Report No. 29946 CentralAmerica Education Strategy Paper November 30, 2005 Human Development Department Latin America and the Caribbean Region Document of theWorld Bank TIMSS ThirdInternationalMathematics and Science Study TSL Tutorial LearningSystem UMCE Unidad de Medici6n de la Calidadde la Educaci6n UNESCO UnitedNations Educational, Scientific and Cultural Organization UPNFM Universidad Pedag6gica Nacional Francisco MorazAn WDR World Development Report ... 111 TABLE OF CONTENTS PageNo . CHAPTER I:A DIAGNOSISOFEDUCATIONALPERFORMANCE INCENTRALAMERICA ........................................................................... 1 ICOVERAGEOFTHEEDUCATION SYSTEMS...................................................................... . 1 A.Average years of education attained....................................................................... 2 3 11.INTERNAL EFFICIENCYOFEDUCATION SYSTEMS.......................................................... B. Gross and Net Enrollment Rates............................................................................. 7 A. Completion Rate...................................................................................................... 7 B. Survival Rate......................................................................................................... 10 15 111QUALITY OFEDUCATION SYSTEMS............................................................................ .C.Age-by-grade Distortion and Repetition Rates..................................................... 20 A.Standardized testing.............................................................................................. 20 w.BRATESOF .Learning outcomes................................................................................................ 22 ............................................................................25 V.EQUITY OFEDUCATIONAL OUTCOMES RETURNTO EDUCATION ........................................................................ 31 A. UrbadRural divide ............................................................................................... 31 B.Socio-economic differences in educational pe$ormance ..................................... 35 C. Educational Outcomes and Gender...................................................................... VI.DCONCLUSIONS............................................................................................................ .Educational outcomesand indigenous and afro-descendentpopulation.............41 42 4~ CHAPTER 11:DETERMINANTS OF (AND CONSTRAINTS TO) LEARNINGINCENTRALAMERICA .................................................. 46 I.AREVIEWOFCENTRALAMERICANNATIONAL EDUCATIONAL ASSESSMENT .................................................................................................................................. 47 SYSTEMS 11.EXISTINGEVIDENCEONDETERMINANTS OFLEARNINGCENTRAL AMERICA IN ..........49 A.StudentBackground Constraints to Learning in Central America ....................... 50 B.Teacher Quality..................................................................................................... 53 C.Pedagogical Practices .......................................................................................... 67 I11.D POLICY IMPLICATIONS . Education spending............................................................................................... 72 and school management....................................................... 73 A Learning Determinants of Effective Schools......................................................... . B. Background Characteristics Policy Implications .............................................. 73 79 C Teacher Quality Policy Implications.................................................................... 90 IV.DCONCLUSIONSANDPOLICYIMPLICATIONS................................................................ ..Pedagogical Practice Policy Implications ........................................................... 91 93 CHAPTER 111:EXPANDINGSECONDARY EDUCATIONIN CENTRAL AMERICA .............................................................................. 97 I.WHYISEXPANDING COVERAGE OF SECONDARY INA BROAD-BASED MANNER A REGIONAL PRIORITY?................................................................................................ 97 11.CONSTRAINTS TOEQUITABLE SECONDARY COVERAGE .............................................. 99 i v 111.ALTERNATIVES INCREASINGSECONDARYCOVERAGE...................................... FOR 106 A. Public and Private Financing ............................................................................. 106 B.Demand-side Interventions ................................................................................. 111 C .Alternative 113 D. Quality and Relevance........................................................................................ Delivery Models ............................................................................... 116 E. What Policy Combination Should beAdvanced in each Country ?.................... 117 Iv.HOW MUCHWOULD IT COST TO INCREASE SECONDARY COVERAGE? SIMULATIONS ........................................................................ 117 A Targets and Main Assumptions of Simulations................................................... FOR ELSALVADOR AND HONDURAS . 118 V.CONCLUSIONS........................................................................................................... 126 B. Scenarios and Cost Projections .......................................................................... 119 CHAPTERIV: A REVIEW OF EDUCATIONEXPENDITUREIN CENTRALAMERICA ............................................................................ 129 1PUBLICEDUCATION SPENDING: LEVELS EVOLUTION . AND ......................................... 129 11.DISAGGREGRATION OFEDUCATIONAL EXPENDITURES EDUCATION LEVEL PER- BY AND ........................................................................... 111.EQUITY STUDENT AND CHILDUNIT COSTS CHALLENGES INTERMSOFPER-CAPITA EDUCATION A. Thedistribution of public enrollment and school-age population across income EXPENDITURE..........131 136 quintiles.................................................................................................................... . . 136 Iv.BPRIVATEHOUSEHOLD .Public subsidyper school-agepopulation by quintile........................................ 140 EDUCATION EXPENDITURES DEMONSTRATE ALSO INEQUITABLE EDUCATION 143 V.EFFICIENCYOF PUBLIC EXPENDITURE FINANCING.................................................................... ...................................................................... 145 A Cost-Effectiveness ............................................................................................... . 146 B.Expenditure distribution by area......................................................................... 148 VI.CTHEROLEOF EXTERNAL .Casestudy: Low cost-efectiveness and high ineflciency in Honduras .............151 . FUNDS .............................................................................. 153 154 CONCLUSIONS ................................................................................................................ A Magnitude and composition of externalfunds .................................................... 157 CHAPTER V: A COMPARATIVEANALYSIS OF SCHOOL-BASED MANAGEMENTINCENTRALAMERICA ........................................ 160 ........................................................................................ 161 11.COMPARINGEDUCATION I.CONCEPTUALFRAMEWORK A. History and Overview of Decentralization Reforms in Central America ...........163 DECENTRALIZATION REFORMS INCENTRALAMERICA .....163 I11.BIMPACT REFORMS .Comparing theDifferent Reform Designs........................................................... 168 OF DECENTRALIZATION .............................................................. 177 A. Comparing School Environments with and without Decentralized Educational Provision.................................................................................................................. B Comparing Enrollments and Student Outcomes Withand Without Decentralized . 178 .Educational Provision ............................................................................................. 187 EXPLAINING THE IMPACT ......................................................................................... 195 A. Explaining impacts andprogress in terms of assets available and linking teacher 196 B. Explaining the impact in terms ofthe reform design .......................................... effort to learning outcomes...................................................................................... 201 V 205 v.CONCLUSIONSAND C.Explaining the impact in terms of contexts of implementation ........................... POLICY RECOMMENDATIONS ..................................................... 208 REFERENCES .................................................................................................................. 213 INDEX OFTABLES PageNo . TABLE1.1: COMPLETIONRATESAT THEPRIMARYAND SECONDARY LEVEL .................... 8 TABLE1.2: PRIMARY COMPLETIONRATEFORDIFFERENT COHORTS ............................... 9 TABLE1.3: STANDARDIZED 22 TABLE1.4: PRIVATERATES OF RETURN.......................................................................... TESTINGINCENTRALAMERICA.......................................... 27 TABLE2.1:SUMMARYFINDINGSONFAMILYAND COMMUNITYRELATED DETERMINANTSEDUCATION OF AMERICANLABORATORY ............................................................................. QUALITY INCENTRALAMERICAAND THE LATIN STUDY 51 TABLE2.2: SUMMARYFINDINGS TEACHER ON RELATEDDETERMINANTS OF EDUCATION LABORATORY ................................................................................................ QUALITY INCENTRALAMERICAAND THELATINAMERICAN STUDY 55 TABLE2.3: HIGHEST AMERICANCOUNTRIES .............................................................................................. LEVEL OFEDUCATIONATTAINED BY TEACHERSINCENTRAL 56 TABLE2.4: MEASURES OF TEACHER EFFORT .................................................................. 61 TABLE2.5: HOURLY SALARIESACROSS TEACHERS AND PROFESSIONSWITH SIMILAR EDUCATIONALATTAINMENT ...................................................................................... 63 TABLE2.6: HONDURAS -SALARYPERHOURFORPRIMARYTEACHERS........................... 64 TABLE2.7: ELSALVADOR- EVOLUTION REALSALARIESFORPUBLIC AND PRIVATE OF TEACHERS .................................................................................................................. 64 TABLE2.8: HONDURAS-EVOLUTION OFREAL SALARIES FORPUBLICAND PRIVATE PRIMARYTEACHERS ................................................................................................... 65 TABLE2.9: SUMMARY FINDINGSOFPEDAGOGICALANDCURIUCULUMRELATED DETERMINANTSEDUCATION OF AMERICANLABORATORY............................................................................. QUALITY INCENTRALAMERICAAND THELATIN STUDY 68 TABLE2.10: EFFECTIVE CLASS HOURS INCENTRALAMERICAN COUNTRIES ..................71 TABLE2.11:TEACHERANDPRINCIPALCHARACTERISTICS INEFFECTIVE AND CONTROLGROUP SCHOOLS ....................................................................................... 75 TABLE2.12: PEDAGOGICALPRACTICESINEFFECTIVE AND CONTROL GROUP SCHOOL ..76 TABLE2.13: SCHOOLMANAGEMENT SCHOOLS ................................................................................................................... PRACTICES INEFFECTIVE AND CONTROL GROUP 76 TABLE2.14: NICARAGUA -TESTSCORES INMULTIGRADE TRADITIONALMODEL AND TABLE2.15: GUATEMALARETENTIONANDPROGRESSIONRATES FORmu AND SCHOOLS ................................................................................................................... 82 - TRADITIONAL SCHOOL STUDENTS ............................................................................. 83 TABLE2.16: TESTSCORESIN EBIAND CONTROL SCHOOLS INGUATEMALA (1986-1991) .............................................................................................................. 86 vi TABLE3.1: REASONS NONATTENDANCEIN THE 13-15AND 16-18AGE RANGES....100 FOR TABLE3.2:TOTAL EDUCATION EXPENDITURESECONDARY EDUCATION SOURCE IN BY ....................................................................................... 106 TABLE3.3: PUBLICENROLLMENTSHAREINSECONDARYEDUCATION(%) .................... OF FINANCING(US$. 000) 107 TABLE3.4: CURRENTAND TARGET SECONDARY GER (REPETITIONADJUSTED) FOR COST SIMULATIONS .......................................................................................................... 118 TABLE3.5: UNITCOSTSPERDIFFERENT TYPEOFPUBLICPROVIDERAND FINANCING MECHANISM, (2002)........................................................................................ US$ 119 TABLE3.6: ELSALVADOR- SECONDARY TARGETS SIMULATION RESULTS .................... 123 TABLE3.7: HONDURAS- SECONDARY TARGETS SIMULATION RESULTS....................... 125 TABLE4.1: PUBLICEDUCATIONEXPENDITUREPROPORTION GDP ..................... IN OF 129 TABLE4.2: FUNCTIONAL DISAGREGATION ........................................ 132 TABLE4.3: PER-CAPITA PUBLIC SUBSIDY INPFUMARY................................................. OF EXPENDITURE 142 TABLE4.4: PERCAPITA PUBLIC SUBSIDY INSECONDARY ............................................ TABLE4.5: ANNUALPRIVATESPENDINGPER STUDENTBYINCOMEQUINTILE ............142 144 TABLE4.6: PRIVATESPENDINGPER STUDENT BY INCOMEQUINTILE AS A PROPORTION ANNUALHOUSEHOLD ...................................................... OF INCOME 144 TABLE4.7: ECONOMIC DISAGREGATION OF EXPENDITURE .......................................... 149 TABLE4.8: PROPORTIONOFNATIONAL EXTERNAL AND FUNDSIN PUBLICEDUCATION EXPENDITURE, 2002us$MILLIONS (UNLESS OTHERWISESPECIFIED) ..................... 154 TABLE4.9: ALLOCATION EXTERNAL OF FUNDS 2002 us$MILLION........................................................................................................... ACROSSECONOMIC CATEGORIES - 155 TABLE5.1: ORGANIZATIONAL AND IMPLEMENTATIONRESPONSIBILITIES..................... 170 TABLE5.2: CHARACTERISTICSOFSCHOOLCOUNCILS................................................... 172 TABLE5.3: FUNCTIONSDECENTRALIZEDSCHOOL COUNCILS .................................. TO 173 TABLE5.4: FINANCIALTRANSFERS ............................................................................... 174 TABLE5.5: TEACHER . SCHOOLS ................................................................................................................. COMPENSATIONAND TENUREIN S B M VS TRADITIONAL 176 TABLE5.6: PARENTALINVOLVEMENTINCOMMUNITYSCHOOLSVERSUS 6 6REGULAR"SCHOOLS .............................................................................................. 179 TABLE5.7: SCHOOLANDCOMMUNITYAUTONOMY COMMUNITYSCHOOLS IN VERSUS "REGULAR" SCHOOLS. NICARAGUA. HONDURASGUATEMALA AND ....................... 182 TABLE5.8: TEACHER ATTENDANCE WORKHOURS AND .............................................. 185 TABLE5.9: AVERAGE STUDENT-TEACHER RATIO AND ENROLLMENT .......................... 186 TABLE5.10: COVERAGEEXPANSION: EDUCO, 1991-2001......................................... 188 TABLE5.11: PRONADE:ACCOMPLISHMENTS 1996,2000-2004................................. 190 TABLE5.12: NICARAGUA: EFFICIENCYINDICATORS FORAUTONOMOUS ANDNON- AUTONOMOUS SCHOOLS BY GRADE, 2003 ................................................................ 192 TABLE5.13: THIRDFOURTHGRADETESTSCORE AVERAGES SCHOOL TYPE.........^^^ BY vii TABLE OFFIGURES FIGURE1.1:STRUCTUREOFTHEEDUCATION SYSTEMS INCENTRALAMERICA.................2 FIGURE1.2:AVERAGE YEARS OFEDUCATION ATTAINEDINCENTRALAMERICA.............3 FIGURE1.3: EDUCATIONAL ........................ 3 FIGURE1.4: GERAND NERINPRIMARY(DIFFERENT DATA SOURCES) ............................. ATTAINMENTAND INCOMEPER-CAPITA, 2000 5 FIGURE1.5: GERAND NERINSECONDARY (DIFFERENT DATA SOURCES) ........................ 5 FIGURE1.6: THE"SECONDARY EDUCATION 6 FIGURE1.7: COMPARATIVEEVOLUTION NERs............................................................ GAP"............................................................ OF 7 FIGURE1.8: GERsAND GCRsINPRLMARY ..................................................................... 10 FIGURE1.9: GERs ANDGCRsIN SECONDARY ................................................................ 10 FIGURE1.10:COMPARATIVE SURVIVORPLOTS................................................................ 12 FIGURE1.11: RCMANDSFPINELSALVADOR............................................................... 14 FIGURE1.12:RCMANDSFPIN HONDURAS .................................................................... 14 FIGURE1.13: RCMANDSFPINNICARAGUA .................................................................. 15 FIGURE1.14:AGE-BY-GRADE FIGURE1.15: COMPARATIVESURVIVORFUNCTIONPLOTSBYAGE................................... DISTORTION AND PROPORTIONOF OUT OFAGE CHILDREN 16 17 FIGURE1.17: INCOMEPER-CAPITAANDREPETITIONRATEINPRIMARY ........................... FIGURE1.16: COMPARATIVEREPETITIONRATESACCORDINGTODIFFERENTSOURCES....19 20 FIGURE1.18: TESTSCORESRESULTSINELSALVADOR.................................................... 24 FIGURE1.19:TESTSCORESRESULTSINHONDURAS ........................................................ 24 FIGURE1.20:TESTSCORESRESULTSINNICARAGUA ....................................................... FIGURE1.21:TESTSCORESRESULTSINCOSTARrCA....................................................... 24 24 FIGURE1.22: THIRD GRADEMATHEMATICSRESULTSAND GDPPERCAPITA FOR TEST SELECTEDCOUNTRIES............................................................................................... 25 FIGURE1.23: THIRD GRADELANGUAGERESULTSANDGDPPERCAPITAFOR TEST SELECTEDCOUNTRIES............................................................................................... 25 FIGURE1.24: m A T ERATES OFRETURNINCENTRALAMERICAAND OTHER COUNTRIES ................................................................................................................. 28 FIGURE1.25: PRIVATERATES OFRETURNANDEDUCATIONALATTAINMENTIN FIGURE1.26: RATEOFRETURNBYLEVELOFEDUCATION................................................ CENTRALAMERICA................................................................................................... 28 29 FIGURE1.27:UPPERSECONDARYRETURNSANDEDUCATIONALATTAINMENT ................29 OF 30 FIGURE1.29: EVOLUTION GERANDNERACROSSURBANANDRURALAREAS .........32 FIGURE1.28: EVOLUTION RATESOFRETURNACROSSCOHORTS ................................ OF FIGURE1.31:SECONDARYGER,NERAND GCRINURBANANDRURALAREAS .............33 FIGURE1.30: PRIMARY GER,NERAND GCRINURBANANDRURALAREA ..................32 FIGURE1.32: HONDURAS-LEVEL OF URBANIZATIONAND COMPLETION RATE, BY DEPARTMENT(2002) .................................................................................................. 35 FIGURE133:EDUCATIONAL 36 FIGURE1.34: GER,NERAND GCRACROSSQUINTILEINPRIMARY .............................. ATTAINMENTOFTHE 25 YEARCOHORT............................... 36 FIGURE1.35: GER,NERAND GCRACROSS QUINTILESINSECONDARY ......................... 37 FIGURE1.36: AVERAGE SECONDARYGER AND GERBY QUINTILEINELSALVADOR AND NICARAGUA ............................................................................................................... 37 FIGURE137: SURVIVORPLOTSBYQUINTILE................................................................... 38 VI11 ... 39 FIGURE1.39: PRIVATERETURNSTOSCHOOLINGBYEARNINGQUINTILE ........................ FIGURE1.38: REPETITIONBY QUINTILE........................................................................... 39 FIGURE1.40:EVOLUTION ACROSS COHORTS OFRATES OFRETURNBY QUINTILE .............40 FIGURE1.41:PRIMARY COMPLETIONRATEBYGENDER................................................ 41 FIGURE1.42: PRIMARY GER.NERANDGCRFORINDIGENOUSANDNON- INDIGENOUS ............................................................................................................... 43 FIGURE1.43: SECONDARY GER.NERANDGCRFORINDIGENOUSANDNON- INDIGENOUS .................................................................................................................................. 43 FIGURE2.1:HONDURAS: DISTRIBUTION OFTEACHERSACROSS URBANAND RURAL AREAS BY LEVELOFEDUCATIONATTAINED ............................................................... 57 FIGURE2.2: ELSALVADOR- EVOLUTION REALSALARY. HOURSWORKED AND SALARY OF PER HOURFORPUBLICTEACHERS ............................................................................. 65 FIGURE2.3: HONDURAS- EVOLUTION REALSALARY. HOURSWORKEDAND SALARYPER OF ...................................................................... 66 FIGURE2.4: NOTIONAL EFFECTIVECLASSHOURSPERYEAR .................................... HOUR FORPUBLICPRIMARYTEACHERS AND 72 FIGURE3.1:ILLUSTRATIONOFCENTRALAMERICANEDUCATION DISTRIBUTION...........98 FIGURE3.2: PRIVATECOSTSOFSECONDARY ARESIGNIRCANTLY HIGHER THAN THEPRIVATECOSTS OFPRIMARY ............................................................................ 101 102 FIGURE3.4: COMPARATIVESURVIVORFUNCTIONPLOTSBYAGE................................... FIGURE3.3: COMPARATIVESURVIVORPLOTSBYGRADE............................................... 102 FIGURE3.5: PUBLICANDPRIVATEENROLLMENTBYINCOMEQUINTILES....................... 108 FIGURE :EDUCATION 4.1 FIGURE4.2: EDUCATIONEXPENDITURE SCHOOLAGEPOPULATION.2001/2002....131 SPENDING AS A PROPORTIONOFGNP. 1999/2000.................130 PER FIGURE4.3: FUNCTIONALEXPENDITURE 132 FIGURE4.4: EXPENDITURE CHILDAND STUDENT. 2001/02..................................... DISAGGREGATION................................ 2002 PER 134 FIGURE4.5: PUBLICEDUCATIONEXPENDITURE STUDENT IN%OFGDP PER PER CAPITA. 2000 ........................................................................................................... 135 FIGURE4.6: DISTRIBUTION PUBLICENROLLMENTAND POPULATIONSHARESBY OF INCOMEQUINTILEAND LEVELS OF EDUCATION ........................................................ 138 FIGURE4.7: COMPOSITIONOFPRIVATEEXPENDITURE THEPUBLICSECTORFOR IN SECONDARY EDUCATION .................................................................................. (%) 145 FIGURE4.8 :EDUCATION EXPENDITURE/GNP AND SECONDARY GROSS ENROLLMENT RATE.2001/2002..................................................................................................... 147 FIGURE4.9: EDUCATION EXPENDITURE/GNP 2001/2002............................................................................................................... AND PRIMARYCOMPLETIONRATE. 147 FIGURE4.10: PRIMARYCOMPLETIONRATEANDEDUCATIONEXPENDITURE. 2001/2002............................................................................................................... 148 FIGURE4.11:SECONDARY COMPLETION RATEANDEDUCATION EXPENDITURE. 2001/2002............................................................................................................... 148 FIGURE4.12: RECURRENT AND CAPITALEXPENDITURES. 2002 .................................... 150 FIGURE4.13: SALARY. NON-SALARY RECURRENTAND CAPITALEXPENDITURES. 2002 ................................................................................................................................ 151 FIGURE4.14: EVOLUTION EDUCATIONAL OF EXPENDITURE. TESTSCORESAND PRIMARYCOMPLETIONRATEIN HONDURAS (1998-2001)...................................... 152 ix FIGURE4.15: HONDURAS-DETERMINANTS BILL INCREASE OF WAGE ........................... 153 FIGURE4.16: INVESTMENT.NONSALARYRECURRENTAND SALARYEXPENDITURE FINANCEDBY EXTERNALAND TOTAL FLiNDS ............................................................ 156 FIGURE5.1: CONCEPTUALFRAMEWORK ....................................................................... 163 X ACKNOWLEDGMENTS This report was prepared by Emanuela di Gropello under the general supervision of Helena Ribe (previous Sector Leader, LCHSD), Eduardo Velez-Bustillo (Sector Manager, LCHSD) and Laura Rawlings (new Sector Leader, LCSHD). The task team also included Ilana Umansky (LCHSD) and Paula Savanti (LCSHD). Julie B. Nannucci was the Language Program Assistant. Peer reviewers were Manny Jimenez (EASHD) and Alec Gershberg (HDNED). The report incorporates the comments of the Ministries of Education of El Salvador, Guatemala, Honduras and Nicaragua (its four focus countries), largely provided through a regional workshop held in Dominican Republic on March 14-17, 2005 to present the main results of the report. Special thanks go to Nina Heinsohn (PRMPR), Jeff Marshall (Consultant, Stanford University), Renan Rapalo (Director of UMCE, Honduras) and Darlyn Meza (Minister of Education, El Salvador) for their help in putting together Chapter V on School-Based Management in Central America. Mike Walton (Regional Advisor, LCRCE) and Ariel Fiszbein (Lead Economist, LCHSD) also provided very useful comments on that same chapter. Thanks also to the members of the Central America HD team, Suhas Parandekar, Carlos Rojas, Joel Reyes, Angela Demas, Andy Mason and Manuel Salazar for support, advice andusefulcomments on this report draft. Finally, thanks to my special group of advisors who provided invaluable comments on education sector issues in Central America. The group includes Ian Walker (ESA Consultant, Honduras), Cecilia De Can0 (Ex-Minister of Education, El Salvador), Regina Caffaro de Morenos (Director of PRONADE, Guatemala), Abigail Castro de Perez (Director of the OEI for Central America), MarvinHerrera Arraya (General Secretary of the CECC), Arbelio Garcia (Director of UNESCO for Central America) and Donald Winder (Consultant, The World Bank). x i Chapter I: A Diannosis of Educational Performance in Central America The purpose of this first chapter is to compare basic education outcomes and indicators in the four Central American countries, which will then be examined and explored in the subsequent chapters. At least five main dimensions of educational performance should be considered in any education sector diagnostic such as this one: (a) educational coverage, measured by enrollment rates; (b) internal efficiency, measured b y student cohort survival rates and other indicators; (c) educational quality, measured by the acquisition o f cognitive skills; (d) external efficiency, measured by private (and ideally, social) rates of return to schooling at the various levels; and (e) equity, measured by the distribution among urban-rural areas, socio-economic groups and ethnic groups o f all the previous indicators. We will make the attempt below to compare the countries along these dimensions, using similar indicators, and, when applicable, triangulating indicators across multiple sources (official Ministry of Education sources and household surveys). When possible, we also provide longitudinal comparisons of these education indicators for each country. A key conclusion of the chapter will be that some urgent priorities remain in spite of several accomplishments undertaken in the past decades, in particular related to quality and learning, primarycompletion andsecondary education coverage. 1. Coverage of the EducationSystems Although geographically and culturally similar, countries in Central America present marked differences in socioeconomic terms. According to WDI data, per capita income levels in the region vary from a Costa Rican level of US$ 4,000 to less than one fourth that amount in Honduras or Nicaragua (US$ 900 and US$ 700 respectively). These differences are present across Central American countries in terms o f educational indicators and, to a lesser extent, educational structure. The different educational systems vary only slightly in terms of the structure of primary and secondary levels, official entrance ages and number o f years o f compulsory education (see Figure 1.1). The Nicaraguan system i s made of three years of voluntary preschool, six years of compulsory primary education, and a voluntary five-year secondary cycle. Technicalhocational and teacher modalities require an additional year (grade 12). Preschool education in Guatemala i s divided into two levels: initial and pre-primary education. The primary level consists of 6 years and secondary education i s divided into a basic "c011lfnon~' cycle o f three years and a diversified cycle that lasts between 2 to 4 additional years. El Salvador has a nine-year compulsory basic education level consisting of three, three- year cycles (traditionally thought of as primary and lower secondary). Secondary 1 education then consists of 2 or 3 additional years, depending on the modality. Like in the previous countries, preschool covers children aged 4 to 6. The system in Honduras was structured like that of Guatemala but has recently been reformed to resemble that of El Salvador, where instead o f having a primary six-year level, there is a basic level that consists of 9 grades, with a short subsequent secondary level o f 2 years (academic stream) or 3 years (professional stream). Costa Rica also follows the same pattern, with a basic education level consisting of three, three-year cycles, and a diversified secondary lasting 2 to 3 additional years. There are three options for diversified secondary in Costa Rica: academic (2 years), artistic (2 years) and technical (3 years). The official school entrance age i s 6 in Costa Rica. In contrast with the other Central American countries, Panama has a mandatory 2-year preschool cycle, then followed by 6 compulsory years o f primary, and 3 compulsory years o f pre-secondary level. The secondary level i s not compulsory and covers years 10 to 12. Figure 1.1:Structureof the EducationSystemsin CentralAmerica A. AVERAGEYEARS OFEDUCATIONATTAINED Advances in Educational Attainment. Average educational attainment has improved in all four Central American countries over the past 30 years (see Figure 1.2 below and Table 1.1 in Annex I). The improvements have been strongest in El Salvador, followed by roughly equivalent increases in Guatemala and Nicaragua, with the weakest advances in Honduras. Nonetheless, the Central American countries still compare poorly to the Latin American average in terms o f educational attainment although gains in educational attainment have, in general, been faster in Central America (with the exception of Honduras) (see Annex Table 1.2). Central American countries also perform quite poorly compared to countries with similar per capita income (see Figure 1.3). 2 Figure 1.2: Average Years of Education Attained in Central America Source: El Salvador: EPHM2002; Guatemala, ENCV 2000; Honduras, EHPM 2003; Nicaragua, EMNV 2001. Notes: (a) simple average of the four countries underanalysis Figure 1.3: Educational attainment and income per-capita, 2000 1 10.00 9.00 8.00 7.00 6.00 5.00 4.00 . 3.00 GUA 2.00 1.oo 0.00 0 1000 2000 3000 4000 5000 I Income Der-caDita Source: Barro and Lee andWDI B.GROSS NETENROLLMENT AND RATES Gross enrollment rates (GER) compare the total number of children enrolled in a specific education cycle as a percentage of the school-age population that corresponds to the same education cycle. Net enrollment rates (NER), on the other hand, consider in the 3 numerator only the number of students of the official age that actually attend that particular level or grade. The difference between the two thus provides a rough estimate of the share o f students in a specific cycle that are under or over aged. Figures 1.4 and 1.5 present NER and GER data from both MINED data and from household survey data (see also Tables 1.3 and 1.4 of Annex I). Small discrepancies between the two sources of data are likely to occur primarily because of the different points of collection of the information (household versus school), which can lead to different sorts of biases or deficiencies in data collection. For instance, households may have the incentive to over-estimate their schooling/access to school for social recognition. Or, statistics on age-per-grade may be poorly collected at the school level (if collected at all), leading to discrepancies in the NERs across the two sources. The ranking among the countries should be insensitive to the use of one or the other source. One exception concerns Honduras, where the ranking between countries depends on which data source i s used. This i s probably due to weaknesses in data collection in Honduras in both the ministry of education' as well as with the household survey. It is, therefore, imperative that Honduras focuses on improving, as it i s currently attempting, its data collection system. El Salvador and Nicaragua have been more successful in this effort (as indicated by more consistency in their indicators) although indicators calculated at the sub-national level are not always fully consistent. Entire data series have disappeared after the last hurricane. Attempts have been made up to 2002 to collect statistics in a consistent way (in particular a school census was carried out in 2000), with, however, dubious results. From 2002, new more ambitious school mapping exercises are under way, with an improvement in the 2003 data (this i s certainly the case at the secondary level, although the data seem to be over-estimated at the primary level). 4 Figure1.4: GER andNER inPrimary (differentdata sources) 120 100 80 60 40 20 0 El Salvador Guatemala Honduras Nicaragua Figure 1.5: GER and NER in Secondary (differentdata sources) 100 90 80 70 60 50 40 30 20 10 0 ElSalvador IGuatemalaI Honduras I Nicaragua I Costa Rica - ' I - secondary. At the primary level, GERs are uniformly higher than loo%, while NERs, High and homogeneous GER in primary Low and heterogeneous GER in although they vary slightly more across countries, are always about 80%. At the secondary level, both GERs and NERs are much lower and vary quite a lot across countries. Low efficiency in primary. A difference o f about 20 percentage points between GER and NER at the primary level combined with GERs over 100 percent suggests that the education system could accommodate the entire age-appropriate student population. 5 These figures also indicate that there i s substantial room for efficiency improvements - i.e. less repetition and dropoutheentry patterns - at the primary level in all countries.2 Low coverage in secondary. At the secondary level, a key issue for the countries under analysis i s low coverage, even in terms of GER. As will be discussed in subsequent sections, many students dropout after (or before) finishingprimary. It is clear from Figure 1.5 that both GER and, to a slightly lesser extent, NER are much lower in all four countries than the Latin American average. Figure 1.6 also indicates that there is a "secondary education gap" compared to countries with similar per capita income. The gap i s even worse when comparing NERs. Figure 1.6 also shows that while Nicaragua compares favorably with the other Central American countries in terms o f GER considering its per capita income, Guatemala compares very poorly. Figure 1.6: The "Secondary EducationGap" 1 1007 I 90 80 70 60 K 8 50 40 30 20 10 0 0 500 1000 1500 2000 2500 3000 3500 4000 4500 Income per-capita Source: Table 1.3Annex Iand WDI Figure 1.7 depicts the evolution of NERs in primary and secondary in El Salvador, Guatemala and Nicaragua since the mid nine tie^.^ Information comes from either official MINED statistics or household survey data depending on a~ailability.~Overall, we see that there was a positive trend in the GER and NER in all countries, although varying across countries. * This margin for efficiency improvements is even higher at the Latin American level. I t is difficult to find fully reliable comparative data before for most of the countries analyzed. It should be noted that the coverage of the secondary level does not always strictly correspond to the one used in Tables 1.3 and 1.4 of Annex Ifor limitation in the availability o f past data, but the trend should not quite differ. 6 Figure 1.7: Comparative Evolution of NERs 1995I 2000 I 2002 1996 I 1998 I 2001 1993I 1998 1 2001 ElSalvador Guatemala Nicaragua IHPrimary NER mSecondary NERl Source: El Salvador, EHPM, various years; Guatemala, MINED official statistics; Honduras, EHPM, various years; Nicaragua, EMN, various years; Costa Rica, MINED, officialstatistics. (a) Primaty is grades (1-6);secondary is grades (7-11); (b) lower secondary without brackets and upper secondary (grades 10-12)within brackets. NER increased particularly fast in Guatemala and Nicaragua, where the starting point was lower, and, at the secondary level, particularly fast inNicaragua. Interms of GER, El Salvador and Guatemala have also experienced high growth since the mid-nineties (see Table 1.5 in the Annex). In El Salvador, this trend was related to the comprehensive education reform started in 1995.5 Honduras i s the country where enrollment rates increased the least over this time-period. II. InternalEfficiencyof Education Systems This next section examines indicators of internal efficiency in Central American education systems. In particular, it looks at completion rates, survival rates, repetition rates, and age-by-grade distortion. A. COMPLETIONRATE The aggregate indicators presented above provide a snapshot o f the coverage situation, but they do not provide enough information for a full educational coverage diagnostic. This is because GERs and NERs refer to averages across all grades in a schooling cycle: a similar enrollment rate can thus be the product of high access to a certain cycle and high dropout during the cycle or low access to the cycle and low dropout during it, we cannot disentangle this. A useful and widely monitored indicator i s the Gross Completion Rate (GCR) that captures the number of children, of any age, who successfully complete the cycle as a proportion o f the population o f theoretical completion age for the cycle. See The World Bank (2004b). 7 We present two completion rate indicators. The first (Table 1.1)uses MINED data and i s the proportion between the enrollees in the last grade of each cycle minus the repeaters divided by the children of the official age of graduation. This figure, therefore, does not actually measure graduates of the level as MINEDs rarely produce this information. Because of this, the indicator tends to over-estimate the true completion rate. It i s still useful because it captures, at a given moment in time, the children "completing" a cycle whatever their age, and i s in fact being usedfor monitoring the MDG goals by the World Bank. Table 1.1: Completion Rates at the Primary and Secondary The second indicator (Table 1.2) Level uses household surveys to I Primary CR (Yo) SecondaryCR ("/.) explicitly measure cycles completed. Instead of capturing El Salvador (2002) 55 (grade 9) 40 (grade 11) 75 (grade 6) 22 (grade 12) completion at a given moment in Guatemala (2001) 51 16 (grade9) time, this indicator measures the primary cycle completion of a Honduras (2001) 69 19 (grade 11) (2001) 14 (grade 12) certain age-cohort over different (a) (2003) 80 31 (grade 11) periods o f time. This indicator Nicaragua (2002) 69 33 (grade 11) does not provide information on how things stand presently, but it Costa Rica (2001) 89 na does capture actual completion LatinAmerica (2001) and can be easily used to Brazil 90 61 (grade 8) compare completion across Chile 101 61 (grade 8) cohorts and time. Jamaica 95 61 (grade8) Mexico 100 61 (grade8) Central American countries Paraguay 78 61 (grade8) compare poorly at an interna- Peru 97 61 (grade 8) Ecuador 96 tional level. Both primary and Bolivia 72 secondary GCRs are lower than Colombia 85 Costa Rica, the Latin American 1 average, and countries with Countrieswith similar incomep/c (1995-2001) similar per capita incomes (see Peru 97 also Figure 1.1inthe Annex). Colombia 85 Jordan 104 Romania 98 GCRs are lower than the Algeria 91 GERs. When comparing GCRs Paraguay 78 to GERs (see Figures 1.8 and Ecuador 96 1.9), we note that in all countries Philippines 92 (except Costa Rica) GCRs are Bolivia 72 substantially lower at all levels. Source: MINED Official Data; Di Gropello, Dubey, Winkler (2003); WDI. This suggests that there i s high Notes: (a) 2003 is likely to be over-estimated at the primary level. 2001 was addedfor comparisonwith other countries. dropout during both educational cycles (although care needs to be taken when comparing GERs and GCRs because we are comparing different student cohorts). 8 GCRs are less than 80percent inprimary. Despite improvements incoverage, primary completion rates are lower than 80 percent6in all four countries. Guatemala, with a PCR of 51percent, is the lowest of the four. The target 100percentcompletion rate established under the MDG goals will require major efforts on the part of all countries. These efforts will need to focus on improving education quality (see Chapter 11) and reducing other barriers such as late entrance. Low GER and GCR in secondary. At the secondary level, GCR does not exceed 40 percent in any country. Part of the reasonbehindthese disturbingly low figures is the low enrollment rates seen above. GCR are also, however, due to high dropout rates in the cycle.7 1 Table 1.2 : Primary Completion Rate for Different Cohorts j 18year-old cohort 25 year-oldcohort 35 year-oldcohort El Salvador, 2002 76 68 54 1 1Guatemala, 2000 I 56 I 45 I 36 I Honduras, 2002 71 65 57 Nicaragua, 2001 55 56 42 Costa Rica, 2000 85 80 82 Source: Household Surveys. GCR evolution shows progress. On a more positive note, the evolution of gross completion rates for the primary cycle shows steady progress, particularly in Guatemala. The four countries are catching up with Costa Rica. El Salvador, for example, is only lagging by ninepercentagepoints. Using the more reliable and comparable 2001 data for Honduras. The reasons for the current low coverage and completion of the secondary cycle and possible options to improve them will be dealt with in Chapter 111. 9 I Figure 1.8: GERs and GCRs in Primary 120 71 100 80 60 40 20 0 El Salvador Guatemala Honduras Nicaragua Costa Rica IBPrimary GER HPrimaly GCR I Source: Table 1.3 Annex land Table 1.1 Figure 1.9: GERsand GCRs inSecondary I70 I 60 50 40 30 20 10 0 El Salvador Guatemala Honduras Nicaragua IBSecondary GER HSecondary GCR I Source: Table 1.3 Annex Iand Table 1.1 B.SURVIVAL RATE Survival rates measure the percentage o f a cohort of students enrolled in the first grade of primaryin a given school year who are expected to reach a specified grade level or cycle. Several methodologies are used to calculate survival rates, depending on the data source.8 The "reconstructed cohort analysis" technique (or pseudo-longitudinal analysis) is typically applied using official MINED data and, making use of two consecutive enrollment years and data on repeaters for the most recent year, simulates the schooling life o f an initial cohort of students, determining repetition, drop- out and survival rates. A second one is a technique that can be applied using household survey data, which involves estimating the probability of remaining in school for specified cohorts. Within this setting, simple survival function plots can be drawn by making use of two basic pieces of information contained in the household surveys: (a) the education years attained; and (b) the current attendance or not to a school. Both techniques have methodological limitations, in particular the reconstructed cohort analysis i s based on a 10 The comparison between the two mainmethodologies (reconstructedcohort analysis and survivor plots) can cast light on the on-going evolution of the internal efficiency of the education system because survivor plots illustrate more recent trends in survival than does the reconstructed cohort method (see footnote for a methodological explanation of each technique). Figure 1.10 graphs the current survival rates of the 7 to 17 cohort in the four Central American countries under analysis, according to the survivor function plots (see also Table 1.6 inAnnex I). Although each country is different, there is a constant loss of students throughout primary. Figure 1.10clearly illustrates that studentsdrop out throughout the educational cycle in all countries, but also that there are significant differences across countries in terms of the prevalence and characteristics of the dropout phenomenon (individual countries' survival plots are shown in Annex I). For instance, all countries currently have a sixth grade survival rate of between 70 and 80 percent, but El Salvador and Guatemala have a continuous and steady decrease in survival along the primary cycle, while Honduras and Nicaragua have a more irregular dropout pattern, with higher dropout rates in grades 4, 5 and 6. For these latter countries, this suggests that quality issues, supply- constraints (for example in Honduras, where close to 20 percent of primary schools have fewer than 6 grades), and late-entrance may in part be responsible for the irregular dropout rates. numberof restrictive assumptions and is very much dependent on the quality of the data on enrollment and repeaters collected by the MINED, while the survivor function plot measures a probability which relies on limited information and on the quality o f the households' replies. Additionally, beyond the data source, there is an essential methodological difference between these two techniques which, as we will see, generates substantial differences between them while also making their comparison meaningful. Survivor function plots make use of retrospective information on the survival rates o f all the previous cohorts in working out the probability o f survival of the most recent cohort, while at the same time updating/adjusting the survival probability to the higher grades by the increased survival probability to the lower grades, which ensures that the evolution in survival rates across cohorts and grades i s taken into account. In contrast, a reconstructed cohort analysis adopts a more retrospective approach by assuming fixed in time the probability of survival to each o f the grades. This second methodology which tends to under-estimate the current survival rates, although mostly for the higher grades, tends to generate a survival rate, which adjusted by the gross intake rate, provides a figure close to the current gross completion rate (it will in fact tend to be higher if the gross intake rate increased in time or if the internal efficiency parameters between grade 1 and grade 2 improved in time). A comparison between the two methodologies will be instructive as it can cast light on the on-going evolution of the internal efficiency o f the education system. 11 Figure 1.10: Comparative survivor plots SURWOR FUNCTION PLOT BY YEARS OF EDUCATION 1.oi :' ................ ElSalvador ........ 0.9: ........ -. -- .-..1 U .......... ---? ........ i5 i ....... ........ j . ,L 0.8: _ _ _ _ _ _-- ........ F----i I Honduras . !.. ... i_---' ...................... ...................... Is i i : 0.7; _ _ _ _ _, ~ .....-..........1 ......... V 9 a :' L--- Nicaragua I ----I I---- 0.6 1 i--- ----a I_-- !. - ._-.. ........: >. ..... ... l... Guatemala 0.5: I------; 1- ------ ~. ....... i_____ 0.4 , , , . , , , , , , , , , , , , , , , , , . , , , , , , , , , , , , , , , , , , , , , , , , , , , , r Source: El Salvador: EPHM 2002; Guatemala, ENCV 2000; Honduras,EHPM 2002; Nicaragua,EMNV 2001. Transition rates vary across countries. The differences across countries are even larger at the secondary level. A critical difference concerns the transition rate between primary and secondary. On the one hand, El Salvador has basically no change in survival rates through grade 9 (Le. decreases in survival rates continue at a similar trend). On the other hand, Honduras loses approximately 25 percent of students between the sixth and the seventh grade. Nicaragua and Guatemala fall somewhere in between. The steady progression of students from what are traditionally considered primary school grades (1- 6) through what are traditionally considered lower secondary school grades (6-9) in El Salvador is likely due to the country's education reform that reorganized grades into a basic cycle (grades 1-9) and in a secondary cycle (grades 10-11 or 10-12). This reform also included substantially increasing the number and availability of basic education schools. In Honduras, by contrast, there remain important supply-side constraints at the secondary level (Le. lack o f access to secondary schools). Furthermore, late entrance i s likely to be a cause o f highdesertion between 6th and 7th grade. It is striking to note that while Honduras has a roughly 10 percent higher survival rate through the last year o f primary, its survival rate to grade 11i s 10percent lower than that of Nicaragua. Honduras urgently needs to address this transition issue. Indeed Honduras, Guatemala, and Nicaragua could all benefit from compulsory nine-year basic education cycles such as that of El Salvador. Honduras, in fact, has recently introduced such a reform, and will, therefore, need to focus now on ensuring appropriate implementation and compliance. Few students make it to upper secondary. Looking specifically at survival rates in secondary, Guatemala, Honduras and El Salvador experience the greatest survival decline in the later years of secondary (transition to grade 10 specifically) while Nicaragua 12 experiences the greatest decline in the earlier years of secondary (grades 8 and 9). In all countries, grade 10i s the start of the upper secondary cycle, which offers a "diversified" curriculum - i.e. separate technical and academic tracks. Lower survival rates from grade 9 to grade 10 in Guatemala, Honduras, and El Salvador indicate that the transition into upper secondary may be problematic due to the pressure to work, lack o f space or lack of quality or relevance o f this cycle for some students. Ultimately, El Salvador has the highest survival rate to grade 11 (close to 70 percent) while Guatemala and Honduras have the lowest (less than 50 percent). Efficiency has improved. Figures 1.11, 1.12 and 1.13 juxtapose survival function plots (SFP) with the reconstructed cohort method (RCM) in El Salvador, Nicaragua and Hondurasg B y comparing these two indicators we can see that survival rates are improving in all three countries. Survivor plots have been adjusted to start from a level o f 100 percent in grade 1 (i.e. taking enrolled students as a basis and not the population). These improvements are likely to result in increased enrollment and completion rates. Efficiency improvements vary, however, across countries. Honduras appears to be making the most progress (where RCM survival rates are significantly lower than in the other two countries) indicating that it i s catching up with Nicaragua and Guatemala." El Salvador has a regular declining trend. Honduras now has a lower drop out rate between grades 7 and 8 while Honduras and Nicaragua both show higher dropout rates between grades 6 and 7." Under-estimated repetition rate in grade 1. Figures 1.11, 1.12 and 1.13 also plot each estimation along with a third estimation that i s the Reconstructed Cohort Analysis in which we have adjusted up first grade repetition to around 30 percent. We do this because although we expect a divergence between the two types of curves for higher-grade students due to their methodologicalpeculiarities (see footnote 8), we are surprised b y the substantial divergence already noticeable at lower grades, where the improvements in internal efficiency should still not have had such a strong impact on the relative survival rates. By adjusting the repetition rates upwards to about 30 percent in all three countries (that i s about double what the official figures report, which, we know, are bound to be under-estimated - see below), we obtain adjusted curves which provide much closer results than the ones generated by the survivor plots up to grade 3. We, therefore, make the plausible assumption that repetition rates in grade 1 may be under-estimated in MINEDdata. We did not have the official data to undertake a reconstructed cohort analysis for Guatemala. loCare should be taken in interpreting this result in the case o f Honduras, however, because the official MINEDdata, upto 2003, underestimatedenrollment at the secondary level. "Itwouldbeusefulto understand what iscausingthis:hasthe shortageofschools atthe secondary level become more acute ? Has late entrance increased so that when they finish grade 6 children need to go to work ? 13 Figure 1.11: R C M and SFP inEl Salvador El Salvador: RCM and SFP 1200 1000 - 80 0 50 g 600 & 40 0 20 0 0 0 1 2 3 4 5 6 7 8 9 10 11 Grades (+RCM +RCM (Rep Adi) +SFP j Source: MINED and HS Figure 1.12: RCM and SFP inHonduras i Honduras: RCM and SFP Source: MINED and HS 14 Figure 1.13: RCM and SFP inNicaragua C. AGE-BY-GRADE (Source: MINED and HS) DISTORTION AND REPETITION RATES I Nicaragua: RCM and SFP I 1200 The survival profiles 1 analyzed above tell 1000 you what proportion 80 0 of a given cohort reach any given grade 5 600 level, but they do not 400 provide any informa- ' ~ tion on how long it 200 takes to the students I I to reach that certain O 0 1 2 3 4 5 6 7 8 9 10 11 grade nor on the Grades average age of I-CRCM +RCM (Rep Adj) +SFPI students who reach that grade. Age-by- grade distortion and repetition ratesfill that gap. a)Age-by-gradedistortion Age-by-grade distortion measures the average number of years a student in a particular grade i s older (or younger) than the official age for that grade. Age-by-grade distortion can be broadly assessed by comparing the average age corresponding to each grade with the official age for that grade in each of the countries. The comparison between GERs and NERs presented earlier showed us that there i s a substantial proportion of over-age (or under-age) students inCentral American countries at the primary and secondary level. Age-by-grade distortion is high. The plots in Figure 1.14 report the age-by-grade distortion as well as the proportion of children out of age at each grade for the four countries using household survey data (see also Table 1.7 in the Annex). These plots show that the majority of students in every grade are not the official age for their grade. Inall four countries under analysis, the official entrance age is 7 and, if studentsproceed regularly through the grades without repeating, they should reach grade 11at the age of 17 (or grade 12 at the age of 18). Our evidence shows that, for each grade, between 50 and 80 percent of children are 'out of age' in all countries. This translates into an average age-by-grade distortion of between 5 and 20 percent higher than the official one. Several further points canbe madehere. 15 Figure 1.14: Age-by-grade distortionand proportionof out of Different intensity and age children patterns of age-by-grade (Source: Table 1.7 in Annex I) distortion and out of age students by country. The p-i magnitude of age-by-grade .- distortion varies across a countries, with Honduras and D 1150 pj Guatemala having, overall, an em B average distortion higher than 1100 Nicaragua and El Salvador.12 1050 ~36 n There i s an interesting 1wo convergence in the age-by- 0 950 I grade distortion in the highest I 2 3 4 5 6 7 8 9 10 11 12 Grades grades (at least up to grade 1l), I+ El Salvador +Guatemala -A- Honduras --Nicaragua 1 which i s not accompanied by 900. an equivalent convergence in 1 I the proportion of out of age 80 0 children. In Honduras, both the 700 age-by-grade distortion and the 2 out of age tend to increase, 500 c suggesting that there continue 400 to be high repetition rates even 300 in the higher grades (see below). In Guatemala, the proportion of out of age 1 2 3 4 5 6 7 8 9 10 11 12 students decreases slightly in Grades I+ the last grades of secondary - El Salvador +-Guatemala -A- Honduras +Nicaragua] this is likely to be due, in part, to older students dropping out once they reach the higher grades. Nicaragua and El Salvador are also interesting to analyze here. The increased age-by- grade distortion in Nicaragua, with a similarly increasing, although at a slower pace, proportion of out of age children, may indicate high repetition and/or less dropout occurring than in other countries and/or drop-outs more evenly distributed between out of age and age-appropriate students. Finally, a decreasing age-by-grade distortion in El Salvador, combined with an increase in the out o f age population, suggests fewer drop- outs than elsewhere, drop-outs more evenly distributed between out of age an age- appropriate students and older students dropping out relatively more. l2It is interesting to note that, while the lower relative age distortion in El Salvador can be clearly associated with a lower proportion of out o f age children, in Nicaragua, the lower relative age distortion is likely to be due either to smaller deviations from the official age in the out of age children or to over-age children being partly compensated by a high proportion of under-age children (since the proportion o f out of age i s very substantial in that country). There i s in fact evidence of a substantial proportion of children that enroll in grade 1at 6 in Nicaragua (see Arcia, 2003). 16 Age-by-grade distortion i s largely the product of late (or early) entrance to primary, and repetition along the educational cycle, and cycles of dropout and reentry. It i s difficult to disentangle the magnitude of these aspects by simply looking at the figures on age by grade, which are generated through official or household survey data. Attempts to disentangle these root causes of age-by-grade distortion have been undertaken for Honduras and El Salvador, showing, respectively, an estimated proportion of late entrants of 38 percent and 15 percent, and an estimated proportion of repeaters of 30 percent for Honduras and 35 percent for El Sa1vad0r.l~Along a similar line, an exploration of the main causes of grade-by-age distortion was undertaken for Guatemala by estimating the determinants of grade-by-age (see Box 1.1 in Annex I), revealing that repetition is the main cause o f grade-by-age distortion in the country. Figure 1.15: Comparative survivor function plots by age ElSalvador SURVIVOR FUNCTION PLOT BY AGE 0 ""-II.kiw . . ~ ~ . ~ . ~ ~ ,..... . . . ............... --& ..?i2i.... ..................: _ _ _ _... _ _.I...... ----.................. ? 0.9 __,.-......... - -- -- ~ .......................... ......................... Honduras Nicaragua -----I r---- Guatemala I-- --- --I I------- I 0.2 0.1 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , I 7 8 9 10 11 12 13 14 15 16 17 edad ,ource: El Salvador: EPHM 2002; Guatemala, ENCV 2000; Honduras, EHPM2002;Nicaragua, EMNV - 2001. Over-age leadsto lower educational attainment. Whatever the exact relative magnitude of these causes, late entrance, repetition, and dropout-reentry cycles are bound to have negative consequences on the schooling cycle of children. Furthermore, repetition l3 The applied methodology is a simple one. We mapped out the household surveys by age and grade in these two countries, calculated the proportion o f seven-year-old children enrolled in the primary cycle, multiplied the remaining proportion by the amount of eight-year-old children currently enrolled (which gives us the maximum amount of 8-year-old children who could have started late), and, therefore, obtained the late entrants and, by converse, also the repeaters of eight or more years (to take account of the seven year repeaters we also added to the repeaters the amount of grade 1 studentsof age 6 who did not make it to grade 2 the year after). Both numbers were then divided by the total enrollees in grade 1to obtain the late entrance and repetition rate. To be noted that this estimation, similar to a methodology applied by Schiefelbein (see Schiefelbein (2004)), assumes that all children get enrolled at either 7 or 8, which might not necessarily be true, and that the proportion of non-enrolled 7-year-old children has been constant for the last two years. 17 imposes a financial strain on the educational system as a whole. Survivor functions by age (not by grade this time) for all the countries under analysis (see Figure 1.15) show clearly that the probability of dropping-out of school increases quite dramatically for all countries once students turn 14 (probably at least in part due to work-related reasons). This implies that students who have successfully completed fewer grades by 14 are more likely to have fewer years of schooling when they leave school and lower educational attainment as adults. In fact, the higher age-by-grade distortion and proportion of over age students found for Honduras and Guatemala also contributes to explain why these two countries, in spite of having a similar structure of dropouts by age than the other countries, have lower attainment rates to grade 7 and each of the following grades. To illustrate this point further in Honduras, a thorough analysis undertaken by Edwards, Fuller and ParandekarI4 using a specially designed dataset which makes it possible to follow a specific cohort o f students (5,204 children) over two years, determined that 13 year-old and older children are more likely to drop-out at some point of the cycle, even if they have not completed primary, and, therefore, that school attainment i s substantially determined by the age of initial enr01lment.I~Among the main causes for late enrollment, the study finds family socio-economic status, family composition and the quality of schooling.16 b) Repetition High repetition rates in grade 1. For the reasons mentioned above, it is important to estimate the correct amount of repetition existent in the system. Figure 1.16 reports different repetition estimates, calculated according to different data sources and methodologies (see also Table 1.8 in the Annex). The evidence suggests that repetition i s particularly strong in the first year of primary education for all countries (higher than 10 percent everywhere), while it tends to decrease for the later grades of primary (as shown by the lower primary averages) and, with the exception of Honduras, for secondary. These repetition rates are, however, under-estimated. As shown in Figure 1.16, the direct estimates of repetition provided by the MINED official data and the household surveys are reasonably close, but the analysis of the age by grade that we undertook to infer repetition for Honduras and El Salvador, and estimates for Nicaragua generated according to similar method~logies'~,show that the real repetition rate i s likely to be at least twice as high for the first grade. Our estimates use some simplifying assumptions'*, but it is not surprising that the direct estimates on repetition produced from both the official and household survey data under-estimate real repetition rates." Repetition rates l4See Edwards, Fullerand Parandekar (1997). l5The study finds that every additional year of age at enrollment increases the probability of dropping out of school by 25 percent! l6Over-crowded schools might delay the age of enrollment. "SeeArcia(2003). Inparticular, that there are no late entrants of more than 8 years and that only in school over-age students are accounted for. l9With the official data, under-estimation would be due to misreporting of the schools' directors, who would consider drop-outs students who left their school, although they might have re-enrolled in another school (during the same year or after a lapse of time) at the same level and, therefore, should be really 18 of about 30 percent in grade 1 are also confirmed by the comparison of survival rates2' undertakenpreviously (see Figures 1.11-1.13). Figure 1.16: Comparative repetition rates according to different sources In any case, even by official standards, repetition is an issue in all the Central American countries under analysis. It tends to be higher than countries with similar income per- capita (see Figure 1.17), although it is in line with the Latin American average, and i s particularly high for the first grade of primary. The very high repetition rates in the first grade are particularly disturbing, and suggest several possible hypotheses21: (a) the curriculum or the standards may be unrealistically demanding; (b) schools are too full and, therefore, lower grades are usedas filters; (c) teaching quality might be too low; and (d) inherent characteristics, such as a language barrier or lack of academic support at home, might makeit impossible for studentsto perform well. considered repeaters. With household survey data, under-estimation would be due to the fact that, as with the age by grade disaggregation exercise, these data can really only capture the in-school students who have repeated but not the out-of-school ones (and will typically lead us to compare the students who were enrolled last year in one grade and repeating it this year with the current student enrollment in that grade, somewhat underestimating repetition); but also, and importantly, to the fact that they rely on parents' responses which are notoriously unreliable when it comes to repetition (in particular in the lower years). In fact, we note, interestingly, that directors' misreporting in the official data produces an under-estimation that turns out to be similar to the one produced using the direct question on repeaters of the household surveys. Overall, it is likely that the age-for-grade analysis provides us with a more realistic, although maybe slightly over-estimated repetition prediction, while using direct information on repetition provides under-estimates of the true amount. 2o To be noted here that survivor function plots do not depend on the information on repetition, while the reconstructed cohort analysis does (although, for the construction of the model, only the first grade repetition will really have an unambiguous impact on the survival), making the first methodology sounder in this sense. 2'This is basedon Edwards (2004). 19 Figure 1.17: Income per-capita and repetition rate in primary Some of these (Source: Table 1.8 in Annex Iand WDI) hypotheses will be tested in Chapter 11. 16 The next section of _-I_____- - + GLIA this chapter examines 14 - + comparative indi- 12 - cators of learning + outcomes in Central 10 - NlCA *--- America. a - I - +ES I -HON ~ 111. Quality of ----I 6 -. + Education 4 - A + Systems 2 - -- Just because a child Y 0 1 i s enrolled in school does not mean that he or she i s learning the skills and knowledge intended. There are many cases of countries with close to full completion rates in primary or secondary, but low student learning outcomes. Additionally, quality itself will be an important determinant of permanency in school and, ultimately, earnings. Papers such as Marshall (2003a)22 and Bedi and Edwards (2002)23on Guatemala show that indicators of education quality such as days o f instruction, teacher experience, teacher training or school infrastructure are also significant determinants o f desertion and earnings; or even that gains in tests scores themselves are a significant determinant of school attendance (Marshall, 2003b).24 A. STANDARDIZEDTESTING Learning outcomes are thought to be an appropriate measure of education quality and learning outcomes are frequently measured through standardized exams. Most countries now have some form of national (census or sample-based) standardized assessment given at the end of a schooling cycle that measures literacy and numeracy skills. The participation in regional assessments or, even, international assessments, such as the TIMSS (the Third International Mathematics and Science Study implemented by the International Association for the Evaluation o f Educational Achievement and the PISA (the recently OECD launched Program for International Student Assessment)26, also provide a country with some measure of learning outcomes, and allow comparisons with other countries, providing some objective benchmarking of the country's performance. For these reasons while standardized exams - both national and 22Marshall (2003a). 23Bedi and Edwards (2002). 24Marshall (2003b). 25The test is designed to assess achievement levels in mathematics and science at grades 4 and 8. 26The test is designed to measure reading, mathematical and scientific literacy attained by 15-year-olds. 20 international - do not measure all the facets of student learning, they are a useful means of measuring learning. Countries may choose to use standardized exams in conjunction with other measures o f learning and education quality. All Central American countries, as part of their development of national educational assessment systems, which we will review in more detail in Chapter 11, have now adopted standardized testing at the primary andor secondary level. Table 1.3 summarizes the existing standardized exams in the countries under analysis with some characteristics. Central American assessment systems cannot be compared. Overall, as we will develop in Chapter 11, each o f the Central American countries has a different testing system. On the one hand, we have cases like Costa Rica where standardized exams cover several grades, are used (in some cases) for accreditation, and have been introduced, at least for the last grade of secondary, as long ago as 1988 or 1994.On the other hand, we have cases like Nicaragua, where exams are only used at the primary level and were only introduced in 2002. Additionally, and very im ortantly, all exams tend to be designed differently and are based on national standards! which also implies that the score scales and the interpretation of the different performance levels (say basic, intermediate, etc) are different across countries. These different characteristics, together with the absence of a Central American regional exam based on Central American standards (such as the one existing for the Caribbean for instance28), make it fundamentally impossible to compare test scores across countries. Very limited participation in regional exams; no participation in cross-regional international exams. Only Honduras and Costa Rica participated in the Latin American regional assessment exam (the UNESCOLLECE exam undertaken at the Latin American level in 1997)29,while both Honduras and Guatemala also participated in a special study which combined the test instruments developed by UNESCO, with items developed by the Universidad del Valle for the ministry o f education in Guatemala and the Universidad Pedugogica for the education secretariat in Honduras (the results for Honduras are still not available). Unfortunately, these two exams are not comparable. Furthermore, none of the Central American countries have, to date, participated in a cross-regional international assessment exam. *'Asystemof Central American standards has recently been approved at the primary level, promoted by the CECC and the OEI, but only El Salvador has started to adapt its national standards to these regional standards. 28 *'The so-calledCXC (CaribbeanExaminationCouncil). CostaRicadid not releaseits results. 21 Table 1.3: Standardized Testing in Central America Grades tested Areas tested Years Frame of Sample or census reference based El Salvador Grades 3, 4, 6, Math, Social 1994-1998 up to 2001: Sample based 9 Sciences, and2001 (b) Natural Norm based3' Sciences and exams Language 1997-2001 Since2001: Census based and 2002 (b) Criterionbased. National standards. Guatemala Grades 3, 6 Language and 1998-2001 Norm based. Sample based Math National Standards. Honduras Grades3,6 Language 1997-2002 "Rendimientos Samplebased Math basicos" up to Natural 2004. National Sciences standards establishedin the new National Basic Curriculum from 2005. Criterionbased. ~ ~~ Nicaragua Grades3,6 Language, 2002 National standards Sample-based Math established in the current curriculum. Criterion based. Costa Rica Grade 6 Language, 1998-2002 Since 1995: Sample-based Math, Social Nationalstandards Sciences and established inthe Grade 9 (c) Natural National Sciences, 1996-2002 Curriculum. Census-based Foreign Objective based Grade 11 or 12 Languages 1988-1994 ( 4 Census-based (c) and 1995- 2002 Since 1999: Norm based. Notes: (a) frueba de Aptitudes y Aprendizaje para Egresados de Education Media, which accredits the end of the secondary cycle; (b) the tests undertaken in 2001 and 2002 are not directly comparable with the other years because of the introduction of the new evaluation system (SINEA); (c) is also an accreditation exam; (d) combination between criterionand norm based. B.LEARNING OUTCOMES Low learning outcomes on national exams. Given all the caveats, Figures 1.18 through 1.21, which depict some test score results at the primary and secondary level, should merely be used to compare each country against itself.31 Table 1.9 in the Annex gives 30 I n norm based exams, the result obtained by the student i s determined taking as a basis the results of the whole student group being examined. In criterium based exams, the results obtained are compared with benchmarks that students need to achieve to dominate a number of skills and knowledge areas. ''Although exams are not always fully comparable in time for the same country - this is for instance the case of Guatemala. 22 more primary and secondary results for each country in different time periods. Overall, what i s clear when analyzing the results in each of the countries i s that: (a) in all countries, except Costa Rica, average test scores are considered to be either low (basic) or low-intermediate vis-&vis the country's established standards; and (b) in all countries where exams are comparable over time, there i s no clear improvement trend in the results over the second half of the 1990sor early 2000s. Itis important to keep inmindthat being "intermediate" inEl Salvador is notthe same as being "intermediate" in Nicaragua, for example. In El Salvador, students are considered to be in the intermediate achievement range if they can identify and understand but cannot elaborate or apply their knowledge. In Nicaragua, this description corresponds more closely to what the country defines as a "basic" level of achievement. In any case, in terms of skills, and more particularly reading skills, it is important to understand that the low learning achievement reported for all countries i s associated with low levels of reading comprehension (which i s often taken as the most relevant indicator of educational quality at the third grade). In Honduras, it appears that in grade 3 a majority of students cannot identify characters or ideas in what they read nor identify parts of s eech. In grade 6, students test slightly higher in Spanish but slightly worse in mathp'. Results are substantially lower in disadvantaged schools. In El Salvador, around 25 percent of students were not able to identify the main message o f a simple written test in 2001 and about 40 percent were not able to make slightly more complicated inferences about it.33 These percentages increase to 50 and 60 percent respectively, if we consider the fraction of correct answers that are the result of mere chance, and they substantially increase for the students whose overall result to the exam i s below average. They are also substantiallyhigher for the poorer socio-economic groups. Progressing through the grades, we also see that grade 9 results are particularly low inthe countries for which there i s testing at the lower education level (El Salvador and Costa Rica). 32 Care needs to be taken that the Honduras standardized exam may not reflect fully the curriculum contents. 33 See Schiefelbein (2004). 23 Figure 1.18: Test scores results in ElSalvador (Source: Table 1.9 inAnnex I) El Salvador Grade 3,2003 I El Salvador Grade 9,2002 650 _ _ 600 1400 550 1350 500 1300 450 1250 400 1200 350 1150 300 1100 Spanish Math I Spanish Math Figure 1.19: Test scores results in Honduras Figure 1.20: Test scoresresults inNicaragua Hondurasgrade3 and 6,2002 Nicaragua grades 3 and 6,2000 100 I 90 400 350 70 300 60 I Spanish Math Figure 1.21: Test scores results in Costa Rica Low learning outcomes on regional exams. The results Costa Rica grades 9 and 11,2001 of the 1997 Latin American IO0 assessment (see Figures 1.22 90 and 1.23) show that 80 Honduras under- performs 70 vis-&vis other Latin 60 American countries, even 50 considering its low GDP. 40 The Latin American 30 countries, inturn, perform 20 10 poorly vis-&vis the OECD, 0 as illustratedby Chile, one of Spanish Math the best performers inLatin America, which ranked poorly on the PISA exam. 24 Overall, this brief analysis suggests that quality, as measured by learning outcomes i s an issue that requires urgent attention in Central America. The barriers to quality, which will involve analysis of learning determinants in Central America, will be addressed in the next chapter. Figure1.22: Third GradeMathematicsTest Resultsand GDP per Capita for Selected Countries Figure 1.23: ThirdGradeLanguageTest ResultsandGDP per Capita for SelectedCountries , i 400 350 10,000 300 te8,OOO 250 u) 36,000 200 150 4,000 100 2,000 50 0 0 Source: UNESCO/LLECE (2000) and US Government Central Intelligence Agency, The World Fact Book. 2001. IV. Rates of Returnto Education While test scores are used as a frequent proxy for education quality in that they measure student learning, longer-term outcomes such as individuals' income, health, and happiness, and countries' growth, prosperity, and peace, are larger goals for education systems. These long-term outcomes are nearly impossible to measure because they are affected by multiple other factors that are difficult to control for, and because longitudinal data i s very difficult to come by. One measure of longer-term educational outcomes that 25 can be analyzed i s the rate of return to education. Rates of return measure the impact of schooling on individuals' future earnings. They are important not just as a way of measuring the quality and effectiveness of schools, but also because people's earnings are a critical link between education and national growth, poverty reduction, and income distribution. This section analyzes the private rates of return to schooling in Central America. W e find that while private rates of return in Central America are generally comparable to other parts of the world, they are low compared with much of Latin America and have been declining in recent years. We also find, as expected, that rates of return are generally higher for the higher education levels but that they are oddly low for the lower secondary level across Central America signaling problems at this level. While many factors contribute to individuals' future earnings, calculation of private rates of return to schooling provides a widely used method to separate the impact of co-varying effects on wages. This allows us to estimate the impact of educational attainment on later earnings. Annex I1provides a description of the models that were specified to estimate private rates o f return for the four countries in the report. In essence, we address two main questions: (i) what is the value of an extra year o f schooling; and (ii) i s the what value of the different educational levels (primary, secondary, and tertiary). Unfortunately, due to data and time restrictions we do not investigate social rates o f return to schooling.34 Social rates of return measure the net benefit to societies of educating their citizens. Part of the social rate of return to schooling i s the private rate return - how individuals benefit from schooling - but other benefits, such as higher national productivity, technological and intellectual development, and reductions in social costs due to imprisonment, health problems, high birth rates, and general poverty, are not evident in estimates of private rates o f return. It is highly probable that in Central America social rates o f return to schooling are considerably higher than private rates o f return. We report private rates of return to schooling, overall and by education level, in Table 1.4. To make the analysis as comparable as possible among countries, we undertook these estimations using only the salary of individuals' main occupation, and usingthe cohort o f 18-60 year-old dependent workers. We also divided this group into three sub-cohorts to assess the evolution o f private rates o f return over time.35 Econometric details on the estimations are also reported in Annex II. To facilitate the interpretation of the rates of return, we also present a number o f instructive figures (Figures 1.24-1.30). Private rates of return in Central America tend to be lower than the Latin American average. Our returns indicate that an additional year of education generates an increase between 9 and 11 percent in private earnings in the four Central American countries (Figure 1.24). While this i s not an insubstantial figure, it i s lower than the average 12 percent in Latin America. What makes this particularly concerning i s that 34Studies of the social returns to schooling are much rarer than those of private returns to schooling. One recentstudy is Junankar and Liu (2003). 35An F-test was performed to assess if the rates of return are significantly different across cohorts, finding the differences statistically significant. 26 private rates o f return tend to be lower when larger proportions of the population are educated because there i s a large supply of skilled labor.36 Because average educational attainment is higher in Latin America as a whole than it is in Central America we would expect to see higher private rates of return inCentral America. The fact that the rate of returnis lower inCentral America raises some concerns about the quality o f schooling as well as the availability o fjobs and demand for skilled labor in the region. The lower rate of return in the Central American countries suggests that an additional year o f education provides less added benefit to would-be employers, probably because o f the low quality of schooling discussed previously. It may also mean - either alternatively or in addition - that there are not enough skilled jobs available for people as they complete their schooling. There i s little evidence, however, that the demand for skilled labor i s flat in Central America.37 Table 1.4: Private Rates of Return Cohorts: 18-60 18-30 31-45 46-60 IEl Salvador, 2001 Average 9.4 6.8 10.8 11.4 Primary 6.2 2.2 6.5 9.4 Lower Secondary 4.8 4.5 6.0 7.5 Upper Secondary 11.9 8.9 14.3 18.0 Tertiary 18.9 17.9 18.9 13.0 Guatemala, 2000 Average 11.5 9.9 12.0 13.7 Primary 8.2 5.7 7.5 12.3 Lower Secondary 6.2 7.5 7.6 1.9 Upper Secondary 24.3 21.4 26.0 33.0 1 Tertiary 10.4 7.0 10.9 9.4 Honduras, 2002 Average 10.4 8.7 11.2 11.7 Primary 8.7 6.4 9.9 10.1 Lower Secondary 7.9 6.0 10.8 11.7 Upper Secondary 13.8 13.3 12.4 15.1 Tertiary 13.3 10.9 13.9 12.4 Nicaragua, 2001 Average 9.3 7.6 10.2 10.4 Primary 5.8 4.4 6.3 7.3 Lower Secondary 6.2 4.2 7.8 9.8 Upper Secondary 10.7 11.6 11.6 2.2 + Tertiary 18.2 17.5 16.3 24.0 Average 10 Latin America Average 12 Asia Average 9.9 Average 7.5 Source: Household Surveys;SANIGEST (2003);and Psacharopoulosand Patrinos (2002). 36This is evidenced by the lower private returns to schooling in the OECD and Asian countries recorded in Table 1.4 as well as from the higher private returns in Guatemala and Honduras where average years of schooling is lower than in Nicaragua and El Salvador (see also Figure 1.15). 37See AED (2004) for El Salvador and The World Bank (2004d) for Honduras on this subject. 27 While private returns to schooling in Central America are more or less on par with those in Costa Rica and the Asian countries the higher educational attainment rates in these comparison countries suggest that private returns are low inCentral America. Figure 1.24: Private rates of return in Central America and other countries I1 4 7 I1 I E l S a l v a d o r Guatemala Honduras Nicaragua Costa Rica LAC Asia O E C D I Source: Table 1.4 Figure 1.25: Private rates of return and educational attainment in Central America 9 9.5 10 10.5 11 11.5 12 Coefficient on Years of Schooling Source: Table 1.4and Table 1.2 in Annex I. Rising returns with the education cycle. When we disaggregate rates of return b y education cycle (also Table 1.4), we see that in El Salvador, Honduras and Nicaragua returns increase from primary to upper secondary and again to tertiary (ignoring for the moment lower secondary). In Guatemala, the pattern i s different. Private returns increase dramatically between primary and upper secondary, but then decrease at the tertiary level. This pattern of generally rising returns with the education cycle (also visible in Figure 1.26) i s consistent with those found in other Latin American countries although the 28 returns to secondary education tend to be slightly higher in Central America.38 This pattern of rising returns can, again, be explained by supply issues. Central American countries, as we have seen, tend to have low enrollment at the secondary and tertiary level, compared to a muchhigher coverage at the primary level making individuals with secondary and tertiary education a limited commodity. Figure 1.27 illustrates this. It shows that for the cohort of individuals 25 years old or older higher educational attainment i s associated with lower private returns. The particularly low educational attainment inGuatemalamay, inpart, explain the high returns to upper secondary. Figure 1.26: Rate of return by level of education I I 30 Source: Table 1.4 Figure 1.27: Upper secondary returns and educational attainment Source: Table 1.4 and Table 1.2inAnnex I I 5 10 15 20 25 30 Coefficienton Upper Secondary 38See, for instanceThe World Bank (2003). 29 Central America is characterized by low private returns to lower secondary. As we noted above, countries tend to have higher private returns to higher levels of education. But in Central America the private returns to lower secondary are lower than the returns to primary in all countries despite significantly lower enrollment rates at that level and despite the fact that students should be learning valuable skills in lower secondary. The low private returns to lower secondary are largely responsible for the unfavorable difference between private returns in Central America and those of Latin America. The low private returns to lower secondary signal a problem at this cycle in terms of quality and relevancy. This will be further discussed in the next chapters. Here let it suffice to say that lower secondary i s a sub-cycle that tends to be seen as a preparation for upper secondary, rather than a stand-alone cycle that provides students with important skills. This has a somewhat important implication: relatively low private returns at this level are likely to serve as a disincentive for students to enroll and complete lower secondary causing a barrier to educational progress and making it impossible for those who drop out to continue on to upper secondary where they would reap higher returns. Returns to schooling have declined across time as attainment has increased. Figure 1.28 illustrates how rates of return evolved across three different cohorts. Although private returns remain substantial, younger generations do not reap the same education premium as older generations. This trend may be explained by the increase in educational attainment that occurred in all countries within a similar time frame. This attainment increased, so did the supply of educated workers, causing some wage compre~sion.~~ Part of the explanation may also lie in a somewhat "flat" demand for skilled labor. As mentioned above, there appears, however, to be sustained demand for skilled labor in Central America. Annex I11 presents a more detailed analysis of the evolution of rates of return, also by education level. Figure 1.28: Evolution of ratesof return acrosscohorts (Source: Table 1.4) 16 14 12 The next section looks at equity in Central American 10 education. Later in that section 8 we look specifically at 6 differences in private returns 4 to education across income quintiles. 2 0 39This is one of the mainreasons advancedby Edwards (2004) for Guatemala. He also finds, usingdata on 15 different cohorts, that rates of return have tended to decrease in Guatemala and that wage compression, following the increasein the supply of educated workers, could be partly behindthis observeddrop. 30 V. Equity of EducationalOutcomes Upto this stage, educational outcome indicators have only beenpresented as averagesfor each of the countries, not allowing us to appreciate the differences in outcomes within each of the countries. Key dimensions to be looked at in an equity diagnostic include the urbdrural discrepancy, the differences across socio-economic strata, the gender gap, and, when applicable, the differences across ethnic or linguistic groups. This section reveals that inequalities ineducation are a pervasive and urgent challenge. A. URBANmURALDIVIDE Looking at how urban and rural populations differ across educational outcome indicators i s particularly important in Central America where large proportions of the population continue to live in rural areas. Populations are slowly and steadily moving toward urban areas, (see Box 1.2 on San Salvador) which also creates challenges for urban centers that must take in new groups of students each year from a very different socio-cultural background. Table 1.10 in Annex Ireports the disaggregation of a number of core educational indicators across urban and rural areas in all countries under analysis (largely constructed usinghousehold survey data). We compare the enrollment ratios that we obtain usingthe official MINED data and the household survey ones in two countries, El Salvador and Honduras, for which we had both data sources available, to show the discrepancy that may result from using these different sources because of migration patterns across areas (see Figure 1.3 in the Annex). Household and MINED data are very similar for GERs at the primary level but they differ somewhat at the secondary level. MINED data reports higher proportions of students inurban schools than does household data inboth countries, although more so in El Salvador. This is likely due to children from rural households going to live in urban areas for secondary school (and living with relatives, for example). In MINED data the child i s considered urban while inhousehold data the child is considered rural.40We will use the household data primarily, as it more accurately captures the dominant urban or rural identity of the child. The urban-ruralenrollmentis shrinking. Usinghousehold survey data, the enrollment and attainment gap between urban and rural areas has been shrinkingover recent years in all countries for which we have past data (Figure 1.29). The rate of this change differs 40 What i s more unexpected is that, in El Salvador, household survey data show a similar primary GER in the two areas and a decrease in time in both indicators, which seems to indicate higher efficiency in the two areas, while official data show a much higher level o f the rate in rural than in urban areas and a marked decrease in urban areas and marked increase in rural areas, which suggest decreasing coverage in urban areas and higher coverage and inefficiency in rural areas. What this may imply is that household surveys are less effective in collecting information in rural than in urban areas (population more spread out, difficult to reach), or still, that official data under-estimate enrollment in urban areas compared to rural ones (schools' directors take less time to respond, etc). In any case, this discrepancy has important policy implications since the decreasing trend in enrollment in urban areas showed by the official data would suggest that urgent action is required to counteract this trend, while the household survey data would not lead us to the same conclusion. 31 across countries, being strongest in Nicaragua. The progress i s due to a proportionally stronger increase in rural than in urban areas in educational attainment and enrollment. Figure 1.30 illustrates this point for primary. GERs in primary education are either equivalent or close to equivalent between urban and rural areas in all countries (a little less so for Guatemala where the GER in rural areas i s still lower than 100percent). Most countries invested heavily in expanding access to schools inrural areas in recent decades. The development of rural community-based schools in Honduras, Guatemala and El Salvador discussed in Chapter V explain, inpart, the coverage increase. Figure 1.29: Evolutionof GER and NER across urban and rural areas (Source: Table 1.10 Annex I) Lower efficiency in rural areas. Nonetheless, Figure 1.30 also illustrates that primary completion rates (for the 12-21 year-old cohort) are still much lower in rural areas than in urban ones. This points to challenges in getting students to progress successfully through school and remain in school. Additionally, differences in NERs tend to be slightly larger than the ones in GERs, indicating that there are higher levels of inefficiency in rural areas, and both late entrance and, above all, repetition rates are higher inrural areas (see Figure 1.4inAnnex I). Figure 1.30: Primary GER, NER and GCR inurban and rural area (Source: Table 1.10 Annex I) Secondary coverage gap between rural and urban areas. At the secondary level there remains a huge gap in terms of both GERs and NERs between urban and ruralareas (see Figure 1.31). Keep in mind that temporary migra- tion is accounted for when we use household survey data. The gap i s even larger in completion terms. 32 Lower educational achievement in rural areas. Finally, large discrepancies in educational achievement exist between urban and rural areas (as measured by test scores -SeeTable 1.11inAnnexI).Guatemala, HondurasandElSalvador, thesedifferences In systematically favor urban areas. These discrepancies do not wholly favor urban areas in Nicaragua (nor in Costa Rica), however, where rural students score just slightly lower than urban students in math andjust slightly higher than urban students in Spanish. The lower differential between tests scores in Nicaragua may contribute to explain the lower repetition gap across areas. Figure1.31: SecondaryGER,NER and GCR inurbanand ruralareas 90 80 70 60 50 40 30 20 10 0 GER NER GCR GER I I I NER IGCRGERI NER I GCRGERI NER I GCR El Salvador Guatemala Honduras Nicaragua Source: Table 1.10 Annex I I s the primary enrollment rate falling in urban areas? There i s some indication that primaryenrollment and attainment may be falling inurbanareas. InEl Salvador, primary completion rate of the 12-21 year-old cohort between 1995 and 2002 have fallen from 76 to 73 percent inurban areas according to household survey data. Similarly, GER has been falling duringthis same time inurban areas according to MINEDdata. This couldbe seen as an indication that urban areas are being unattended. As more and more of the rural population moves to the large cities in El Salvador, it i s plausible that urban areas are getting increasingly overcrowded and poor. This i s likely to be contributing to the declining enrollment indicators. 33 that surround th ation. Whereas t heWorldBank (20 A similar trend exists in Honduras. The plots of Figure 1.32 graph each of the departments o f the country by their level of urbanization and the level of their primary gross completion rate (calculated from official data as the current enrollees minus repeaters in grade 6 over the 12 year-old population), both overall and by urban and rural areas of the departments. Looking at the figures, two observations come to mind. First, there i s a strong positive relationship between level of urbanization and primary completion rate but this relationship i s driven by the rural areas o f the departments rather than by the urban areas of the departments (inthe most urban departments, rural areas do much better than in the most rural ones, while differences are less clear-cut in urban areas). In generally urban departments there i s no clear relationship between how urban a particular area is and the PCR inthat area. Second, the two most urban departments inthe country, Cortes and Francisco Morazan, have significantly lower completion rates than the rough pattern in the rest of the country in both urban and rural zones of the two departments. The main implication of this finding i s that addressing completion in urban and rural areas in Honduras may require a dual strategy. Inparticular, rural completion may require a targeting of resources in the most rural departments, while completion in urban areas seems to require a major effort in the departments of Cortes and Francisco Morazan, on the one hand, and a somewhat global approach, not targeted per level of urbanization of 34 the departments, on the other one. El Salvador may require a dual strategy as well, to address the increasing coverage issue in San Salvador.41 Figure 1.32: Honduras-Level of urbanization and completion rate, by department (2002) ' I 20000_ _ _100 _ . 200 300 400 500 600 700 800 0.0 100 200 300 40.0 500 600 700 800 Level of urbanization(%) Levelof Urbanization(%) 120 , - 25 80 - * * A e* Morazan g! 60 K m .--540 L f 20 0 0 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 Level of urbanization("A) Source: The World Bank (2004~) B.SOCIO-ECONOMIC DIFFERENCES INEDUCATIONAL PERFORMANCE Educational indicators in Central America are sharply divided along socio-economic lines with children from poor families suffering the inferior outcomes. 4' See The World Bank (2004 b). 35 Figure 133: EducationalAttainment of the 25 year cohort (Source: Table 1.12 Annex I) Still substantial inequities in I400 coverage and completion. ~ Figure 1.33 illustrates, in a 12 00 snapshot, the substantial 1000 difference in educational 8 00 attainment across quintiles of 6 00 the current 25-year cohort. 4 00 Figure 1.34 shows that, while 2 00 for most countries, inequities 0 00 are relatively small in terms of the current gross enrollment rates at the primary level (another encouraging result), they tend to increase slightly in terms of the net enrollment rates (indicating higher incidence of late entrance and repetition in poorest quintiles) and, even more, completion rate42,at this same level. Table A.12 in Annex Ireports the current level of a number of educational indicators across income quintiles for the countries under analysis. Figure 1.34: GER, NERand GCR across quintileinprimary It should be noted that in Guatemala 140 0 even GERs differ 120 0 sharply over io00 quintiles. It shows 80 0 signs of being the 60 0 most inequitable country in this 40 0 sense. Recall that 20 0 we found that 0 0 GER INER IGCR GER I NER IGCR GER I NER IGCR GER I NER IGCR Guatemala was El Salvador Guatemala Honduras Nicaragua the only country where rural GER (Source: Table 1.12 Annex I) was below 100%. These inequities are then magnified at the secondary education level (where even GERs are very different across quintiles) (see Figure 1.35). In El Salvador - arguably the most equitable of the four countries - approximately five times as many children from the richest quintile complete the secondary cycle than from the poorest one (vs. about 2.5 times for primary). In Nicaragua - the least equitable country of the four in terms of 42 Calculated as the proportion of 12-21 year-old children having completed the primary cycle. At the secondary level, the 17-21year-old cohort was used. 36 secondary education - 28 times more children of the richest quintile complete the secondary cycle than of the poorest quintile (versus four times for primary)! Figure 1.35: GER, NER and GCR across quintilesinsecondary (Source: Table 1.12 Annex I) 120.0 These findings on El 100.0 Salvador and Nicaragua provide us with the 80.0 interesting insight that, 60.0 although Nicaragua does almost as well as El 40.0 Salvador when it comes to secondary coverage, this 20.0 expansion seems to have 0.0 occurred in a much more GER NER GCR GER NER GCR GER NER GCR GER NER ~ G C R I I I I I I I "elitist" way - i.e. El Salvador Guatemala Honduras Nicaragua benefiting the higher socio- economic brackets much more than the lower ones (see Figure 1.36). Part of the explanation may simply lie in the fact that secondary education expansion has been quite recent in Nicaragua and, therefore, that the current 17-21 year-old cohort had less access than the same cohort in El Salvador to the secondary cycle. Or, still, a lower completion rate of the primary level (particularly visible when we compare the completion rates of the 18 year old cohort - see above) suggests that in Nicaragua the expansion of secondary started from a smaller basis. Further exploration of this interesting difference between the two countries will be needed. While secondary coverage appears to be relatively more equitable inEl Salvador, a comparison with Costa Rica indicates that the lowest SES quintile still has insufficient access to secondary education (Figure 1.5 inthe Annex). Figure 1.36: Average secondary GER and GER by quintile inEl Salvador and Nicaragua (Source: Table 1.12 Annex I) I120.0 j 100.0 80 0 60 0 40.0 20 0 0 0 El Salvador Nicaragua 1 1 G E R +-Aver GER/ 37 Finally, the survival plots shown in Figure 1.37 reveal that while there may be nationwide improvements in student survival and drop out rates, these educational outcomes are strongly related to student socio-economic background. Again, these inequities are greater at the secondary level than at the primary level. We note that the distance among plots i s smaller for El Salvador, indicating less inequity, and larger for Guatemala and Nicaragua, indicating more inequity. Our picture o f substantial inequity at the secondary level in Nicaragua i s confirmed here. Policy implications of these findings for the expansion of secondary education, including the issue of trade-offs between primary and secondary, will be discussed in Chapter III. Figure 137: Survivor Plots by quintile El SALVADOR 2002: SURVIVOR FUNCTION PLOT by Quintiles 'G 101- 1 0 - 0.9- ............. o,8. .. ............ 1 ........... ', - - - - - -...... ,. ...... 0.7; - - . ---- s . ____. : L..... ............ $50.6: 1 I +...... .... ., ....... 8.. .... 0.5 5E 0.5{- i . . . . . . 0.4 $ 0.4: . . 1 : I. ..... L.......... 0.3 0.3: LL . . . . . . . . 0.2 : 0.1 4 0.1 : 0.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . , . . . . , . . . . o . o \ , , , , , , , , , , , , . , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 5 quanlilesof ingpe - 1 .......YEW2 3 - - - 4 5 HONDURAS 2002: SURVNOR FUNCTION PLOT by quintiles ......... .,............. ...................... ......... . . . . . . ., ........ - 1 ...................... 1.. ... - - - - - ; L---,L. .... . .I !__._ 1 , ......: . , ............. ,...... $E0.5: 8.. - ...... ...... 04: ...... ...... 1.................. ' 0.4- ------L --L 0.31 ..-. ...... 1 .......... 1 02: ....... 0.1j 0.1 0 0 1 jJ , , ,, 0.0 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 0 1 2 3 4 5 8 7 8 9 1 0 1 1 1 2 1 3 1 4 1 0 1 2 3 4 5 6 7 8 9 10 11 YEDU ScnoaiNoATTAINMENT INGBADESCOMRETED 5 quantiles01 INGPC - 2 3 - 4 5 NTlLESolCONSUPPC __ ~ I 2 ....... 3 ....... 4 5 Source: El Salvador: EPHM2002; Guatemala, ENCV 2000; Honduras, EHPM2002; Nicaragua, EMNV 2001. 38 Inequity in repetition. It is also clear that repetition, in particular in primary, i s much higher for the poor (Figure 1.38). This fact confirms our finding that low educational attainment i s also related to high age-by-grade distortion. Inequity in repetition mirrors strong inequities in educational achievement that will be exemplified in Chapter 11. Figure 1.38: Repetition by quintile I1 6 1 1 14 12 10 8 6 4 2 0 El Salvador Guatemala Nicaragua Source: Table 1.12Annex I Unequal rates of return disfavor the poor. The analysis of private rates of return presented earlier estimated average private returns to schooling. But average returns do not represent the impact of education on all workers' earnings. In fact, in all four Central American countries private returns to schooling are higher for wealthier students than they are for poor students, suggesting, among other factors, lower learning levels. Figure 1.39: Private returns to schooling by earning quintile (Source: Table 1.14 Annex I) This is the result of two types l41 I o f observations. First, marginal returns to education are higher for workers in the upper quintile o f the conditional earning distrib- ution than they are for the lower quintile in all countries (Figure 1.39 and Table 1.14 in Annex I). there are a Second, number of possible contributing factors that explain this difference (quality 39 of education, labour market connections, etc) and it i s likely that their distribution i s biased against low income household^^^. In particular, it i s likely that the education receivedby the upper income quintiles i s of higher quality than that receivedby the lower income quintiles, that cultural capital passed through families and schools favours upper income groups in the labour market, and that upper income groups have more access to highpayingjobs through personal connections. This finding points to a magnification of the impact of education on income inequity: poorer people do not only have lower levels of educational attainment but they also seem to be reaping less benefits than wealthier people for a similar level of attainment. These lower private rates of return for the poor are likely to affect the incentives of many poor people. For example, lower private rates of return in primary for the poor in Nicaragua (shown in Table 1.14 in Annex I) could well contribute to explain why the completion rate of the primary cycle i s so unequal, leading then also to high inequity in secondary coverage. Similarly, in Guatemala, the low private returns to lower secondary for the poor (Table 1.14 in Annex I)may explain why secondary coverage i s so unequally distributed there. The fact that rates of return are unequally distributed only at the upper secondary level in El Salvador (Table 1.14 in Annex I) may contribute to explain why educational coverage i s more equally distributed in that country. Figure 1.40: Evolution acrosscohorts of rates of return by quintile 16 14 12 10 8 6 4 2 0 I II18-30 31-45 464 I I 18-30 31-45 46-60 18-30 31-45 46-60 18-30 31-45 46-60 I I 1 I I I El Salvador, 200; Guatemala, 2000 Honduras, 2002 Nicaragua, 2001 Source: Table 1.14 Annex I This being said, it is good to note in Figure 1.40 that the divergence in rates of return across earning quintiles has tended to decrease over time in Nicaragua and Honduras, 43See, for instance, Arias (2004). 40 pointing to a reduction in the "un-equalizing" impact of education (although it still remains strong inNicaragua). Trends are less clear-cut in Guatemala and El Salvador. Equity has generally improved. In many of the indicators we have just examined there has been an equity-enhancing convergence of outcome indicators across SES quintiles over time at the primary education level (see Tables 1.12 and 1.13 Annex I). This evolution i s largely due to faster-paced improvements in rural areas than in urban ones. The evolution of primary NER across quintiles for El Salvador and Nicaragua i s quite illustrative of this convergent trend, pointing to an equalizing impact of the increased enrollment in rural areas. The evolution of primary NER and PCR over time can be seen in Figures 1.6 and 1.7 in the Annex. The trend was actually the opposite - although weakly - in Honduras, where the bottom quintile lost ground between 1995 and 2002 in primary NER. Finally, on a longer time horizon, we also see that the enrollment push in rural areas that occurred in all countries, although not necessarily at the same time, combined with generally improving social and economic conditions in the 1990s, was particularly fruitful in ensuring a more democratic distribution of educational opportunities at the primary level. This is illustrated by the more equal pattern of completion of the primary cycle across quintiles for the 18 year cohort, which is very much the product of the 1990s in all countries, than for the 25 and 35 year-old cohorts, whose schooling life was disrupted by economic recessions, social disorder and military conflict. The improvements were particularly noticeable in El Salvador and Guatemala over the 1990s and less clear in the case of Nicaragua (where they seem to have happened after). With the exception of Nicaragua, the improvements in the distribution of educational opportunities were less clear-cut between the 25 year-old cohort (product of the second part of the 1980s) and the 35 year-old cohort. c.EDUCATIONAL OUTCOMESAND GENDER Figure 1.41: Primary Completion Rate by gender 70 0 60 0 50 0 40 0 30 0 20 0 - 1 0 0 I 0 0 El Salvador Guatemala Honduras Nicaragua I m m a l e I ~ -female (Source: Table 1.15 Annex I) 41 Guatemala i s the only country where there i s a marked gap in educational performance between boys and girls. In all other countries girls do as well, or better, than boys (see Table 1.15 in Annex I). In Guatemala, however, females have much lower educational attainment (4.9 years versus 6.2 year for males), lower GER and NER in primary and secondary, and much lower completion rates in primary and secondary (see Figure 1.41). Repetition, age-by-grade distortion and academic performance are similar across the two genders in Guatemala. Lower enrollment, completion and attainment outcomes coupled with equitably distributed educational efficiency indicators suggest that there are cultural barriers to girls education, namely that families are less likely to enroll their daughters in school and are more likely to pull them out of school regardless o f their performance in school. Inthe three other countries, it is the low performanceof boys at all levels that seems to be more of a concern. Lack of interest for schooling, high repetition rates and need to work are all reasons o f low completion that may be prevalent in males. D.EDUCATIONAL OUTCOMES AND INDIGENOUSAND AFRO-DESCENDENT POPULATION We conclude this section on equity by assessing if there are noticeable differences in some educational indicators across specific ethnic groups. In Central America, we can distinguish three groups o f countries according to the relative importance o f their indigenous and afro-descendent population: (a) Guatemala with over half o f its population, which can be identified as indigenous or afro-descendent; (b) Honduras with about 15 percent of its population which can be considered indigenous or afro- descendant; and (c) a third group of countries, composed of El Salvador, Nicaragua and Costa Rica, with between 1-7 percent indigenous or afro-descendent population. Figures 1.42 and 1.43 below as well as Table 1.16 and Table 1.17 in the Annex present key educational indicators for the indigenous and afro-descendent populations of Central America. Official MINED data generally do not report separate information for these groups making it difficult to present definitive information on educational outcome differences across ethnicity. In the future, MINEDs should separate their data along these lines.44 Lower educational outcomes for indigenous populations. Indigenous students have lower educational attainment than non-indigenous students. In particular, differences are stark in terms of primary completion rates and are even starker for secondary completion rates. Differences are confirmed for educational achievement in Guatemala and Honduras. It i s also clear, however, that there are a few differences across countries, with a smaller overall gap between indigenous and non-indigenous in Nicaragua than in Guatemala and Honduras. It would be interesting to explore why this i s the case, particularly in view of the fairly developed/established Guatemalan policies in multiculturalism and bilingual education (see Chapter II), it i s hoped are which supporting the learning needs o f indigenous students in that country. The more recent 44Although this is changing in Honduras with the new 2003 school mapping which is explicitly collecting and reportinginformationfor all the indigenous groups (Mosquitas, Mayas, Lencas, etc). 42 bilingual education strategy may be particularly effective in Nicaragua (see Chapter 11)or policies to improve the performance of indigenous populations easier to apply in countries with much smaller shares of indigenous populations. Importantly, Nicaragua has four main ethno-linguistic populations while Guatemala has between 24 and 52 separateethno-linguistic groups. Figure1.42: PrimaryGER,NER and GCRfor indigenous Figure 1.43: SecondaryGER,NER and GCR for andnon-indigenous indigenousand non-indigenous "1 - 100 80 60 40 a I I 0 GER I NER IGCR GER INER IGCR GER I NER I GCR GER INER IGCR GER 1 NER IGCR Honduras Guatemala Honduras Nicaragua Source: Table 1.16 Annex I Nicaragua and Guatemala show limited progress toward rectifying existing educational inequities for indigenous students over the past decade. In Nicaragua indigenous PCR plummeted from 67 to 51 percent between the 35 and 25 year-old cohorts before bouncing back to 64 percent for the 18 year-old cohort. In Guatemala indigenous PCR has climbed steadily but still only reaches 37 percent while in Honduras, indigenous students made significant progress between the 35 and 25 year-old cohort but then lost two percentagepoints between the 25 year-old and 18 year-old cohorts. 48 percent of the 18 year-old cohort has completed primary school (see Table 1.17 and Figure 1.8 in the Annex). VI. Conclusions Important challenges lie ahead. This chapter has shown that educational attainment, primary and secondary enrollment rates, internal efficiency of education systems, and the gap between urban-rural areas and socio-economic groups (in primary), have generally improved in recent decades in Central America. It also showed, however, that important challenges remain in all areas in Central American countries. In particular: (a) primary completion rates are troublingly low in most of the countries under analysis; (b) there is a serious secondary education enrollment gap in all countries, as evidenced by low 43 transition to andor completion of this cycle; (c) repetition i s still too high in primary (in particular in grade 1); (d) quality i s still unsatisfactory at all levels; and (e) inequity between urban-rural, socio-economic and other population sub-groups remains substantial. Challenges vary somewhat across countries. This general conclusion varies somewhat across countries. We present below some key findings by country: El Salvador: El Salvador performed well over the 1990s,consolidating and improving its results in most areas. The country has better educational indicators than all other countries in terms of timely entrance to primary, primary completion rate, transition to secondary education, secondary enrollment rate, urban-rural gap and socio-economic gap. Its regular survival rate curve, showing a gradual decrease in attainment but no sharp transition rate to grade 7, i s particularly interesting and suggests that the country's basic education reform has been successful. However, repetition i s still very high in grade 1 and the quality of the teachinglearning process i s clearly insufficient, raising the need for urgent interventions at this level. The transition rate to upper secondary is still low and there are persistent gaps, in particular at the secondary level, between urban-rural areas and socio-economic strata, which make it also urgent to address the issue of broad secondary coverage. Additionally, preliminary evidence indicates that primary enrollment rates are falling in marginal urban areas. Finally, it i s important to keep in mind that while El Salvador compares favorably to the other three countries examined in this report, it does not compare well with other Latin American countries. Costa Rica, for example, has much higher educational indicators. Continuous effort is, therefore, still necessary. Guatemala: In spite of clear progress over the 1980s and 1990s (not always adequately measured due to the lack o f good quality comparable information), Guatemala i s at the other extreme of the spectrum, with the highest age-by-grade distortion, the lowest primary completion rate, the lowest secondary coverage (associated with both low transition and low completion of the cycle), and the worst combined (primary and secondary) socio-economic gap in terms of completion and coverage, which also reflects a persistent substantial gap between indigenous and non-indigenous population. Honduras: Overall, Honduras i s the country that has made the least progress over the past decade. Despite mediudlong-term improvements in the primary completion rate and its distribution across socio-economic groups (the two are linked), many challenges remain. The proportion of overage children i s high, late entrance i s widespread and increasing, repetition rates are high throughout the education cycle, the transition rate to secondary education i s very low, academic results are clearly insufficient and, in particular at the secondary level, the urban-rural and socio-economic gaps are persistent. Nicaragua: Finally, Nicaragua, like El Salvador, has also made substantial progress over the 1990s, in particular in terms o f timely entrance to primary, primary net enrollment rate (and its distribution across income quintiles), secondary enrollment rate, and the urban-rural gap in primary. Despite these gains the primary completion rate remains low 44 as does the distribution of primary completion across income quintiles. Repetition i s still very substantial, academic achievement unsatisfactory and the urban-rural and socio- economic gap high in secondary (the worst gap among the four countries). Three mainpriorities. The diagnostic makes it clear that the challengespresentedabove can be loosely grouped into three main priorities: (a) learning; (b) primary completion; and (c) secondary coverage. These can be further distilled to the first and third priorities (learning and secondary education covera e) considering the strong link between education quality and primary completion4' (although reaching full completion will require particular focus on the quality of the education for the poor). Addressing the grave inequities that poor, rural, indigenous, and other marginalized populations face i s fundamental to achieving all of these goals. Chapter I1will be devoted to the determinants of learning in Central America, suggesting possible options to improve it. This chapter also focuses in on specific barriers to and options for advancing quality education for disadvantaged sectors of the population, as this is necessary as well for attaining universal primary completion. Chapter I11 will deepen the analysis of the causes of low secondary enrollment, which, as foreshadowed in this chapter, are diverse, and suggest policy options to address them. The last two chapters will concentrate on two cross-cutting instruments, education spending and management which, if well used, have the potential of greatly improving education delivery, both quality and coverage. 45 Illustrating this point, the internal efficiency analysis suggested that over-age, caused by late entrance and/or repetition, is a leading cause of non-completion o f the primary cycle and low transition to secondary becauseof the increasing tendency to drop-out (for work or other related reasons) in children over a certain age. Similarly, low quality o f primary and lower secondary is not only important for itself, but because, by generating lack of interest in families and students, it is also likely to explain why children drop-out at some point of their schooling cycle. Hence, overall, policies to promote timely entrance or even lower the official entrance age, together with policies which decrease the impact o f repetition such as flexible promotion or accelerated education programs, may be useful; but even more important, however, will be quality improvements at all levels to improve cognitive skills, increase education relevance and education returns, and help reduce repetition. 45 Chapter II: Determinantsof (and Constraintsto) Learning in Central America With the current global emphasis on access and enrollment in school, a focus of the Millennium Development Goals, for example, there i s the risk of pushingto the side what i s perhaps the most important objective of schools - to teach children meaningful skills and knowledge. This report puts that fundamental objective back into the limelight because without adequate learning outcomes people's ability to live healthy, enjoyable lives, compete on the labor market, and contribute to national growth are severely restricted. Learning can be thought of as a result of three factors: the quality of schooling; family, and community background and contextual factors; and personal characteristics o f each student. This chapter will examine the determinants of and constraints to learning in Central America with emphasis on the first and second factors - the quality of schooling and background, or demand-side factors. Poor learning outcomes are often thought of as a problem distinct from educational attainment and access. But low levels of learning directly impact educational attainment. This occurs in three primary ways: (a) inadequate learning outcomes can result in repetition and grade failure if students do not gain sufficient skills and knowledge to pass to the next grade level; (b) the quality o f learning directly affects students' interest and enjoyment of school which contributes to their decisions regarding whether to stay in school; and (c) the quality o f learning impacts the private rates of return of education, which inturn creates incentives or disincentives for attendan~e.~~ Because o f this, learning i s key to achieving primary cycle completion as well as greater enrollment in higher levels of schooling. This i s witnessed in several phenomena in Central America including high repetition rates associated with higher dropout rates, lack o f interest in schooling as a factor for school abandonment, and lower test scores associated with lower completion rates.47 In all cases, these phenomena acutely affect disfavored segments of the population, to a large extent comprised of poor, rural, and indigenous students. This chapter will, therefore, focus on the reasons behind poor learning outcomes in Central America, at all levels, with emphasis on the learning of the poor. As discussed previously, this challenge i s identified as the single most important priority for educational reform in Central America. This priority i s increasingly widely espoused by countries and development stakeholders. General quality improvements and focus on quality of delivery for the poor are the focus of the recent Education For All Fast Track Initiatives implemented in both Honduras4*and Nicaragua.49 46The link between quality as well as relevance and secondary education coverage will be further explored in the next chapter on the expansion of secondary education. 47See, for instance, Marshall (2004b) on Guatemala. 48 EFA-FTI in Honduras focuses on educational efficiency improvements, over-age children, teacher management reform, and education for indigenous peoples. 49InNicaragua, EFA-FTIfocuses on performance-based delivery and school management. 46 This chapter begins with a discussion of the national assessment systems in the four Central American countries and how to improve them. We begin with this discussion because we argue that effective national assessment systems are key to understanding the learning challenges that confront each country and that improvingthese systems will help countries plan methods to tackle these challenges. The remainder of the chapter is then divided into two sections. The first examines some of the main determinants of and constraints to learning in Central America while the secondfocuses on policy options that respond to these determinants and constraints to learning. In both sections three main areas are investigated. These are: (1) student background characteristics, (2) teacher quality, and (3) curriculum and pedagogical practices. In the second section, we begin with an analysis of what differentiates high-performing low-income schools from low- and average-performing low-income schools as a means of introducing some of the subsequent policy recommendations. The conclusion of the chapter reviews the main challenges and policy recommendations put forward in the chapter towards improving learning inCentral America. 1. A Review of Central American National EducationalAssessment Systems Improving student learning and completion will depend on ministries of education and communities having clear and accurate information on the strengths and weaknesses of education systems. This section, therefore, discusses the state of national assessment systems in Central America and proposes some ideas for strengthening evaluation and assessment in the region. Table 4.1 in Annex IV summarizes some of the main characteristics of these systems. National assessment systems generally have two primary roles, (a) a diagnostic and action role: to provide inputs to specific audiences (education authorities, school directors, teachers, parents) for diagnostic and policy refondcorrection purposes; and (b) an accountability role: to create accountability mechanisms for education authorities, schools and teacher^.^' To a lesser extent, some national systems also have a formal student accreditation role in which assessments are used to award certificates or diplomas to students based on their performance. The primary role of the Central American national educational assessment systems is diagnostic rather than accountability. This focus on diagnosis presents both positive and negative implications that are discussedin other research.51 The different roles of national assessment systems necessitate distinct assessment design and dissemination practices. For diagnostic purposes, teachers should be provided with detailed and accurate information on their teaching practices and their students' learning outcomes. For accountability purposes, information should be available to a wider 50This dual objective is also developed in Ravela (2002). 51See Ravela (2002) for a discusion along these lines. Essentially, the argument is that the diagnostic role of national assessment systems is very useful and should be the primary objective of an assessment system but, without accountability for results, there is no security that the diagnostic will trigger changes. On the other hand, using the assessment system to evaluate schools and teachers in a "strong way", creating clear accountability, could be seen as a threat and de-naturalize the primary objective of the evaluation. 47 audience and should be linked to clearly defined repercussions (awards and/or sanctions) used to promote behavioral changes in specific groups of actors. Heterogeneity across systems. While all four of the countries covered by this study have national assessment systems, their institutional history and integration differ considerably. El Salvador and Costa Rica have systems that date to the late 1980s/early 1990s and have evaluation departments fully integrated within their ministries of education. Nicaragua, Honduras and Guatemala have begun national assessment more recently and have lower levels of institutional i n t e g r a t i ~ n . ~ ~ Standardized testing. All countries undertake standardized testing at different levels of the educational cycle (see Chapter I,Table 1.3 for more details) and accompany these assessments with analyses of school, teacher, and background characteristics that are associated with higher and lower achievement. The implementation of standardized testing for assessment purposes i s a major development in Central America in the past decade. Lack of dissemination. Despite recent advances in standardized testing, a weak point of all the countries' assessment systems i s the insufficient dissemination o f results on both tests and associated learning determinants. Limited dissemination greatly limits the possibility o f these assessments having an impact on educational quality (as also shown in the last column of the table). Inparticular, it is striking that even in Honduras, where building accountability is among the chief objectives of the assessment system, results have not been widely disseminated. Typically, a limited number o f reports are printed and the report i s posted on the Internet, but that i s the extent of dissemination. Perhaps most importantly, the results of the tests are not fully and systematically shared with the schools that took part in the testing in most countries. Misalignment with curricula and standards. There is also a lack of timely alignment between tests, standards, texts and curriculum that characterizes the education systems in Central America. This, again, limits the potential o f national assessment to contribute to improvements in teaching practices, curricula, and teacher motivation. Tests are often adjusted only much after curricular changes or changes in standards have been introduced, limitingtheir effectiveness inproviding feedback. An example i s El Salvador, where tests are still not fully reflecting the new standards introduced in 2000.53The situation in Honduras i s similar. Potential for regional collaboration. Up until recently, educational standards differed across each of the Central American countries. Now that regional Central American standards have been developed at both the primary and secondary level, there i s a possibility of developing a regional assessment and/or accreditation system that could 52A current debate is, for instance, under way on the sustainability of the UMCE in Honduras. The feasibility of creating a specialized institution with its own legal framework, organization and responsibilities is under analysis, as is the one of fully absorbing UMCE within the structure of the MINED. j3See Rapalo (2004). 48 greatly expand the potential impact of as~essment.~~Regional accreditation systems could even be used to facilitate students' movement across studies for educational opportunitie~.~~ The costs andbenefits of these different possibilities should be assessed. In sum, while there have been major strides in the creation of national testing, assessment, and evaluation systems in the past decade or two, these systems for the most part lack sufficient institutionalization and as of yet have only had a limited impact on improvingthe quality of education and learning in Central America.56With support, these evaluation and assessment systems will become a basis for identifying challenges and improving education. In confronting barriers to learning, Central American nations should seriously address the needfor more and better assessment. II. Existing Evidenceon Determinantsof Learning in Central America The following section explores some of the most important factors contributingto student learning. For this section, a review of studies of factors associated with student achievement in Central America (including those mentioned above conducted by national assessment systems) was performed.57Findingsfrom these studies were compared across countries and compared to Latin American and other international meta-analysesof such education production function ~tudies.~'In this section we discuss the main factors that were found to limit or enhance student learning in the region and offer some diagnostic informationon each. First, however, a few caveats are necessary. The findings from these Central American studies are not conclusive; there are a number of limitations to and weaknesses of education production function studies, particularly in developing countries. One such example is that education production function analyses face substantial methodological limitations, which are particularly well illustrated in a recent article of Glewwe (2002).59 Secondly, these studies look at how different factors are related to student achievement on standardized exams, but standardized exams are themselves an incomplete and arguably, problematic, measure of learning and education quality. Using test scores as a proxy for learning outcomes is a common practice, but many researchers and theorists 54A Central American Assessment Institute is currently under discussion. 55A system of education title equivalence is already being coordinated by the CECC for Central America, but this effort would be greatly facilitated ifcounty standards were more aligned on regional standards See PREAL (2003) for a similar conclusion. 57See Umansky (2004). Comparison studies included Fuller and Clarke, 1994; Hanushek, 1995; and Velez, Shiefelbein and Valenzuela, 1994 as well as the Latin American Laboratory UNESCOLLECE, 2000. 59We will not review these limitations here, but maybe the most important one is related to the many variables that can affect learning and are not observed by the analyst. If some of these variables are correlated with some of the included variables, the impact o f the included variables may be biased and inaccurate. Several methodological remedies have been proposed to address this issue, but none completely resolves the problem. In particular, student, family and community background variables are often insufficiently controlled leading to over or under-estimations o f the impact o f school-related factors. Additionally, there are school-related factors such as teachers' motivation, characteristics of teaching practices and directors' leadership that are typically not well captured in these types o f studies. 49 argue that it may be incomplete, inaccurate, and unreliable as a proxy. A test measures only very limited areas o f knowledge being a sample of certain areas of learning which in turn only comprise certain areas of educational outcomes. Objectives for educational outcomes are much broader, comprising not only wide subject matter knowledge, but also social, emotional, creative, physical, and intellectual development. Tests have also been foundto be vulnerable to corruption, inflation, and measurement error.6o Lastly, the studies undertaken on each of the countries under analysis differ substantially in the quality of their assessment. The El Salvador study uses more than one method of multiple regression to triangulate findings for example, while the Nicaragua study relies only on Pearson Correlation o f a very limited number of variables without controls.61 Nonetheless, the studies provide initial evidence on the importance of student background, teacher, curricular/pedagogical, and school management factors which are developed in this section. A. STUDENT BACKGROUND CONSTRAINTSTO LEARNINGCENTRAL IN AMERICA Student background constraints to learning have been recognized as critical barriers to educational effectiveness for several decades by policy-makers and academics, and since time immemorial by teachers and principals. The seminal work, A Nution ut Risk, turned researchers' and policymakers' attention to the fact that students' families, communities, and backgrounds played decisive roles in students' educational outcomes. Studies in Central America corroborate these findings. Children from poor, ethnic minorities, and less educated families and communities all tend to do significantly poorer on exams than their more advantaged fellow students. Table 2.1 below provides some initial evidence on demand-side constraints to learning. The table synthesizes the findings of the recent Central American education production function studies, discussed above, in terms of how student, family, and community- related characteristics impact and are related to learning. In all studies learning was measured by student achievement on standardized exams given at the primary level. The table also presents the findings from UNESCO's regional Latin American Laboratory study as a point of comparison. While not investigated in these studies, many of these same background factors have been linked to secondary school learning achievement, as well as to other educational indicators such as enrollment, entry-age and attainment. 6o These and other critiques of the use of test scores as a primary means of measuring learning are raised in papers such as Murnane and Cohen (1986), Koretz (2002), Kane & Staiger (2001), and Jacob & Levitt (2002). 61See Umansky (2004). 50 merica an 9 I Familv recommendations for out of school I I 1 E time Student Characteristics 2 Understandslittle 1 G 3 Perceivesfavorable learning environment** 0 Pos 51 Notably, there i s more consistency in the findings regarding these student background determinants of learning than for any other group of determinants (teacher, school, pedagogy, etc) underscoring that student background has particularly strong and clear effects on learning. The table shows that both student's family and community characteristics as well as personal attributes impact learning. In the area of family and community background characteristics there are clear economic, social, and cultural constraints to learning. Students from poorer backgrounds have inferior performance on exams. The economic factors show that at the family, community, and even state level, lower income and increased poverty are associated with inferior test results. There are multiple reasons for this. Poor families are more likely to lack basic resources that help children learn. Lack o f a clean, quiet, well-lighted space at home prevents children from doing their homework effectively. Poverty frequently results in malnutrition and hunger, which affect children's mental capacity, concentration, and interest in school. A recent study b y the Pan-American Health Organization reports dangerously high levels o f severe Fowth retardation in first-graders in Central America due to chronic malnutrition.6 Poor students are more likely to suffer psychologically at school if they are teased because of their tattered uniform or castigated for not paying voluntary school fees. Parental education level and literacy are strongly related to student achievement. In all four countries measures of parental literacy and education are positively related to student outcomes. Educated, literate parents can help their children with homework and explain academic questions and doubts more easily than non-educated or illiterate parents. The cultural capital (values, vocabulary, knowledge, etc) o f parents is much more likely to be harmonious with the cultural capital taught in schools if a student's parents were also educated in schools. This shared culture makes it much easier for children to learn because they have to learn one way of life rather than two. Parents who did not attend school are more likely to have very different practices and values than those the child experiences in school, a situation which makes adapting and learning the ways of either system more difficult.63 Finally, educated and literate parents are much more likely to feel comfortable talking with teachers and principals about how their child i s doing and working with them to support their child's learning than parents who, not educated themselves, feel uncomfortable or unwelcome in schools. Social and cultural characteristics also influence learning. Several of the factors examined in the Central American and UNESCO studies find that social and cultural attributes and practices of families and communities are also related to educational achievement in predictable ways. Whether families have computers, and the number o f books and overall belongings at home appear to be positively related to achievement in studies in El Salvador and Guatemala (although these factors are not statistically significant in all studies). Relatedly, UNESCO finds that parents that read to their children tend to have children with higher test score results. Ethnicity also seems to be 62 From PAHO (2004). Severe growth retardation is shown to be particularly acute in Honduras (15.2 percent) and Guatemala (14.5 percent) and less so in El Salvador (3.1 percent). 63See the work of Pierre Bourdieu on cultural capital. 52 important. Studies for Guatemala and Honduras indicate that students who speak an indigenous or non-Spanish languageor who live in areas where more people do not speak Spanish tend to do poorer on exams. Recent studies of indigenous poverty in the Americas also corroborates that once controlling for other background factors, indigenous students are less likely to be enrolled in These background characteristics not only create barriers to learning but also may relate to decreased demand for quality schooling. Many if not all of these factors create barriers to learning, and, therefore, also barriers to regular attendance, school attainment and completion. It is also likely that in addition to creating actual barriers, many of these family and student background characteristics are associated with decreased demand for schooling (because of lack of interest in schooling, low relevance of the schooling process or high direct and opportunity costs). Children from poorer families contribute relatively more to family income than those from wealthier families. The opportunity cost of having a child in school is therefore higher in poor families. Add to this the fact that poor children usually attend lower quality schools further decreases motivation to keep children enrolled, attending, and succeeding in school. Similarly, some indigenous families may have lower demand for quality schooling. If families do not think the education their child will receive is relevant to or respectful of their culture, history, and values, they may be less inclined to enroll or support their children in school. In sum, demand-side constraints are predominantly a barrier to learning for disadvantaged sectors of the population. All of these findings regarding how background characteristics affect learning point to a fundamental conclusion: demand- side barriers to learning almost exclusively relate to marginalized segments of the population. It is important to note that this does not mean that in all cases it i s the students'families or communities that needto change to improve their children's learning. In some instances this may be true; for example, malnutrition is a terrible and avoidable constraint to learning. In other cases the opposite may be true. Indigenous students may have a harder time succeedingin school, and an appropriate policy responseis not to strip them of their indigenous identity but rather transform schools such that indigenous studentscan and do thrive intheir studies. B.TEACHER QUALITY Teachers are vitally important to education and the learning outcomes of students. This i s intuitively obvious but also increasingly demonstrated in academic s~holarship.~~ Nonetheless, teacher quality is by itself a difficult variable to measure, as are the determinants of teacher quality. It is generally considered that teacher quality should include both a teacher's capacity to effectively transmit the curriculum to the students and the will, effort and continuity to do so. The capacity to teach (difficult to measure) is often consideredto be related to the level and quality of the initial education received, the See various country-specific background papers for the Indigenous People, Poverty, and Human DevelopmentinLatinAmerica, 1994-2004study (World Bank, forthcoming). 65See, for example, Rivkin, Hanushek, and Kain (1998), Park and Hannum (2003), or Wright, Horn, and Sanders (1997). 53 quality o f the in-service support and training, the existence o f incentives to attract and retain talented and skilled individuals to the teaching profession, and the amount of experience a teacher has in the classroom. Teacher effort and motivation (also difficult to measure) i s frequently found to be related to performance incentives such as evaluation from direct supervision systems, salary bonuses, or hiring and firing authority by some local actor or actors (such as a parent council).66 Clearly, however, there are other factors that contribute to teaching capacity and teacher effort, many of which are very difficult to measure or even identify.67 B y and large, findings on the relationships between teacher characteristics and student outcomes in the Central American countries are less clear-cut than in larger meta- analyses. This could be due to data or analysis limitations as discussed above or reflect real contextual differences in these countries, such as particularly low quality pre-service or in-service educatiodtraining. Following the structure of Table 2.2 we will start by reviewing some basic teachers' characteristics which influence teaching capacity, and then move on to more complex dimensions, related to incentives, which influence capacity, effort and motivation. a) Teacher Capacity: Education, Professional Development, and Experience i)Education Level Teacher education i s the teacher characteristic most consistently related to achievement across the studies under analysis, although this impact is less strong than the one detected in many meta-analyses. The weaker impact of teacher education in Central America may not mean that teacher education does not matter but rather it may point to low quality teacher preparation in the region, or teacher education that does not adequately prepare teachers to work with disadvantaged student populations. Relatedly, teacher subject knowledge i s consistently found to be positively related to educational achievement in the Guatemalan and Honduran study as well as in numerous other studies. 66 This last type of incentives will be more typically associated with decentralized education systems where an increasing role is given to the creation of "external" accountability relationships, by empowering parents and providing them with the information needed to enforce accountability. 67 Benabou and Tirole (2003) and Villegas-Reimers and Reimers (1996) write about the intrinsic motivation of teachers. 54 Table 2.2: Summary Findings on Teacher Related Determinants of Education Quality in Central America and the Latin American Laboratory Study (N= Nicaragua, E= ElSalvador, H= Honduras, G = Guatemala) Num' Of Mixed** Significant Not UNESCO Studies Positive Negative LLECE IBasic Teacher Characteristics 1. Educationlevel 3 N G* H Pos 2. Training 1 E NS 3. Bilingualed training 1 G 4. Experience 4 H N H, E NS 5. Age 2 N E 6. Gender (female) 3 N H E ~~ 17. Subject knowledge 1. Temporarycontract 1 H 2. Morethan one job worked 1 E Neg 3. Satisfactionwith salary 1 E Pos 4. Satisfactionwith school 1 E 5. Experienceat school 1 E 6. Attends adminlpedagogical meetings 1 E 7. Absence 2 H G 8. Mid-yearturnover 1 H ITeacherPerceDtions I 1. Attributesstudent performance to student skills/interest 1 E Pos 2. Attributes student performance to family conditions/SES 1 G* Neg 3. Attributes student performance to teaching & ed. opportunity 1 G* NS 4. Prognosisof student success 1 E 5. Values multiculturalism 1 G 6. Positiveclassroomclimate 1 E Further data on teacher education i s provided in Annex V. Table 5.1 in Annex V provides an institutional description of pre-service and in-service teacher training, which can help us understand the structure, requirements and contents of teacher education programs in 55 Central America. Table 2.3 shows the education level of a sample o f teachers using household survey data in four Central American countries. This analysis i s only illustrative as household surveys are not representative of the full teacher population so care needs to be taken in interpreting the results too literally, although formal data collected by the MINED tend to provide similar evidence on educational levels. In most of the four Central American countries, national guidelines require that primary school teachers have completed at least secondary or teacher-training school (escuela normal) while secondary school teachers are required to have studied at the tertiary level (see Table 5.1 inAnnex V). El Salvador has the highest teacher preparation requirements. In El Salvador, all teachers, both primary and secondary, need to be certified from a higher education institution. Inpractice, these requirements are met to varying degrees in each of the countries, however. Teacher education level could be better assessedwere we able to divide primary and secondary teachers but, unfortunately, we do not have this information disaggregated by school level. Additionally, the comparison between countries i s made a little difficult by the fact that in Nicaragua, teacher education is presented as a separate education level in the household survey, while in Honduras and El Salvador it i s included either in upper secondary (Honduras) or in tertiary technical (El Salvador). F a b l e 2.3: Highest level of education attained by teachers in Central American countries I Sources: The World Bank (2004b,c);Arcia (2003); Guatemala, 2000 ENCOVI HouseholdSurvey. Notes: (a) Including teacher school; (b) Includestechnical and universitytertiary. Gap between required and actual educational attainment. The difference in formal teacher education prerequisites i s noteworthy (see Table 5.1) as i s the strikingdifferences between countries in terms of the degree to which the formal teaching requirements are applied. In El Salvador, only 12 percent o f the teachers teach without the minimum requirement of a university degree. In Nicaragua, by contrast, 55 percent of active teachers have not attended university or teacher training school. In Guatemala and Honduras it i s not possible to judge what precise percentage of teachers do not have the proper requirements because we can not disentangle upper secondary from teacher training institutes (but these fractions are probably about 45-50% and 30-35% respectively). In Honduras and Nicaragua, due to limited access to teacher training 56 schools and insufficient supply of teachers in rural areas, highnumbers of under-qualified teachers are employed in the classroom. In Honduras, a substantially higher fraction o f rural teachers have less than the required teacher qualifications than urban teachers (see Figure 2.1). This difference i s much smaller in El Salvador. Although allowing a certain flexibility may be necessary to ensure higher teacher coverage in rural areas, it is clear that this i s a sub-optimal situation, particularly because the higher poverty levels and lower educational performance in rural areas necessitate, if anything, higher-qualified teachers rather than lower-qualified teachers. Figure 2.1: Honduras: Distribution of teachers across It should be emphasized that the urban and rural areas by level of education attained level of qualifications is not the (Source: The World Bank (2004~)) only aspect which is relevant to teach in rural areas, experience, 12001 I motivation, quality of training, sensitivity to cultural difference, etc, are all factors that are also important. This point will be made clear in the effective school analysis for Nicaragua and Honduras. In any case, an effective policy to attract and retain competent teachers in rural areas is needed in Honduras, Nicaragua and Guatemala (some steps have begun to be taken in that direction). Heterogeneous educational attainment. The four countries vary considerably in terms of the educational attainment of teachers. El Salvador has the highest educated teachers whereas Nicaragua has the lowest educated teachers, on average. Overall, about 90 percent of the teachers have tertiary level education in El Salvador, while this proportion falls to 52 percent in Honduras and only 29 percent in Guatemala and 13.5 percent in Nicaragua. Teacher entrance exam requirements. El Salvador i s the only country that has implemented an explicit policy to enhance teacher selection criteria. Since 2001, prospective teachers must meet a minimum cut off point on the secondary exit exam in order to study education. In addition, candidates must also, since 2001, pass a special accreditation exam called the ECAP ("Evaluacidn de Competencias Acadkmicas y Pedagdgicas") at the end of the third year of teacher training This is a positive step in the direction of higher teacher quality, which other countries may want to consider. To a certain extent the possibility o f teacher requirements is facilitated by an excess supply of teachers in El Salvador, also reported in other Central American countries (although less so in rural areas). Exams such as the ECAPare a very valuable diagnostic tool to assessthe quality of teacher education. 57 Low quality of teacher education. For the most part, teacher preparation is of very low quality in Central America. InEl Salvador, where as we have seen, teachers are arguable better prepared than the other countries, only about 40 percent of eligible students passed the ECAP exam in 2001. Similarly, in Guatemala, results of the achievement tests of teacher-training students in teacher training institutions carried out in 2000 by the PRONERE program show that the mean test scores obtained by students were only 28.5 for math and 54.3 for reading, out of a possible 100points, another indication of the low effectiveness of teacher training.69 The quality and relevance o f teacher education in the region i s an extremely important challenge to confront. Disconnect between national teacher needs and teacher education. A gap between demand and supply o f teacher specializations frequently leads to teachers working in positions they were not prepared for and can further diminish teaching quality. In El Salvador, in the schools/universities where teachers are formed, a number of specialties are available: pre-primary education, primary education, and several different areas of secondary education. There i s evidence that there i s a certain discrepancy between demand and supply in terms of the specializations pursued both with respect to the educational level considered and to the specific subject^.^' Pre-primary (26 percent) and social sciences (18 percent) appear to prevail, while primary education (8 percent) and secondary education (7 percent) are less sought after. In Guatemala, teacher education at the teacher school level has 12different programs; the most attended being urban primary teacher education (60 percent), followed by pre-primary (14 percent), and rural primary (7 percent).71It i s striking that such a low share of teachers gets trained in rural primary (which includes special training in multi-grade and bilingual teaching) in a country where 66 percent of the primary enrollment i s in rural areas.72This could possibly explain why teacher education has an ambiguous impact in Guatemala. ii) On-going Professional Development (or In-service Training) Teacher training i s another variable that has been traditionally analyzed in educational achievement studies. It was also tested in three of the studies under analysis (see Table 2.2). The mixed evidence concerning in-service training and student achievement found in the meta-analyses is confirmed in the Central American context. In-service professional development i s generally o f very spotty quality in much o f the developing world, a fact that may account for its mixed and non-significant effects. Overall, on-going professional development seems to have improved in all Central American countries over the last decade. Box 2.1 summarizes some characteristics of teacher training in all countries. In particular, the systems have become more structured, while at the same time also being more flexible and decentralized to increase their responsiveness to different training needs. Flexibility refers, for instance, to the offer of in-site and distance education programs. Decentralization o f professional development to 69 See GuatemalaGC. `O See, notably, PREAL (2002a). '*This See The World Bank(2004a). point is also taken by PREAL (2003). 58 regional and local bodies, school-based delivery, and teacher reflective practice grou s and teacher centers have been a main thrust o f in-service training in recent years p3 . Increasingly, professional development has also focused more on participatory methodologies and classroom innovation. In-service professional development i s being actively used in Honduras and Nicaragua to provide certification and critical skills to teachers that lack teacher preparation requirements (ducentes empiricos). icular interest are the Notwithstanding these positive developments, there are still a number of challenges ahead to improve the relevance and quality of these training programs. In particular, it i s generally highlighted that: (a) too little effort i s put on changinghpdating classroom 73A little less so in Honduras where teacher training is still quite centralized. 59 practices; (b) too little attention i s paid to non-standard education contexts (multi-cultural environments, rural areas, etc); (c) too few formal evaluations exist of the effectiveness o f professional development programs; (d) there i s a lack of follow-up on trained teachers to see if they are applying what they have learnt in the classroom; and (e) in some of the countries, training implies missing substantial time from the classroom, leaving children without their primary teacher.74 Other variables, in particular teacher experience, are frequently put forward as determinants of teaching quality and student educational achievement. However, although teacher experience i s normally found to be positively related to learning in international studies, its impact appears to be insignificant or, at best, mixed in the context of the Central American studies (although it i s a significant difference between high- and low-performing low-income schools in Nicaragua as will be discussed below). This seems to indicate that, at least for the average school, experience is less relevant to student learning than teacher qualifications and motivation. Why teacher experience does not appear to be more important in Central America requires future research. b) Teacher Effort and Motivation: Incentives to Perform i) Teacher efSort Typically, unobserved variables in educational achievement studies are teacher effort and motivation, although education production studies sometimes attempt to measure them by measuring teacher effective working hours and absences, the level of teacher satisfaction with the job, or teachers' perceptions on student success. Two Central American studies attempt to measure teachers' absences and one measures teacher satisfaction with the school, and results are generally non-significant (with one noticeable exception being that, as predictable, teachers' absence in Honduras is negatively related to student achievement). The general lack of significant findings may be due to possible measurement errors or to the fact that teachers' absences or satisfaction with the school fail to adequately proxy for teacher effort and motivation. Teachers' perceptions on student success are more systematically related to student achievement, although it i s difficult to establish a causal relationship. Where do Central American countries stand in terms o f teacher effort as measured by effective teaching hours and teachers' absences? Overall, no national reliable data are available, making it necessary to use information coming from small databases, self- reporting, and household surveys (see Table 2.4).75 74See, for instance, PREAL (2003). 75This data should not be interpreted too literally as it is subject to measurement error. For instance, all the surveys collect self-reported estimates of number o f absences and, often, corroborate this data with the schools' registers. Nonetheless, it i s very likely that only excused absences are declared or reported, which most likely underestimates actual teachers' absences. Additionally, we generally do not have comparable indicators across countries as we had for other variables. 60 Low teacher effort. It appears that teacher absenteeism is a serious problem in the countries for which we have information. In particular, in Honduras 40 percent of teachers missedone month or more of school in 2000 according the school census.76Also problematic it appears the gravity of the absenteeism problem may not be fully acknowledged in the region. Few principals in Honduras (7 percent) consider high absenteeism to be a problem in a recent school survey.77Teachers in Honduras also report working seven hours fewer than their official workweek, even including out of classroom hours worked (30 of 37).78Teachers in El Salvador and Guatemala report that they work even less than in Honduras. Average weekly work hours are significantly overestimated in that they do not correct for teachers' absences. For example in Honduras, assuming an average teacher absence of four weeks per year would lead to about 11407' effective hours, which would translate to substantially less hours in terms of teaching). We will analyze effective instructional time inthe next section. Honduras El Salvador Nicaragua Guatemala Teachers'absences Teachers'yearly absences-days (a) 16 18 Teacherswith no month away (%) (b) 60 Teachers with 1 monthaway (YO)(b) 26 I Teacherswith 2 or more months away (Yo) (b) I 14 I I I I ~ ~~ ~ Average declaredhours per week Average hours per week (c) 30 I Hours worked at school each week (d) I I I I 22 I Teachers (e) 27 Teachers (f) 27 Public sector teachers (e) 28 76To be noted that data on teacher attendance in a small rural sample (N=l32) provide a somewhat more positive evidence with, on average, about 16 days o f teacher absences per year. 77See UMCE (2003b). 78Which includes not only weekly hours teaching, but also preparing classes, meeting with parents, undertaking administrative tasks, grading, etc. 79Using the school survey figure of 30 hours per week (slightly lower than the figure of 31.3 reported in the household survey). 61 Public sector teachers (f) 27 In sum, teacher effort appears to be low in several of the countries and, therefore, improving this variable i s a key priority. To be able to make policy recommendations, we need to identify a number of key underlying factors that influence teacher effort and motivation and develop reliable means of collecting data on these indicators. ii) Salaries In Central America, we show below that when factoring in hours worked and time off, teachers earn as much, or more, than professionals with similar levels of qualifications. This implies that salary level should not cause any disincentive to work as a teacher. Secondly, we also show that there appears to be a disconnect between average salaries and teacher effort in that there are few mechanisms that link salary increases with measures of teacher effort such as effective work hours or teacherhtudent performance.80 The relationship between salary level and student achievement was not tested in the Central American studies. El Salvador did investigate whether satisfaction with salary i s related to student outcomes and found this relationship to be insignificant. Other research, has, however, investigatedthis subject more extensively.81 Average teacher salaries are on par with those of similar professions, or higher, when factoring in hours worked and time off. Table 2.5 shows comparative hourly salaries. As salary generally varies according to academic qualifications (see Box 6.1 in Annex VI), we control for different educational attainment records by comparing teacher salaries with the salaries of professionals with similar educational attainment levels. The 8o Salary level is generally pinnedto teacher education and training level and years o f experience. These are usefulproxies for teacher quality to a certain extent but they do not link with teacher effort. 81 A recent detailed survey on incentives and the labor market for teachers finds that, across different studies (covering European countries, the US, but also some developing countries), there is evidence that the level o f teacher pay relative to other professions is an important determinant of teacher quality and performance. For instance, Nickel and Quintini (2002), on the UK, find that the relative pay of male teachers has an influence on their results. Also in the European context, Dolton and Van de Klaauw (1999) find that better wages in the non-teaching professions increase the tendency among the most qualified teachers to leave the profession for a non-teaching job. In Latin America, the LLECE study also finds satisfaction with wages to have a positive impact on student achievement. Essentially, higher teacher salaries are likely to attract more talented and skilled new teachers into the profession, as well as help retain them. I t i s also found, however, that the intensity of the link between teacher pay and teacher performance depends on peculiarities of the teacher labor market and remuneration structure and, that, in many cases, the link will be only weak. In particular, teachers are generally rewarded according to experience and qualifications, without taking actual performance or teacher effort into account: in such a system higher wages may attract more skilled and hard worker new entrants and foster further qualification, but not necessarily create incentives to boost student performance or increase effort by teachers who are already practicing. To illustrate this point, some studies show a clearer impact of teacher pay on education achievement when performance is directly taken into account in the salary scale, by introducing, for instance, team-basedperformance-pay (see Glewwe and others (2003) and Lavy (2002)). 62 information comes from household surveys and, therefore, i s again purely illustrative.82 Primary teachers earn substantially more than professionals with a similar level of education in Honduras, and overall teachers earn somewhat more than similar professions in Guatemala. Teachers earn generally more than other equivalent professionals in El Salvador. Only in Nicaragua average salaries are similar or slightly lower for teachers. With the exception of Nicaragua, this tells us that teachers' salary levels are generally satisfactory within the economic framework of each of the countries. Table 2.5: Hourly salaries acrossteachers and professionswith similar educational attainment Category Honduras, El Salvador, Nicaragaua, Guatemala, 2002 2002 2001 2000 ITEACHERS I I I I I Primary teachers 2.2 0.6 Secondary teachers 2.3 1.o Public sector primaryteachers 0.5 Public sector secondarvteachers 0.5 ITeachers(a) I I 3.4 I I 2.1 I With upper secondary 1.7 0.9 1.3 With tertiary university 9.6 2.9 2.4 With tertiary non-university 2.3 0.9 Hourly salaries are likely to be even more favorable to teachers when one takes into consideration the fact that teachers, on average, work fewer weeks per year than non- teachers. In Nicaragua, if one assumes a 36-week year for teachers versus a 46-week year for non-teachers, we find that teachers are paid roughly equally to non-teachers. Re- calculating the net salary per hour on the basis o f 42 weeks per year, we find that the 82It would be more correct to compare the official MINED data with official data on other professions' salaries, on census information and for professionals of a similar level o f experience. On this last point, using the household survey, we are just assuming that the randomly selected teachers and other equivalent professionals have, on average, a similar level of experience. 63 salary per hour increases in Honduras from US$2.2 to US$2.7. Calculating in teachers' absences, we find a further increase to US$3.0 per hour, just 10 percent less than public employees with university education (see Table 2.6). The same pattern of comparable wages in El Salvador, Honduras83, and Guatemala84i s true when comparing average monthly, rather than hourly, salaries for teachers and non- teachers. Thus, if teachers in Central America do work fewer hours than non-teachers, it does not appear to be resulting in significantly lower salaries for teachers in these three countries (Nicaragua has lower monthly salaries, just as it also had generally lower hourly salaries). Table 2.6: Honduras - Salary per hour for primary teachers Net salary per hour Declared hours: 1628 hours per year (a) 2.2 Declared hours: 1314 hours per year (b) 2.7 Effective hours: 1189effective hours per year (c) 3.0 Notes: (a) calculated on the basis of the weekly hours reported in the Household Survey and 52 weeks; (b) calculatedon the basis of the weekly hours reportedin the HouseholdSurvey and 42 weeks; (c) hours reported in the Household Survey adjustedassumingan average teacher absence of 4 weeks per year. Disconnect between salaries and teacher effort. Teachers in Honduras (1997), El Salvador85(1995), and Guatemala (1996) all benefited from recent increases in the level of real salaries. In most cases, these salary adjustments followed periods of declines in real teacher salaries. Stagnant indicators such as low test scores (see Chapter I), high teacher absenteeism, and stagnant or even decreasing hours worked, do not suggest any obvious improvements in teacher quality generated from these salary improvements. At the same time there are sure to be many factors that contribute to stagnation in teacher work hours, absenteeism, and testing outcomes that may be entirely independent o f salary level. Furthermore, increases in salary levels may have resulted in positive changes not captured in these indicators or there may be a lag period before positive changes are evident. I Table 2.7: ElSalvador- Evolution of real salaries for public and private teachers IPC I Nominal Salaries I Real Salary I worked per I Hourly real wage 1992=100 Private Public Private Public Private Public _Private _ _ _ _ Public I1991-92 100.00 1,704.10 1,614.58 1,704.10 1,614.58 -____ 1995 I142.91 I1,459.36 I2,132.43 1 1,021.20 I 1,492.19 I ----- 23.43 124.75 I 10.90 1 ____- I 15.07 83 I t might seem surprising that after the substantial salary increase that Honduras had since 1997 teacher salaries are just in line, or even lower, than other salaries. A salary increase was probablyjustified to fillthe gap between teachers and other public employees. 84 Salaries are, however, somewhat lower for secondary teachers in Honduras and public sector teachers with university degrees inEl Salvador. 85 The increase compensated for eroded purchasing power in the previous years in Honduras and El Salvador (see The World Bank 2004 b,c). 64 1999 I 153.91 1 2,155.99 I 3,632.76 I 1,400.78 I 2,360.25 I 24.90 1 26.24 I14.06 122.49 2000 157.42 1,973.71 3,778.81 1,253.80 2,400.49 23.69 26.08 13.23 23.01 2001 163.31 2,376.10 3,862.77 1,454.93 2,365.24 24.79 25.90 14.67 22.83 2002 166.36 3,517.77 3,772.57 2,114.54 2,267.70 24.63 27.48 21.46 20.63 Evidence taken from El Salvador on changes in hours worked and real salaries shows that, up to 2001, although hours worked have increased for public teachers, real salaries increased much faster resulting in a substantial increase in hourly salaries (see Table 2.7 and Figure 2.2). Since 2001, a reverse trend in real salaries and a substantial increase in hours worked has resulted in a substantial decrease inhourly salaries. Figure 2.2: El Salvador- Evolution of real salary, hours worked and salary per hour for public teachers (Source: Table 2.7) In Honduras, real salaries of primary teachers have 160 substantially increased 140 since 1997 with no 120 equivalent increase in hours worked which, in 100 fact, decreased over this 80 period, leading to an even 60 higher increase in hourly 40 wage by 2002 (see Table 20 2.8 and Figure 2.3)86. A similar trend can be seen 0 for private teachers. A ~ 1995 1996 1997 1998 1999 2000 2001 2002 I t R e a lsalary (index: 1995=100)+Hours worked per week -.*-Hourly real wagell possible interpretation of these trends i s that salary increases led to a dominant income effect rather than substitution effect: Le., that the salary increase, through its positive impact on income, led teachers to work less instead than increasing their effort. This is, however, only one possible explanation, more research needs to be done to confirm or reject this hypothesis. I Table 2.8: Honduras-Evolution of real salaries for public and private primary teachers Monthly nominal Monthly real salaries Hours worked per Hourly Salary (in 1990 lempiras) 1week public private public private public private public private 1995 1670.1 1234.8 590.33 436.47 34.1 32.4 4.33 3.37 ]1996 I 2115.8 I1410.9 I 606.45 1 404.40 1 33.0 I34.6 I 4.59 I2.92 86 An indirect consequenceof the salary increase inHonduras was however the one of attracting unemployed university professionals into the teaching profession. 65 1997 I 2113.3 I 1905.0 I 517.30 466.31 I 36.3 I 33.9 I 3.56 I 3.44 1998 2543.8 2521.2 548.29 543.42 34.8 34.0 3.94 4.00 1999 3216.6 2784.4 627.62 543.29 35.3 37.5 4.44 3.62 2001 4497.1 3859.5 718.40 616.55 33.4 32.0 5.38 4.81 2002 4848.5 4020.9 718.27 599.37 31.5 30.9 5.70 4.85 Source: EPHM several years. Notes: 1 Lempira=0.06 US$. Figure 2.3: Honduras- Evolutionof real salary, hours worked and The tenuous link between salary per hour for public primary teachers teacher salary, teacher (Source: Table 2.8) effort and student performance, suggests an absence of mechanisms to link salary increases to 120.00 teaching quality or hours worked in Central 100.00 America. 80.00 This is further evident in 60.00 ~ the salary structures of the different countries (see Box 6.1 in Annex VI). In all countries salaries are driven b y seniority and 1995 1996 1997 1998 1999 2000 2001 2002 academic qualifications, l+Real salary (index: 1995400) *Hours worked per week +Hourly real wageI although to somewhat different extents across countries. This structure encourages teachers to remain in the field and to seek higher levels of formal education, both potentially positive incentives. Yet salary structures pinned tightly to education and years o f service fail to reward good teaching and learning. Furthermore, under the fiscal constraints faced by many countries in the region, large benefits related to academic title can tax governments and ultimately generate a disincentive for national policies on ~ertification.~~ Performance-based and other targeted incentive reforms are few and far between. Incentive reforms attempt to promote specific choices or behavior on the part of teachers by institutingsystems o f rewards or penalties. While incentives have been shown in cases to boost student achievement, especially on the precise indicators rewarded by the bonus, they have in other instances been shown to be ineffective at prompting the desired behavior or choices, and in other instances to result in undesirable responses. Only El Salvador has introduced, and only recently, a merit-pay mechanism (el bono ul bum desempeiio institucionul), which by linking salaries to school efficiency indicators in several areas i s designed to provide a performance incentive. As explained in Box 6.1, 87 This is the case in Honduras where an urgent reform of the Teacher Statute is needed to, at least, eliminate the title allowance for primary teachers. 66 this mechanism works as a monetary incentive for teachers in a school to work together to improve school outputs.88 The effectiveness of this incentive still needs to be assessed. Additionally, none of the Central American countries penalize teacher absenteeism nor reward teacher attendance or punctuality. In part, this is because teacher hours and attendance are difficult to accurately monitor. Also, there is very little culture of teacher assessment or accountability in Central America. A possible way to promote teacher effort might be to simply reward the number of hours worked by increasing the value of the hourly wage with the number of hours worked. This would, however, require accurate and reliable means of measuring teacher hours. A positive development in Central America has been the introduction in El Salvador, Honduras, and Nicaragua (as well as Costa Rica) of pay incentives for rural or otherwise disadvantaged geographic areas. These incentives are designed to encourage skilled teachers to work in marginalized areas. Notably, Guatemala does not have a similar incentive for its rural teachers. Strong inspection systems, decentralized systems of teacher monitoring, and fixed-term contracts are other types of measures which could enhance teacher performance but have been less consistently evaluated. Reflecting the weak teacher assessment culture mentioned above, centrally driven inspection systems are generally weak in all countries, with the possible exception of Costa Rica. On the other hand, several Central American countries are moving towards increasingly involving parents and principals in teacher monitoring in an effort to compensate for the lack of central inf~rmation.~~ This "indirect" supervision system will be fully evaluated in the chapter on community-based school management. Finally, more flexible contracting options are also increasingly being usedwithin a decentralized framework. This will also be examined in chapter V. C.PEDAGOGICALPRACTICES Learning i s not only about teachers but also, as the education production function literature tries to capture, about what is taught and how it is taught. Teaching strategies, in particular, together with teacher capacity and effort, will ultimately influence the way curriculum i s transmitted to students. In general, Central American countries have nationally determined curricula, with limited regional differentiation. In contrast, teaching strategies are developed fairly autonomously by teachers. Teaching strategies, and to some extent curricula, can also differ according to the context in which they are applied. In particular, rural areas may require different strategies than urban ones, indigenous populations may have particular needs that traditional curricula and teaching strategies cannot address, and students who fall behind may require special teaching strategies to catch up with other students. 88 This incentive is similar to Chile's SNED (Sistema Nacional de Evuluucibn de Docentes) although student scores are not directly taken into account. 89 This isthe "asymmetric information" rationalefor decentralization. 67 The lack of common indicators of pedagogical practices between studies and the difficulties in measuring these practices in the first place make it difficult to discern clear trends (see Table 2.9 below). The meta-analyses reviewed do not find clear patterns of significance either, perhaps for the same reasons. An exception to this i s that frequency of homework, length of instructional time, curriculum coverage and use of active pedagogies seem to have, overall, a positive impact on achievement. Pedagogical variables may also have an impact on repetition and grade failure, independent of educational achievement measured by test scores. Table 2.9: Summary Findings of Pedagogical and CurriculumRelated Determinants of Education Quality in Central America and the Latin American Laboratory Study (N=Nicaragua, E = ElSalvador, 1 1 1 1 1 H=Honduras. G=Guatemala) UNESCO Studies Num Of Positive Negative Mixed Significant Not LLECE IPedaaomf I I I I I I 1.Text use 1 H 2. Homework frequency 2 H, G 1 3. Blackboard use 1 1 I H l I I I ~ 4. Teacher explains and asks questions 1 H 5. Research-based teaching 1 E 1 6. Participative methods I 3 I H,G I I I E l 7. Traditional methods 1 E 8. Use of pedagogical supplies 1 E I 9. School-wide pedagogy 1 1 1 E l I I I 10. School has education project 1 E Curricula I I. Curriculum type I l l I I I E l 2. On schedule with curriculum 2 E, H 3. Bilingual curriculum 1 G All the countries of the region implemented far-reaching education reforms inthe 1990s, generally coinciding with the return to peace and democracy. Curricular and pedagogical reform i s one of the main thrusts o f these reforms, along with decentralization to school- and community-based management, increases in the Ministry o f Education budget, and the creation of national assessment systems. Box 7.1 in Annex VI1presents a review o f curricular and pedagogical reform measures in all countries. Overall, except in El Salvador, the curricular reforms focused on primary schools, were based on constructivism, and were implemented flexibly and gradually (going through different phases o f elaboration, validation and generalization). A more flexible focus also made it easier to incorporate multicultural elements closer to local 68 realities. In all cases, these curricular reforms were accompanied by the creation of new textbooks and teaching guides. In some cases, these new teaching and learning materials were produced nationally in order to more closely reflect national curricula while in other cases they were imported from other countries. What follows is an attempt to provide an initial summary of evidence of the impact of these reforms, both of the new curricula per se and of the implementation of the curricula and teaching strategies in the classroom. This section offers diagnostic information on the state of pedagogical practices and instructional time in Central America and how they impact student learning. a)Active/ParticipativeTeachingStrategies On curricula per se, we have practically no assessments, with the exception of El Salvador (see Box 7.1 inAnnex VII). We have, however, some evidence on the impact of the related introduction of active/participative teaching practices that can help determine if the introduction of constructivism inthe curricula was a positive factor. There is some evidence across the studies that we have reviewed that more active/participative pedagogical practices have a positive impact on educational achievement as measuredby test scores (Table 2.9). In particular, one of the two studies on Guatemala shows that a higher teacher-student interaction, typical of more participative techniques, is positively associated with educational achievement and the Honduran study reports a positive impact of participatory techniques on learning outcomes. The evidence from El Salvador i s less clear-cut with a surprisingly positive impact found for traditional techniques. This evidence conflicts, however, with qualitative evidence collected on 24 urban and rural schools which shows that best performing schools tend to apply participative practices." Additionally, the impact of active practices may be stronger for particular sub-groups of students, such as students with particular learning difficulties or over-age students. This would not be captured by the studies reviewed here." There has been some evidence along these lines for El Salvador (see Box 7.1). A recent study on Honduras also suggests that traditional teaching methods may exacerbate repetition for repeating student^.'^ Finally, the much more extensive literature on Latin America reveals that there is some evidence of a positive impact of active strategies on learning (Annex 11.1). If this is the case, the move towards constructivism is a positive trend in Central America. Weak execution of the constructivist curriculum. However, implementation of constructivist techniques has been weak at best. Box 7.1 in Annex VI1 reveals that in most countries classroom practices did not change substantially following the pedagogical reform. There are several possible reasons for this, none of which has been examined rigorously across countries. Among the most commonly mentioned reasons are: (a) the lack of teacher expertise in participatory pedagogies; (b) the lack of textbooks available on a wide basis; (c) large class sizes; and (d) teacher opposition to the reforms. The lack of teacher expertise in the application of active pedagogies was already recognized as a weakness of in-service training above when it was mentioned that still See Kraft (2003). 91 See Schiefelbein(2004)for an argument along these lines. 92 See Marshall, J (2002): "Grade repetition in Honduran primary schools" - mimeo (Stanford University). 69 too little effort is put on changinghpdating classroom practices. This i s confirmed here. Applying a new curriculum requires substantial training and support to enable teachers to effectively change their methods. The magnitude of the effort needed was probably under-estimated in the Central America countries under analysis. Class size, at least within a certain reasonable range, has not been consistently related to educational achievement, but active and participative teaching methods particularly may be easier or more effectively implemented in small to medium size classes. Finally, on teacher opposition, it appears that in many cases teachers have not been sufficiently consulted on the reform process.93 b) Instructional Time and ClassHours Another major weakness in the execution of the curriculum and quality o f education in Central America i s the relatively low number of actual teaching hours. Instructional time (also related to coverage of the curriculum, as measured in Table 2.9) i s frequently found to be positively related to student achievement and pass rates, including in two recent studies of Honduras94and G ~ a t e m a l a . ~ ~ All countries in this report have schooling calendars of 180-200days and a school week of 25-30 class hours in primary. This results in roughly 900-1,200 class hours a year, similar to the rest o f Latin American.96 However, there i s a widespread perception in Central America that actual instructional time is much lower. Teacher and student absences and school closings considerably reduce the effective class time o f students. Evidence at this point, however, i s largely anecdotal. It i s notoriously difficult to assess effective class hours due to information constraints on teachers' absences and the magnitude of teaching hours. The ministries o f education do not usually collect this information, which i s constrained by the absence o f systematic reporting of attendance and hours at the school level. In this vacuum, we made an attempt to assess effective class hours in the Central American countries under analysis97by: (a) assessing average weekly teaching hours (using school-based observations on teaching hours generated on school samples of different size - often small, unfortunately); (b) assessing the effective number o f weeks of instruction, using information from a similar source on teachers' absences and, when available, school closings98; and (c) obtaining the effective yearly class hours by multiplying the effective number o f weeks by the teaching time. 93 This has been the case in Guatemala where teachers were consulted at the beginning of the reform but not during the reform process itself and in Nicaragua where teachers have not been part of the reform designand teachers' unions have been generally kept aside from the reform. The experience was different in El Salvador where teachers were consulted since the inception of the reform and have taken part in the whole process mostly through national consultations. Teacher inclusion and national consensus may also explain why El Salvador could pass the reforms on teacher selection and remuneration discussed earlier. 94 See Di Gropello and Marshall (2004). 95 See Marshall (2003a). 96 900-1,200 hrs/yr is significantly less than the averages for Europe or Southeast Asia (1,200-1,750 hrs/yr) as cited in P E A L (2002). "AlthoughnoinformationcouldbegatheredonNicaragua. 98Or, in the case of Guatemala, directly on the effective school year. 70 This methodology is a very simplified one99,but provides us with some insights that are better than the anecdotal evidence. Low instructional time. Table 2.10 reveals that in all countries there would be between 500 and 800 effective class hours at most, which represent between 58 and 80 percent of the notional class time (see Figure 2.4). It i s obvious that even 700 class hours per year compare very unfavorably with the more than 1,200 notional class hours of Europe or other countries and constitute a major barrier to educational achievement in Central I Table 2.10: Effective class hours in Central American Countries I Honduras El Salvador Guatemala Weekly Teaching Hours (a) 21 21 20 Weekly Class Hours (b) 21 21 20 NormalWeeks per Year 42 42 36 EstimatedTeacher'sweekly absences (c) 4 3.5 na School closings due to non-teacher absence reasons (d) I 3 l n a I na I Effectivenumber of weeks (e) 35 38.5 26 Effectivenumberof class hours per year (f) 735 808 520 Effectivenumberof class days per year (9) 122 161 104 America. Honduras and Guatemala seem to have the largest gap between notional and effective class hours, while El Salvador has the smallest gap. 99For instance, it makes the assumption that teaching hours and class hours are similar, which will be valid only if one teacher teaches all subjects and, if there is more than one shift, one teacher only teaches one shift. Additionally, the estimated effective class hours per year will generally be an under-estimation because o f the lack of information on unexcused teachers' absences in Honduras and El Salvador and, for El Salvador, on schools' closings (beyond normal holiday closings) due to reasons others than teachers' absences. On the other hand, considering all teachers' absences may somewhat under-estimate class hours if a substitute or replacement teacher is used. Finally, the figures that we obtain provide an estimate of effective class hours, but not of time-on-task, which is expected to be even lower if we consider that a class hour includes a 15-minute break and that only a portion of the remaining 45 minutes is spent on the task itself (time frequently is spent on discipline, grading, organizing, setting up, etc). 71 D. EDUCATION SPENDING AND IFigure 2.4: Notional and effective classhours per year SCHOOL MANAGEMENT 14001 I 1200 A third category of factors, increasingly taken into account 1000 in educational achievement 800 studies, relates to broader variables such as education 600 spending and school 400 management, which are expected to have an impact on 200 educational achievement by 0 impacting the behavior and characteristics of teachers, the school learning environment, instructional and physical resources, and stakeholder responsibilities. These factors are only cursorily addressed in this chapter as both topics have their own respective chapter later in this volume. Education spending i s addressed in Chapter IV while school management i s addressed in Chapter V. a) Education Spending Education spending has not been tested in the Central American education production function studies reviewed for this chapter. Education spending is, however, significantly and positively related to student achievement in about half the cases cited in the meta- analyses o f education in Latin America and developing countries generally. It stands to reason that the impact of educational spending depends on the effectiveness of resources allocation. There i s evidence that school supplies, textbooks and school infrastructure are beneficial to student learning in Central America specifically, and developing countries generally. Other factors related to education quality, such as teacher education discussed earlier, also have spending implications. Since education spending i s related to both quality and coverage outcomes and i s currently a key issue for all Central American countries, we will deal with it in a separate chapter. b) SchoolManagement How schools are managed also affects education quality. Several of the Central American education production function studies examine school factors relating to school management. There i s little consistent evidence on the impact of directors' characteristics on student learning, although director leadership appears, for instance, to be positively related to educational achievement in El Salvador, confirming the general recognition that directors' leadership i s a key factor for effective school administration and classroom environment. Here again, the inconsistent findings are likely to be due to limited data, poor quality data, and unobservable factors. Surveys should explicitly measure more 72 principal characteristics and behaviors in order to increase understanding of the impact of, say, director selection and training, director commitment/leadership, and other factors. More evidence is available on school autonomy and community participation and, interestingly, these variables, in many cases, have a significant and positive association with student achievement. In particular, school autonomy is positively related to educational achievement in both El Salvador and Honduras (which may also contribute to explain why private schools perform better in El Salvador), while it has a more ambiguous -positive or negative- impact in Guatemala. Within the current decentralization strategies adopted in Central America, it is particularly important to be able to assess the impact of autonomy and community participation. To address this issue special studies have been undertaken comparing community-based schools with traditional ones that we will explicitly review inChapter V. A separatechapter on school- based management is required not only because of the increasing relevance of decentralization in the countries under analysis, but also because decentralization is, like education spending, a crosscutting area which affects quality, efficiency and coverage. 111. Policy Implications The previous section detailed many of the constraints to learning in Central America. This next section takes each of these areas and proposes policy options to overcome these barriers. We begin with an analysis of the qualities of high-performing low-income schools and what differentiates them from less effective low-income schools. This analysis is offered in order to begin to look at what policies, characteristics, and practices being usedcurrently in Central America may be valuable to consider. The following sub- sections then look at specific policy options to address the learningconstraints identified inthe previous section: (a) studentbackground characteristics; (b) teacher quality; and (c) pedagogical practices. We do not look at policy options concerning education financing or school managementbecause each of these sections is dealt with in subsequent chapters of this report. A. LEARNINGDETERMINANTS EFFECTIVE OF SCHOOLS In Central America, where poverty continues to be a severe problem affecting large segments of the population, insuring high quality education for the poor must be a top priority. Unfortunately, to date, schools serving the poor tend to be of significantly lower quality than those serving more privileged social sectors, and student achievement continues to parallel gaps between social class and racial divisions in ways that underscorecontinued urgent inequalities. The goal of the following analysis is to identify the characteristics of high-performing low-income schools in Honduras and Nicaragua that differentiate these schools from average or low-performing low-income schools (no equivalent databases were available for El Salvador or Guatemala). a) Methodology and Data Annex VIII contains a description of the data that were used and the methodology applied to undertakethis effective school analysis. Here, we just point out that groups of effective 73 and "control" schools were identified in each of the countries with effective schools being defined as schools that scored substantially higher than the mean on multiple exams and significantly below the mean in terms of socioeconomic status. The tables below report t-tests on sample means of control and effective schools. b) Findings and Discussion Before discussing specific results from the analysis, a couple of general findings are worth mentioning. First, the cut off boundaries to identify effective schools in Nicaragua was much less rigorous than that of Honduras. This was necessary because there simply were very few high-performing, low-income schools in the Nicaraguan dataset. This may be partially due to sample size, but i t i s also likely that socioeconomic status (SES) and test scores are more tightly linked in Nicaragua than in Honduras; in other words, educational outcomes are more inequitable in Nicaragua. Evidence of this sharp inequality in Nicaragua was also highlighted in Chapter I. Second, many more schools met the criteria as effective in math than in Spanish in both countries. This phenomenon has been discussed in other literature and has been attributed to the fact that children learn more literacy skills than math skills at home. This means that children enter school on more equal footing for math than for language. The occurrence of more effective schools in math than in Spanish in both Honduras and Nicaragua demonstrates that the inequalities in children's backgrounds are not being sufficiently compensated for in schools. i) Teachersandprincipals Skilled teachers matter. There are important differences between the teachers and principals of effective and control schools in Honduras and Nicaragua (Table 2.1l), but these differences are not entirely consistent across country. In Honduras, the education level of teachers and principals i s significantly higher in effective schools but these same teachers and principals have fewer years o f experience (although this variable i s not significant). In Nicaragua the situation i s reversed. Although limited to the sixth grade, teachers and principals in effective schools have significantly lower education levels but more years of experience. This evidence suggests that the policy of hiring teachers with low qualifications in rural areas i s problematic in Honduras, where teachers with university degrees obtain better results in low-income schools. In contrast, in Nicaragua, hiring teachers with limited educational background may be all right (at least for grade 6), so long as the teachers hired have a lot of professional experience. Inboth cases, these findings illustrate the need for policies that attract and retain competent teachers in poor areas. 74 Honduras Nicaragua3 Nicaragua6 Effective Control Effective Control Effective Control Teacher Education ILevel I I I 2.18 1 2.10 I 1.89 I2.35*** I Only Normal 0.71 0.84** University degree in 0.12 0.05** pedagogy Universitydegree 0.12 .06* Teacher Experience Teaching 5.98 7.40 6.36 7.17 16.55 11.03*** Teachina Grade 3.55 3.28 13.71 4.78*** 1 Principal Education I I I I I I I Level 2.24 2.47 2.00 2.41** Universitydegree 0.13 .05** PrincipalExperience 6.40 7.28 5.28 4.39 7.55 7.9 ii) Pedagogy Positive impact of active methodologies. There are significant differences in teaching methods and practices reported in effective and control group schools, particularly in the sixth grade in Nicaragua (see Table 2.12). Effective schools appear to use traditional pedagogical practices significantly less often than control group schools. This would indicate that more active methodologies are employed in successful low-income schools in Nicaragua (confirming the importance of this type of teaching). Also in Nicaragua, effective schools were much more likely to use flexible promotionthan control schools in both the third and the sixth grades. Flexible promotion allows students to re-enter or pass between grades levels even if they have been absent for a significant portion of the school year. This i s an interestingfinding since flexible promotion policies are also particularly useful for ensuring completion. This finding suggests that automatic promotion does not necessarilyjeopardize the quality of learning. InHonduras no classroom practices appear significantly different in effective and control schools, but teachers in effective schools give more homework, on average, than teachers in control schools. Part of the reason why there are more pedagogical differences in Nicaragua than in Honduras is that Nicaragua has spent more than 10 years focusing on changing classroom practice to a more dynamic and interactive constructivist model of learning. It is probable that Nicaraguan teachers have been exposed to more training and skill development in using a broader range of teaching methods although only a minority of teachers seems to have applied these practices. 75 Honduras Nicaragua3 Nicaragua6 Effective 1 Control Effective 1 Control Effective 1 Control Traditionalpedagogy Dictation 0.1 0.06 0.71 0.81 0.55 0.75** lCopy into notebook I 1 1 0.96 I 0.94 1 0.92 1 0.97* I /Teacherreadsfrom text 1 I 1 0.92 I 0.92 I 0.92 1 0 . 9 6 * 1 IHomeworkFrequency I 3.51 1 3.33* I I I I I Flexible Promotion 0.5 0.03*** 0.43 0.10* iii) School managementpractices Teachers and parents are more engaged and supported in effective schools. Some very interesting differences emerge looking at school management practices in effective and control group schools (see Table 2.13). Namely both teachers and parents appear to be more engaged and supported in effective schools. Table 2.13: School Management Practices inEffective and Control Group Schools 1 I Honduras I Nicaragua3 I Nicaragua6 1 Effective Control Effective Control Effective Control Teachers' role in management of school Teacher responsibilityindex 0.57 -0.33** 0.83 0.17 Indexfor teacher support 0.19 -0.08* Parental participation Indexof parentalparticipationin 1.46 0.72* 0.64 0.99 school 1 Parentscollaborate with education 2 1.79* I I I Percent of time principal spends 11 48.07 1I 41.57* I1 teaching I II School-based management AutonomousSchool 0.6 0.42 0.55 0.49 ISchoolautonomyindex 1 0.09 I 0.05 I I I I I Source: Nicaragua: MINED Database; Honduras: UMCE Database. Notes: Asterisks refer to significance (two-tail)for t- test comparisons of independent sample means assuming equal variances. *** Significant at 0.01 level; *' Significant at 0.05 level; * Significant at 0.10 level. 76 In Nicaragua teachers have a wider range of responsibilities in school management in both third and sixth grade, although the difference is only significant at the third grade level. There was no equivalent data for Honduras, but in that country teachers reported receiving more support and appreciation for their work. Similarly, there is indication that parents may be more involvedin school functioning and activities as well as their own children's learning in both countries. InNicaragua, parents score significantly higher on a composite index measuring parental involvement in schools in the third grade. This relationship did not hold for the sixth grade however. In Honduras, teachers reported that parents collaborate more with the education of their children. Both of these measures indicate that there may be a closer relationship between families and schools/classrooms in high-performing low-income schools inHonduras and Nicaragua. If accurate, this i s a promising finding for community-based school management (discussed in ChapterV). Finally, in Honduras, effective schools principals spend more time, on average, teaching. This difference suggests that principals in those schools may be more involved in instructional or pedagogical leadershiprather than having more administrative functions. Unfortunately, we cannot test community-based schooling (PROHECO) directly in Honduras because the Honduran sample consisted of too few PROHECO schools. We will see in Chapter V (where we use a specially generated database) that PROHECO does, indeed, have a positive impact on many schools (although PROHECO schools are weak on the teacher involvement side). We can only test a general measure of school autonomy, which, although higher in effective schools, does not reach significance. The results for the autonomous schools in Nicaragua'ooindicate that, though not significant, effective schools are more likely to be autonomous than their control group counterparts. Chapter V will show that the impact of the Nicaraguan decentralization reform is somewhat constrained by low parent and teacher involvement. The model may, however, holdpromise for rural areas. We should add that school type such as single teacher schools (Le. multigrade) in Honduras and multigrade schools in Nicaragua were also tested and, although effective schools were more likely to be single-teacher and multigrade, the difference was not significant. We will see that, when focusing on rural areas, the benefits of multigrade teaching incertain countries emerge more clearly. iv) Student repetition and absences One final area of differences between effective and control group schools emerged from this analysis concerning the reported levels of repetition and absences of students in the two groups of schools. Lower repetition and student absenteeism rates. In Honduras, students were significantly less likely to have repeated years of school in the effective schools and the '00 Where the sample of schools in Nicaragua was designed to allow for stratification in terms o f participation in the Autonorniu Escolar reform. 77 average sum of years repeated by each child was significantly lower than in comparison schools. A similar trend, although not statistically significant, i s true for the number of students repeating the current year in Nicaragua in both third and sixth grade. Not surprisingly, these results point to a relation between achievement and repetition: high- achieving schools also have lower repetition rates, confirming the importance o f quality education for completion. Students also appear to be absent less frequently in effective schools in both countries. In Nicaragua, both girls and boys were less likely to have been absent the week prior to the survey in effective schools for both third and sixth grade although the difference i s only statistically significant for sixth grade girls. In Honduras, students were also reported to be absent significantly fewer days in the effective schools. v) Conclusions andpolicy implications Although there i s a fair amount of variation among significant variables between grade levels and countries there are some clear and intuitively meaningful differences that emerge between high-performing and lower-performing low-income schools in Honduras and Nicaragua. These differences center around four areas. First, teacher and principal education and experience both appear important for effective learning, suggesting the need for a policy which brings and retains competent teachers to poor areas. Second, effective schools are less likely to use traditional passive teaching methods and give more homework to students, suggesting the need for promoting new pedagogical approaches. Third, substantive and supportive involvement of both teachers and parents in school management and a potentially instructional leadership role for principals are present in effective schools, pointing to a possible high potential for school-based management or other forms o f community involvement. Fourth, students in effective schools are less likely to repeat grades and are in school more. The results on absenteeism suggest that policies to limit students' absences, for instance by improving teacher effort, offering nutritious meals, or making schools more accessible, could improve achievement in low income schools. These differences emerged between effective and control group schools that had very similar background characteristics and therefore may represent avenues b y which to improve learning in low-income schools. Not surprisingly, the areas in which effective schools differ from their less-effective counterparts mirror the main factors found to be determinants o f and constraints to learning in Central America. These include student background characteristics, seen in the effective schools analysis in the findings on student absenteeism and repetition; teacher and teaching quality, seen in the effective schools analysis in terms o f teacher education and experience; pedagogical practices, seen in the above analysis in the form of interactive teaching methods and more homework; and finally school management, seen inthe analysis as greater parent, teacher, and community involvement and empowerment. The next sections suggest specific policy recommendationsfor the three first areas. 78 B. BACKGROUNDCHARACTERISTICSPOLICYIMPLICATIONS We saw earlier that constraints to learning that relate to student background characteristics almost exclusively disfavor marginalized segments of the population, in particular, poor, rural, and ethnic-minority students. Strategies for improving education for the poor are varied. This section outlines a variety of supply-side and demand-side policy options to remove or compensate for these barriers, including compensatory programs, cross-sectoral policies, conditional cash transfers, decreasing the private costs of schooling, and publicity campaigns. a) CompensatoryEducation Programsfor Rural, Indigenous and Over-age Students in CentralAmerica Central America currently has multiple compensatory programs in place that support equality of educational opportunity for rural students, students whose first language is not Spanish, and over-age students. These groups of children have suffered long-lasting educational and socio-economic disadvantages (see Chapter I). Their educational needs are both great and distinct from the educational needs of the 'typical' student generally considered in educational planning. As countries increasingly recognize the enormous barriers to equality of educational opportunity and the severe disadvantages these groups of students face, there are a growing number of programs and policies that attempt to remedy these weaknesses. Some of these educational adaptations and compensatory programs have achieved considerable success and have become models for other countries, although there are still challenges ahead in ensuring equality of educational outcomes. We do not hope to report on all of the compensatory policies and programs for the region but rather to focus on three main areas: 1) multigrade schooling; 2) bilingual education; and 3) specialized learning programs for over-age youth. For each area we will describe some of the region's programs, report some evidence of their impact (if any exists), and present a few summary reflections. i)Multigrade schoolingfor rural populationdo' Providing quality and equitable education in rural areas is a challenge faced by all the Central American countries. Large portions of the national populations continue to live in remote areas. These rural populations are, on average, much poorer than their urban lo'Multigrade schools are not the only option for expanding educational access in remote areas. More and more frequently countries are using distance education techniques, through radio, television, and computers, to offer or expand educational opportunities where they previously did not exist, to lower the unit cost of providing education in remote areas, and to improve the quality and content of education offered in these areas. Distance education is not typically used for primary education in these countries but is more frequently used to provide secondary, technical, or adult education, as we will see in Chapter 111. Another policy aimed to improve education delivery in rural areas is flexibility and adaptations to the academic calendar. In Nicaragua, for example, policies are in place to allow rural schools to run on academic calendars which mirror agricultural cycles in the surrounding area, enabling students who help their families during harvests or planting not to miss school due to this work. 79 counterparts and have fewer available material and human resources. In addition, populations are widely spread making access to school much more difficult and costly. Teacher shortages in rural areas are also common. As a response to these serious challenges many countries are using a multigrade model of education in which one teacher simultaneously leads class for more than one grade level. This cuts down on the number of teachers needed to provide more complete schools. But because of this structure, and the rural context of these schools, multigrade schools have unique needs, in particular teachers must be able to successfully facilitate the learning of students o f different ages and skill levels, materials must be adapted for students who receive less frontal teaching, and schedules and curricula must be appropriate for the rural multigrade setting. At their best, multigrade schools have been shown to cost-effectively address the unique context o f remote rural areas and provide rural children with not only access to all primary education grade levels but also ensure relevance to the local setting, and successful student outcomes.'02 At their worst, however, multigrade schools can perpetuate and exacerbate the inferior education provided to rural comm~nities.'~~ Program descriptions. Although all four countries have introduced multigrade schooling programs, evidence on impact i s limited and of varying quality. O f all the programs, the Guatemalan New Unitary School (NEU) project has been the subject of the most academic study, while several of the other programs have received limited evaluation by education officials, donors, and external evaluators. Box 2.2 includes synopses on the range o f multigrade. lo* In UNESCO's Latin American Laboratory study of education quality in eleven Latin American countries,only Colombia hadhigher studentachievementin the ruralsector than in the urban sector.This is largely thought to be due to the Escuelu Nuevu schools, a rural multigradeeducation programbegun in the 1970sthat is the inspiration behindmany of the CentralAmerican multigradeprograms(PREAL, 2003a). lo3BenvenisteandMcEwan(2000). 80 Improvements in test scores. There i s evidence that when successful, students in multigrade schools can outperform traditional rural students and even urban students. The NEUprogram evaluation found that NEUschools students have better reading skills than traditional rural school students.' lo Similarly, indicators of student achievement in Nicaragua's model multigrade schools are promising. Test scores were higher for math but lower or similar for Spanish than in regular model schools in 2002. Girls, however, IO4PREAL (2003a). IO5AED andJuirez andAssociates (2003). IO6The Alternative Classrooms Project is implemented in EDUCO schools, community managed rural schools inEl Salvador. They will be discussed moreextensivelyin Chapter V. lo'PREAL (2002). lo*MEP (2002); MEP (2001). '09MINEDUC, OEI, & World Bank, no year. 'loKraft (1998). 81 scored lower in both Spanish and math in multigrade schools indicating that more attention may need to be directed at gender inequities in rural areas (see Table 2.14). Finally, in Honduras, when examining only rural schools, single-teacher schools score significantly higher in math and science than multiple-teacher schools (there i s no significant difference in Spanish test scores). '' Table 2.14: Nicaragua Test Scores inMultigrade and Traditional - Model Schools I SchoolType Subject Gender Grade 3 Grade 4 I 2002 2002 Male 60 81 Spanish I Female I66 I82 I Model Total 63 82 Regular Male 48 64 t- Math Female 43 62 Total 45 63 Male 60 84 Spanish Female 58 80 Model Total 59 82 Multigrade Male 52 73 Math Female 41 66 I Total 47 69 Source: AED and Juarez and Associates, 2003 Greater student retention. Measures o f student retention and completion also appear to be improving in some of the programs. In Guatemala, fewer NEUschool students repeat and more students get to fifth grade than in traditional rural schools. Still, 95 percent of monolingual Mayan students continue to drop out o f NEU schools before completing primary education (see Table 2.15). While not disaggregated by multigrade or traditional school type, student completion in model schools in Nicaragua i s significantly higher than national averages.' l2 The alternative classrooms program in El Salvador also appears to be reducing the proportion of over-age students in rural schools. Between 1992 and 2000 the progress in increasing the proportion o f age-appropriate enrollment was three times faster inrural schools than in urban ones. 'I'UMCE (2003a). '12AEDand JuBrez andAssociates (2003). 82 Table 2.15: Guatemala Retention and Progression Rates for NEUand - Traditional School Students Advance without RetentionRate School Type repeatingto Advance to for Monolingual fourth grade fifth grade Mayan Students NEU Schools 30% 154% I 5% I ControlGroup II10% I25% 15% ~~ Source: Kraft, 1998 Limitationsto the multigrade model. The experiences of multigrade in Costa Rica and Honduras highlight the existence of problems in multigrade schools. Single-teacher schools in Costa Rica have inferior pass, completion, repetition and dropout rates compared with traditional rural schools. They typically offer less instructional time, lack specialized multigrade teacher training or preparation, have few educational resources, and poor infra~tructure."~The Honduran Ministry of Education further reports that these schools often suffer from inadequate infrastructure, weak management, and constrained budgets. Policyimplicationsfor multigradeschools From the experiences that we have reviewed, it would seem that when implemented purely as a cost saving or access-expandingmeasure, multigrade schools are likely to fail. These schools necessitate an entirely different education model than the traditional frontal model still dominant in Central America. Teachers in multigrade schools need sufficient high-quality educational opportunities to be successful and schools need to be equipped with appropriate and relevant textbooks and other educational materials. It appears that Guatemala, Nicaragua and El Salvador have been more successful in providing multigrade schools with significant support aimed at making them successfully different from traditional schools. But significant challenges remain even in countries with broad support for multigrade schools. InGuatemala, monolingual Mayan students are still extremely disadvantaged in terms of educational completion. In all countries, as i s clear from Chapter I,students in rural areas continue to be more likely to be over-age and have lower average educational attainment and academic achievement than urban students. Multigrade schools are not managing to fully compensate rural children for the geographic, economic, and social barriers they face. Providing continuous pedagogical support and educational resources will help strengthen the impact of these models. Alternatively, cost-efficient options for education delivery such as organizing rural schools in networks, in particular if associated with elements of community-based management, may also help organize education delivery in rural areas. This option may be particularly promising if including all nine grades of basic education. * I 3MEP(2002); MEP(2001). 83 ii) Bilingual and multicultural educationfor ethnic and linguistic minoritypopulations The Central American countries have diverse ethno-linguistic populations, primarily indigenous Native American groups (largely of Mayan affiliation) and Afro-Caribbean populations along the Atlantic coast. The very distinct histories of each country, however, have resulted in very different population compositions among the countries. As indicated in Chapter I,roughly half of the nation's population in Guatemala i s indigenous and one quarter of the nation's population does not speak Spanish at home. InNicaragua, nearly 1 in 10 people i s Afro-Caribbean, an ethnic group whose primary language i s English. Other countries, such as El Salvador, have much more ethno-linguistically homogenous populations. In all of the Central American countries, however, ethnic minorities, especially indigenous populations, face severe economic, political, and social disadvantages compared with the majority mestizo populations. Increasing the human capital of indigenous people has been shown to decrease these ineq~alities."~At the same time, increased education levels have been shown to have a negative association with indigenous language proficiency.' l5 Supporting cultural and linguistic diversity while combating the severe inequalities indigenous and other ethnic minorities face i s a critical challenge in most Central American countries. Only recently have the Central American governments begun state-initiated and supported bilingual education programs. All countries, except El Salvador, have by now established these types of programs. Here we describe the Guatemalan program, as this is by far the most advanced of the bilingual programs in the region and the only one that has been evaluated. Annex IX includes a box describing the programs implemented in Nicaragua and Honduras. Guatemala BilingualInterculturalEducation(DIGEBI): descriptionand impact - Despite the fact that the Ministry of Education only began bilingual education in Guatemala in 1980, they have made significant progress in the area and are frequently regarded as a model for other countries developing bilingual intercultural education programs.' l6 Due to its large indigenous population, Guatemala, of all the Central American countries, faces the most pressing need for quality bilingual education. Complicating the matter, however, i s the fact that Guatemala's indigenous population i s not monolingual. There are between 24-52 indigenous languages used in Guatemala, although 75 percent of the total indigenous population speaks one of four main Mayan languages. These different groups not only speak different languages but also have unique education needs as evidenced by their distinct dropout, repetition, and pass rates.' '' In addition, the diverse Guatemalan population means that there is a need for all ethnic groups to learn respect and understanding through intercultural education. Patrinos andVelez (1996). The World Bank (2002). Dutcher (1997); CummingsandTamayo (1994). 117Cummings and Tamayo (1994); Carvajal, Morris and Davenport(1993). 84 Improvement and expansion of bilingual intercultural education has a rominent place in the Guatemalan Peace Accords and the Indigenous Peoples' Accord.' Multiculturalism and interculturalism are foundations of Guatemala's long-term national education strategy. This strategy calls for deepening knowledge about national ethnic groups, providing bilingual and intercultural education at all education levels, strengthening cultural understanding and respect, and insuring equitable educational access and participation to all Guatemalan populations.' l9To implement this strategy, the 2000-2004 Education Plan divides the country into ethno-linguistic areas with decentralized authority to develop and implement culturally and linguistically relevant educational adaptations.l2O Substantial investment in bilingual intercultural education (EBI) in Guatemala has gone to the development o f a national infrastructure to support the program. This infrastructure includes experts in linguistics, curriculum development, and educational planning as well as trained bilingual teachers and bilingual intercultural education materials.12' EBI schools now serve 18 percent of the indigenous Guatemalan population. Textbooks and supplementary materials have been created and produced in many languages. Guatemala has trained thousands of bilingual teachers, and offers diplomas and degrees for these teachers in applied linguistics, education administration, curriculum development, reading and writing, and special instruction for teaching bilingual first grade.'22 Positive effect on student achievement. Studies have shown that the EBI program in Guatemala has had positive effects on student achievement, parental attitudes, and school quality. Unfortunately most of these studies are fairly dated. A 1986-1991 evaluation of student achievement on standardized exams showed that students in bilingual intercultural schools outperformed students in a control group on almost all exams (see Table 2.16). B y grade three, when children have received four years o f bilingual intercultural education, students in EBI schools scored significantly higher in math and social studies. In Spanish, there was no significant difference in scores, Another evaluation indicated that EBI students in schools that had fully implemented the program had even greater achievement over traditional schools.'23 'IsMarques and Bannon (2003). 119ComisidnConsultivapara la Reforma Educativa(1999). 120MINEDUC (2000). 12'Dutcher (1997). The World Bank (2004a); Vawda and Patrinos(1999); Cummings and Tamayo (1994). 123Chuta& Chuta (1987) as cited inDutcher(1997). 85 Subject Year DlGEBl Control Significant Grade One Mayan 1988 77 74 Yes Spanish 1989 33 30 Yes Spanish 1991 86 82 Yes Grade Three Math 1986 34 30 Yes Math 1987 37 36 No 1 Social Studies I1987 I42 I 3 9 IYes I Spanish 1987 35 35 No Early evaluations in the 1980s show that students in EBI schools had improved rates of attendance, repetition, promotion, dro out, and test scores compared to a control group of Mayan children taught in S anish." Parents' attitudes toward school were also more favorable in bilingual schools.1 E Policy implications Most Central American nations are just beginning to offer bilingual education. Serious challenges remain, as indicated by the persistent indigenoushon-indigenousgap in all countries (Chapter I). Most strikingly, access to these programs needs to be expanded. As stated earlier, even in Guatemala the vast majority of indigenous children are not in bilingual schools.126127 Expansion may require a significant increase in financing for these programs, as implementation requires teacher preparation as well as curriculum and material development for new ethno-linguistic groups. It would also be timely to start evaluating the programs inNicaragua and Honduras. 124 Morris and Davenport (1993); and Scott and Chuta, n.d. as cited in Patrinos and Velez (1996). 125 USAID (1993) as cited by Patrinos and Velez (1996). 126 Silvestre, no year. 12' A recent report from a Nicaraguan Atlantic Coast non-governmental organization criticizes Nicaragua's bilingual education program (see Annex 11.5) for not serving more non-Spanish speaking communities. However, the smaller size of the indigenous population in that country probably makes relative access higher than in Guatemala contributing to the explanation o f the minor indigenouslnon-indigenousgap in that country. 86 iii)Alternative programsfor over-age youth As already discussed, over-age students are more likely to drop out, repeat, or fail and require specific targeted programs and policies addressedto their needs. Programs in the region for these populations include flexible (or automatic) promotion, accelerated education and alternative schedules. Flexible promotion allows students to continue into subsequent grade levels irrespective of whether their achievement qualifies them to pass to the next grade. This not only directly lowers repetition rates but is also designedto decrease age-by-grade distortion, a factor that is highly associated with future school abandonment. Accelerated education permits students who are over-age or who for some other reason need to get through school quickly to condense their education into fewer years. This creates an incentive for students who might otherwise drop out or remain out of school to study and complete basic education. Finally, flexible schedules, such as night shifts, weekend classes, or alternative calendars also target studentswith special needs such as working studentswho cannot come to school during typical school hourddays. At the primary level these programs are, in many ways, a "second best" compared with interventions that aim at increasing educational attainment by improving quality of schooling, but they can help increase primary completion, particularly in the short term. At the secondary level, when the pressure for students to drop out is much stronger, alternative schedules are in fact often highly advisable. The following chapter on secondary education will discuss this further. Unfortunately, there is practically no research/assessment on these compensatory programs for over-age students and therefore we just provide a summary description of a couple of interesting programs on accelerated education and automatic or flexible promotion which have been developed inEl Salvador and Nicaragua. Program descriptions ElSalvador: Pilot Program in Accelerated Education and Flexible Promotion. This recent innovation speeds up grade transition for older, more mature or motivated students. It employs a special curriculum, flexible promotion, and trains and provides materials for teachers in specialized teaching methods. This program is still small. In 2000, 47 sections of accelerated education were serving 1410 students in the second through fourth grades.12* Student learning manuals have been developed and distributed to program participants. These manuals promote cooperative learning, problem solving, creative and critical thinking, and a broad range of activities linked to c~rricu1a.l~~ El Salvador also uses flexible promotion in its EDUCO (community run) schools. Flexible promotion allows students to reenter school without having to repeat or drop back a grade level if they have been absent for any length of time. Flexible promotion is dependent on curricula and teaching methods that allow the student to continue where he 12'PREAL (2002). `29Kraft(2004). 87 or she left off. This method is particularly important for decreasing repetition and preventing dropout among working students, rural students, or other groups of students who are more likely to miss large segments of the school year.13' Nicaragua: Automatic Promotion and Overage Primary. In 1998 Nicaragua began a policy of automatic promotion from the first to the second grade. This was extended to the second to third grade transition the following year. The Ministry of Education has also adopted a policy of placing the best teachers into the earliest grades to attempt to curb the early roots of student failure, repetition, and lack of learning. Oddly, repetition in the first two grades continues at rates similar to those of 1997, indicating that the policy has not yet been implemented as planned (see Chapter Iand MECD website). Over-age Primary in Nicaragua allows children and teens age 9-15 to complete primary education in three years instead of six. In many cases these classes are offered in regular primary schools but during an evenindnight shift so that normal-aged students are not present and so that studentscan work during the day (MECD website). b) Conditional Cash Transfers Conditional cash transfer programs, such as those in Mexico and Brazil, have proven effective at increasing attendance in school, educational attainment, and boosting enr01lment.l~~ These demand-side programs give money to families - usually targeted to poor families only - on certain conditions such as a target attendance rate in school or a minimum number of doctors' visits per child per year. They address both direct and opportunity costs of schooling. Conditional cash transfer programs have limited applicability for the primary level in Central America because they could easily be prohibitively expensive. They may be more appropriate for the secondary level and are discussed further in the chapter on secondary. However, CCT programs have been applied in Honduras and Nicaragua, with varying levels of success (Box 2.3). In general, they had a positive impact on school attendance and early drop-out in both countries, while their effect on enrollment was more ambiguous. There i s evidence that the CCT program of Honduras was poorly designed and implemented, while the program in Nicaragua was better developed. Poor quality of the supply can also constrain these programs, requiring coordination with supply-side interventions. Conditional cash transfer programs are not designedto generate quality improvements in schools and there is no evidence that they do so. I3OKraft (2004). 13'See, for example Skoufias and Parker (2001) on Progresa (Opportunidades) in Mexico, and Bourguignon et a1(2002) on Bolsa Escola (Bolsa Familia) in Brazil. 88 Box 2.3: Conditional Cash Transfers Can Help Increase c) Cross-SectoralInterventions Human Capital Nicaragua's Social Protection Network (RPS) gives Many of the student background poor families cash transfers in exchangefor attending constraints to learning identified in health workshops, bringing their young children to this chapter are most directly dealt regular healthcare appointments, and keeping their 7- with cross-sectorally. If students are 13 year-old children enrolled in and attendingschool. Initstargetpopulation,RF'S hasresultedin: unable to learn and come to class regularly because they are 0 18% increase inprimary school enrollment; frequently sick and malnourished, 0 23% increase in school attendance; then an excellent policy option i s to 0 7% increaseinprogressionthrough grades; ensure that communities have access to clean water, medical care, 0 5% decreasein child labor (age 7-13); safe homes, and other basic 0 Improvedchild healthandnutrition; requirements for life. 0 Greater improvements for the extreme poor and families in areas going through economic downturn. Education ministries in some countries internationally have paired (Source: Maluccio, 2004) up with ministries of health or other bodies to offer nutritious meals in InHonduras, the PRAF (Programade Asignaciones schools to attending students. This Familiares) implements a program to improve can address problems of hunger and Honduran population's human capital through malnutrition and can also create an encouraging poor familiesto useeducation, healthand incentive for children to go to nutrition services, as well as improving the quality of such services. The program consists of several school and for parents to keep their interventions, including demand-side interventions kidsin school. which comprise a set of vouchers for nutritional status and school attendance (for children attending school d) Decreasing the Private Direct until4" grade-Le, 6-12 years of age) [85,000 families Costs of Schooling covered]. A recent impact evaluation shows that, in its target population,PRAFhas resultedin: One straightforward way of 0 A growth of 4.5% in assistance rates; relieving barriers to greater demand for schooling among disadvantaged 0 A reductionin early drop-outs; population groups i s to decrease the 0 No significant impact on enrolmentrates private costs of schooling which, (Source: IPFRI, 2003) especially for poor and very poor families, can not only be a discouragement for keeping children in school, but can be prohibitively expensive. Tables 4.5 and 4.6 in Chapter lV estimate private spending for primary and secondary by income quintiles. They show that a family in the lowest income quintile would have to spend over a third of their monthly income to keep two primary level and one secondary level child in school in El Salvador. In Guatemala the same family would have to spend 85 percent of their monthly income. Eliminating school fees, fees for exams, and providing free textbooks could considerably alleviate some of these private direct costs. Nicaragua has recently eliminated mandatory secondary school fees and i s attempting to enforce the voluntary-nature of primary and secondary school fees. For comparative 89 purposes, Kenya also recently abolished school fees at the primary level and the response from families was massive.'32 e) Publicity Campaigns Finally, publicity campaigns of various natures can also increase demand for quality schooling. Radio announcements, billboards, television commercials, and town meetings are a low cost means of increasing awareness about the importance o f education. These campaigns can address demand constraints in terms of enrollment but also in terms of learning by encouraging families to enroll their children in school, giving important information about exams, enrollment, or school, or providing tips on how to support children's academic success in school. C. TEACHER QUALITY POLICYIMPLICATIONS a) Pre-Service and In-Service Teacher Education Policy Implications As seen above, there is substantial room for improvement of pre-service teacher education in the region. This improvement could be focused along four main dimensions: (a) reducing the gap between required and de-facto academic qualifications for teaching; (b) developing selection criteria and diagnostic tools such as accreditation exams, which can contribute to increase teacher quality while rationalizing teacher supply; (c) improving the quality of teacher preparation; and (d) reducing the gap between demand and supply of teacher specializations. Resolving the first gap will require addressing the lack of teacher incentives to work in rural areas, which ends up leading to the hiring o f insufficiently skilled local teachers. The development o f selection criteria may also imply moving towards a system of full teacher certification in all countries, as is currently done in El Salvador and Costa Rica (in fact, both Guatemala and Nicaragua have plans to transfer teacher preparation schools to the tertiary level), although this policy may be an expensive one if salary increments are directly related to academic title. To improve the quality of pre-service teacher education involves ensuring talented committed teacher- educators, updated and relevant curriculum and resources, and adequate opportunities for practicum. Finally, addressing the demand and supply gap will require a careful assessment of the relevance o f subjects taught to the current socio-economic and cultural context, with particular focus on the most disadvantaged student groups. In terms of in-service teacher professional development, improvements will require, at a minimum: (a) putting more effort on changingupdating classroom practices and taking into account non-standard education contexts (multi-cultural environments, rural areas, etc); (b) undertaking more formal evaluations of the effectiveness of professional development programs; (c) providing more follow-up with teachers post-training to support the often challenging application o f what they learned; and (d) improving even further the programs' flexibility. 132 See,for example, The World Bank (2004e). 90 b) TeacherEffort Policy Implications Two main policy implications emerge from the findings on teacher effort. These are (a) future salary increases should be more accurately tied to actual improvements in teacher performance and effort such as hours worked, student test score gains, or improvements in teaching skills; (b) second, attention should be paid to avoid a possible income effect in which teacher wage increases result in a diminishment of hours worked. Two sorts of mechanisms that could be explored are merit-pay schemes or salary scales that promote higher effective hours of work. A limitation of merit-pay schemes lies in the "distorted" use that could be made of student achievement scores (to obtain higher students scores, schools may be encouraged to select their students or cheat) and in "moral hazard" issues which are raised by having teacher pay dependent on outcomes not entirely under the control o f teachers. Using changes in test scores rather than raw test scores as well as intermediate quality/efficiency outcomes, such as i s being done in El Salvador, could help address these issues. A limitation of salary scales promoting higher effective hours of work i s that they rely on teachers' assessments while these are rarely done. Using simple increases in the hourly wage depending on the total number of hours worked may be a simpler mechanism than salary penalties. Additionally, it would also be important to analyze the feasibility of other incentive mechanisms, such as effective teacher monitoring systems. Finally, it also appears important to continue (or introduce in the case of Guatemala) salary incentives for disadvantaged geographic areas as a policy to attract high-quality teachers in these areas. D.PEDAGOGICAL PRACTICE POLICYIMPLICATIONS a)Active Participative TeachingStrategies Policy Implications Education reforms have probably taken the right direction in introducing constructivist curricula but effective implementation of this type of curriculum is fairly demanding both financially and technically. Teachers need to be prepared to teach new materials in new ways and then supported as they make those changes in their practice; a wide range of interactive educational materials needs to be available; classes may need to be smaller; etc. Central American countries do not seem to have fully prepared themselves for this challenge. It is, therefore, essential to ensure that the Ministries of Education be fully aware of limitations in effective implementation and commit themselves to adequate implementation. Without this, it is unlikely that these reforms will succeed. A particularly important set of measures will refer to teacher capacity in executing the new curricula. Possible measures to improve teacher capacity will have to include greater emphasis on classroom practices in both pre-service and in-service education and innovative ways of organizing in-service teacher support such as the creation of spaces for teachers to share new ideas and methodologie~'~~,following, for instance, the "teacher microcenters" concept being applied in Nicaragua (see Box 2.1), or an effective 133Teachers Training Teachers (TTT) programs have had effective success around the world in providing supportfor teaching. 91 system of school supervisors which ensures that teachers obtain continuous support in the application and ongoing application of new method~logies.'~~ Finally, it is also important to ensure that curricular reform, standards, training and supervision be implemented at the secondary education level as well as at the primary level. Most countries have tended to leave secondary education aside in their education reforms with negative consequences on the quality of the secondary cycle. This shortcoming also ends up jeopardizing the transition from primary to secondary. It i s particularly imperative to improve the quality o f grades 7 to 9 in all countries. The move towards incorporating lower secondary education into the definition of `basic education', undertaken in El Salvador and, more recently, Honduras, goes in the right direction by shifting the focus from grades 1 to 6 to grades 1 to 9, and supporting a number of accompanying measures (integrated curriculum for the three cycles, introduction of standardized testing at grade 9, etc). More, however, i s needed. b) Instructional TimePolicy Implications The reasons for low effective class hours vary somewhat across countries. With only 180 class days, the amount of official class hours itself seems to be a constraint in Guatemala and Nicaragua. InHonduras and Guatemala, teachers' absences and school closings seem to be particularly problematic. In all four countries the relative proportion o f teacher's time spent on teaching i s too low. What can we suggest to increase effective class time? To start with, Nicaragua and Guatemala would be well advised to follow the example of El Salvador and Honduras and introduce an official 200-day school year. Second, and importantly, more effective and reliable means should be used to record teachers' absences and school closings at the school level, through, say, continuously-updated schools' registers. Duly completing the school register should be made compulsory in all countries. Third, an efficient monitoring system of teachers' absences should be put in place, involvingthe department and district level.135 Fourth, communities themselves should be fully involved in monitoring teachers' absences. As we will see, communities are already involved in community- based schools (Chapter V). They could also have a more responsible role in traditional schools. As a minimum, community associations should record teacher absences and report them to district offices which should then take some form of disciplinary action. El 134These two last measures may, in some ways, appear as an illustration of two different approaches, one more compatible with flexibility and pedagogical decentralization, and the other with result orientation and pedagogical centralization. Countries have indeed followed preferably one or the other approach, with some having introduced very centralized frameworks where the Ministries of Education determine how the curriculum will be taught to the students (this is, for instance, the case in Cuba), and others where teachers are given substantial autonomy on how to teach the curriculum (such as Chile for instance). These a proaches do not however need to be incompatible with one another. lg Honduras, for example, is in the process o f strengthening its department and district offices and delegate to them the management of databases on human resources. Monitoring teachers' absences should also be part of the functions of the newly strengthened offices. Deconcentrationof, say, the teacher payroll can also reduce teachers' absences directly by the introduction of a more efficient payment system which does not oblige teachers to travel to get paid. 92 Salvador already has had some success in monitoring teachers' absences through school governments. Finally, there should be a reallocation of teachers' time towards more time spent on teaching itself, trying to limit, for instance, time spent on school administrative activities. Improved directors' leadership and higher community involvement should help reduce the time teachers spend on these activities. IV. Conclusions and Policy Implications The overarching conclusion that can be extracted from this chapter i s that, in spite o f on- going improvements, there are still serious barriers to learning in Central America. Five main general constraints to student learning have been identified: (i) insufficient culture of evaluation, assessment, and accountability; (ii) demand constraints associated with student background characteristics; (iii) insufficient and poor teacher preparation and education; (iv) low teacher effort and insufficient instructional time; (v) weak execution o f the new constructivist curricula; These conclusions disproportionately affect the poor as revealed by the analysis o f effective schools. It is difficult to suggest concrete reform paths to address these constraints without further examination o f each of the countries' contexts, but we can at least summarize some o f the policy implications developed in the chapter, trying, when possible, to address their application to specific countries. Some suggested recommendations to address the constraint "insufficient culture of evaluation" include: (a) increase the level of institutionalization of the assessment systems, by making sure that academic assessments are carried out routinely and have the necessary financing from the Ministry of Education (an external institution could also work if the two above conditions are fulfilled). It would also be advisable to start some evaluation at grade 1, where grade failure i s very frequent; (b) develop effective information dissemination mechanisms at all levels. Report cards to disseminate information on school performance to different audiences could be particularly promising; (c) improve the timely alignment of the tests with standards and curricula. It would also be worth trying to align national standards to the regionally established ones to provide the basis for a well-functioning regional accreditation system (see below); (d) establish a well-functioning Central American accreditation system, which could also lead, if necessary, to the implementation o f a regional assessment system. However, the costs and benefits of establishing such a system should be assessed. 93 Participating in international exams may be sufficient to promote higher levels of quality. Some suggested recommendations to address demand-side constraints include: (a) where appropriate, alleviate the roots o f demand constraints. Some of the student background characteristics associated with lower demand for schooling and poorer academic performance are things such as parent illiteracy, malnutrition, and non-Spanish first language. Offering parent literacy classes, ensuring access to health care and sufficient food and clean water, and offering at least the first grades of primary education inthe primary language of students can remove many of the root barriers to enrollment and learning. (b) improve multi-grade schooling and improve and expand school-based management modalities. Chapter V focuses on SBM. The potential of multi-grade schooling could be more fully realized by making sure that teachers are specifically and adequately trained and experienced in multi-grade teaching techniques and that they obtain the on-going support necessary to successfully apply these methods. (c) consider conditional cash transfers as a means of boosting enrollment, attendance, and completion for disadvantaged sectors, particularly at the secondary level. (d) reduce the private direct costs of schooling. Central American countries should consider the feasibility o f eliminating some of the direct costs of schooling such as enrollment fees, exam fees, and fees for textbooks. (e) consider using publicitv campaigns to expand awareness and acceptance of practices linked to higher enrollment and learning (such as having reading material available inthe house and reading to children). Some suggested recommendations to address the constraint "insufficient and poor teacher preparation and education" include: (a) reduce the gap between required and actual academic qualifications for teaching (which will largely require to address the shortage o f skilled teachers in rural areas - see below); (b) introduce effective policies to provide incentives for skilled teachers to work in rural and low-income areas. This will be particularly important in Honduras, where some monetary incentives to move to disadvantaged areas have been introduced, but the disparity between levels o f qualifications in urban and rural areas remains large and i s detrimental to low income schools. An effective policy to attract skilled teachers (more experienced) to rural or low-income areas will also be particularly useful in Nicaragua to increase the number of effective 94 schools. Overall, existing monetary incentives, while necessary, may not be sufficient in and of themselves or may not be sufficiently large. Other measures, such as the obligation for all teachers to spend a period of time in rural areas (as has been introduced in Honduras) or a teacher monitoring system, which limits unjustified teacher transfers between urban and rural areas, may be useful; (c) develop selection criteria (in terms of academic achievement) and accreditation exams (to become a teacher), which may require a move towards a system of full teacher certification in all countries, as i s currently done in El Salvador and Costa Rica; (d) reduce the gap between demand and supply of teacher specializations, undertaking a careful assessment o f the relevance of subjects taught considering the socio-economic and cultural context of schools, with particular focus on the most disadvantaged student groups. This will be particularly important in Guatemala, where gaps are very substantial between rural and urban areas and ethnic groups, and not all these differing needs are still addressed. Teacher education for multigrade teaching should be strengthened in all countries that have multigrade schools. Some suggested recommendations to address the constraint "low teacher effort" and "insufficient instructional time" include: (a) avoid decreasing work hours as a primary result o f teacher salary improvements; (b) develop mechanisms which explicitly link salary increases to teacher performance and effort, analyzing the feasibility of introducing team-based merit-pay schemes such as in El Salvador (which, however, still needs to be fully assessed) or salary scales which promote higher effective hours of work; (c) analyze the feasibility and timeliness on a large scale o f other incentive mechanisms such as: (a) strong inspection systems; (b) decentralized systems of teacher monitoring; and (c) fixed-term contracts or local authority to hire and fire teachers. The results o f the effective schools analysis suggest that parental participation in school management may have very good potential for improving teacher effort (see Chapter V). (d) introduce an official 200-day school year inNicaragua and Guatemala; (e) develop better ways o f recording teacher absences and school closings at the school level, through, continuously updated schools' registers for example; (0 implement an efficient monitoring system of teachers' absences, by involving the department and district level, as well as the communities themselves organized in school governments; 95 (g) reallocate teachers' time towards teaching itself, trying to limit, for instance, time spent on school administrative matters (which could be to a large extent decentralized to directors and parents). Teachers should be granted the necessary autonomy and support within the school to perform their pedagogical tasks. Some suggested recommendations to address the constraint "weak execution of the constructivist curriculum" include: (a) put more emphasis on classroom practice in both pre-service and in-service training; (b) develop innovative ways of orpanizing in-service teacher support such as the creation of spaces for teachers to share new ideas and methodologies, following, for instance, the "teacher microcenters" in Nicaragua; (c) develop an effective system of school supervisors which ensures that teachers obtain continuous support in the application of new methodologies, while making sure that they continue using these methods (the system of "asesores pedagdgicos" of El Salvador should be assessed). Principals should be more actively used in a role of academic and pedagogical leadership; (d) ensure that curricular reform, standards, training and supervision are implemented at the secondary level as well as at the primary level. 96 Chapter 111: Expanding Secondarv Education in Central America With the dramatic increase in primary school enrollment rates across Central America, countries are turning their attention to secondary schools. This attention i s the result o f a convergence of factors including increased societal demand for secondary education, growing acknowledgement of the importance of post primary education for personal and national development, serious challenges in secondary coverage and quality, and some fiscal space for investing more resources in secondary now that many o f the fixed costs for primary are largely in place. At the same time secondary education faces challenges of access, equity, quality and relevance as was described in Chapter I. Addressing these challenges is one of the most urgent needs in Central American education. In deciding policies for secondary expansion and improvement countries' first major task i s to determine the primary objectives o f secondary education in their country. The objectives of secondary education as posited by Ministries of Education in Central America revolve around two major themes: the first i s to equip youth with knowledge and skills to support national economic and technological growth and, the second, to prepare students to successfully enter the labor market. This chapter begins by briefly presenting why it i s important to expand secondary education in the region. Second, the chapter examines the reasons for low coverage and options to expand it, including why investment of more public monies in the cycle i s crucial. A final section offers cost simulations to reach target secondary enrollment rates inEl Salvador and Honduras. I. Why is Expanding Coverage of Secondary in a Broad-based Manner a RegionaI Priority? Chapter Ishowed that there i s a severe coverage gap in Central America at the secondary level. Gross enrollment rates in the four countries range from 33 percent in Guatemala to 61 percent in El Salvador. These rates are well below those of comparison regions with similar per-capita income. This coverage gap, paired with quality and equity problems in secondary education in the region, create a major stumbling block to individual opportunity and national development. Secondary education helps individuals improve their lives and it also helps countries grow and develop. The rate of return analysis in Chapter Iaffirmed that people who go to secondary school, particularly upper secondary school, earn significantly more than those who do not. These additional earnings allow individuals to improve their health, nutrition, and overall quality of life. But recently, research i s showing that secondary education also benefits national growth and development in meaningful ways. A population with full secondary education is key to boosting national competitiveness and GDP in an increasingly technological Secondary enrollment in developing countries has been positively related to GDP levels over the past three decades and mass secondary schooling i s critical to the ability o f countries to attract and benefit from trade I3'See Fuller and Holsinger (1993) and The World Bank (2003). 97 and foreign direct investment. As good jobs increasingly rely on skilled labor, an undereducated population will fail to attract new investment. Employment opportunities in countries with low or highly skewed education levels are likely to only exacerbate existing inequalities and face severe skill bottlenecks. On the other hand, skilled workers are much more productive than unskilled workers in today's technologically advanced firms. Skilled labor generates a beneficial cycle in which higher-skilled workers, by being more productive, generate greater profits for firms, who reinvest a portion of those profits in higher salaries, thereby increasing the assets of workers who, in turn, can invest more in developing more skills. There is, therefore, both a direct link between the relationship o f education and wages at the individual level (as captured by private rate of return analysis) and the relationship of education and aggregate income at the country level. What is more, this skill upgrading should have a larger effect on growth rates now than before because of the increasing need for skilled, educated labor. Thus, countries with low levels o f education remain in a trap of technological stagnation, low growth and low demand for education. Looking over time, secondary enrollment in developing countries has been positively related to GDP levels over the past decades.'37 The high returns in upper secondary education in Central America suggest a combination of high demand for secondary graduates and low supply o f secondary graduates. The demand for educated workers i s likely to only increase in Central America with the CAFTA (Central American Free Trade Agreement) because reforms to liberalize trade regimes, encourage FDI and facilitate licensing of technologies may further unleash the demand for schooling. CentralAmerican countries have pursued lopsidededucation development paths. Figure3.1: Illustration of Central The most successful examples of nations AmericanEducationDistribution increasing their populations' education level, notably in East Asia in recent years and the United States some decades ago, expanded their education systems one level at a time - first ensuring that there was broad enrollment in primary, then in secondary, and finally in tertiary.13* Latin America, on the other hand, has for the most part not followed this approach. Instead, many Latin American countries have expanded andor invested heavily in the higher levels of education without first ensuring that there was broad access to and completion o f lower levels. A balanced transition is necessary because it minimizes inequalities, ensures that the most gifted students go to university, and takes advantage of the I3'See The World Bank (2003) and Loayza, Fajnzylber and Calder6n (2002) on Central America. This second study, by regressing the change in the average per capita growth rate between the 1980s and the 1990son a set of structural and cyclical variables, finds that changes in educational attainment, measured by the gross enrollment rate in secondary education, significantly contributed to the differential growth rate inCentral America (about 15% of the projectedper capita growth rate between the 1980s and 1990sis explained by education). 13'See The World Bank (2003). 98 complementarities between the phases in an educational transition and those in the technological transition of a ~ 0 u n t r y . l ~ ~ Central American countries represent no exception to this pattern of limited secondary enrollment and, subsequently, inadequate skills. Costa Rica and El Salvador have education distributions that resemble a "lopsided anvil" where the bulk of the population have only primary schooling, and fewer individuals have secondary schooling only than have tertiary level schooling (see Figure 3.1).l4' In Nicaragua, most people have only reachedprimary school, very few have reached lower secondary, and more have reached upper secondary and tertiary. Finally, inHonduras and Guatemala, the largest share of the population has only reached primary and similarly fewer people have reached lower secondary, upper secondary and tertiary education. East Asian countries, on the other hand, have a distribution resembling a diamond in which more adults have secondary schooling than primary or tertiary. This diamond shape is more desirable for a developing country in that it represents less educational inequality and indicates that most of the population has, at the least, attended secondary. II. Constraintsto Equitable Secondary Coverage There is no one single obstacle blocking greater secondary coverage. Rather, there are multiple different causes, ranging from students' need to work and constraints such as high private direct costs of schooling, to students'lack of interest in attending secondary due to low quality and relevance of schooling. To a more limited extent there are also supply shortages, particularly in rural areas. This section combines findings from household surveys, survivor function plots, rate of return analysis, and education production functions to investigate some of the reasons for low secondary enrollment figures in Central America. Table 3.1 summarizes the primary responses from household surveys in three countries regarding why teenage students do not attend scho01.l~~These responses are a good starting point from which to discuss the main reasonsfor limited secondary coverage. 139Latin America i s the most unequal region in the world and unbalanced transitions between education levels can perpetuate inequality. An unbalanced transition, with low secondary education levels, would relegate the bulk of workers to low-skill, low-wage jobs while a privilegedfew earn the high wages that come from a university education. Because of a close relationship between schooling and earnings, an unequal distribution of education tends to perpetuate an unequal distribution of earnings. 140This figure is adapted from The World Bank (2003). 14'Household surveys provide useful insights on the reasons for low coverage. These surveys ask families why teens are not in school. Care needs to be taken, however, when analyzing this type of information. Firstly, replies are subjective and, as such, are subject to bias, mistakes and miss-interpretations. Secondly, a thorough study of demand relationships can only be done when there is availability of supply (we could then observe the revealed preference of a family that has access to schooling). If supply is not available, it will be more difficult to disentangle the true reasons for non-attendance and the responses obtained may under or, more likely, over-estimate the weight of demand-side factors. Also, information is only available on reasons for non-attendance by age, providing an incomplete explanation of reasons for non-attendance by grade (which the comparison between the survivor plots by grade and age could somewhat illustrate). Finally, this question addresses non-attendance in school rather than non-enrollment. Non-attendance most likely includes both non-enrollment as well as school absences. 99 Source: Household Surveys. (a) Includes no place available. Economic constraints: private costs of secondary. Household surveys suggest economic constraints are a main determinant of secondary school non-attendance. In El Salvador and Guatemala roughly one in four families report that the cost of schooling prevents youth from attending school. Indeed the private costs of secondary education in Central America are markedly higher than other education levels (see Tables 4.5 and 4.6 in Chapter IV and Figure 3.2). Sending one child to public secondary school in 100 Guatemala costs, on average, almost 50 percent of the annual income of a poor household and 15 percent of the annual income of a middle-lower income household. However, in all four countries, families with multiple school-age children may face prohibitively high private costs of schooling (in the private but also in the public sector). Figure 3.2: Private Costs of Secondary are Significantly Disaggregation of the household Higher than the Private Costs of Primary surveys by urban and rural location (not shown here) reveals that urban families are more likely to cite schools as "too expensive'' than are rural families. This may have to do with the large proportion of private 300 secondary schools throughout 250 Central America inurban areas. 200 I I 150 Need to work. Relatedly, the declared need for youth to work, 100 whether at home or in the formal so or informal labor markets, is one 0 of the most significant factors for non-attendance, both for urban Source: Table 4.5. and rural areas. Predictably, its importance increases, as youth get older. Survival plot functions (Figures 3.3 and 3.4) suggest that dropout rates between education cycles and mid-cycle dropout rates in upper secondary, which are generally quite high, may be related in part to age/work reasons142.The need to work is much less relevant before age 13 in all countries. Disaggregation by income quintiles reveals that the need to work at home is more prominent in the poorer quintiles, while the need to work in the labor market i s basically equivalent. This surprising last finding may hint at the fact that youth pertaining to the wealthier echelons have an easier time finding employment. Although wealthier students have higher private rates of return from secondary education (see Chapter I), may they also have a higher opportunity cost of attending, in particular in the urban area, if they have access to relatively well-paid jobs. B y contrast, youth from the poorest quintiles may have lower private rates of return but face less enticing immediate earning prospects. Certainly, however, the earnings contributions of youth from poor and very poor families i s likely to result in a larger proportional contribution to family income and the higher declared need to work in the poorer quintiles (summing both domestic and labor market work) probably reflects this. 14*The average age of completion of primary is between 13 and 13.5 in all countries, which is precisely when children start dropping-out of school for work reasons. As age increases, drop-outs increase even more. 101 Figure 3.3: Comparative survivor plots by grade SURVIVOR FUNCTION PLOT BY YEARS OF EDUCATION ElSalvador 4 - + 1 n 0 9 . . . . . . > . . . . ~L ........ ......................... Honduras !0 I 1....... + 1.--.-.- - !s 0.8: I: ___7 5 : Nicaragua 0.7{ ------ -.... 1................. -_---I r---- 9 E 1 ! 0.6 : 1 ......... I. ........ ,~ 0.5 1 x ,_ i ~. ~ I i.. ..... 0 1 2 3 4 5 6 7 8 9 10 Figure 3.4: Comparative survivor function plots by age SURVIVOR FUNCTION PLOT BY AGE ElSalvador ....... _ _ _ _ I U ~ 0.8 - 1T.T 1.7 I.. ..... A : ........................ : E 0.7- .......... ......... ......- .......... --_-- Nicaragua ........... r----i j-Z 0.6- g !0.5- I.. ...... 2 2 1 . ... 0.4 - 0.3 - 0.1 . . . . . , . . . , , . . . . , . . , , , , , , , , , . . . , , . . ~, . . . . , , , , , , , , , . 7 8 9 10 11 12 13 14 15 16 17 &ad Source: HouseholdSurveys. Another, more analytical way, of analyzing the determinants o f enrollment and completion i s through the estimation of demand and completion functions. We undertook this exercise in the case of El Salvador. Without spending time on methodological 102 caveats, we report the main results of this exercise in Box 3.1 be10w.l~~ this analysis In socio-economic status i s the first determinant of enrollment and completion (proxied by parents' education, dwelling characteristics and household income). Lower socio- economic status, as seen above, is associated with lack of money to attend school, need to work in the household and lower access to quality education. Additionally, we can also see that there i s a positive but decreasing impact of age on enrollment and completion: this hints at the fact that while older children are more likely to be enrolled in secondary than younger ones (due to repetition, etc), it is also true that, as they become older, their probability of not being enrolled increases due to increasedopportunity cost of staying in school. A widespread lack of interest in attending school points to major obstaclesin terms of the quality and value of attending lower and upper secondary. Another of the most prominent reasons families state prevent school attendance is lack of interest. While this i s clearly a subjective opinion whose root cause i s impossible to determine, it i s likely that much of the lack of interest is due to a belief that schooling is not worth the time, monetary, and energy investment it requires. This i s tied into questions of low rates of return to schooling, little relevancy to students' lives, and poor quality of schooling. The education production function on school enrollment and completion in El Salvador (Box 3.1) supports these findings. School quality, as proxied by pupil-teacher ratios, class size and public spending per child is a particularly important determinant of enrollment and completion. Survival functions show that student dropout i s particularly acute at the lower secondary level (Figure 3.3). There are pieces of evidence which point to an acute problem in education at the lower secondary level. First, lack of interest is the most important reason for non attendance in the 13-15 year old range, age of transition between primary and secondary. Secondly, low relevance and quality are also confirmed by rates of return and achievement scores particularly low at that level. As the opportunity cost of attending school increases, for the 13-15 year old range, low quality and relevance constitute another disincentive for attendance. This i s true for all countries, and is, in fact, most frequently cited in household surveys in El Salvador, the country that has put significant effort into improving its lower secondary cycle. This indicates that despite efforts and relatively high transition rates from primary to lower secondary, even El Salvador has to pay close attention to ensuringthat students enter, and remain in lower secondary. Lack of interest continues to be a problem in the 16-18 year old range and the survival plot graph (Figure 3.4) shows that as youth get older they are more and more likely to be out of school. The survival plot by years of schooling shows that transition from lower to upper secondary is also problematic. Nicaragua i s the only country that does not have a large lower to upper secondary transition gap, perhaps because both cycles are generally offered in the same schools. In the other countries, where lower secondary schools and upper secondary schools are often separate, there appears to be challenge in facilitating student transition to upper secondary. This is true despite mid-cycle dropout in lower secondary beingrelatively low in all countries (less so in Nicaragua). '43Methodological limitations include biasesproducedby unobservable factors, measurement error, etc. 103 Box 3.1: Determinants of enrollment andcompletion at the secondary level inElSalvador Enrollmentand completionfunctions,2002 Completionfunctions,2000 Net Net Net Completion Enrollment Enrollment Enrollment basic>=15 Secondary 16-Secondary2 Secondary3 Completion Completion 18 16-18 16-18 basic, 15-21 secondary, 18-21 age2 20.3884*** 20.1429*** 20.1864*** 9.8564*** Householdincomeper capita(1og) 0.1929*** 0.4240*** (4.4793) (4.4849) (4.4823) (2.0140) (0.0486) (0.1188) age2 -0.5946*** -0.5874*** -0.5886*** -0.2693*** Fees 0.0037*** 0.0015* (0.1315) (0.1317) (0.1316) (0.0605) (0.0008) (0.0008) Female 0.1950 0.1815 0.1929 0.1244 Drinking water 0.3459*** 0.5013*** (0.1294) (0.1294) (0.1294) (0.1171) (0.0786) (0,1274) Birth Order -0.3392*** -0.3276*** -0.3377*** -0.3301*** Electricity 0.8075*** 0.6046*** (0,0581) (0.0577) (0.0580) (0.0533) (0.1253) (0.2313) Father'sEducation 0.0581*** 0.0591*** 0.0577"** 0.0707*** Floor 0.5356*** 0.7147*** (0.0203) (0.0205) (0.0203) (0.0185) (0.0859) (0,1482) Mother's Education 0.1146*** 0.1154*** 0.1147*** 0.1376*** Sanitation 0.6050*** 0.4541 (0.0208) (0.0210) (0.0209) (0.0184) (0.1902) (0.4133) Public 1.5750*** 5.1496*** 1.0625 0.3997 4 s 4.7000*** 6.9929*** (0.5432) (0.6343) (1,3504) (1.1422) (0.3879) (2.0642) HHHeadFemale -0.4127 -0.4010 -0.4133 -0.5103** 4ge squared -0.1200*** -0.1627*** (0.2664) (0.2611) (0.2657) (0.2460) (0.0107) (0.0527) Mother Self- employed 0.4491*** 0.4627*** 0.4552*** 0.3645*** Female 0.3657*** 0.2417** (0.1514) (0.1513) (0.1516) (0.1349) (0.0684) (0,1035) HHIncome pc (log) 0.7470 0.6951 0.7432 0.3948 Father's Education 0.1280*** 0.1158*** (0.5184) (0.5011) (0.5172) (0.4803) (0.0102) (0.0151) HHIncome pc squared (log) -0.0731* -0.0698* -0.0728* -0.0357 Mother'sEducation 0.0962*** 0.1330*** (0.0438) (0.0424) (0.0437) (0.0413) (0.0121) (0.0182) Lognumber children <=21 yr 0.6227*** 0.5976*** 0.6218*** 0.5279*** HH Head Female -0.1194 -0.3886** (0.1776) (0.1770) (0,1777) (0.1605) (0.1323) (0.1924) Dfloor 0.5878*** 0.5808*** 0.5836*** 0.6202*** Total number children e 2 1 yr -0.1278*** -0.1729*** (0.1951) (0.1949) (0.1961) (0.1525) (0.0191) (0.0347) Delect 0.2464 0.2069 0.2366 0.2957 Birth Order 0.5775* 0.2084 (0.2272) (0.2254) (0.2298) (0.2240) (0.2987) (0.3938) Dwater 0.4975*** 0.5114*** 0.4937*** 0.4918*** Single parent 0.0119 0.2682 (0.1691) (0.1677) (0.1699) (0.1469) -0.6761*** (0.1301) (0.1836) Dsanit 0.4958 0.5514 0.4912 0.6737** Fostering -0.6875*** (0.1212) (0.1856) Pupil-reacherrho, Basic -0.0271*** (0.0079) Subsidy Basic per child 7-15 0.0452*** (0.0154) Pupil-teacherratio, Secondary -0.1035*** (0.0234) Subsidy Secondaryper child 0.0366 16-18 (0.0364) 0.2514 0.1603 0.2087 0.2929** (0.1545) (0.1589) (0.1618) (0,1440) Zonstant -49.6571*** -78.4513*** Constant -180.5303*** -174.1645*** -179.2265*** -95.1737*** (3.5160) (20.2462) (38.1210) (38.1396) (38.1276) (16.7252) 3bservations 9209 4782 Observations 2715 2602 2602 3571 Robuststandarderrors inparentheses tobust standarderrors in parentheses Other (nonsignificant) variables included in the specification but not reported *significant at 10%;** significantat 5%; ***significant at 1% here include: parents'employment status and occupation, regional dummies *significant at 10%; **significant at 5%; ***significant at 1% 104 School availability is less of a constraint. A fourth constraint to secondary coverage that should be addressed is lack of supply. Pure supply-related factors, such as the lack of presence of a school or grade, account for only a minor share of the explanation for non- attendance in household surveys, with the exception of the 13 to 15 age range in Nicaragua. This may illustrate the fact that school availability is not a major issue anymore in Central American countries. In El Salvador, where lower secondary grades are now part of the basic cycle and are taught on the same campus as primary school grades there is probably very little supply-constraint at the lower secondary level. Meanwhile in Nicaragua, the large number of incomplete primary schools in rural areas may explain the more acute supply constraint finding in the 13 to 15 age range. The survival plots suggest that supply shortage (or lack of cheap public supply) may also be an issue in Honduras, where there are particularly low education cycle transition rates. On the other hand, supply constraints may be under-estimated by the household surveys. Since these surveys give no actual information on school supply other responses such as lack of interest may camouflage actual access problems and economic constraints may reflect limited availability of low-cost schools. Certainly at the upper secondary level there are enough spaces for only a fraction of the number of age-appropriate youth. The education production function analysis of El Salvador demonstrates that school supply, as proxied by the number of classrooms per children, is an important determinant of school enrollment and completion at the upper secondary level (public spending per child may also proxy for availability of supply). The fact that school availability is hardly cited as a problem in the 16-18 year old range suggests that the household surveys are, indeed, under-estimating supply constraints. As other constraints are relieved the lack of sufficient supply of low-cost lower and upper secondary schools may become an increasingly prominent problem. Disaggregation by urban and rural location predictably demonstrates that school availability i s more of a probleminrural areas than inurban ones. All of these factors are likely to disproportionately affect the poor. In Chapter I survival plots by income quintile illustrated that transition rates between cycles and mid- cycle dropout rates are worse for lower income groups. Many factors are likely to contribute to this troubling pattern including greater repetition rates and proportions of over-age students, more health related factors such as early pregnancy, more supply constraints in terms of quantity and quality of services, and greater need to contribute to family income. Even worse, the private rate of returnanalysis inChapter Ifurther showed that when poor students do stay in school their returns to schooling are lower than students from higher socioeconomic groups. In sum, there are multiple factors contributing to inadequate and inequitable enrollment in secondary schools in Central America. These factors include high private and opportunity costs of schooling, low education quality, and insufficient supply of schools in certain areas. Because of the diversity of causes of inadequate enrollment policy 105 responses will also have to be varied and multidimensional. The next section of this chapter explores these options. 111. Alternativesfor Increasing Secondary Coverage This section reviews four different areas of policy responses to the constraints discussed above. We look at (1) secondary financing, (2) demand-side interventions, (3) alternative delivery mechanisms, and (4) quality enhancing interventions. A. PUBLICAND PRIVATE FINANCING Public financing of secondary education should expand in Central America and ensure a favorable environment to equitable secondary coverage. Household surveys suggest that private financing (household expenditure in public and private schools) represents between 50 and 80 percent of total secondary education spending in Central American countries (see Table 3.2). High shares of private financing also translate into high shares in terms of supply. Indeed, between 20 and 55 percent of secondary enrollment is provided by the private sector in Central America (see Table 3.3).'44 Private schools in Central America are attended disproportionately by students in the top income quintile (see Figure 3.6) Table 3.2:Total Education Expenditure in Secondary Education by Source of Financing (US$, 000) El Salvador Guatemala Nicaragua 2002 2001 2002 Total educationexpenditure 219,000 215,000 52,200 Estimatedpublic expenditure 108,000 40,400 13,900 Estimatedprivateexpenditure 111,000 174,300 38,400 Percentof estimatedprivate exp. 51% 81Ya 74% Source: Table 4.2 and Household Surveys. '44These figures change slightly when considering official data of the MINED in all countries. For instance, the official 2002 figures report a public share of 67% and a private share of 32% in El Salvador in upper secondary. InHonduras, differences are more marked at the lower secondary level. It is likely that the private share be under-estimated in the Honduras household survey due to the lack of precision of the question which was asked for the first time in March 2003. 106 I Table 3.3: Public enrollment share in secondary education (%) El Salvador, 2002 Guatemala, 2000 Honduras,2003 Nicaragua,2001 7-11 7-9 lo-7-11 11 7-9 l,:- 7-11 7-9 lo-7-11 11 7-9 10- 11 Public 80 84 72 78 82 68 sector 46 55 24 73 75 66 (80) (86) (68) (75) (79) (68) Private 20 16 28 22 18 32 sector 54 45 76 29 25 34 (20) (14) (32) (25) (21) (32) Source: Household Surveys. Notes: Between brackets: Dataof the MINED, 2002 for El Salvador and 2003 for Honduras. Additionally, within the same public sector, per capita private expenditure on secondary i s higher than public expenditure in all countries for which we have data, with the exception of El Salvador (see Figure 3.9, which contributes to explain why public schools are attended predominantly by students from upper middle-income groups. The plots in Figure 3.6, which show what percentage of secondary enrollment i s from students ineach income quintile, suggest that El Salvador has recently made great strides toward achieving broad-based secondary coverage. Private secondary schools, in recent years, have become even more targeted to the upper two income quintiles but public schools are now largely attended equitably by the three middle-income quintiles. Nicaragua, on the other hand, which has a similar lower secondary enrollment rate to that of El Salvador, is still far from achieving equitable broad-based coverage. Despite progress toward greater equity since 1995, secondary distribution in Nicaragua i s still skewed in both public and private schools towards the upper income quintiles. A similar situation i s found in Guatemala and Honduras. Figure 3.5: Public and Private Expenditure per School-Age Population in Public Secondary Education I I I I I I 1 I 0 20 40 60 80 100 120 140 us $ Source: Household Surveys and Table 4.4. 107 Figure3.6: Publicandprivateenrollmentby incomequintiles El Salvador Guatemala 45 60 E40 35 - 8 50 5 30 40 25 B 20 --$,30 0 E, n' '"5 e 15 c i n 20 2e 5 10 0 I I1 111 IV V 0 IncomeQuintiles I II ill iv V -+public enrollment2002 +public enrollment 1995 Income Quintiles -.?-private enrollment2002 "+private enrollment 1995 /+Public enrollment*Pnvate enrollment1 Honduras Nicaragua ~ 60.00 50.00 40.00 30.00 r 20.00 E,20 22 i o 10.00 0.00 0 I II Ill IV V I II 111 IV V Income Quintiles IncomeQuintiles +Public enrollment2002 +-Public enrollment 1995 l+Public enrollment4 ~ r i v a t eenroilmentl -*Private enrollment2002 -*e,Private enrollment1995 Source: Household Surveys. The high share of private financing in secondary indicates the need for increased priority of secondary education in the national budget. Increased public financing could be used to expand traditional as well as alternative secondary models (discussed below), balance public-private costs in the public sector (by exempting low-income students from paying fees or by providing scholarships), and promote quality improvements. The case for publicfinancing and trade-offsamong regionalpriorities. There are two main reasons why we argue increasing public financing of secondary i s important. First, as discussed in Chapter I,private rates of return to lower secondary are low in all countries, creating a major disincentive for enrollment, especially when coupled with the increasing opportunity costs of attending school in the 13-18 age range. A public intervention should not be decided on the basis o f the private rate of return but on the basis of the difference between the social and private rates of return.'45 Social rates of 145On this point, see also Hammer (1996). 108 return are, however, very difficult to Yet the disconnect between low private rates of returnfor lower secondary and quantitative evidence that broad-based secondary education has a strong impact on the accumulation of human capital and national growth, suggests that social rates of return are higher than private ones and, therefore, without sufficient public intervention, a socially sub-optimal equilibriumlevel of lower secondary education will be reached. A public intervention can act in different ways to ensure the social optimum. First public financing can be used to implement policies that improve the quality and relevance of lower secondary encouraging higher demand from families and the labor market for this cycle.'47 Second, public spending can reduce the opportunity costs of schooling by providing demand-side subsidies and/or developing alternative secondary education modalities that make it possible to combine work with schooling. Second, as was shown in Chapter I, rates of return to secondary are particularly low for low-income groups, while direct private costs are particularly highfor these same groups, creating additional disincentive to attend for low-income people. Therefore, there is justification for additional public intervention on pure equity grounds. Public intervention may encourage higher participation of low-income groups in secondary education by implementing policies that improve the quality of the schools that they attend and/or reducing the burden of private costs through demand-side subsidies or other policy measures (other public policies such as improving the access of the poor to highly paid jobs would also be useful but are beyond the scope of this report). With low income levels and high debt levels, all four Central American countries face, however, strong budgetary constraints in the funding of government services and spending areas. Because of these fiscal constraints, there will necessarily be some trade- offs among the regional priorities laid out in this report (learning improvements, universal completion and broad-based secondary coverage). The nature of the interrelation of these three education priorities suggests an order of intervention: first, the quality of learning should be improved; next, universal primary completion should be assured; and finally, secondary education should be expanded for all population groups. While education quality should be the first and foremost priority for all four countries, the later two priorities could be addressed simultaneously or even in reverse order in the short or medium term if there is substantial room for improvement in the primary to lower secondary transition rate. Expanding secondary coverage can not happen overnight. But nor should this priority go ignored. At the very least, countries should expand lower secondary to accommodate all children finishing primary. Nicaragua, Guatemala and Honduras should aim at encouraging broad-basedexpansion of lower secondary by expanding equitable access to '46Social rates of returninclude private benefits and external effects (or externalities). j4'This approach is also supported by Glewwe (1996). He takes Ghana as an example and as he finds that there are low returns in middle schools (grade 7 -1) suggeststo increase investment, on quality grounds, to raise the rate of return. 109 secondary education for those children who are now completing primary.'48Additionally, all countries should make an effort to expand upper secondary education, although, with the exception of El Salvador, these efforts may need to be more selective to save on scarce resources. El Salvador, a country with higher and more equitable lower secondary enrollment, should now aim at expanding upper secondary (while also pushing for full transition of the poor and very poor to lower secondary). In addition to balancing these trade-offs, it will be essential to envisage strategies which allow increased budgetary priority in all (or at least three) of the four countries to the education sector and promote higher resource mobilization also from other sources (private sector, households). It will be equally important to introduce measures that encourage higher efficiency in the use of resources (more effective teacher management, higher shares of non-salary recurrent costs, re-allocation of spending across education levels within the education budget, cost-efficient delivery systems, effective demand-side interventions, lower repetition rates, etc). Greater private financing of upper secondary may be an appropriate short-term option in Guatemala, Honduras and Nicaragua. In Guatemala, Honduras, and Nicaragua Ministries of Education are not likely to have enough funds for the expensive tasks of expanding both lower and upper secondary in the short term. Because of this, these countries may wish to explore private financing options for upper secondary (involving civil society as a whole). Increased private financing may also be appropriate in that becausefew studentsare making it through lower secondary, limitedexpansion of upper secondary is a reasonable goal for the short term. Inaddition, private rates of return are high in upper secondary education, and may therefore make up for the higher opportunity and/or direct costs of schooling at that level. Highprivate rates of return may also suggest that the price elasticity of demand for upper secondary education will be lower than for lower education levels (in other words, higher direct schooling costs - as i s generally the case for private schooling - would have less of an impact on the demand for schooling at that level). Increasing private financing does not mean abandoning or cutting public support. Public support remains important for creating an environment that i s conducive to private sector participation and expansion, and for mobilizing private resourcesfrom companies, NGOs, etc. In addition, it is important to keep in mindthat as a larger and more diverse pool of students graduate from lower secondary schools, the government will need to take increasing responsibility to ensure wider access to upper secondary. In El Salvador, by contrast, greater public financing at the upper secondary level is a more immediate need given current lower secondary indicators. Efforts should, however, be made to minimize crowding-out existing private financing (in other words, upper- income families who are willing to pay for upper secondary should continue to do so). 148As the most disadvantaged groups of children begin to complete primary school in higher proportions due to reforms of primary quality and completion rates secondaryexpansion will becomeincreasingly more equitable. 110 B.DEMAND-SIDE INTERVENTIONS We have seen above that the reasonsfor low coverage go beyond school supply issues. In fact, school capacity may exist but students may still not go to school because schooling i s too expensive, they need to work, or they simply do not have any interest in a poor quality schooling system. All these reasons, we have seen, have a role to play. Two main options exist to address lack of money and need to work issues. The first option is to intervene on the demand-side by providing demand subsidies to families and students. The second option i s to intervene on the supply-side by improving school quality and relevancy or by providing flexible delivery options that make it possible to combine studying and working. Below we review possible demand-sideinterventions. Chapter I1discussed some demand-side mechanisms. Traditional scholarships that cover fees and other costs of schooling (uniforms, books, transportation, meals, etc) can be awarded to select (need-based or merit-based) students. Conditional cash transfer programs can provide cash to families or students subject to certain pre-established condition^.'^^ Targeted vouchers provide funds for families to select the school of their choice, usually including both public and private schools (see the Colombian case in Box 3.2).l5O These interventions generally impact demand directly but some can have an impact on supply as well, such as the voucher program, which, by generating competition among schools for student enrollment, can create greater school accountability to local communities as well as spur quality improvements. ation based on sound criteria, such as 1 disbursements, burden of program monitoring) led to the discontinuation of the program at the national this, however, survived at the sub- `49 "Oportunidades" in Mexico and "BolsaEscola" in Brazil are probably the most well known examples to date. A detailed description of demand-sidefinancing mechanismsinLatinAmerica if provided by Patrinos, H(2002): "A Review of Demand-SideFinancingInitiativesin Education", BackgroundPaperfor the WDR 2003. 111 -During Chile: the 1980s, following the introduction of the per-student subsidy, enrolment increased considerably inthe private voucher schools (from 14% in 1980 to 32% in 1990) and, controlling for socio- economic status, private subsidized schools seem to perform better than municipal schools, pointing to positive effects of competition on private schools' performance. However, it is generally studies that are largely centred on the 1980s period that competition among private and was hamperedby the lack of incentives and capacity of the municipal sector to compete budget constraints at the municipal level, restrictions on municipal decision-making and lack of autonomy pedagogical municipal school autonomy, introduc student attendance improve the operation of vouch central compensatoryprograms. To date there are few examples of demand-side interventions in Central America in secondary education. The PRAF in Honduras and the RPS in Nicaragua are conditional cash transfer programs applied at the primary level which have been briefly reviewed in Chapter 11. Scholarship programs exist in Costa Rica and El Salvador (see Box 3.3). FANTEL) and for secondary ated in 1999, and FEPADE has established Source: the World Bank (2004b) 112 C. ALTERNATIVE DELIVERYMODELS Increasingly, Central American countries are using innovative delivery mechanisms to increase the supply, quality, and cost-effectiveness of secondary schooling. This sub- section looks at basic education centers, public-private partnerships, flexible delivery programs, and school-basedmanagement. Mandatory basic education cycles can improve the transition from primary to secondary. Mandatory basic education cycles can improve the transition to lower secondary by incorporating primary and lower secondary schools on the same campuses, smoothing curricular and pedagogical transitions between the cycles, and generating public acceptance of lower secondary as a compulsory part of basic education. In Central America, this innovation has been piloted in El Salvador and Honduras where both countries have incorporated the full three years of lower secondary (grades 6-9) into the basic education cycle (see Figure 1.1 in Chapter I). Results remain mixed, with more favorable results in El Salvador than in Honduras. These reforms address lower-education supply constraints by bringing lower secondary to primary school campuses as well as quality constraints created by disconnects between primary and secondary curricula. They also have the potential of decreasingunit costs by combining schools although there may be significant initial costs.'51 There may be opportunities for public financing-private delivery schemes. Public- private partnerships are another policy option for increasing secondary enrollment capacity as Central American countries push to expand secondary coverage. Key factors to be considered in the decision whether to support public or private secondary expansion include the current existence of private secondary supply, the quality differential across public and private schools, the different capacity utilization of the two sectors, and the priority given to social heterogeneity and integration. Ifprivate schools in aparticular country tendto be higher quality than public schools and have idle capacity, then public subsidization of these schools may make sense. Inthe cost simulation for El Salvador later in this chapter we see that it would be less costly to expand secondary coverage by increasing or at least maintaining its current private sector share than by increasing public schools disproportionately. Many examples exist of public-private partnerships in which public funds support private education delivery. Charter schools in the United States, voucher-funded private schools in Chile, and schools run by private institutions and organizations on public school campuses in Colombia (Concessionary Schools) are a few examples. These methods can support private delivery of education to students regardless of their socioeconomic background. Publicly subsidized, privately run schools in Central America could increase secondary enrollment while improving the equitable distribution of youth in secondary if, for example, private schools were granted public funds to enroll low-income students without charging them tuition fees. This option could also foster more mixed socioeconomic distribution of students across schools and could potentially generate Lewin and Caillods (2001). 113 improved educational outcomes.'52 Alternatively, public subsidies could be provided to private schools to enroll middle class students at a subsidized rate. This could create new spaces in public schools for poorer students as middle class students shifted from public to private institutions. Both of these scenarios would increase secondary coverage though the first would have a greater per-pupil cost and better equity implications. In Central America, there are a couple of examples of public-private partnerships in Costa Rica and El Salvador, but these have limited coverage. The partnership approach developed in El Salvador consists of providing public resources to private organizations, mostly education non-governmental organizations or religious groups, to help establish and administer schools. In some cases, the Ministry o f Education pays teachers' salaries and the counterpart institution provides the remaining funds (infrastructure, materials, administration, etc). In other cases, this partnership involves contracting out the administration of the school, with MINED providing most o f the teachers and some material resources and parents providing the rest. Specific examples of these partnerships include the Fe' y Alegriu schools, the Dominicus de la Anunciutu Congregation, and the Muristus schools. In instances where these schools are located in poor neighborhoods and admit poor students public investment in these schools i s socially progressive. In other instances middle and upper income students enroll in the schools making public financing of these schools socially regressive. Public-private partnerships can also support the expansion o f flexible and alternative secondary delivery models, particularly if civil society or the private sector has some value-added in establishing these types o f distance or non-traditional programs. Alternative delivery mechanisms can expand secondary accessibility in rural and low-income areas and can provide specific and relevant skills and knowledge. There have been few interventions in Central America that have aimed to overcome the inequitable enrollment distribution in the public sector by reducing the opportunity costs or private direct costs of schooling. An exception i s the recent development of alternative or flexible secondary modalities in rural and low-income areas, which can reduce the opportunity costs o f schooling, address supply-side constraints and provide a more relevant education to non-traditional students. Common alternative delivery mechanisms are distance education programs and programs offered at worksites. These programs are increasingly common in Central America. The Honduran program, Educutodos, for example, served more than 85,000 people in 2003. Alternative delivery mechanisms use new pedagogical approaches designed to better serve rural and low-income communities as well as working individuals. They meet the needs of low-income communities by offering a free or low-cost alternative to traditional secondary. Many expand access to rural communities that have not had access to secondary school, and they avoid the high costs of building and staffing a traditional school. Flexible delivery programs are frequently designed to allow students to continue "* Although evidence is mixed, private schools, at times have been found to be more cost-effective in terms of student learning outcomes. 114 working or raising a family while also studying and earning their secondary degree.'53 Flexible delivery mechanisms would be particularly advisable in Honduras (due to existing fiscal constraints), as well as in Nicaragua and Guatemala, where they would constitute a valuable tool to improve coverage with equity. That said, it is also important to ensure that such models are subjected to the same rigorous quality requirements that should be applied throughout the system. If these delivery mechanisms do not ensure quality, they can, in effect, set up a parallel and inferior education system for the poor. This should be vigorously avoided. One way to measure this would be to analyze the relative rates of return to traditional and alternative secondary programs. Developing a strong testing and accreditation system will help ensure more uniform quality standards across secondary education options. A key challenge will also be ensuring steady quality of programs and outcomes as programs go to scale. This may require additional per pupil expenditure. Additionally, it will be important to assess and expand flexible programs for marginal urban areas where the opportunity costs of schooling are particularly high due to more dynamic labor markets. To date, although these programs have had some success in improving secondary enrollment, particularly in marginalized areas (see Box 3.4), they have not received sufficient assessment. ance Educationin El Salvador. Distan inistry of Education's old that cannot g of Saturday of Sunday. Teachers ducationin Hon an alternative s 153A discussion of the benefits of alternative models for secondaryeducation is also provided in Figueredo and Anzalone (2003). 115 into the secondary range. In 2001 it had 7,510 participants from 7 to 8 grades, and was in the The Tutorial Learnin s still a small-scale operation. In evaluation of these two programsi s still needed. School-based management models can also improve cost-effective secondary coverage, particularly in rural areas. Chapter V focuses exclusively on the school- based management education models currently functioning in Central America. These models give management authority to school staff and community members and are shown to be a cost-effective means of expanding coverage in rural areas although to date, most countries use the school-based management model principally at the primary school level (Nicaragua being an exception). D.QUALITY ANDRELEVANCE Finally, it i s clear from the previous discussion on the causes of low secondary coverage that quality improvements will be key to the expansion of secondary education, in particular in lower secondary. Raising the private rates of return would make investing in education today more attractive, for the same forgone income and discount rate. We will not go in any detail on quality issues here since Chapter I1 already reviewed quality constraints and areas for reform and intervention. Here, we shouldjust add that there are challenges specific to secondary education that need to be further analyzed, but are beyond the scope of our analysis. A key issue i s a curricular one, which refers to the offering of different modalities o f secondary for academic and vocational preparation. This is a hotly debated issue. Different modalities enable schools to become centers of excellence in particular areas, allow schools to meet the unique needs of different types of students, and can result in graduates with specific and appropriate skills. However, 116 academic and vocational streams often end up segregating students of upper and lower classes and, frequently, academic secondary schools are considered superior to vocational schools.'54 E.WHATPOLICY COMBINATIONSHOULDBEADVANCEDINEACHCOUNTRY ? There is no one appropriate policy combination to achieve broad-basedlower secondary coverage in Nicaragua, Honduras and Guatemala and broad-basedupper secondary in El Salvador. Each country's decisions should ultimately depend on the relative importance of the different causes of low enrollment, which will determine the relative impact of each intervention on coverage, and on the relative costs of each of the interventions. In other words, a cost-benefit or cost-effectiveness analysis is important in order to design the most cost-effective policy package for each country. In all countries there seems to be scope for a combination of policies. El Salvador was reasonably successful in expanding lower secondary education in a relatively equitable way by putting emphasis on public financing and delivery (including basic education centers and autonomous schools) and introducing demand-side subsidies and flexible delivery mechanisms.Nicaragua was also successful in expanding secondary education, albeit substantially less equitably (see Chapter I). Nicaragua used less public financing and delivery, made little use of distance education or demand-side subsidies, and focused on public delivery in urban areas (autonomous schools). Because of the inferior progress in terms of equity in Nicaragua, Guatemala and Honduras would be well advised to follow El Salvador's example when expanding secondary education. In Honduras, Guatemala and Nicaragua there should be more emphasis on public financing and supply in lower secondary education. Basic education centers should be further encouraged or developed, as should flexible delivery mechanisms. Quality improvements should be promoted everywhere. Demand-side subsidies may be particularly important in Guatemala. InEl Salvador, there i sjustification for larger public intervention in upper secondary, especially interms of cost-effective modalities, demand- side subsidies and mixed public-private schemes. As in the other three countries, quality improvements need to be encouraged at all levels. Below, we present cost simulations for El Salvador and Honduras that explore thesedifferent options. IV. How much would it Cost to Increase Secondary Coverage? Simulations for El Salvador and Honduras In this section we run cost simulations for El Salvador and Honduras. InEl Salvador we simulate how much it would cost to achieve universal coverage (Gross Enrollment Rate - GER - of 100%) in lower secondary and wide coverage in upper secondary (70% GER). InHonduras we simulate costs to achievebroad-basedcoverage in lower secondary (70% GER) and, although not considered an urgent objective, a substantial expansion of upper secondary (50% GER). For each country we run multiple models representing the various '54Several studies have found rates of return to academic secondary are higher than those to vocational secondary, and the quality of learning outcomes is better in academic streams (Castro, Carnoy and Wolff, 2000; Heuser, Jacobs and Umansky, 2003). 117 policy options discussed above. This allows for comparison of cost-effectiveness between various policy combinations. It i s important to note, however, that the simulations do not take into account possible quality differentials among options. A. TARGETS MAINASSUMPTIONSOFSIMULATIONS AND Table 3.4, below, summarizes the target indicators set for the simulations in the two countries, For both countries, targets are set below national goals (Education for All Plan in El Salvador and the Poverty Reduction Strategy in Honduras) in that our indicator is Gross Enrollment Rate (adjusted by the repetition rate) rather than the Net Enrollment Rate (NER) targets set by the countries. That said, as GER increases so, too, will NER provided that efficiency levels do not get worse at the primary level. We chose GER because it i s an easier indicator to monitor - and is, therefore, more reliable - since it i s unaffected by the cumulative impact of delays in completing primary education. While these goals are below national targets they would bring both countries up to par with other countries with similar per capita incomes. El Salvador Honduras Current (2001) Target (2015) Current (2001) Target (2015) Lower Secondary Cycle (7 to 9) 69% (a) 100% 41% (C) 70% Upper Secondary Cycle (10 to 12) 38% (b) 70% 22% (d) 50% Estimates of the fiscal impact of meeting the education coverage goals for both countries are based on the average current cost o f providing service under each type of program. These costs include teacher salaries, other administrative costs incurred by the Ministries of Education, non-salary recurrent expenditures, plus an estimation of capital costs, when appli~able.'~~For El Salvador, a scholarship and private sector subsidy were also estimated based, respectively, on the direct costs of schooling (fairly high in secondary), and on average teacher salary per student in the private sector (assuming the public subsidy covers teachers' costs). Since the emphasis of this analysis i s on the fiscal impact of the different scenarios, the cost o f privately funded and provided services i s not estimated (only the public subsidy is). As stated earlier, there i s no weight given in the 155The models suppose that for each incremental student in the traditional public education system, the government will have to build 2m2 of school construction. This is based on parameters used in Honduras, which should also be valid for El Salvador. It is assumed here that this i s not necessary under the alternative non-traditional programs, which usually work in local community facilities. For EDUCO schools, infrastructure costs are already included in the unit costs. I t i s clear that new infrastructure will not always be needed and using more intensively the existing one will also be an option in all types of education centers. The current infrastructure constraints in public upper basic and secondary education in Honduras and over-crowding in the public urban sector in El Salvador will, however, probably make this necessary as the school-age population increases. 118 modeling to possible differences in teaching quality or learning outcomes between the different programs included in the model. Table 3.5 summarizes unit costs per education level by providedfinancing mechanism: Provider/Financing Mechanism: Lower Secondary 1 Upper Secondary I El Salvador 1 I I Traditional Urban/Rural EDUCO rural 202 (b) 1IDistance Education Rural 168 168 Scholarships 1I 206 (c) II 237 (c) II Honduras Traditional Urban/Rural 280 280 (d) I CEB 196 Educatodos/SAT 150 Telebasica 328 Source: The World Bank (2004 b, c, d); Notes: (a) This is a simplifying assumptions, should be slightly higher for rural schools; (b) Includes some infrastructure costs, admittedly small, which we could not disentangle since they are included in the MINED transfers to EDUCO with teacher costs. (c) For lower secondary, we proxy the value of the scholarship by the subsidy amount calculated by FUSADES; for upper secondary, the scholarship amount comes from FEPADE; (d) This is a simplifying assumption as costsshould be higher in uppersecondary andfor ruralschools. B. SCENARIOSAND COST PROJECTIONS This sub-section presents service expansion projections and cost estimates for achieving the 2015 target coverage goals in both countries. It is supposed that both countries advance at a steady rate from the present situation to reach these goals b y 2015. The GER needed in each year i s estimated from the total population projections and the relevant population group for each cycle. This enrollment quota also takes into account repetition, adding the projected number o f repeaters to the total number o f places necessary to achieve the "net" coverage rate desired for each year. a) Scenarios El Salvador The policy options section above revealed several logical policy options for upper secondary expansion in El Salvador. These include private subsidized schools in urban areas, further expansion o f school-based management in rural areas (EDUCO schools), quality-enhancing interventions, and expansion of the targeted scholarship program. The rationale for promoting both public and private delivery through public financing is based on the fact that private enrollment has been decreasing at a continuous trend since the 119 early 1990s creating unutilized capacity in the private sector in urban areas.156 Additionally, teachers' salaries are slightly lower in the private sector while student performance in private schools i s still higher than in public schools. While the expansion of public delivery has contributed positively to broad-based lower secondary expansion, further expansion could probably be undertaken in a less costly way b y also making use of this unutilized private sector capacity. This rationale for public-private partnerships i s not currently valid in Honduras, where there i s no evidence of under-utilized capacity in the private sector.157 Based on these arguments, four main costing scenarios are modeled for El Salvador (also summarized in Table 3.6). All scenarios assume that coverage expands simultaneously with advances in quality and equity. On the quality side, we assume that spending per student in lower and upper secondary increases in relation to per capita GDP such that El Salvador `catches up' with the Latin American average (see Chapter N).On the equity side, we assume an increasing share of low-income students i s covered by scholarships. We do not model different scenarios based on different repetition rates, since repetition is already quite low at the secondary level inEl Salvador. The four scenarios include: 1. A "status quo" scenario in the public sector delivery structure, with a decrease in the share o f the private sector enrollment in the urban area. In this case, in future years, private provision i s assumed to continue its decline (which has been going on since 1993) at a -1 percent yearly rate in grades 7 to 9 and -3 percent in grades 10 to 12. As a result of this assumption, since the total number of places must grow, the projections show a steady decline inthe share of private places by 2015; 2. A "status quo" scenario in the public sector, with a constant share of the private sector, obtained through the subsidization of private schooling by the public sector; 3. A scenario where the relative importance of traditional and non-traditional provision (EDUCO, distance education) changes in the rural areas, with a decreasing share of the private sector; and 4. A scenario where the relative importance of traditional and non-traditional provision changes in the rural areas and the share of the private sector i s maintainedconstant through public subsidization. While private schools representabout 45% of upper secondary institutesthey only represent about 33% of the enrollment in 2002, versus about 45% in the early 1990s (the percent of private institutesremained similar). Small class size (about 13 students per classroom vs. 44 in the public sector in 2000) and low pupil-teacherratios (about 20:1vs. 26: 1in the public sector in 2000) are indicationsof the excess-capacity that there is now inthe private sector (seeThe World Bank (2004b). '57Although the quality of the informationis poor in the country and, therefore, trends in enrollmentshares are difficult to monitor, it seems that private sector enrollment has kept on increasing, at least in absolute terms, representing a somewhat stable 30/25% proportion of total enrollment in secondary. The proportion of private institutes, about 50% of total institutes, has also been quite stable. Essentially we have a fairly stable share of small private schools. 120 Honduras The discussion in the previous section on policy alternatives to expand secondary coverage suggests that in Honduras it makes sense to expand public secondary delivery while attempting to maintain the historic growth rate of the private sector as well as develop more alternative or cost-efficient delivery mechanisms. Because of this, the simulated models assume an increase in non-traditional delivery in rural areas and further development of basic education centers - two key strategic policy options for Honduras. Internal efficiency improvements, through lower repetition, are also simulated since there i s sufficient space for efficiency improvements in the country. Quality improvements are also necessary in Honduras. These are modeled, in part, with the expansion of the TeZebasicaprogram in lower secondary.158 Three main scenarios are envisaged in Honduras (also summarized in Table 3.8). In all o f the scenarios for Honduras the growth rate of private provision i s held constant at its historic rate o f 3.6 percent a year, and all other necessary growth is assigned to the public sector.159As a result of this assumption the projections show a decline in the share o f private secondary enrollment, to 19percent of the total by 2015. The three scenarios are: 1. A "status quo" scenario in the public sector (no efficiency gains, same delivery structure); 2. A scenario with efficiency improvements (this scenario allows for a gradual increase in the levels of efficiency, decreasing the repetition rates from the initial level of, respectively, 16 percent in grades 7 to 9, and 18 percent, in grades 10 to 12, to half these levels by 2015); and 3. A scenario like scenario 2, except with changing delivery structure in lower secondary that allows for greater participation of alternative or more cost-efficient programs (Basic Education Centers (CEB), EducatodoslSAT) and quality improvements (through TeZebasica).160 b) Cost Simulations The results of the cost projections are summarized in Tables 3.6 and 3.7 for each o f the scenarios.161 TV Education is often usedto strengthen lower secondary education in Central America. 15' This might be considered over-optimistic: an increased effort to provide more public places and improve quality in the public sector could be expected to crowd out some of the private sector growth in the future. To the extent that this were to happen, the resulting fiscal cost estimates would be under-estimated. To be noted that the simulations were done on the basis of 2000 official data on relative public and private shares which were indicating a share of 30% for the private sector. It i s very unlikely that this share was reduced from 30% to 25% in only three years, indicating deficient data either in 2000 or in 2003. If the rate of growth of the private sector was somewhat lower than 3.6%, then there was some crowding-out, which could lead Honduras to face similar issues than ElSalvador in a longer run. Given that there are no alternative service provision programs for the diversified (upper secondary) cycle in the country, there are no sub-scenarios for that cycle (although the introduction of some form of distance education may have been simulated). 16'Complete analyses of the simulations can be found in The World Bank (2004b, c, d). 121 El Salvador A substantial increase in public financing is necessary to reach target enrollment rates in El Salvador. As a percentage of GDP, the combined cost of both lower and upper secondary cycles will need to go from 0.60 percent in 2002 to between 1.45 percent and 1.61 percent, depending on the scenario, in 2015. Overall, the upper secondary share of the education budget will need to increase to three or four times its current level, and the share for lower secondary (termedUpper Basic in El Salvador due to the basic education reform that included grades 1 through 9 in basic education) will need to double or more than double, depending on the scenario. It i s also clear, however, that substantial savings are possible through the use of public subsidies to finance some urban private schools (scenarios 2 and 4).'62Scenarios 3 and 4, finally, also lead to savings in lower secondary by enrolling students in EDUCO schools where the unit costs and infrastructure costs of these schools are lower than those of traditional schools. The share o f distance education, however, does not need to grow inEl Salvador, in fact in upper secondary it should decline, as there are currently more than enough spaces to cover all the appropriate target population groups.'63 The number of scholarships in traditional secondary, however, should expand to support more low-income students. Both of these changes increase the per pupil cost of upper secondary. 162 This is because if the public sector subsidizes private schools to keep their enrollment share it only covers the salaries of private teachers, while if the public sector has to absorb an increasing share of private sector enrollment (scenario 1) it needs to be paying for both the salaries of public teachers (higher than private teachers' salaries) and new infrastructure. `63 The share is assumed to be constant at a 15% of public rural enrollment, as `heeding to work" is a reason for non attendance for only 13% o f the out-of-school children in rural areas and a share of the children non attending for "cost reasons" (too expensive) would be dealt with through scholarships. In the scenarios the share of distance education in upper secondary decreases to 30% and the share of traditional education increases to 70%, resulting in higher costs. To generate equity improvements the proportion of low-income students receiving a scholarship would be increased from 20 to 25%. The cost of the sub-cycle would increase because of the replacement of distance education by more expensive traditional one and the increased use of scholarships. 122 I Table 3.6: El Salvador- Secondary Targets Simulation Results I I (US$2OOZ millions) 2002 I2005 I2008 12011 I2015 Secondary (upper basic) Total Costs Upper 28.3 40.7 60.3 89.9 157.4 Secondary Lower Secondary 12.2 14.3 16.5 19.0 26.0 CostsEducation Budget Upper Secondary 6.0 I8.0 I 10.1 I 14.6 I22.6 CostsEducation Budget III I23 I 2002 I 2005 I2008 I 2011 I2015 CostsEducation-Budget Upper Secondary II 6.0 I 7.9 I 10.8 I 14.9 I23.9 Secondary (upper basic) Total Costs Upper 28.3 39.2 56.7 82.7 141.6 Secondary Lower Secondary 12.6 13.8 15.6 17.6 23.6 CostslEducationBudget Upper Secondary 6.0 7.6 10.0 13.4 20.4 CostsEducation Budget Lower and Upper 0.60 0.70 0.84 1.14 1.45 Secondary Costs/GDP source: The World Bank (2004b). Honduras The budget simulation analysis for Honduras suggests that this country will need to spend between 0.3 and 0.5 percent more o f GDP on secondary education by 2015 to achieve broad-based coverage in lower secondary. A substantial expansion in upper secondary would require another 0.5% of GDP. As a percentage of GDP, the combined cost o f both cycles will need to go from 1.05 percent in 2002 to between 1.78 and 2.11 percent, depending on the scenario, in 2015.'64 Overall, the secondary share of the education budget would be between 70 percent and 100percent higher, depending on the scenario. Efficiency improvements (scenario 2) at the secondary level bring some savings. The expansion of alternative delivery programs (scenario 3) i s the most cost-effective o f the three scenarios. It also may respond best to the unique needs of Honduras' large rural and poor populations. It is interestingto notehere that the cost simulations for Hondurasand El Salvador are roughly equivalentdespite the fact that the policy combinations simulated in the El Salvador scenarios are more ambitious and far-reaching. The main reason for this i s that the growth rate of the school-agepopulation in El Salvador is significantly lower than that of Honduras. 124 Table 3.7: Honduras-SecondaryTargetsSimulationResults (US$ millions) I 2002 I Aver. IAver. I Aver. I Aver. I2015 Budget Total Costs Lower I42 I 5 7 I 87 I 128 I 187 1240 Secondary Total Costs Upper 126 1 37 I 60 I 9 5 1150 I201 Secondary Total Costs Secondary 68 94 147 223 337 441 Total Costs Secondary 17.2 % 19.3% 23.2% 27.0% 29.9% 34.5% EducationBudget Total Costs Secondary 1.05 % 1.18% 1.42% 1.65% 1.83% 2.11 % I/cmP . __. - 1 I I I I I I Scenario2: Same as scenariobut with efficiency gains in both levels 0 Repetitionrate decreases from 16% in2002 to 8% in2015 (lower) 0 Repetitionrate decreasesfrom 18% in 2002 to 9% in 2015 (upper) Total Costs Lower 42 57 84 121 176 223 Secondary Total Costs Upper 26 36 58 90 139 185 Secondary Total Costs Secondary 68 93 142 211 315 408 Total Costs Secondary 17.2% 19.2% 22.4% 28.0% 31.59% EducationBudget I 25.5% Secondary Total Costs Secondary 68 87 129 190 283 372 Total Costs Secondary 17.2% 17.9% 20.4% 23.0% 25.1% 29.1% EducationBudget Total Costs Secondary 1.05% 1.10% 1.24% 1.40% 1.53% 1.78% /GDP In conclusion, most of the enrollment growth inboth countries will needto be funded by the public sector (national and/or international), through direct provision in traditional public centers, the set up of alternative service delivery mechanisms, and quality improvements.Additionally, the increase in public financing will need to be substantial. In El Salvador, supporting publidprivate partnerships appears to be a useful policy as 125 does expanding demand-side subsidies.'65 InHonduras, in contrast, the primary focus of the country should remain on providing infrastructure and expanding alternative delivery mechanisms. Efficiency improvements will also be particularly important in Honduras. V. Conclusions Expanding equitable coverage of quality secondary education i s a critical priority for Central America. Secondary education is integral to both individual and national development and opportunities. The high rates of return to upper secondary suggest that this level of skilled labor i s highly demanded on the labor market. While expanding the supply of educated citizens can have the effect of decreasing the rate of return there i s significant reason to believe that skilled labor would attract foreign investment and promote growth in Central America enough to compensate or even outstrip any downward effect of additional supply. Secondary education should expand in the most broad-based way possible to minimize adverse effects on educational and socioeconomic inequality and maximize growth. Because of fiscal constraints there will be some degree of trade-off between improvements in primary and secondary. Efficiency improvements in primary can reduce the extent of this trade-off but the necessary quality improvements at all levels will place additional burden on education budgets. At the very least, Nicaragua, Guatemala and Honduras should aim at encouraging broad-based expansion of lower secondary by expanding equitable access to secondary education for those children who are now completing primary. El Salvador, a country with higher and more equitable lower secondary enrollment, should now aim at expanding upper secondary (while also pushing for full transition of the poor and very poor to lower secondary). In order to expand secondary, countries will have to address the unique combination of constraints that currently limit enrollment. Inthis chapter we found the key constraints in Central America to include high private and opportunity costs to schooling, lack of motivation due primarily to low relevance and quality o f lower secondary, and supply shortages. Low test scores and low rates o f return to lower secondary are testament to the low quality and relevance of this level, creating another barrier to enrollment. Supply of secondary schools has expanded in recent decades but remains a challenge in rural areas. 165 This partnership could take the form of a public subsidy to some private schools to help them cover recurrent costs, such as teacher salaries, or to help them attend low-income students. Inthe first case, which i s the one depicted in our scenarios, teachers' salaries of some private schools would be entirely financed by public funds which would allow them to keep on functioning and, hopefully, provide good quality education, retaining the existing upper-income students and attracting new middle-income students as coverage increases. The double advantage of this intervention would be that, on the one hand, the public sector would still be saving because of lower delivery costs in the private sector, on the other one, public sector delivery could be better targeted to lower income students, decreasing its middle-income bias. In the second case, closer to a voucher system, private schools would receive a subsidy that would cover the schooling life of a low-income student, and, as such, cover tuition fees and other private costs, as well as a portion of schools' salaries: less students, overall, would be covered, but this system would still make it possible to reduce the burden on the public sector (although to a lesser extent) and, above all, foster more socio-economic diversity in schools rather than the socio-economic segmentation that the first model would produce. 126 It is particularly acute at the upper secondary level. In all cases these constraints disproportionately affect lower income people. Lower income groups attend poorer quality schools, have lower rates of return, higher age by grade distortion, and face more acute economic and work-related barriers. Because of the diverse range of barriers to equitable high-quality secondary coverage, countries need to address this regional priority with policy packages that respond to the unique needs of their specific country. Inthe policy options section and cost simulations section of this chapter we saw that the most effective and efficient policy combination by country. The main components of these policy combinations include: An increase in public financing for secondary education. This is a key condition for broad-based secondary coverage in all countries. Simulations undertaken for El Salvador and Honduras show that reaching satisfactory coverage targets in both countries will require an increase by about 1 percent of the secondary education share in terms of GDP over the 2002-2015 time period. An increaseintraditional public supply. The bulk of secondary expansion should be provided by public sphere in Central America although innovations in public service delivery are a promisingform of public supply. Promotingbasic education centers, with all three sub-cycles of basic education, is a promising intervention that has been relatively successfulin El Salvador. Inrural areas, promoting school networks up to grade 9 or even grade 11 or 12 andor community-based schools could also be a viable cost-efficient option. An increase in public-private sector partnerships. In countries where private schools are more cost-effective than public schools there may be scope for encouraging public financing/private delivery systems. Inpractice, this may mean covering part of the recurrent costs of private schools to have them accept middle or low-income students. The presence of idle capacity in the private sector, as in El Salvador, makesthese partnerships more promisingfinancially, since little new investment is necessary. The El Salvador simulation suggests that this is a good policy option for that country. An increase in non-traditional public supply, such as distance education, in rural areas, can ensure coverage with flexibility. A program like Educatodos seems to be particularly relevant in this context, but others alternative education programs, in urban areas, for example, should be further explored. An accreditation system, ensuring comparable quality between traditional and non-traditional delivery models, needs to be put in place. Non-traditional supply appears to be a valid option for all four countries. An increase in the quality and relevance of secondary education. Improving the relevance of lower secondary i s particularly urgent in all countries. All the measures suggestedin the previous chapter on education quality apply to this sub- cycle. Upper secondary quality and its relevance to labor market needs should 127 also be improved. Countries need to think carefully about the structure and main objectives of secondary, particularly in terms of the age-old debate on technical- vocational versus academic secondary. This subject, however, has not been discussed in this report. (f) An increase in demand-side subsidies, including targeted scholarship programs and conditional and targeted cash transfers. These programs would be particularly useful in countries with relatively high tuition levels or other private costs of schooling. 128 Chapter IV: A Review of Education Expenditure in Central America How much money countries devote to their education systems, how that money is distributed between different education levels and programs, and the efficiency with which it is used to promote the final objectives of education are critical areas to understand and assess. The amount, distribution, and efficiency of education expenditure affect educational coverage, quality, and equity. In this chapter we look at some of the different aspects of educational expenditure. The first section looks at the overall level of public education funding in the four Central American countries, comparing them with the Latin American average and Costa Rica. The second section examines how public education funding is distributed across education levels. The third section looks at equity challenges inthe region that arise from per student expenditure and enrollment distributions at the various education levels. The fourth section deals with household education expenditure. The fifth section investigates the efficiency of education spending, focusing in on cost-effectiveness in the four countries and on the effectiveness of the distribution of expenditures across spending- type. Finally, the sixth explores the issue of external education funding. The conclusion of the chapter reviews main findings and presents several areas of intervention. 1. Public Education Spending: Levelsand Evolution While government spending on public education has been growing over the past decade in Central America it is still low both as a proportion of GDP as well as on a per-child basis. Table 4.1: Public EducationExpenditure inProportion of GDP (%) (a) Notes:(a) Public expenditure on education includes all public funds channelled to the educationalsystem through public institutions, therefore includingboth national and external public funds. (b) New classification (only education expenditure, and excludes social security and public debt). (c) with official GDP data: (d) with "unofficial" GDP data. There is some agreement that GDP figures may be underestimated by as much as 70%, this second data series reflects this under- estimation (e) Public expenditure includes public funds channelled only through the Ministry of Education (includingboth national and external public funds). Sources: El Salvador: WB WDI and LAC Database: Ministeriode Hacienda: CEPAL population estimates Guatemala: Banguat, Ministry of Finance, and Planning Unit, Ministry of Education; 2001-2003: CIEN. New classification: MFP and Bank of Guatemala. Honduras: 1990-1996:"EstudioSectorial Plan Decenal'; 1997- 1998: UNAT 2001; 1999-2002:SIAFI; IMF Financial Statistic differentvolumes Nicaragua: 1995-2000: MCHP and Central Bank, as reported in Arcia (2003); 2001-2002: Directorate of Finance, MECD. Costa Rica: Central Bank, Ministry of Finance and ICEC (National StatisticalOffice). 129 Table 4.1 above presents data on public education expenditure as a ratio of Gross Domestic Product (GDP) for the four countries under analysis as well as for Costa Rica from 1995 through 2003.'66 Education expenditure is growing, but it is still too low in Central America. The proportion of GDP devoted to education has been growing throughout Central American since the mid-1990s. This reflects growing prioritisation o f education. Education spending does vary considerably, however, across countries, being particularly high in Honduras (7 percent) and substantially lower in El Salvador, Nicaragua, and Guatemala (between 2.5 and 4 percent). Honduras i s also the country where education spendinghas increased the most in the last decade. El Salvador, Nicaragua, and Guatemala all spend significantly less than the Latin American average (see Figure 4.1) - close to 5 percent of GDP - and less than comparison countries with similar per capita GDP.'67 Figure4.1: EducationSpendingas a Proportionof GNP, 1999/2000 'I 7 0 1 1 6 0 5 0 ~ 8c 4 0 3 0 0) n II 2 0 10 0 0 Source:WB, WDI database, 2003 On average, Latin American countries spend roughly US$300 on public education each year for every age-appropriate child. Central American countries spend much less (see Figure 4.2). The equivalent figure in Nicaragua, for example, is less than US$70. Much of the difference i s due to lower per capita income in Central America (Honduras spends more than the Latin American average in terms of education spending as a percentage o f 166Public education spending includes all education spending from any public institution (except for Costa Rica where only Ministry of Education spending is reported). In Central America, non-Ministry of Education spendingon education is not insubstantial (15 to 30 percent). Most public primary and secondary education is provided by Ministries of Education, but other educational levels such as tertiary, adult or special education are frequently co-financed and co-provided with other institutions. Primary and secondary is also, at times, financed and executed by institutions outside of the Ministries of Education such as by social funds. 167 An average expenditure ratio of about 4.2 percent was reported for a sample including Philippines, Jordan, Bolivia, Algeria, Peru, Colombia, Paraguayand Ecuador (based on UIS 2000 data). I30 GDP, but their yearly spending per child is far below the Latin American average due to lower per capita income), but lower priority ratios can also explain it. Figure 4.2: Education Expenditure per School Age Population, Because of the sharp 2001/2002 difference in per capita income between Central American countries and some of the more affluent countries in Latin America, Central America should aim towards spending an e 150 equivalent proportion of n GDP rather than equivalent 100 per student expenditure. In this way Central American 50 countries would spend as much per child as other 0 Latin American countries El Salvador Guatemala Honduras Nicaragua Costa Rica with similar per capita GDP. Of course, the Source:Table 4.1 and HouseholdSurveys. effectiveness and efficiency of education spending is, in the end, more important than the dollar amount o f per child spending bit the lack of investment ineducation in Central America i s problematic on its own. In sum, in spite of a positive evolution since the mid-1990s in terms of education expenditure vis-&vis GDP, Central American countries still tend to spend relatively little on public education in terms o f GDP or income per-capita. An exception is Honduras, where, however, there are major efficiency issues. Expenditure per child i s lower than US$200 in all countries and is below US$ 100in Guatemala and Nicaragua. II. Disaggregration of Educational Expenditures by Education Level and per-Studentand Child Unit Costs This section disaggregates public educational investment by level (primary, secondary and tertiary) and then examines per-child and per-student expenditure at each o f these levels. The analysis reveals that while primary education logically occupies the bulk of Ministries of Educations' budgets, spending on secondary appears to be low while spending on tertiary i s high in most Central American countries. Inadequate per-student and per-child spending i s particularly apparent in Nicaragua. Table 4.2 and Figure 4.3 show estimates of public education expenditure in 2002 by education level. Several factors make precise figures difficult to identify. Non-Ministry of Education spending i s often impossible to divide by level, as are capital expenditures. Again, Costa Rica i s offered as a point o f comparison. 131 Pre- Primary Basic Secondary Tertiary Admin. Other Total primary El Salvador (a) 31.6 247.6 27.9 (b) 27.1 27.8 22.0 (c) 384 ~~~ ~~ El Salvador (d) 37.5 284.5 49.2 37.7 (e) 62.0 (f) 471 Guatemala(9) 36.8 250 40.4 61.5 36.5 74.5 (i) 500 IGuatemala,2001(!$ I 37.0 I 259.9 I 1 40.4 I 74 . 5 I 36.9 I 71.1 fi) I 520 I Honduras(h) 16.0 150.1 70.1 67.1 74.5 18.2 (i) 396 Honduras,2001 (h) 15.1 156.2 69.2 72.0 55.1 12.3 (i) 380 Nicaragua(I) 4.1 72.0 13.6 43.0 (m) 7.1 3.9 (n) 144 Nicaragua(0) 6.0 83.0 13.9 43.0 (m) 7.1 5.0 (n) 158 Costa Rica (%) (P) 41.8 (4) 24.2 20.2 7.6 6.2 100 (a) All recurrent public expenditure channeled through public institutions is included; (b) Includes grades 10-12; (c) Includes the portion of recurrent costs financedthrough external funds, which could not be allocated to any of the levels; (d) Includes all public expenditure channeled through public institutions; (e) Includes amount directed to Education Development; (f) Includes investmentexpenditure that cannot be allocated across levels, as well as the Teacher Welfare and Strengthening of Access to Education programs; (9) Includes all recurrent public expenditure channeled through public institutions +unallocatedcapital expenditure; (h) Includes all recurrent public expenditure channeled through the Ministry of Education + unallocated capital expenditure. 2002 is budget; (i) Includes all investment and recurrent expenditures that cannot be allocated across levels; (I) Includes all recurrent public expenditure channeled through the Ministry of Education; (m) Includes all university expenditure (transfers and non transfers); (n) Includes adult education, special educationand teacher training; (0)Includesall public expenditure channeled through the Ministryof Education; (p) Only MINED public expenditures; (4) Includespre-primary.Sources: see Table 4.1. Figure 4.3: FunctionalExpenditureDisaggregation, 2002 100% 90% 80% 70% administration 60% 50% 40% 30% 20% Notes: Prirnayand Secondary 10% Shares have been estimated in El Salvador; Primary includes Pre- 0% primaly in Costa Rica; Tertialy includes Universityown budget in El Salvador Guatemala Honduras Nicaragua Costa Rica Nicaragua. Sources: See Table 4.2 132 Primaryeducationis the bulkof educationbudgetsinall countries.All four countries spend between 40 and 50 percent of the public education budget on primary education. This high share is logical in that primary education is clearly the first priority of all countries and it has by far the largest enrollment. Low sharesspentonsecondary.Incontrast, the share directed to secondary education i s much lower, fluctuating between roughly 8 and roughly 25 percent of the total public education budget. Secondary shares are higher in El Salvador and Costa Rica where secondary i s more prioritized and has higher enrollment rates. All Central American countries spend a smaller roportion on secondary than the Latin American average, which is about 30 percent.16' This fact is probably both a contributing factor toward and a result of low enrollment in secondary in Central America (see Chapter HI). Relatively high shares are spent on tertiary in Central America. The tertiary share fluctuates substantially across countries, ranging from 7 percent in El Salvador to 30 percent in Ni~aragua.'~~s interesting to note that the three poorest countries spend as It i much or even more on tertiary as they do on ~econdary.'~' Administration costs are likely to be under-estimated in all countries in that some administrative costs are covered by institutions outside the Ministries of Education and are, therefore, not easy to track. From the limited data we have, however, it appears that administration costs are relatively contained in all countries, except in Honduras where more i s spent on educational administration than is spent on secondary education. Per-child and per-student spending. Missing from the previous information on how education expenditure i s divided by level is an understanding of what this means for per child and per-pupil public expenditure. As we've seen throughout this paper enrollment rates are very different in primary, secondary and tertiary institutions. By comparing recurrent public expenditure for each of the levels to the number of school-age children and students enrolled in the public system for each of these levels we obtain expenditure per school-age population and per student by education level, two indicators that provide a more insightfulpicture of the actual level of resources available for each level. Figure 4.4 (and Table 10.1in Annex X) compare expenditure per child and student across levels and countries. "?See ECLAC (1994, 1996). 16'InNicaragua, education level budgetdisaggregationdefinedtertiary educationconsidered quite broadly, contributing, inpart, to the high percentageof spending identified as goingto this level. Argentina, Chile, Uruguay and Colombiaspent approximately20% of their educationbudget on tertiary education in the time-period 1994-2000 (see various World Bank Country Public ExpenditureReviews). This is inthe range of what these three, muchpoorercountries, spend. 133 Figure4.4: Expenditure per Childand Student, 2001/02 Primary. For educa- tion levels that lack (Source: Table 10.1Annex X and HouseholdSurveys) universal enrollment (particularly secondary and tertiary) we would expect per- student expenditure to be higher than per- child expenditure. In a hypothetical situation universal enrollment and zero repetition per-student per-child expenditure would be As enrollment decreases per- student will be higher than expenditure per child. As repetition rates increase expenditure per-student will drop relative to expenditure per-child. Similar expenditure per child and student in Honduras and Nicaragua suggests high repetition rates. Reducing repetition would increase per student expenditure, funds that could be invested in improving the quality of education and ensuring that all students complete the cycle. The higher per-student expenditure in Guatemala, in spite of high repetition rates, reinforces the continuing unmet challenge of universal primary enrollment in the country'71. Per-student ex enditure at the primary level has increased in most countries over the last several years.R Significant rises in teachers' salaries explains, to a large extent, this 2 increase in per-pupil primary expenditure. In spite of this increase, however, an international comparison suggests that all Central American countries, except Honduras, spend below average on primary given their per-capita income (Figure 4.5). 17'For El Salvador, expenditureper student is higherbecause we are considering basic education, where there are drop-outs, insteadof primary. "* Per-pupil public primary expenditure rose roughly 20 percent in Guatemala in the 1996 to 2001 time period (CIEN, 2003), 40 percent in Costa Ricabetween 1998 and 2002 (Sanigest, 2003) and 35 percent (in basic education) inEl Salvador in the 1996to 2002 time period(The World Bank, 2004b). 134 Figure 4.5: Public Education Expenditure per Student in % of GDPper capita, 2000 40 35 30 25 20 15 I 10 5 1 &6$G-O +&@$8 of significant cost. $9" 6."e5&$ ,6+&9 +e+Go 'b b'b' ,0 .$b *o 9 Nicaragua is, however, an exception. Source: OECD (2002) and Table 10.1 -Notes:Central America 2002 or 2001 data Figure 4.4 shows that per- student spending in Nicaragua is higher in primary than in secondary. One explanation for the low per-pupil unit costs of secondary in Nicaragua is simply that the country has invested insufficiently at this level. A second explanation relates to the relative priority given to coverage versus quality in secondary education: although expenditure per-child in secondary i s also lower in Nicaragua than in the other countries, the gap is much greater for expenditure per-student, suggesting that the limited resources devoted to secondary have been focused at expanding enrollment rather than spending more on each child who is enrolled. The low per-pupil spending is sure to have clear implications on secondary education quality. In all countries expenditure per child is dramatically lower than expenditure per pupil at the secondary level. This reflects the severe secondary coverage gap discussed in the previous chapter and underscores again the importance o f the priority of increasing secondary enrollment in a broad-based manner. Notably, there has been little or no gain in per student expenditure at the secondary level over the last several years in most o f Central America.'73Expanding secondary coverage, although still limited, contributes to explain this stagnation. Tertiary. Finally, although not shown in the graph, per-student expenditure is particularly high at the tertiary level (see Table 10.1 in Annex X). In Honduras, for example per-student public recurrent expenditure i s US$933 at the tertiary level whereas the equivalent figure for per-child expenditure i s US$ 140. Certainly, unit costs at the tertiary level are bound to be higher than at the secondary or primary level but this huge public investment in tertiary education is socially regressive as will be discussed further below. '73Secondary spending has stagnated in Nicaragua, Guatemala and El Salvador since 1998. 135 111. Equity Challenges in Terms of Per-Capita Education Expenditure The distribution of public expenditure across socio-economic strata, geographic areas, and other divisions has important implications on the equity of education delivery. This section looks at how per child spending at the various educational levels compares with how enrollment i s distributed across income groups at those same levels. This analysis reveals that much of educational provision in Central America i s socially regressive in that it invests disproportionately in children of wealthier backgrounds. This i s particularly true at the secondary and tertiary levels, whereas public primary provision i s relatively progressive. This fact raises serious equity concerns since public education i s financed with general tax revenues. The section first looks at the distribution of public enrollment according to income quintiles in order to determine the levels of participation of the different income quintiles in the sector. This analysis is followed by a calculation of levels and distribution of spending across income groups, based on unit costs per educational level. A. THEDISTRIBUTIONOFPUBLICENROLLMENTAND SCHOOL-AGEPOPULATIONACROSS INCOME QUINTILES There are many fewer poor children in public secondary and tertiary schools than in public primary schools. The lines in Figure 4.6 represent the proportion of public enrollment that serves children from each income quintile (see also Tables 10.2-10.5 in Annex X). In primary schools and pre-primary schools the poorer income quintiles tend to have the highest proportion of enr01lment.l~~This i s because children from wealthier families are more likely to enroll their children in private schools. At the secondary and tertiary levels, however, enrollment i s concentrated in the top income quintiles. Confirming the results presented in Chapter 111, the distribution o f secondary enrollment appears to be concentrated in the middle-upper class in all countries (quintile 4), with El Salvador having a somewhat more balanced distribution favoring the three middle quintiles. At the tertiary level the highest proportion o f students come from the wealthiest quintile. These patterns are true for all four Central American countries. Public delivery is well targeted in primary, but poorly targeted in secondary and tertiary. Public education, which should provide access to schooling for all children, particularly those whose families can not afford to send them to private schools, i s fulfilling its role fairly well at the primary level but much less satisfactorily in secondary and tertiary education. The situation i s aggravated when we compare the distribution of public enrollment with the distribution o f school age population by income quintiles. '74 Each of the enrollment levels, which were produced on the basis of the household surveys for distinct years, are not the ones that were used for the calculation of the expenditure per student, where MINED official data were used for sometimes also different years. Discrepancies can then occur when recalculating unit costs on the basis of these enrollment data (in particular in Honduras). Note that the 2003 household survey was used in Honduras instead of the 2002 one because it is the only survey which differentiates between private and public enrollment. 136 It is usually the case that school-agepopulation, inparticular inthe younger age range, is not distributed evenly across quintiles, but has a higher incidence in the lowest income quintiles. This is largely because poorer families tend to have more children. By comparing the relative public enrollment shares of each quintile with its school-age population share, we capture with more precision how well or badly targeted i s public delivery. The bars in Figure 4.6 represent the proportion of the age-appropriate population that belong to each income quintile. This Figure illustrates that at the primary level the two lowest income quintiles tend to make a more than proportional use of public delivery, compared to their school-age population share (although this is less true for Guatemala); while the highestquintile makes a less than proportional use of public delivery, compared to its school-age population share (and, by converse, it makes a much more proportional use of private delivery in all countries - although these figures are not shown here). As a result, we are not surprised to find a gross enrollment rate roughly equitably distributed across quintiles in all countries (with the exception of Guatemala), as indicated in Chapter I. In contrast, at the secondary level, the two lowest income quintiles make a less than proportional use of public delivery, compared to their school-age population share (although this is less true for El Salvador); while the fourth quintile makes a much more than proportional use of public delivery, compared to its school-age population share (also less so in El Salvador). The highest quintile also tends to make a more than proportional use of public delivery although less so because of its much more than proportional access to the private sector (see Chapter III).As aresult, the secondary gross enrollment rate is fairly inequitably distributed across quintiles, as shown in Chapter I (again, less so in El Salvador). Finally, the highest quintile makes a much more than proportional use of public delivery at the tertiary level. 137 ? If 0 , ; / I t l (050) 0 0 ~ ( 0 m t m ( U ~ 0 0 0 0 0 0 0 0 u - U ? * L m P '0 m ii 8E I - d B.PUBLICSUBSIDY PERSCHOOL-AGE POPULATIONBY QUINTILE In Tables 4.3 and 4.4 we calculate the public subsidy per school-age population by quintile for each country andfor primary and secondary (wherewe have the mostreliable unit costs). We find that while the public education subsidy at the primary level is slightly progressive(Le.investsmoreheavilyinthe poor), the public educationsubsidy at the secondary level is tremendously regressive, awardingthree to four times the amount of public dollars to wealthy children as it does to poor children. To calculate the public education subsidy for different income quintiles at different education levels we make the im erfect assumption that per-student expenditure per student across quintiles is uniform!75 We then divide total education expenditure across income quintiles according to their relative public enrollment share. Finally, we obtain per-capitaexpenditure by dividing the amount per quintile by the school-age population of each of the quintiles. The tables show the distribution of public spending for primary and secondary education, the percenta e of the spendingthat is capturedby each quintile andthe resultingper capitasubsidies.1% The primary education public subsidy i s progressive but the secondary education subsidy is regressive. The per-capitapublic subsidy at the primary level i s fairly similar across the first four quintiles (this is because the enrollment share generally parallelsthe population share by income quintile as seen above at the primary level). This is somewhat less so in Guatemala. Ingeneral across countries, the public primary subsidy is lower for the richest quintile becauseof their lower enrollment proportion. Therefore, in per capita terms, public primary spending continues to be progressively distributed, although not highly so. The situation changes quite dramatically at the secondary level, where, in all countries except El Salvador, the per capitapublic subsidy is three or even four times higherfor the fourththan for the first quintile (andabout 3 times higher for the fifth quintile) due to the disconnectionbetweenenrollment and population shares. In El Salvador, the public subsidy is relatively more evenly distributed, although it is still regressive. Fair access to primary schooling does not mean fair access to quality schooling. Although the public subsidy seems to be relatively progressively distributed at the primary level, and the gross enrollment rate is quite homogenous across quintiles, there are still reasons to believe that even at the primary level public education expenditure i s not distributedprogressivelyenough. Unit costs are higher for poor students than they are Infact, the differentquality of the schools attended by the poor and the possible targeted interventions to the poor should both be considered and would alter unitcosts per quintile. 176Care should be given when comparing the total amounts per level with the ones of Table 4.2. These total amounts have been obtained by multiplying expenditure per student by the total public enrollment as reported in the household surveys. Additionally, the year o f the per student expenditure and of the enrollment do not always coincide exactly. Discrepancies occur in particular at the secondary level and for both Honduras (whose enrollment figures were particularly high in 2003) and El Salvador (where the unit cost in upper secondary i s multiplied by the enrollment o f the whole secondary cycle). In any case, the distribution o f the public subsidy across quintiles is what really matters here. 140 for wealthy students as poor children do not have the cultural and economic capital provided to wealthy students in their homes and communities. Concretely, poor children need more learning materials and better teachers than wealthy students in order to make up for disadvantages they bringwith them to the classroom. Recent analysis showing that poor students are much more affected by educational inputs (teacher quality, educational infrastructure, etc) confirms this. In contrast, available evidence suggests that the public schools poor children attend tend to be o f lower quality than those that wealthy children attend. Highdropout and repetition rates in poor students signal that education systems are frequently failing these children. Lower rates o f return for the poor further indicate that their schools may be o f inferior quality. It i s likely, in fact, that the assumption we made to calculate the public subsidy estimates is problematic and that per-student education expenditure i s lower, on average, for poorer students. If this is the case, our analysis is over-estimating the progressiveness of public expenditure. If we assumed that public expenditure was about 20 percent lower in the first quintile (and 5 percent higher in the remaining four quintiles to ensure a similar average expenditure per student), we would find that there i s only a marginal difference between the first and fifth quintile and that the public subsidy per capita is higher for the middle-class than for the poor in all countries. Poor students need more public investment at the primary level. Public delivery, therefore, provides fair access to the poor in primary as judged by equal proportionally correct enrollment in primary according to the age-appropriate population (although this i s somewhat less true in Guatemala). But, judging from educational outcomes o f the poor (in particular in terms of completion), public education delivery is not necessarily providing fair quality schooling. Although expenditure may not be the first determinant o f quality (many improvements, in particular related to management, do not require more funds), there is ground for ensuring higher expenditure per poor student (in terms of teaching materials, teacher education, etc). Furthermore, in Guatemala, fair access to primary schooling, inparticular for the second quintile, has still not been achieved. Broad-based expansion of coverage is needed in secondary education in order to correct the current regressive public subsidy. At the secondary level, the public subsidy per capita i s already very inequitably distributed, as are gross enrollment and completion rates. If our calculations over-estimate the public subsidy to the poor than this inequality i s even greater. The key issue for the secondary level i s how to provide wider access to schooling for the middle-lower and lower classes. This topic has already been dealt with in Chapter 111. Essentially, in all Central American countries, greater public spending at the secondary level should be funneled into expanding access to quality secondary to, at least, the three middle quintiles - as is currently the case in El Salvador. 141 Q. I Q.II Q.III Q.IV Q.V Total HONDURAS2003 Public Subsidy 47,780,247 42,l38,533 33,808,373 25,535,896 13,623,937 162,886,986 G b i c Subsidv 29.3 25.9 20.8 15.7 8.4 100.0 IPer Capita Public Subsidy II 144.6 II 146.6 II 145.5 I1 136.2 II 101.6 II 139.0 II IAGU 2001 I I I I I 1 Public Subsidy II19,811,143 19,287,171 II16,952,143 13,823,743 5,962,286 75,836,486 G b l i c Subsidv 26.1 II 25.4 22.4 18.2 7.9 100.0 I Per Capita Public Subsidy 1I 92.1 I 99.2 II 93.1 III 84.9 III 50.9 II1 87.0 III IEMA 2000 I Public Subsidy II 63,667,050II 58,416,600 60,087,750 II II 49,142,100II 24,629,100II 255,942,600II Yo of Public Subsidy 24.9 22.8 23.5 19.2 9.6 100.0 Per Capita Public Subsidy 139.5 128.8 142.5 141.2 94.0 131.8 EL SALVADOR2002 Public Subsidy 50,190,937 46,810,l 19 39,300,239 29,204,823 16,871,860 182,377,979 ~ Yo of Public Subsidy 27.5 25.7 21.5 16.0 9.3 100.0 Per Capita Public Subsidy 213.0 215.8 209.9 187.9 121.7 195.3 Table 4.4: Per Capita PublicSubsidy in Secondary I I HONDURAS2003 Public Subsidy 10,142,564 15,260,341 22,320,651 27,306,573 22,158,753 97,188,881 Yo of Public Subsidy 10.4 15.7 23.0 28.1 22.8 100.0 ~ Per Capita Public Subsidy 47.9 68.9 106.0 136.5 132.0 96.1 NICARAGUA2001 Public Subsidy 1,203,676 2,435,933 3,320,749 4,515,087 2,814,707 14,290,152 Yo of Public Subsidy 8.4 17.0 23.2 31.6 19.7 100.0 Per Capita Public Subsidy 8.1 19.2 24.5 32.7 25.0 21.6 GUATEMALA2000 Public Subsidy 2,993,312 5,536,744 8,838,992 13,437,704 10,481,560 41,288,312 Yo of Public Subsidy 7.2 13.4 21.4 32.5 25.4 100.0 Per Capita Public Subsidy 9.8 19.5 26.0 34.7 34.3 25.4 EL SALVADOR 2002 Public Subsidy 14,255,336 21,916,623 23,524,654 23,786,184 17,712,328 101,195,125 Yo of Public Subsidy 14.1 21.7 23.2 23.5 17.5 100.0 Per Capita Public Subsidy 86.2 117.8 140.2 141.6 112.2 119.8 142 Reduction in the public subsidy for tertiary education. Finally, due to the highly regressive nature of the public subsidy in tertiary education it would make sense in countries such as Honduras, Nicaragua and Guatemala, which spend as much or more in tertiary as they do in secondary education, to decrease the amount of public expenditure at that level. This is, however, a very sensitive and political subject in Central America and will be difficult to modify in the short or medium term. Ideally, an effective system would be implemented in which students who could afford to do so would contribute more co-financing at public universities and the share of quality private universities would expand. This is easier said than done given the pressures that exist in all countries for the financing of the public university system. This measure seems, however, unavoidable if countries wish to significantly improve the equity of public education subsidies. IV. Private Household Education Expenditures Also Demonstrate Inequitable Education Financing An important factor that contributes to low enrollment of the lower income quintiles - particularly at the secondary level - both in the public and private sector, is the magnitude of both direct and indirect private costs of schooling. This section looks at the magnitude and composition of household costs of education. Included in private household costs are such things as school fees, learning materials and books, exam fees, and uniforms. Tables4.5 and 4.6 show annual private household expenditures on education by income quintile in real terms and as a proportion of annual household income.177They show that, while contained at the primary level, private costs to schooling are more substantial at the secondary level. Private household education costs can be prohibitive to the poor, particularly in Guatemala and El Salvador. In Guatemala, the financial burden of secondary education is heavy for the first (poorest) quintile and substantial for the second one. In El Salvador, the burden i s substantial for the poorest quintile. Compounding this difficulty is the fact that most families throughout Central America, particularly poor families, tend to have multiple school-age children. In Guatemala, an average household has more than three children between ages 1and 15. A poor family with three children of respectively 8, 13 and 16 would have to spend 145 percent of its household income to send their children to In El Salvador, a similar family would have to spend about 40 percent of its income. In Nicaragua, private costs appear significantly 10wer.l~~There, a family with three children would only spend about 10percent of household income to keep their children in school. The removal of all compulsory fees in Nicaragua in 2002 probably contributes to explain these findings. Recall, however, from Chapter 111, that household expenditure for 177 No data was available for Honduras, unfortunately. 17* Such an analysis for Guatemala is confirmed by Edwards (2004), who concludes that a key reason for low coverage at the secondary level is undoubtedly related to the high direct, out-of-pocket cost of schooling. Edwards also highlights the sudden discontinuity in direct costs as key factor crucial in the decision of poorer students to not continue their studies. Using our data to illustrate this concept, a poor student having attended public primary schooling and wanting to continue onto secondary schooling in the public sector would have to go from private costs of about US$ 22 per year to US$ 155 per year. If he were to continue in the private sector for lack of room in the public sector, he would go from US$22 per year to US$ 300 per year! 179 The figures obtained for Nicaragua are broadly confirmed by Arcia (2003), who finds that private costs represent less than 3% of the poor non-food consumption in primary and about 6% of the poor non-food consumption in secondary. 143 secondary in Nicaragua is higher than public expenditure highlighting a serious problem in the I overall secondary financingstructureinthat country. Table 4.5: Annual PrivateSpendingper Student by IncomeQuintile (in US$) Total 144 Composition of private household costs. Uniforms, books and materials and school fees are the most important sources of expenditure in secondary education although the relative weight of each cost differs by country (Figure 4.7). The cost of books and materials is particularly important in Guatemala, where it is by far the major source of private spending. Developing scholarship programs, subsidizing textbooks for the poor, or other private cost-cutting measures should be considered in Guatemala. In El Salvador, private costs do not come predominantly from one area. El Salvador has already been quite innovative in introducing scholarships, but other options to address private costs may include strengthening exemptions from the payment of fees for low- income students or introducing conditional cash transfers. Figure 4.7: Composition of Private Expenditure inthe Public Sector for Secondary Education (%) I ElSalvador Guatemala 1materials OTransport W Others1 1 Nicaragua IPSchool Fees =Uniforms 0Books and materials OTransport and others Source: Household surveys V. Efficiencyof Public Expenditure Clearly, overall public expenditure levels have important implications on issues such as education quality and equity. But perhaps more important than the mere level or distribution of public investment i s how effectively that money is used to support learning for all children at all levels. 145 This section looks at two main subjects, first, the cost-effectiveness of education spending interms of educational outcomes per dollar spent, and second, the distribution of public education expenditure across spending-area such as salary, non-salary, and capital costs. A final section highlights Honduras as a case example. Overall we argue that the efficiency of public education expenditure i s low: Central American countries tend to perform poorly considering their education investment level and they divert too few resources to key inputs such as teaching and learning materials while spending the bulk of their resources on teacher remuneration without linking remuneration to teacher performance or effectiveness. A. COST-EFFECTIVENESS A simple way o f gauging the cost-effectiveness of public expenditure can be undertaken by comparing expenditure indicators with educational outcomes. In Figures 4.8 and 4.9 we plot secondary gross enrollment rates and primary gross completion rates by education expenditure as a proportion of GNP for a group of Latin American countries with similar per-capitaincome levels to those of this study (Costa Rica, with higher per-capita income is also plotted).'*' These figures do not give any definitive evidence but they do show that given similar income levels the Central American countries compare poorly to other Latin American countries such as Peru, Ecuador and Bolivia, which, on average, tend to get have better coverage indicators considering their education investment levels. Unfortunately, there was not comparable data on per capita expenditure to do the same exercise with this indicator of expenditure. 146 It is also interesting to compare the Central American countries themselves. With similar investment levels Costa Rica has higher educational outcomes than Honduras and El Salvador has higher outcomes than Figure4.8: EducationExpenditureIGNPandSecondary Gross Guatemala. Nicaragua also fares EnrollmentRate,2001/2002 somewhat poorly although as we saw earlier per-child and per- student expenditure in that 1 90 country i s particularly low I 80 (Figures 4.10 and 4.11 will L clarify this point). I o j --i!60 Figures 4.10 and 4.11 expand &a the comparison between Central LD 540 American countries by a I I comparing the primary and $20 ( E secondary completion with t820 measures of per-student public 10 education expenditure. In these graphs Nicaragua does especially well while Guatemala 1 2 3 4 5 6 EducationExpenditurdGNP does especially poorly.'81 Plots of secondary net enrollment rate Source:Educationministriesfor CA, WBMIDI for other countriesand expenditure. and secondary survival rates (not Figure4.9: EducationExpenditure/GNPandPrimaryCompletion shown here) point to a similar Rate,2001/2002 scenario. Overall, this analysis can offer some insights on the relative efficiency of expenditure but it i s still very simplistic. In particular, educational achievement, as measured by exam results, and equity outcomes should also be considered. If it is true that 2 Nicaragua manages to obtain good coverage and completion 1 2 3 4 5 6 7 given what it spends in secondary, it is also true, for EducationExpenditurdGNP instance, that these outcomes are Source: Education ministries for Central America, WBMIDI for other countries particularly inequitably and expenditures. distributed in that country (see '*'Assuming that we expect to find, as illustrated by Honduras, El Salvador and Costa Rica, a somewhat increasing but convex relationship between expenditure and completion in primary, which indicates that after a certain level increasing completion may be more difficult. 147 Chapter Iand expenditure incidence analysis below). Figure4.10: PrimaryCompletionRate and Education Expenditure,2001/2002 100 90 2 80 70 .-2c60 2 50 e 6 E 40 e 30 20 10 0 0 50 100 150 200 250 300 EducationExpenditureper Student(US$) Source: MINED Dataand Table 10.1 Annex X Figure 4.11: Secondary CompletionRateand Education Expenditure, 2001/2002 0 50 100 150 200 250 300 350 400 450 EducationExpenditureper Student (US$) Source: MINED Data and Table 10.1 Annex X B.EXPENDITURE DISTRIBUTION BY AREA In Table 4.7 and Figures 4.12 and 4.13 we report public education expenditure disaggregated by spending categories.182 On average, roughly 87 percent of education expenditure goes to recurrent costs. This is a similar pattern to that of much of Latin American and OECD c~untries.''~ lS2Unfortunately, we could not undertake this disagregation by education level due lack of data but we assume that the picture looks somewhat similar in primary and secondary. Another limitation concerns the fact that for some of the 148 recurrent recurrent(non- capital total (salaries) salaries) El Salvador (a) 320 63 aa 471 Guatemala (b) 467 (c) 33 500 Guatemala (d) 327 46 0 373 I Honduras(e) I 403 ( 4 I Honduras (f) 370 15 11 396 Nicaragua(9) 60 30 14 104 Nicaragua (h) 101 (c) 14 115 Costa Rica (in percent) (i) 91.6 8.4 (I) 100 Low share of non-salary recurrent costs. Figure 4.13 plots the shares of capital expenditure and recurrent expenditure, sub-divided into salary and non-salary items. In all countries, except Nicaragua, the share of non-salary recurrent costs, which includes teaching materials, teacher training, equipment, etc, represents less than 15 percent of total expenditure and less than 20 percent of total recurrent expenditure. These figures are below OECD averages.184 185 countries we only have information on the MINED, under-estimating capital expenditure. Accounting of recurrent versus capital expenditure is also not always precise in the different countries. In particular, as we will also see in the next sub-section, expenditure financed by external funds often tends to be accounted as investment while it is non- salary (or even salary) recurrent spending. An extreme case occurs in Nicaragua where all external sources are labeled as investment even when they are used for recurrent expenditure. Arcia (2003) made a careful re-classification of expenditure across economic categories in Nicaragua that we are using here (and which we also used in the above functional classification of expenditure). Finally, by being largely financed by external funds, non-salary recurrent and investment expenditures can fluctuate significantly across years making a one time cross-section analysis not as meaningful as a time-series analysis, which it was however difficult to undertake for all our countries. Our analysis shows, however, that beyond a couple o f exceptions the share of external funds was fairly stable in our countries over these last three years making this issue somewhat less serious. This percent is probably over-estimated in Nicaragua and Costa Rica, where only data on the MINED budget are available, but it may be under-estimated in El Salvador, Honduras and Guatemala due to accounting issues. The OECD average was about 20% of total education expenditure in 2000 (see OECD, 2002). It was possible to make this disagregation only on the funds channeled through Ministries of Education (with the exception of El Salvador). This share is, however, likely to be under-estimated in El Salvador, where capital costs are likely to also include some non-salary recurrent expenditure - see section on external funds. 149 Figure 4.12: Recurrent and Capital Expenditures,2002 Much of the recent increase in per- student educational spending at the primary level in Honduras, El Salvador, Costa Rica and Guatemala i s due to increases in the teacher wage bill. Teacher salaries have increased in real terms over the last several years. Common knowledge and research have shown that teacher salaries should be competitive with those of other professionals with similar educational backgrounds in order to attract and retain high quality teachers.'86 But teacher pay should also reward good teaching, and as in most other professions, pay and Source: Table 4.7 position should be contingent on teachers showing up and doing their job. Unfortunately, the recent salary improvements in Central America do not sufficiently reward good teaching nor avoid corruption or gaming. In most cases the recent salary increases in Central America corrected for longstanding low relative wages - and to a lesser extent class sizes have decreased causing the need for more teachers to be hired.' 87 These increases in the salary-share o f recurrent educational expenditure have been coupled with stable or even decreasing share for non-salary recurrent costs. This i s particularly true in Honduras as will be discussed in the case study on Honduras below. In the other countries the increase was less sharp and, at least in the case of El Salvador, salaries were, at least in part, linked to measures o f teacher performance. The cases of Costa Rica and Guatemala needto be further analyzed to see if salary increases resulted ineducation quality improvements. See for example Psacharopoulos,Valenzuela and Arends (1996) or Loeb and Page (2000). "'See The World Bank (2004b), Sanigest (2004) and CIEN (2003). 150 Figure 4.13: Salary, Non-salary Recurrent and Capital In sum, although the increase in Expenditures, 2002 expenditure can be partially explained by the, still limited, 100% expansion o f secondary coverage 90% (through the financing o f new 80% investment and teachers - in 70% particular in Nicaragua), and by the trend in capital expenditure (related 60% to both new investment and 50% reconstruction - in particular in El 40% Salvador"'), this analysis shows 30% that salary increases have probably 20% been the most significant driving 10% force behind the increase in public 0% El Salvador Guatemala Honduras Nicaragua 1 education expenditure. Little focus has generally been put on teaching materials, school supplies and Source: Table 4.7 comprehensive teacher training, which are potentially related to educational a~hievement.''~We can therefore conclude that there would be scope for: (a) increasing non-salary recurrent expenditure, both in absolute and relative amounts, in all countries (even in Nicaragua, for secondary); (b) maintaining the wage bill under control by maintaining efficient pupil-teacher ratios (which, in our Central American countries, are generally in line with the Latin American average) and holding stable real salaries (which, as seen in Chapter 11, are more than in line with the ones of professionals with similar qualifications); and (c) linkingreal salary increases to improved teacher performance and real qualifications (this could be particularly advisable in Nicaragua where expenditure per student i s still quite low). Finally, as a pre-condition for all these measures, it would be greatly necessary to improve the way recurrent and capital expenditures are accounted for (see below). The case of Honduras highlightssome of these challenges. c.CASE STUDY: LOW COST-EFFECTIVENESSAND HIGHINEFFICIENCY INHONDURAS In the international cost-effectiveness comparisons shown earlier (Figures 4.8 and 4.9) Honduras appears quite cost-ineffective. This i s further evident comparing the evolution of primary level test scores and completion rates with per student expenditure in recent years (see Figure 4.14). Between 1998 and 2001 per student spending increased significantly in Honduras. Education expenditure increased sharply in relation to GDP and the same happened with the expenditure per student per education level (which, in real terms, in primary, went from US$ 78 in 1997 to US$ 151 in 2001; and in secondary, from US$205 in 1997to US$280 in 2001). Where substantial external funds were directed to reconstruction after the earthquakes. 18'The availability or use of textbooks and school supplies is significantly associated with achievement in a few countries (El Salvador, Honduras, Guatemala - see Chapter 11) and achievement of the poor, through the multi-grade and bi-lingual education programs, in all countries. 151 Although we would not necessarily expect a clear improvement in outcomes over this same time period as improvements take time to observe, it is, nonetheless, worrisome that educational indicators got worse over the course of these years. Disaggregating the data to the regional level we further find that the generalized increase in real spending across regions was accompanied by a decrease in average test scores in all but three regions (see Figure 10.1inAnnex X). Figure 4.14: Evolution of Educational Expenditure, Test Scores andPrimary Completion Rate inHonduras (1998-2001) 160 140 120 100 *completion rate =$a 80 60 40 20 Source: Table 4.2, MINED data, UMCE 0 1998 1999 2000 2001 Honduras has had a substantial wage bill increase due to increasing teacher wages and inflated payrolls. Of the four countries under analysis, Honduras is the one where the distribution between salary and non-salary recurrent costs i s by far the most dramatic. An analysis of the evolution of personnel recurrent costs in real terms, illustrated in Figure 4.15 below, clearly shows that these costs were the driving force behind the recent increase in education spending in Honduras. Meanwhile, the share of non-salary recurrent costs has decreased. What was behind the wage bill increase? An analysis of teacher management and remuneration in Honduras makes it clear that the driving force behind the spending increase was a combination of increased teachers or class hours on the payroll and increased wage rate, for both primary and secondary.190While official figures show stable or even increasingpupil-teacher ratios, the payroll figures show substantial increases in the number o f teachers (at the primary level) and class hours (at the secondary level) which suggest the presence of "ghost" teachers and an over-estimation of class hours. It is highly advisable for Honduras to "clean up" the payroll records and establish an effective supervisory and monitoring system to eliminate ghost teachers and over-estimated class hours. This could be done by involving regional and/or local Ministry of Education offices and schools themselves. I9'See The World Bank (2004~). 152 On the salary side, the approval and implementation of the Teacher Statute in 1997 led to an increase in the wage rate through three main channels: (a) the indexation, in 1998, of the base salary to the minimum wage; (b) the introduction of new salary allowances in 1998 and in 2000; and (c) from 1998 to 2002, an increase o f the base salary, and, therefore, also of the allowances, at a nominal rate higher than the inflation rate. Overall, the combination of these factors resulted in real increases o f the yearly gross wage o f about 40 percent over the 1998-2002 time period which, at least in primary, were the main driving force behind the wage bill and the unit cost increase over that same time-period.191 192 Figure 4.15: Honduras-Determinants of Wage Bill Increase 180 160 140 120 *Wage bill (per student) 4X - 100 E 80 -Pupil-teacher ratio 60 40 Source The World Bank (2004~) 20 0 1998 1999 2000 2001 2002 The salary increase was largely necessary to make up for the erosion in real salaries that had occurred during the 1990s and to fill the gap between the salaries o f teachers and those of other public employees. However, the two last increases have more than filled that gap (Chapter 11).The impact on teacher qualifications has been limited by an inadequate application of the academic allowance, and the current salary scale does not seem to have adequate mechanisms for monitoring or evaluating teacher effort and performance (Chapter 11).Additionally, the payroll i s full of civil servants who do not teach. In this context, it is not surprising that there is no observable improvement in educational outcomes. VI. The Role of External Funds External funds - both grants and loans - play an important role in education in Central America, particularly in the four countries under examination in this report. When poorly used, they can worsen the inefficiencies o f the sector and increase foreign debt, but when used strategically, they can trigger positive changes in cost-effectiveness, sector priorities, and the magnitude of total 19' Depending on the source (official data or survey data), deflator used and salary concept (gross or net), real increases fluctuate between 30 and 42 percent over the 1998-2002 time period, which explains the small discrepancy with Chapter 11. 19* The trend in the salary increase has recently been reversed. The realignment o f the base salary increase on the inflation rate has been just negotiated with the IMF as well as a freezing of the salary allowances to their current nominal level. 153 financing for the sector. We start below by reviewing some basic facts about external funds (magnitude and composition), to then reach a few conclusions and recommendations regarding the changing characteristics and role of external aid. A. MAGNITUDEAND COMPOSITION OFEXTERNALFUNDS Table 4.8 reports the relative share of external funds in all countries, focusing, when possible, on the time period 2000-2002.'93A limitation of this information i s that, with the exception of El Salvador, it only covers public expenditure channeled through Ministries of Education, whereas significant support for the education sector, in particular for investment expenditure, i s also provided through public sector actors (such as social funds) or non-governmentalorganizations. Table 4.8: Proportionof National and External Fundsin Public Education Expenditure, 2002 US$millions (unlessotherwise specified) El Salvador (a) Honduras (c) Nica. (c) Guatemala (c) 2000 1 2001 (2001(b)l 2002 2000 I 2001 I 2002 2002 2000 2001 2002 I ~~ =finds 1 356 I 388 I 393 1 412 I 279 1 342 1 358 1 75 1 338 II380 II 302 External funds 30 40 126 59 36 34 38 40 11 10 71 Total 386 428 519 471 315 376 396 115 349 390 373 % of external funds 8 9 24 13 11 9 10 35 3 3 19 Central American countries have different degrees of dependence on foreign education funding. Central American countries differ substantially in terms of their use of foreign funds. Nicaragua gets the highest proportion of its education budget from external funds, accounting for one-third of all the expenditure channeled through the Ministry o f Education. External funds representing about 10-15 percent of the public education expenditure over the time period 2000- 2002 in Honduras and El Sa1~ador.l~~Guatemala has the lowest dependence on foreign funds. Substantial sharesallocated to capital and non-salary recurrent expenditure and to primary. Table 4.9 and Figure 4.15 report a disagregation of external funds by economic categories for three countries, also reporting how much of each category of expenditure i s financed through these funds.'95It should be kept in mind, however, that it is difficult to establish how external funds are allocated due to poor accounting procedures in the region. In Nicaragua all external funds are considered capital expenditure, for example. This happens to some extent in El Salvador as well. In Honduras, a substantial fraction of the external funds are classified as transfers with no further disagregation. Given these limitations, any analysis of the use of external funds needs to be taken with a grain of salt.'96 193 There was no recent data with which to compare Costa Rica. 194 2001 was an exceptional year for El Salvador if we add rehabilitation loans that also benefitedthe education sector. '95 Unfortunately we could not obtain detailed information on Guatemala and Costa Rica. 196 Insome cases we had to make estimations to be able to get closer to the true allocation. 154 Table 4.9 and Figure 4.15 reveal that external funds are primarily allocated to capital and non- salary recurrent costs and finance a substantial extent of the capital and non-salary recurrent expenditure of Nicaragua and Honduras. Evidence from Honduras and Nicaragua further suggests that external funds are largely invested in primary or at most basic education. In El Salvador, where there was a large World Bank project for secondary education, we expect the allocation to be less concentrated in basic education. b r Table 4.9: Allocation of External Funds across Economic Categories 2002 US$ million (unless otherwise - I `ndicated) El Salvador (a) Honduras (c) Nicaragua ( 4 2000 2001 2001 (b) 2002 2001 2002 2002 Investment 30 (22) 40 (18) 126 (104) 56 (22) 15 11 12 In % of total investment 73 (67) 71 (51) 81 (78) 64 (41) 79 100 86 Recurrentnon-salary (7) (19) (19) (29) 7 12 18 In % of total recurrent (13) (24) (23) (31) 100 80 60 non-salary Salary (1) (3) (3) (5) 12 15 8 In % of totalsalary (0) (11 (1) (2) 3 4 13 Total 30 40 126 56 34 38 38 Source: Own elaboration on the basis of the sources of Table 4.1 (Nicaragua: elaborationsof Arcia (2003) on the basis of the sources of table 4.1); (a) In current US$ million. Between brackets are the figures we obtain when we assume that the expenditure classified under human capital finances for 85% non salary recurrent expenditures and for 15% salaries; (b) Includes loans for rehabilitation after earthquake; (c) Only MINED expenditure. In 2001 assumes 15% of transfers go for salaries, 75% to capital and 10%to recurrent non salary; in 2002 assumes 15% of transfers go to salaries, 50% to capital and 35% to recurrent non salary; (d) Only MINED expenditure (pre-primary, primary, and secondary). Additional evidence i s provided by a recent effort undertaken by the OEI (Organizucidn de Estudos Zberoarnericanos) to map out the areas where the different donors intervene in Central America.lg7The OEI map shows a strong concentration of donors inthe sector "quality and equity of basic education." Within this sector, some areas, such as national quality assessment systems, pedagogical and curricular innovation, teacher training, infrastructure, equipment and teaching materials, multicultural-bilingual education and multi-grade schooling, are particularly invested in, with several cases of duplication between activities and projects of the different donors. Other areas, such as specific compensatory programs, educational statistics systems, institutional strengthening and decentralization, including school based management, are relatively less covered. There is much less concentration in the sector "youth, secondary education and preparation for the labor market", and, within that sector, the most dynamic area i s youth at risk. Little emphasis, finally, i s being put on tertiary education, or science and technology. OEI (2004a). 155 Figure 4.16: Investment, non salary recurrent and salary expenditure financed by external and total funds El Salvador,2002 Honduras, 2002 ~ 400 350 300 / 100 I 50 50 I 0 0 Investment I I Financed by external Recurrent non-salary Salary Investment Recurrent non-salary Salary funds +Total I I I Financed by externalfunds +Total I I ' Nicaragua, 2002 I 60 50 v) I 'e$ 30 a 20 Investment Recurrent non-salary Salary i 'I byexternalfunds+Total Financed 1 Source: Table 4.9 Moving beyond traditional aid. This basic evidence suggests a few conclusions and recommendations, but we should be generally careful in interpreting it. International cooperation in education in Central America is driven by a number of key donors that predominantly invest in specific activities of the education sector. Decisions where to allocate the funds are taken jointly with the governments, but with strong influence o f the donor. Inthis donor-driven context the way external funds are reported and accounted for and how donors coordinate are critical. In other words, it will be imperative to improve the reporting and accounting of external funds in all countries to obtain reliable overall expenditure data. It would not be too difficult to agree with the different donors on a few simple guidelines for reporting education expenditure and to work more closely with Ministries of Finance to establish a consistent accounting system which records all expenditures (including donations) under a few clear expenditure ~ateg0ries.l~~ Poor coordination of donors, which produces duplications in certain activities and insufficient funds in others, reduces efficiency in the use o f external funds. It i s important to work toward improved coordination among donors to avoid duplication between activities. The new Education A reform of the SIAFI in Honduras is, for instance taking this direction. 156 For All-Fast Track Programs in Honduras and Nicaragua at the primary level represent a shift from traditional project-based approach to a programmatic sector-wide approach, with five main focuses: (a) policy dialogue; (b) results; (c) sustainability; (d) donor harmonization; and (e) in- country dialogue. It i s hoped that by helping design and implement a consistent and coordinated framework for intervention in the primary sector these initiatives will maximize the impact o f external funds in the sub-sector. Similar initiatives may, however, also be needed for the secondary cycle. The allocation of external funds across spending categories and education levels is not necessarily critical in that money is to some extent fungible. In other words, higher allocation of external funds on, for instance, non-salary recurrent expenditure, may lead the countries to spend less on these items from their own national funds and free up those funds for other purposes. From the evidence presented above, this may in fact play out negatively if we want countries to be absorbing more of these items in their national funds for sustainability. It may then make more sense to use a greater share of external funds to support salaries. Overall, what will really count i s having countries reach specific targets on their overall financial envelopes, both in terms of magnitude and composition of expenditure; and getting governments and donors to agree on the same targets. The new programmatic sector-wide approaches will ensure that this happens by establishing sector-wide targets on educational outcomes (in terms o f completion, quality, etc) and expenditure. As we shift towards multi-sector programmatic approaches, such as those being supported by the World Bank in the poverty-reduction strategy credits and social sector programmatic loans, donors will not even have to come to agreement with governments on specific activities to be financed anymore butjust on a set o f conditionalities. In Honduras, for instance, the Poverty Reduction Support Credit (PRSC) established strong conditionalities related to expenditure reporting, efficiency and composition across education levels. Conclusions The magnitude of public education expenditure is still an issue in Central America. Although all countries have increased public education expenditure since the mid-1990s, the level of education expenditure in terms of GDP and per age-appropriate child i s still below the Latin American average (with the exception o f Honduras' GDP percentage). A similar picture holds true when comparing per-student education expenditure inprimary and secondary with countries with similar per-capita income (an exception i s again Honduras, where, however, there are major efficiency issues). Additionally, the secondary education share of the education budget i s low in all countries, reflecting the still insufficient priority given to that education level in all countries. On efficiency grounds, Central American countries do not perform particularly well. El Salvador and Nicaragua are more efficient than Honduras and Guatemala when comparing primary and secondary enrollment and completion rates with education expenditure in terms of GDP or per- student expenditure. To some extent, this difference in cost-effectiveness can be associated with higher shares of non-salary recurrent and capital expenditures in the budgets of these two countries, which ensure higher investment in infrastructure and in learning materials and teacher training. In terms of equity and educational achievement outcomes, however, Nicaragua i s clearly 157 not performing particularly well.'99 Efficiency is low in Honduras where the increase in expenditure per-student was generated by a substantial increase in personnel costs not associated with improved educational outcomes (the share of non-salary recurrent expenditure in Honduras i s very low). Honduras and Nicaragua also appear to be particularly dependent on external funds for non-salary recurrent expenditure and investment.200 Donor coordination issues have limited the impact of external funds somewhat, but the new sector-wide programmatic approaches have the potential o f addressing this problem while also helping to establish common outcome indicators and expenditure targets for the education sector. Finally, while the public education subsidy appears to be reasonably progressively distributed in primary, and the gross enrollment rate quite homogeneous across quintiles, a different situation holds for secondary and tertiary where the public subsidy i s regressively distributed, reflecting large inequities in access and completion across quintiles. Even at the primary level, the fact that a much higher proportion of children of the lower quintiles still leave school without completing the cycle suggests that the subsidy i s probably less progressively distributed than it should be. Additionally, private costs are a considerable burden in Guatemala and a substantial one in El Salvador. The main (general) policy implications of this chapter can be summarized as follows: Central American countries should spend at least the Latin American average in terms of GDP to reach an acceptable public expenditure level per school-age population. This is valid for Nicaragua, El Salvador and Guatemala; The share of secondary education should increase in all countries, although this should happen within an increasing overall budget in terms of GDP to ensure more resources also for primary education. It i s particularly important that Nicaragua increases its expenditure per student in secondary education to ensure higher quality education at that level; The share of non-salary recurrent costs should be increased to at least 20 percent of total education expenditure (also by education level), to ensure the school supplies, teaching materials and teacher training necessary for improving educational achievement; The wage bill should be kept under control by maintaining efficient pupil-teacher ratios and either stable real and relative salaries or salary increases effectively linked to teacher performance and quality. It i s imperative for Honduras to revise its payroll figures and implement an effective teacher monitoring system. Real salary increases should be accompanied by improved teacher performance or meaningful teachers' qualifications (see Chapter Ifor more concrete actions); In all countries, it is imperative to improve the reporting and accounting of external funds, to obtain reliable overall budget data. Donors and countries should agree to a few simple guidelines for reporting education expenditure and to work more closely with the 199As will also be shown inChapter V analyzing autonomous schools. 2ooThere are no data on Guatemala. 158 Ministries of Finances to establish a consistent accounting system which records all expenditures under a few clear expenditure categories; (f) External funds should be better coordinated and leveraged. The new Education For All- Fast Track programs (EFA-FTI) are promising, as are the new multi-sector programmatic approaches, to ensure higher donor coordination and establish common outcomes and expenditure targets for the education sector; (g) Better access to aualitv schooling should be provided to the Door (which should be partly reflected in higher unit costs per poor student) to ensure higher completion; (h) Broad-basedexpansionof secondarv education should be encouraged; (i)Public subsidies to tertiary education should be reduced in Honduras, Nicaragua and Guatemala, by leaving higher co-financing responsibility to public universities and more room for private universities; (j)The burdens of private costs should be addressed in Guatemala and El Salvador. Developing targeted scholarship programs and subsidizing textbooks for the poor could be possible options for Guatemala. El Salvador should strengthen exemptions from the payment of fees for low-income students and introduce conditional cash transfers providingfull coverage of private costs inpublic or private schools. 159 Chapter V: A Comparative Analvsis of School-Based Manaqement In Central America Introduction This chapter analyzes and compares the impact of education decentralization reforms in El Salvador, Guatemala, Honduras and Nicaragua. Decentralization i s the process of re-assigning responsibility and corresponding decision-making authority for specific functions from higher to lower levels of government and organizational units.20'In the context of education reforms, decentralized functions frequently include policy-making, revenue generation, curriculum design, school administration, and teacher management. Typically decision-making authority for these types of functions i s decentralized either to regional or municipal government offices or to schools themselves. Decentralization reforms take many forms and correspondingly, have many objectives. In Central America a primary objective of these reforms has been to improve the efficiency and effectiveness of education by increasing school autonomy and community participation and the autonomy and capacity of local and regional education offices and stakeholders. Decentralization through municipalization has also taken place in some moments of time, for instance in Nicaragua, but has never reached the breadth and depth that decentralization to the school level has, with the consequence that municipalities are generally only a minor player in education delivery in Central America202. Internationally, a large amount of research exists that analyzes the impact of decentralization reforms which shift certain decision-making powers to the school level (also called school-based management reforms, or SBM) on education outcomes203.O f the Central American reforms more research has been conducted on the Nicaraguan and Salvadoran reforms, and less on those of Honduras and Guatemala, as their adoption and implementation i s more recent. Research i s substantially more limited on how the reforms have expanded community participation and empowerment and on to what extent and in what ways schools exercise meaningful autonomy. Finally, very few studies compare empowerment and education outcomes across these four countries. This chapter will be organized in five main parts: section Ipresents, very succinctly, the analytic framework that has guided this chapter; section I1discusses and compares the design o f the SBM models in the four countries under investigation; section I11analyzes the impacts the reforms have had on both empowerment and education results; section IV discusses how variations in reform design, country contexts and actors' assets can explain differences and similarities in result; and section V provides recommendations on how to improve the impact of the SBM on empowerment and quality of education and guarantee the sustainability of the reforms. Paqueo and Lammert (2000). *02 In some countries such as Honduras, however, municipalities play a noticeable "informal" role in education delivery by financing some infrastructure and, at times, even teachers. 203Although little of this large amount o f research is systematic or definitive. Generally speaking, the reader will need to keep in mind, throughout this chapter, that it i s often difficult to associate reforms with outcomes, even if the most sophisticated technical tools are available. I n this sense, all results needto be interpreted with caution. 160 1. Conceptual Framework The conceptual framework draws on the World Bank's recent work on empowerment and accountability as well as on principal-agent literature in order to assess the effects of decentralization reforms on empowerment and education results. In summary, the framework argues that: 1. Decentralization reforms have a positive influence on the efficiency204and effecti~eness~'~ of education service delivery largely because decentralization: (a) enables the service provider (in this case the school) to make use of information about local preferences (asymmetric information argument), and (b) increasesthe opportunities for the service receiver (inthis case the community) to hold the service provider accountable (accountability argument), which, in turn, can improve teaching and learning; 2. The ability of the school to make use of its knowledge about local preferences and of the community to hold school staff accountable de ends on the extent to which the reforms have led to both community206and school empowermentFO7; 3. School and community empowerment will not be enough to ensure the fulfillment of some national objectives, such as higher coverage or national targets in learning achievement, due to the "positive externality" argument208.Equity concerns and local institutional weakness may also not be fully addressed, as risks of capture by local elites. In this context, the government will need to find mechanisms which lead the agent (Le. the school) to pursue its national interests. This will require an effective accountability relationship between the policy maker and the organizational provider - Le. the school. The recent WDR209,which also highlights the importance of the ~~ '04 We can define efficiency in two different ways, as technical (productive) efficiency or social (allocative) efficiency. Technical efficiency is about producing a higher output for similar costs or the same output for lower costs; social efficiency i s about choices that reflect more closely consumers' preferences. '05 Effectiveness (although not necessarily a very precise concept) is more about impact on outputs and outcomes, such as the coverage of the services, their quality (measured for instance by learning achievement), their impact on poverty reduction and social development, the equity o f delivery, etc. '06 Community empowerment refers to parents' ability to have a stronger voice vis-&vis school staff, for instance by giving parents the power to hire and fire teachers or to have a say in teaching methods. Strengthening accountability to local communities, through community empowerment, i s particularly important because, borrowing from the principal-agent literature, the agent, or school, once given decision-making autonomy will often have the temptation to use it opportunistically. This means that the school might put its own interests before the achievement of the national objective for which decentralization was undertaken in the first place (i.e. improving student learning), thereby taking advantage of the fact that the principal (in this case the ministry of education) will typically not observe the true effort and ability of the agent (moral hazard issue).In this setting, the community can act as a second principal (or client) and help solve the incentive issue. This approach i s also shared by the recent World Development Report (The World Bank, 20040, which highlights the key role of the accountability relationship between the clients (in this case the community) and the providers (in this case the school staff), defined as client power, in having decentralization working in practice. *07 School empowerment (also referred to as school autonomy) involves strengthening the school's decision-making ower (vis-&vis the education authorities), for example on pedagogical or administrative matters. `Os In other words, education has broad societal benefits (on growth, social development, etc) that are not fully captured by private benefits. As such, without national intervention, education delivery might reach a socially sub- optimal equilibrium level. *09 The World Bank (2004e). 161 accountability relationships between policymakers and organizational providers, refers to this accountability relationship as the compact; 4. Finally, the degree to which schools and communities become empowered and the effectiveness of the accountability relationship between policymakers and education providers i s influenced by both the: (a) assets of actors and communities, which include skills, and information as well as organizational, psychological, human, financial and material assets, and (b) the context in which the school and community exist. In fact, as we will see, assets and context also have a direct impact on the teaching and learning environment as well as on students' education outcomes. On the formal side, the context includes the specific reform design (dejure provisions regarding the operation of the program, most notably the functions and responsibilities decentralized to the school210). On the informal side, the context consists of a range of socio- political factors that determine whether the reform, once adopted, can be sustained, and to what extent it i s implemented according to the dejure provisions.211Context and assets are interrelated. The framework i s graphically presented in Figure 5.1. The chapter will have a focus on the impact o f assets and reform design on empowerment and education results. *lo Reform design is important for community and school empowerment. For instance, the extent to which parents become empowered vis-&vis school staff is influenced, among other things, by the types of rights and responsibilities that are devolved to them or the degree to which the reform makes provisions for relevant parental capacity building. Similarly, school empowerment vis-&vis the education authorities will be enhanced by a clear-cut decentralization of administrative functions, including extensive teacher management responsibilities and special training programs for school directors. 2 " Hthe decentralization reform works in practice or defacto is influenced by a series of socio-political factors. ~ ~ Factors that have a bearing on school empowerment, for example, may include the relation between the government and the teachers' unions, the efficiency o f the administration, and informal spaces for autonomous decision-making. In turn, factors that may influence the extent of community empowerment include a community's tradition of social mobilization, the extent of social division between parents and teachers, and the implementation and operation of policies, such as standardized evaluation and social auditing systems, which ensure the availability o f information about educational outcomes at the national and local levels. 162 Figure 5.1: ConceptualFramework Formal Context: 1 Reform Design affects School Empowerment 2 ReformDesign affects Community Empowerment 3 Reform Design affects accountability directly (compact) and information 4 lnformal Context affects SE 5 Informal Context affects CE 6 Assets affect SE 7 Assetsaffect CE 8 Informal Context and Assets influence eachother 9 Formal Context and Assetsinfluence eachother 10 Formal Context affectsTeaching and Learning Environment 11 Formal Context affects Outcomes 12.Informal Context affectsTeaching and Learning Environment 13 Informal Context affects Outcomes II. Comparing EducationDecentralizationReformsin Central America A. HISTORY OVERVIEW OFDECENTRALIZATION AND REFORMSINCENTRAL AMERICA This sub-section provides an overview of each of the four countries' education decentralization reforms, discussing the historic background and rationale that led to the adoption of the reform, and describing reform features such as institutional arrangements and the criteria communities have to fulfill to participate in the programs. Sub-section B compares specific aspects of the reform designs in more depth. These include the nature and extent o f community participation, the frequency and formula for financial transfers, and teacher salaries and benefits. Overall, the community-based school management programs implemented in Central America have been aimed at increasing enrollment, strengthening community participation, and improving efficiency. A less frequently cited objective has been the improvement of education quality through increased responsiveness to local needs and interests. In three cases-EDUCO (El Salvador), PRONADE (Guatemala), and PROHECO (Hondurasbthe main objective has been to increase enrollment in isolated rural areas affected by conflict, poverty or natural disasters. The School Autonomy Program in Nicaragua aims instead to give voice to parents and civil society on educational issues and, in this way, increase operational efficiency.212 212Arciaand Belli (1999). 163 a) The Case of El Salvador El Salvador was the first among the four Central American countries under investigation to decentralize when, in 1991, the government officially launched its school-based management program, known by its Spanish acronym EDUCO (Educacidn con Participacidn de la Cornunidad - Education with Community Participation). Prior to the establishment of the program, it was clear that the country's basic education sector faced a number of serious challenges, such as low enrollment, high dropout and repetition rates, inefficient management, and low fiscal allocations for primary education.213In part this was due to the civil war that had ravaged the country in the preceding decade. At the same time, lack of access to public schools during the civil war had led many rural communities to organize their own schools -- staffed, financed and administered by the community membersthemselves. The Ministry of Education regarded community participation as key to expanding access, and, under the EDUCO initiative, decided to institutionalize these, already existing, community-run schools and open up similar ones throughout the country. The EDUCO initiative makes provisions for decentralizing some of the responsibilities of the Ministry of Education (MINED) to the regional and the school level, and for the participation of community members in school management. Both aspects create de jure spaces for school autonomy and community empowerment. The organizational structure underlying the EDUCO initiative is as follows:214 0 MINED enters into a contractual partnership with community education associations, known by the acronym ACE (Asociacidn Comunal para la Educacidn), so that these ACEs become responsible for education delivery to a given community on the ministry's The MOE remains responsible for transferring funds to the ACEs, establishing guidelines for the distribution of funds, producing teaching and learning materials, supporting the organization, legalization and training of the ACEs, and supervising schools. 0 EDUCO's coordinating unit, which reports directly to the MoE, provides the following: policy and technical design; promoting, supervising and evaluating implementation; coordinating with MOE with regards to training for school staff and supervisors; and providing training and technical support to the regional offices. 0 EDUCO regional offices have recently been phased out216.Their original mandate was to provide immediate technical and administrative assistance to the schools, including the provision of supervision, training, curriculum development and programs for school improvement. 0 District supervisors, replaced by Technical Liaisons (t&cnicosde enlace), serve as the link between the regional and central EDUCO offices and the EDUCO schools. They are responsible for promoting community participation through the following activities: administrative and 'I3Edge (2000). 'I4Since its inception, the organizational and implementation structure of the EDUCO program have changed. The two most important changes are that the system of the supervisors has been replaced with that of the so-called te'cnicos de enlace and that the ACEs have been assigned greater decision-making powers, including the hiring of the te'cnicos de enlace and of administrative assistants who are responsible for organizing and monitoring the school councils. These changes will be discussed in more detail in sub-section 1I.B. '15This section draws on Edge(2000), and Meza and Guzman (2004). 'I6To streamline the structure of the Ministry o f Education. 164 technical support to the ACEs, technical assistanceto teachers, dissemination of informationabout EDUCO and MINED, assistance with the creation of schools for parents, and liaison between ACEs and MINED. 0 ACEs are school-level legal entities whose membershipis drawn from the local community. They are responsible for administering the funds transferred to them by MINED, selecting, hiring and monitoring teachers, building and maintaining schools, and mobilizing the rest of the community in support of the school. EDUCO schools are exclusively established in rural areas and provide for pre-school and basic education (grades 1-9). Communities qualify for the EDUCO initiative if there are at least 28 studentsper grade in the community and no other education services are available.217 b) The Case of Guatemala Guatemala's school-basedmanagement reform is the National Community-Managed Program for Educational Development (Programa Nacional de Autogestidn para el Desarrollo Educativo, PRONADE). PRONADE, like EDUCO, is a SBM program that seeks to increase access in poor, rural, isolated areas, and foster community participation in school administration. Initially piloted in the early 1990s, the program's expansion was linked to the 1996 Peace Accords, which were signed at the end of almost forty years of civil war. Stipulated in the Peace Accords was the demand that the government decentralize its education system and that all children complete at least the third grade. In an effort to fulfill these objectives, the MINED and the Commission for Education Reform focused on the PRONADE experience as an effective education delivery system. They did so for three reasons: (1) pilot programs, under implementation since 1992, appeared to be successful, though a formal evaluation was lacking; (2) the MINED believed that community involvement was essential to rapidly increase access to education, and (3) PRONADE's objectives - decentralization, community participation, and indigenous and linguistic diversity - were in line with the Peace Accords and the overall strategic direction of the government.218 PRONADE decentralizes important functions to community school councils, the COEDUCAs (Comitks Educativos), dejure giving the school a greater degree of decision-making autonomy in its relation with the education authorities and enabling parents to have a stronger voice in their relation with school staff. Functions and responsibilities are divided among the different organizational layers as follows:219 oPRONADE's implementation unit, headquartered in the capital, is responsible for planning, directing and evaluating the program's administrative and operational activities, including determining the geographical areas the program covers, hiring and supervising the ISEs (see below), and signing the legal covenant with the COEDUCAs. *The 21 departmental offices are in charge of coordinating policy implementation at the departmental level, assisting with the identification of communities that lack schools, processing '"Cuellar Marchelli (2001). *18 Vaierio and Rojas (2004). ' I 9Information in this section is drawn from MINEDUC (2003a), MINEDUC (2003b), MINEDUCDP Tecnologia (2002), and Valerio and Rojas (2004). 165 the financial and administrative data gathered by the ISEs, and supervising the school feeding program. *Educational Service Institutions (ISEs) are NGOs hired by the implementation unit to identify communities' educational requirements, organize and assist COEDUCAs in obtaining legal status, provide capacity building opportunities to teachers and school councils, and maintain updated information on the school's administrative and financial matters. oCOEDUCAs are legal entities that draw membership from the local community. Their responsibilities include: selecting, hiring, paying and supervising teachers, monitoring student attendance, managing funds transferred from the central authorities, organizing school feeding programs, and defining the school schedule and calendar. Similar to EDUCO, PRONADE schools are established in remote, rural areas and provide for pre- school and primary education. Communities may qualify to participate in the PRONADE program iftheymeetthefollowingfourcriteria: (1)theydemonstratetheabilityandinteresttomanagethe new school, (2) the nearest public school must be at least three kilometers away, (3) there are at least 20 pre-primar and primary school-age children, and (4) there is presently no official teacher inthe community.dl c) The Case of Honduras Honduras' community-based education program, PROHECO (Proyecto HondureAo de Educacio'n Cornunitaria) was launched in March 1999 with the objective of enhancing access to education and fostering community participation in school-related decision-making. Studies the Ministry of Education had carried out in 1997 showed that more than 14 percent o f school-age children were not enrolled in schools, 85 percent of these in rural areas. Building on the experiences in El Salvador and Guatemala, the Honduran government decided to use a school-based management model to address these deficiencies and establish new pre-school and primary schools in remote rural villages. B y the end o f 1999, PROHECO schools existed in more than 500 communities. Responsibilities for the implementation o f PROHECO are divided among stakeholders as follows:221 0 PROHECO's coordinating unit, located at the central level, as part o f the MINED, i s responsible for establishing the overall operational policies and strategies, including curriculum design, and for coordinating the program's technical and financial activities, including capacity building,community participation, collection of statistical data, and monitoring and evaluation. 0 MINEDdepartmental offices are responsible for raising awareness about PROHECO in the communities, and for overseeing the organization and fostering o f the AECOs (see below). They should also coordinate and oversee the work of the promoters. 0 MINED district offices support schools with the application of the national curriculum and the collection of school data, such as enrollment and drop-out rates. They also help identify communities in which PROHECO schools will be established. 220The World Bank (2001a). 221This section draws on Secretaria de EducacidnPdblica (2003). 166 0 SocialPromoters (promotores sociales) are in close contact with the communities andthe school councils, helping with the identification of communities, providing training to the AECOs, and advising on questions of community participation and budget oversight. 0 The school council, AECO (Asociacio'n Educativa Comunitaria) is a legal entity that draws its membership from the community and that has the following functions: selecting and paying teachers, monitoring teacher and student attendance and performance, managing funds for school materials and school improvement activities, and buildingand maintainingthe school. Communities qualify for a PROHECO school if they meet the following criteria: (1) they are located in a rural area, (2) there are at least 25 pre-school and primary-school age children, (3) the nearest school i s at least 3 kilometers away, and (4) the village is located in an area that was affected by hurricane Mitch222. d) The Case of Nicaragua Nicaragua's school decentralization reform was introduced under the presidency of Violetta Chamorro and her coalition government that replaced the Sandinista regime in the 1990 election.223The goals of the reform included raising the efficiency and effectiveness of education services. Greater participation and decision-making authority of parents and teachers was regarded as central to this end. First steps towards the "Autonomous School Program" (ASP) were taken in 1991, when so-called consultative councils (consejos consultivos) were established in all public schools to ensure the participation of the educational community, particularly parents, in school- related deci~ion-making.~~~ The initiative was deepened as a pilot in 1993 with the transformation of the consultative councils into full-fledged governing councils (consejos directivos) at the school level in 24 well-functioning secondary schools. The objective of this pilot was to test the feasibility of a program that not only envisioned community participation but also increased autonomous decision-making at the school level. The autonomousschool program divides responsibilities among different actors as follows225: 0 MINEDis in charge of instituting norms for the operation of schools and the basic content of educational programs, setting standards for the quality of school materials, the qualifications of teachers and schools' physical facilities, determining the national curriculum, and channeling the national budget for the financing of public education. 0 The municipal delegate of the MINED informs schools about ASP, guides schools through the application process, supports the program in the community while also acting as a liaison with the central ministry, and provides capacity buildingto the membersof the school council. 0 School councils (consejos directivos), both at the school and the municipal level, have legal status and are in charge of conducting the school's business, including hiring teachers and 222Secretariade Educaci6nPdblica(2003). 223Gershberg(1999). 224MECD(1993). Thisis adaptedfromThe World Bank(1995) andGershberg(1999). 167 administrative staff, maintaining the school building, making and overseeing budget allocations, generating additional financial resources (student fees), overseeing teacher performance and making pedagogical choices. Unlike the cases in El Salvador, Guatemala and Honduras, the Nicaraguan reform initially targeted secondary schools in urban areas. It was not so much concerned about the establishment o f new schools but rather the transformation of the governing structures of already existing schools. Since 1995 the autonomous school program has also targeted primary schools, using two different models: (1) one for urban schools which i s similar to the secondary school model, and (2) one for rural schools which are organized into municipal nuclei and where a centrally located school performs the management functions and operates the bank accounts for the school cluster226.This means that inurban primary and secondary schools councils operate at the school level, whereas in rural areas they operate at the municipal level. B. COMPARINGTHE DIFFERENTREFORMDESIGNS This sub-section explores specific features of the different reform designs, comparing a series of aspects that relate to institutional arrangements, the organization, membership characteristics and functions of the school councils, financial transfers, and teacher salaries and benefits. It should be noted here that programs like El Salvador's EDUCO and Nicaragua's Centros Auto'nornos have evolved over time: EDUCO more in its organizational arrangements, Nicaragua in its target schools and financial arrangements. When necessary, we will attempt to capture this evolution. The information on the characteristics and responsibilities o f the councils i s particularly important for understanding and comparing the de jure provisions for community empowerment and school autonomy. Overall, however, all the aspects of reform design analyzed in this sub-section will be key to understanding the impact of the SBM models on empowerment and educational outcomes, as will become clearer in section IV. a) Organizational and ImplementationArrangements SBM models can differ according to the organizational arrangements that are made for project implementation, including contracting and supervision of council trainers, teacher training, the identification of the communities that are to participate in the reform, as well as the organization, training and monitoring of the school councils. Table 5.1 shows how respective responsibilities are divided among different actors, indicating which actors are most closely in contact with the school councils and to what extent representatives or delegations o f the MINED are involved with them. El Salvador's EDUCO program, as the one longest under implementation, has undergone a series of organizational rearrangements which are discussed in the subsequent sections and/or Table 5.1. We will see in section lV that school councils' training and support i s a key factor of success of an SBM model, because it is key to the working of the models that parents get adequate training and support in a variety of areas, starting from administrative matters. King, Ozler and Rawlings (2001) and Fuller and Rivarola (1998). 168 Contracting of school council trainers. Initially, the El Salvador, Guatemala and Honduras programs did not differ substantially as to the actors responsible for contracting and supervising school council trainers (such as promoters, supervisors or ISEs). In each case, this function was carried out by the program's central coordinating unit, located within the MINED. While this is still the case in Guatemala and Honduras, in El Salvador the respective responsibility has been shifted to the school councils directly. El Salvador's school councils themselves are now in charge for contracting and supervising their trainers and therefore have substantially more powers than their counterparts in the other countries. Nicaragua's ASP was different from the start as, unlike the case in the other three programs, no system separate from the government was established to provide training and supervision to the school councils. In Nicaragua, the capacity buildingunit of the MINED- rather than an intermediary or non-government actor such as the promoter or supervisor -provides training to the school councils. In three of the four countries, El Salvador, Honduras, and Guatemala, teacher training is mainly organized at the regional or district level. Incontrast, Nicaraguan teacher training i s organized at the central level. 169 Table 5.1: Organizationaland ImplementationResponsibilities Organizational/ Implementation ElSalvador Guatemala Honduras Nicaragua Responsibility Contracting and Initially, contracting and PRONDADE's PROHECO's An equivalent supervision of supervision were done by the implementation coordinating unit non-government school council MINEDand EDUCO's unit system to that of trainers (e.g. coordinating unit, respectively. the supervisors, supervisors, More recently these promotores etc. in promoters, ISEs) responsibilities have been charge of council shifted to the ACEs. training and organization was not established in Nicaragua. Training of Initially, this took place in MINED's PROHECO MINED teachers EDUCO's regional offices. departmental coordinating unit More recently, training is offices plus MINED's provided by MINED's regional district offices training centers. Identification of Initially, this was done by the ISEs together with Promoters Teachers vote to participating supervisors with the help of the MINED's together with enter the program, communities central coordinating unit. These departmental municipalities and principal then have been replaced by the offices MINED'soffices submits a petition te`cnicosde enlace. operating at the to MINED's district and municipal delegate departmental level Organization of Initially, supervisors carried out ISEs Promoters MINED's school councils this function. More recently, municipal delegate these have beenreplaced by administrative assistants who are hiredby the ACEs. Training for Originally, MINED's regional ISEs Promoters MOE school councils training centers provided training. This function has now shifted to the tecnicos de enlace, operating at the department level. Monitoring of Initially, supervisors carried out ISEs Promoters Municipal school councils this function. They have now delegates been replaced by administrative assistantshired by the ACEs. Identification of schools. Differences in reform design play a role with regards to the way schools are identified and selected to participate in the SBM reform: while in El Salvador, Guatemala and 170 Honduras villages have to meet certain requirements to be eligible for a community-managed school, it i s at the discretion of Nicaraguan schools themselves whether or not they want to participate in the autonomous school program. The former have to meet such general criteria as location in remote, rural areas and often such specific criteria as a minimumdistance to the nearest school or minimum number of school-age children. Communities are identified on the basis of information available to the MINED as a result of relevant assessments and/or information exchanges with lower levels of government. Intermediary figures such as Honduras' promoters or El Salvador's tkcnicos de enlace then visit the communities to inform the community leaders of the SBM reform and agree with them to return to the community on a given date when further information will be shared with the rest of the community and the council will be elected. In Nicaragua, schools do not have to meet any specific criteria to participate in the ASP. In order to become autonomous, teachers, along with the school principal, must vote on the matter. If the vote i s in favor of joining the ASP, school staff sign a petition that indicates their inclination to participate in the program and then send the petition to the municipal delegate of the MINED for Organization of school councils. Differences in reform design are also evident with regards to the legal organization of the school councils. Once participating communities are identified and school councils elected, they typically receive support inorder to foster their organization and gain their legal status (persona juridica). The underlying idea for continued organizational support is that relevant procedures will be new to the communities and that they require support in legal and administrative matters. Without the legal status, school councils cannot open bank accounts, receive financial transfers, and perform functions on the government's behalf. In the case of Guatemala and Honduras, organizational support is provided by the ISEs and promoters, the intermediary figures most directly in touch with the school councils. In El Salvador, the district supervisors initially carried out this function. In the context of the program's rearrangements, however, responsibility for school council organization has shifted to the so-called administrative assistants who are hired by the ACEs. The Nicaraguan model also differs from the Honduran and Guatemalan ones as organizational support is provided by a government body, namely the municipaldelegate of the MINED. Training and monitoring of school councils. Following their election and legal recognition, school councils in all four countries are also provided training on budgetary and administrative matters. Differences exist with regards to whether the training is provided by intermediary figures or directly by the MINED. The latter is the case in Nicaragua. In addition to source of training provision, there also are some variations regarding the extent and duration of the training. In Nicaragua, content and amount of training provided vary on an annual basis but typically cover such topics as budgeting, accounting, planning and evaluation, and pedagogical matters. In El Salvador, the ACEs receive a total of 5 days of training on such issues as community development, budgeting, accounting, and personnel management.228Guatemala's COEDUCAs receive three days of training upon their legalization, plus an additional nine days as part of a 12- month follow-up service. The least amount of training is provided to Honduras' AECOs: training amounts to three days and covers a series of legal and administrative matters related to the school councils' functions. In all countries the same entities that are responsible for assisting school 227Fuller and Rivarola (1998). 228Meza (1997). 171 councils with their legalization are also in charge for monitoring their performance. Tasks that are subject to monitoring may include: book and record keeping, and procedural matters such as making announcements to the community at large and holding regular council elections. To sum up, in terms of organizational and implementation arrangements, the School Based Management reforms o f Honduras, El Salvador, and Guatemala are all quite similar while Nicaragua's ASP differs substantially from the others. Most notably, the Nicaraguan reform does not use intermediary actors, such as NGOs, to support local autonomy but rather schools interact directly with the MOE. b) Characteristics of School Councils School councils themselves vary greatly across countries and, indeed, across schools. They vary in size, constituents, and leadership. Table 5.2 details the basic characteristics of school councils in the four reforms. In El Salvador, Nicaragua and Honduras school councils are roughly the same size, between five and seven people, while in Guatemala they are substantially larger, at around 15 members. In Honduras, El Salvador, and Guatemala school council members are entirely drawn from the local community whereas inNicaragua the principal as well as one teacher representative also must sit on the council. In all four countries local communities members have the voting majority within the council. We will see that the composition of school councils has an impact on community empowerment with communities being somewhat less empowered in Nicaragua due to a predominant role of the school principal. Table 5.2: Characteristics of School Councils Characteristics EDUCO,El PRONADE, PROHECO, ASP, Salvador Guatemala Honduras Nicaragua Size of Council Five members Fifteen Six members Five council members members in school with less than 500 students. Seven in schools with more than 500 students. Council All are All are All are Five member council: Membership members of the members of the members principal, one community. community. of the teacher, three parents. community Seven member councils: principal, two teachers, four parents . Key decision- Community Community Community Principal and maker within members members members Community members council 172 c) Functions and Responsibilities of School Councils The nature and extent of the functions relocated to the school level determine - at least in part - how much space the decentralization reform creates for parental participation and school autonomy, in other words for their de jure empowerment. Decision-making powers that are devolved typically fall into one of the following categories: personnel management, pedagogy and curriculum, school maintenance and infrastructure, and budget matters. As Table 5.3 indicates, the decentralization reforms in El Salvador, Guatemala and Honduras are fairly similar, granting the school council a certain degree of autonomy in the areas of personnel management, school maintenance and infrastructure, and budget oversight (with some limited power on pedagogy in Guatemala). The SBM reform in Nicaragua reaches further, however, in theory granting the council some decision-making powers over pedagogical matters and budget allocation and size. De jure responsibilities will have a clear impact on de facto empowerment, although implementation issues can complicate the transition from theory to practice. Buying school material * * * * Budget oversight229 * * * Budget allocation230 ** Establishingschool fees 4 229Budget oversight refers to the fact that school councils keep track of transfers and expenditures by checking account statements. 230Budget allocation refers to the fact that it is at the school council's discretion how the transferred funds are used, i.e. which amount i s spent on salaries and which on school materials. 173 d) Financial Arrangementsand Teachers'Benefitsand Status This sub-section is concerned with the nature of the financial transfers made to the school councils and with the salaries, other benefits and contractual arrangements o f teachers. Composition and frequency of financial transfers. As highlighted in Table 5.4, formulas to calculate transfer amounts in El Salvador, Guatemala and Honduras are similar in that they all include a predetermined fixed sum for teachers' salaries and benefits. There i s also a fixed sum for school maintenance and supplies, except in Guatemala where funds for school and teacher materials depend on the number o f teachers and students. This i s not the case in the Nicaraguan SBM model where transfer amounts depend on school-specific factors such as student enrollment and drop-out rates and the location of the school rather than on predetermined sums. The total school transfer in Nicaragua i s obtained by multiplying a cost per student by the number of students attending school (calculated by resting drop-outs from initial enrollment). We will see that the incentives inherent in these different types of formulae are very different, making the formula applied in Nicaragua promising from an enrollment and efficiency perspective. The formula has evolved in time, becoming more equitable and efficiency-enhancing in 2002. In addition to the financial transfers, Nicaraguan schools are also allowed to collect school fees to supplement their state allocation. Table 5.4: FinancialTransfers Aspect of transfer ElSalvador Guatemala Honduras Nicaragua Formulausedto The amount The amount The amount Since 2002: a calculatetransfer includes: (a) includes: (a) includes: (a) per-student teacher salary teacher salary teacher salary amount i s (USD418/month), (USD (USD215/month calculated, (b) additional funds 240/month), (b) perprimary school addin up salary for benefits (rural additional teacher andUSD and bonus of USD funds to cover 72/month per pre- operationand 40/monthper benefits, (c) school teacher), (b) maintenance teacher, 14percent USD28 per contributions for costs per student of each teacher' s teacher/yr to social security (11 and an extra salary for social buy teaching percent of each amount for security), and (c) supplies, (d) teacher's salary), smaller andrural USD35/year for USD7 per (c) provisions for schools. This school maintenance studentlyr for school materials amount per and supplies. school (USD 100/yr.) and studenti s materials, and school maintenance multiplied by (e) USD 12per (USD65/yr). students child/year for attendingschool school (enrollees minus breakfast. drop-outs). 231Obtained by dividing the average teacher salary by a standard pupil-teacher ratio. 174 Frequency of Monthly Quarterly Monthly transfers I Transfer Initially, EDUCO's The Ministry of The Ministry of The Ministry of arrangements coordinating unit Public Finance Financetransfers Financetransfers depositedfunds transfers funds funds to funds directly into local bank into PROHECO's into local bank account registered PRONADE's coordinating unit accounts opened by the ACE. trust fund which then makes up by the Followinga series account which transfers to AECO consejos. of organizational then deposits bank accounts. rearrangements,it paymentsinto i s now the Ministry local accounts of Finance, through openedup by i t s departmental COEDUCAs. offices, who depositsthe funds to the ACES. Financial transfers to the school councils are made on a monthly (El Salvador, Nicaragua) or quarterly basis (Guatemala, Honduras). Transfer arrangements differ as to whether the transfer i s made directly from the Ministry of Finance to the school councils' bank account (El Salvador, Nicaragua) or whether the funds pass through the program's central office (Guatemala, Honduras). Teacher salaries, benefits and contract tenure. Comparing teacher salaries, benefits and contract tenure across countries as well as between SBM and traditional schools i s important because it sheds light on the imbedded incentive structures that, among other factors, influences whether it i s attractive for teachers to join and remain in their profession as well as their day-to- day work as teachers. Who teaches and teacher performance, in turn, i s directly related to educational achievement. Comparing teacher salaries and benefits across SBM and traditional schools, we see that, for a similar levels of qualifications, in El Salvador and Nicaragua salaries are identical across the two systems (the possibility of raising fees would in fact allow salaries to be higher in the autonomous schools of Nicaragua), while in Guatemala and, particularly, Honduras, salaries are higher in the traditional system. All other monetary benefits are generally higher in the traditional system (where, in particular, there are more fringe benefits), with the exception of Nicaragua. Lower average teachers' benefits in the SBM schools will contribute to explain why teacher education i s lower in these schools. Finally, contract tenure i s fixed-term, yearly renewable, for SBM models, which constitutes one of the most salient features of the models, designed to create an incentive to perform well. 175 Table 5.5: Teacher Compensation and Tenure in SBM vs. Traditional Schools Teacher Teacher Benefitsin Benefitsin Contract Contract salary in salary in SBMschools traditional ;enurein tenure in SBM traditional schools 3BM traditional USD 5016/yr 5016/yr Teachers Rural bonus, h e year 3pen ended working in social security .enewable rural areas benefits, other receive a fringe benefits monthly "rural (medical and bonus" of 40 life insurance, USD.As maternity social security leave). benefit, they also receive 14 per cent of their yearly salary. USD 2880/yr USD3084/yr Teachers Teachers 3ne year Open ended receive an receive an renewable annual bonus annual bonus of USD 720, of USD 514, paid in three paid in two installments, installments, plusUSD 157 social security as a benefits and contribution to other fringe a pension plan benefits. (no full social security). USD243l/yr USD 3018/yr Teachers Two annual One year Open ended receive two bonuses, in renewable annual June and bonuses, in December. June and Teachers also December. receive social Teachers also security and a receive social pension, plus security. other fringe benefits. USD 1350/yr USD 1350/yr As a bonus, As a bonus, One year Open ended (primary (primary teachers teachers renewable school school receive one receive one teachers), teacher), USD additional additional USD 1488/yr 1488/yr salary per year. salary per (secondary (secondary They also have year. They school school social security also have teachers) plus teacher). benefits, in social security funds part paid for benefits, in potentially by MINED part paid for raised by and in part by by MINED parents. the teachers. and in part by the teachers. I76 111. Impact of DecentralizationReforms The purpose of this sub-section i s to review and to systematize the existing analytical evidence on the progress and impact of the school-autonomy models in Central America. The evidence relies mostly on secondary sources and, for Honduras, Guatemala and Nicaragua, recently collected data on autonomous schools. In all countries the information was complemented by interviews with key actors on specific aspects of the impact of the models and by additional secondary data covering aspects such as coverage, internal efficiency, etc. Before presenting the results, we should made clear at the outset a few caveats. Firstly, as already mentioned, community-based school management programs implemented in Central America have been aimed at increasing enrollment, community participation, efficiency and, very marginally, on improving the quality o f education through more parental and local participation. Raising quality has not necessarily been at the center of these initiatives, and the quality improvement objective has instead been considered as an indirect outcome or simply been added on as the programs have matured. Secondly, three factors complicate the use o f outcome measures (learning, years of schooling, etc.) to evaluate the impact of education decentralization: (a) time series o f these measures are seldom available; (b) these school outcomes usually change slowly in response to any kind of educational intervention, including decentralization; and (c) it i s very difficult to control for external shocks ranging from natural disasters and fiscal crises to teacher strikes and changes in national education leadership-that may also influence school outcomes. Third1 selection bias i s another major concern when assessing the effectiveness of decentralization?" If the kinds of communities who choose to participate in the school based management program are made up of especially motivated individuals then evidence o f marginally higher test scores or more frequent student attendance may be incorrectly attributed to the decentralization experience itself. Also, differing rates of student attrition between decentralized and non-decentralized school types may make comparisons of outcomes for enrolled students very difficult to interpret. We are well aware of the inherent difficulties o f decentralization research in developing countries, and we compensate for the lack of perfect data by bringingin multiple sources o f information that include secondary evidence from other studies as well as recently collected data from three o f the participating countries. Following our conceptual framework, our approach to understanding the effects of decentralization in Central American schools focuses on two kinds of variables. The first group, which we refer to as the "first order" or direct impacts of decentralization, includes improvements in things like community and school empowerment, the provision o f supporting inputs and teaching and learning processes. B y empowering communities and making teachers and directors more accountable to parent councils the SBM model predicts more effective schooling environments. These are the most immediate goals of the S B M approach. But the story does not end here, since it i s hoped that first order effects like improved teacher attendance will translate into second order improvements in things like student learning. The first and second order distinction i s necessary from an evaluation standpoint, and it i s important to first establish an impact of decentralization in things like learning and teaching environments in order to explain any observed differences in student outcomes. 232Jimenez and Sawada (1999). 177 This i s not necessarily a requirement for establishing that decentralization works, since differences insecond order outcomes may result from things that are not measured well inthe data or, for non learning outcomes such as coverage and equity, from factors related to reform design which do not affect first order results. But a more causal argument in support of decentralization is built if the evidence shows that community participation or autonomous schools are more effective in areas that, in turn, are related to outcomes like test scores. This point has been somewhat neglected in the decentralization literature, at least in empirical analyses where the focus i s almost always on student outcomes such as test scores or drop out rates and less attention has been paid to understanding first order differences in school environments. A. COMPARING SCHOOL ENVIRONMENTS AND WITHOUT DECENTRALIZEDEDUCATIONAL WITH PROVISION a) School Autonomy and Community Empowerment Community Empowerment Mixed evidence, although community empowerment is generally greater in S B M schools. There i s some encouraging evidence that communities have been empowered in Central American SBM models. In El Salvador, parents of EDUCO students participate more in school affairs. ACE members (parent councils) in EDUCO schools visit classrooms almost 4-5 times more often than their traditional counterparts. At the parental association meetings, 80% and 79% o f ACESdiscussed teacher discipline and attendance of school personnel, respectively, while corresponding figures of schools in the traditional system are 62 percent and 38 percent, respectively.233 EDUCO parents are also three times more likely to engage in day to-day classroom activities than parents in traditional schools.234 Another benefit o f EDUCO, cited by Lindo (2001), has been the growth of community social capital in communities with EDUCO. B y empowering parents, i t i s expected that other improvements will result in areas such as democratic participation and state relations. There i s also positive evidence of community empowerment in Guatemala where each community i s represented by the locally-elected school council, COEDUCA . A positive spillover effect of the program has been the community's more frequent and effective participation in other local and civic affairs.235This has also been confirmed by other studies such as Asturias et a1 (2001) that found that COEDUCAs have used the knowledge acquired in training to lobby for projects to develop their communities such as roads, schools, electricity, running water, credits and others. A prior study by CIEN (1999) also found that COEDUCAs are a key means of increasing social capital. 233 Sawada (2003). 234 Jimenez and Sawada (1999). 235 Valerio and Rojas (2004). 178 Table 5.6: Parental Involvement in Community Schoolsversus "Regular" Schools El Salvador Guatemala Nicaragua Honduras Control EDUCO Control PRONADE Non- Autono Auto m o w Public PROHECO Parent-Teacher Meetings (Parents) 4.8 5.0* 3.43 3.44 Parent-Teacher Meetings (Teacher) 4.2 4.7* Parent-Teacher Meetings (Director) 1.56 5.65* 5.41 4.92* 2.62 1.71* Director-Parent Meetings 4.5 4.8* Meetings Coverage 0.55 0.58* Parental Participation 1.65 1.73 jource: Nicaraauadata 2~02;UMCE 2003; PRONA idata 2002; Sawadaand Ragatz, 2004.1 res: Data are not comparableacross countriesdue tomeasurementdifferences. In all countries except Guatemalathe data are take rom third/fourthgrade questionnaires. For Parent-Teacher Meetings the numbers refer to frequency, whereas Meetings Coverage measures percentage of topics covered during meetings. Parental Participation (Hondurasonly) is the degree of parental participation in the school activities according to the Director. Asterisk indicatescomparisonis significantlydifferentat p0.01, 0.05 or 0.10 level. Table 5.6 presents more evidence on parental participation in the region. The results are not entirely positive. In Nicaragua and Honduras parents in SBM schools appear to meet less frequently with school personnel than parents in traditional schools while in El Salvador and Guatemala the opposite i s true. This inconsistency is somewhat surprising, and we cannot rule out measurementissues since parents at SBM school may not consider all of their trips to the school to be "visits." Butfrequency is only part of the story, and it couldbe arguedthat the more important indicator of parental involvement relates to what parents actually do at schools and the amount of control they have, rather than how frequently they come around. On this count there i s more evidence of an impact, as parents reported a (slightly) greater coverage of issues in Nicaraguan autonomous schools and PROHECO parents, according to Directors, have a greater degree of participation in school processes. However, neither difference is particularly large and, in the case of PROHECO, it is not significant (measurement error is likely to be particularly relevant in Honduras). In Nicaragua, a study on the concluded "the evidence that parents' participation in school affairs has improved with the reform is weaker. About one-half of parents and teachers think that it (their participation) has become stronger, but a larger proportion of directors and council members think that it has not changed, at best". While this is evidence of 1996, these findings are confirmed in more recent 236King and Ozler (1998). 237See King, Ozler and Rawlings (2001); and Florez, Ruiz et al. (2003). 179 School Decision-Making Autonomy Mixed evidence, although decision-making has generally increased. The next first order outcome to be considered i s decision-making autonomy. The SBM model predicts greater involvement by parents in day-to-day school operations, and the evidence (detailed above) i s generally supportive of this contention. The effects on autonomy are less predictable, since decision-making power may shift from directors and teachers to parents without necessarily increasing overall. For instance, the ability to fire teachers may shift from directors to parents under S B M but the locus of control versus the center may remain unchanged. This does not seem likely, however, since with more community participation and control o f schools, the overall decision-making autonomy should increase too, especially when parents have financial power. These factors, like those demonstrating parental participation, are also difficult to measure. On the one hand S B M stakeholders may not consider many o f their activities as evidence o f increased decision-making power and may under-report their real authority. On the other hand there may be significantly more dejure autonomy than defacto autonomy. The ability to hire and fire teachers i s an important mechanism linking community participation with decision-making autonomy in all of the SBM reforms inthe four Central American countries. Parents may feel less inclined to participate in administrative processes or enter classrooms to instigate procedural and/or curriculum changes. They may instead use their ability to replace teachers as a means for improving autonomy and making schools more responsive to communities. In all community school models parent councils have the power to hire and fire teachers. While certainly an important power, it i s less clear whether communities members are skilled at choosing high-quality teachers. Additionally, communities are often faced with only a very small pool of applicants considering the common attributes of SBM schools - lower salaries, remote locations and one-year contracts. These supply side constraints in turn make it hard to fire teachers because readily-available replacements are largely absent. It is also important to consider how well parents and community members can evaluate the work of their teachers. Parents may be limited to evaluating teachers on limited measures, such as absenteeism, and teachers may be resentful of beingjudged by non-professionals. The PRONADE data from Guatemala reveal that less than 2 percent of the parent councils in PRONADE schools report firing a teacher, which i s not significantly different from the non-PRONADE schools (PRONADE data). However, PRONADE schools report nearly three times as many instances of teachers quitting the school. This, again, may be due to the more attractive salaries, job security, and working conditions intraditional schools. The first evaluation of Nicaragua's autonomous schools238found that stakeholders in private schools report making 16out of the 25 decisions pertaining to school operations and management at the school level (as opposed to a higher level), compared with 10 for traditional state schools and 13 for autonomous schools. At the primary level, respondents in both types of autonomous schools report that the school makes 10 out of the 25 decisions, as compared with seven for traditional public schools. These differences among the types of schools are statistically significant. There i s some variation in the responses by school directors, teachers, and council members within each school type, but these differences are not significant; thus, there i s general 238King and Ozler (1998). 180 agreement among the members of the school community regarding the degree of autonomy their school possesses. The evidence suggests that the reform has indeed successfully expanded the role of the school in its governance. King and Ozler (1998) also found that autonomous schools make significantly more schooling decisions than do traditional schools, especially on personnel matters and in determining the school plan and budget. The Nicaraguan data also demonstrate that the degree o f decision-making actually exercised by autonomous schools varies greatly, and depends on i m p l e m e n t a t i ~ n . ~Overall, the evidence for Nicaragua shows that the reform has indeed ~ ~ successfully expanded the role of the school, although, in contrast to the other countries, we will see that this came more from an increasing role of the school director. In El Salvador, Sawada and Ragatz (2004) find that EDUCO parent councils have more power to hire and fire teachers and directors. Some additional data on autonomy are reported in Table 5.7. The variables are not comparable and questions about measurement and quality abound. So the table is used only to highlight some of the issues that need to be considered in the context of autonomy and implementation of the SBM model. For example, there is some evidence that teachers feel more "incentive-ized" in Nicaraguan autonomous schools, which i s a potentially important mechanism o f local control and autonomy if school councils and parents control the incentives. In Guatemala, the evidence from PRONADE is that parent councils have more control over aspects of schooling such as the work calendar and schedule, teacher supervision and even teaching methods (suggesting higher defacto than de jure autonomy). This last point i s particularly interesting, since the ability o f parents to enter classrooms, observe processes and instigate changes would appear to be limited in these contexts, especially rural Guatemala where many parents have never attended school. Finally, the results in Table 5.7 indicate that PROHECO teachers report less control over various aspects of their work, although the results are not very significant (discussed below). Unfortunately, no evidence was specifically collected on the degree of control o f the parents' councils but defacto control is likely to be higher at that level (recent evidence extracted from a comprehensive assessment confirms that parents feel that they have substantial control over teachers in PROHECO schools240). 239Parker (2004). 240See ESA (2004). 181 Table 5.7: School and Community Autonomy in Community Schoolsversus"Regular" Schools, Nicaragua, Honduras and Guatemala Nica igua Honduras temala Non- Auto PRONADE Variable Autono mous Sum of teacher-reported incentives 1.6 2.6* Degreeof school control over 0.25 0.38* promotions According to Parent Councils: Controlover Calendar 0.24 0.37* Controlover Work Hours 0.25 0.46* Control over Teaching Methods 0.18 0.27* Have Fired Teacher 0.012 0.015 According to Teachers: Controlover teaching 2.8 2.6 Controlover priorities 2.6 2.2* Control over decisions 2.7 2.7 Control over planning L 2.7 2.6 Source: Nicaraguadata, 2002; PRO 4DE Data2 Notes: Dataare not comparable across countries due to measurement differences. In all countries the data are taken from third/fourthgrade questionnaires. Asterisk means that comparison is significantlydifferent at ~ 0 . 0 1p<0.05 or , pcO.10 level. Lack of teacher autonomy. The evidence on PROHECO highlights an important point: teacher autonomy does not appear to increase in community schools. This has also been found in Nicaragua where King, Ozler and Rawlings (2001) find that teachers feel less empowered in Nicaraguan autonomous schools. Increasing autonomy in community schools may involve shifting responsibilities, which can result in empowerment for some and relative disenfranchisement for others. Hannaway (1991) argues that decentralization can diminish, rather than increase, the feeling of autonomy of local agents such as teachers because of the greater control that is wielded by parents and the local community. There i s some evidence on community control and autonomy from multivariate analysis. For example, Sawada and Ragatz (2004) present results for the marginal effect of EDUCO participation on perceptions of control among teachers, directors and parents. The results show that parents feel much more empowered, especially regarding teacher and director management, whereas directors generally feel less control over processes. We should highlight that the internal distribution of power i s somewhat different in Nicaragua, where the experience shows that school directors dominate school councils in many schools and there are concerns about the low participation level of community members and parents.241In the case of Nicaragua stakeholder participation appears to depend integrally on the leadership style of the school director242 and teacher participation appears to wane when financial incentives are smaller.243 241Castillo(1998);King and Ozler (1998);Fuller and Rivarola (1998). 242Florez, Ruizet al. (2003). 243Gershbergand Winkler (2000); Gershberg(2003). 182 How can the impacts of SBM on community participation and school autonomy be summarized ? With the available data only a few, tentative conclusions can be put forth. First, the fact that hundreds of these schools are operating in the region provides some evidence of community empowerment, at least as measuredby participation. There is also some evidence that parents are participating more in day-to-day operations in community schools, although the most certain mechanism appears to be through teacher management (hiring, paying, supervising, firing) and budgeting. What is less clear is the extent to which parents are actively involved in the teaching and learning environment of SBM schools. These are difficult aspects to measure, and in some cases (especially Guatemala) the parents are reporting more involvement in these facets of school life. Final conclusions about school autonomy levels are also difficult. Decentralization may, in some cases, have the effect of switching decision-making power from teachers and directors to parents, rather than increasing overall autonomy vis-a-vis the center. However, the fact that parents are participating more in teacher management in community schools is an important finding that may signal greater overall autonomy. b)Teachingand LearningEnvironments In this section we consider the relationship between decentralized control of schooling and teaching and learning environments. With the exception of certain observable and clear-cut school and home factors and teacher characteristics these are difficult variables to measure. It is also not entirely evident whether we would expect SBM schools to have better or worse teaching and learning environments compared with traditional schools. On the one hand the remote and impoverished settings of many SBM schools may lead us to believe they would have inferior teaching and learning environments. On the other hand, increased community participation and school accountability may lead these schools to have superior teaching and learning environments. Learning Materials and School Infrastructure The relative state of school resources and infrastructure in SBM schools compared with traditional schools varies across countries due to program design and national priorities. In El Salvador, EDUCO schools and traditional schools are equally likely to have sanitation facilities, but fewer EDUCO schools have access to electricity or piped water. There are no differences in access to textbooks between the two types of schools, but EDUCO classrooms tend to have fewer students and a larger library.244Despite this advantage EDUCO schools have fewer overall resources than traditional or private schools and teachers use fewer resources in the classroom to improve learning.245 The quasi-experimental study of PRONADE schools in Guatemala attests that PRONADE schools are more likely to lack water and sanitary facilities than a control group of schools.246Another study carried out by CIEN (1999) reports that people generally believe PRONADE schools to be worse off than traditional counterparts. 244Sawada (2003). 245MINEDLJES(2003). 246MINEDUCDPTecnologias(2002). 183 The situation i s different in Nicaraguawhere, largely due to program design, Autonomous Schools tend to have more resources and better infrastructure than traditional schools. Parker (2004) finds that Autonomous Schools are more likely to have curriculum standards and teacher guides and that school conditions and resources tend to be higher. This is also true for Honduras where PROHECO schools have more resources and better infrastructure, on average, than their traditional school counterparts. In terms of infrastructure, the program has benefited from access to external funds to improve infrastructure, and construction of PROHECO schools has been a priority of the Honduran Social Fund (FHIS). In terms of learning materials, PROHECO schools also tend to have more than traditional schools perhaps because school councils receive money directly to purchase materials.247 PROHECO schools are not altogether more advantaged then traditional schools however, traditional schools tend to have more schools services and larger libraries than PROHECO schools.248 TeacherEfSort If community schools have greater autonomy and parental participation it seems likely that they will do a better job of maximizing existing capacity, perhaps b y eliciting more effort from teachers.249 Teacher effort i s yet another elusive idea to measure but typically researchers use proxies such as teacher attendance, hours worked, and frequency o f homework assigned. Eliciting greater or improved teacher effort i s one of the key ways in which school-based management may improve student learning and development. Teacherattendance and work hours There is some evidence that community school teachers work more days and more hours than traditional school teachers, although this is more the case in El Salvador, and Guatemala than in Honduras or Nicaragua. In the case of El Salvador, Jimenez and Sawada (1999) argue that close community monitoring and greater job insecurity in EDUCO schools result in less teacher absenteeism and fewer school closings than in traditional schools. There is indication that school councils in EDUCO schools are more likely to discuss teacher attendance and disciplinary measures than parent associations in traditional schools.250 EDUCO teachers report that they are also more likely to visit students' homes if students are absent. This may also be at least inpart a behavior stimulated by greater accountability to parents inthese schools. 251 Table 5.8 summarizes the evidence from El Salvador, Honduras, Nicaragua and Guatemala. Similar measures of teacher attendance and work hours are considerably more mixed in Nicaragua. While autonomous school teachers and principals tend to report that teachers work more hours and are less absent than traditional school teachers these results are not reflected in parent or student surveys. 252 247Durston (1999). 248DiGropelloand Marshall(2004). 249Sawada 250Jimenez and Sawada (1999). . (2003); DiGropelloand Marshall(2004). 251 Ibid. 252Kingand Ozler (1998). 184 In Honduras, the PROHECO teachers report fewer hours of weekly work, although when decomposed into various categories the results show they spend the same amount o f time on teaching activities.253 PROHECO teachers appear to be absent more often, the difference emerging largely due to more PROHECO teachers receiving more training. Interestingly, students report that PROHECO teachers are less frequently absent than traditional school teachers. Finally, PROHECO schools are more insulated from labor problems as their teachers are not union members. During the last two years, PROHECO schools have remained open when others have been closed due to strikes. School directory also confirms that PROHECO schools are closed less freauentlv. Table 5.8: Teacher Attendance and Work Hours t Nicaragua El S; lador Hc duras Guatemala EDUCO Public PROHECO Public PRONAD Autono mous mous Work Hours 27.3 29.0* 33.7 36.8' 39.7 36.4 TeacherAbsences (a) 1.4 1.2* 13.9 17.4 0.15' TeacherAbsences (b) 0.34 0.35 1.94 1.54* DaysWorked in School 111.0 113.W School Closings 29.5 20.5* nrshall,20t 3,2003b; 5 wada and Rag; ,2004. Notes: (a)According to teachers or parents; (b) according to stui nts. Dataare not comparable across countries due to measurementdifferences. In all countriesthe data are takenfrom third/fourth grade questionnaires. Asterisk meansthat comparison is significantly differentat pcO.01, pc0.05 or pcO.10 level. Using the PRONADE evaluation data254we see that parent committees in PRONADE schools report fewer teacher absences, meanwhile Marshall's (2004a, 2004b) data from three Guatemalan departments show that PRONADE schools report significantly more days worked during the school year. Parents also report that they prefer PRONADE schools because of more effective days of class and because teachers are more responsible to their duties and resp~nsibilities.~~~ Finally, EDUCO teachers inEl Salvador report more weekly hours and fewer absences. Teaching Pedagogv In conclusion, while there is mixed evidence it appears that by and large SBM schools are less frequently closed and S B M teachers are less likely to be absent and may work more hours that traditional school teachers. These findings are more consistent in Guatemala and El Salvador and are somewhat less so in Honduras and Nicaragua. Few school factors are as difficult to measure for quantitative analysis as classroom pedagogy. Furthermore, it i s not clear that parent councils actively monitor teaching practices, especially given the targeting of school decentralization in poor areas where parents themselves have low 253DiGropello andMarshall (2004). 254MINEDUCDP tecnologias (2002). 255Asturias et al. (2001). 185 levels of schooling. Nevertheless, it should not be dismissed that SBM teachers may be more inclined to adapt their teaching methods to better suit local needs and contexts by using more dynamic and interactive pedagogies or assigning more homework for example. As shown in Chapter I1there i s evidence that more interactive pedagogies may enhance learning although this i s contingent on several factors such as teachers' mastery o f these methods. Data on teaching methods comes from teacher and students surveys, both of which have their drawbacks. Finally, intervening variables-like class size or the number of grades in the classroom-can also influence the use or success of particular methodologies. Little evidence that teaching methodologies are different. There i s no evidence that S B M schools tend to differ significantly in pedagogical methods in Nicaragua or Honduras256.The best data available on this subject may come from Marshall's (2004b) analysis of student learning and attendance in rural Guatemala. This study involved classroom observations of teaching methods rather than surveys and found that while there were differences between PRONADE and traditional schools, both types employed traditional teaching methods. PRONADE teachers tended to use direct instruction methods while traditional schools tended to use individual work methods. In terms of homework there is equally weak evidence. Nicaraguan SBM schools tend to review homework more often while in Guatemala, PRONADE students report that they receive less homework. Class Size and Grades in Classroom Generally smaller class size, but more grades taught. Two final components of teacher effort are less actual indicators of teacher supply of effort and more likely to be indicators o f the demands placed on teacher effort. These are class size and the number of grades taught in the classroom. Class size i s an important variable for evaluating community schools and probably an endogenous one. For example, parents may intentionally limit class size in order to increase efficiency, or good schools with good reputations may find themselves with large class sizes as more parents enroll their children. The number of grades is also subject to parental manipulation if they have concerns about the demands placed on the teacher. In both cases these variables are likely to have an intervening effect on implementation of SBM in the classroom. Nicaragua El Salvador Honduras Guatemala Variable Non- Autonomo Public EDUCO Public PROHECO Public PRONADE Auto us Studentsper teacher 25.9 30.6* 28.0 20.8* 37.6 29.7* 29.0 26.V Multi-gradeschool or grades per teacher 0.20 0.39* 0.65 0.73* 1.5 2.7* Total Enrollment 98.5 68.7* 84.1 78.4 256 Parker (2004) and Di Grope110 and Marshall (2004). 186 For EDUCO, class sizes in the community schools are about seven students per teacher smaller than a group o f control schools. The same pattern i s apparent in Honduran and Guatemalan community schools (see Table 5.9). The exception i s the Nicaraguan autonomous schools, which have larger class sizes than non-autonomous schools. Honduras, El Salvador, and Guatemala SBM schools all tend to serve more grades per classroom than control group schools. B y themselves these results may suggest that community schools are more concerned about class size, perhaps because they want to get the most out of the teacher's ability especially considering the multigrade context of the classroom. With a link between learning and class size we can put this (supposed) concern into an actual learning context. Another important issue concerns costs. If community schools are getting more effort out of teachers then the cost effectiveness i s increasing, but this may be counter-acted somewhat by having fewer students per teacher. And once again the relationship between class size and learning (or other outcomes perhaps) i s crucial because if class size really is not a conditioning variable linking teacher effort with student performance then parental efforts to maximize learning by limitingclass size may be misplaced. In sum, the evidence on teaching and learning environments is mixed. First, with regards to materials and physical conditions the evidence i s inconclusive, as some autonomous schools (such as Nicaragua) appear to be better equipped while in other areas (such as PRONADE schools in Guatemala) the opposite i s true. In terms of teacher effort the results are somewhat more conclusive, as autonomous and community-based schooling environments appear to maximize teacher effort to a greater degree (at least in terms of teachers' work hours). Once again it must be stated that these conclusions are tentative, and based mainly on simple t-tests using data that may be of low quality. Despite the somewhat inconsistent results it i s possible to begin to form a picture o f SBM implementation in the region, or at least begin to prepare the ground for a more demanding analytical framework. The evidence that schools get more out o f teachers, despite having generally lower qualifications (as we will see below) and being paid less (see sub-section 11.2), points to a more efficient schooling model. This relative efficiency i s attenuated, however, by the generally lower student-teacher ratio in these schools. In cost effectiveness terms we can of course only speculate, especially since we do not know how any of these measures (teacher effort, class size, etc.) translate into student outcomes like learning. Finally, given the numerous differences between the community school "treatments" in the region, we must once again highlight the role of local context, assets and implementation schemes in determining these outcomes. B. COMPARING ENROLLMENTS AND STUDENT OUTCOMES W I T H AND WITHOUT DECENTRALIZED EDUCATIONAL PROVISION a) Impact on Enrollments Ministries of Education have used three administrative models to rapidly increase primary school enrollment in rural and marginal urban areas, with local and multilateral financing. The first is the 187 "traditional" model o f opening schools. The two most difficult aspects of this model are the creation of a teacher position and school construction. A second model, supported by the World Bank, was the establishment of afternoon shifts in existing schools in marginal urban areas. This model has the same teacher position challenge but avoids the need to build additional schools. The third model has been that of School Based Management, this model has also been supported by the World Bank and other international agencies. SBM models have been successful in increasingenrollment. In El Salvador there has been a steady increase in enrollments attributable to EDUCO. In 1991 the program was administered by 263 ACEs with 263 teachers and 8,416 students (see Table 5.10). 10 years later, in 2001, the program included 1,970 ACEs and 322,432 students in preschool, primary and middle schools. This represents 50 percent of the public rural preschool enrollment and 37 percent of all students in grades 1-9 in rural El Salvador (38 percent of students from grades 1-6, and 25 percent of grades 7-9). This growth i s impressive and can certainly be related to the increased coverage in rural areas highlighted in Chapter I, it i s necessary to put it in the context of generally although growing enrollments across the region in rural areas. Source: MINED,2002, and DatabaseDireccidn de Evaluacidn The program has also contributed to improve equality o f access to educational service. As Figure 5.1 shows, the program developed more in the poorest departments o f the country (on the right hand side o f the graph). 188 Figure 5.1:Enrollment in EDUCO as a percentageof the public enrollment in the rural areas by departmentsrankedby poverty levels.2002 60 .-0 - 50 E 40 c --E C 30 : 20 w 10 0 Source: MINED,database 2002; UNDP, 2001. In Guatemala, PRONADE's main objective was to open educational opportunities for 50,000 out- of-school children per year until full primary school coverage was attained. The immediate goal was to ensure that at least 70 percent of primary school-age children in each one of the 22 departments received services by the year 2000 (Figure 5.2). Since most departments had already achieved this goal, priority was given to departments that had not yet reached the minimum coverage, which, coincidentally, are also the departments with the highest proportions of indigenous peoples. Today PRONADE provides services to children in virtually all regions of the country in 20 department^.^'^ In a relatively short period of time PRONADE moved from a small innovative pilot program in 19 rural communities to a nationwide program reaching almost 3,600 communities. By the beginning of the current school year (2004), student enrollment reached almost 400,000 at the primary level. As of December of 2002, PRONADE schools represented21 percent of primary school enrollment in rural areas and accounted for 14 percent of the total enrollment in primary education at the national PRONADE clearly had a significant role incontributing to the increaseof the enrollment rate inrural areas pointed out inChapter I. 251Valerio and Rojas (2004) 258MINEDUC (2004). 189 Figure5.2: Percentof PrimarySchool-agepopulation in school in 1995 andtarget for 2000 I Source: Valerio and Rojas(2004) Table 5.11: PRONADE: Accomplishments1996,2000-2004 ear Primar School Preschool Primary School Schools Department ISEs served Children served Teachers hired Actual 55,000 67,734 327 13 13 258,000 249,024 43,848 7,373 3,419 21 17 261,397 48,635 7,903 3,420 21 17 272,820 48,739 8,325 3416 21 17 328,448 57,564 10,001 4,161 21 17 004 382,588 62,651 11,688 4,554 21 17 Source: Informationfrom While not specifically targeted to rural communities, the autonomous schools in Nicaragua have also had an impact on coverage, in particular at the secondary level, where the enrollment rate increased substantially over the 1900s. In 1997, 73 percentof all secondary students and 37 percentof all primary students were under the autonomy program. In terms of teachers, 78 percentof all secondary teachers and 41 percentof primary teachers were under the autonomy program as well.259B y 2002, 37 percent (1,781) of all Nicaraguan primary and secondary schools were autonomous; autonomous schools served 63 percentof the students (501,OOO), and had 13,419 teachers.260According to the Participation Law approved in 2002 all schools in Nicaragua must be autonomous by the year 2006. 259Arcia and Belli (1999). 260MECD(2004). 190 In Honduras, where PROHECO targeted the poorest and most isolated rural communities, the program had 820 schools in 2000 with 5,396 students at the pre-basic level and 34,144 primary school students261.According to the Central Unit of PROHECO, in 2004, 87,310 students were enrolled at the pre-basic and primary level in PROHECO, representing about 11 percent of the total enrollment in rural areas. PROHECO appears to be a somewhat more targeted program than the other SBMreforms. b) Impact on Internal Efficiency (Student Flows) As just shown, the SBM models have helped the countries in the region to improve their enrollment rates. Yet, one of the vexing problems for policy makers and other stakeholders of the education systems i s how to keep children in school once they enroll. We have seen that repetition and dropouts are a problem in Central America, in particular in rural areas (Chapter I). There is some evidence that community schooling leads to better student flows suggesting that internal efficiency, in particular in rural areas, may have improved over these last few years, but it remains a problem. There is some evidence of improved student flows. Once again the EDUCO program has received the most attention among researchers. Jimenez and Sawada (2003) find that the EDUCO program contributes significantly to decisions to remain in school beyond grade three. Additionally, students in EDUCO schools are less likely to repeat grades than students in traditional schools (although the result was not statistically significant). There i s less evidence for the PRONADE program in Guatemala. Based on 1995 data, PRONADE's indicators of internal efficiency are neither above nor below those of traditional rural schools, although PRONADE indicators have gradually improved in time. It is of particular concern that, as in traditional schools, repetition in the first grade of PRONADE schools was as highas 34 percent in 1999. More recent data confirm that PRONADE's indicators are very similar to the average rural indicators. More positive evidence, at least on student retention, comes from Marshall's (2004a) data from rural Guatemala, which make it possible to conduct very precise comparisons between cohort grade completion and retention rates in a regional sample of rural schools in Guatemala. His results show that PRONADE schools have done a betterjob at keeping studentsin school. Furthermore, in an econometric analysis using a multinomial outcome (passing, failing and desertion) Marshall (2004a) finds that boys enrolled in PRONADE schools are significantly less likely to drop out of school. For Honduras, Di Gropello and Marshall (2004) find some evidence that community schools have lower rates of dropout and grade failure than traditional schools in rural Honduras in their multivariate analysis of school averagesfor various student flows. Finally, in Nicaragua, recent data of the MINED (see Table 5.12) indicate that autonomous schools have slightly lower promotion rates, except in sixth grade; lower repetition rates, except in fourth grade; and slightly higher drop out rates in all grades, except in sixth grade. Overall, however, the results are very similar between the two groups. Higher drop-outs may be the SE (2000). 191 consequence of the relatively little impact of the autonomous model on teacher attendance. The new allocation formula, which includes an adjustment for drop-outs, will hopefully help improve student flows. Autonomous schools Non-autonomous schools Rates 1 2 3 4 5 6 1 2 3 4 5 6 Promotion 65.95% 81.75% 80.52% 81.63% 82.26% 97.06% 66.44% 81.84% 81.31% 82.26% 82.82% 96.61% Repetition 17.76% 10.25% 10.57% 7.89% 6.28% 2.94% 20.02% 11.46% 11.86% 9.13% 6.78% 3.39% Dropout 15.43% 9.34% 9.99% 11.26% 12.49% 0.00% 12.76% 7.62% 7.61% 9.42% 11.35% 0.00% 192 . c) Impact on Student Achievement TheEvidence It is commonly believed that community schools lead to more student learning. But it should be restated that none of the SBM models studied here has as a main goal improving quality o f education. Despite the lack o f a consensus in education policy circles on the relationship between student learning and community schooling, there i s some evidence from Central America on this subject. Mixed results on learning, but the impact of different assets needs to be considered. Table 5.13 summarizes, in descriptive form, the evidence on learning in the region. For each country standardized test scores (mean 0, standard deviation of 1 except for El Salvador) are presented together with indicators of family socioeconomic status and parental education. The results show that in Guatemala the PRONADE students are scoring significantly lower than their public school counterparts, while in El Salvador and Nicaragua the differences are insignificant. In Honduras, the PROHECOstudents have higher averages in science. Table 5.13 also addresses the issue of community assets (which we will develop further below). The pattern here is more consistent, as in all countries except Nicaragua the community schools are located in communities with lower levels of parental education and SES, usually measured b y the quantity of services or items in the home. These data provide an empirical justification for an important point that has only been referred to until now: community school initiatives in Central America are being targeted in areas that have fewer resources. The issue of community assets i s an important one, since implementation o f any decentralization initiative requires local capacity to be fully effective. The data in Table 13 demonstrate the need for multivariate analysis in order to asses the true effect of SBM on learning. In the next section on explaining impacts we will undertake our own analysis using data from Honduras, Guatemala and Nicaragua. But there are other studies-many using these same data-that also need to be summarized first and reveal a somewhat ambiguous pattern. Table 5.13: Third/J?ourth Grade Test Score Averages By School Type Guatemala Hon iras El Sa Nicara ua Control PROHE Control EDUCO Control AUTO Model: Control PRON Control PRONA co 1 I A DE DE Spanish -0.33 -0.39 0.02 -0.16* -0.04 0.03 1.75 Math -0.27 -0.38" 0.01 -0.07 -0.04 0.03 3.74 3.59 -0.01 Science -0.12 0.08* SES 2.5 2.2* 0.31 -0.46* 0.49 -0.42* 0.66 0.28 4.1 4.5* Parental Education 4.2 3.0* 2.5 1.7* 3.2 2.5* 0.53 0.50 5.5 6.2' Source: Nicaragua data 2002; UMCE2003; World Bank 2002; Marshall, 2004b; Sawada and Ragatz, 2004. Notes: For El Salvador average SES refers to percentage of homes with electricity, while Average Parental Education is the percentage of motherswith basic education. Asterisk meansthat comparison is significantlydifferent at p0.01, 0.05 or 0.10 level. 193 Existing studies indicate SBM schools support improved learning in El Salvador and Honduras and, to some extent, Guatemala and Nicaragua. In an evaluation of El Salvador's EDUCO program, Jimenez and Sawada (1999) compare teacher absenteeism and student achievement in EDUCO schools with that of traditional schools, controlling for student characteristics and selection bias (since the EDUCO schools were not randomly selected). They find that despite the fact that EDUCO schools have fewer days of teacher absenteeism than traditional schools, the student achievement in EDUCO schools was no different from that of traditional schools. The EDUCO analysis has continued with Sawada and Ragatz's (2004) recent paper that attempts to link community school implementation with student outcomes. They make the case that teacher behavior has changed for the better in EDUCO schools. The challenge lies in linkingthese changes with improvements in student outcomes, and the resulting statistical analysis provides a link-if somewhat tenuous-between improvements that result from EDUCO participation and improved student outcomes like test scores. The already mentioned evaluation of Nicaragua's autonomous schools by King and Ozler (1998), investigating the impact of two types of autonomy (de jure, based on whether the public school officially had a contract with the Ministry of Education to transform the school council to a consejo divectivo; and de facto, based on the proportion of key decisions made by the school council rather than central or local government), find that de jure autonomous schools do not appear to have an impact on student achievement. The authors attribute this lack of test score improvement to a possible lag time inbecoming truly autonomous. Interestingly, they find that de facto autonomy - ie nun-autonomous schools that practice decision-making power - is associated with higher student achievement. B y disaggregating the defacto autonomy variable into two types of decision-making areas (decisions that are related to teachers, including such decisions as hiring and firing, evaluation, supervision, training, and relations with the teachers union; and decisions related to instruction, including decisions such as class size, curriculum, textbooks, educational plans, and the school hours and calendar), they find that the variable on pedagogy gives mixed results; while the effect of the variable on teacher-related issues i s positive for both education levels and subjects, and i s statistically significant throughout, except for math scores at the secondary level. Recalling Levin (1995), these results suggest that the schools that are more active in tracking and monitoring teacher activity and in controlling staffing issues are likely to be more successful in increasing student achievement. Parker (2004) also addresses the autonomous reform in Nicaragua and finds that autonomous schools, on average, have higher Spanish scores than centralized schools (only significant in the third grade). Unlike Spanish, student math scores are not predicted by the factors associated with school autonomy, namely teacher incentives, material resources, and professional development. Her overall conclusion i s that the study has not shown clear differences in outcomes between centralized and autonomous schools (school type is only a significant predictor in one example, third grade classroom mean Spanish scores). An initial hypothesis of the paper was that the effects of decentralization contribute to changes in teacher behavior, with resultant changes in student outcomes, but the study has not shown a clear relationship between these three262. Supporting the findings of Parker, Gershberg and Kaestner (2002) show some skepticism at the results of the King and others papers because of the non-random participation of schools in the reform program which, according to these two authors, was addressedusing statistical procedures demonstrably not effective. 194 Despite PRONADE's expansion by 1994, very little i s known about its real impact on the ground or on student learning achievement. In 1997, MINEDUC launched the National Assessment System (PRONERE) to monitor students' achievement in mathematics and Spanish. Although a formal evaluation of student achievement in PRONADE schools has not taken place, the PRONERE'S tests carried out in 1999 and 2001 included some PRONADE schools in its sample of students.263Although average scores for PRONADE students are not among the highest, the results are encouraging given the lower base from which many of the PRONADE students begin, and the relative disadvantagedposition of the communities from where the students come.264 A World Bank study on student achievement in mathematics and reading was conducted in Guatemala in 2002. The study combined the test instruments developed by the UNESCO's Regional Office in Latin American and the Caribbean (OREALC) in 1997for a comparative study in Latin America, with items developed by the Universidad del Valle for the MINEDUC in Guatemala. Without controlling for the characteristics of students, teachers or schools, the results show that PRONADE schools get some of the worst average scores, even within rural schools.265 However, controlling for these background factors these negative findings disappear. Infact, they found that in comparison with urban public schools and regular rural schools, PRONADE helped studentsachieve higher scores in Spanish. Finally, Di Gropello and Marshall (2004) assess the PROHECO impact on student learning in Honduras using data from the UMCE's 2002 and 2003 data collections. As with the Sawada and Ragatz paper (2004) they make a link between improvements in teacher behavior (measured by work hours, frequency of homework, school closings, etc) and student learning, and the results are generally robust to corrections for selection bias. IV. Explainingthe Impact As was shown in the previous sections, the SBM reform has had an important impact in several areas of the education system in the Central America sub-region. The impact, however, has not been the same in all of the indicators and shows similarities and differences across the four countries. The degree of the impact of SBM on different indicators and the similarities and differences across the countries can be explained to a large extent by the following factors: 1) the assets of the actors and communities involved; 2) the provisions that are made in the reform design; and 3) the context of implementation such as socio-political factors operating at both national and local levels. All these factors mutually influence each other. But for analytical purposes, the discussion will focus on one at a time. As discussed in section I,assets include multiple resources such as skills and informational, organizational, human, material and financial assets; reform design has to do with both the objectives and the de jure provisions regarding the operation of the program; and the context of implementation consists of a series of sociopolitical factors that determine whether the reform, once adopted, can be sustained, and whether it is implemented according to the dejure provisions. Accordingly, in what follows, an effort is made to explain how and to what extent the 263 *@de Baesa(2001). Valerio and Rojas (2004). 265The World Bank (2004a). 195 differences in availability of assets, reform design and contexts of implementation help explain the progress and impact of the SBM models in terms o f school and community empowerment, teaching and learning environments, and final outcomes such as student achievement within and across the four countries. A. EXPLAINING IMPACTS AND PROGRESS IN TERMS OF ASSETS AVAILABLE AND LINKING TEACHER EFFORT TO LEARNING OUTCOMES As was said before, the success of the SBM models depends to a large extent on the assets that the actors and communities involved already have and the ones they receive according to the provisions of the reform design. a) First OrderEffects Community empowerment is high in spite of poor assets. For the SBM models to work properly, the school or community councils should be able to carry out the functions that have been delegated to them. B y assuming decision-making powers that they did not have previously, or by increasing their influence in areas they already affect, the councils are empowered. The effectiveness o f this empowerment depends on the assets that the members of these councils have. Because the reforms target different types of communities in the various countries, there i s variation in the type of assets available in each case. InEl Salvador, Guatemala and Honduras, the programs targeted poor, rural areas that lack material, financial and educational assets. In the case of Guatemala and Honduras very poor and isolated rural areas that did not have a school were targeted, which means there was little pre-existing knowledge of how schools work in these communities. In the case of El Salvador EDUCO was originally a grassroots experience, so the sense of ownership and motivation might be stronger than in Honduras and Guatemala. The Nicaraguan ASP model seems to be the better placed in terms o f assets since the initial targets were larger secondary schools located in relatively well-to-do communities that had more financial resources. The schools were already operating and many parents were participating, so they had some previous experience and knowledge o f how a school functions. Finally, since joining the program was voluntary, all the actors involved had the right motivation to participate.266Summing up, becauseof the type of communities they target, the SBM models of ElSalvador, Honduras and Guatemala look similar in terms o f community assets, although, across these countries, the communities targeted in Guatemala and Honduras have fewer resources. Higher levels o f SES are found in the ASP of Nicaragua. Across the region, thousands of SBM schools are functioning, attesting to at least some degree of community empowerment. Parents are making decisions on matters related to school management (hiring, supervising and paying) and budgeting that they did not affect before, more so in El Salvador, Guatemala, and Honduras, and less so in Nicaragua where parents seem to be less empowered. This suggests that even without previous experience in managing schools, poor rural 266Butthis has changed since the introduction of the Education Participation law of 2002. Beginning in 2006 all the schools in the Nicaraguaeducation system must be participatory schools. 196 communities have certain skill-sets and motivation to manage their schools and that the reform was able to harness and develop local capacity despite the lack of parental education, resources and experience in management. In Nicaragua, reform design was somewhat more focused on enhancing school autonomy through the role of the school director rather than on providing all the necessary support to communities. The SBM models in Central America demonstrate that, with the right type o f assistance, even very poor and undereducated parents are able to run schools. Poor community assets frequently mean poor school conditions. How does the physical and material conditions o f the schools relate to the assets already possessed by the communities and other key actors in the schools? As was pointed out before, if community schools are concentrated in the poorest and most rural parts of the country it seems likely they have fewer resources and learning materials. This i s certainly the case in Guatemala and Honduras. Communities have to meet certain requirements to be eligible such as not having another public school nearby. In Guatemala and Honduras, and even in El Salvador, schools begin to function in private houses or in huts built by the communities. So, SBM schools in these countries frequently begin with insufficient physical and material conditions, which are not always compensated by SBM programs. A different story holds for Nicaragua where students come from slightly better-off fami1ies. Poor teacher assets can be a constraint for innovative teaching methods. An important component of decentralization in the context of education i s to maximize teacher effort. The evidence reviewed above suggests that the schools in the SBM models are maximizing teacher effort to a greater degree. If parents have more say in how the school i s run, or more power to sanction poor performing teachers and directors, then it seems likely that school personnel will put forth more effort. Indeed, community school teachers work more days and hours. However, their teaching methods, overall, seem to be no different from the ones used in traditional schools. This may raise an issue o f reform design (lack of community participation in pedagogical matters and/or lack of teacher autonomy) which will be further explored. It may also arise due to a lack of teacher assets (skill, knowledge, experience, creativity). Teaching methods are likely to depend to a large extent on the assets already possessed by teachers. Teacher "assets" such as education and experience are considered important in educational delivery. In fact, much of the financial incentives that teachers get in the region are based on such assets. The skills and training that teachers bring with them to work are likely to help explain the degree to which the decentralization scheme i s implemented inside the classroom, just as parental education and community assets will affect the overall implementation. Table 5.14 reviews some basic teacher assets in the four countries. The results are, once again, mixed across countries in the region, although they are generally less favourable for autonomous schools. For Nicaragua the autonomous school teachers are more experienced but less likely to have a college degree. InEl Salvador the opposite holds, as EDUCO teachers are highly educated and comparatively in-experienced. For PROHECO, targeted in very small communities, the teachers are less likely to have a degree from the national teacher's university (UPNFM), are less experienced and in a majority o f cases did not attend a normal school. Finally, in rural Guatemala, the data from Marshall (2004b) show that PRONADE teachers are also less educated, and have less experience, than control school teachers. Overall, there does not seem to be a very clear cut relation between teachers' assets and teaching methods, although having a high 197 proportion of teachers with no teaching certificate (such as in Honduras) and little professional experience i s likely to make it more difficult to implement innovative teaching methods. Table 5.14: Teacher Education and Experience NicE .gua El SE ador Honl 1ras Guat nata Non- Autonom Public EDUCO Public PROHE Public PRONA Variable Auto ous co DE PercentTeachers 44.5 38.2* 30.0 74.0* 6.4 2.9* 2.5 o.o* with University PercentTeachers 3.9 54.8* without Certificate Teacher Experience 5.8 7.2* 9.5 3.9* 7.6 2.2* 9.7 4.6* Source: Nicaraguaaata 2002; UFntiE2003; Marsnall, 2004b; Sawada and Ragatz (2004). Notes: Asterisk means that comparison is significantly different at pc0.01; pe0.05, or pe0.10 level. b) Second Order Effects: Multivariate Analysis of Impact on Achievement The evidence presented in previous sections on first order effects resulting from SBM implementation and differences in assets (especially for teachers) helps to form a picture of the dynamics of community schooling in Central America. Inthis section we continue this discussion by replicating the analysis of Di Gropello and Marshall (2004). Using multivariate analysis that introduces groups o f variables one by one we can not only pinpoint the marginal effect of community school participation on student learning but also consider how other factors interact with this variable. This kind of analysis allows us to consider community and teacher assets together with other variables (like teacher effort) that we expect to be affected by implementation. We then can move on to larger, macro-type influences such as program design and contexts (next sections). What are the factors that condition the results in academic achievement of the students in SBM schools compared to their traditional counterparts? Following Di Gropello and Marshall, the coefficients presented in Table 5.15 represent the marginal effect (in standard deviations) of studying in a SBM school relative to the control categories. We then gradually add up groups of variables (community SES, school and teacher controls and teacher effort) to see how this marginal effect varies when controlling for these variable. In the first model student test scores are regressed onto student gender only (and the community school dummy), which provides the most basic comparison o f how community school students compare with their regular school counterparts. Not surprisingly, we see that in Guatemala the PRONADEschools start out lower when only accounting for these basic controls. 198 c Table 5.15: Effect of Community/Autonomous Schoolson Spanish and Mathematics, Summary of Regression Analyses, Guatemala, Honduras and Nicaragua Guatemala Honduras Nicaragua IModel: World Bank 2002 Spanish Math Science Spanish Math Spanish Math 1. Student gender -0.22 -0.24 0.08 0.15 0.29 0.01 -0.02 (-1.49) (-1.82)* (0.67) (1.16) (2.42)*** (0.07) (-0.26) 2.Student and community -0.04 -0.08 0.17 0.22 0.34 -0.02 0.01 SES (-0.35) (-0.75) (1.04) (1.50) (2.45) ** (-0.22) (0.05) 3. Plusschool and 0.14 0.07 0.23 0.44 0.44 -0.01 -0.01 teacher controls (1.11) (0.59) (1.12) (2.06)** (2.34)** (-0.10) (-0.16) 4. Plus teacher effort 0.16 0.13 0.11 0.37 0.16 -0.04 -0.01 (1.24) (1.04) (0.54) (1.65)* (0.84) (-0.36) (-0.16) Sample Size 3,103 I 3,103 975 I 952 974 2,596 2,575 Source: Nicaraguadata 2002; JMCE 2003;World Bank,2002. i4otes: Dependentvariable i standardi; zero and standarddeviationof 1.O. All parametersreferto dummy variable of community/autonomousE 1001,and each is interpreted in relationto "control" (Le. non community)schools. Variables are added in each model, and the sample size is the same for each estimation. For Guatemalathe controls also includethe incomingtest score inthe subjectfrom 2001. All results referto thirdor fourth grade students.t-ratiosare betweenbrackets.Asterisks indicatestatistically significant at 0.10r),0.05(**)and 0.01(***) levels. When controlling for community, teachers and school assets, the impact of community schools generally improves. In the second estimation household and community controls for SES, parental education, child labor and school attendanceare added. If community schools are serving different clients (Le. poorer) then we should expect the relative effectiveness to be greater (or less negative) once we take into account these differences; the results in Table 5.15 demonstrate this is the case in Guatemala and Honduras. In Guatemala, the entire difference between PRONADE and control disappears once we take into account differences in student background. In Honduras, the difference in favor of PROHECO schools tends to increase when controlling for the different socioeconomic status (not as much as expected because an effort had already been made to identify a control group close enough to the PROHECO As expected there is little and ambiguous change in Nicaragua, where the autonomous schools are slightly more affluent than the control group. In the third estimation the teacher and school controls are added. As shown above, community school teachers in these three countries (Honduras, Guatemala and Nicaragua) have lower educational qualifications and are also, with the exception of Nicaragua, less experienced. For the school controls like teaching materials and total enrollment the differences are less clear cut between SBM school and control in the region. As expected, the inclusion of these variables in Guatemala results in the PRONADE marginal effectiveness increasing vis-a-vis the control schools. In fact, although insignificant, PRONADE schools score higher than the control schools once these factors are added. In Honduras, the trend is also to increase the community school effect. Finally, in Nicaragua there is almost no movement in the coefficients when taking into account these variables, which is consistent with what has been detailed earlier since the 267See Di Gropello and Marshall (2004). 199 autonomous schools have similar, or even better assets in terms of teaching materials, compared with control schools. The link between teacher effort and academic achievement i s not consistent across countries. Indicators of teacher effort are added in estimation 4. These include measures o f days worked in the school and/or teacher absences and total work hours per week, school closings (in Honduras), homework frequency or frequency o f going over previous material (in Nicaragua), teaching methodology, and class size and number of grades. If community schools have succeeded in making changes in these kinds of outcomes-and they are positively associated with test scores- then we should expect the marginal effect of community schooling in the production functions to decrease. In other words, the positive coefficients for community schools in Guatemala and Honduras should get smaller; and the negative ones for Nicaragua get larger. However, the results inTable 5.15 fail to provide a uniform picture. There i s strong evidence in Honduras that higher teacher effort, associated with autonomous schools, i s also associated with higher test scores, but such evidence i s much weaker for Nicaragua (although the dummy coefficient in Spanish moves in the right direction) and, even more, Guatemala. This can cast some doubt on the degree to which community schools have made improvements in teacher effort characteristics that, in turn, translate into differences in student achievement in Guatemala and Nicaragua. From the previous evidence, there i s ground to think that teacher effort may not have improved very substantially in Nicaragua; while the results for Guatemala are more su risingsuggesting that increased teacher effort did not translate into higher scores in that country." However, before we rule out this link for these two countries altogether, we need to make an important point. The measure o f teacher effort adopted in the regressions is somewhat heterogeneous, capturing different aspects which may not have evolved in the same direction. For instance, we know that teaching methodology changed little in the countries under analysis; or, still, class size, an imperfect measure of teacher effort, may have unpredictable impacts on achievement (as appears with class size in Honduras, which i s quite strongly positively related to achievement, driving part of the results on the teacher effort variables). The various estimation results in Table 5.15 are not easily summarized. For Guatemala and Honduras the community school treatment clearly needs to take into account the very different environments that these schools operate in. This relates not just to student background but also teacher assets. In Nicaragua the targeting scheme was very different, and it is not the case that autonomous schools are serving poorer communities. Capacity is maximized and related to academic achievement in Honduras and El Salvador, capacity is also maximized in Guatemala but the relation with achievement is less clear cut. Overall, the evidence presented for Honduras, and the one that was presented for El Salvador b y Sawada and Ragatz (2004), suggest that autonomous schools have been successful in providing learning environments that are equal or even higher than their public school counterparts, although targeting poorer communities, by maximizing existing capacity (mainly by working more and limiting class size) to make up for the lack of teacher and parental capacity (in particular in Honduras). The link between capacity utilization differences and student outcomes i s encouraging, It is also true that different databaseshave been used for Guatemala, which can produce somewhat inconsistent results. 200 and provides prima facie support for community schooling on pure efficiency grounds. This is especially true in Honduras where PROHECO teachers are paid less and have fewer qualifications. The evidence provided for Guatemala, while less positive, still shows that, in spite of lower teacher assets, PRONADE schools manage to provide a learning environment as good as their public counterpart for students with similar characteristics, although the link between improved capacity utilization (which higher teacher effort also illustrates in Guatemala) and achievement cannot be directly made. Overall, across these three models, we seem to have a picture where SBM school parents concentrate their oversight to the most visible aspects of teacher effort-their attendance and hours worked-without entering classrooms and instigating changes in teaching methodology and classroom management, or choosing the best teachers. In other words, SBM parent councils may not be striving to create the best schools; they may simply want the most efficient. Lower levels of efficiency in Nicaragua. The picture is somewhat different in Nicaragua where autonomous schools perform as well as non-autonomous ones but with slightly more community and teacher assets, which suggests somewhat lower levels of efficiency (although class size is kept larger counter-acting this trend). B.EXPLAINING THE IMPACTINTERMS OFTHE REFORMDESIGN So far the analysis of the impact of SBM on student outcomes has focused on linking differences in implementation and assets with outcomes like student achievement. What remains is an analysis of additional "conditioning factors" that are likely to determine not only the success of the SBM in affecting first order changes in schooling but also the degree to which these differences translate into observable second order differences. These factors are left for last for the simple reason that they are, of all of the possible influences on schools and schooling outcomes in these countries, the most difficult ones to assess from an evaluation standpoint. Nevertheless, they are likely to provide some additional clues about SBMperformance. The first factor, that we will briefly review here, is reform design.269 Insection 11, it was shown that there are differences in reform design in aspects related to institutional arrangements, organization, membership, characteristics and functions of the school councils, financial transfers, and teacher salaries and benefits. Among the differences in the reform design that are most relevant for explaining the impact of the models on community and school empowerment, teaching and learning environments and final outcomes such as enrollment rates, internal efficiency and student learning the following can be mentioned: the objectives of the reform, its coverage, the composition of the councils, the range of responsibilities delegated to the school councils, the provisions for relevant parental capacity building, and the system of transfers and teachers' incentives. Objectives of the reforms. In terms of the reform objectives, as already noted, none of the SBM models considered in this study had community empowerment or quality enhancement as primary 269A more detailed analysis of the link between reform design and outcomes can be found in Rapalo and Marshall (2004). 201 objectives o f the SBM reforms. As such, any impact on empowerment and quality i s more a by- product of the reform than an explicit target. B y the objective of the reform design, the Nicaragua's SAP seems to have had the highest potential to contribute more to the empowerment of the school and communities, as school autonomy was a key instrument to increase social accountability and operational efficiency. However, for other aspects of reform design (see below), recent qualitative analysis of autonomous schools found that the reform has not reconstructed the social pact to increase arental participation in their children's education, but has had a "financial-administrative'' focus.?' In general, because of their lack of focus on quality, the SBM did not make adequate pedagogic provisions to improve school quality, as will be discussed below. Coverage of the programs. In terms of coverage of the program, we have already noted that some reforms are more targeted than others to poor areas, with consequences on community and teacher' assets and educational achievement. Another distinction concerns the extensiordcoverage of the models in terms of schools and students. While EDUCO, ASP, and PRONADE all cover a substantial fraction of schools, PROHECO i s much more focused, serving a relatively small number of schools. This may partly explain the better results of the model, which may be due to more efficient implementation, perhaps becauseof better supervision from program operators. Composition of school councils. Interms o f composition o f the councils, section 11makes it clear that all countries have granted a key role to parents in the management of the autonomous schools, which has effectively translated into higher levels of empowerment. However, interviews revealed that no program has been able to involve all the community members all the time271.Mothers of students and community leaders are most frequently involved in school councils. There i s little evidence to show that parents in the councils rotate as expected to give chance to other leaders elected in a democratic way. The case of Nicaragua i s different insofar as the director has a leading role in the councils, limiting somewhat, as we have seen, parental empowerment (and the 2002 participation law gave even more power to school principals making them executive directors and allowing them to runthe councils almost at their will). Responsibilities of school councils. All countries also granted extensive responsibilities in teacher management, school maintenance and budget oversight to the councils which resulted in higher defacto decision-making in these areas and, overall, higher teacher effort. However, little or no transfer o f pedagogical responsibilities to parents led to little or no decision-making in this area and unchanged teaching practices, which, in part, may have also been produced by little expertise or interest of parents themselves to instigate changes in the classrooms. In turn, no or little change in classroom processes hampered the impact o f all the models on learning, compensated by a positive impact of small class sizes, which can be to some extent associated with more individualized learning, and higher work hours in Honduras and El Salvador. Another important highlighted aspect of all models i s the decreased level o f teacher autonomy which essentially occurred by design by shifting the balance o f power towards parents: it is likely that this also explain the lack of pedagogical innovation of SBM models. ~~~ 270 Florez, Ruizet al. (2003); Gershberg,et al. (2001). This hasbeen confirmed in a recentassessmentof PROHECO(ESA, 2004). 202 Provision for parental capacity building. In El Salvador, Honduras and Guatemala, the SBM models have all made provisions for relevant parental capacity building and technical assistanceto compensate for the lack of assets and experience in running schools, with some success if we judge from the level of community empowerment achieved in all these countries. In fact, almost all the organizational structure underlying the management of these programs revolves around the issues of parental capacity building and assistance and the transfer of financial resources. In particular, one key element that has promoted community empowerment in Honduras and Guatemala has been the role played by the promoters and the ISE, respectively (although they provide largely administrative training with very little pedagogical training). In all countries, except inEl Salvador, the content and duration of the training for school councils is decided by the central ministry or coordinating unit. As part of a general policy, recently, the Ministry of Education of El Salvador has given the power to all schools to decide about their own training and to contract trainers directly. This arrangement, although apparently instrumental in providing more power to school councils, seems to be working against them in the EDUCO schools because little funding is given and the location of the schools provides little access to hire quality trainers.272 Despite the differences and the generally satisfactory results, in all countries there are some complaints about the content, relevance, duration and follow-up of the training. Again, interms of technical assistance in administrative matters, the role of the ISE in Guatemala and the promoters in Honduras seems to be particularly important, while EDUCO schools, although they can hire accountants, complain about the little help they get from the departmental offices. Both in El Salvador and Nicaragua, many actors think that the administrative process should be simpler and more transparent. Speaking about Nicaragua, section I1makes it clear that in this aspect as well the ASP program is the one that most differs from the others because no specific external facilitator/trainer is hired to help the councils. Because the schools are already running, the communities of autonomous and non-autonomous schools are similar and the principal of the school plays such a large role in the managementof the school, Nicaragua has not seen the need to create intermediary actors to provide training and technical assistance as has been done in the other countries. Transfers. Another important element of reform design is the design and implementation of the monetary transfers to schools (see section II) the provisions for budget generation. Here again and there is substantial difference between El Salvador, Honduras and Guatemala which allocate fixed amounts per teacher to the autonomous schools, plus a pre-determined amount for teaching materials, and Nicaragua which allocates funds on a per-student basis leaving the schools free to decide how to allocate them. Additionally, in Nicaragua, the schools had (initially) the right to charge fees and still have the possibility of undertaking fund raising. The evidence presented on the availability of teaching materials and equipment suggests that autonomous schools are decently provided with materials and equipment in Nicaragua, much less so in the other countries, suggesting limitations in the current allocation of funds in El Salvador, Honduras and Guatemala. Infact, ininterviews carried out for this study, schools council membersin El Salvador, Honduras and Guatemala argued that funding i s insufficient and that they are only able to buy the most basic materials. Moreover, there are legal restrictions as to what can be bought with the transfers. Given that, it is unlikely that the strategy applied in Nicaragua can work in the communities covered by the other countries, where the socio-economic characteristics of the parents and the size of the communities limit additional fund-raising activities by parents or other community members, *'*Nochez (2003). 203 requiring a somewhat different strategy for the allocation of resources, which should include a share o f non-salary recurrent costs proportional to the needs of the schools (see section on policy recommendations). Interestingly, transfers seem to be designed to produce stronger incentives for maintaining small class sizes in the case of El Salvador, Honduras and Guatemala, and incentives for increasing enrollment and class size in Nicaragua. In El Salvador, Honduras and Guatemala the amount transferred increases with the number of teachers hired and in Nicaragua transfers and user fees increase with enrollment. None of the transfer systems appear to create direct incentives to improve the quality of education, although, to some extent, small class size seems to favour individualized learning which has a positive impact on achievement; and, by subtracting last year drop-outs from this year initial enrollment, the formula applied in Nicaragua encourages internal efficiency and, indirectly, quality. Additionally, only in Nicaragua transfer formulae take into account equity concerns, although the new law, approved in 2002, does not have provisions to compensate small schools if they do not have enough students and fund raising activities will benefit more those schools that are located in well-to-do neighborhoods, creating equity problems. Equity adjustments are clearly less important in the other three countries where autonomous schools are all rural and poor, but may become necessary if the models are scaled-up to incorporate larger and wealthier schools (as i s already happening with EDUCO). Teachers' incentives: monetary benefits and their link with teachers' assets. Finally, a key element o f all the reforms i s also the role of teachers' incentives. Teachers' incentives will be an important determinant of teachers' assets and teacher effort. Table 5.5 of Section I1 makes it clear that, while salaries are similar across SBM and traditional teachers in Nicaragua and El Salvador, they are lower for SBM teachers in both Guatemala and, even more, Honduras. Other monetary benefits are also lower in autonomous schools in these countries. There seems to be a clear relation between teachers' assets as measured by education and experience and the lower monetary benefits in Honduras and Guatemala. In fact, these countries are already prone to have the least skilled teachers teaching in community schools because of the location of the schools and the existing teachers' incentives are doing nothing to counteract this tendency. Although Honduras may be particularly successful in maximizing the impact of its available community teachers, it seems clear that achievement results would be even better if PROHECO schools had higher- quality teachers. This i s also likely to be true for Guatemala. Additionally, we have seen that teachers' assets are also likely to be related to teaching methodologies which, in turn, are likely to be related to learning: by investing in more skilled teachers, the S B M models in Honduras and Guatemala may be able to increase class size moderately, possibly generating savings. Contract tenure and teacher effort. Finally, teachers' incentives will also affect teacher effort. Low salaries (or large delays in salary payment) may for instance have a negative impact on teacher motivation, although this does not appear to be the case in Honduras and Guatemala. More interestingly, fixed- instead of open-ended contract tenure, used in all four countries, may tend to motivate teachers to do better if contract renewal i s based upon fair and accurate evaluations of teacher performance. Additionally, in the cases o f Honduras and Guatemala, the reform design includes provisions penalizing teacher absences: in Guatemala teachers must compensate for days lost and in Honduras absences are deducted from salaries. In the case o f Nicaragua, it is less likely that teacher effort is affected by the reforms because, although teachers are also appointed 204 through fixed-term contracts, they continue to be protected by a general law that regulates the whole system (Ley de Curreru Ducente). On the other hand, user fees in Nicaragua may provide an incentive for teachers to work in autonomous schools. Teachers incentives and teacher permanency in the non-traditional system. Overall, judging from generally higher teacher effort, these mechanisms, together with the community empowerment that they confer, seem to be working. Two limitations of the models, however, are that: (a) none of the programs have established an adequate system of evaluating and monitoring teacher and school performance; and (b) the combination of lower monetary benefits and few professional development prospects could end up hampering the motivation of teachers who are doing well and would otherwise stay into the autonomous system .The first point will be further developed in the sub-section on contexts. On the second point, there is some evidence that, in Honduras and Guatemala, teachers move to traditional schools as soon as they get the opportunity. All SBM programs have put in place a system of teacher training, but this does not seem to be enough. Strategies to improve the medium term incentives for teachers to stay in the SBM system need to be found and applied. In fact, all SBM models (including Nicaragua, where a separate system for teachers in autonomous schools would be advisable) face the challenge of introducing some kind of professional development system, including creating a merit system (esculufun) for their teachers. Finally, we should also note that none of the SBM models has made provisions for special training for directors to develop leadership and managerial skills, which appears to be an important characteristic of effective schools. c.EXPLAINING THE IMPACT INTERMS OF CONTEXTSOFIMPLEMENTATION The discussion in the previous sub-sections shows that the impact and progress of the SBM models in the education systems of the region have been affected by the assets of the communities and actors involved and by the provisions that are made in the reform designs. In what follows an effort is made to show that the impact, progress and sustainability of the SBM models have been also affected by the contexts of reform implementation. Inthis section we will briefly review some of the socio-political factors which appear to be particularly relevant for the implementation of the SBM reforms: state-related variables such as the level of political commitment, state and social linkages such as the inclusion of different groups in the policy reform, traditions of local participation, institutional capacity and efficiency in reform administration, and availability of information. We should also point out that the success of the SBM models will also depend on what goes on in the rest of the education sector because of the linkages between the different education levels and public and private delivery (for instance, an expansion of the private sector can free teachers for the public sector makingmore teachers available to autonomous schools).273 Political commitment. In terms of state-related variables, there i s a consensus that successful implementation of social policy reforms depends on state commitment. In the case of the SBM reforms in the region, support was strong at the beginning and has continued steadily in El Salvador and Nicaragua, while it has shown some variability in the cases of Honduras and Guatemalawhere there have been changes in the ruling parties and inkey bureaucratic positions in the ministries of education and in the coordinating units. In El Salvador, the government has, in 273Illustratingthis point, a massive expansion of the tertiary system muchthrough the private sector inEl Salvador helpedEDUCOto hire lots of teachers. 205 effect, adopted EDUCO as its educational policy for rural areas. In Guatemala and Honduras the SBM programs were initially designed to improve enrollment rates in poor and isolated rural areas. In Honduras, the scope of the program has remained the same but in Guatemala the program was quickly taken to scale. In the Portillo administration, however, there was an open attempt to terminate the program and this slowed down its implementation. In general, turnover rates have affected the expansion and implementation of the programs but have not changed their main objectives and their "thrust" (even in Guatemala), indicating that local communities have developed a strong preference on behalf of safeguarding the reform (stake-ownership). Lately, however, governments in Nicaragua and Honduras are discussing the possibility of shifting the focus o f the decentralization reform from SBM models to municipalization. It i s also interesting to note that, in all cases, the World Bank and international donors in general have played a constructive role in combating the tendency of new administrations to scrap the programs of the previous administration. Because the new administrations realize that these programs are Bank- sponsored, they hesitate to dismantle them. Inclusion in the reform. Regarding state-society linkages, there is also evidence that including diverse groups from civil society (civic, business, university and religious leaders, notable personalities, opinion-makers, NGOs, consumer groups, teachers' unions, opposition parties) into policy discussions, both prior to the reforms and throughout the implementation process, generally helps implementation. In both aspects the SBM reforms have come up short. El Salvador seems to be the most inclusive case of all four. There were national dialogues at the start of the reform and repeated consultations during the implementation period. In Guatemala, there was a dialogue and consultation prior to the implementation of the reform but this has not continued during the implementation phase. In Honduras, PROHECO was created outside the MINED and was initially run by a commission of notables appointed directly by the Honduran president. There has been very little dialogue and consultation prior and duringthe implementation, and the same can be said inthe case of Nicaragua. In no country have the reforms included the teachers' union who are arguably the biggest cost- bearers of decentralization reforms, although some effort to do that has been attempted in El Salvador, perhaps explaining why relationships with the teachers' unions are more harmonious there than elsewhere. In general, teachers, who have not developed "stake-ownership", remain the main opposition and can potentially derail the reform, especially in Honduras and Guatemala where teacher unions are relatively strong. As we have seen, reform designs in these same two countries have also not been favourable to the creation o f incentives for teachers in the autonomous system, aggravating the situation and creating serious issues of sustainability in spite of the success of the models in terms of empowerment and coverage. Traditions of local participation. Another contextual aspect that may hinder the implementation of the decentralization reforms, especially in terms of community empowerment, may be traditions of local participation. Despite the apparent advantage o f the Nicaragua program in terms o f the type o f communities they target, the programs in El Salvador, Guatemala and Honduras have advantages in terms o f the tradition of social participation and the creation of social capital. Durston (1999) suggests that programs like PROHECO "have the advantage o f not having to concentrate efforts in fighting against institutional and traditional habits to establish the new process." This i s to say that since the relationship between parents and teachers i s brand new, there 206 i s a better possibility to establish relationships of mutual respect, accountability, trust and compromise with the educational outcomes. Moreover, according to this author, the rural communities have better contextual conditions for creating social capital and for making possible that the school councils may influence some other social processes in the communities.274In general, parents in the rural areas have more sense of community, more experience in social participation and more willingness to sacrifice time and money to participate in these types of programs. Parents of these communities might also feel more empowered because they know that this is probably the only way that their children may have access to education. In the case of Nicaragua, parents and community members know that whether or not they participate they have a school where they can send their children. Institutional capacity and efficiency of the administration: sustainability of institutional arrangements. The technical capacity and efficiency of state agencies both at the central and local levels are crucial to ensure the proper implementation of decentralization reforms. A first issue is the sustainability of institutional arrangements. In the SBM models discussed here, the state agencies provide schools or community councils with the necessary resources, such as money, training and information. State agencies function differently in the four cases. In El Salvador and Nicaragua the coordinating units are within the ministries and work with the school councils with the help of the local MINED offices (departmental or municipal) (although in El Salvador, councils also work with assistants that they hire themselves). In the cases of Honduras and Guatemala, the coordinating units belong to the Ministries of Education but work more independently. Their work with the councils is carried out almost exclusively with the help of intermediary agents that do not belong to the MINED, with little contact with the local units of the MINED.This i s why in these countries the programs are seen more as parallel programs than programs belonging to the MINED, with possible issues of sustainability. It has, however, been recognized that the intermediaries have been particularly useful in these two countries, suggesting that an hybrid type of solution (involving for instance both the deconcentrated units and fixed- term external agents) may be needed. Technical capacity of coordinating units. The composition and capacity of the coordinating units is important to ensure their proper functioning. In all four cases the personnel of the coordinating units are selected by the MINED. Most of the time the personnel are selected for their technical capacity, although politics and personal connections may also enter into the selection process, especially in Honduras and Guatemala. Judged by their performance, the coordinating units have done a relatively goodjob, especially inEl Salvador and Nicaragua. If we review some evidence of institutional strength in terms of financial transfers, provision of training and availability of information, we find, however, that: (a) there have been problems with the regularity of the transfers in some countries, which can indicate a difficult relationship between the ministries of education and finance but also low technical capacity of the coordinating units (this is particularly the case in Honduras and Guatemala); (b) the changes in government and personnel in the coordinating units have affected the provision of training in all the countries because they have changed the rules (community and schools actors in all the countries feel that the provision of training was stronger and better at the beginning of the programs); and (c) the 274Durston(2002). 207 models have not generated enough and relevant information to evaluate and monitor the performance of their schools.275 Availability of information and monitoring of teacher behavior. Related to this topic, another question that can be asked is whether the SBM models are providing school councils with the needed information to carry out their functions. In all countries the programs provide school councils with information regarding their functions, especially interms of managementof material and human resources. However, parents also need information on teacher performance to exercise effectively their task of hiring and firing teachers. In most cases, it appears that the council members do not have enough information to select the best candidates and technically evaluate them. A related point is that parents usually find it difficult to monitor curriculum and pedagogy. The only criteria used by parents seem to be teachers' attendance and good behavior. These are still important dimensions which ensure, as we have seen, higher teacher effort in terms of hours worked, but having a more comprehensive picture of teacher and school performance would be helpful to enhance the impact of the models on learning. None of the programs provide information about school performance and teacher performance; and although schools do produce some information on promotion, drop-out and repetition, they often send it to higher levels, without using it or sharing it with the community. If information on teacher performance is incomplete and it i s difficult to relate school performance to teacher performance, "monitorability" of aspects such as pedagogical practices will still be a problem and other mechanismswill need to be found to make teachers accountable on this ground such as: (a) setting-up effective supervision/inspection systems, involving directors and deconcentrated levels of the ministries of education in the direct observation of classroom practices; or (b) providing incentives for pedagogical improvement such as school competitions for the best school improvement project. V. Conclusionsand Policy Recommendations The school-based managementmodels considered in this study seem to be a potentially promising means to promote more civic engagement in education and to cost-effectively get better or similar educational results than traditional programs. Inparticular, community and schools appear to have been generally empowered and teacher effort strengthened, resultingin a better use of the existing limited capacity. This is more true for El Salvador, Honduras and Guatemala than for Nicaragua. Additionally, the models have had a very substantial impact on enrollment and are somewhat associated with better student flows. There is also evidence that academic achievement i s at least as high in autonomous schools as in traditional schools.276However, while some aspects related to school quality such as teacher/school effort have been prioritized in the models, some others such as availability of learning materials, teacher skills and pedagogical innovation have not been ~ '''For instance, coordinating units have not had the technical capacity to conduct studies to evaluate the performance of autonomous schools vis-&vis traditional schools, although they have received technical assistanceby international agencies and consultants. InEl Salvador, despite the proliferation of information, most of the information used to evaluate the program with a comparison group comes from 1996,in Nicaragua from 1997 and in Guatemala from 1999.I n the case of Guatemala, the quasi-experimental study was not completed and did not use standardized tests. In Honduras, becausethe program i s more recent, it i s only until this past year that a systematic study of its impact has been undertaken. Additionally, the available statistical information to monitor schools' characteristics and performance is often of poor quality and incomplete. 2'6 InGuatemala, only when taking into account the poorer communities in which these schools are located. 208 sufficiently developed, limiting the potential impact of the models on the quality of education and learning. Additionally, the sustainability of these models remains at stake in most of the countries. The evidence presented in this chapter, including the links that have been established with reform design, assets and the socio-political context, allows us to provide a set of recommendations for consolidating the good results of the models, improving their impact on the quality of the teaching and learning process and ensuring their sustainability. ReformPriorities. Empowerment and quality should be introduced as explicit objective of all the models; Pedagogicalmodel. At the outset, the way forward in community schools should be to provide an integrated package (called modelo pedugdgico propio) to give pedagogical strength to the model. The package should offer, in addition to community participation, bilingual education (ifneeded), multi-grade teaching, student assessment, teacher guides and in-service training, etc; Pedagogical innovation: teacher autonomy. There is international evidence that increased teacher autonomy in pedagogical practices and more skilled teachers can be conducive to higher pedagogical innovation and improved teaching-learning processes. The effective school analysis undertaken in Chapter I1 confirms this finding. Increased teacher autonomy in pedagogical practices could benefit all model designs. At a minimum, given the importance of teacher participation in improving the quality of education, it is necessary to give teachers a more meaningful role by establishing support committees that can tackle issues and allow for their voices to be heard. More skilledteachers should also be attracted through the implementation of a better incentive framework (see below). Teachers' excellence also depends to a large extent on on- going professional development opportunities. For Guatemala and Honduras it would be imperative to create special incentives (such as scholarships) for teachers who do not have a degree to get into a flexible program that allows them to be trained in pedagogical matters, while also working; Pedagogicaldecentralization to communities. Another option would consist in providing more pedagogical responsibility to the communities themselves to foster changes in teaching practices. There is less clear-cut international evidence that decentralizing this type of decisions to community would lead to better teaching practices. Additionally, by empowering parents in this area, teachers may feel even less valued. If it is decided to provide more power to the school councils in pedagogical matters, the models should proceed with some caution, gradually extending authority to communities as they acquire and demonstrate capacity in this area. They should establish mechanisms of accreditation before they grant this power and determine what kind of power they wish to hand over. There are some pedagogical matters such as the setting of policies regarding curricula, programs and evaluation which should remain, in any case, centralized; Transfer formulae. Transfer formulae should be modified, by basing the allocation of funds on the real demands that may come from the school councils themselves (through their School Education Plans), ensuring therefore higher shares for non-salary recurrent expenditure for materials and the like, and introducing allocation criteria that take into account equity, when 209 necessary, and performance (provision for small schools in Nicaragua, measures of student attendance or retention in Honduras, Guatemala and El Salvador); Parental support. Continuous parental support through flexible and transparent mechanisms needs to be provided. Although the use of intermediate agents, such as NGOs and social workers, needs to be more fully assessed in Guatemala and Honduras, there i s some positive evidence on their role in providing administrative support. In any case, these agents should receive training on a continuous basis, keep close ties with other local officials from the ministries of education, and use standard procedures and provide timely advice and support. It i s also imperative that in future strategic discussions regarding the programs' expansion and evolution, the longer-term role and status o f these external agents be addressed. It would also be important, in particular for the sustainability o f the models, to define with precision and strengthen the role o f the deconcentrated levels of the ministries of education (departmental and district offices). These levels could have particular relevance in pedagogical support and supervision to parents, teachers and directors. Additionally, it would be useful to assess the feasibility of having these levels hiring the external agents (in particular in Honduras and El Salvador), to contribute to the institutionalization and acceptance of the models within the ministry o f education, while not losing the economies o f scale that can be made by hiring external agents at an intermediate level (in contrast to the school); Informationon school and teacher performanceand supervision mechanisms. A substantial effort should be made to provide good information on the performance o f schools and teachers and make it available to all education actors (directors, teachers, parents, department and district education offices). The MINEDs could create a very simple model with the most important indicators (orientuciones) to measure teacher and school performance. Standardized testing should be used and disseminated. The model should be shared and validated with the parents and local community members and information should be produced in a systematic way to measure those indicators on a regular basis. A good information system will ultimately allow parents to obtain information on the performance o f their children relative to the rest of the school, the performance of the school relative to other schools of similar characteristics, and the performance o f the system against itself over time. Setting-up effective supervision mechanisms, involving directors and the deconcentrated levels of the MINEDs, would also be key to monitor teacher behavior in cases of imperfect information. Teachers' skills. Financial incentives are key to make the programs attractive vis-&vis the traditional system and to attract skilled teachers. Teachers' salaries and social security benefits should be balanced across traditional and non-traditional systems in Honduras and Guatemala. In all countries, it i s important to ensure that fringe benefits be more similar between the traditional and non-traditional sector (teachers in community based programs usually lack access to medical and life insurance benefits and female teachers do not have access to maternity benefits and leave). In all countries, there would be room for introducing performance bonuses in the non-traditional sector. The key here i s to be bold and rely on the local school council to determine the type and size o f the bonus to be distributed at the end o f each school year, and how to allocate the bonus among teachers. The idea i s that school councils should guard against linkingpermanent rewards (such as salary level) to performance, but rely instead on cash or in-kind bonuses that are used to 210 specifically recognize the work of good teachers, and openly express the appreciation of the community;277 Teacher motivation and permanency in the non-traditional system: timely payment. Due to administrative and financial difficulties, some of which are beyond the control of the coordinating units or MINEDs, there have been frequent delays in payments of teacher salaries, especially in Honduras and Guatemala. In order to ensure that teachers remain motivated and interested in working in autonomous schools, it is imperative that salaries be paid in a timely manner. To do this, having the school basedmanagementprograms' budgets integrated into the general budget of the ministry of educations may probably help. Secondly, in all countries, the forms and paper work to liquidate the funds should be simplified. Thirdly, explicit priority should be given to the payment of salaries within the transfer of funds. Fourth, to strengthen transparency, the programs should also contract independent firms to do the auditing of the transferred funds. Fifth, if the liquidation of funds has to be done every three months, it would seem more appropriate to pay teachers' salaries in advance; Professional development. While maintaining fixed-term contracts, which i s at the very core of the incentive structure of autonomous models, a merit system (esculufh), similar to the one applied in the private sector, should be designed and implemented for all countries in the non- traditional sector to retain teachers (including regular salary increases based on seniority and qualifications, increased responsibilities, training options, etc). It is also important to change the perception of instability which is part of the non-traditional system. Stability should not be understood in terms of time but in terms of performance. Teachers should feel confident that if they perform well they will keep their jobs. This can happen if they know what is expected from themand how their work will be evaluated. Directors: role. A comparison of Nicaragua with the other three countries suggests that directors should probably not be part of the school councils, or, if they are part, that their relative responsibilities be very clearly specified, to prevent communities from not being truly empowered in administrative matters (teacher and budget management). Taking advantage of the involvement of the communities in these matters, directors could take up a key role in academic matters in the schools, providing pedagogical support and supervision and ensuring a sound and constructive academic environment. Directors should probably be granted higher authority in larger schools (such as in Nicaragua), where monitoring school and teachers may be more complex. However, it will be necessary to ensure a democratic type of leadership which will require a more transparent way of electing directors and closer monitoring of their performance as well as special training and technical assistanceto develop their leadership skills. Training. Specific training should be introduced for directors to strengthen their management skills and leadershipin a variety of areas, in particular in academic matters. Institutional capacity of coordinating units. Peoplein charge of the coordinating units should be selected according to their merit and not for political or other reasons. Their performance should be evaluated not only according to their individual actions but also according to the overall results of the community program. Their capacity of evaluating and monitoring schools should be *"Also see Arcia and Belli (1999). 211 strengthened. Finally, to help institutionalize the programs, it should also be gradually envisaged to substitute the coordinating units by departments or offices that are fully part of the structure of the ministries of education; Consensus and institutionalization. As already discussed, consensus is a decisive factor in the successful expansion and institutionalization of any educational reform. Where it i s impossible to achieve absolute consensus, it i s important to make sure that the main political forces know the program's goals and its achievements. When this i s not possible, as was the case in Guatemala in the previous administration, it is important to make sure that other important stakeholders (such as the parents themselves, private sector, majors and others) are ready and organized to defend the programs. The S B M models should a have a network of key people ready to act. More studies are also needed to determine the impact of the models on student achievement, repetition and dropout rates, as well as teacher effectiveness. To institutionalize and create consensus around the programs it i s also necessary that they appear as part of the general policies and programs implemented by the government to improve the quality o f education. As such, they should be included in the legislation on education, in laws regarding participation, in the development plans and, especially, in the national budgets. They should be seen as part o f the routine o f the ministries of education without loosing their innovative character; Teacher inclusion and sustainability. Finally, it is key to implement strategies to include teachers and teacher's unions in the reform process, by disseminating information on the reform, negotiating more favorable employment conditions for teachers in the non-traditional sector (see above), or granting higher teacher pedagogical autonomy. It i s necessary to reiterate that teachers, who have not developed "stake-ownership", remain the main opposition and can potentially derail the reform, especially in Honduras and Guatemala where teacher unions are relatively strong. It is imperative for the same sustainability o f the models to address the issue of teacher opposition. Regional collaboration. A mechanism should be established that allows for community based programs to share their experience in the region. This can start with a summit of the ministries o f education (or vice-ministries) of the four countries to analyze the results o f this study. The purpose should be to develop a common agenda to solve the most urgent problems and to plan the long term sustainability of the models. 212 References Academy for EducationalDevelopment(AED) (2004): "Calidad y Equidad: Renovandola Visi6n de LargoPlazoen la Eduaci6nPreescolary Basicade El Salvador", USAID, unpublished document. AED and Jutirez and Associates (2003): "The NicaraguaBasicEducationProject, BASE11:2002 Annual Study", AED, unpublisheddocument Anderson, MH(2004): "Overview of the educationsectorin Guatemala" in The world Bank (2004a), Volume 11. Arcia, G (2003): "The Financingof PublicEducationinNicaragua", consultingreport for the World Bank. Arcia, G and Belli, H(1999): "Rebuilding the School Contract: SchoolAutonomy inNicaragua", LCSHD Paper Series No.40, The World Bank, Washington,DC Arias, 0 (2004): "Poverty inEl Salvadorduring the 1990s:Evolution and Characteristics", backgroundpaper for The World Bank PovertyAssessment of El Salvador. Arrien, Jed. (1998): "La Educacidn y la Reforma de la Educacidn en Cinco Paises Centroamericanos", PREAL, Managua, Nicaragua. Asturias, L. et a1 (2001): "Diseiio de un Estudio Etnografico piloto sobre la Influencia de PRONADEen el DesarrolloLocal", MINED, Guatemala. Bedi, A and Edwards,J (2002): "The impactof schoolquality on earningsand educational returns-evidencefrom a low-incomecountry", inJournal of Development Economics, Vol. 68 (2002) BCnabou, R, and Tirole, J (2003). "Intrinsic and Extrinsic Motivation." Review of Economic Studies70: 489-520. Benveniste,L, and McEwan, P(2000): "Constraints to ImplementingEducational Innovations: The Case of MultigradeSchools", InternationalReview of Education, 46(1/2) Bommier,A and Lambert, S (2000): "Education Demandand Age at SchoolEnrollmentin Tanzania", inJournal of Human Resources, 35(1):177-203. Bourguignon, F, FFerreira, and PLeite (2002): "Ex-Ante Evaluationof ConditionalCash Transfer Programs:The Case of Bolsa Escola", Working Paper N.516, William Davidson Institute, Michigan. CardD (2001): "Estimating the returnsto schooling: progresson some persistenteconometric problems", inEconometrica 69(5). 213 Carvajal, M,Morris, F and Davenport, L (1993): "Economic Determinants of Academic Failure and School Desertionin the Guatemalan Highlands", Economics of Education Review, 12(1). Castillo, M.(1998): "La Descentralizaci6n de 10s Servicios de Educaci6n en Nicaragua", Serie Reforma de Politica Publica, Naciones Unidas Comisi6n Econ6mica para AmCrica Latina y el Caribe, Santiago de Chile. Castro, C., Carnoy, M and L.Wolf (2000): "Secondary School and the Transition to Work in Latin American and the Caribbean", IADB, Washington DC. CIEN (1999): "La Experiencia del Programa Nacional de Autogesti6n Para el Desarrollo Educativo PRONADE", Resdmen Ejecutivo. Guatemala City, Guatemala. CIEN (2003): "Revisi6n del gasto en Educaci6n en Guatemala", Background Paper, Guatemala Costa Rica -Ministerio de Educaci6n Pdblicd Departamento de Pruebas Nacionales (2001): "Informes Nacionales sobre Resultados de las Pruebas Nacionales de la Educaci6n Formal, ciclo I1 y 111", San JosC, Costa Rica Costa Rica- Ministerio de Educaci6n Pdblica (MEP) (2002): "Equidad y calidad para 10s nifios y la niiias que asisten a las escuelas unidocentes", San Jose', Costa Rica. Costa Rica - Ministerio de Educaci6n Pdblica (MEP) (2001): "Educaci6n Integral el las Escuelas Pdblicas y Privadas de Costa Rica: UnFinN o Alcanzado", San JosC, Costa Rica. Cuellar Marchelli, H (2001): "Decentralization and Privatization o f Education in El Salvador: Assessing the Experience", Occasional Paper No. 19, National Center for the Study of Privatization inEducation, Columbia University, New York Cummings, S and Tamayo, S (1994): "Language and Education in Latin America: An Overview", Human Capital Working Paper N. 13068, The World Bank, Washington DC. De Baessa, Y. (2001): "Factores del Hogar en el Rendimiento Acadkmico", Unpublished document, Universidad del Valle, Guatemala. DiGropello, EandMarshall, J (2004): "Teachers' IncentivesinRural Honduras", The World Bank, mimeo DiGropello, E, Dubey, A and Winkler, D (2003): "Reaching Basic Educationfor All inLatin America", The World Bank, mimeo DiGropello, E(2005): "Education Decentralizationand Accountability RelationshipsinLatin America", Policy Research Working paper N..., The World Bank, Washington DC Dolton and Van Der Klaauw (1999): "The turnover of teachers: a competing risks explanation", Review of Economics and Statistics, 81(3). 214 Durston, J. (1999): "La Gestidn Comunitaria de la Educacidn Rural en PROHECO', PNUDKEPAL. Durston, J. (2002): "El capital social campesino en la gesti6n del desarrollo rural. Diadas, equipos, puentes y escaleras", CEPAL, Santiago de Chile. Dutcher, N.(1997): "The Use of First and SecondLanguagesinEducation: A Review of International Experience", The World Bank, Working PaperN. 18803, Washington DC. ECLAC (1994, 1996): "El Gasto Social en America Latina: unExamen Cuantitativo y Cualitativo", in Cuadernos de la CEPAL, Santiago, Chile Edge, K (2000): "El Salvador: EDUCO", Decentralization and School Based Management Resources Kit Case Studies,The World Bank, Washington DC Edwards, J (2004): "Education and Poverty in Guatemala", in The World Bank (2004a), Volume 11. Edwards, J, Fuller, B and Parandekar, S (1997): "Primary Education Efficiency inHonduras. What remainsto be done ?", LatinAmerica HumanDevelopment Working Paper Series N.7, The World Bank, Washington DC. ESA (2004): "An Evaluation of PROHECO" -Consulting report for The World Bank - mimeo ElSalvador- Ministerio de Educaci6n (2002): "Factores Asociados a1rendimiento de 10s estudiantes que se sometieron a la PAES 2000", San Salvador, El Salvador. El Salvador - MINEDKJES (2003): "Plan Especial deEstimulo a la Labor Educativa Institucional para el AAo 2002", Ministerio de Educacibn, Universidad de El Salvador, San Salvador. Figueredo, V and S, Anzalone: (2003): "Alternative models for secondary education in developing countries: rationale and realities", American Institutefor Research, mimeo Florez, C., R. Ruiz, et al. (2003): "Lecciones aprendidasde la autonomia escolar nicaraguense", New School University, New York City. Fuller,B andDHolsinger (1993): "Secondary Education inDevelopingCountries", The World Bank,ESPDiscussionsPapersSeries N.7, Washington DC. Fuller, B and Riverola, M (1998): "Nicaragua's Experiment to Decentralize Schools: Views of Parents, Teacher and Directors", Working Paper Series on Impact Evaluation of Education Reforms, PaperNo. 5, The World Bank, Washington, DC Gershberg, A I(1999): "Fostering Effective Parental Participation in Education: Lessons from a Comparison of Reform Processes in Nicaragua and Mexico", in World Development, Vo. 27 (4), pp. 753-771 215 Gershberg, A and Winkler, D (2000) "Education Decentralization in Latin America: The Effects on the Quality of Schooling", LCSHD Paper Series No. 59, The World Bank, Washington, DC Gershberg, A. et al. (2001): "Lecciones Aprendidas de la Autonomia Escolar Nicaragiiense ",The New School University, Managua. Gershberg, A. and Kaestner, R (2002): "Lessons learned from Nicaragua's School Autonomy Reform: A review of Research by the Nicaragua Reform Evaluation Team of the World Bank", Milano Graduate School of Managementand Urban Policy, mimeo Gershberg, A. I. (2003): "Empowering Parents while Makingthem Pay: Autonomous Schools andEducation Reform Processes in Nicaragua", paper written for the project fundedby the Tinker Foundation. Glewwe, P (1996): "The relevanceof standard estimates of rates of return to schooling for educationpolicy: a critical assessment", inJournal of Development Economics, Vo1.5 1,267-290 Glewwe, Ilias and Fremer (2003): "Teacher Incentives", NBERWorking Paper 9671. Guatemala-MINEDUC (2000): "Plan de Gobierno: Sector Educaci6n 2000-2004", Guatemala. Guatemala - MINEDUCDP Tecnologias (2000): "Estudio Cuasi-Experimental de Resultados del Aiio 1999", ResumenEjecutivo, Ciudad de Guatemala, Guatemala. Guatemala-MINEDUCDPTecnologia (2002): "Estudio Quasi-Experimental de Resultados de PRONADE 2001",GuatemalaCity, Guatemala Guatemala - Ministry of Education (MINEDUC) (2003a): "Programa Nacional de Autogestidn para el Desarrollo Educativo -PRONADE", Guatemala, Guatemala Guatemala - Ministry of Education (MINEDUC) (2003b): "PRONADE - Actores Principales del Programa", Available at www.mineduc.g;ob.g;t. Guatemala - MINEDUC (2004): "Anuario Estadistico 2002", Guatemala. Hammer, J (1996): "The Public Economics of Education", Policy ResearchDepartment, The World Bank, mimeo Hannaway, J. (1991): "The organization and management of public and catholic schools: Looking insidethe black box", InternationalJournal of Educational Research, 15 Hanushek,E., (1995): "Interpreting Recent Researchon Schooling inDeveloping Countries", in The World Bank Research Observer, 10(2), 227-246. Heuser, S., Jacobs, Iand I. (2003) :Unpublished Base Document for the 2003 Umansky Education MinisterialMeeting, Mexico, August 2003, OAS, Washington DC. 216 Holsinger, D and Cowell, R (2000): "Positioning Secondary School Education in Developing Countries", in Series Policies and Strategies for Secondaryeducation, International Institutefor Educational Planning, UNESCO, Paris Honduras - Secretariade Educaci6n Pitblica (SE) (2000): "Estadisticas de PROHECO", Unidad Coordinadora, Tegucigalpa. Honduras - Secretaria de Educaci6n Pitblica (2003): "Revisado y Actualizado Manual de Organizaci6n y Funcionamiento de las AECOs", Tegucigalpa, Honduras Honduras -UMCEKJniversidad Pedag6gicaNacional Francisco Morazan/SE (2003a): "Znforme Nacional de Rendimiento Academic0 2002", Tegucigalpa, Honduras. Honduras -UMCEKJniversidad Pedag6gicaNacional Francisco Morazin/SE (2003b): "Znforme sobrefactores asociados a1 rendimiento acadkmico Honduras 2002", Tegucigalpa, Honduras. IFPRI (2003): "Proyecto PRAF/BID Fase 11:Impacto Intermedio", Sexto Informe. Ichino, A and Winter-Ebmer, R (1999): "Lower and upperbounds of returns to schooling: an exercisein instrumentalestimation with different instruments", inEuropean Economic Review N. 43. Figueredo, V and Anzalone, S (2003): "Alternative models for secondary education in developing countries: rationale and realities", American Institute for Research, mimeo. Forster, P. andJ. Martinez (2002): "Evaluaci6n de Impacto de Proyecto de Mejoramiento de la Educai6n Bisica (PROMBEB)", Informe Final PKF Worldwide, Fairfax, Virginia -unpublished document. Fuller, B, and P. Clarke (1994): "Raising School Effects While Ignoring Culture?Local Conditions andthe Influence of Classroom Tools, Rules, and Pedagogy", inReview of Educational Research 64(l), 119-157. Fuller, B and Riverola, M(1998): "Nicaragua's Experiment to Decentralize Schools: Views of Parents, Teacher and Directors", Working Paper Series on ImpactEvaluation of Education Reforms, Paper No. 5, World Bank, Washington, DC Jacob, B, and Levitt S (2002). "Rotten Apples: An Investigation of the PrevalenceandPredictors of Teacher Cheating." NBER Working Paper 9413. Cambridge, MA: National Bureauof Economic Research. Jimenez, E., and Sawada, Y (1999): "Do Community-Managed Schools Work? An Evaluation of ElSalvador's EDUCOProgram", in The WorZdBank Economic Review 13(3): 415-441. Junankar, P N, Liu, J (2003) "Estimating the social rate of return to education for indigenous Australians." Education Economics 1l(2): 169-192. 217 Kane, T, and Staiger D (2001). "Improving School Accountability Measures." NBER Working Paper 8156. Cambridge, MA: National Bureau of Economic Research. King, E, and Ozler, B (1998): "What's Decentralization Got to Do with Learning? The Case of Nicaragua's Autonomy Reform", Working Paper Series on Impact Evaluation of Education Reforms, No. 9, The World bank, Washington, DC. King, E Ozler, B and Rawlings, L B. (2001): "Nicaragua's School Autonomy Reform: Fact or Fiction?", Working Paper Series on Impact Evaluation of Education Reforms N. 19, The World Bank, Washington, DC. Koretz, D (2002). "Limitations in the Use of Achievement Tests as Measures of Educators' Productivity." Journal of HumanResources 37: 752-77. Kraft, R. (1998): "Rural Education Reforminthe Nueva Escuela Unitaria of Guatemala", AED, Washington DC. Kraft, R (2003): "Caracterizacidn del Iy 11Ciclo de educacidn bcisica en El Salvador, un enfoque cualitativo ". Kraft, R. (2004): "Educaci6n BBsica" inAID (2004), Chapter 111. Lavy, V (1996): "School Supply Constraints and Children's Education Outcomes inRural Ghana", inJournal of Development Economics, 51:291-314. Lavy, V (2002): "Paying for performance: the effect of teachers' financial incentives on students' scholastic outcomes" - mimeo Lewin, K and Caillods, F(2001): "Financing the Development of Secondary Education in Developing Countries", inProspects, vol. XXXI, N.I,March 2001 Loayza, Fajnzylber and Calder6n (2002): "Economic Growth in Latin America and the Caribbean Stylized Facts, Explanation and Forecasts", in International Monetary Funds Seminar Series N.2003, Washington DC. Lloyd, C., Mensch, B. and W. Clark, (2000): "The Effects of Primary School Quality on School Dropout among Kenyan Girls andBoys", in Comparative Education Review, 44(2): 113-147. Lloyd, C, ElTawila, S, Clark, W and B. Mensch, (2003): "The Impact of Educational Quality on School Exit inEgypt", in Comparative Education Review, 47(4): 444-467. Loeb, S and M.Page (2000): "Examining the Link betweenTeacher Wages and Student Outcomes: The Importance of Alternative Labor Market Opportunities and Non-pecuniary Variation." Review of Economics and Statistics 82: 393408. Maluccio, J, (2004). "Using quality of interview informationto assess nonrandom attrition bias in developing country panel data." Review of Development Economics 8(1): 91-109. 218 Marques, J. and Bannon, I(2003) : "Central America: Education Reformin a Post-Conflict Setting, Opportunities and Challenges", CPR Working Paper 26667, PaperNo. 4, The World Bank, Washington DC. Marshall,J (2003a): "If you Buildit Will They Come ? Primary School Quality andGrade AttainmentinRuralGuatemala", Stanford University, mimeo Marshall, J (2003b): "School Attendance and Academic Achievement: a Simultaneous Equations Approach", Stanford University, mimeo Marshall, J. (2003~):"Academic Achievement, School Attendance, and Education Policy in Honduras" - mimeo. Meza, D (1997): "Descentralizacibn Educativa, Organizacih y Manejo de las Escuelasa Nivel Local: El Caso de El Salvador", EDUCO, InformeNo. 9, The World Bank, Washington, DC Meza, D (2001) "El Salvador: El Casode EDUCO', Presentationheld at the Taller Internacional sobre Participacibn y Empoderamiento paraunDesarrollo Inclusivo, July 9-11, 2001, World Bank, Lima, Peru Meza, D and Guzman, J (2004): "EDUCO: A Community-Managed Education Program in Rural Areas of El Salvador (1992-2003)", Paper preparedfor the 2004 Shanghai Conference on Scaling- UpPoverty Reduction Mella, 0 (2002): "Estudio de FactoresAsociados a1ResultadoEscolar", mimeo MINEDUC (Guatemala), OEI, & World Bank, no year: "Material Didictico en Aulas Multigrados", San Salvador, El Salvador. Murnane,R, and Cohen, D (1986). "Merit Pay and the Evaluation Problem: Why Most Merit Pay Plans Fail and a Few Survive." Harvard Education Review 56: 3-17. Nicaragua - Ministerio de Educacibn, Cultura y Deportes (MECD) (1993): "Reglamento General de Educacibn Primaria y Secundaria", Managua, Nicaragua Nicaragua- Ministerio de Educacibn, Cultura y Deportes (2003): "Znforme de Resultados 2002", Managua, Nicaragua. Nicaragua- MECD (2004): "Politicas de educaci6n basica y media" -Versibn preliminar, Managua, Nicaragua. Nicaragua Reform Evaluation Team (1996): "Nicaragua's School Autonomy Reform: A First Look", Working Paper Series on Impact Evaluation of Education Reforms N. 1, The World Bank, Washington DC Nickell, T and Quintini, J (2002): "The consequencesof the decline inpublic sector pay in Britain:alittle bit of evidence", Economic Journal, 112 (477) 219 Ndchez,M.(2003): "Evaluacidn de la Descentralizacidn del Fondo de Desarrollo Professional Docente-aAo 2001", San Salvador: Ministeriode Educacidn. OECD (2002): "Education at a Glance - OECDIndicators", Paris, France. OEI (2003): "Organizacidn y Estructurade laFormacidnDocenteen IberoamCrica", documento de trabajo. OEI (2004): "Diagndstico de 10s Sistemas de EvaluacidnEducativade CentroamCrica", consulting report for OEI, San Salvador. OEI (2004a): "Mapa de CooperacibnInternacionaldel Sector Educacidnen CentroamCrica", unpublisheddocument, San Salvador PAHO (2004): Epidemiological Bulletin, 25(l), 9-13. Paqueo, V and J Lammert (2000): "Decentralizationin Education", in K.EdgeDecentralization and School-BasedManagement Resource Kit, The World Bank, Washington, DC Patrinos, Hand Velez, E (1996): "Costs and Benefitsof Bilingual EducationinGuatemala:A PartialAnalysis", HumanCapitalDevelopmentWorking Paper N.74, The World Bank, WashingtonDC Patrinos, H(2002): "A Reviewof Demand-SideFinancingInitiativesinEducation", Background Paperfor the The World Bank World DevelopmentReport 2003. Paxson, C, andN.Schady, (2002): "The Allocation and Impactof SocialFunds: Spendingon SchoolInfrastructurein Peru", in WorldBank Economic Review, 16(2): 297-319. Park, A, and HannumE(2001). "Do Teachers Affect Learningin DevelopingCountries?: Evidencefrom MatchedStudent-TeacherDatafrom China." Paper preparedfor the conferenceon RethinkingSocial ScienceResearchon the DevelopingWorld inthe 21st Century, Park City, Utah, June 7-1 1. Parker, C (2004): "Teacher Incentives and Student Achievement in Nicaraguan Autonomous Schools", WashingtonDC, The World Bank, mimeo PREAL (2002): "El Tiempo de Aprendizaje en America Latina", Serie Politicas, Santiagode Chile. PREAL (2002a): Informe de Progreso Educativo: El Salvador". " PREAL (2003): "Time to Act: A Report Card on Education in Central American and the Dominican Republic", Task Force on EducationReforminCentralAmerica. PREAL(2003a): "Desarrollo de laeducacibnen sectores rurales.PREAL Formas y Reformasde la Educacibn", Serie Politicas5 (13). 220 Psacharopoulos,G, J. Valenzuelaand M.Arends (1996): "Teacher Salaries inLatin America: A Review." Economics of Education Review 15(4): 401-6. Psacharopoulos,G and Patrinos, H(2002): "Returns to InvestmentinEducation:a Further Update", World Bank PolicyResearchWorking Paper N.2881, The World Bank, WashingtonDC Rapalo, R (2004): "Calidad de la educaci6ncomo mejorade 10s aprendizajes" inAED (2004), (ChapterV). Ravela,P (2002): "Como presentansus resultados10s sistemas nacionalesde evaluacioneducativa en America Latina", PREAL, Working PaperN.22 Rivkin, S, Hanushek,Eand Kain, J (1998). "Teachers, Schools and Academic Achievement." NBER Working Paper No. w6691. Cambridge, MA: NationalBureauof EconomicResearch. SANIGEST (2003): "Fortalecimiento del SectorEducativoen CostaRica: Situaci6nActual y Opcionesde Politica", consultingreportfor the World Bank Sawada, Y (1999): "Community Participation, Teacher Effort and Educational Outcome: The Case of El Salvador's EDUCOProgram", Working Paper No. 307, The World Bank, Washington, DC Rivkin, S, Hanushek,E and Kain, J (1998). "Teachers, Schools andAcademic Achievement." NBER WorkingPaper No. w6691. Cambridge, MA: NationalBureauof Economic Research. Sawada, Y (2003): "Community participation,teacher effort and educational outcomesin a developingcountry", The World Bank, mimeo Sawada, Y and Ragatz, A (2004): "Decentralization of Education,Teacher Behaviour and Outcomes: the Case of El Salvador's EDUCOProgram", The World Bank, mimeo Schiefelboin, E (2004): "Se Reduce la Repeticibn, Deserci6ny Extraedad, per0 Continuael Bajo Aprendizaje, Especialmenteen 10s Grupos de MenoresIngresos", in AED (2004), ChapterI. Skoufias,E. and S Parker (2001): "Conditional CashTransfers and Their Impacton ChildWork and Schooling: Evidencefrom the PROGRESA Programin Mexico", DiscussionPaper N. 123, InternationalFoodPolicyResearch Institute,WashingtonDC The World Bank (1995), Staff Appraisal Report of Nicaragua's Basic EducationProject, Report No. 13705-NI,The World Bank, Washington,DC The World Bank (2002): ImplementationCompletionReportP007399,WashingtonDC. The World Bank (2003): "Closing the gap in education and technology", World Bank Latin American and Caribbean Studies, WashingtonDC. The World Bank (2004a): "Guatemala-Equity and Student Achievement inPrimaryEducation", Volumes Iand 11,Report N.22691, WashingtonDC. 221 The World Bank (2004b): "El Salvador -Education Strategy Note", unpublisheddocument. The World Bank (2004~):"Honduras - Education Strategy Note", unpublisheddocument. The World Bank (2004d): "Honduras. Development Policy Report. Accelerating Broad Based Growth", Report N.28222,Washington DC. The World Bank, (2004e): "Strengthening the Foundation of Education and Training in Kenya", Sector Report No. 28064-KE, Washington DC. The World Bank (20040: "Making service delivery workfor poor people ",The World Bank and Oxford UniversityPress UIS (2000, 2003): "Global Education Digest: Comparing Education Statistics across the World", UNESCO Institute for Statistics, Montreal, Canada. Umansky, I(2004): "Factors Associated with Education Achievement in Central America: An Examination of Existing Research", background paper for the World Bank Central America Education Strategy Paper. UNESCO (2003): "Primer Foro Centroamericanode Educacidnpara todos 2003", InformeFinal Relatoria. UNESCOLaboratorioLatinoamericano de Evaluaci6n de la Calidad de la Educacidn (LLECE) (1998): "First International Comparative Study of Language, Mathematics, and Associated Factorsfor Students in the Third and Fourth Yearsof Primary School: First Report", UNESCO, Santiago, Chile. UNESCO/Laboratorio Latinoamericano de Evaluaci6n de la Calidad de la Educacidn (LLECE) (2000): "First International Comparative Study of Language, Mathematics, and Associated Factorsfor Students in the Third and Fourth Yearsof Primary School: Second Report", UNESCO, Santiago, Chile. Valerio, A and Rojas, C (2004): "Education Decentralization and School Management Local Communities: The Case of PRONADE', in The World Bank (2004a), Vol. 11. Vawda, A and Patrinos, H(1999): "Producing Educational Materials in Local Languages: Costs from Guatemalaand Senegal", inInternational Journal of Educational Development, 19,287-299. Velez, E, Schiefelbein, E, and Valenzuela, J (1993). "Factors Affecting Achievement in Primary Education." Human Capital Working Paper Series 12186. Washington, DC: The World Bank. Villegas-Reimers, E, and Reimers, F(1996). "Where Are 60 Million Teachers?The Missing Voice inEducational Reforms around the World." Prospects 26(3): 469-92. Wright, S, Horn, S, and Sanders, W (1997). Teacher and Classroom Context Effects on Student Achievement: Implications for Teacher Evaluation. Journal of Personnel Evaluation inEducation 11: 57-67. 222 Younger, S (2000): "The Incidenceof PublicServices and Subsidies in Peru: Evidencefrom HouseholdSurveys", WorkingPaperN. 103, Ithica, Cornel1Foodand Nutrition PolicyProgram, NY. Younger, S (2003): "Benefits on the Margin: Observationson the MarginalBenefit Incidence", in World Bank Economic Review, 17(1): 89-106. 223 LIST OFANNEXES (Tables andFigures) Annex I:Tables andFiguresfor Chapter I ............................................226 Table 1.2: Average Years of Educational Attainment (Barro and Lee Dataset) ..................226 Table 1.1:Average Years of Education Attained................................................................. 226 Table 1.3: GER and NER in Pre.Primary. Primary and Secondary (MINED).................... 227 Table 1.5: Evolution of GER and NER ................................................................................ Table 1.4: GER and NER in Pre.Primary, Primary and Secondary (Household Surveys) .. 227 Table 1.6: Survival Rates for the 7 to 17 Cohort (survivor plots) ........................................ 228 228 228 Table 1.8: Repetition Rate (different sources)...................................................................... Table 1.7: Age-by-grade distortion....................................................................................... Table 1.9: Standardized Test Scores..................................................................................... 229 230 Table 1.10. Distribution of Educational Outcomes by UrbanRural ................................... 231 Table 1.12: Distribution of Educational Outcomes by Quintiles ......................................... Table 1.11: Disagregationof test scores across urbanhral area ......................................... 232 Table 1.13: Primary Completion Rate by Cohort ................................................................ 233 234 235 Table 1.15: Distribution of Educational Outcomes by Gender ........................................... Table 1.14: Rates of Return per Income Quintile................................................................. Table 1.16: Distribution of Educational Outcomes by Indigenous/Non-Indigenous...........236 237 Table 1.17: Primary completion rate for indigenous and not indigenous and different Cohorts.......................................................................................................................... 237 Figure 1.1:Per capita income and GCR inprimary ............................................................ 238 Figure 1.3: GERs in urbanand rural areas according to different sources........................... Figure 1.2: Survivor function plots....................................................................................... 238 239 Figure 1.5: Secondary NER by quintile inEl Salvador and CostaRica............................... Figure 1.4: Efficiency Indicators inUrban and Rural Areas ................................................ 239 240 240 Figure 1.7: Primary Completion Rateby quintile and cohort .............................................. Figure 1.6: Evolution o f primary NER by quintile............................................................... Figure 1.8: Evolution across cohorts o f primary completion rate ........................................ 241 242 Annex 11: Methodologyfor the Calculationof PrivateRates of Return 244 .. Annex 111: Evolutionof Rates of ReturninCentral America ...................245 Figure 3.1:Evolution across cohorts o f rates o f return and educational attainment for several countries ............................................................................................................................. 245 ............................................................................................................................................ Figure 3.2: Evolution of private returns to schooling b y educational level and cohort...... 246 Annex IV: NationalAssessment Systems in Central America ..................247 Table 4.1:Characteristics of National Assessment Systems in Central America .............. 247 224 Annex V: Teacher EducationinCentralAmerica ...................................... 249 Table 5.1:Teacher Education............................................................................................... 249 Annex VI: Teacher Salary in CentralAmerica ........................................... 251 Annex VII: Curricular and PedagogicalReformsinCentralAmerica .....253 Annex VIII: Methodological Aspects on the Analysis of Effective Schools inHonduras and Nicaragua ........................................................ 255 Annex IX: InterculturalBilingualProgramsinNicaragua & Honduras 257 . Annex X: Tables andFigures for Chapter I V .............................................. 259 Table 10.1: RecurrentExpenditureper Child and Student ................................................ 259 Table 10.3: Distributionof Public Enrollmentby IncomeQuintiles-Primary...................259 Table 10.2: Distributionof PublicEnrollmentby IncomeQuintiles-Preschool................ 259 Table 10.4: Distributionof Public Enrollmentby IncomeQuintiles-Secondary...............260 Figure 10.1:Test Scores and EducationalExpenditureby RegioninHonduras .................260 Table 10.5:Distributionof Public Enrollmentby IncomeQuintiles-Tertiary ..................260 ************ 225 Annex I:TablesandFiguresfor Chapter I Table 1.1: Average Years of EducationAttained 25 year 40 year 55 year cohort cohort cohort El Salvador (2002) 8.05 6.41 4.04 Guatemala (2000) 5.36 3.85 2.11 Honduras (2003) 6.45 5.56 3.80 Nicaragua (2001) 6.23 5.36 2.98 001. Table 1.2: Average Years of Educational Attainment (Barro and Lee Dataset) Population Population over 15 over 25 years old years old population >25 F ~~ ~ 1990 5.32 4.97 Latin America and the Caribbean 1995 5.74 5.38 15% 0.76 2000 I 6.06 I 5.73 I I I 1990 4.20 3.69 Honduras 1995 4.50 3.89 10% 0.39 2000 4.80 4.08 1990 3.65 3.60 Nicaragua 1995 4.09 4.01 22% 0.82 2000 4.58 4.42 1990 5.55 5.57 Costa Rica 1995 5.77 5.82 8% 0.44 2000 6.05 6.01 Table 1.3: GER and NER inPre-Primary,Primary and Secondary (MINED) I Pre-Primary I Primary I Secondary Year GER NER GER NER GER NER El Salvador 12002 I 4 8 I 4 3 1 y:5 182 ('-') 1i! 124 (10-11) 1 85 (1-6) 47 (7-11) Guatemala 2001 55 41 103 85 33 22 (7-12) Honduras (a) 2001 57 na 108 na 39 (7-11) na 2003 54 na 119 na 54 (7-11) na 48 (7-12) na 2001 26 26 104 82 54 36 (7-11) Nicaragua(a) 2002 28 28 109 86 57 38 2003 29 29 106 84 59 39 Costa Rica 2002 87 86 (b) 105 99 66 59 (7-12) Ecuador 2000 69 115 99 57 48 Philippines 2000 na 113 93 77 53 Bolivia 2000 46 116 97 80 68 ITable 1.4: GER and NERinPre-Primary,Primaryand Secondary (HouseholdSurveys) I I I Pre-Primary I Primary I Secondary I El Salvador(2002) 1 85 (1-9) 30 (10-12) z; 44 141 I ;:5 I87(1-6) li I 52 (7-11) 28 (7-11) Guatemala (2000) I 2 3 121 I 9 9 I 7 8 11 125 (7-12) ~ Honduras (2002) 37 30 105 85 50 35 (7-11) Nicaragua(2001) 34 28 111 81 58 39 (7-11) Source: Household Surveys (El Salvador: EHPM2002; Guatemala, ENCV 2000; Honduras, EHPM2002; Nicaragua, EMNV 2001). 227 Table 1.5: Evolution of GER and NER Source: El Salvador, EHPM, various years; Guatemala, MINED official statistics;Honduras, EHPM, various years; Nicaragua, EMN, various years; Costa Rica, MINED, official statistics. (a) Primary is grades (1-6);secondary is grades (7-11); (b) lower secondary without brackets and upper secondary (grades 10-12)within brackets. 1Table 1.6: Survival Rates for the 7 to 17 Cohort (survivor plots) I SR to last grade of SR to first grade of SR to last grade of primary secondary secondary 69% (grade 11) El Salvador (2002) 75% (grade 9) 70% (grade 10) 83% (grade 6) 79% (grade 7) Guatemala (2000) 68% 54% 45% (grade 11) Honduras (2002) 80% 53% 47% (grade 11) Nicaragua(2001) 76% 67% 55% (grade 11) Table 1.7: Age-by-grade distortion Age for grade distortion(average age/official age) 1-6 7-12 lo l2 average average El Salvador 1.128 1.124 1.114 1.113 1.073 1.073 1.068 1.062 1.056 1.066 1.066 1.056 1.101 1.062 Guatemala 1.203 1.201 1.190 1.167 1.126 1.133 1.134 1.158 1.126 1.110 1.115 1.089 1.100 1.083 Honduras 1.081 1.105 1.114 1.099 1.077 1.089 1.127 1.110 1.122 1.113 1.098 1.175 1.094 1.125 Nicaragua 1.043 1.085 1.083 1.058 1.074 1.068 1.112 1.081 1.076 1.082 1.114 1.110 1.026 1.063 Proportionof children out of age (%) 1-6 7-12 lo l1 l2 average average El Salvador 52.8 55.0 59.8 60.8 63.3 58.5 62.0 61.0 62.7 67.1 63.9 69.0 57.8 63.6 Guatemala 66.3 73.2 76.5 74.5 75.1 74.8 72.2 75.9 74.1 73.0 73.3 70.6 72.4 73.4 Honduras 76.0 74.1 76.9 75.0 75.3 80.8 76.6 78.5 79.1 78.5 79.1 82.2 76.1 78.5 Nicaragua 76.4 75.8 80.2 77.2 79.8 80.8 81.8 78.6 76.9 78.7 80.1 81.0 78.0 79.5 Source: HouseholdSurveys 228 Source RR First Grade RR RR First Grade RR Secondary Primary Primary Secondary El Salvador MINED,2002 16% 6% (1-9) ;7% (1-6) 3% 3% (7-11) HS, 2002 22% 11% (1-6) 4% 3% (7-11) Estimates (a) 2002 35% Guatemala MINED,2000 na 15% 4% (7-9) 1% (10-12) HS, 2000 21Yo 12% 6% 5% (7-11) Honduras MINED,2001 14% 9% 16% (7-12) HS na na 16% na Estimates(a)2001 30% Nicaragua MINED, 2002 15% 9% 9% 6% (7-11) HS, 2001 12% 8% 9% 6% (7-11) Estimates(b)2002 30% Costa Rica MINED,2002 (c) 11% (16%) 6% (8%) 11Yo (1OYO) 16% (13%) LatinAmerica, WDI2000 12% Peru UIS 2000/01 11% 5% Colombia UIS 2000/01 5yo 4% Jordan 1 uIs2000/01 I I I I Romania UIS 2000/01 3% 2% Algeria UIS 2000/01 13% Paraguay UIS 2000/01 8% 1Yo Ecuador UIS 2000/01 2% 4% Philippines UIS 2000/01 2% 2% Bolivia UIS 2000/01 4% 4% Notes: (a) Based on estimates produced on the basis of age for grade disaggregation. In Honduras, householdsurvey data were used; while in El Salvador, MINED official data were used; (b) Estimates based on Arcia (2003); (c) urban (rural); (d) Come from UIS (2003). 229 ITable 1.9: StandardizedTest Scores Grade3 Grade4 or 6 Grades 11 and (9) Year Spanish IMath Spanish 1Math Spanish I Math El Salvador 1996 (a) na 4.0 2.2 2.9 1998 (a) 4.2 4.0 2.7 1.6 4.5 (2.4) 5.0 (1.1) 1999 5.2 5.2 2001 (b) 498 488 890 865 2002 (b) 1705 (1301) 1672 (1264) 2003 (Grade 6) (Grade 6) 1689 1678 Guatemala 1998 (c) 55.1 46.2 (National) 2000 61.5 48.7 2001 na na (OREALC) 2002(d) 51.8 38.9 58.3 46.0 National (Grade 4) (Grade4) Honduras 1998 (e) 41 43 47 40 1999 42 43 na na 2000 41 43 46 39 2002 38 44 43 39 Nicaragua 2000 (f) 247 251 244 248 Costa Rica 1999 (9) 79.1 67.8 2001 78.0 71.1 74.2 (69.2) 70.3 (55.6) 230 x 2 c o z L I I 00 - m - l e I ee2 I I I m .. m .. aa, m .. (d C (d C IIp N m m r\l m e Q\ m m 2E .e 2 E 2e, E c) -EE Y 2 e, Y e, e Y EE 3 W E m w cd m C c cd c 5 8C2 5 e, L L L L e , e , e , e , 3 3 3 3 - - e - 0 0 0 0 8 8 8 8 e , e , e , e , e , e , e , e , 3 3 3 3 C l C Y Y e , e , e , e , a C ) D C ) D I , 2 aQ c 8 rx E2 h h aQ x8 e E 13 2, k .L B i32 h x .L k B EL7 L Guatemala Honduras Nicaragua 2000 2002 2001 NonIndig Indig Non Indig Indig Non Indig Indig Average years of EducationAttained (25 year 6.57 3.59 6.55 5.29 6.31 5.54 GERPrimary 104.39 90.74 108.23 103.56 111.47 107.01 NERPrimary 83.96 70.89 87.39 84.15 81.75 77.07 GER Secondary 48.95 23.92 53.08 33.08 59.10 45.72 NER Secondary 36.03 16.13 36.60 20.72 39.98 26.64 Gross Completion Rate Primary (cohort 12-21) 54.24 26.88 60.06 42.05 52.60 39.03 - Gross Completion Rate Secondary (cohort 17-21) 13.44 5.31 13.73 5.53 18.80 11.02 Repetition rate primary 10.63 13.19 na na a.07 7.14 Rate of Return of an additional year of schooling 11.44 10.36 na na 9.25 9.28 [ Not Indig. 1Indig. 1 Not Indig. 1Indig. Not Indig. 1Indig. II 18 years cohort 25 years cohort 35 years cohort Guatemala, 2000 1168 I 37 1156 128 1I49 116 II Honduras,2002 70 48 65 50 58 23 Nicaragua,2001 54 64 57 51 44 67 237 Figure 1.1: Per capita income and GCR inprimary 0 500 1000 1500 2000 2500 3000 3500 4000 4500 ~ Source: WDI andMINED Figure1.2: Survivor functionplots HONDURAS 2002: SURVIVOR FUNCTIONPLOT COHORT 12-17 NlCARAGUA2001: SURVIVOR NNCTION PLOT CHILDREN 7 to 17 1.0: ,.._ i i 9 0.81 0.7 g 0.81 L c . . k-. 2 0.5:- b $ :I 0.4; 1 u i i - 0 1 4 i 0.0 , - 0.4 , . . , , , , , , , . , , , , , , , , . . , , , , , , . , . , , , , . . , , , , , , , , 0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 0 1 2 3 4 5 6 7 8 9 10 E D " SCHMUM AlTAINMENTINWDESCOMPLETED El SALVADOR 2002: SURVNOR FUNCTION PLOT CHILDREN 7 to 17 GUATEMALA:SURVNOR NNCTlON PLOT CHILDREN7 to 17 1.0: .- . o2 0.9j i - --k 0 9 02- 0 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . , , , , , , , , , , , , , , , , , , . , , , , , , , , , , , ,, , , ,, , ,, , ,,, , , ,,, , , ,,, , , , , , , ,,, , , ,,, . , ,, ,. , ,,, , ,, , , , , ,, , , ,- 0 1 2 3 4 5 6 7 6 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 0 1 2 3 4 5 6 'Ye '"*7 8 9 10 11 12 13 Source: Household Surveys 238 Figure 1.3: GERs in urban and rural areas according to different sources primary I secondary primary 1 secondary ElSalvador Figure 1.4: Efficiency Indicators inUrban and Rural Areas 14 12 10 8 6 4 2 0 age for repetition age for repetition age for age for repetition grade rate grade rate grade grade rate distortion distortion distortion distortion (grade 1) (grade 1) (grade 1) (grade1) ElSalvador Guatemala Honduras Nicaragua Source: HouseholdSurveys 239 Figure 1.5: Secondary NER by quintile in ElSalvador and Costa Rica 80.0 ~ 1 50.0 40.0 I II 111 IV V /BE1Salvador NER mCosta Rica NERI Source: Household Suweys Figure 1.6: Evolution of primary NER by quintile Honduras El Salvador 900 I 900 800- . 80 0 70 0 60 0 50 0 4 a o i 400 30 0 30 0 200 -. I 1 20 0 1 0 0 - I i o 0 0 0 Qi Q2 Q3 Q4 Q5 1+2002 3-1995 1 /+ZOO2 3-19951 Nicaragua Source: HouseholdSurveys ~ 1 0 0 0 - 60 0 ::: 50 0 ::: ~ 1 0 0 Qi Q2 Q3 04 0 5 I 1+2001 3-19931 240 II Nicaragua El Salvador io0 00 10000 I I 90 00 90 00 80 00 80 00 7000 70 00 60 00 60 00 50 00 50 00 40 00 40 00 30 00 30 00 20 00 20 00 io00 10 00 0 00 0 00 I II 111 IV V I II 111 IV V 1 [+I8 year +25 year +35 year] 1+18year+25year-t35yearI Guatemala Honduras I 90 00 io0 00 90 00 ~ 8000 70 00 80 00 60 00 7000 50 00 60 00 50 00 40 00 ~ 4000 1 3000 3000 20 00 20 00 i o 00 0 00 1 I II 111 IV V I I+ 18 year +25 year-t35 yearI /+primary 18year *primary 25 year +primary35 year] 241 Figure 1.8: Evolution across cohorts of primary completion rate Honduras I Not Indigenous Indigenous /+I8 +25 1 - 3 5 ] Guatemala I E' s F M BPE8IO a 8 so Not Indigenous Indigenous [+la +25 -351 Nicaragua W E " Z W -E D1 to E " P ae re Not Indigenous Indigenous [+Nicaragua I 8 +Nicaragua 25 -Nicaragua 351 Source: HouseholdSurveys 242 rI Box 1.1: Determinants of grade-foi tge distortion inGuatemala analyze ove lment new Determinants of grade-for-age (t-statistics) Dependentvariable: gradefor age Z-score oach that esti the -0.08 (-19.99) Enrolledlate -0.15 (-4.12) Years delay -0.31 (-25.00) Heightfor age 0.13 (15) HH size -0.02 (-4.37) HH Cons. Per capita 0 (9.28) Female 0.05 (2.79) Kiche 0.02 (0.46) Q'ueqchi -0.16 (-3.96) Kaqchiquel 0.12 (3.3) Mam -0.02 (-0.39) Other Maya 0 (0.01) Repeater -0.78 .25.53) Constant 1.22 (21.1) IObs 5862 R-square 0.49 that grade repetition transla pita consumption (an indicato well-being) are both associated istics, girls do slightly bette ethnic variation. Ethnicity tion. The K'ichb, the Mam an stricted sense indistinguishable fron 1/3 of a year behind similarly situated Ladinos while the tenthof ead. (Source: Edwards (2004)). 243 Annex 11: Methodologyfor the Calculationof Private Rates of Return We specify the following model: yi = Cp(Si, Xi, Zi) +Ui (1) where yi i s the logged (hourly) labor market earnings for individual i; si stands for completed years of schooling, Xi i s a matrix of personal characteristics other than schooling, namely, age, age squared, gender, and zi is a matrix of context-specific circumstances, namely urbanhral and regional location The last component, ui, is a random disturbance term that captures unobserved characteristics. The functional form for education has been deliberately left unspecified in equation (l), because the model will be specified in two different ways to address two main questions: (i) valueisattachedtoanextrayearofeducation; and(ii)what valueisattachedto what different educational levels (primary, secondary, or tertiary). The specifications are the following. Respectively: Linear formulation of education, where the model assumes a constant return to one additional year of schooling regardless of the level: psch could be interpreted as the returns to one year of additional schooling278and Sscho stands for the years of schooling for individual i. Spline form of Years o f schooling, where the average return to one additional year of schooling can vary between levels of education: The functions are estimated using mere OLS techniques, which do not adjust for the fact that unobservable characteristics, such as ability or the quality o f education, could generate a correlation between the error term and the schooling variable biasing the rates of return; and for selectivity biases generated, for instance, by the fact that many adults are not wage workers279. A recent review (Card 2001280), focusing on the causality debate between schooling and earnings, reaffirms the conclusion that the effect of ability and related factors does not exceed 10% of the estimated schooling coefficient. and finds that instrumental variable estimates of the returns to education based on family background are higher than classic OLS estimates281. Psacharopoulos and Patrinos (2002), therefore, conclude that the estimation method makes little difference. 278This is assuming that the only cost of additional schooling is forgone wages. Inthis case, the above functional form will yield Psch as the private return to schooling by strainghtforawrd differentiation. The annual private rate of return i s the annual increase in income (yi-yi-1) divided by the annual cost of the investment (yi-I): (yi-yi.l)/yi.l = yJyi.1 - 1 = ePo + Psch Sscho,i + xi'p+ zi'pz +ui / ePo + Psch (Sscho,i -1)+xi'p +zi'pz +ui -= ePsch-,- Psch. (see Glewwe, 1996). 279For a usefulpaper summarizing these issues, see Glewwe (1996). 280See Card (2001). 281For the use of instrumental techniques in the estimation of rates of return, see Arias (2004); and Ichino, A and Winter-Ebmer, R (1999). 244 Annex 111:Evolutionof Ratesof ReturninCentralAmerica Figure 3.1 shows an interesting trend in which private returns decreased less than attainment increased for the 31-45 year-old cohort and more than attainment increased for the 18-30 year-old cohort (in all countries except Guatemala). This could be due to an educational attainment threshold after which rates of return start declinin more steeply?82 Or it could reflect a decline in the quality of the education inrecent years.f 8 3 Figure3.1: Evolutionacross cohortsof ratesof returnand educationalattainmentfor several countries El Salvador Nicaragua 12 I : 10 10 8 4 2 ! I 31-45 18-30 18-30 I Guatemala Honduras I 16 14 12 10 8 6 4 4 2 2 I I 0 46-60 31-45 18-30 I551 j WI [251 Cohorts Cohorts l+Coeff on Years of Schooling +Educational Attainment) /+Coeff on Years of Schooling +Educational Attainment] Source: Household Surveys Figure 3.2 illustrates how rates of return evolved across educational cycles and cohorts for each of the countries. In all countries, with the big push for primary education, rates o f return decreased across cohorts for this cycle; and this also happened at the upper secondary level in Guatemala and El Salvador, and at the lower secondary level in El Salvador, Nicaragua and Honduras284.Again, the increase in educational attainment can also explain this trend. The trends for upper secondary in both Nicaragua and Honduras 282I t would be useful to analyze the experience of the OECD countries to corroborate this explanation. 283This would be difficult to test as national standardized tests have only been introduced recently. 284 It should be noted that also in Costa Rica, although adopting a different methodology, it is found that the premium on completed primary, and secondary incomplete and complete, have been decreasing since the early 1980s (Sanigest, 2003). 245 are, however, quite interesting. After an initial decrease, rates o f return have been increasing in Honduras for the latest cohort: i s it due to the comparatively lower progress in secondary enrollment and completion, to increased demand for workers with secondary education, or to increased quality of the cycle? In Nicaragua, returns increased quite markedly for the last two cohort groups. The high rate of return for the 31-45 generation may be due to increased quality of the secondary cycle and higher demand for secondary school graduates after the Somoza dictatorship, which had left the education sector in a situation o f disarray and a weak and protected economy at the end of the 1970s. The persistently high rate of return for the 18-30 generation, in spite of growing secondary enrollment, may be explained by a further improved education and economic situation after the Sandinistaperiod inthe 1980s. Figure3.2: Evolution of private returnsto schoolingby educational level and cohort 18 30 16 14 25 12 20 10 8 15 6 10 4 5 2 0 0 primary Ilower secondary I upper secondaryI tertiary primary Ilowersecondary 1uppersecondary I tertiary El Salvador Guatemala I 3 0I I 14 25 12 20 10 8 15 6 10 4 5 2 0 0 primary 1lower secondary1 uppersecondary 1 tertiary primary 1lowersecondary 1upper secondary 1 tertiary Honduras Source: HouseholdSurveys 246 Annex IV: National Assessment SystemsinCentralAmerica Table 4.1: lharacteristicsa gational Assessment Svstemsin CentralAmerica Main Objectives Institutional Main Components Dissemination of Use/lmpact of the of the Systems Aspects the Information Information El Salvador 1)Diagnosticand In 1991,the Standardized strategq Lack of s stematic Some impacton action: provide Sistema Nacional Testingfrom 1993 8: curriculum information on de Evaluacionde development educational 10sAprendizajes (grades3, 4, 6, 9, quality to (SINEA)was 11 and 12) National education established. It is publicationonly of Some impact on authorities, managed by the Analysis of grade 11,12 exam teacher schools and Direccion Nacional Associated Factors development teachers; de Monitoreoy (for grade 4, 6, 11, Results are shared (grade 11, 12) 2)Accreditation Evaluacidnin the 12) with the schools (PAES, grade 11, Ministryof since 1998 Accreditationof the 12) Education end of secondary National reports on cycle More elaborated associatedfactors analysis is (very little undertakenoutside disseminated) the Ministryby foreign consultants Guatemala 1)Diagnosticand The management Standardized Lack of systematic Limitedfeedback action: provide of the system has Testing from 1998 strategy informationon been contracted (grades3, 6, 9) educational out to the quality to schools Universidaddel No national and teachers. Vale by the Analysis of publication Ministryof associatedfactors Education (grades3) Results are shared with the schools Honduras 1)Diagnosticand The system is Standardized National report on Littlefeedback in action: provide managedby the Testing since 1997 the results (limited the curriculum informationon Unidadde (grades3 and 6) printing, but, educational Medicionde la recently, also on quality to Calidadde la UMCE website) Some impacton education Educacidn Analysis of teacher authorities, :UMCE), Associated Factors development schools and 3utonomousunit (for grades 3 and Departmental teachers. ocated in the 6) reports (limited Limited impact on 2)Accountability: Yniversidad printing) accountability generate a Dedagdgica. culture of Resultsshared evaluation where with schools and education vlore elaborated teachers authorities and malysis is National and schools are indertaken in departmental accountable for JMCE with help of reports on their results nternational :onsultants associated factors (limited printing) 285A thorough review of the El Salvador National Assessment System is provided in Rapalo (2004). 247 Nicaragua 1)Diagnosticand Managedby the Standardized Lack of systematic action: provide DireccionGeneral Impacton Testing since 2002 strategy informationon de Capacitacion, awareness. (grades 3, 6) educational Evaluaciony quality to Planificacidn,in National report on education the Ministry of Analysis of the results, authorities, Education Associated Factors including schools and (for grades 3 and correlation analysis teachers. 6) Reportcards will be producedto share results at different levels. Costa Rica 1)Diagnosticand ~ Managedby the Standardized National reports on action: provide Divisionde Limitedfeedback Testing since 1988 the resultsfor all informationon Desarrollo (grades 6,9, 11, grades educational Curricular, in the quality to 12) Ministryof education Education Resultsshared authorities, Analysis of with schools and schools and associatedfactors teachers teachers. (for grades 6 and 2)Accountability: 9) Reportpublished generate a at the nationaland culture of departmentlevel evaluationwhere on associated education factors authorities and schools are accountablefor their results. 3)Accreditation :grades 9, 11, 12) Notes: This table was put together on the basis of indklual country studies :e El Salvador), pro OEI (2004). ct informationand 248 .-C on .-C e, cd 2 am6 Y .-3 v1 w C e, 0 cd G Y am -? i Y- e, O L/1 m .-C 0 .-f Y Y Y v1 w C Annex VI: Teacher Salary inCentral America gement and the "de-linking" between the educ ich the teacherteaches ble below summarizes average monthly wages of quality receive yet works full time the base salary is determined as the product between Seniority 139.30 57.84 101.20 26.31 Merit 16.66 6.92 5.43 1.41 Position 58.57 24.32 9.00 2.34 Geographic area (zonaje) 241.02 100.08 4.85 1.26 School Category 72.31 30.02 1.71 0.45 Title 121.61 50.50 15.14 3.94 each additional 5 years of experience); and geographic area 251 accomplishments like publications, participation in seminars, languages skills, etc. ase salary for category A on length of service. Since 2002, teachers Source: Arcia 252 Annex VII: Curricularand PedagogicalReformsinCentralAmerica stent with the new rature, math, natura months. The education hours in the academic earning outcomes in studen ce that most teachers do not teach more than four ho of 133 days per months, and 30 class hours per 286See Shiefelbein (2004), which reports figures of about 500 class hours a year (however, based on old data). *" See Rubio, F(2000): "Presentation on the Status of Bilingual Education", MEDIR Project, USAID. 253 inistry fully replaced. Th contents and structure that had been established un was overall an improvement over the past s ion of its different internal components and ~ 288See Di Gropello and Marshall (2004). 289See Honduras-Secretary of Education (2002): "Todos con educacion: Honduras 2003-2015", which reports up to 333 yearly hours (versus the official 1200 ones) in a worse case scenario. 290See MECD (2004): "Politicas de educacion basica y media" - Version preliminar 254 Annex VIII: MethodologicalAspects on the Analysis of Effective Schools inHonduras and Nicaragua Data. The national assessment datasets usedfor the education production function studies reviewed earlier in this chapter were used for both Honduras and Nicaragua. In Honduras, the Ministry of Education administered tests in Spanish, math, and science for both third and sixth graders along with surveys of students, parents, teachers, and principals in a randomly chosen group of 1090 schools. In Nicaragua, one stratified random sample of roughly 250 schools for the third grade and another of the same size for the sixth grade took Spanish and math exams and parents, students, teachers and principals completed surveys. Effective and control schools. A group of "effective schools" and control group schools were then identified for each country. For Honduras, effective schools are defined as those schools that score at least three quarters of a standard deviation below the mean on a constructed socio-economic status (SES) composite variable and at least three quarters of a standard deviation above the mean on a minimum of three of the six exams. The control group of schools are simply the remainder of schools whose SES index rank was three-quarters of a standard deviation or more below the mean. While this is a somewhat rough method of choosing a control group, the effective and control groups were compared across multiple background indicators and appeared to be quite similar. In Nicaragua, there was not a sufficient number of schools that qualified as effective schools using the same criteria as Honduras. This is probably at least in part due to the fact that the sample sizes were less than a quarter the size of that of Honduras as separate samples took the third and sixth grade exams. Effective schools were therefore defined as those schools scoring at least one quarter of a standard deviation above the mean on both exams and at least one quarter of a standard deviation below the mean on a constructed SES index. The control group was again defined as the remainder of the schools with SES index rankings one quarter of a standard deviation below the mean. For the sixth grade schools a lower bound cut off point was also established for the control group as without it the effective and control groups were statistically significantly different on a number of backgroundmeasures. By and large the background characteristics of the control group of schools matched the background characteristics of the effective schools. We did not want to match schools on school characteristics as it was those very characteristics that we wanted to observe for potential differences across high-performing and lower-performing low-income schools. At the same time we wanted to be sure that effective schools were not simply schools where parents had higher education levels, more access to services, or were located in more urban locations. Once creating the control groups we found no statistically significant differences between these types of background indicators for any of the groups under analysis. Methodology of analysis. A series of indexes were created for each data set using principal components analysis. These were based on the inter-correlation between related variables as well as on previous literature on identified characteristics of high- 255 performing, low-income schools. Paired t-tests were then conducted on all individual and composite variables to identify what characteristics were significantly different between the effective schools and the control groups. There were hundreds o f variables in each dataset, far to many to report here, so we have decided to only report those variables that were statistically significant as well as a few others that are interesting. There were a few variables that were statistically significant that are not reported here because they did not seem particularly enlightening or important. 256 Annex IX: Intercultural BilingualProgramsinNicaraguaand Honduras alEducation3- program i s limited, inpart becausethe current a itable education to incorporated into EIBplanningleaving scarceresourcesfor programimplementation. Issuesstill 291L a Prensa (2004) and MECD website. To be added that two Bank-funded projects support Intercultural and Bilingual Education program of MECD. The APRENDEI1(1999-2003) financed the construction of 58 new primary schools and 7 preschools, and the rehabilitation of 21 schools on the Atlantic Coast. Likewise, the Social Investment Fund(FISE) has built and equipped approximately 170 schools (including Institutes) since 2000 on the Atlantic Coast region. 292Forster and Martinez (2002). 257 requiring attentionincluded EIB teacher training, national EIBleadership, EIBevaluation, and classroom implementationzg3 . (ADEL-PEC);(v) a reschools were organized se, a total of 222 PROHE 293Forster and Martinez(2002); The World Bank (2002). 258 Annex X: Tables and Figuresfor Chapter IV Table 10.1: Recurrent Expenditure per Child and Student (in 2002 US $, unless otherwise specified) Pre-primary Primary Secondary Tertiary Exp. per Exp. per Exp. per Exp. per Exp. per Exp. per Exp. per Exp. per child student child student child student child student El Salvador (a) 66.8 171.8 180 217.8 69.5 260.7 55 246 -Guatemala na 113 Na 142 na 184 na 684 Guatemala,2001 34.5 110 133.9 145 24.9 184 97.6 549 IHonduras I 29.5 I 84.2 I 144.5 I 142 I 78.1 I 238.4 I na I na I Honduras,2001 28.2 79.5 152.5 151.2 78.3 280.2 139.8 932.6 Nicaragua 9.9 24 82.7 85.3 17.3 47.4 na na 1 7 ~~ 6staRica 1 na I na I Na I 250 I na I 390 I na Sources: Table 2 and MINEDofficialenrollment data (exceptCosta Rica:Sanigest,2004, and Honduras2002: Household Survey). Notes: (a)Primary= basic and secondary = uppersecondary. Source:HouseholdSurveys. HONDURAS2003 NICARAGUA 2001 GUATEMALA 2000 EL SALVADOR 2002 Source: HouseholdSurveys. 259 ITotal I 407,611 I 301,480 I 100.0 I 224,393 I 100.0 I 388,096 I 100.0 I100.0 I Source: Household Surveys. Source: Household Surveys. Figure 10.1: Test Scoresand Educational Expenditure by Region inHonduras * * ** * n n n n "." I Source: -100.0 -50.0 0.0 50.0 100.0 150.0 UMCE and Standardized changes in Spanish score, 2002-2000 (%) SlAFl 260