 percentage points-while the share o f workers inindustry and incommerce and other services rose (table 6.3). Yet agriculture remains dominant, employing four out of every five workers. Thus, while the prospects for work elsewhere in the economy have improved between 1994 and 2001, such jobs are still relatively rare. - 164- Table 6.3. Distributionof Employmentby EconomicSector, Ethiopia, 1994,1999and 2001 1994 1999 2001 Yumber of Number of Percentage workers Percentage workers Percentage Economicsector share thousands] share thousands) share Agriculture, hunting, forestry and fishmg 89.3 19,869.1 79.6 21,143.5 82.1 Mining and quarrying 0.1 15.7 0.1 26.3 0.1 Manufacturing 1.8 1,107.2 4.4 914.9 3.6 Construction 0.3 228.5 0.9 162.6 0.6 Electricity, gas and water supply 0.1 28.0 0.1 33.9 0.1 Wholesale, retail trade and catering 4.2 2,341.9 9.6 1,948.1 7.6 Transport, storage and communication 0.6 122.9 0.5 111.0 0.4 Financial intermediation 0.1 19.6 0.4 16.8 0.1 Other services 3.6 1,163.7 4.4 1,390.6 5.4 All sectors 100 24,896.6 100 25,747.6 100 Sources: data for 1994, 1994EthiopiaPopulationand HousingCensus; datafor 1999,1999 LFS; data for 2001,2001 ESACLFS. 6.7 With regard to types of jobs, the data in table 6.4 indicate that most jobs are in the informal sector: self-employed and unpaid family workers account for about 90 percent o f all workers in2001, about the same as in 1994. The unpaidworkers continue to make up the single largest block ofworkers with a share o f about 50 percent inof all workers in2001. Salariedjobs remain rare, even though their share grew from about 6 percent o f all jobs in 1994 to nearly 9 percent in 2001. Salaried jobs requiring the highest qualifications-Le,, managerial jobs and professional andtechnicaljobs-represent only 1.4percent o f alljobs in 1999. - 165 - Table 6.4. Distributionof Employmentby Type, Ethiopia, 1994,1999 and2001 I Formal :ctor + 1999 Informa sector Bothse :ors No. of No. of workers workers workers Type ofEmployment ?A0 (`000) % Salariedjobs 3.9 2,031.8 8.2 2,212.8 Executives 103.0 0.0 104.8 0.4 109.8 Techniciandassociate professionals 190.8 15.4 0.0 199.6 0.8 255.9 Others 823.7 3.8 1,727.4 6.9 Apprentices 0.0 3.5 0.0 12.0 0.1 Employers 2.6 3.6 0.8 192.6 0.8 Self-employed 39.5 38.9 45.7 10,837.7 43.5 Unpaid family workers 51.1 20.0 49.4 11,697.5 47.0 Others I 1.1 I 55.4 0.3 124.9 0.5 376.0 ~ -Alltypes 100.0 1,238.9 100.0 24,896.6 100.0 Dataunavailable. Source: for 1994, the 1994EthiopiaPopulationand HousingCensus; for 1999, authors' estimates basedon the 1999LFS; for 2001, authors' estimates basedon the 2001 ESACLFS. 6.8 Educational attainment o f the workforce. The number o f educated workers in Ethiopia has risen over time, thanks to the government's massive investment since 1994. Yet the gains have made only a modest impact on the illiteracy rate, which fell from 77 percent in 1994 to 71 percent in 2001 (table 6.5). The educational profile o f workers differs greatly across sectors, however. In agriculture, nearly 80 percent o f the workers are illiterate, and 11 percent have between one and four years o f primary schooling; workers have on average some 1.7 years of schooling. By contrast, in the formal sector, only 28 percent o f the workers are illiterate and workers have on average more than 7 years of schooling. - 166 - Table 6.5. EducationalAttainment of Workers Aged 10 and Above, Ethiopia, 1994,1999 and 2001 1999 Formal Informal sector All 1994 sector Overall IAgriculture sectors 2001 Distributionby educational attainment (%) t Illiterate 27.6 75.9 78.5 73.5 70.5 Literacy campaign and non formal education 4.9 4.5 4.6 4.5 3.6 Grades 1-4 7.0 10.7 10.5 10.5 12.9 Grades 5-6 7.7 4.4 3.7 4.6 5.2 Grades 7-8 7.8 2.6 1.8 2.8 3.2 Grades 9-12 25.4 1.8 0.8 3.0 3.2 Above grade 12 19.5 0.1 0.0 1.1 1.3 All educationlevels 100.0 100.0 100.0 100.0 100.0 Average years o f schooling 7.1 1.7 1.5 2.00 1.5 Employed population (inthousands) 25,732.2 1,238.9 23,657.7 19,869.1 24,896.6 25,747.6 The MarketReturnsto Education 6.9 In this section, we present evidence on the returns to education in agriculture and in salaried employment. This evidence has been distilled from the results reported in the most reliable pblished reports accessible at the time o f report writing. It is supplemented by original estimates o f the retums to education in salaried employment that were prepared for this report using data from the 2001 Ethiopia Stand-Alone Child Labor Force Survey. Although the focus of this survey was child labor, it contains information on the earnings o f all members o f households sampled, and as such provide the data needed for the estimates on 6.10 Education and amicultural moductivitv. Ethiopian farmers are in general technically inefficient: a careful study by Suleiman (1995) suggests that outputs would have increased by 50 percent had the farmers used the right te~hnology.'~'For our purpose here, the issue is whether or not education makes a difference to their performance. Ifinternational experience is a guide, we should expect educated farmers to be more willing to adopt new farming techniques, to be able to manage their production better, and to be generally more productive (see, for example, `26 By contrast, the 1999/2000 Household Incomeand ConsumptionExpenditure Survey (HICES) and the Welfare Monitoring Survey (WMS) contain earnings information only for the household head. Although such data can also be used to estimate earnings functions, the underlying data suffer from obvious limitation because of selectionbias. For the results based on these data see table T3.3 intechnicalnote 3. 12' The results reported in Suleiman (1995) were generatedby applyingthe analyticaltechniqueof data envelopment analysis to data from the first round of the Ethiopia Rural HouseholdSurvey (ERHS) conducted in 1994 by the Department of Economics, Addis Ababa University, in collaboration with the Center for the Study of African Economies, Oxford University, and the InternationalFood Policy Research Institute, Washington D.C. The data covered 1,477 households associatedwith 18 peasant associations, in 15 woredus and six regions.Datafor about 250 farmers were usedinSuleiman's analysis. The ERHS, which is a nonrepresentative panel survey, captures information on the family background in each household as well as their activities pertainingto production, consumption, andaccess to anduse ofcredit. - 167- Jamison and Lau [19821and Rozenzweig [19951for well-known summaries on the international evidence in this regard). For Ethiopia, we summarize the results reported in Weir (1999), perhaps one o f the most accessible works on the topic at this writing12* As in Suleiman, Weir relies on data from the first round o f the Ethiopia Rural Household Survey (ERHS). He estimates standard production functions that relate the value o f cereal crops to a set o f inputs (land, human, and animal [ox] labor, and capital) as well as to a set o f household characteristics, including the educational attainment o f the household head and all the adult workers on the farm. Table 6.6 displays a summary o f the results pertaining to the education variables. They show that farmers' average years o f schooling have no influence on agricultural productivity-a disappointing but not surprising finding, given the extremely low stock o f human capital among farmers. When educational attainment is specified in terms of dummy variables, however, the regression estimates indicate that farmers with four to six years o f primary schooling are significantly more productive than those who are either illiterate or have three or fewer years o f schooling. On the assumption that investing in four to six years o f primary schooling involves three years o f forgone production, the magnitude o f the estimated regression coefficients imply a retum o f between 5 and 7 percent a year, depending on the regression specification. Table 6.6. Regression Estimates of the Effect of Education on Farmer Productivity, Ethiopia, 1994 Specification 1: Education of Specification 2: Education of all adult rorkers on 1 e farm Effect on Effect on the the log of Specification of the education log of the Rates of the value of Rates of variable inthe regression Sample value of returns Sample cereal returns equation average cereal crops (%)' average crops (%)a' Years o f schooling 1.14 -0.01 n.s. 1.45 0.01 n.s Level of education Illiterate (omitted category) 0.63 - - 0.56 - - Grades 1-3 0.25 -0.06 n.s. 0.33 0.05 n.s. Grades 4-6 0.08 0.12** 6.6 0.09 0.18* 4.6 Grade 7 andhigher 0.04 -0.33 ns. 0.02 -0.14 n.s. -Not applicable. n.s. Not statistically significant. ***Statisticallysignificant at the 10%level Statistically significant at the 5% level. Note: Estimatesare basedon 616 observations; see Weir 1999for complete results on coefficient estimates for other regression variables. a. Computed on the assumption that workers with an educational attainment of grades 4-6 forgo three years of eamings. Source: Weir 1999-tables 3a and 3.b, equation Sa for years of schooling; tables 4a and 4b, equation 1l a for education dummies. 6.11 The positive relation between education and farmer productivity works through many channels. Among them is the widely documented tendency o f better-educated farmers to adopt modem farming technologies and inputs (see Lau and Jamison 1982; Rosenzweig 1995). The behavior o f Ethiopian farmers i s no different, as the findings in various studies confirm (e.g., Other citations include Asfar andAdmassie (1996) and Croppenstedt, Demeke and Meshi (1999). Unfortunately neither of them was accessible at the time o f writing. - 168 - Croppenstedt, Demeke and Meshi [1999]; Tesfaye and Shiferaw [2001]; and Weir and Knight [2000]). To illustrate, the results in Weir and Knight (2000) on the impact o f education on fertilizer use-based again on the ERHS data-are summarized intable 6.7 below.'29 They show that each additional year o f schooling above the sample mean raises the probability o f fertilizer use by 1.3 percent. Interestingly, peer group effects are very large: an additional year o f schooling among adults inthe locality o f the farm raises the probability o f fertilizer use by nearly 22 percent, regardless o f the fanners' own level o f educational attainment. The magnitude o f this estimate is comparable to the finding reported in Croppenstedt, Demeke, and Meshi (1999) that the rate o f fertilizer use is 29 percent more intensive among literate farmers than among their illiterate counterparts. The role o fpeer group effects suggests that if education is to help improve farmer productivity, the strategy would necessarily involve broadening access to basic schooling as much as possible. Table 6.7. Education's Influence on Fertilizer Use, Ethiopia, 1994 Estimated marginal effect on Indicator Sample mean the probability of fertilizer use Average years o f education o f all adults Infarmer's household 1.47 0.013* Inthe locality ofthe farm (i.e., peasantassociation) 1.46 0.217* Percentageof farmers using fertilizer 46 n.a. n.a. Notapplicable. * Denotesstatisticalsignificanceat the10%level. Note: Estimatesare basedondata for a sample of 1,191 farmers. Source: Weir and Knight 2000-table 1 for sample average; table 7, equation 1, for estimatesof the marginaleffects. 6.12 Education and eaminas among salaried workers. Consider first the pattem o f mean earnings shown in table 6.8. It reflects the expected positive relation between eamings and education: on average, those with higher education eam 93 percent more than those with secondary education; those with secondary education eam 47 percent more than those with only a Grades 5-8 primaryeducation; those with a Grade 5-8 education, inturn, eam about 76 percent more than those with only a Grade 1-4 education; and the latter eam about 72 percent more than workers who are illiterate. That eamings rise with educational attainment i s a3 attem found among menandwomen workers, as well as among those inurban andrural areas.1 9 12' The analysis involves estimation o f a probit model in which the regressors include the main characteristics of the farmer's household and his land, as well as the average years of schooling of adults in the farmers' household, and in the locality of the farm(i.e., peasant association). I3O The pattern is consistent with those reported inKrishnan, Selassie and Dercon (1998). The data from this study also reveal the characteristic wage compression in the public sector: at low levels of qualification, workers earn more than their private sector counterparts, but as the level o f education rises, the opposite is true. Thus, while the govemment pays workers with less than a primary education 24 percent more than similarly qualified workers in the private sector, it pays workers with secondary education 17 percent less, and workers with higher education, 14 percent less, than workers in private employment with a corresponding level o f education. - 169 - Table 6.8. Monthly Mean Earnings by Levelof Education, Ethiopia, 2001 (Current Birrs) Workers' level of educational Whole Gender attainment sample Male Female Urban Rural Illiterate 75 96 61 84 63 Primary education Grade 1-4 129 174 77 129 131 Grade 5-8 227 281 136 238 147 Total 193 248 112 203 139 Secondary education 334 378 266 347 256 Higher education 645 681 552 701 378 No. of observations insample 9,145 5,221 3,924 Note: Wage eamers paidon amonthly termofpayment only; excludingworkers with a secondary occupation. Source: authors' estimatesbasedon the 2001 ESACLFS. 6.13 The private returns to education. The foregoing pattems provide a first hint that education is an attractive investment for individuals seeking salaried employment. Several studies have attempted to quantify the magnitude o f the retums but as technical note 3 which make it difficult to draw solid conclusions from the result^.'^' Below we report on new documents and explains, none o f them is free from serious data limitations and other constraints results which are based on the recently released 2001 Ethiopia Stand-Alone Child Labor Force Survey inwhich earnings data were collected from all adult members o f the sampled households (table 6.9). The returns are computed from the coefficient estimates in standard Mincerian eamings e q ~ a t i 0 n s . l ~ ~ 6.14 For the country as a whole, the private returns follow a U-shape, in which they are highest at the lowest levels o f educational attainment as well as at the top end. Given these pattems, it is not surprising that the demand for education i s strong throughout the system. The U-shape i s also characteristic o f the returns to education among men. Compared with men, women receive higher returns on their education at all levels except the lowest. Finally, the pattem o f returns in urban and rural areas also add new insight: for Grades 1-4, the returns in rural returns are much higher than those in urban areas; but they also drop off much more precipitously than in urban areas as the level of education rises, falling to only 8 percent a year for Grades 9-12, compared with a still reasonably high 12 percent inurban areas. We can thus expect the demand for schooling in rural areas to weaken considerably beyond grades 1-4. Further, it is possible that even ingrades 1-4, the lack o f information about schooling's benefits 13'See technical note 3 for a more detailedcommentary on the available studies that were publishedafter 1990. See especially table T3.4 which summarizes the estimate of retums from the most reliable of the studies. Data for 1994 suggest that in urban areas, an additional year of schooling yields a retum of about 7 to 9 percent a year, depending on sector of employment, comparedwith as high as 18 percent a year inrural areas; estimatesbased on data for 2000 indicateareturnof 16percent ayear. The latter data also suggest that the private retums average 33 percent a year for primary education, 19 percent for secondary education, and 22 percentfor higher education. 132See table 6.9, footnotea. - 170- may lead parents, particularlythose who have themselves never beento school, to underestimate the returns andthus under invest intheir children's schooling, particularlythat o f their daughters. Ethiopia Men Women Urban Rural Average Average Average Average Average Educational years of years of years of years of years of attainment schooling Returns schooling Returns schooling Returns Illiterate 0 - 0 - 0 - Grade 1-4 3.0 15 2.9 22 2.8 38 Grade 5-8 6.6 16 6.6 15 6.4 11 Grade 9- 12 11.3 11.3 11.4 17 11.3 12 11.2 8 Diploma 14.0 14.0 14.0 40 - - - - Degree 17.0 22 17.0 20 17.0 30 - - - - a/ The reiums are calculated as [(exp(bj-bi)-l]/[mj-mi] where bj i s the regression coefficientof the educational dummy variable j and mj i s the corresponding averageyears of schooling (see appendix A6.2 for the relevantregression estimates). Source:analysis of the 2001 EthiopiaStand-Alone Child Labor Survey (ESACLFS). TheImpactof Educationon PovertyandNonmarketOutcomes 6.15 Education also produces benefits through its impact on poverty and other indicators o f the population's well-being. In this section, we briefly review the evidence on the relation between educational attainment and poverty, fertility behavior, and child health and nutritional status. 6.16 Educational attainment and poverty. The recently completed World Bank (2004~)study on poverty in Ethiopia underscores that while many factors are involved in reducing poverty, education plays a key role. Table 6.10 shows, for example, that whereas the overall poverty rate i s around 38 percent, the share of households living inpoverty falls to 32 percent among those headed by someone with some primary education, and 21 and 9 percent, respectively, among those whose householdheadshave completed secondary andpostsecondary education. - 171 - Percentage living in Characteristic of household head poverty a' Male 0.38 Female 0.37 Completed some primary education 0.32 Completed secondary 0.21 Completed postsecondary 0.09 a. Refers to share of householdswhose consumptionexpenditure(food andnonfood) per adult equivalent falls below a lowerboundthreshold estimatedin World Bank 2004c; see World Bank 2004c, box 1.2, for more details onthe estimation method. Source: World Bank 2004c-table 1.8, which is based ondata fromthe 1999Ethiopia HouseholdIncome and ConsumptionExpenditure Survey. 6.17 Regression analysis confirms the positive impact o f education on poverty reduction: on average, increasing a household head's schooling by one grade would raise the consumption expenditure per adult equivalent in male-headed households by around 1.6 percent, and in female-headed households by between 1.5 and 2.5 percent (table 6.1 1). These impacts are o f the same order o f magnitude as the annual growth rate o f Ethiopia's per capita GDP during the past 12 years. They imply that raising the average educational attainment o f household heads by an extra year would boost consumption expenditure by roughly the same amount as a year o f economic growth at its current pace. It is interesting to note that increases from even modest levels o f educational attainment would achieve a large impact: an extra grade o f primary schooling would expand household consumption expenditure inmale-headed households by 1.4 percent compared with an increase o f 1.8 percent associated with an extra grade o f secondary schooling; andby the corresponding rates o f 1.7 and 2.1 percent infemale-headed household^.'^^ Simulations based on these regression estimates suggest that if all Ethiopian adults, men and women, received a grade 4 education, the overall poverty rate would fall by 18 percent. The reduction would be smaller, 12 percent, if only the women were so educated; it would be even smaller, 6 percent, ifonly the men attained this level. 133Additional results show that female education generally reduces poverty to a greater extent in urban than in rural areas; in urban areas, its impact is about twice as large as that of male education. There are also important differencesacross regions: in Amhara and SNNPR, female educationexerts a substantiallylargerimpact than male education, whereas inOromiya, the impacts are comparable; inTigray, female educationhas, on average, no impact, while male educationachievesa statisticallysignificant, thoughquantitativelymodest, impact. - 172- Table 6.11. Regression Estimates of the Impact of Education on Poverty, Ethiopia, 1999 Regression results Simulated impact of Marginal endowing all adults with Indicator Sample impact of a grade 4 education mean education (% change) Consumption expenditure per adult equivalent, 1995-99 (Birr) 1,435 8 Meanyears o f schooling o f adult males 1.5 to 1.6* 3 Meanyears o f schooling o f adult females 1.5 to 2.5* 5 Probability o f living inpoverty, 1995-99 (percent) Men 38 n.a. -6 Women 37 n.a. -12 All adults 38 n.a. -18 6.18 Education and fertility behavior. Fertilitybehavior plays a major role inchanging the size and structure of the population. According to a 2000 estimate, Ethiopianwomen of child-bearing age average an estimated 5.9 births, which makes their total fertility rate one of the highestinthe world today. As in many other Sub-Saharan African countries, female education in Ethiopia i s inversely related to fertility (table 6.12). Women with no education expect to average 6.2 live births each, compared with only three live births among women with at least some secondary education. Because o f the higher fertility rates among the less well educated, these women are also at greater risk o f increasedmaternal mortality. r Table 6.12. Fertility Rate and Number of Children Per Woman by Level of Education, Ethiopia, 2000 Total fertility Number of children ever Educational attainment rate a' born to women ages 15-49 No education 6.2 7.1 Primary education 5.1 5.9 Secondary and higher education 3.1 4.6 Overall average 5.9 7.0 a. Number of live birthsper woman between ages 15 and49 givencurrentage-specific rates ofbirthover this age span; refers to the ratefor the five yearsprecedingthe surveyin2000. Source: 2000 EthiopiaDemographic and HealthSurvey. 6.19 Education and child health and nutritional status. Infant and child mortality rates in Ethiopiahave been declining since the 1 9 6 0 but they remain hightoday: 97 childrenper 1,000 ~ ~ who are born die before their first birthday and 188, before their fifthbirthday. According to the World Bank and Ministryof Health's 2004 country status report on health andpoverty, mother's - 173 - education is associated with health behaviors that affect the outcomes o f children's health in important ways, including the intake o f iodized salt and use o f immunization vitamin A supplementation. According to regression analysis, an extra year o f mother's schooling reduces the under five mortality rate by an estimated 16 deaths per 1,000 live births inrural areas, andby 6 deaths per 1,000 live births in urbanareas-significant reductions from the current average o f 188 deaths per 1,000 live births (table 6.13). 6.20 Part o f the association between mother's education and child health comes from its influence on children's nutritional status. Ethiopian children are generally at higher risk o f malnutrition than other children in Sub-Saharan Africa: about 45 percent are underweight compared with the region's average of 33 percent during the 1990s. A study by Christiaensen andAlderman (2003) usingdata from the 1995, 1997, and 1998 Welfare MonitoringSurveys, as well as data from the 1998 Health and Nutrition Survey, confirms maternal education as a key determinant o f chronic child malnutrition in Ethiopia: among children between age 3 and 60 months, each year o f schooling o f the most educated female in the household raises a child's height-for-age score by 0.03, all else being the same. It is noteworthy that the marginal impact is about twice as large as the corresponding impact for paternal education. The authors also show that the age-for-height score increases by an average o f 0.23 points inhouseholds where the most educated woman has received postsecondary schooling relative to the score inother households, When thus defined, maternal and paternal education exert a comparable influence on child nutrition. Based on the regression results, the authors estimate that having at least one female adult per household attain primary education would reduce the incidence o f stunting inEthiopia by as muchas 6 to 11percent. Table 6.13. ParentalEducationand ChildMalnutrition,Ethiopia, 1995-98 I Marginal effect a/ status of childrenages 3 Under five child mortality IEducationalattainment of mother or adultsinhousehold to 60 monthsb/ Rural Urban Mother's years of schooling -0.016 *** -0.006** Females adults Years of schooling the most-educated woman has 0.031*** - - Most educated woman has postsecondary education 0.230* - - Male adults Years of schooling the most-educated manhas 0.016*** A - Most-educatedmanhas postsecondary education 0.246*** - - ~~ -Not estimated in source. ** statistically * statisticallysignificant at the 10percent level. *** statistically significant at the 5 percent level. significant at the 1percent level. a. Refers to the regression coefficient on the indicated education variable. b. Measured in terms of height-for-age Z scores. The score i s computed as (h-hr)/sdr, where h is the child's observed height, and hr and sdr are, respectively, the medianheight and the standarddeviation in the populationo f children inthe child's sex and age group. c. Measured as the probability of a child death. Source: or impact on child nutritional status, Christiaensen and Alderman 2003, table 3; for impact on child mortality, World Bank 2004c, table 4.5. - 174- Issues for Policy Development 6.21 The evidence presented above raises issues for policy development from two perspectives-the long and short term. Ensuring that all Ethiopian citizens attain at least four years o f schooling is essential for the country's long-term economic and social progress. Because it lays the foundation for not only raising productivity in agriculture and salaried employment, but also for reducing poverty and improving human welfare, this minimum investment in the country's human capital i s arguably an immediate priority. Beyond this minimum, the main issue is to set a pace of expansion that is appropriate to the country. Two considerations are relevant in this regard: overly rapid expansion o f postprimary levels o f schooling may outstrip the economy's capacity to create jobs for the correspondingly large output of highly educatedgraduates that the systemwill produce; and it may divert public funds away from educational investments that need to be prioritized (e.g. universalization of primary education). As a contribution to the discussion, we present below some information on the employment experience o f recent graduates in the Ethiopian labor market; and estimates o f the social returns to education. 6.22 Output o f nraduates and their absorption into the workforce. Although a full assessment would require more detailed data than were available for this report (such as tracer surveys that track recent graduates' transition from school to work), some insights nevertheless emerge from a review of unemployment among workers by age group. For example, secondary school graduates entering the workforce for the first time are likely to be concentrated among workers ages 20-24, while those with vocational andhigher education are likely to be concentrated inthe 25-29 age bracket. Thus, the incidence o f unemployment among workers in these age groups can provide indirect evidence on the labor market absorption o f graduates by educational attainment. 6.23 Table 6.14, which i s based on the 2001 Ethiopia Stand-Alone Child Labor Force Survey, provides some interesting results in this regard. Unemployment in 2001 i s more widespread among those with secondary education than among the lesser qualified, as well as among higher education graduates. But the incidence o f unemployment among recent labor market entrants (whose unemployment rates are denoted inthe shaded cells inthe table) i s even higher, reaching 34 percent among graduates attaining grades 11-12, and 25 percent amon those attaining grades 9-10, compared with about 9 percent among higher education graduate^."^ Among new workers with grades 5-8, the unemployment rate averages 10 percent, compared with just 4 percent among those with an attainment o f only grades 1-4. Taken as a whole, these results suggest that the most highly qualified workers-those with higher education-enjoy good employment prospects at present, whereas other educated workers, especially those with secondary schooling face substantially greater difficulty inlanding their first job. While the situation may improve as the modem sector o f the economy develops, these pattems caution against an overly ambitious plan for expanding postprimary education. Even for higher education, the favorable picture could well deteriorate as graduate output from the systemrises (because o f planned expansion o f 134 Datafrom the 1999 Ethiopia Labor Force Survey show similar pattems of graduate unemployment. That survey indicate that unemployment rate among recent graduates of vocational and technical education is about 25 percent, compared with 44 percent among general secondary school graduates. - 175 - the system); jobs ingovernment administrationwould then level off as vacancies associated with decentralization are filled and the pace o f job creation in the private sector o f the modern economy remains tepid. Table 6.14. Unemployment Rate by Educational Attainment and Age, Ethiopia, 2001 15-19 I 20-24 I 25-29 30-34 >34 Allages 4.7 3.6 3.7 (7,116) (6,587) (6,078) (4,426) (20,030) (58,487) 3.5 2.9 4.3 ~ 4.5 5.8 7.8 4.0 4.3 5.7 (2,122) (4,366) (23,379) 10.7 6.6 16.4 12.6 9.4 22.1 11.9 8.3 19.9 (1,307) (1,897) (8,755) Highereducation - I 3.7 3.3 7.1 (485) I (776) (598) (1,374) (3,258) -Too few observations to compute reliable ratesof unemployment Note: Shaded cells refer to the age groups where firstitime entrants Erom each education category are concentrated. Number of observations inthe sample inparentheses. Source: Authors' estimatesbased on the 2001 ESACLFS. 6.24 The social returns to education. These data provide another perspective for prioritizing public investments in education. Unlike the private returns to education presented earlier, the results shown in table 6.15 refer to social rates which take into account both the private and public costs o f ed~cati0n.l~~.Because o f data constraints, we apply the method only to the data for primary and secondary education. The results indicate that inurban areas, the social returns for grades 1-4, grades 5-8 and grades 9-12 are comparable, ranging from 11to 13 percent a year. By contrast, inrural areas, the retums for grades 1-4 are muchhigher than insubsequent grades, averaging 20 percent a year compared with only 8 percent a year for grades 5-8 and 9 percent a year for grades 9-12. These patterns imply that expansion o f Grades 1-4 inrural areas warrants particularly strong support; and that the case for expansion o f subsequent levels is generally weaker, though less so inurbanthan inrural areas. See Chiswick 1998 for a detailed description of the method used here. Inkeeping with custom, the retums are called social retums even though the calculations ignore possible extemalities associated with the various levels of education. Incorporating extemalities inthe calculation is often not attempted because o fthe difficulty of estimatingthem unambiguously and accurately. - 176 - Table 6.15. Social Rates of Returns to Education inEthiopia, 2001 (percent per year unless otherwise indicated) Level of educational Average years attainment of schooling Urban Rural Grades 1-4 3.0 12 20 Grades 5-8 6.6 11 8 Grades 9-12 11.3 13 9 a/ The retumsare calculated, basedon resultsfrom the Mincerianregressionsreportedin appendix table A6.2, as [(exppj-bi)- l]/([mj-mi] x k) where bj is the regression coefficient of the educational dummy variable j and mj is the corresponding averageyears of schooling, and k given by the ratiobetween the total costofthe averageyear of schooling(includingdirect costs and forgone earnings) and the forgone earnings from a full year of work had the person been in the labor market instead of at school. For estimatingthe privateretums,the directcosts referonly to whatthe individualincurs, while for the socialretums,italso includes what the government spends. For the purpose of these calculation we assume that students continue to be able to contribute to household production while in school: those in Grades 1-8 contribute an assumed 10 percent of their time to productiveactivities, while those in Grades 9-12 contribute 25 percent. The impact of these assumptionsis to reduce the forgone cost of schoolparticipationand raisethe retumsby 1to 2percentagepointscomparedto the resultsobtainedwhen itis assumedthat students makenocontributionto work while enrolled. For further detailsonthe methodofestimationseeChiswick 1988. Source:authors' calculationsbasedon analysisof the 2001 ESACLFS. Conclusion 6.25 At the start of the 21Stcentury, Ethiopia's population is among the least educated inthe world, averaging fewer than two hllyears o f formal schooling among the adults. Strengthening the country's human capital base to he1 its economic and social performance will require sustained investments ineducation. An important priority is to ensure that all Ethiopians acquire at least four years o f primary schooling. Such a strategy can be expected to produce highsocial retums as well as boost social equity. Beyond this minimumtarget, additional investments will be required to help meet the expected increase in demand for educated labor as the economy expands and diversifies out o f agriculture. What i s at issue is the pace o f expansion o f the system beyond the first four years o f primary schooling. In this regard, it is important to recognize that the demand for educated labor is a derived demand, one that is driven by how fast the modern sector of the economy grows. This sector has indeedbeen expanding inrecent years inEthiopia, butits growthhasproceededmodestly; as a result, agriculture continues to dominate the economy with more than 80 percent o f the workforce. Two other features o f the labor market are also relevant: (a) most workers engage in self-employment or work as unpaid family labor; and (b) unemployment among recent labor market entrants, especially among the moderately well educated, i s relatively high. Taken together, these characteristics suggest that while it makes good economic and social sense to universalize four years o fprimary schooling as soon as possible, the pace of expansion in subsequent grades may need to proceed in tandem with the economy's capacity to absorb well-educated graduates into jobs for which their training has prepared them. - 177- TechnicalNote 1: Preparationofthe Dataon Current Public ExpenditureonEducationinEthiopia T1.1 This note explains the preparation o f the data reported in chapters 1 and 3 on public current spending on education inEthiopia. The main source o f the aggregate expenditure data are the consolidated figures from the Ministryo f Finance and Economic Development (MOFED).'36 These data are coded in sufficient detail to reveal broad categories o f spending by level o f education and function at the federal and regional levels. Yet because the reporting system has been evolving, no doubt reflecting the progress o f efforts by the government to improve it, the reported aggregates sometimes suffer from changes in definitions, misclassifications, and periodic gaps in certain items o f expenditure or in the reporting o f regional spending. For the purpose o f this report, an effort was therefore made to: (a) "clean" the reported MOFED data to ensure time-series consistency in the data underlying the aggregate amounts; and (b) for one year, 1994 E.C. (Le., 2001-02), to supplement and compare them with information on spending culled from other sources (especially the school census for that year and reports from specific institution^).'^^ Below we provide details on how we have dealt with missing data, and how we have estimated items o f spendingneeded to construct a more detailed picture o f the composition of the government's recurrent spending on education. GeneralDefinitions T1.2 Data for the period 1986-92 E.C. (1993-94 to 1999-2000) refer to actual current expenditures, while those for 1993 E.C. and 1994 E.C. refer to the voted budget. In chapter 3 where detailed estimates o f spending are made for 1994E.C., aggregate spending on the salaries o f teachers and administrative staff in primary and secondary schools refer to actual spending, based on data reported by each school through the school census; the same item for teacher training institutes, teacher training colleges, and institutions o f higher education refer to estimates based on reports from individual institutions inthe correspondingcategories. 13* T1.3 To construct our time series on current spending on education, we make the following adjustments to the reported data available from MOFED: (a) we exclude spending on items that should in fact be classified as capital spending (e.g., purchase o f equipment and motor vehicles; and o f materials for school construction [see appendix table A1.5 for the amounts excluded]); and (b) we minimize double counting by excluding the amounts shown under "retained `36 Prior to 1986E.C. (Le. 1993-94), current expenditureson educationare based onthe Countrywidebudgetary revenue and expendituredatafrom the Ministry of FinanceandEconomicDevelopment(MOFED). For subsequent years, the main source of dataare MOFED's electronicbudgetdata files. 13' Specifically, we useddata from the Ministryof Education's SchoolCensusfor 1994E.C., andregionalreports submittedto the Ministry on physicalandfinancial performance in 1994E.C. under the EducationSectorDevelopmentProgramme(ESDP); anddatasuppliedby individual teacher training institutes(TTIs), teacher training colleges (TTCs), and other institutionsof higher education. 13' Becauseof dataconstraints,the same approachcouldnot be usedto estimatethe salarybill for technicalandvocational education andtraining; for this subsector,the spendingis thus simply the budgetedamountreportedinMOFED's budget data files. - 178 - revenues" where these are likely to be transfers between spending units (e.g., between regional governments and the National Organization o f Examinations). Our calculations include grants to institutions and individuals, but exclude current spending financed by external sources o f funding, the latter treatment necessitated by the incompleteness o f data on external hnding. Whenever possible, they exclude spending on noneducation services that are included in the reported budget data; the most important example is spending on Black Lion Hospital which appears under the budget for Addis Ababa University. Economic Classification T1.4 Total current expenditures include spending on the followingitems: Stagremuneration. This itemrefers to the expenditure on the salaries, allowances, andpensions o fteaching andnonteaching staff. The amounts for the latter two items are estimated from their proportionate share o f the wage bill as reported inMOFED's data for the Ministry o f Education. Expenditures on contract and casual stafl Note that the data are available only for 1994 E.C. and at the federal level. Miscellaneous payment to stafl The data are again available only for 1994E.C. and at the federal level. Education supplies. Foodstufl Uniforms, clothing, and bedding. Grants to individuals. Other expenditure. This category is defined as the difference betweentotal current expenditure net o fthe spending on the items listed above. Functional Classification T1.5 The current expenditures detailed above are disaggregated along the following functional categories: e Administration.Under this rubric are included expenditures on: (a) administrative and general services at both the federal and regional levels; (b) the National Organization o f Examinations (or its predecessor organization, the Examination Agency) at the federal level; (c) the Higher Education Main Department (for 1986 E.C. only); and (d) the Curriculum and Supervision Institute (for 1986 E.C. and 1987 E.C. in Oromiya Region). e Primary education. This item includes expenditures at the federal andregional levels on grades 1-8, as well as expenditures incurredby the Ethiopian Community School. e Secondary education. Inthis category are expenditures at the federal andregional levels on grades 9-12, as well as expenditures incurredbythe Bole Boarding School. e Technical and vocational education and training (TVET). TVET expenditures include those at the federal and regional levels to provide technical and vocational training. e Teacher Training Institutes (TTIs). Under TTIare expenditures on the institutes at Adwa, Arba Minch, Bonga, Hosana, Assela, Debre Berhan, Dessie, Gambella, Harar, - 179 - Metu, Nekemte, Jijiga, and Robe; since some o f these institutions were created only recently, the data naturally exclude them for the years precedingtheir establishment. Teacher Training Colleges (TTC). This item comprises expenditures o f the colleges as Kotebe, Abi Adi, Awassa, Gondar, and Jimma. As with the TTIs, some o f these colleges are new and the data naturally exclude them if the year in question precedes their date o f creation. 0 Higher education. Higher education expenditures refer to the following institutions: Addis Ababa Junior College o f Commerce, Bahar Dar Polytechnic, Ambo Agricultural College, Jimma College o f Agriculture, Addis Ababa University (excluding spending on Black Lion Hospital, which was reported only in 1994 E.C.), Mekele University College, Alamaya Agricultural College, Nazareth Technical Teachers Education College, Mekele Business CollegeMekele University, Arba Minch Institute o f Water Technology, Wondogonet College o f Forestry, Jimma Health Science College/Jimma University, Awassa College o f Agriculture, Bahir Dar Teacher Training CollegeBahr Dar University, Gonder College o f Medical Science, Dila Health and Education College, Ethiopian Civil Service College, Education Mass Media Training Institute, Art College (Tigray region). Due to lack o f data, expenditure on Maichew Technical College (Tigray region) is excluded. 0 Other. Under this rubric are included the expenditures o f the following institutions: Educational Mass Media (federal and regional levels); adult education (federal and regional levels); kindergarten teacher training institutes (regional level); special education at the regional level; Study and Research Center (Amhara region); Adama Teacher Training College (for E.C. 1993 and 1994 only; Oromiya Region). T1.6 Inorder to fillgapsinthe reportedbudget data, we took the followingmeasures. The data for the National Organization o f Examinations was completed with information supplied directly by the institution itself. The expenditures o f the Teacher Training Institutes at Harari and Amhara, and o f the Teacher Training Colleges at Abi Adi, Awassa, Gondar, Jimma, and Kotebe were missing for some years; in response to a request made in the context o f the present study, the institutions supplied the information. The data for two regions, Benshangul-Gumuz and Gambella, proved particularly problematic. Reporting was incomplete; the data that were reported were presented indifferent formats over time and were aggregated inways that made it difficult to prepare a complete time series with the same economic and functional classifications as for the other regions. For these two regions, only a functional breakdown was possible to construct for the period under analysis. For the years 1986 E.C. to 1988 E.C. (i.e,, 1993-94 to 1995-96), the budget data were complete and had the required format. For the other years, the data were pieced together as follows. For 1994 E.C. (2001-02), the hnctional breakdown was constructed from ad hoc information contained ineach of the two regions' report on physical and financial performance under the Education Sector Development Programme (ESDP) for 1994 E.C. For the years 1989 E.C. to 1993 E.C. (i.e. 1995-96 to 2001-02), the breakdown o f spending was estimated usingthe following information: (a) the trend intotal current expenditure between 1988 E.C. and 1994 E.C. (Le., 1994-95 to 2001-02); and (b) the trends in key cost-related education sector variables in the two regions (e.g., unit costs, pupil-teacher ratio, teacher- administrative staff ratio, and so on). - 180- DetailedEstimatesof the Compositionof Spendingin 1994E.C. (2001-02) T1.7 To better understand the pattern o f education spending, we made further estimates o f the economic breakdown o f spending by level and type o f education, using the data for 1994 E.C., the most recent year for which the relevant data were available to us at the time o f data analysis. The estimates pertain to expenditures at the federal and regional levels on the following: grades 1-4, grades 5-8, grades 9-12, TTI, TTC, andhigher education. For the reasons explained above, we excluded Benshangul-Gumuz and Gambella from the exercise because o f insurmountable data constraints. Our method involved combining budget data for nonsalary expenditures with estimates o f staff remuneration based on data collected through the Ministry o f Education's 1994 E.C. school census (for primary and secondary education) and on those collected (as part o f this study) from the teacher training institutes and colleges and the various institutions o f higher education. This approach implies that the spending on nonsalary items pertains to voted budget amounts, while the expenditure on staff salaries refers to the estimated actual spending. NonsalaryExpenditures T1.8 For each level o f education, expenditure on educational supplies, foodstuff, uniforms, clothing, and bedding, we took grants to individuals and others directly from the reported federal and regional budgets. For Addis Ababa, the breakdown o f spending on grades 1-4 and grades 5- 8 are not shown separately. We estimate the breakdown between these cycles by prorating the aggregate amount for grades 1-8 according to each cycle's share o f total enrollments. Wage BillofTeachingandNonteachingStaffinGrades 14,543, and9-12 T1.9 The 1994 E.C. school census contains information for 11,444 schools (out o f a total of some 12,000 schools nationwide) on the distribution o f teachers assigned to teach ingrades 1-4, 5-8, or 9-12 and, within each instructional cycle, the distribution o f the teachers by salary range (see table 17.2 o f the school questionnaire). This information makes it possible to estimate the average remuneration o f the teachers ineach o f the three instructional cycles.'39 For schools with missing data in the school census, we imputed the average salary o f teachers at the school level by first estimating a regression equation relating the average remuneration o f teachers to the vector o f variables consisting of: (a) average o f teachers' age; (b) teachers' age squared; (c) percentage o f teachers by rank; and then using the regression equation to fill the wage data gaps. This approach failed in Somali, however, because none o f the schools in this region were included inthe school census. We thus made the assumption that the average wage o f teachers in this region is the same as in Afar, the other region where teachers routinely receive hardship allowances. Once we have obtained estimates o f the average wage o f teachers assigned to each level o f education, the desired aggregate wage bill by level o f schooling in each region is obtained in two steps: (a) topping up the estimated average wages by payments for pensions .(assumed to be the same percentage o f the total wage bill as in the reported budget) and for allowances (which are reported as a lump sum in the budget for teaching and nonteaching staff, 13'The format of the school census questionnaireis suchthat teachers are counted inonly one cycle of instruction, even ifthey haveteachingduties inmore than one cycle.However,becauseinformation is lacking on the number of teachersinvolved, it is impossibleto adjustthe estimates ofaverage remunerationaccordingly. - 181- andwhich we split betweenthe two categories according to their share o f the wage bill); and (b) multiplyingthe resultbythe corresponding number o fstaff. T1.10 We estimated the salaries for nonteaching staff in a similar fashion. However, because administrative staff are allocated to schools and not to a particular subcycle within each school, we distributed the wage bill o f these staff members in schools offering multiple cycles o f instruction (Le., those offering grades 1-8 and grades 1-12) across subcycles by prorating it according to the distribution o fteachers assigned to each subcycle. TeachingandNonteachingStaffRemunerationinTTIs, TTCs, andHigherEducation Institutions T1.11 We basedour estimates on data suppliedby eachinstitution inresponse to a request made in the context o f this study. Because the data were patchy, we made various assumptions to complete the gaps. T1.12 For TTIs, we used the data supplied by 6 o f the 13 institutions (Abar Minch, Assela, Nekemte, Gambella, Harar, and Jijiga) to estimate the average wages o f teachers and administrative staff insuch institutions, and then applied the average to the staff inthe remaining 7 institutions with missing data. This approach does not take into account differences in staff qualification between the two sets o f institutions. T1.13 For TTCs, we first estimated the average salary o f teachers by rank from the records o f individual staff inJimma TTC. We then made an adjustment to include payment o f pension and allowances (by an amount suggested by the share o f such items inthe total wage bill reported in the MOFEDbudget), andmultiplied the result bythe total number o fTTC teaching staffby rank as reported in the Ministry o f Education's Statistical Abstract for 1994 E.C. For administrative staff inTTCs, we used a similar approach, but applied the average salary o f staff by qualification observed in all institutions o f higher education rather than in Jimma TTC alone. This broader data source is justified by the fact that whereas TTC teachers are not paid at the same rate as teachers inuniversities and other institutions o f higher education, administrative staff tend to be. T1.14 For higher education, we obtained our estimates o f the aggregate spending on staff remuneration as follows. For Addis Ababa University, the available data pertain to the individual staff, their assignment to teacher and nonteaching functions, and their salaries. This information allows us to obtain an aggregate o f spending on teaching and nonteaching staff at this institution. For the other institutions o f higher education, we first estimated the average remuneration of staff by qualification using the data from the Jimma University and the Abar Minch Institute, the only two institutions for which we have data on the rank o f individual staff and their salary. We applied the average from these institutions to the number o f staff by rank inthe other institutions o f higher education, and added the resulting wage bill estimate to the wage bills o f Addis Ababa University, Jimma University, and Abar Minch Institute, thus obtaining the aggregate for higher education as a whole. From the data for Addis Ababa University, supplemented by data from the Ministry of Education on the amount o f transfers to institutions to cover the cost o f expatriate - 182- staff, we were also able to compute the average remunerationo f expatriate teacherspaid on local andinternational pay scales. - 183 - Technical Note 2: EstimatingStudent Flow Profiles T2.1 Below we elaborate on the three methods o f estimation that were presented inthe text: the pseudo-cohort method, the UNESCO reconstructed cohort method, andthe composite cohort method. The BasicData T2.2 All three methods rely on the same basic data on enrollments andrepeaters (table T2.1). In Ethiopia, the authorities make a distinction between repeaters and readmitted students: the former are students who take and fail the year-end examination, and return to the same grade the next school year; readmitted students are those who drop out during the school year without taking the year-end examination and return to the same grade the next school year. The distinction supposedly allows educators to isolate two factors associated with grade repetition: inefficiencies in the teaching process itself, and demand-side constraints that result in students dropping out before the end of the school year; this in turn retards their learning and makes it necessary for them to repeat the grade. While the separation of the two types o f students has its use, it is not relevant for the student flow calculations presented below. The two groups are thus combined into a single group, which for lack o f a better term we also call repeaters, following standard education terminology. Treating readmitted students as repeaters is important to avoid serious underestimation of the cohort survival rates. Two other reasons make it important to combine the two groups o f students into one. First, it renders Ethiopia's data comparable to those o f other countries. Second, since the distinction between repeaters and readmitted students was made only from 1993 E.C. (i.e., school year 2000-01) onward, combining the two groups makes it possible to construct an internally consistent time series for Ethiopia as well. - 184- Table T2.1. Number of Students andRepeaters,Girls and Boys, Ethiopia, 1992-93 to 2002-03 tedgrade 1992-93 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 2002-03 ,nrollments 1 623,663 907,576 1,107,751 1,424,754 1,611,928 1,751,600 1,840,185 1,955,035 2,052,607 2,053,135 2,092,159 1 2 331,767 394,095 500,007 636,395 858,043 951,944 1,070,382 1,235,823 1,363,249 1,448,462 1,445,630 3 261,426 305,826 356,448 445,780 550,862 721,587 818,086 940,891 1,096,122 1,208,636 1,262,631 4 236,185 252,422 298,508 348,153 418,716 515,300 639,276 741,934 871,320 1,015,721 1,124,084 5 207,945 212,979 232,970 273,268 316,265 384,756 466,222 576,199 673,280 786,058 902,320 6 194,908 210,741 226,508 251,718 249,894 282,546 333,431 397,377 497,980 580,773 667,612 7 184,660 178,225 188,129 218,552 255,394 264,518 299,355 346,831 419,904 537,454 633,588 8 164,143 179,203 188,101 189,299 207,192 218,419 235,296 268,413 299,659 352,521 444,291 9 120,144 134,778 151,028 167,000 172,453 191,408 210,375 211,278 265,664 312,058 354,965 10 88,234 87,097 93,525 103,516 110,176 116,479 135,035 156,103 112,463 156,088 193,108 11 86,562 75,709 77,036 84,352 91,396 103,772 111,790 131,598 175,712 32,953 49,965 12 68,746 59,611 49,327 47,885 52,470 56,010 64,528 72,740 95,382 183,531 28,676 Gr. 1-12 2,568,383 2,998,262 3,469,338 4,190,672 4,894,789 5,558,339 6,223,961 7,034,222 7,923,342 8,667,390 9,199,029 1 - 95,804 126,234 197,054 192,952 269,676 303,256 356,938 374,769 387,191 391,140 2 - 28,871 29,539 41,693 44,644 77,770 94,465 131,628 161,220 186,698 203,655 3 - 18,069 16,912 20,580 20,208 30,563 59,191 91,606 117,830 141,028 159,165 4 - 19,794 21,765 25,353 23,823 31,824 39,807 74,760 106,089 129,413 163,063 5 - 12,801 13,551 14,029 18,097 26,255 33,128 60,554 83,959 98,164 124,711 6 - 21,516 28,791 29,583 14,039 14,444 16,252 31,542 44,402 53,774 67,005 7 - 23,108 26,222 33,405 33,860 41,743 43,025 58,807 76,940 97,077 121,555 8 - 34,904 43,460 42,965 44,192 38,992 48,400 57,879 62,389 63,661 77,885 9 - 25,743 30,208 36,855 43,550 40,020 44,400 56,214 77,756 80,208 91,492 10 - 14,466 16,025 17,527 19,948 19,784 18,361 28,402 15,050 5,762 6,069 11 - 13,223 14,927 16,972 17,719 20,604 26,303 30,032 33,419 201 903 12 - 0 595 0 0 0 0 2,207 8,136 4,166 36 Gr. 1-12 I - 308,299 368,229 476,016 473,032 611,675 726,588 980,569 1,161,959 1,247,343 1,406,679 I -Datanotavailable. Note: Datainclude students in govemment and nongovemmentschools inregularprograms only (Le., they excludeeveningstudents); data for grades 11-12 for 2000-01 and 200142 show unusualtrends. The number of grade 12repeatersin200142 is probablyinaccurate,beinginconsistentwith the numberreported in grade 11in2000-01 and ingrade 12in200142. The inconsistency is beingcheckedby the authorities at the time ofreport writing. a. Readmittedstudents refer to students who droppedout beforethe end of a schoolyear andretumed the next year. Separate data on repeaters and readmitted students availableonly from 1993 E.C. [or 2000-01] onward. Source: Govt. of Ethiopia 1995,1996, 1997, 1998, 1999,2000,2001,2002; and UNESCO for data on repeaters for 1993-94 to 1996-97. Pseudo-CohortMethod T2.3 True cohort survival rates can only be computed from data for an actual cohort whose year-to-year or grade-to-grade progression is traced. Because we only have cross-sectional data at our disposal, our calculations are an approximation o f the true pattern, yielding what we call pseudo-cohort survival rates. Table T2.2 shows the data on nonrepeaters that are used in the calculation. For pupils entering grade 1 for the first time in 1994-95, we could compute the pseudo-survival rate to grade 8 by dividing the number o f nonrepeaters in grade 8 in 2001-02 (288,860) by the number of nonrepeaters in grade 1 in 1994-95 (981,517)-which yields 29.4 - 185 - percent. The result is inaccurate to the extent that the nonrepeaters in 2001-02 are unlikely to be the same group as the new grade 1entrants in 1994-95. The former group will inevitably include many who entered grade 1in years prior to 1994-95 and were straggling along and by chance found themselves in grade 8 in 2000-01; and it will exclude the new first-graders starting in 1994-95 who have yet to reach grade 8 in2001-02 because of grade repetition. 365,224 470,468 594,702 813,399 874,174 975,917 1,104,195 1,202,029 1,261,764 1,241,975 287,757 339,536 425,200 530,654 691,024 758,895 849,285 978,292 1,067,608 1,103,466 155,117 161,907 185,147 221,534 222,775 256,330 288,024 342,964 144,299 144,641 146,334 163,000 179,427 186,896 210,534 237,270 9 109,035 120,820 130,145 128,903 151,388 165,975 155,064 187,908 231,850 263,473 10 72,631 77,500 85,989 90,228 96,695 116,674 127,701 97,413 150,326 187,039 11 62,486 62,109 67,380 73,677 83,168 85,487 101,566 142,293 32,752 49,062 12 59,611 48,732 47,885 52,470 56,010 64,528 70,533 87,246 179,365 28,640 T2.4 The longer is the distance between the two years to which the calculation refers, the fewer the number o f pupils in the two years who belong to the same cohort, and the less the result will resemble a pseudo-cohort survival rate. One way to reduce the distortion i s to take a shorter time span of, say, four years (which conveniently matches perfectly the current structure of the Ethiopian education system). In the table above, the shaded cells exemplify the calculations ofpseudo-cohort survival rates: for 1993-94, the survival rate from grade 1to grade 4 is 48.6 percent (=394,893/811,772), and for 1997-98, the survival rate from grade 5 to grade 8 is 81.4 percent (=237,270/358,501). Table T2.3 shows the result o f applying this algorithm to the rest o f the data. Note that the pseudo-cohort survival rates can only be computed up to 1999- 2000 because the first-graders, fourth-graders, and ninth-graders in subsequent years have yet to reachthe corresponding end grades in2002-03, the last year for which data are available at time o f data analysis. - 186- IPseudo-Cohort Survival Rates, Ethiopia, 1993-94 to 199940 Year when cohort ,dof cycle started in the cycle Grades 1to 4 Grades 5 to 8 Grades 9 to 12 1993-94 48.6 81.4 48.1 1994-95 49.3 81.8 46.4 1995-96 48.8 72.1 49.6 1996-97 47.0 70.6 54.7 1997-98 51.6 66.2 57.6 1998-99 57.7 66.7 a/ 1999-00 60.1 71.1 18Sb' la. The available dataimply a rate of 108.1percent, which is impossibleby definition. The result signals probable inaccuracies that have arisen inthe counts of repeatersandreadmitted students as the new policy was implemented, beginningin 2001-02, to restrict entry to grades 11and 12. b. The drastic decline in survival between grades 9 and 12 reflects the impact of the new policy noted in footnote a. Source: Computed from data in table T2.2; see main text for further details. UNESCOReconstructed-CohortMethod T2.5 To make the calculations more up-to-date, educators have developed other methods to compute cohort survival rates. We explain here the UNESCO reconstructed-cohort method. The Ethiopian Ministry o f Education also reports this calculation inits annual statistical abstract, but the calculations do not count readmitted students as repeaters (effectively treating them as dropouts), thus underestimating the survival rates, Here we replicate the calculations counting readmitted students as repeaters, and examine as well the implications o f two alternative assumptions about repetition behavior. Computingthe Repetition,Promotion,andDropoutRates T2.6 The first step is to compute these rates basedon data for enrollment and repeatersintwo consecutive years. Table T2.4 illustrates the calculation with data for 1992-93 and 1993-94. In 1992-93, 623,366 pupils were in grade 1; in 1993-94, 394,095 pupils were ingrade 2. Because in 1993-94 there were there were 95,804 repeatersinfirst grade and 28,871 repeatersinsecond grade, it can be deducedthat o f the 623,366 first-graders in 1992-93,365,224 advancedto grade 2 (=394,095 -28,871); and 162,635 dropped out (= 623,663 -365,224 -95,805). Table T2.4. Example Calculation of Promotion, Repetition, and Dropout Rates The repetition, promotion, and dropout rates ingrade i h o l l m e n t (Boys and Girls) 1 1are therefore as follows: /Grades ,1992-93 1993-94 Grade 1 623663 907576 0 Repetition rate = 0.154 (=95,804/623,663) ;Grade2 0 Promotion rate = 0.586 (=365,224/623,663) - 331767 394095 ' 0 iil jRepeatem(Boys andGirls] Dropout rate =0.261 (= 162,635/623,663) /Grades j : 1993-94 95804 Table T2.5 shows these rates for each grade, each 28871 year since 1993-94; this i s the first year that data are Source: Authors' construction. available both on total enrollments andthe number of repeatersamong them - 187- Table T2.5. Promotion, Repetition, Dropout Rates, Ethiopia, 1993-94 to 2001-02 idicator & rade 1992-93 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 romotion rate 1 0.586 0.518 0.537 0.571 0.542 0.557 0.600 0.615 0.615 0.605 2 0.867 0.862 0.850 0.834 0.805 0.797 0.793 0.792 0.783 0.762 3 0.890 0.905 0.906 0.886 0.878 0.831 0.816 0.813 0.809 0.795 4 0.848 0.869 0.868 0.856 0.856 0.840 0.807 0.794 0.789 0.766 5 0.910 0.928 0.953 0.863 0.848 0.824 0.785 0.787 0.783 0.764 6 0.796 0.768 0.817 0.880 0.891 0.907 0.864 0.863 0.884 0.882 7 0.781 0.812 0.778 0.746 0.703 0.707 0.703 0.684 0.688 0.682 8 0.664 0.674 0.692 0.681 0.731 0.760 0.659 0.700 0.774 0.747 9 0.605 0.575 0.569 0.540 0.561 0.610 0.607 0.461 0.566 0.599 10 0.708 0.713 0.720 0.712 0.755 0.734 0.752 0.912 11 0.689 0.644 0.622 0.622 0.613 0.622 0.631 0.663 1.021 0.869 12" tepetition rate 1 0.154 0.139 0.178 0.135 0.167 0.173 0.194 0.192 0.189 0.191 2 0.087 0.075 0.083 0.070 0.091 0.099 0.123 0.130 0.137 0.141 3 0.069 0.055 0.058 0.045 0.055 . 0.082 0.112 0.125 0.129 0.132 4 0.084 0.086 0.085 0.068 0.076 0.077 0.117 0.143 0.149 0.161 5 0.062 0.064 0.060 0.066 0.083 0.086 0.130 0.146 0.146 0.159 6 0.110 0.137 0.131 0.056 0.058 0.058 0.095 0.112 0.108 0.115 7 0.125 0.147 0.178 0.155 0.163 0.163 0.196 0.222 0.231 0.226 8 0.213 0.243 0.228 0.233 0.188 0.222 0.246 0.232 0.212 0.221 9 0.214 0.224 0.244 0.261 0.232 0.232 0.267 0.368 0.302 0.293 10 0.164 0.184 0.187 0.193 0.180 0.158 0.210 0.096 11 0.153 0.197 0.220 0.210 0.225 0.253 0.269 0.254 12 0,000 0.010 0.000 0.000 0.000 0.000 0.034 0.112 Dropoutrate 1 0.261 0.343 0.285 0.294 0.290 0.270 0.206 0.193 0.197 0.205 2 0.046 0.063 0.066 0.096 0.104 0.104 0.084 0.078 0.080 0.098 3 0.041 0.040 0.037 0.069 0.067 0.087 0.072 0.061 0.063 0.073 4 0.069 0.045 0.047 0.075 0.068 0.082 0.076 0.063 0.062 0.074 5 0.028 0.008 -0.014 0.071 0.069 0.090 0.085 0.067 0.071 0.077 6 0.094 0.095 0.052 0.064 0.051 0.035 0.042 0.025 0.008 0.003 7 0.093 0.041 0.045 0.099 0.134 0.131 0.100 0.094 0.081 0.092 8 0.123 0.083 0.080 0.086 0.081 0.019 0.095 0.067 0.014 0.032 9 0.181 0.201 0.187 0.199 0.207 0.158 0.126 0.171 0.132 0.107 10 0.128 0.103 0.092 0.096 0.066 0.108 0.038 -0.008b' 0.658 0.647 11 0.159 0.159 0.158 0.168 0.162 0.125 0.100 0.083 -0.022b' 0.103 12" We: Datarefer t he uromotion,reuetition,and drouout rates betweenthe war in each colunw and the following.year. Shaded cells hiehliehtthe imact of the - - newpolicy beginning 2001-02 to regulate studentflow at grade 10. in I. a. Dataunavailableon transitionto postsecondaryeducation, makingit impossible to distinguishbetween those who stop after grade 12 and those who go on to higher education. b. The two negativeentries, for grade 11in2002-01 and grade 12 in2001-02, are an indication of underlying inaccuraciesinthe data onrepeaters and readmitted.Coincidentally,anew policy was being implementedthat year that radically reducedaccess to grade 11and 12. Source: Computedfrom data in table T2.1. - 188 - SettingUp the Calculations T2.7 Once the repetition, promotion, anddropout rates havebeencomputed, they canbe set up as infigure T2.1 to reconstruct a cohort's movement from grade to grade in each year. Suppose 1,000 students enroll in grade 1 in 1999-00. Using the grade-specific flow rates computed for 2000-01 (as reported in the previous table), we would expect 192 pupils to repeat grade 1 the next school year, 615 pupils to advance to grade 2 the next school year, and 193 pupils to drop out. Figure T2.1. ExampleCalculation in the Reconstructed- T2.8 In 2000-01, of the 192 repeaters in Cohort Method gradel, we would expect 37 (= 192x Grade 1 Grade 2 0.192) to repeat grade 1in2001-02,118 (= 192 x 0.615) to advance to grade 2, and 37 (= 192 x 0.193) to drop out. In2000-01, there would be 615 grade 2 pupils-those who advanced from grade 1in 1999-2000. Year 1 1999-2000 O f these we would expect 80 (= 615 x 0.130) to repeat the grade in 2001-02, 487 (= 615 x 0.792) to advance to grade 3 in 2001-02, and 48 (= 615 x 0.078) to drop Year 2 2000-01 192repeaters 615promotees out. The calculations are repeated in a similar way insubsequent grades and years Source: Authors' construction, basedon data for Grade 1in 2000- to yield a complete picture o f the student 01. flow by grade through the system. The ResultsUsingthe ClassicalAlgorithm T2.9 The classical calculation calls for the foregoing algorithmto berepeatedas manyyears as are needed for members o f the initial cohort to exit the grade, either through promotion to the next grade or through dropping out. Figure T2.2 summarizes the result of the calculations. The last row shows the pattern o f cohort survival: of the 1,000 entrants to grade 1, the number who eventually reach grade 2 i s simply the sum of pupils promoted from grade 1 to grade 2 in the four years it took for everyone inthe cohort to pass out of grade 1, which i s 761 (= 615 + 118 + 23 +4 + 1). Expressed as a percentage o f the 1,000 pupils inthe initial cohort, the survival rate would thus be 0.761 percent. The survival rates to grade 4 are 0.644 percent, and to grade 8 are 0.469 percent. - 189- Figure T2.2. Computing SurvivalRatesUsingthe UNESCO Reconstructed-CohortMethodwith Repetition AlgorithmApplied untilAll RepeatersinEach Grade Exit the System 76I 1 -I-427 16 65 sm 469 + M 41 116 3 311 279 The Resultswith a ModifiedAlgorithm T2.10 The Ethiopian Ministry of Education uses a modified approach to compute survival rates. In particular, the repetition algorithm is applied only twice in each grade, and students who have not been promoted at that point are treated as dropouts. Although this calculation departs from the logic of the reconstructed method, it is viewed as beingmore realistic giventhe current rule that allows children to repeat the same grade no more thantwice. FigureT2.3 shows the adjusted calculations; these are the same as those reported in the Ministry of Education's annual statistical abstract, with the exception that readmitted students are treated as repeaters rather than as dropouts.'40 Strictly speaking, we would more properly describe this approach as a simulation rather than a reconstruction o f student flow. As expected, the simulation yields an appreciably lower survival rate: 0.626 instead o f 0.644 from grade 1 to grade 4; and 0.399 insteadof 0.469 from grade 1to grade 8. For this reason, the results show a survival rate of 0.399 insteadof the much smaller value o f 0.275 reportedinthe Ministry of Education's 2001-02 Annual Statistical Abstract. - 190- FigureT2.3. ComputingSurvivalRatesUsingthe UNESCOReconstructed-CohortMethodwithRepetition AlgorithmAppliedOnly TwicePer Grade a: " - -I. m 86 65 IY 6 44 Note: Repeatersstill left are thus treated as dropouts inthis calculation. CompositeCohortMethod Figure T2.4. ExampleCalculationinthe CompositeCohort T2.11 The idea behindthis method is simple. Using data on nonrepeaters Nonrepeaters, Nonrepeaters, from two adjacent years, we compute. the proportion of pupils in one grade Grade 1 1,598,097 1,677,838 that advances to the next grade. In figure T2.4, the survival rate from 1,104,195 1,202,029 grade 1 to grade 2 is 0.752 (=1,202,029/ 1,598,097), and that from grade 2 to grade 3 is 0.889 I \ Grade 4 667,174 I I 765,231- 1 t (=978,292/1,104,195). The cohort survival rate from grade 1 to grade 3 ~ would then simply be the result of Source: Authors' construction based on data intable A2.2. multiplying the two grade-to-grade survival rates: 0.669. By computing the survival rates for all adjacent pairs of grades, we can compute the survival rates from grade 1to any subsequent grade. Table T2.5 shows the data and the results basedon them. - 191- Table T2.6. Number of Nonrepeaters, Boys and Girls, Ethiopia, 1993-94 to 2002-03 Indicator 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 2002-03 Grade 1 811,772 981,517 1,227,700 1,418,976 1,481,924 1,536,929 1,598,097 1,677,838 1,665,944 1,701,019 Grade 2 365,224 470,468 594,702 813,399 874,174 975,917 1,104,195 1,202,029 1,261,764 1,241,975 Grade 3 287,757 339,536 425,200 530,654 691,024 758,895 849,285 978,292 1,067,608 1,103,466 Grade 4 232,628 276,743 322,800 394,893 483,476 599,469 667,174 765,23 1 886,308 961,021 Grade 5 200,178 219,419 259,239 298,168 358,501 433,094 515,645 589,321 687,894 777,609 Grade 6 189,225 197,717 222,135 235,855 268,102 317,179 365,835 453,578 526,999 600,607 Grade 7 155,117 161,907 185,147 221,534 222,775 256,330 288,024 342,964 440,377 512,033 Grade 8 144,299 144,641 146,334 163,000 179,427 186,896 210,534 237,270 288,860 366,406 Grade 9 109,035 120,820 130,145 128,903 151,388 165,975 155,064 187,908 231,850 263,473 Grade 10 72,63 1 77,500 85,989 90,228 96,695 116,674 127,701 97,413 150,326 187,039 Grade 11 62,486 62,109 67,380 73,677 83,168 85,487 101,566 142,293 32,752 49,062 Grade 12 59,611 48,732 47,885 52,470 56,010 64,528 70,533 87,246 179,365 28,640 Grade-to-aade survival rates Grade 1, - 1.ooo 1.000 1 .ooo 1.ooo 1.ooo 1.000 1.ooo 1.ooo 1.000 Grade 2 - 0.580 0.606 0.663 0.616 0.659 0.718 0.752 0.752 0.746 Grade 3 - 0.539 0.548 0.591 0.523 0.572 0.625 0.666 0.668 0.652 Grade 4 - 0.518 0.521 0.549 0.477 0.496 0.550 0.600 0.605 0.587 Grade 5 - 0.489 0.488 0.507 0.433 0.444 0.473 0.530 0.544 0.515 Grade 6 - 0.483 0.494 0.461 0.389 0.393 0.399 0.467 0.486 0.450 Grade 7 - 0.413 0.462 0.460 0.368 0.376 0.363 0.437 0.472 0.437 Grade 8 - 0.385 0.418 0.405 0.298 0.315 0.298 0.360 0,398 0.363 Grade 9 - 0.322 0.376 0.357 0.277 0.292 0.247 0.322 0.389 0.331 Grade 10 - 0.229 0.268 0.247 0.207 0.225 0.190 0.202 0.311 0,267 Grade 11 - 0,196 0.233 0.212 0.191 0,199 0.166 0.225 0.105 0.087 Grade 12 - 0.153 0.179 0.165 0.145 0.154 0.137 0.193 0.132 0.076 -Data on survivalratesnot computedfor lack of data on enrollments andrepeartersfor 1992-3 Note: Figures in bold font refer to transition rates between cycles; shaded cells signal probable errors in the counts of repeaters (including readmitted students) since transition rates by definition cannot exceed unity. As indicated in table T2.4, the inaccuracy is probably associated with introduction of a new policy in 2001-02 to regulatestudentflow after grade 10. a. Estimatesreflect studentflow amongstudentsenrolledinthe precedingyear. Source: Datain first block, table T2.2; the other two blocks, computedfrom data in the first block - 192- Technical Note 3: Selected Studies on Rates of Return to Education inEthiopia T3.1 We focus on the seven studies published since 1990 that we have been able to access. Using data for 2000, we complement the studies' findings with our own estimates o f the returns to education inEthiopia. Inbrief, most o f the studies, including our own, suffer from serious data and other shortcomings, which makes it difficult to place a highdegree o f confidence inmany o f reported results. Below we provide a short review o f the studies and summarize what are, inour view, the most meaningful estimates as o f this writing on the rates o f return to education in Ethiopia. An Overview of Seven Studies T3.2 The seven studies and our estimate are listed in table T3.1, along with selected details about each study and the underlying database, as well as the corresponding regression model specification and coefficient estimate on the education variables. Following i s a brief comment on each study. T3.3 Study No. 1: Krishnan (undated). The results refer only to young people ages 15-29 in urban areas. Because the data are for 1990, the last year o f the Marxist regime, the results may no longer be relevant inthe country's current economic context. Although separate estimates o f returns are made by gender and sector o f employment, the very small sample sizes-ranging from 59 to 213 observations, depending on gender and sector-are a serious drawback, one that considerably weakens the reliability o f the results. T3.4 Study No. 2: Venvimp (1996). The results are based on a small sample o f 422 male wage earners employed inthe public and private sectors in 1994. An extra year o f schooling yields an estimated return o f 15 percent. T3.5 Study No. 3: Zinash (2001). The data are undated but probably refer to 1994. They pertain to a small sample o f 550 wage earners in urban areas. The coefficient estimate for primary education was not statistically significant; those for secondary and tertiary education have the expected positive sign. The magnitude o f the returns to these two levels o f education i s not documented, however, inthe version o fthe paper to which we had access. T3.6 Study No. 4: Mengistae (1998). The data pertain to 1994 and relate to a sample o f 1,170 full-time male and female employees o f public agencies or private firms in major towns. However, the sample is unlikely to be a representative sample o f Ethiopianworkers inthe formal and informal wage sector since the mean years o f schooling inthe sample were about ten years. According to this study, an additional year o f schooling yields, on average, a private return o f 8.5 percent a year in the public sector and 7.4 percent in the private sector. These figures are probably good estimates o f the returns to educationinthe formal sector. I -* W 2 * & g I 0 *** ** W 2s zq * I W 9 e4 *** 50 ' 4 *** r - W N 2 2 - 9 0 * c c m 2 t 04 m v! m h x0' W G. 0' 9 G d VI W F. m - 194- T3.7 Study No. 5: Woldav (1997). The study relies on data for 1996 for 843 workers ina state- owned enterprise (Edget Cotton Factory) and 170 workers in a formal sector private enterprise (MOENCO). According to this study, an additional year o f schooling yields, on average, a return o f 5 percent inthe public sector and 8 percent inthe private sector. Note that these results stand incontrast to those reported inStudy No. 4 above, which finds that the returns to education are comparable inboth sectors-8.5 and7.4 percent a year, respectively. For employees inthe state- owned firm, the study estimates a return o f 3.3 percent for primary education, 5.5 percent for junior secondary education, and 2.8 percent for senior secondary education; in contrast to these low numbers, the returns soar to 18 percent for diploma holders and 11.3 percent for those qualified beyond the diploma level.'41 T3.8 Studv No. 6: World Bank (1998). The study uses data from the 1995/96 Ethiopia Household Income and Consumption Expenditure Survey. It restricts the observations to a subsample o f households that have sole wage earners ages 15 to 65 who reported positive labor market earnings. The resulting subsample is thus highlyrestricted and is inno way representative o f the population. This study finds that an additional year o f schooling yields a return o f 23 percent a year, which is very high compared with estimates reported in the other studies (e.g., Studies No. 4 and 5 report returns ranging between 5.0 and 8.5 percent a year). The wage specification involving dummy variables for education (estimated only for males) also implies very high returns by level o f education: 37 percent a year for primary education, 55 percent for secondary education, and 48 percent for higher education.'42 These magnitudes are sufficiently out of line with most o f the results inthe other studies as to raise doubts about their reliability. T3.9 Study No. 7: Krishnan, Selassie, and Dercon (1998). Based on data from the 1997 Ethiopian Urban Household Survey, this study estimates separate earnings functions for urban men and women ages 14-65 by sector o f employment. The sample sizes are relatively small, rangingbetween 125 and 336 observations depending on gender and sector o f activity. The study estimates the returns to education by introducing a set o f three dummy variables for education: (a) whether or not a person has completed primary education; (b) whether or not he or she has completed secondary education; and (c) whether or not he or she has completed higher education (as such, the omitted category comprise individuals with less than primary schooling). Because o f the small number o f observations for women, the estimated returns to education for them are probably not reliable. For men employed inthe public sector, the study estimates a return o f 10.6 percent for primary education, and 15 percent for both secondary and higher education (see table T3.2); for those working in the private sector, the returns are virtually 0 for primary education, while they are estimated at 8.2 percent a year for secondary education and 21.5 percent for higher education. 14'As reportedinWoldayAmha (1997, table3). 14' Rates of retums to different levels of education are calculated as [exp(bj-bi)-l]/(sj-si) where bi and bj are the regression coefficients of two successive levels of completededucationand sj and si standfor the total number of years of schoolinginboth levels iandj. As inWorld Bank (1998), we have assigned3 years of forgone earnings for wage earnerswith primary education completed, 6 years for those with secondaryeducation, and4 years for thosewith higher education. - 195 - Table T3.2. Returns to Education inEthiopia, Computed Using Regression Coefficients inKrishnan, Selaissie, and Dercon (1998) Male Female Educationlevel Publicsector Privatesector Publicsector Privatesector Primary education 10.6 ns. ns. ns. Secondaryeducation 15.0 8.2 ns. 20.9 Higher education 15.1 21.5 ns. 9.0 T3.10 Study No. 8: Our estimates based on dat,a from the 1999/2000 Ethiopia Household Income and Consumption Expenditure SurvevNelfare Monitoring Survey. As in Study No. 6 described above, the data are such as to require restriction o f the sample to men ages 14 to 65 who are the sole wage earners in their households and who reported positive labor market earnings. As in that study, this restriction drastically reduced the sample size, cutting it down from 17,000 observations to 2,000. The resulting subsample contained an overrepresentation o f public sector workers: 54 percent compared with 13 percent-the share o f comparable workers inthe universe of wage earners. We thus draw a random sample of the public sector workers in the data, in order to reduce their share to 13 percent. As a result o f this procedure, the final subsample o f data contained only 1,024 observations. On average, the workers inthis subsample have three years o f schooling. The regression estimates imply that an additional year o f education yields a private retum of 15.6 percent a year-a return that is smaller thanthe estimate reported inthe earlier World Bank study (Study No. 6) and comparable to that reported in Study No. 2. However, as before, the fact that the data refer to a specific type o f household (i.e., households with only one wage earner), the estimated returns are probably biased; a further difficulty i s that little can be said about the direction andmagnitude o f the bias. - 196- Sample Regressors average Coefficient Coefficient Age 37.92 -0.0047 -0.0049 AgeZ - 0.0002 0.0002 Years o fschooling 3.54 0.1563* - Educationdummies" Primaryeducation 0.31 - 0.6800* Secondaryeducation 0.11 - 1.6112* Higher education 0.06 - 2.2517* Intercept - 3.7158* 3.7378* RZ - 0.2455 0.260 Numberofobservationsinthe subsample analyzed 1,024 n.a. n.a. Numberofprivatesector workers inthe survey 16,834 n.a. n.a. Summary of the Estimatesof theReturnsto Education T3.11 None of the estimates based on the eight analyses discussed above is perfect. Nonetheless, the shortcomings in Studies 1, 3, and 6 are sufficiently serious for us to discard the estimates. W e also exclude the estimates reported in Study No. 5, since the data refer to workers inonly two enterprises; we exclude those for urbanworkers inStudy No. 7 because o fthe tiny sample on which the estimates were based. Table T3.4 summarizes what we consider to be the most meaningful estimates fiom the studies we reviewed. - 197- Table T3.4. Summary of the Rates of Returns to Education, Ethiopia I Urban areas Public and private sectors Private sector Formal and Rate of return Public sectora' Formalb' informal" Average rate ofreturn 8.5 7.4 18.0 15.6 (1994; 10.3) (1994; 9.2) - (1994) (2000; 3.5) Rateofreturnbylevel' Primary 10.6 - ns. 32.5 Secondary 15.0 - 8.2 19.2 Higher 15.1 - 21.5 22.4 (1997) (1997) (2000) Note: Figures in parentheses refer, respectively, to the year of the data and the average years of schooling in the sample of observations. a. Estimates of average rate of retum pertain to full-time employees of both sexes working in public and private enterprises or organizations in major towns, based on the results in Mengistae 1998, table 3.6; estimates of the retums by level of education pertain to urban male workers ages 14-65, computed by the authors from the results reported in Krishnan, Selassie, and Dercon 1998, table A.3, usingthe methoddescribedinfootnote g below. b. Estimatesrefer to the retums for full-time employees of both sexes in public and private sector enterprises or organizations in major towns, basedon the results inMengistae 1998, table 3.6. c. Estimatesrefer to urbanmale workers ages 14-65, computedby the authors from the results reported in Krishnan, Selassie, and Dercon 1998, table A.3 usingthe methoddescribedinfootnote f below. d. Estimatesrefer to ruralmale full-time nonfarmers, asreported inVenvimp 1996, table 3. e. Authors' estimatesbased on the 1999/2000HICES/WMS (subsample of male householdheads ages 14-65 in householdswhere they are the only wage eamers andthey reportpositive labormarket eamings). f. The rates of retums to each level of education are calculated as [exp@j-bi)-l]/(sj-si) where bi and bj are the regression coefficientsof two successive levels of completed education and sj and si stand for the total number of years of schoolinginboth levelsiandj. We have assigned 3 years of forgone eamings for wage eamers with primary education completed, 8 years for those withsecondary education,and4 years for those with highereducation. Source: See footnotes above. - 198 - Appendix Tables Table Al.l. GDP andOverallGovernmentRevenues,Ethiopia, 1981-2002 Governmentrevenues llionsof c rrentB r) GDP 2DP (millions deflator Total Total Per capita Estimated Ethiopian lregorian of current (1987 including ?xcluding revenues population Calendar calendar Birr) E.C.=100) grants grants Zrants ?xcl.grants) (millions) 1973 1980181 10,721.3 40.3 1,947.3 1,756.9 190.4 45.3 38.8 1974 1981182 11,280.9 41.9 2,138.3 1,876.7 261.6 47.1 39.9 1975 1982183 12,540.0 43.2 2,433.8 2,174.5 259.3 53 41 1976 1983184 11,849.3 41.8 2,547.5 2,293.8 253.7 54.4 42.2 1977 1984185 13,876.2 55.3 2,954.3 2,323.0 631.3 53.6 43.4 1978 1985186 14,4 14.5 52.4 3,249.2 2,806.1 443.1 62.8 44.7 1979 1986186 15,407.7 49.1 3,247.9 2,925.9 322.0 63.5 46.1 1980 1987188 15,901.5 50.6 4,103.0 3,467.1 635.9 72.8 47.6 1981 1988189 16,773.9 53.5 4,698.2 3,899.2 799.0 79 49.3 1982 1989190 17,788.4 55.3 3,544.0 3,142.6 401.4 61.4 51.2 1983 1990191 19,688.3 66.0 3,169.6 2,706.3 463.3 51.1 53 1984 1991/92 20,792.0 76.4 2,751.1 2,208.1 543.0 40.3 54.8 1985 1992193 26,671.4 86.4 3,657.4 3,191.3 466.1 59.9 53.3 1986 1993194 28,328.9 88.7 4,926.1 3,938.9 987.2 71.8 54.9 1987 1994195 33,885.0 100.0 7,044.6 5,912.9 1,131.; 104.6 56.5 1988 1995196 37,937.6 100.9 8,063.0 6,966.2 1,096.8 119.6 58.2 1989 1996197 41,465.1 104.7 9,034.1 7,530.1 1,504.C 126 59.8 1990 1997198 44,840.3 115.4 9,373.9 8,100.6 1,273.: 132.2 61.3 1991 1998199 48,687.5 118.2 10,4 15.0 8,653.0 1,762.C 137.8 62.8 1992 199912000 52,074.2 121.9 11,222.5 9,498.5 1,724.( 147.7 64.3 1993 2000101 52,871.7 114.1 12,819.0 10,191.O 2,628.( 154.8 65.8 1994 2001102 59,108.5 106.0 13,893.0 11,306.0 2,587.( 167.9 67.3 ource: Africa EconomicDevelopment. - 199- Table A1.2. Evolution of Revenue-Sharing F `mula, Ethiopia, 1994-95 to 2001-02 Wei its assignec o indicator yo.) Indicators inweighting formula 1994-95 1996-97 1997-98 2001-02 Population 30 33.3 60 55 I-distanced 25 Own revenue to budget ratio 20 33.3 15 Capital budget allocation in 1993-94 15 Area 10 Level of developmentlexpenditure needsb' 33.3 25 20 Levelofpovertyd 10 Revenue-raising effort and sectoral output performance - - - 15 a. Refers to an index of development distance, computed on the basis of indices for education, health, road density, electric power sales, telephoneline density, and safe drinkingwater. b. Computed usinga series of indicatorsrelating to education, health, access to safe drinking water, roads, andrecurrentbudget implication, as follows. For education, the indicators include:(i)inverseof studentto classroomration; (ii)unitcost of primary school construction; (iii) inverseof pupil-teacher ratio; (iv) primary and secondary schoolparticipation rates. For health, they include:(i) center to population health ratio; (ii)hospital bed to population ratio; (iii) unit cost of constructing health center; (iv) under 5 mortality rate; (v) doctors per 100,000 population; (vi) healthassistantsper 100,000, and primary healthservicescoverage.For access to safe drinkingwater, the indicator is simply percentageof the populationcovered. For roads, it is the unit cost of constructing a kilometerof ruralroad. For recurrentbudgetimplication, the indicatori s the region's share of the numberof woredusand area. c. Measuredby: (i)own revenueto incomeratio adjustedby populationshare; (ii)change in primary schoolparticipationrate; (iii) changein the number of health centers; and (iv) changein the lengthof ruralroadconstructed. Source: Govt. of Ethiopia2000a. - 200 - Table A1.3. Government Expenditures, Ethiopia, 1981-2002 (millions of cu ent Birr) ~ Goveri lent expenc ures Intere payment n debt Recurrer spending Net of Net of all mxternal debt Ethiopian hegorian Iomestic ixternal .ebtinterest interest Calendar calendar Total Recurrent Capital U1deb1 debt debt payment payment 1973 1980181 2,28 1.5 1,776.4 505.1 84.2 65.0 19.2 1,692.2 1,757.2 1974 1981182 2,780.8 2,065.8 715.0 83.8 66.1 17.7 1,982.0 2,048.1 1975 1982183 3,786.1 2,540.8 1,245.3 91.3 70.5 20.8 2,449.5 2,520.0 1976 1983184 3,347.8 2,414.8 933.0 147.4 120.5 26.9 2,267.4 2,387.9 1977 1984185 3,883.4 2,696.4 1,187.0 225.4 196.4 29.0 2,471.0 2,667.4 1978 1985186 4,073.2 2,601.4 1,471.8 194.0 155.2 38.8 2,407.4 2,562.6 1979 1986186 4,023.7 2,640.6 1,383.1 222.9 151.8 71.1 2,417.7 2,569.5 1980 1987188 4,893.0 3,43 3.9 1,459.1 256.2 156.9 99.3 3,177.7 3,334.6 1981 1988189 5,731.5 3,791.9 1,939.6 248.7 164.3 84.4 3,543.2 3,707.5 1982 1989190 5,280.9 3,840.8 1,440.1 227.6 180.4 47.2 3,613.2 3,793.6 1983 1990191 4,849.4 3,635.4 1,2 14.0 263.2 212.7 50.5 3,372.2 3,584.9 1984 1991192 4,205.3 3,253.5 951.8 307.3 246.0 61.3 2,946.2 3,192.2 1985 1992193 5,219.2 3,434.3 1,784.9 530.3 408.9 121.4 2,904.0 3,312.9 1986 1993194 7,094.0 4,399.5 2,694.5 956.9 809.9 147.0 3,442.6 4,252.5 1987 1994195 8,371.9 5,2 15.4 3,156.5 838.6 596.8 241.8 4,376.8 4,973.6 1988 1995196 9,144.7 5,582.0 3,562.7 922.5 609.6 312.9 4,659.5 5,269.1 1989 1996197 9,906.0 5,738.0 4,168.0 918.7 635.3 283.4 4,819.3 5,454.6 1990 1997198 11,229.0 7,082.0 4,147.0 835.7 526.1 309.6 6,246.3 6,772.4 1991 1998199 14,9 16.0 10,126.0 4,790.0 957.0 588.0 369.0 9,169.0 9,757.0 1992 1999100 17,183.6 13,741.6 3,442-0 1,122.0 723.0 399.0 12,619.6 13,342.6 1993 2000101 15,370.0 10,352.0 5,018.0 1,080.0 575.0 505.0 9,272.0 9,847.0 1994 2001102 17,394.0 11,366.0 6,028.0 Source: Africa Live Database, World Bank 2003b; Govt. of Ethiopia various; Budgetary Revenues and Expenditures; Ministry of Finance and Economic Development;WorldBank 1998; PublicExpenditureReview.All figures are incurrentprices. - 201 - Table A1.4. Current and CapitalGovernmentSpendingonEducation,Ethiopia, 1980-81 to 2001-02 Spendingon education Spendingon education it Birr) E.C. Birr) GDP GDP Ethiopian Gregorian lmillion current deflator Calendar calendar Birr) Total :1987=100) Total 1973 1980181 10,721 228.3 40.3 567.2 1974 1981182 11,281 270.6 41.9 645.1 1975 1982183 12,540 276.7 43.2 639.9 1976 1983184 11,849 295.6 41.8 706.8 1977 1984185 13,876 327.0 55.3 591.5 1978 1985186 14,415 344.8 52.4 658.6 1979 1986186 15,408 375.3 49.1 763.6 1980 1987188 15,902 403.0 50.6 795.9 1981 1988189 16,774 484.7 53.5 905.7 1982 1989190 17,788 495.6 55.3 895.6 1983 1990f91 19,688 489.7 66.0 742.5 1984 1991/92 20,792 528.5 76.4 691.8 1985 1992193 26,671 - - 745.8 86.4 - - 862.8 1986 1993194 28,329 848 274 1,121.5 88.7 955.4 308.8 1,264.1 1987 1994195 33,885 913 287 1,199.9 100.0 913.1 286.8 1,199.9 1988 1995196 37,938 979 457 1,436.8 100.9 970.2 453.1 1,423.3 1989 1996197 41,465 1,084 427 1,511.5 104.7 1,035.5 408.2 1,443.6 1990 1997198 44,840 1,175 408 1,582.4 115.4 1,018.3 353.3 1,371.6 1991 1998199 48,688 1,266 454 1,720.1 118.2 1,070.6 384.2 1,454.8 1992 1999100 52,074 1,343 296 1,638.8 121.9 1,102.2 242.5 1,344.7 1993 2000101 52,872 1,564 623 2,186.8 114.1 1,370.5 546.3 1,9 16.9 1994 2001102 59,109 1,799 570 2,369.2 106.0 1,696.8 537.4 2,234.2 ible. Note: Datafor the last two years refer to budgets; see appendixtable A1.5 for the rearrangementof sourcebudget on capital spending. Source: Africa Live Database, World Bank2003b; Govt. of Ethiopia various; BudgetaryRevenues and Expenditures; Ministry of Financeand Economic Development; World Bank 1998; Public ExpenditureReview.All figures are incurrentprices. - 202 - Table A1.5. CapitalSpendingonEducation,Ethiopia, 1993-94 to 2001-02 (millions of current Birr) Capitalspendingincludedin reported Capital Educationbuilding spendingas Ethiopian Gregorian Vehicles & and construction reportedin Total capital calendar calendar equipment material budgetdata spending 1986 1993-94 15.4 2.6 256.0 273.9 1987 1994-95 14.8 2.8 269.2 286.8 1988 1995-96 13.7 1.9 441.9 457.4 1989 1996-97 6.4 0.9 420.1 427.3 1990 1997-98 12.2 0.0 395.4 407.6 1991 1998-99 12.3 0.1 442.0 454.3 1992 1999-00 7.8 0.0 287.8 295.5 1993 2000-0 1 17.7 0.1 605.5 623.2 1994 2001-02 30.1 0.0 539.8 569.9 Source: Analysis o f budget data files from the Ministry of Financeand Economic Development(MOFED). See technicalnote 1 on the preparationof the data onpublic spendingon education. c c c c c 4 C .f4 b eC s P 7 iI c9 i 3eC I I 3 -i b de ; I 5 C I0 .Ec S I d cC 9L ..5 c 3P a - 204 - TableA2.2. Percentageof PopulationEver EnrolledinGrade 1 at EachAge, Ethiopia, 1996 to 2000 LaborForce Welf e MonitoringSI veys Survey Jan./Feb. 1996 darJApri11998 Jan./Feb. 2000 May 1999 Age ( S Y 1995-96) (SY: 1997-98) (SY: 1999-00) (SY: 1998-99) 7 12.6 17.3 20.6 21.9 8 17.0 24.7 31.5 29.3 9 23.2 32.2 38.4 37.7 10 27.6 38.7 46.3 45.7 11 30.4 44.1 51.0 50.8 12 32.1 43.3 51.9 51.4 13 34.2 43.7 53.2 54.4 14 36.7 45.5 51.8 54.6 15 37.4 41.4 49.7 - 16 37.0 41.1 46.1 - 17 36.9 42.6 46.7 - 18 35.2 38.1 39.8 - SY Schoolyear. -Datanotavailable Note: The first date in each column refers to the date of data collection in each survey; the second, to the corresponding school year (SY). Source: Authors' estimatesbased on the indicated surveys. Table A2.3. Cohort CompletionRates, Ethiopia, 1995-96 andcirca 2000 Cross-sectional Compositecohortrates rates Grade 1995-96 2001-02 2000 1 37.2 60.5 90.70 2 24.6 45.1 81.32 3 22.0 39.4 74.03 4 20.4 35.5 65.56 5 18.9 31.2 52.72 6 17.2 27.2 41.80 7 17.1 26.4 38.91 8 15.1 22.0 30.92 9 13.3 20.1 21.90 10 9.2 16.2 17.82 11 7.9 5.3 14.49 12 6.1 4.6 6.39 table A2.2 and survival rates reported in technical note kble T2.5; see- text seckon on completion rates for an explanation o fthe difference between cohort and cross-sectional survival rates. - 205- Table A3.1. RecurrentPublicSpendingonEducationby Region,Ethiopia, 1993-94 to 2001-02 (inmillions ofcurrentBirr) 1986E.C. 1987E.C. 1988E.C. 1989E.C. 1990E.C. 1991E.C. 1992E.C. 1993 E.C. 1994 E.C. Region 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 Tigray 39.0 43.5 52.6 59.9 64.0 70.3 72.2 82.0 96.7 Afar 15.2 16.6 13.2 8.0 8.8 12.5 16.3 17.7 24.5 Amhara 163.7 179.0 205.5 212.9 222.0 232.4 229.8 259.0 280.2 Oromiya 259.4 283.1 321.6 344.3 370.2 408.0 416.4 460.2 540.3 Somali 19.1 21.2 15.4 14.2 13.0 13.7 16.8 22.7 26.1 Benshangul-Gumuz 12.4 13.7 14.3 17.7 21.0 22.4 22.4 25.0 25.3 SNNPR 129.8 141.2 147.3 184.0 195.5 218.7 218.0 250.8 261.6 Gambella 6.7 7.6 8.2 10.3 11.8 13.6 16.7 16.2 20.3 Harari 7.5 7.7 8.1 10.0 12.3 11.9 13.2 16.7 16.1 Addis Ababa 75.7 83.4 76.8 81.5 94.5 93.9 109.1 126.0 146.1 Dire Dawa 5.7 6.2 7.2 6.8 7.8 8.6 8.4 11.6 14.3 Federal 113.4 109.9 109.3 134.4 154.0 159.7 203.9 275.6 347.8 Ethiotia 847.6 913.1 979.4 1,084.1 1,174.8 1,265.8 1,343.2 1.563.5 1.799.3 Memorandum: GDP deflator 88.7 100.0 100.9 104.7 115.4 118.2 121.9 114.1 106.0 SNNPR Southem Nations and Nationali i Peoples Re blic. Source: Electronic data files from MOFl ,supplemen1 withestimates basedonannual school censuses. Table A3.2. CurrentExpendituresby Function, Ethiopia,1993-94 to 2001-02 (1986 E.C. to 1994 E.C.) (inmillions ofCUI 1986E.C. 1987E.C. 1988E.C. 1989E.C. 1990E.C. 1991E.C. 1992E.C. 1993E.C. 1994 E.C. Indicator 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 1999-00 2000-01 2001-02 Administration 42.6 48.0 67.8 102.0 119.3 151.1 157.0 169.3 190.0 Primary & secondary 685.4 737.3 783.6 820.3 860.1 901.2 926.4 1,062.7 1,134.7 Grades 1-8 583.8 628.8 670.7 722.7 761.0 798.8 822.0 915.1 979.7 Grades 9-12 101.6 108.5 112.9 97.6 99.1 102.4 104.3 147.6 155.0 TVET 6.3 8.2 7.1 7.7 15.2 17.0 16.2 25.1 91.4 TTI 12.8 12.7 13.6 15.6 14.3 19.0 19.5 16.9 21.4 TTC 3.7 4.0 3.9 7.7 11.5 16.3 15.8 18.1 20.4 Higher education 78.8 86.0 89.0 119.1 140.3 145.6 190.2 247.1 310.5 Other 18.1 16.8 14.6 11.7 14.2 15.5 18.1 24.3 31.0 Total 847.6 913.1 979.4 1.084.1 1,174.8 1,265.8 1,343.2 1,563.5 1,799.3 Memoranda: I GDP Deflator 88.7 100.0 100.9 104.7 115.4 118.2 121.9 114.1 106.0 GDP 28,328.9 33,885.0 I 37,937.6 41,465.1 44,840.3 II 48,687.5 52,074.2 52,871.7 59,108.5 TTC Teacher training inst! tes:, TTI Teacl.trainingcolleges; TVET Tec iical and voc mal education and training. Source: Electronic data files fromMOFED, supplemented with estimatesbasedon annual school censuses. - 206- Table A3.3. Recurrent Spendingby Subsector andRegion,Ethiopia, 1993-94 to 2001-02 (inmillionsofcurrentB&) - . 1986 1987 1988 1989 1990 1991 1992 1993 1994 Regiodsubsector E.C. E.C. E.C. E.C. E.C. E.C. E.C. E.C. E.C. 1993-94 1994-95 1995-96 1996-97 1997-98 1998-99 199900 2000-01 2001-02 Tigray Administration 2.5 3.2 4.1 4.5 4.7 4.8 4.9 5.8 6.4 Primary & secondary 33.4 36.7 45.2 53.0 55.4 59.9 61.3 69.7 82.8 Grades 1-8 29.6 32.3 39.9 48.1 50.7 54.9 56.9 64.2 73.2 Grades 9-12 3.9 4.5 5.3 4.9 4.7 5.o 4.4 5.4 9.6 TVET 0.4 0.6 0.7 0.7 1.1 1.2 1.7 1.7 2.0 TTI 1.o 1.3 1.4 1.2 1.1 1.2 1.4 1.6 1.7 TTC 0.0 0.0 0.0 0.0 0.6 1.9 1.6 1.8 2.1 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.5 0.8 Other 1.6 1.7 1.1 0.5 1.1 1.3 1.2 1.o 1.o Total 39.0 43.5 52.6 59.9 64.0 70.3 72.2 82.0 96.7 Afar Administration 0.7 0.9 1.o 1.4 2.1 3.0 6.2 3.O 4.4 Primary & secondary 14.4 15.5 12.0 6.6 6.7 9.4 10.1 14.7 19.1 Grades 1-8 12.1 13.1 10.1 6.0 5.2 7.2 7.4 8.7 10.8 Grades 9-12 2.3 2.5 1.9 0.7 1.5 2.2 2.7 5.9 8.4 TVET 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTI 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTC 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Other 0.1 0.1 0.1 0.0 0.0 0.0 0.0 0.0 1.o 15.2 16.6 13.2 8.0 8.8 12.5 16.3 17.7 245 Administration 3.8 5.3 23.4 23.9 30.8 33.4 25.0 29.9 36.4 Primary & secondary 154.6 166.9 175.0 180.2 182.6 187.7 197.2 219.3 230.0 Grades 1-8 136.7 147.6 155.1 163.3 166.2 170.9 180.9 200.1 210.4 Grades9- 12 17.9 19.3 19.9 16.9 16.3 16.8 16.3 19.3 19.6 TVET 0.6 0.8 0.7 1.o 0.8 0.9 0.8 1.4 4.8 2.1 3.1 3.4 3.9 2.1 3.1 1.6 1.8 1.8 0.0 0.0 0.0 0.7 1.9 2.6 1.3 1.8 1.8 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.7 2.9 3.O 3.3 3.8 4.6 3.8 4.9 5.4 163.7 179.0 205.5 212.9 222.0 232.4 229.8 259.0 280.2 Administration 4.7 6.5 7.4 27.2 35.6 50.8 54.1 40.3. 44.2 Primary & secondary 245.8 265.4 302.4 305.3 320.0 339.7 346.8 395.0 434.5 Grades 1-8 214.1 230.1 261.3 272.3 287.3 307.5 313.3 338.9 370.2 Grades9-12 31.7 35.3 41.1 33.1 32.8 32.2 33.5 56.1 64.3 TVET 0.8 1.o 1.1 1.2 3.0 4.1 4.2 10.7 42.6 TTI 3.2 4.6 4.9 5.6 5.6 5.7 4.8 4.7 7.8 TTC 0.0 0.0 0.0 1.3 1.7 2.6 2.9 3.1 3.1 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Other 4.9 5.6 5.7 3.6 4.4 5.1 3.6 6.2 8.1 Total 259.4 283.1 321.6 344.3 370.2 408.0 416.4 460.2 540.3 Somali Administration 1.1 1.3 2.9 5.4 4.8 6.2 5.5 2.8 2.9 - 207- Primary& secondary 17.9 19.7 12.4 8.7 8.3 6.8 9.9 18.2 20.8 Grades 1-8 15.7 17.0 10.9 7.7 7.5 6.3 9.0 15.8 18.7 Grades9-12 2.2 2.7 1.6 1.o 0.7 0.5 1.o 2.4 2.1 TVET 0.0 0.0 0.0 0.0 0.0 - - - 0.7 TTI 0.0 0.0 0.0 0.0 0.0 0.8 1.4 1.7 1.7 TTC 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Other 0.2 0.2 0.0 . 0.0 0.0 0.0 0.0 0.0 0.0 Total 19.1 21.2 15.4 14.2 13.0 13.7 16.8 22.7 26.1 Benshangul-Gumuz Administration 0.9 1.2 1.6 2.1 2.5 2.9 3.4 4.1 4.5 Primary& secondary 11.3 12.3 12.5 15.3 18.1 19.2 18.7 20.6 20.2 Grades 1-8 10.3 11.1 11.4 13.5 16.3 16.7 16.7 18.1 18.2 Grades 9-12 1.1 1.2 1.1 1.8 1.9 2.5 2.0 2.6 2.0 TVET 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTI 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTC 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Other 0.2 0.2 0.1 0.3 0.3 0.3 0.3 0.4 0.7 Total 12.4 13.7 14.3 17.7 - 21.0 22.4 22.4 25.0 25.3 SNNPR Administration 4.9 8.5 7.7 15.0 15.3 23.8 22.8 23.7 25.0 Primary& secondary 120.9 126.6 135.1 161.7 168.0 179.0 179.6 207.2 211.4 Grades 1-8 107.1 111.8 119.6 148.2 154.9 166.9 167.3 190.9 194.9 Grades9-12 13.7 14.8 15.5 13.5 13.1 12.0 12.3 16.4 16.5 TVET 0.0 0.7 0.8 0.8 3.5 4.1 3.5 4.3 6.1 TTI 1.6 2.4 2.4 3.4 4.1 6.9 7.1 4.0 5.2 TTC 0.0 0.0 0.0 1.5 3.0 4.3 4.3 5.6 7.0 Highereducation 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Other 2.4 3.1 1.3 1.6 1.5 0.8 0.7 6.0 6.9 Total 129.8 141.2 147.3 184.0 195.5 218.7 218.0 250.8 261.6 Gambella Administration 0.7 1.o 1.4 1.9 2.4 2.9 3.6 3.8 4.0 Primary& secondary 5.5 6.0 6.1 7.6 8.4 9.6 10.5 10.4 13.4 Grades 1-8 5.0 5.4 5.5 7.0 7.5 8.5 9.5 9.5 10.2 Grades 9- 12 0.5 0.6 0.6 0.6 0.9 1.1 1 .o 0.9 3.2 TVET 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.2 TTI 0.4 0.5 0.6 0.7 0.9 1.o 2.4 1.9 2.1 TTC 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Highereducation 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.1 0.1 0.1 0.2 0.2 0.5 ' 6.7 7.6 8.2 10.3 11.8 13.6 16.7 16.2 20.3 Administration 0.4 0.4 0.4 0.7 1.o 1.1 1.1 1.2 1.3 Primary& secondary 6.2 6.5 6.9 8.4 10.7 10.3 11.2 14.1 13.6 Grades 1-8 5.2 5.5 5.6 6.5 8.8 8.4 9.2 10.4 10.3 Grades9-12 1.o 1.o 1.2 1.9 1.9 1.9 2.0 3.7 3.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 'TTI 0.8 0.7 0.8 0.8 0.5 0.4 0.8 1.2 1.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 E i e reducation 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ~ - 208- Ither 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.1 0.1 rota1 7.5 7.7 8.1 10.0 12.3 11.9 13.2 16.7 16.1 iddisAbaba idministration 2.0 2.8 4.5 7.2 8.7 9.7 18.2 27.5 25.6 'rimary & secondary 69.6 75.4 68.6 66.6 74.8 72.5 74.3 84.9 79.1 Grades 1-8 43.3 49.9 45.4 44.6 50.9 46.1 46.7 52.8 56.2 Grades 9-12 26.3 25.6 23.2 22.1 23.9 26.4 27.6 32.0 22.9 rVET 2.5 4.6 3.1 3.3 6.2 6.0 5.4 6.2 33.3 l-TI 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTC 0.0 0.0 0.0 4.1 4.4 4.9 5.6 5.9 6.4 3igher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3ther 1.6 0.6 0.5 0.3 0.5 0.7 5.6 1.6 1.7 rota1 75.7 83.4 76.8 81.5 94.5 93.9 109.1 126.0 146.1 Dire Dawa 4dministration 0.1 0.2 0.2 0.3 1.o 1.2 1.3 2.1 2.9 ?rimary & secondary 5.0 5.3 6.1 5.9 6.2 6.6 6.6 8.6 9.7 Grades 1-8 3.9 4.1 4.6 4.7 4.7 5.0 5.1 5.7 6.6 Grades 9-12 1.1 1.2 1.5 1.2 1.4 1.6 1.5 2.9 3.2 NET 0.4 0.5 0.7 0.6 0.6 0.7 0.5 0.9 1.7 lT1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 lTC 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Higher education 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3ther 0.2 0.2 0.2 0.0 0.0 0.1 0.0 0.0 0.0 rota1 5.7 6.2 7.2 6.8 7.8 8.6 8.4 11.6 14.3 Federalgovernment Administration 20.9 16.7 13.1 12.4 10.5 11.2 11.0 25.0 32.6 Primary & secondary 0.7 1.o 1.1 1.o 1.o 0.4 0.1 0.1 0.1 Grades 1-8 0.7 1.o 1.1 1.o 0.9 0.4 0.0 0.0 0.0 Grades 9-12 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.1 0.1 WET 1.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 TTC 3.7 4.0 3.9 0.0 0.0 0.0 0.0 0.0 0.0 TTI 3.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Higher education 78.8 86.0 89.0 119.1 140.3 145.6 190.2 246.6 309.7 Other 4.0 2.2 2.3 2.0 2.3 2.5 2.6 3.9 5.4 Total 113.4 109.9 109.3 134.4 154.0 159.7 203.9 275.6 347.8 Ethiopia Administration 42.6 48.0 67.8 102.0 119.3 151.1 157.0 169.3 190.0 Primary & secondary 685.4 737.3 783.6 820.3 860.1 901.2 926.4 1062.7 1134.7 Grades 1-8 583.8 628.8 670.7 722.7 761.0 798.8 822.0 915.1 979.7 Grades 9-12 101.6 108.5 112.9 97.6 99.1 102.4 104.3 147.6 155.0 TVET 6.3 8.2 7.1 7.7 15.2 17.0 16.2 25.1 91.4 TTC 12.8 12.7 13.6 15.6 14.3 19.0 19.5 16.9 21.4 TTI 3.7 4.0 3.9 7.7 11.5 16.3 15.8 18.1 20.4 Higher education 78.8 86.0 89.0 119.1 140.3 145.6 190.2 247.1 310.5 Other 18.1 16.8 14.6 11.7 14.2 15.5 18.1 24.3 31.0 Total 847.6 913.1 979.4 1084.1 1174.8 1265.8 1343.2 1563.5 1799.3 Memoranda: GDP deflator 88.7 100.0 100.9 104.7 115.4 118.2 121.9 114.1 106.0 -Datanotavailable. SNNPR Southem Nations andNationalities PeoplesRepublic. Source: Electronicdata files from MOFED, supplementedwith estimates basedon annual schools census - 209- Table A3.4. Earnings Function for Teachers inGrades 1-12 inGovernment Schools, Oromiya, 2002-03 Variable Sample ,mean Coefficient t-statistic Experience (inyears) 12.2 0.060** 248.89 Experience2 213.8 -0.00 1 ** -135.79 Teacher certification` TTC (omitted category) 2.0 n.a. n.a. ' TTI 84.8 -0.183** -42.27 BA/MA 2.9 0.308** 53.37 Diploma 7.6 0.03 1 6.53 1,2, or 3 years ofuniversity 2.0 -0.020* * -3.46 Grades 9, 10, 11, 12, 1211 0.5 -0.221** -24.46 Other 0.3 -0.133 ** -10.9 Teaching assignment Grades 1-4 (omitted category) 59.5 n.a. n.a. Grades 5-8 31.0 0.029** 21.48 Grades 9-12 9.6 0.109** 34.55 6114.0 1371.13 R-squared n.a. 0.84 n.a. Number of observations 45,439 n.a. n.a. Memorandum on the dependent variable b/: In(monthly salary) 6.5 n.a. n.a. Monthly salary (excl. pension; inBirr) 683 n.a. n.a. BA Bachelor's degree; MA Master's degree; TTC Teacher traininginstitutes; TTI Teacher trainingcolleges; WET Technical andvocationaleducationandtraining. n.a. Not applicable ** Statisticalsignificanceat the 1%confidencelevel. a. The certificationcategoriesinclude universitydiploma; incomplete university studies lastingone to three years; and secondary school completion at grades 9-12, with additional sundry training in some cases; and a residual "other" categorycomprising teacherswithqualifications other than those listedabove. b. The dependentvariable is the monthlysalaryexpressedon a logarithmscale. Source: Analysis of a data file on teachers suppliedby the OromiyaBureauof Education iflthecontext ofthis study. - 210- Table A3.5. Teacher Salary ScaleinPrimary and Secondary Education, Ethiopia (effective February 2001) I Level of instr lion Secondary and teacher Salarv Scale (Birr Der month) training institute (TTI) - 4 502 630 1672 716 Junior 1 1 1 teacher 475 502 530 560 595 630 672 716 760 805 850 Teacher Beginner teacher 560 595 630 672 716 760 805 850 895 942 990 Senior Junior teacher 1I1 1II Teacher 672 716 760 805 850 895 942 990 1,145 Associate Beginner 805 850 895 942 990 1,310 teacher teacher Junior teacher 942 990 1,040 1,092 1,145 1,500 Associatc Teacher 1,092 1,145 1,200 1,255 1,310 1,370 1,435 1,500 1,707 teacher Senior teacher teacher 1,255 1,310 1,370 1,435 1,500 1,565 1,635 1,707 1,780 1,853 1,926 Associate II ll ll I I leadteacher 1,435 1,500 1,565 1,635 1,707 1,780 1,853 1,926 2,000 2,075 I Leadteachei 1,635 I 1,707 l 1,780 l 1,853 1,926 I III 111 2,000 2,075 2,150 2,225 2,325 2,425 1 I Note: Salaries reflects the new scale following an 18 percent raise that became effective February 2001. Source: Federal Civil Service Commission; 1994E.C., as reported in ECORYS-NEI2003. - 211- Table A3.6. Data for Decompositionof Spending Per Student inGovernment Primary ndSecondarySchools, Eihiopia, 2001-02 Component f spendingo teachers E student Overall Share of Average Students' Teachers' spendingper recurrent emunerationof Average weekly weekly student spendingon teacher Pupil- section instructional teaching kades and Region (Birr /year) teachers (YO), (Birr /year) eacher ratio size hours Dad (hours) irades 1-4 Ethiopia 86 92.5 5,987 75.2 77.0 24.6 24.0 Tigray 77 89.8 4,882 70.3 64.9 24.1 26.1 Afar 302 71.9 7,037 32.5 42.5 25.6 19.5 Amhara 86 96.4 6,259 75.6 78.4 24.3 23.4 Oromiya 84 90.4 6,O 17 79.2 76.8 24.9 25.6 Somali 148 84.5 7,037 56.3 - - - Benshangul-Gumuz - - 5,628 46.4 57.3 25.9 - SNNPR - 78 97.7 6,062 79.6 84.4 24.5 23.1 Gambella - 5,326 40.2 60.8 25.6 - Harari 254 73.2 5,079 27.3 56.2 24.1 11.7 Dais Ababa 174 75.2 5,440 41.7 62.0 22.7 15.3 DireDawa 199 80.5 6,486, 40.5 61.4 24.6 16.2 irades 5-8 Ethiopia 160 89.8 7,005 48.8 68.0 24.0 17.2 Tigray 126 82.2 7,192 69.2 67.1 23.8 24.6 Afar 456 65.5 8,383 28.1 41.4 25.5 17.3 Amhara 129 94.8 7,131 58.4 70.0 23.7 19.8 Oromiya 177 89.3 7,251 45.9 66.3 24.3 16.8 Somali 402 87.4 8,383 23.9 0.0 - - Benshangul-Gumuz - - 6,746 78.2 63.4 25.9 - SNNPR 159 93.8 6,855 45.9 70.2 24.2 15.8 Gambella - - 6,382 34.7 64.8 25.5 - Harari 412 82.1 7,682 22.7 53.9 24.0 10.1 Dais Ababa 200 76.7 5,365 35.0 69.1 22.7 11.5 DireDawa 190 91.4 7,522 43.3 66.2 23.9 15.6 3rades 9-12 Ethiopia 283 71.0 10,341 51.5 82.6 24.08 15.0 Tigray 205 66.3 12,102 89.1 80.5 24.01 26.6 Afar 1852 13.4 13,604 54.8 108.0 25.63 13.0 Amhara 239 82.7 8,720 44.1 77.2 24.00 13.7 Oromiya 372 69.2 12,022 46.7 79.9 24.09 14.1 Somali 557 73.5 13,604 33.2 - - - Benshangul-Gumuz - - 10,595 23.6 65.7 25.25 - SNNPR 168 73.1 7,468 60.8 91.5 24.23 16.1 Gambella - - 14,233 39.0 62.7 26.25 - Harari 579 64.0 12,299 33.2 83.0 23.63 9.5 Dais Ababa 251 78.8 10,262 51.8 87.4 23.63 14.0 DireDawa , , . 73.3 10,178 40.5 76.7 23.63 12.5 ,, 343 -Datanotavaille;dataforEt piaexcludedataforBenshangul umuz,Gambella,a Somali. SNNPR. SouthemNations and ionalities PeoulesReuublic. Source: 'halysis of the Ministryof Education's Ethiopi; 2001-02 schoolcensus, supplementedwith estimates of student's instructionalhoursbased on distributionof schoolsby shiftsof operationandhoursofinstructioninthe cumculum, andestimatesof teachingloadsbasedon the data inthe restof the table. See text discussionfor algebraic relationships underlyingthe calculations. - 212- Table A4.1. Percentage of the Population Ever Enrolled in Grade 1at Each Age by Locality, Gender, and Wealth Group, Ethiopia, 2000 B y ' :nder By ICality By weal groupa` `ge Poorest Richest Whole years) Male Female Urban Rural quintile quintile sample 7 21.4 19.6 69.7 14.9 22.6 52.4 20.5 8 33.0 29.8 80.2 25.4 35.3 64.5 31.4 9 40.6 35.7 87.2 31.7 42.8 70.7 38.2 10 47.5 44.8 87.4 39.6 49.8 76.7 46.2 11 55.4 46.2 91.1 44.5 55.4 78.6 50.9 12 57.3 45.8 92.4 45.0 53.9 77.5 51.8 13 61.1 45.0 90.3 46.2 50.7 75.8 53.1 14 58.5 44.5 89.8 44.4 58.7 76.5 51.8 15 58.9 39.7 86.7 42.2 51.3 76.2 49.7 16 54.9 36.6 85.4 37.9 46.4 77.1 46.1 17 56.0 36.9 85.7 36.6 54.5 78.2 46.7 18 50.4 30.5 83.2 30.1 44.3 74.5 39.8 ICohort entry Irate to grade 1" 59.8 44.8 90.0 45.3 54.0 76.1 52.4 meholdsby householdconsumptionexpenditureper adult equivala Becauseof IT sing consumption expenditure data, t 6 data by wealth group is basdd on fewer observafions than the rest'of the table, and the percentages appear to be biased upward, by about 15 to 20 percent. To the extent that the upward bias is greater among the poorest quintile, the rich-poor gap would be larger than impliedby thedatashown here. b. Computed as the average of the percentage ever enrolled among 13-14 year olds, the ages at which the percentage tends to peak. For the percentageamong thepoorest quintile, the averageis basedon the data for ages 11-15 to smoothoutthe somewhat erraticpattem. Source: Analysis of the 2000 Ethiopia Welfare Monitory Survey and the 2000 Ethiopia Household, Income and Consumption Expenditure Survey. - 213- Table A4.2. Student Flow Indicators inPrimary Education acrossRegions, Ethiopia, circa 2000 'ercentage o ~~ Composite cohort survival Composite cohort Repeaters as I cohort ever rate from ~ 'ade 1(%) completic rate (YO) percentage of enrolled in enrollments Region grade 1 To grade 4 To grade 8 Grade 4 Grade 8 (grades 1-8) Tigray 61.4 84.2 69.9 51.7 42.9 10.1 Afar 38.1 20.7 Amhara 46.2 65.5 41.0 30.2 19.0 16.6 Oromiya 51.2 60.6 37.2 31.0 19.1 12.1 Somali 32.1 Benshangul-Gumuz 65.0 47.6 17.5 31.0 11.4 22.7 SNNPR 60.0 45.0 16.5 27.0 9.9 24.3 Gambella 89.8 39.1 16.6 35.2 14.9 20.1 Harari 73.9 55.5 46.8 41.1 34.6 14.0 Addis Ababa 94.5 91.8 87.0 86.8 82.3 10.3 Dire Dawa 79.1 68!3 61.5 54.0 48.7 11.2 SNNPR. SouthernNations an1 qationalitiesPeoulesReuublic. Note: Blanks denote data unreliable; see previous table for notes on eachindicator. Source: Percentageever enrolledin grade 1 basedon analysis of the 1999 Labor ForceSurvey; remainingcolumnsbasedon datasuppliedby the Ministryof Educationon enrollmentsandrepeaters (includingreadmittedpupils) by grade for two schoolyears, 2000-01 and2001-02. - 214- Table A4.3. Probit Estimates of Changesinthe Probability of Being Registered for School among Children k :es 7 to 14 Years Usinn Data from the 1999/2000-WMS/HICES Sample mean I (standard Change inProbabilitythat child registei ifor school' I I I Rural children .Children Children idicator Girls 0.ofobservations 9,333 8,076 17,409 4,557 4,096 8,676 8,733 ependent variable: Childis registeredfor :hool 2,455 6,263 hange inprobability associatedwith: rbanarea (relative to rural)' n.r. n.r. 0.305* 0.292* 0.253* 0.295* 0.309* n.a. n.a. (0.020) (0.031) (0.029) 10.023) (0.025) .egion(relative toAmhara)' Tigray 0.066 0.08 0.008 -0.123* 0.073 -0.023 0.038 0.034 -0.021 (0.027) (0.035) (0.042) [0.035) (0.031) (0.053) (0.028) Afar 0.003 0.008 -0.095 * -0.151* 0.039 -0.033 -0.160* -0.122* -0.127* (0.037) (0.042) (0.058) [0.049) (0.051) (0.050) (0.040) Oromiya 0.394 0.298 -0.044* -0.073* -0.012 0.024 -0.1 16* -0.101* -0.039 (0.019) (0.030) (0.029) (0.024) (0.025) (0.034) (0.022) Somali 0.01 0.034 -0.271 * -0.276* -0.234 * -0.200* -0.335* -0.193 * -0.183 * (0.032) (0.03 1) (0.053) (0.039) (0.036) (0.045) (0.016) Benshangul-Gumuz 0.011 0.007 0.090* -0.02 0.092* 0.179* -0.017 0.137* 0.084* (0.032) (0.053) (0.042) (0.036) (0.042) (0.058) (0.039) SNNPR 0.249 0.113 -0.049* -0.128* 0.013 0.045 -0.149* -0.083* 0.012 (0.022) (0.03 1) (0.030) (0.026) (0.028) (0.034) (0.025) Gambella 0.002 0.004 0.171* 0.047 0.218* 0.213* 0.114 0.251 * 0.432* (0.044) (0.067) (0.040) (0.043) (0.058) (0.062) (0.066) Harari 0.001 0.008 0.120* 0.083 0.103* 0.195* 0.026 0.059 0.023 (0.025) (0.048) (0.037) (0.03 1) (0.038) (0.050) (0.042) Addis Ababa 0.001 0.227 0.09 1 * 0.113* 0.056 0.110* 0.065* 0.257* 0.051 (0.023) (0.045) (0.034) (0.027) (0.032) (0.081) (0.041) DireDawa 0.002 0.02 -0.1301 -0.166* -0.093 -0,032 -0.233' -0.139* -0.092* (0.036) (0.040) (0.049) (0.036) (0.047) (0.054) (0.026) bistance in kmto nearest facility Primaryschool 3.467 0.868 -0.0203 -0.035* -0.018* -0.017* -0.026* -0.057* -0.0 15* (3.459) (1.214) (0.004) (0.006) (0.005) (0.004) (0.005) (0.013) (0.003) Postoffice 20.927 5.128 -0.0023 -0.001 -0.001 * -0.001 *-0.0021 -0.001 0 (17.540) :13.044: (0,001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) All-weatherroad 10.743 0.415 -0.001 0 -0.001 0 -0.001 -0.001 -0.001 * (14.345) (1.693) (0.001) (0.001) (0.001) (0,001) (0.001) (0.001) (0.001) - 215- luster (i.e., enumeration area) averages Percentage o f other households incluster using nearest primaryschool 0.482 0.654 0.339* 0.229* 0.325* 3.339* 1.342* 0.382* 0.298* (0.254) (0.217) (0.029) (0.044) (0.041) 0.035) 0.037) (0.050) [0.038) Percentage literate among other household heads incluster 0.245 0.579 0.210* 0.275* 0.220* 0.190* 3.233* 0.153* 0.032 (0.167) (0.192) (0.037) (0.055) (0.050) 0.046) 0.047) (0.072) (0.055) ousehold resource dummy variable''d nelative to quintile 1, the poorest) Quintile 2 0.253 0.192 0.042* 0.032 0.043 0.066* 0.018 0.061* 0.043* (0.013) (0.025) (0.023) :0.017) 10.019) (0.023) (0.018) Quintile 3 0.206 0.17 0.059* 0.051 0.029 0.079* 0.037 0.091* 0.050* (0,015) (0.027) (0.024) :0.018) 10.022) (0.028) (0.021) Quintile 4 0.168 0.169 0.073* 0.115* 0.036 0.097* 0.053* 0.083* 0.078* (0.016) (0.030) (0.027) :0.020) 10.021) (0.030) (0.024) Quintile 5 0.096 0.223 0.084* 0.186* 0.023 0.110* 0.054* 0.129* 0.055* (0.016) (0.032) (0.027) [0.020) :0.023) (0.034) (0.028) 0. of adult equivalents inhousehold 5.291 5.47 0.008 0.002 0.022* 0.016 -0.003 -0.003 0.017* (1.718) (2.239) (0.006) (0.01 1) (0.011) (0.008) (0.009) (0.011) (0.009) :ouseholdhead variables Male (relative to female)' 0.807 0.659 -0.034 -0.055 -0.027 -0.048 -0.017 -0.054 0.02 (0.020) (0.039) (0.031) (0.026) (0.028) (0.043) (0.032) Age in years 46.29 45.005 0.013* 0.006 0.014* 0.009* 0.016* 0.009 0.004 (12.278) :12.298) (0.002) (0.005) (0.004) (0.003) (0.003) (0.005) (0.003) Age squared -o.ooo* 0 -0.000* -0.000' -0.000' -o.ooo* 0 0 0 0 0 0 0 0 IS literate' 0.23 0.604 0.114* 0.136* 0.052* 0.113* 0.117* 0.077* 0.069* (0.013) (0.022) (0.022) (0.016) (0.018) (0.025) (0.020) Number o f spouses 0.794 0.641 0.013 -0.013 0.071* 0.029 -0.006 0.024 0.001 (0.440) (0.492) (0.019) (0.035) (0.029) (0.024) (0.0251 (0.035) (0.030) Iousehold structure Share of members younger than 7 years 0.201 0.136 -0.017* -0.012 -0.038* -0,017' -0.015 0.006 -0.007 (0/148) (0.175) (0.007) (0.011) (0.011) (0.009; (0.009: (0.012) (0.008) Share o f male members older than 15 years 0.217 0.214 -0.013 0.001 -0.038* -0.026' 0.004 0.003 -0.017 (0.124) (0.140) (0.010) (0.017) (0.015) (0.012; (0.013: (0.017) (0.013) Share o f female members older than 15 year! 0.228 0.285 0.022* 0.032* -0.001 0.004 0.041* 0.029 -0.017 (0.108) (0.147) (0.008) (0.015) (0.014) (0.011; (0.012; - (0.015) (0.012) lhild characteristics Male (relative to female)' 0.514 0.474 0.104* 0.059* 0,136* 11.8. 0.125* 0.096* (0.010) (0.016) (0.017) (0.019) (0.012) Age inyears (relative to 7 years old)' 8 years 0.148 0.122 0.135* 0.144* n.a. 0.135* 0.136;1 0.113* 0.137* (0.015) (0.017) (0.019 (0.022 (0.03 1) (0.032) 9 years 0.134 0.115 0.217* n.a. 0.235* 0.193' 0.247* 0.202* (0.013) (0.016 (0.021 (0.030) (0.034) 10years 0.122 0.136 0.255* 0.2424 0.271' 0.285* 0.299* (0.013) (0.017 (0.020 (0.03 1) (0.036) - 216- 11years 0.096 0.098 0.254* 0.261* 0.245* 0.256* 0.338* (0.013) (0.016) (0.021) (0.032) (0.039) 12years 0.137 0.154 0.279* 0.293* 0.257* 0.283* 0.362* (0.013) (0.015) (0.020) (0.031) (0.036) 13 years 0.104 0.124 0.248* 0.280* 0.201* 0.326* 0.307* (0.015) (0.016) (0.023) (0,033) (0.039) 14 years 0.109 0.138 0.255* 0,009 0.276* 0.229* 0.317* 0.360* (0.014) (0.016) (0.016) (0.022) (0.032) O.O39[lvr' n.r.Notreportedinsource. n.a. Notapplicable. SNNPR SouthemNationsand Nationalities PeoplesRepublic. * Estimate i s significantly different from 0 at the two-tailed 5 percent level. Note: Descriptivestatistics calculated usingpopulationweights. a. For continuous variables, the estimatesrefer to the derivative of the probability of beingregistered for school with respect to the indicated variable, evaluated at the means of all right-hand-side variables. For dichotomous variables, they refer to the changein probability as the variable is changed from 0 to 1, while holdingall other right-hand-side variablesat their means; standarderrors are shown inparentheses. b. Variable has avalue of 1ifthe indicatedcondition is satisfied, and 0 otherwise.Omitted categoryis reference group. c. Householdresourcequintiles are identified withrespectto the country-widedistribution ofrealper adultequivalent consumption expenditure. Source: Analysis of 1999/2000Ethiopia Welfare Monitoring Survey (WMS) andthe HouseholdIncomeand Consumption Expenditure Survey (HICES), as reportedin Schaffner 2003, table A.1, and Schaffner 2004, tables4-7. - 217- Table A4.4. Estimates of Changes inthe Probability of Currently Attending School, Working in a "Job," or Doing Unpaid Housework in Last SevenDays, Using Data from the 1999 Labor Force Survey Sample mean (standard Probabilityestimates" deviation) Rural sample ages 7-14 Urbansample ages 7-14 Did Did unpaid unpaid Currently Had house- Currently Had house- Rural Urban attending "Job" in work in attending "Job" in work in ndicator sample sample school last7 days last7 days school last7 days last 7 day lo. o f observations 49,554 24,539 49,538 49,538 49,538 24,524 24,524 24,524 )ependentvariables Currently attendingschool 0.312 0.83 Had"job" inlast sevendays 0.428 0.114 Didunpaidhousework in last seven days 0.319 0.43 :hang.einmobability associatedwith: legion(relativeto Amhara)' Tigray 0.063 0.065 0.01 -0.106* 0.021 0.023 -0.009 -0.045 (0.026) (0.028) (0.029) (0.020) (0.015) (0.045) Oromiya 0.421 0.338 -0.015 -0.102* 0.115* -0.012 0.003 0.116* (0.018) (0.024) (0.024) (0.014) (0.012) (0.034) Benshangul-Gumuz 0.013 0.007 0.078* -0.285* 0.160* 0.037 0.013 0.107 (0.038) (0.028) (0.049) (0.032) (0.031) (0.08 1) SNNPR 0.241 0.134 -0.01 -0.158* 0.210* -0.064* -0.007 0.224* (0.018) (0.022) (0.023) (0.017) (0.01 1) (0.034) Gambella 0.002 0.006 0.204* -0.382* 0.255* -0.067 0.039 0.111 (0.094) (0.013) (0.107) (0.063) (0.060) (0.132) Harari 0.001 0.01 0.038 -0.232* 0.189* 0.007 -0.018 0.136* (0.052) (0.044) (0.066) (0.028) (0.023) (0.069) Addis Ababa 0.001 0.3 0.037 -0.240* 0.289 0.011 -0.045* 0.01 (0.100) (0.077) (0.187) (0.022) (0.013) (0.066) DireDawa 0.002 0.003 1.ooo* 1.ooo* -0.998* 0.076 0.026 0.128 (0.000) (0.000) (0.011) (0.042) (0.037) (0.169) 'opulationdensity in woreda 1,000 populationper square kilometer 0.202 5.203 0.316* 0.797* -0.214 -0.007* 0.001 -0.004 (0.585) 12.624 (0.109) (0.150) (0.129) (0.002) (0.002) (0.006) 1,000 populationper square kilometer%00 -3 3.523* -108.57* 31.07 0.013* -0.004 0.009 (12.804) (17.995) (17.134) (0.003) (0.003) (0.01 1) Lvailabilityof school services in woreda No. ofschools with grades 1-4 per 1,000 population 0.184 0.159 0.289* -0.074 -0.205 0.055 -0.128 0.044 (0.074) (0.078) (0.116) (0.145) (0.135) (0.108) (0.095) (0.271) No. ofschools with grades 5-8 per 1,000 population 0.11 0.108 -0.006 0.351 0.156 -0.106 0.212* 0.32 (0.053) (0.050) (0.139) (0.179) (0.162) (0.143) (0.103) (0.343) No. ofschools with grades 9-1 2 per 1,000 population 0.005 0.011 0.714 -0.57 1.604 0.332 -2.3 10* -3.397* (0.006) (0.011) (0.857) (1.095) (1.O14) (0.777) (0.674) (1.628) iumber of teachers per school in woreda Teachers in grades 1-4 6.507 9.797 -0.003 -0.006 0.003 -0.003 0 0.004 - 218- (1.912) 4.616) (0.004) (0.005) (0.005) (0.003) (0.002) [0.006) Teachersingrades 5-8 5.083 3.967 0 0.015* -0.012* 0.003 0,001 0.003 (2.516) 6.263) (0.004) (0.005) (0.004) (0.003) (0.002) (0.005) Teachersingrades 9-12 20.742 14.014 0.001 * -0.001 0.001 0.001* 0 0 :20.002) 16.921) (0.000) - (0.001) (0.000) (0.000) (0.000) (0.001) idicatorsof qualityof school inputsin woreda Share of schools with latrines 0.43 0.44 0.061 -0.153* 0.133* -0.01 1 0.017 0.062 (0.156) 0.134) (0.038) (0,062) (0,050) (0.032) (0.030) (0.075) Share of schools with alibrary 0.182 0.209 0.054 -0.013 -0.032 0.166* -0.014 -0.033 (0.124) 0.127) (0.049) (0.069) (0.059) (0.040) (0.037) (0.101) Share of femaleteachersingrades 1-8 0.276 0.33 -0.064 -0.065 0.087 0.229* -0.066 -0.173 (0.088) 0.096) (0.073) (0.102) (0.091) (0.060) (0.056) (0.148) Shareof teachers ingrades 1-8 with less than 'IT1 0.024 0.04 -0.273* 0.339 -0.261 0.134 0.119 0.134 (0.040) 0.048) (0.132) (0.176) (0.170) (0.113) (0.088) (0.266) Share ofteachersingrades 1-8 with vocational training, BA, or M A 0.002 0.017 0.912 -0.195 -0.783 0.051 0.232 -0.471 (0.005) - 0.030) (0.805) (1.340) (1.235) (0.307) (0.264) (0.761) luster (i.e., enumeration area) averages Share of other householdheads incluster who are literate 0.257 0.631 0.449* -0.141* 0.104* 0.123* -0.023 -0.1 1 (0.131) 0.174) (0.041) (0.060) (0.053) (0.042) (0.032) (0.087) Share of other householdheads incluster who have formal sectorjob 0.01 0.216 0.338 -0.501 * 0.221 0.155* -0.076* 0.207* (0.027) 0.148) (0.210) (0.224) (0.191) (0.042) (0.035) (0.090) Share ofother householdheads incluster who are jobless 0.122 0.237 0.098 -0.522 * 0.146* 0.084* -0.182* 0.117 (0.111) - 0.128) (0.050) (0.073) (0.061) (0.040) (0.035) (0,100) [ouseholdheadcharacteristics Male (relative to female)' 0.813 0.679 -0.041 * 0.02 -0.0 13 -0.057* 0.015* -0.009 (0.010) (0.010) (0,010) (0,009) (0.006) (0.013) Age inyears 46.22 $4.815 0.003* -0.002 0 0.005* -0.002' 0.005* (12.431: 12.571 (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) Age' -0.002 0.003* 0.001 -0.004 * 0.002* -0.005' (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) ISliterate' 0.25 0.629 0.086* -0.006 0.022* 0.106* -0.028* 0.042* (0.007) (0.007) (0.007) (0.008) (0.006) (0.011) Reportsreligion other than Orthodox,Catholic, Protestant, or Muslim' 0.04 0.004 -0.094* 0.042 -0.032 -0.1 17 -0.01 1 0.026 (0.014) (0.023) (0.018) (0.068) (0.042) (0.058) Has ever migrated' 0.29 0.849 0.026* -0.020* 0.017 0.022* -0.01 1 0.023 (0.008) (0.010) (0.009) (0.009) (0.007) (0.013) Has aformal sectorjob' 0.01 0.217 0.145* -0.101 * 0.007 0.061* -0.0503 0.057* (0.040) (0.037) (0.037) (0.012) (0.008) (0.019) Hasnojob' 0.092 0.201 -0.041 * -0.047* 0.014 0.012 -0.044' 0.028* (0,009) (0.011) (0,010) (0.009) (0.005) (0.012) louseholdsize and structure No. ofmembers 6.498 6.578 0.008* -0.009* -0.006* 0.011* -0.002 -0.009* (2.123) :2.593] (0.002) (0.002) (0.002) (0.002) (0.0021 (0.002) Share ofchildrenages 0-6 years 0.202 0.138 -0.076* 0.232* 0.105* -0.201 * 0.047* 0.149* (0.152) :O. 142) (0.024) (0.027) (0.025) (0.025) (0.018: (0.038) - 219- Share ofmale adults 15 years andolder 0.212 0.211 -0.116* 0.072* 0.041 -0.025 -0.029 0.101* (0.125) 10.144) (0.028) (0.030) (0.029) (0.030) (0.020) (0.042) Shareof female adults 15 years andolder 0.224 0.283 -0.009 0.199* -0.1 55 * 0.073* -0.077* -0.049 I (0.108) :0.141) (0.030) (0.034) (0.03 1) (0.030) (0.020) (0.040) Share of adultmemberswith formal sectorjob 0.004 0.059 0.357* -0.052 0.202 0.066 0.043 -0.085 (0.028) :0.110) (0.123) (0.156) (0.142) (0.055) (0.040) (0.067) Childcharacteristics Male (relativeto female) ' 0.514 0.481 0.139* 0.225* -0.212* 0.044* 0.047* -0.176* (0.005) (0.007) (0.006) (0.005) (0.005) (0.008) Age inyears (relativeto age 7 years) 8 years 0.151 0.123 0.096* 0.101* 0.059* 0.078* 0.015 0.082* (0.010) (0.010) (0.008) (0.007) (0.013) (0.014) 9 years 0.131 0.116 0.200* 0.170* 0.118* 0.122* 0.06 1 * 0.185* (0.010) (0.010) (0.009) (0.006) (0.013) (0.013) 10years 0.131 0.13 0.267* 0.365* -0.004 0.138* 0.168* 0.194* (0.010) (0.011) (0.011) (0.006) (0.018) (0.014) 11years 0.085 0.099 0,314* 0,389* -0.001 0.148* 0.218* 0.215* (0.011) (0.010) (0.0 12) (0,006) (0.020) (0.015) 12years 0.135 0.14 0.331* 0.427* -0.008 0.146* 0.278* 0.245* (0.011) (0,010) (0.011) (0.006) (0,019) (0.015) 13 years 0.104 0.13 0.341 * 0.454* -0.013 0.128* 0.321* 0.233* (0.011) (0.010) (0.012) (0.006) (0.021) (0.015) 14years 0.102 0.142 0.333* 0.482* -0.045 * 0.121* 0.355* 0.224* (0.012) (0.009) (0.012) (0.007) (0.021) (0.016) Failsto reportmother alive' 0.056 0.07 -0.057 * 0.017 0.002 -0.071 * 0.002 -0.002 (0.011) (0.013) (0.011) (0.016) (0.009) (0.018) Failsto report father alive' 0.12 0.162 -0.049* 0.02 -0.001 -0.053* 0.021* 0.014 (0.009) (0.010) (0.009) (0.010) (0.007) (0.012) Failsto reportbothparentsaliveb' 0.012 0.026 -0.034 -0.013 -0.004 -0.002 -0.0 18 0 (0.025) (0.027) (0.024) (0.022) (0.013) (0.031) a For continuousvariables, the estimates refer to the derivativeof the probability ofbeingregisteredfor school relative to the indicatedvariable, evaluatedat the means of all righthandside variables. For dichotomous variables, they refer to the change in probability as the variable is changed from to ,while holding all other righthandsidevariables at their means; standarderrors are showninparentheses b Variable has a value of 1ifthe indicated condition is satisfied and0 otherwise. Omitted category is reference group. Source: nalysis of the 1999 EthiopiaLabor Force Survey, merged with data from the Ministry of Education's school census for 199940; as reported in Schaffner 2003, table B.l and 8.7. - 220- Table A4.5. Estimatesof Changesinthe Probabilitythat ChildhasEver Attended School 'hangeinprobability associatedwith: .egion(relativeto Amhara) Tigray 0.062 0.131 -0.012 -0.0 18 -0.016 -0.004 (0.073) (0.072) (0.048) (0.040) Afar 0.01 0.007 -0.073 -0.079 0.033* 0.035* (0.053) (0.052) (0.013) (0.011) Oromiya 0.385 0.282 -0.059 -0.064 0.038* 0.039* (0.042) (0.041) (0.015) (0.013) Somali 0.013 0.037 -0.23 1* -0.236* -0.072 -0.087 (0.026) (0.025) (0.061) (0.065) Bengshangul-Gumuz 0.011 0.007 0.074 0.068 0.021 0.021 (0.052) (0.051) (0.022) (0.020) SNNPR 0.232 0.122 0.012 0.001 -0.014 -0.012 (0.051) (0.050) (0.043) (0.039) Gambella 0.002 0.004 0,227* 0.230* 0.027 0.032* (0.075) (0.077) (0.019) (0.014) Harari 0.001 0.007 0.149* 0.123 0.047* 0.047* (0.067) (0.066) (0.012) (0.010) Addis Ababa 0 0.198 0.002 0.004 (0.027) (0.025) DireDawa 0.002 0.019 -0.003 -0.012 -0.068 -0.055 (0.067) (0.067) (0.044) (0.039) louseholdassets No. ofrooms for sleeping 1.299 1.625 0.044* 0.043* 0.035* 0.033* 0.030* 0.016 (0.522) (0.723) (0.013) (0.013) (0.010) (0.012) (0.012) (0.013) Househasdirt floorc' 0.022 0.323 0.370* 0.351* 0.189* 0.027 0.02 0.019 (0.072) (0.073) (0.043) (0.015) (0.015) (0.022) Househas electricityd 0.004 0.763 0.301* 0.304* 0.166* 0.138* 0.122* 0.250* (0.078) (0.078) (0.063) (0.037) (0.035) (0.032) Househas toiletd 0.08 0.711 0.096* 0.087* 0.069* 0.03 0.025 0.052* (0.030) (0.030) (0.021) (0.018) (0.017) (0.022) Householdowns landd 0.959 0.183 -0.031 -0.025 -0.009 -0.015 -0.013 -0.043 (0.035) (0.035) (0.023) (0.020) (0.018) (0.027) Householdowns animals' 0.892 0.402 0.016 0.013 0.052* 0.01 0.006 0.021 (0.019) (0.020) (0.015) (0.015) (0.014) (0.023) Householdhascash cropd 0.333 0.034 0.001 -0.003 0.01 0.049* 0.045* 0.13 (0.019) (0.019) (0.014) (0.011) (0.011) (0.069) louseholdsize and structure - 221- No. ofmembers 7.139 6.85 0.003 0.002 0 -0.004 -0.004 0.001 (2.141) (2.459) (0.004) (0.004) (0,003) (0.003) (0.003) (0.004) No. ofchildren ages 0-6 years 0.222 0.152 -0.03 -0.03 0.017 0.05 0.054 0.032 (0.133) (0.133) (0.060) (0.060) (0.049) (0.067) (0.065) (0.081) No. of male adults 15 years and older 0.218 0.209 -0.095 -0.1 1 -0.121 * 0.058 0.074 -0.015 (0.117) (0.13 1) (0.064) (0.064) (0.05 1) (0.061) (0.059) (0.075) No. of female adults 15 years andolder 0.218 0.284 0.160* 0.161* 0.044 0.06 0.07 0.037 (0.097) (0.124) (0.071) (0.071) (0.060) (0.071) (0.066) (0.082) Languageof household(relativeto all other languages)" Amarigna 0.302 0.526 0.154* 0.139* 0.130* 0.063* 0.058* 0.038 (0.048) (0.047) (0.032) (0.020) (0.020) (0.027) Oromigna 0.357 0.207 0.067 0.06 0.016 -0.002 -0.005 -0.016 (0.048) (0.047) (0.029) (0.015) (0.015) (0.031) Tigrigna 0.062 0.142 0.038 0.019 0.114 1.058* 9.051* 0.065 (0.078) (0.074) (0.074) (0.011) (0.011) (0.042) Parentalcharacteristics Hasno father" 0.081 0.139 -0.004 -0.006 -0.026 -0.043 -0.04 -0.052* (0.021) (0.021) (0.017) (0.023) (0.022) (0.024) Mother'sage inyears 36.653 35.974 0.022 0.023 0.006 0.011 0.011 0.024 (6.638) (6.017) (0.014) (0.015) (0,008) (0.009) (0.008) (0.014) Mother'sage' 0 0 0 0 0 0 0 0 0 0 0 0 Fatherhas completedgrade Id 0.177 0.456 0.173* 0.165* 0.086* 0.040* 0.035* 0.035 (0.022) (0.022) (0.014) (0.013) (0.013) (0.018) Motherhas completedgrade Id 0.075 0.463 0.129* 0.113* 0.045* 0.051 * 0.037* 0.016 (0.031) (0.03 1) (0.021) (0.017) (0.015) (0.020) Motherwas bominrural aread 0.98 0.546 -0.1 81 * -0.167* -0.092* -0.012 -0.0 13 -0.016 (0.052) (0.052) (0.030) (0.013) (0.012) (0.017) Fatherhas other wives" 0.159 0.065 0.01 0.007 0.01 0.02 0.016 0.083 (0.022) (0.022) (0.017) (0.022) (0.024) (0.058) Mother is not father's first wifed 0.076 0.035 -0.02 -0.019 -0.035 0.024 0.026 -0.058 (0.026) (0.026) (0*022) (0.022) (0.020) (0.072) Child characteristics Male (relative to female)d 0.517 0.49 0.085* 0.084* 0.084* 0.012 0.012 0.022 (0.012) (0.012) (0.009) (0.011) (0.011) (0.015) Age inyears (relative to age 7 years)" 8 years 0.153 0.166 0.139* 0.138* 0.077* 0.049* 0.047* 0.166* (0.025) (0.025) (0.015) (0.010) (0.009) (0.028) 9 years 0.144 0.111 0.251* 0.254* 0.153* 0.061* 0.058* 0.225* (0.026) (0.026) (0.016) (0.009) (0.009) (0.030) 10years 0.126 0.141 0.332* 0.332* 0.225* 0.078* 0.074* 0.276* (0.028) (0.028) (0.016) (0.010) (0.010) (0.028) 11 years 0.098 0.118 0.422* 0.422* 0.289* 0.073* 0.069* 0.287* (0.028) (0.028) (0.018) (0.010) (0.010) (0.030) 12years 0.118 0.123 0.466* 0.467* 0.320* 0.080* 0.076* 0.275* (0.027) (0.027) (0.017) (0.011) (0.01 1) (0.029) 13 years 0.109 0.116 0.467* 0.468* 0.334* 0.079* 0.075* 0.285* - 222- (0.028) (0.028) (0.018) (0.011) (0.010) (0.03 I) 14 years 0.091 0.109 0.480* 0.480* 0.347* 0.075* 0.070* 0.272* (0.031) (0.032) (0.019) (0.010) (0.010) (0.032) Oldestchildc' 0.209 0.235 0.008 0.009 0.005 0.002 0.002 0.027 (0.017) (0.017) (0.013) (0.014) (0.013) (0.020) Oldest sond 0.337 0.352 0.016 0.016 0.01 0 0 0.005 (0.011) (0.011) (0.010) (0.010) (0,010) (0.016) Youngest childd 0.105 0.211 0.093* 0.091* 0.074* 0.01 0.008 0.006 (0.022) (0.022) (0.017) (0.013) (0.013) (0.021) Youngest sond 0.338 0.382 0.006 0.006 0.007 0.008 0.008 0.005 (0.012) (0.012) (0.009) (0.010) (0.010) (0.015) Attitude variables Mother listens to radiod 0.188 0.707 0.068* 0.050* 0.049* 0.068* (0.016) (0.012) (0.017) (0.020) Mother believesbeatingisjustified for goingout without husband'spermissiond 0.629 0.407 -0.025 -0.027* -0.003 -0.001 (0.013) (0.010) (0.012) (0.018) Motherreports religion other than Orthodox Catholic,Protestant,or Muslimd 0.034 0 -0.056 -0.022 -0.333 (0.033) (0.032) (0.222) EA Enumerationarea, SNNPR Southem Nations andNationalitiesPeoplesRepublic. * Estimate i s significantlydifferent from 0 at the two-tailed 5 percentlevel. a. Sample of childrenwhose mothersrespondedto the questionnairefor women. b. For continuousvariables, the estimates refer to the derivative of the probability that a child has ever attendedschoolwith respectto the indicatedvariable, evaluatedat the means of all right-hand-side variables; for dichotomous variables,they refer to the change inprobability when the variable is changed from 0 to 1, while holding all other right-hand-sidevariables at their means. Standarderrors are shown inparentheses. c. Variable has a value of 1ifthe indicated conditionis satisfied, and0 otherwise. d. The methodinvolves including one dummy variable for each EA (except for a referenceEA). The dummy variables absorb all variation inattendancerates across EAs, allowing the estimates ofhouseholdand child characteristicsto be derivedpurelyon the basis of within-EA variation. Source: Analysis of the 2000 EthiopiaDemographicand HealthSurvey, as reported in Schaffner 2003, tables C.1 and C.3. - 223- Table A5.1. PercentageDistribution Government and NongovernmentPrimary and Secondary Schoolsin Urban and Rural Areas by Instructional Program, Ethiopia 2001-02 Rural areas Non- Non- Government government Government government schools schools Both types schools schools Both types ~ Grades 1-4 only 45.2 ' 44.1 45.2 16.3 16.7 16.4 Grades 5-8 only 0.2 0.9 0.2 5.0 1.5 4.2 Grades 1-4 and 5-8 54.6 55.0 54.6 78.7 81.7 79.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 (93751) (211) (9,962) (1,403) (389) (1,792) Secondary education Grades 9-1 0 only 57.7 0.0 55.6 23.7 34.0 25.0 IGrades 11-12 only 0.0 0.0 0.0 1.4 1.9 1.5 Grades9-lOand 11- I 42.3 I 12 44.4 74.9 64.2 73.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 (26) (1) (27) (355) (53) (408) Note: Figures inparenthesesrefer to the number of schools inthe category shown. Source: Analysis of the 200142 EthiopiaSchoolCensus ofthe Ministryof Education. Table A5.2. Distribution of Section Sizes by Level of Instruction, Ethiopia, 2001-02 Level of instruction & range of section size Whole c cle ~~ Primary 4'0 14.8 33.1 50 to 69 27.0 32.9 70 to 89 31.4 23.9 90 to 109 17.4 7.8 > 110 9.3 2.2 All 100.0 100.0 100.0 Secondary <3 0 n.a. n.a. 0.6 30 to 49 n.a. n.a. 50 to 69 n.a. n.a. 70 to 80 n.a. n.a. ii; ~ >90 n.a. n.a. All n.a. n.a. 100.0 Memorandaonnumber of observations: Primary schools 11,001 6,383 11,001 Secondary schools n.a. n.a. 330 n.a.Notapplicable. Source: Analysis of Ministryof Education's 2001-02 school census. - 224- Table A5.3. Average Section Sizes of Primary and Secondary Classes in Government and Nc eovernment Schoolsbv Localitv and Re m, Ethiopia, 2001-02 Government sch 11s Non- wernment s( 001s :ountry/locality/ I egion Grades14 Grades58 I Secondary Zrades 1-4 Grades 5-8 Secondary tthiopia 75.9 I 60.7 80.4 61.9 54.8 51.1 bylocality Urban 74.4 71.1 81.4 55.1 55.1 51.0 Rural 76.1 58.5 66.0 74.4 53.8 59.7 byregion Tigray 66.8 59.3 80.4 53.4 42.5 68.7 Afar 37.6 29.6 67.1 33.8 - - Amhara 78.0 61.0 76.0 55.8 54.9 50.9 Oromiya 75.6 60.4 77.5 68.1 49.4 56.9 Somali - - - - - - Benshangul-Gumuz 55.4 51.0 61.0 47.3 - - SNNPR 84.4 63.5 87.9 74.8 63.1 30.7 Gambella 54.7 49.2 54.6 54.7 - - Harari 52.2 46.3 78.0 32.7 28.4 24.3 Addis Ababa 61.5 66.8 101.5 54.9 56.4 48.9 Dire Dawa 58.2 50.4 77.1 52.6 51.7 60.0 -No dataavailable. SNNPR SouthemNations andNationalitiesPeoplesRepublic Source: Analysis of Ministry of Education's 2001-02 Ethiopia schoolcensus. - 225- Table A5.4. Regression Estimates of the Relation between Numbers of Teachers and Pupils - across Primary and Secondary Schools, :hiopia, 2001-02 Government schools Other schools 1 ,imaryschoc i Secondary schools P mary scho Secondary (grades schools 2egressor Graded-8 Grades14 Grades 5-8 9-1 2) 3radesl-8 2radesl-4 hades 5-8 grades9-12) Pupils 0.016 0.012 0.018 0.009 0.015 0.011 0.016 0.007 (164.73)** (91.58)** (132.49)** (26.86)** (31.96)** (18.82)* * (30.84) ** (2.37)* rigray -27.396 -14.52 -16.202 -4.694 -6.191 -2.759 -3.116 0.909 (36.96)** (28.43)** (30.55)** -1.86 (5.52)** (3.59)* * (3.67)** -0.42 Afar -22.684 -12.45 -11.146 -3.267 -7.519 -3.496 0 0 (25.11)** (19.89)** (12.68)** -0.71 (2.91)** (1.99)* (*I (.I 4mhara -27.444 -14.913 -15.197 0.923 -6.699 -3.267 -3.119 -5.052 (38.07)* * (30.08)** (30.08)** -0.4 (5.94)** (4.27)** (3.70)** -1.54 3romiya -26.905 -15.24 -13.912 0.176 -6.364 -3.555 -1.934 -2.007 (37.46)** (30.86)** (27.76)** -0.08 (10.62)** (8.90)** (4.74)** -1.19 Benshangul-Gumuz -25.053 -12.983 -15.455 -0.551 -8.679 -4.54 0 0 (30.88)** (23.21)** (19.89)** -0.15 (3.36)** (2.59)* * (.I (.I SNNPR -2 6.999 -15.186 -14.14 -1.749 -8.01 -4.344 -3.296 1.481 (37.45)** (30.62)** (27.98)** -0.75 (12.22)** (9.83)* * (7.47)** -0.55 Gambella -22.174 -12.008 -1 1.028 -1.683 -8.852 -4.72 0 0 (23.85)** (18.67)** (12.97)** -0.3 -1.56 -1.22 (.I (J Harari -10.997 -4.596 -2.817 34.843 4.188 1.587 0.981 3.202 (9.16)** (5.36)** (2.81)** (6.32)** -1.91 -1.06 -0.67 -0.7 Dire Dawa -20.997 -10.554 -10.477 44.607 -4.577 -1.486 -3.206 -3.094 (18.41)** (13.33)** (9.49)** (4.91)** (2.92)** -1.39 (3.05)** -0.95 Constant 26.244 15.78 15.417 9.428 9.356 5.888 5.2 8.076 (35.88)* * (31.22)** (30.36) ** (3.96)** (17.24)** (16.30)** (16.05)** (6.58)** Observations 11,122 11,022 5,939 372 600 592 437 53 R2 0.76 - 0.5 0.79 0.75 0.74 0.48 0.76 0.2 ***Regressioncoefficientthat is statisticallys' 1, Regression coefficientthat is statistically pificant at the 1percent confidence level. SNNPR Southern Nations and Nationalities 1 iples Republic. Note: Values of t-statistics inparentheses; omitted region is Addis Ababa. Source: Regression analysis of the Ethiopia 1994E.C. school census. - 226- TableA5.5. RegressionEstimatesofthe RelationbetweenNumbersofTeachers and PupilsinGovernment PrimarySchoolsbyRegion,Ethiopia,2001-02 Regressionresultsand relatedinformation Coefficienton Simulationsfor a school number f pupils R2v ues with 700 pupils,grades 1-8 No. of Grades Number of Pupil- Region Grades Grades Grades 3bservations 1-8 1-4 1-8 1-4 teachers teacher ratio Ethiopia 11,122 0.017 0.012 0.715 0.44 10.9 64.2 Tigray 887 0.014 0.014 0.92 0.74 10.0 69.9 Afar 128 0.022 0.018 0.83 0.69 17.6 39.7 Amhara 2,927 0.015 0.012 0.78 0.55 9.8 71.8 Oromiya 4,4 17 0.017 0.011 0.72 0.44 10.7 65.1 Somali - - - - - - - Benshangul-Gumuz 266 0.013 0.011 0.78 0.47 11.3 62.0 SNNPR 2,223 0.015 0.010 0.59 0.27 10.4 67.4 Gambella 108 0.015 0.008 0.40 0.18 14.9 47.1 Harari 40 0.038 0.030 0.83 0.54 28.4 24.6 Addis Ababa 79 0.020 0.016 0.88 0.65 31.1 22.5 DireDawa 47 0.021 0.019 0.96 0.84 16.4 42.6 -Datanotavailable. SNNPR, Southem Nations and Nationalities Peoples Republic. a. Refersto number of government primary schools offering instruction ingrades 1-8; those offering only grades 1-4 are smaller by a few schools. Source: Analysis o f the Ministry of Education's 2001-02 Ethiopia school census; see table A5.2 for complete regression results. Table A5.6. PrimarySchoolConstructionCosts, Ethiopia, 2002 4-classroombuilding Cost per c issrooma/ Location (Birr) Birr US% Near Addis Ababa 242,528 60,632 7,133 Southregion Near urbancenter 266,781 66,695 7,847 Semiremotesite 434,421 108,605 12,777 Very remotesite 485,056 121,264 14,266 Northregion Accessible site 281,415 70,354 8,277 Verv remotesite 637,874 159,469 18,761 a. Assumes conversion rate of 8.5 Birr per US$. Source: Coyle 2003. - 227- Table A5.7. Number of Candidates Who Took the National Grade 8,10, and 12 Examinations by Region ,Ethiopia, circa 2000 Grade 8 Grade 10 Grade 12 Countrylregion 1999-00'' 2001-02 2001-02 3hiopia 267,642 296,668 182,903 3y Region Tigray 39,768 32,441 11,800 Afar 1,188 719 1,111 Amhara 42,399 26,447 29,660 Oromiya 77,472 44,034 55,374 Somali 1,921 1,205 788 Benshangul-Gumuz 1,950 1,407 1341 SNNPR 47,721 15,949 35,361 Gambela 2,555 818 1,002 Harari 2,07 1 1,121 2,260 Addis Ababa 47,295 27,125 41,449 Dire Dawa 3.302 1,402 2,557 N N P R Southem Nations and Nationalities Peoples Republic. a. Beginning in 2000-01, regional examinations-replacedthe national grade 8 examination. Source: Personalcommunication from the National Organization of Examinations Table A5.8. Results from the 2000 Ethiopia National Baseline Assessment on Grade 4 a IGrade8 Student Achievement, 1 hiopia, 2000 Grade 4 Grade 8 Indicator Overall Urban Rural Overall Urban Rural No. of sample schools 256 n.r. n.r. 136 n.r. n.r. No. o f students tested 10,506 n.r. n.r. 5,099 n.r. n.r. No. o fteachers surveyed 758 n.r. n.r. 469 n.r. n.r. No. o f school directors surveved 250 n.r. n.r. 136 n.r. n.r. Test scores (percentage o f correc answers) Reading 64.3 60.5 64.5 n.a. n.a. n.a. English 40.5 39.4 40.6 38.8 39.5 37.0 Mathematics 39.3 37.9 38.0 38.2 38.8 36.9 Environmental science 48.1 45.7 50.4 n.a. n.a. n.a. Chemistry n.a. n.a. n.a. 40.3 47.1 47.3 Biology n.a. n.a. n.a. 47.2 40.4 39.3 Overall average a/ 48.1 45.9 48.4 41.1 41.5 40.1 n.r.Notreuorted. n.a. Not applicable becausetest was not administered for the subject, a. Simple average across all subjects. Source: Hassenet al. 2001aand 2001b - 228- Indicator Ages 5-9 Ages 10-14 Ages 5-14 Population(thousands) 8,897.4 7,163.3 16,060.7 Percentagenot working 54.8 18.4 38.6 Percentage working 45.2 81.6 61.4 Indomesticactivities 27.1 31.3 28.9 Inproductive activities 18.1 50.3 32.5 Percentageworking inproductive activities Boys 24.7 61.6 41.4 Girls 11.3 38.1 23.1 Urbanareas 2.4 15.7 9.4 Ruralareas 19.9 56.0 35.6 Source: Authors' estimatesbased on the 1999EthiopiaLabor Force Survey (ELFS). ble A6.2 M i n irianEarning functions,Et opia,2001 Whole population Men Women Urban Rural Age 0.147 0.144 0.137 0.135 0.188 (3 1.59)** (24.45)** (16.38)** (26.45)** (15.69)** Age2 -0.002 -0.002 -0.001 -0.001 -0.002 (24,25)** (19.99)** (11.36)** (20.01)** (12.35)** Grade 1-4 0.562 0.696 0.338 0.478 0.726 (16.78)** (15.18)* * (6.97)** (12.93)** (8.30)* * Grade 5-8 0.989 1.049 0.803 0.935 1.033 (33.76)* * (26.59)** (18.03)** (29.28)** (10.64)** Grade 9-12 1.434 1.417 1.414 1.395 1.369 (51.70)** (37.36)** (34.43)** (45.80)** (13.28)** Certificate 2.041 2.003 2.072 1.974 2.198 (5 2.50)* * (39.60)** (33.91)** (46.45)** (21.33)** Diploma 2.148 2.119 2.121 2.105 2.315 (5 3.98)* * (42.71)** (30.13)** (5 1.17)** (8.51)** Degree 2.652 2.579 2.767 2.610 2.912 (45.87)** (39.48)** (18.68)** (45.19)** (5.69)** Constant -3.823 -3.618 -3.740 -3.574 -4.478 (54.46)** (38.99)** (32.13)** (45.35)* * (26.54)** NumberofObservations 10960 6638 4322 9271 1689 R-sauared 0.54 0.50 0.55 0.53 0.49 *-Regression coefficient that is statistically significant at 5 percent confidence level. ** Regressioncoefficient that i s statistically significantat the 1percent confidence level. a! Refersto wage earnersonly aged between 10and 65. The dependantvariable is the hourly salary (including benefits) from the mainjob. The results reflect estimates based on a semi-logarithmic model specification; education (highest level attained) i s expressed by a set of dummies variables (illiterate is the reference group) Note: Absolute values of t-statistics inparentheses; omitted education category is no education. Source:authors' estimates basedon the 2001 Ethiopia Stand-Alone Child Labor Survey (ESACLFS). - 229- Appendix Figures Figure A2.1. Probabilityof BeingLiterateamongAdults with Six Years of PrimarySchooling, Ethiopia and Other Africa Countries, circa 2000 40 50 60 70 80 90 100% t t t t t ChadNiger tI Senegal Togo BurundiEt opia t tP tI Sierra Leone Guinea-Bissau Rwanda Note: Adults refer to thoseages 15-49 in Ethiopia, andages 2244 in all theother countries. Source: For Ethiopia: computed from 2000 Ethiopia Demographic and Health Survey; for the other counties, computed from each country's UNICEF-sponsoredMultiple IndicatorCluster Survey (MICS) surveyas reportedinMingat2003. FigureA3.1. Trends inAverage Teacher Salary inPrimaryEducation, SelectedWorld Regions, 1975-2000 1975 1985 1992 2OOo - 230- FigureA4.1. Cumulative Distributionof a HypotheticalCohort byEducationAttainment and of the Public SpendingBenefitingI t 100 I 80 I P I s n * W fi 8 I / 60 * I I I 4 / I I I I I 20 II I / 0 I I 0 20 40 60 80 100 Populationshare (%) Source: Authors' construction. - 231- Figure A5.1. Percentage of Government and NongovernmentSchools Serving Two or More Shifts of Students egion, Ethiopia, 2001-02 Ethiopia YGovemment 1 schools Urban Rural \Other schools Tigray Afar Amhara Oromiya Ben- shangul SNNPR Gambella Haran Addis Ababa DireDawa 1 I 160 140 120 100 80 60 40 20 0 0 20 40 60 60 100 120 140 160 Secondary schools (percent) Primary schools (percent) Benshangul, Benshangul-Gumuz; SNNPR Southem Nations and Nationalities Peoples Republic. Note: Striped part of each bar refers to the percentageamong government schools; the solid part, to the percentageamong nongovernment schools; data for Somali not shown for lack of data. Source: Analysis of the Ministry of Education's 2001-02 Ethiopia school census data. Figure A5.2. Relation between Enrollments and Number of Teachers inGrades 1-5 and Grades 5-8 across Government Schools, Ethiopia, 2001-02 (a) Grades 1-4 (b) Grades 5-8 1 0 70 70 c: c 2 50 fiP 36 0 0 0 c 50 c- P b 0 b 30 5 30 2 2 10 I O 1 1 10 500 1000 1500 2000 3000 Number of pupils Number of pupils - 232- References ADEA. 2003. MeetingDocuments ofthe 2003 Biennale of the Association for the Development of Education inAfrica (ADEA). Mauritius, December 3-5. Ainsworth, Martha, and Deon Filmer. 2002. "Poverty, AIDS and Children's Schooling: A Targeting Dilemma." Policy Research Working Paper 2885. World Bank, Policy Research Department, Washington, D.C. Asfaw, A,, andA. Admassie. 1996. The Impact of Education on Allocative andTechnical Efficiencyo f Farmers: The Case of Ethiopia. Ethiopian Journal of Economics 5 (1): 1-27. Avenstrup, Roger. 2004. "Free Primary Education and Poverty Reduction: The Case of Kenya, Lesotho, Malawi and Uganda," World Bank, Africa Region Human Development Department,Washington, D.C. Processed. BREDA (UNESCO Regional Office for Education in Africa), World Bank, and UNESCO Institute of Statistics (UIS). 2002. Universal Primary Education Goal For All. Statistical Document. MINEDAFVIII. Document prepared for the Eighth Conference o f Ministers o f Education o fAfrican Member States (MINEDAF), Dar-es-Salaam, Tanzania, December 2-6. Bruns, B., A. Mingat, and R. Rakotomalala. 2003. Achieving Universal Primary Education by 2015: A Chance for Every Child. Washington, D.C.: World Bank. Case, A. 2001. "The Primacy o f Education," Research Program in Development Studies, Princeton University, Working Paper 203, Princeton, N.J. Case, A., C. Paxson, and J. Ableidinger. 2002. "Orphans in Africa." Center for Health and Wellbeing, ResearchProgram inDevelopment Studies, Princeton University, Princeton, N.J. Processed. Chiswick, B. 1988. "Interpreting the Coefficient o f Schooling inthe HumanCapital Earnings Function" Journal o f Educational Planning andAdministration, 12(2), April 1998, pp. 123- 130. Christiaensen, L., andH.Alderman. 2003. Child Malnutrition inEthiopia: Can Maternal Knowledge Augment the Role o f Income? Washington, D.C.: World Bank. Coyle, T. 2003. "Review o f Opportunities for Lower Cost Primary School Facilities." Report prepared for the Ethiopian Education Development Program, Ministry of Education (final report datedDecember 30). - 233- Croppenstedt, A., M.Demeke, andM.Meshi. 1999.An EmpiricalAnalysis ofDemand for Fertilizer inEthiopia. Ethiopian Journal of Agricultural Economics 3 (1): 1-39. Demeke, M.2003. "Education andthe Labour Market," background paper for this report, Addis Ababa, Ethiopia. Addis Ababa. Processed. Demery, Lionel. 2000. "Benefit Incidence: A Practitioner's Guide." World BankAfrica Region, Poverty and Social Development Group, Washington D.C. Processed. Dercon, S. 1997. "Poverty and Deprivation in Ethiopia," Centre for the Study o f African Economies, Department of Economics, Oxford University. Processed. Easterly, W. 2002. "Growth in Ethiopia: Retrospect and Prospect." Center for Global Development, Institute for International Economics, Washington D.C. Processed. ECORYS-NEI. 2003, EducationUnit Cost Study-Ethiopia, Rotterdam, September4. EQUIP. Undated."Education: Implementation of Report Cards," EQUIP2 Policy Brief. Washington D.C.:Educational Quality Improvement Program. Federal Civil Service Commission. 1994 E.C. Personnel Statistics, 1994 Ethiopian Fiscal Year. Addis Ababa: Federal Civil Service Commission. Franz, Jutta. 2003. "Technical and Vocational Education and Training (TVET) Expenditure and Costs." Background Paper for the 2002-03 World Bank Public Expenditure Review: Public Spending inthe Social Sectors inEthiopia. Washington, D.C. Processed. Govt. o f Ethiopia. 1994. Basic Education Statistics. Education Management Information System (EMIS). Addis Ababa: Ministry ofEducation, July. Govt. of Ethiopia. 1994 E.C. Distance Education Program. Consolidated Statistics 1971-1994 E.C. Educational MediaAgency. Addis Ababa: MinistryofEducation. Govt. o f Ethiopia. 1995. Education Statistics Annual Abstract, 1993-94. Addis Ababa: Ministry o f Education. Govt. of Ethiopia. 1996. Education Statistics Annual Abstract, 1994-95. Addis Ababa: Ministry of Education. Govt. o f Ethiopia. 1997. Education Statistics Annual Abstract, 1995-96. Addis Ababa: Ministry o f Education. Govt. of Ethiopia. 1998. Education Statistics Annual Abstract, 1996-97. Addis Ababa: Ministry of Education. - 234- Govt. o f Ethiopia. 1999a. Education Statistics Annual Abstract, 1998-99. Addis Ababa: Ministry o f Education, August. Govt. o f Ethiopia. 1999b. Education Statistics Annual Abstract, 1997-98. Addis Ababa: Ministry o f Education, January. Govt. o f Ethiopia. 2000a, The Federal Grant Formula. Addis Ababa: Ministry o f Finance and Economic Development, March. Govt. of Ethiopia. 2000b. Alternative Routes to Basic Primary Education. Addis Ababa: MinistryofEducation, September. Govt. o f Ethiopia. 2000c. Education Statistics Annual Abstract, 1999-2000. Addis Ababa: MinistryofEducation. Govt. o f Ethiopia. 2000d. Guidelines for Organization of Educational Management, Community Participation andEducational Finance. Addis Ababa: Ministry of Education, August. Govt. o f Ethiopia. 2001. Education Statistics Annual Abstract 2000-01. Addis Ababa: Ministry o f Education. Govt. o f Ethiopia. 2002a. Poverty Reduction Strategy Paper. Addis Ababa: Welfare Monitoring Unit,MinistryofFinanceandEconomicDevelopment, August. Govt. o f Ethiopia. 2002b. Poverty Profile inEthiopia. Analysis Based on the 1999/00 HICE and WM Survey Results. Addis Ababa: Welfare Monitoring Unit, Ministry of Finance and Economic Development, March. Govt. of Ethiopia. 2002c. Education Statistics Annual Abstract, 2001-02. Addis Ababa: Ministry o fEducation. Govt. of Ethiopia and ORC Macro. 2001, Ethiopia Demographic and Health Survey 2000. Addis Ababa: Central Statistical Authority, May. Hassen, Bedm Kedir, Daniel Zewdie and Yohannes Afework with William Cummings, Jennifer Spratt and Tizazu Assare. 2001a. Ethiopia National Baseline Assessment on Grade Eight Student's Achievement. Executive Summary of Study Results. Addis Ababa: National Organization of Examinations, Ministryo f Education. Hassen, BedruKedir, Daniel Zewdie and Yohannes Afework with William Cummings, Jennifer Spratt and Tizazu Assare. 2001bEthiopia National Baseline Assessment on Grade Four Student's Achievement. Executive Summary of Study Results. Addis Ababa: National Organization of Examinations, Ministryo f Education. - 235- Hassen, Bedru Kedir, Daniel Zewdie Eshetie and Yohannes Afewrok. 2001b. Ethiopia National Baseline Assessment on Grade Eight Student's Achievement. Addis Ababa: National Organization o fExaminations, Ministryof Education. Hinchliffe, Keith. 2003. "Notes on the Impact of the HIPC Initiative on Education and Health Public Expenditure in African Countries," World Bank, Africa Human Development Department, Washington D.C. Processed. IIZ/DVV-ETH. Various issues. Focus on Adult and Non-Formal Education in Ethiopia. Newsletter. Addis Ababa: Ministry o f Education and IZZ/DVV-Etho (a branch of the Institute for International Cooperation of the GermanAdult Education Association). Jamison, D.T., and L.J. Lau. 1982.Farmer l3ducation and Farm Efficiency. Baltimore: Johns HopkinsUniversityPress. Kedir, Bedru; Daniel Zewdie and Yohannes Afework 2000. Ethiopia National Baseline Assessment on Grade Four Pupils' Achievement. Addis Ababa: National Organization of Examinations, Ministryo fEducation. Kengne, Valkse Mapto, and Alain Mingat. 2002. (Analyse comparative intemationale de la fkminisation du corps enseignant et de l'impact du sexe de l'enseignant sur la performance des systkmes kducatifs primaires en Afrique.)) Paper presented at the May 23-25 Colloque Association Francophone d'Education Comparke (AFEC), Caen, France. Krishnan, P. Undated. Returnsto Schooling inUrbanEthiopia. Report No. 9a. Oxford: Center for the Studyo fAfrican Economies, OxfordUniversity, England. Krishnan, P., T. Selassie, and S. Dercon. 1998. The UrbanLabour Market during Structural Adjustment: Ethiopia 1990-1997, Report No. WPS98-9. Oxford: Center for the Studyof African Economies, Oxford University, England. Lockheed, Marlaine, Adriaan Verspoor and Associates. 1991. Improving Primary Education in Developing Countries. New York: OxfordUniversityPress. McArthur, Don. 2003. New School Design, Consultant report prepared for the Ethiopia Ministry o f Education. Addis Ababa. Mengistae, T. 1998. Wage Rates andJob Queues: Does the Public Sector Overpay inEthiopia? Report no. WPS/98-20. Oxford: Centre for the Study o f African Economies,Oxford University. Mingat, Alain. 2002. <