THE WORLD BANK W O R L D B A N K O P E R A T I O N S E V A L U A T I O N D E P A R T M E N T Books, Buildings, and Learning Outcomes An Impact Evaluation of World Bank Support To Basic Education in Ghana OPERATIONS EVALUATION DEPARTMENT ENHANCING DEVELOPMENT EFFECTIVENESS THROUGH EXCELLENCE AND INDEPENDENCE IN EVALUATION The Operations Evaluation Department (OED) is an independent unit within the World Bank; it reports direct- ly to the Bank's Board of Executive Directors. OED assesses what works, and what does not; how a borrower plans to run and maintain a project; and the lasting contribution of the Bank to a country's overall development. The goals of evaluation are to learn from experience, to provide an objective basis for assessing the results of the Bank's work, and to provide accountability in the achievement of its objectives. It also improves Bank work by identifying and disseminating the lessons learned from experience and by framing recommendations drawn from evaluation findings. W O R L D B A N K O P E R A T I O N S E V A L U A T I O N D E P A R T M E N T Books, Buildings, and Learning Outcomes An Impact Evaluation of World Bank Support To Basic Education in Ghana 2004 http://www.worldbank.org/oed The World Bank http://www.isdb.org/ Washington, D.C. © 2004 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 All rights reserved. Manufactured in the United States of America 1 2 3 4 07 06 05 04 The findings, interpretations, and conclusions expressed here are those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank cannot guarantee the accuracy of the data included in this work. The boundaries, colors, denomina- tions, and other information shown on any map in this work do not imply on the part of the World Bank any judgment of the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. No part of this work may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or inclusion in any information storage and retrieval system, without the prior written permission of the World Bank. The World Bank encourages dissemination of its work and will normally grant permission promptly. For permission to photocopy or reprint, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Cover photo shows villagers in Volta Region making improvements to a school pavilion constructed with World Bank resources. ISBN 0-8213-5884-7 eISBN 0-8213-5885-5 Library of Congress Cataloging-in-Publication Data has been applied for. Printed on Recycled Paper Contents vii Acknowledgments ix Foreword, Prólogo, Avant-propos xv Executive Summary, Résumen, Résumé Analytique xxvii Abbreviations and Acronyms 1 1 Introduction 1 Scope and Purpose of the Study 2 Evaluation Questions: What Explains Educational Performance? 3 Overview of the Study 7 2 Changes in Basic Education Since the 1980s 7 The Education System Before 1986 8 The 1986 Reform Program 9 FCUBE and Decentralization 10 Budget 11 School-Level Inputs 17 3 The Bank's Education Portfolio in Ghana 17 Overview 18 The Sector Adjustment Credits: EdSAC I and II 21 Investing in Basic Education: PSD and BESIP 23 The Role of Other Donors i i i B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S 25 4 The Bank's Impact on Education Policies and Outputs 25 The Bank and Educational Reform 28 Education Sector Outputs 31 5 Educational Performance Has Improved 31 School Attainment: Higher Enrollments and Better Completion Rates 34 Improved Test Scores 37 6 Causes and Consequences of Improved Educational Performance 37 Better Inputs 38 From Inputs to Outputs (Attainment and Achievement) 40 From Outputs to Outcomes 43 7 Lessons Learned and Progress Toward the MDGs 43 Progress Toward the Millennium Development Goals 43 Lessons Learned 22 Annexes 49 Annex A: Test Examples 53 Annex B: Budget Analysis 63 Annex C: School Costs 71 Annex D: School-Level Changes in Inputs, Management, and Methods 107 Annex E: Tables Of School Quality Variables 125 Annex F: Variable Definition 129 Annex G: Analysis of Test Scores 145 Annex H: Data on Educational Performance 165 Annex I: School Attainment 177 Annex J: Conditions Attached to Bank Credits 191 Annex K: Education and Welfare Outcomes 197 Annex L: Evaluation Approach Paper 207 Endnotes 219 Bibliography Boxes 5 Box 1.1: Evaluation design: costs and benefits Tables 5 Table 1.1: Coverage of data collection instruments 18 Table 3.1: World Bank support to Ghana's education sector, 1986­2003 20 Table 3.2: Allocation of resources under EdSAC I and II 22 Table 3.3: Allocation of resources under PSD project and BESIP 29 Table 4.1: Activities in World Bank basic education projects 33 Table 5.1: Drop-out rates are low (2003) 35 Table 5.2: Student performance is returning to the levels attained 40 years ago (average test scores of primary school leavers by decade) i v C O N T E N T S 41 Table 6.1: Results from studies of education and social outcomes in Ghana 42 Table 6.2: Percentage reductions in welfare outcomes caused by higher school attainment 44 Table 7.1: Progress toward the education related MDGs Figures 2 Figure 1.1: Ghana's education system went into decline in the mid-1970s starting a slow but steady recovery since the mid-1980s 4 Figure 1.2: How educational inputs affect welfare outcomes 9 Figure 2.1: Restructuring of education system 10 Figure 2.2: Government spending on education has risen: central government education expenditure 12 Figure 2.3: Schools in nearly all areas have more material inputs than before: cluster-level material inputs to school quality 13 Figure 2.4: The quality of school infrastructure has improved in most areas: cluster-level physical inputs to school quality 14 Figure 2.5: The proportion of teachers who are trained rose 14 Figure 2.6(a): Efficiency gains have been realized by increasing the pupil-teacher ratio 15 Figure 2.6(b): But the PTR is too high in the Savannah region (PTR by zone, 2003) 19 Figure 3.1: Bank disbursements on an annual and project basis, FY86­02 32 Figure 5.1: More children are attending school (attendance rates by age group) 32 Figure 5.2: And gender and regional gaps are closing (enrollment differentials, 6­11 year olds) 33 Figure 5.3: More children finish school (primary completion rates ) 34 Figure 5.4: School outputs have improved: test scores plotted against years of schooling 36 Figure 5.5: Criterion Reference Test scores in public schools have risen each year, 1992­2000 v Acknowledgments This Operations Evaluation Department report (MoE) in preparing and implementing the survey has been prepared in response to a request -- KB Danso-Manu and Thomas Coleman from the World Bank's Board of Executive Direc- deserve particular mention -- and to the Bank's tors that the department resume work on impact Ghana education team (Benôit Millot and Eunice evaluation. OED has a long history of conduct- Dapaah) for their cooperation. Mary Esther ing such studies, in which the meaning of impact Dakubu, of the Institute for African Studies, and has been interpreted in various ways. In this Kweku Osam, Linguistics Department, University report, impact is taken to imply a concern with of Ghana, were responsible for the preparation final welfare outcomes, and the attempt to estab- of the local language questionnaire used in this lish the counterfactual to isolate the effects of dif- study. The following World Bank staff provided ferent determinants on those outcomes. The comments: Helen Abadzi, Victoria Elliot, Deon study thus traces the causal chain from inputs Filmer, Nils Fostvedt, Patrick Grasso, Benôit Mil- through to welfare outcomes. The data con- lot, Dean Nielsen, Halsey Rogers, and Yvonne straints and methodological challenges facing Tsikata. William Hurlbut edited the report. Admin- such analysis are well known. OED hopes to istrative support was provided by Pilar Barquero. demonstrate that these challenges can be over- Collaboration with GSS and MoE extended come to illustrate how the Bank's activities con- beyond the survey to data analysis as part of tribute toward the alleviation of global poverty. OED's support for evaluation capacity building. This report has been prepared by Howard Staff from both agencies, with support from White with the assistance of Edoardo Masset. OED, conducted analysis of the data collected Preparation of the study was assisted by Alain for this study, which has been drawn on in the Barbu and Martha Ainsworth and contributions preparation of this report. from the peer reviewers Kwame Akyeampong This study was carried out under the part- and Paul Glewwe and from Dean Nielsen. Thanks nership agreement between OED and the UK are due to the co-operation of Ghana Statistical Department for International Development Service (GSS) and the Ministry of Education (DFID). Director-General, Operations Evaluation: Gregory K. Ingram Director, Operations Evaluation Department: Ajay Chhibber Manager, Sector and Thematic Evaluation: Alain Barbu Task Manager: Howard Nial White v i i F O R E W O R D Foreword Prólogo Avant-propos ~NOL A ENGLISH ESP FRANÇAIS Bank support to basic educa- El apoyo del Banco a la edu- L'appui fourni par la Banque à tion has increased greatly over the cación básica ha aumentado consi- l'enseignement de base a considé- last two decades. The Ghanaian derablemente en las últimas dos rablement augmenté au cours des experience provides a useful test décadas. La experiencia de Ghana vingt dernières années. Le cas du case of the effectiveness of this sup- constituye un buen ejemplo que Ghana offre un exemple notable de port. Since 1986 there have been ten demuestra la eficacia de ese apoyo. l'efficacité de cet appui. Depuis 1986, Bank education sector projects in Desde 1986, el Banco ha financiado 10 la Banque a réalisé dix projets dans Ghana, of which five have directed proyectos del sector de educación le secteur de l'éducation ghanéen, support to basic education: the Health en Ghana, de los cuales cinco han dont cinq étaient axés sur l'ensei- and Education Rehabilitation Project, destinado respaldo a la educación gnement de base : le projet de réha- which supplied school learning mate- básica: el proyecto de rehabilitación bilitation des secteurs de la santé et rials; two education sector adjust- de los sectores de salud y educación, de l'éducation, dans le cadre duquel ment credits in support of the reform que proporcionó materiales didácti- des matériels didactiques ont été program; the Primary School Devel- cos a las escuelas; dos créditos para fournis aux écoles ; deux crédits opment Project; and the Basic Edu- fines de ajuste del sector de educa- d'ajustement à l'appui du programme cation Sector Investment Credit. ción en respaldo del programa de de réforme du secteur de l'éducation ; The main questions addressed reforma; el proyecto de desarrollo de le projet de développement des in the OED study are: (1) what has las escuelas primarias, y el crédito écoles primaires ; et le crédit d'in- happened to educational outputs para inversión en el sector de edu- vestissement dans le secteur de l'en- (school attendance and learning); cación básica. seignement de base. (2) what are the main determinants Los principales interrogantes que L'étude de l'OED visait princi- of those outputs; (3) which educa- se plantean en el informe del Depar- palement à déterminer: 1) l'évolution tional interventions have the largest tamento de Evaluación de Opera- du rendement de l'enseignement and most cost effective impact on ciones (DEO) son los siguientes: (taux de fréquentation scolaire et educational outputs; (4) to what 1) ¿qué ha sucedido con los resulta- apprentissage) ; 2) les principaux extent have Bank-supported activ- dos educacionales (asistencia a la facteurs déterminants de ce rende- ities promoted interventions which escuela y aprendizaje)?; 2) ¿cuáles ment ; 3) les interventions pédago- support improved educational out- son los principales factores deter- giques qui influent le plus, et le plus puts; and (5) how do improved minantes de esos resultados?; 3) ¿qué efficacement au plan des coûts, sur educational outputs support better intervenciones educacionales pro- le rendement de l'enseignement ; welfare outcomes? These questions ducen el mayor impacto en los resul- 4) la mesure dans laquelle les acti- were addressed through a variety of tados educacionales de la manera vités appuyées par la Banque ont means, including a nationwide sur- más eficaz en función del costo?; favorisé des interventions à l'appui vey carried out by OED in collabo- 4) ¿en qué medida las actividades de l'amélioration du rendement de ration with Ghana Statistical Service financiadas por el Banco han promo- l'enseignement ; et 5) comment and the Ministry of Education, Youth vido intervenciones que propician l'amélioration du rendement de and Sports. The survey followed up mejores resultados educacionales?, y l'enseignement contribue à améliorer on a living standards survey con- 5) ¿de qué manera los mejores resul- le bien-être de la population. Plu- ducted in 1988 that included data on tados educacionales propician mejo- sieurs méthodes ont été employées test score outcomes and school res resultados en términos de bienes- à cette fin, et notamment une en- quality. The study is thus in a unique tar? Estos interrogantes se abordaron quête de portée nationale réalisée i x B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S position to analyze school- recurriendo a diversos méto- par l'OED en collaboration level changes over the 15 dos, entre ellos, una encuesta avec le service de statistique year period, 1988-2003. nacional realizada por el DEO et le ministère ghanéen de A major finding of the ~NOL en colaboración con el Servi- l'Éducation, de la Jeunesse et A study is that both the quan- cio Nacional de Estadísticas y des Sports. Cette enquête a tity and quality of schooling el Ministerio de Educación, pris le relais de l'enquête sur ENGLISH ESP have improved over the last Juventud y Deportes de FRANÇAIS les niveaux de vie menée en fifteen years. Enrolments in Ghana. La encuesta comple- 1988, qui avait permis de basic education have increased by mentó otra encuesta sobre los niveles recueillir des données sur les résultats over 10 percent compared to 15 de vida realizada en 1988, que incluía obtenus aux examens et sur la years ago. Moreover, 15 years ago datos sobre los efectos directos en los qualité des écoles. La présente étude nearly two-thirds of primary school puntajes de las pruebas y sobre la a donc l'avantage unique de pouvoir graduates were illiterate, as shown calidad de las escuelas. En conse- analyser l'évolution des paramètres by the fact that they scored two or cuencia, el estudio ofrece una au niveau des établissements less on a simple eight question mul- oportunidad sin igual para analizar scolaires sur une période de quinze tiple choice English test ­ the same los cambios que se han producido ans (1988-2003). as guessing. Less than one in twenty en el ámbito de las escuelas en los L'une des principales conclusions do so badly today. Statistical analy- 15 años transcurridos entre 1988 y de l'étude est que le nombre d'en- sis shows that these improvements 2003. fants scolarisés et la qualité de l'en- in learning outcomes are clearly Una conclusión importante del seignement ont augmenté au cours and strongly linked to better welfare estudio es que tanto la cantidad des quinze dernières années. Le as measured by higher income, bet- como la calidad de la escolaridad han nombre d'inscrits dans l'enseigne- ter nutrition, and reduced mortality. mejorado en los últimos 15 años. ment de base s'est accru de plus de Analysis of the economic rate of En este lapso, la matrícula en la 10 % sur cette période. Par ailleurs, return to education shows that there educación básica ha aumentado más il y a quinze ans, près des deux tiers is no return to simply attending del 10%. Además, hace 15 años casi des individus qui avaient achevé le school, but there is a return to learn- dos tercios de los alumnos que ter- cycle d'enseignement primaire ing achievements. The majority of minaban la escuela primaria eran étaient illettrés : ils ne pouvaient ré- children now benefit from attending analfabetos, como lo demuestra el pondre correctement, au plus, à school, both educationally and eco- hecho de que, en una prueba de seulement deux questions sur huit nomically, which was not the case inglés, respondían correctamente posées dans le cadre d'un test cons- 15 years ago. dos o menos preguntas sencillas de titué de questions à choix multiple-- The data show that gains in edu- selección múltiple de un total de soit le résultat que produirait la cational outputs are directly linked ocho, que era lo mismo que adivinar sélection de réponses au hasard. to better school quality, manifested las respuestas. En la actualidad, Aujourd'hui, ce n'est le cas que de in improved infrastructure and menos de uno entre 20 alumnos ob- moins d'une personne sur vingt. greater availability of school sup- tiene semejantes resultados. Un aná- L'analyse statistique montre que cette plies. Today it is the norm to have lisis estadístico muestra que estas amélioration du rendement de l'en- one textbook per child for math mejoras en los resultados del apren- seignement est manifestement et and English, rather than one per dizaje están clara y firmemente étroitement liée à l'amélioration du class as was common before the vinculadas al mayor bienestar medi- bien-être telle qu'indiquée par l'aug- advent of reforms. Text book pro- do en términos de mayores ingresos, mentation des niveaux de revenus, vision is amongst the most cost mejor nutrición y menor tasa de les progrès en matière de nutrition effective means of improving test mortalidad. El análisis de la tasa de et la baisse des taux de mortalité. scores. School building has con- rentabilidad económica de la educa- L'analyse du taux de rendement éco- tributed to higher enrolments. In ción muestra que el simple hecho de nomique de l'éducation monte que, one area surveyed in which a new asistir a la escuela no produce ren- si la fréquentation de l'école n'a pas, school was constructed enrolments tabilidad, pero que sí es rentable el à elle seule, d'effet sur le plan éco- x F O R E W O R D more than tripled. These nivel de aprovechamiento de nomique, l'instruction acquise gains are impressive, but los estudios. Hoy, por asistir a un impact certain. La there remains substantial a la escuela, la mayoría de los majorité des enfants profitent room for improvement. ~NOL niños se benefician tanto maintenant de l'école, que ce A Enrolments lag in some parts educacional como soit sur le plan pédagogique of the country, and while test económicamente, lo que no ou sur le plan économique, ENGLISH ESP scores are improved they are ocurría 15 años atrás. FRANÇAIS ce qui n'était pas le cas il y a still weak. Los datos muestran que quinze ans. Statistical analysis shows that las mejoras en los resultados Les données montrent que les increased school quality can in turn educacionales están directamente progrès accomplis au niveau du be linked to the Bank's support relacionadas con una mejor calidad rendement de l'enseignement sont which has financed the construc- de las escuelas, expresada en una directement liés à l'amélioration de tion of 8,000 classroom blocks and infraestructura más adecuada y una la qualité des écoles, dont témoi- provided 35 million textbooks over mayor disponibilidad de materiales gnent les progrès réalisés au plan des the last 15 years. Nationally, the escolares. Hoy, la norma es contar infrastructures et des fournitures Bank supported school building con un libro de texto por alumno en scolaires. À présent, il est normal and rehabilitation program has matemáticas e inglés, en lugar de que chaque enfant ait un manuel increased enrolments by around un libro por clase, como solía ocurrir pour le calcul et pour l'anglais alors four percent. Moreover Bank sup- antes de las reformas. El suministro que, avant les réformes, il n'en y port helped sustain initially unpop- de libros de texto es uno de los avait généralement qu'un par classe. ular reforms, demonstrating the medios más eficaces en función de La fourniture de manuels scolaires est efficacy of working in partnership los costos para mejorar los puntajes l'un des moyens les plus efficaces au with a government committed to a de las pruebas. La construcción de plan des coûts d'améliorer les well-defined sectoral strategy. establecimientos escolares ha résultats aux examens. La cons- The downside of this positive contribuido al aumento de la matrí- truction d'établissements scolaires a story is that increased reliance on cula. En una zona estudiada donde permis d'accroître le nombre d'en- community and district financing se construyó una escuela nueva, la fants scolarisés. Dans l'une des means that schools in poorer areas matrícula aumentó más de tres veces. régions couvertes par l'enquête où get left behind, especially those in Estos avances son impresionantes, une nouvelle école avait été cons- off-road rural communities. There pero aún hay mucho margen para truite, le nombre d'enfants scolarisés are still some schools with very seguir mejorando. En algunas partes à plus que triplé. Ces progrès sont poor facilities in which little learn- del país la matrícula va a la zaga, y impressionnants mais il reste encore ing takes places. si bien los puntajes de las pruebas beaucoup à faire. Le taux de scola- The lessons drawn from this han mejorado, aún son deficientes. risation reste insuffisant dans cer- study are: Un análisis estadístico indica que taines régions du pays et, si les · increasing the availability and la mejora de la calidad de las escue- résultats des tests sont meilleurs quality of classrooms and instruc- las está relacionada, a su vez, con el qu'auparavant, ils sont toujours trop tional materials directly con- apoyo del Banco, que ha financiado faibles. tributes to both educational la construcción de 8.000 bloques de L'analyse statistique montre que attainment and achievement; aulas y suministrado 35 millones de l'amélioration de la qualité des éta- · supervision of teachers by the libros de texto en los últimos 15 blissements scolaires peut à son tour head teacher and circuit super- años. En el plano nacional, el pro- être lié à l'appui de la Banque qui a visor matter, as do the teaching grama de construcción y rehabilita- financé la construction de 8000 blocs methods adopted by the teacher, ción de escuelas financiado por el de salles de classe et la fourniture de including the language used as Banco ha permitido un aumento de 35 millions de manuels scolaires au the medium of instruction, so la matrícula de alrededor del 4%. cours des quinze dernières années. efforts should also be made to Asimismo, el apoyo del Banco À l'échelle du pays, le programme de retain trained teachers, to contribuyó a que se mantuvieran las construction et de remise en état des x i B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S improve teacher morale, and reformas inicialmente impo- établissements scolaires to expand in-service train- pulares, lo que demuestra la appuyé par la Banque a ing; eficacia de colaborar con un permis d'accroître le nombre · a class of schools in poorer ~NOL gobierno comprometido a d'enfants scolarisés d'environ A communities are very poorly llevar a término una estrategia 4 %. L'appui de la Banque a resourced, so resources sectorial bien definida. également favorisé la pour- ENGLISH ESP should be directed to the El lado negativo de esta FRANÇAIS suite de réformes au départ most needy schools to over- experiencia satisfactoria es impopulaires, ce qui montre come the bias that results from que la mayor dependencia del bien l'efficacité d'une action menée community-based financing; and financiamiento proveniente de la en partenariat avec un gouvernement · the private sector has been neg- comunidad y de los distritos significa déterminé à poursuivre une stratégie lected, although it is of growing que las escuelas situadas en zonas sectorielle bien définie. importance; attention needs to pobres quedan rezagadas, sobre todo Cette médaille a néanmoins un be paid to it in both government aquellas de las comunidades rurales revers car le recours accru au fi- strategy and Bank support. aisladas. Todavía hay algunas escue- nancement des communautés et des las cuyas instalaciones son muy districts signifient que les écoles precarias y donde se aprende poco. situées dans les régions les plus Las siguientes son las enseñanzas pauvres, et notamment dans les com- que se extraen de este estudio: munautés rurales isolées, restent à la · La mejor calidad y la mayor dis- traîne. Il reste encore des écoles dont ponibilidad de aulas y de mate- les locaux sont très insuffisants et où riales didácticos contribuyen di- l'apprentissage est très limité. rectamente a mejorar el nivel de Il ressort de cette étude que : instrucción y de aprendizaje; · Toute amélioration quantitative · La supervisión de los maestros et qualitative des salles de classe por el director de la escuela y el et des matériels didactiques supervisor de la zona es impor- contribue directement à relever les tante, como los son los métodos niveaux d'instruction et les pro- de enseñanza, incluido el lenguaje grès réalisés ; empleado por los maestros como · La supervision des enseignants medio de instrucción; por ello, par le directeur de l'école et también deberían hacerse esfuer- l'inspecteur d'académie a un effet zos por contratar a maestros bien sur les résultats, de même que preparados, elevar la moral de les méthodes didactiques choisies los docentes, y ampliar las acti- par l'instituteur, et notamment la vidades de capacitación en el langue en laquelle l'enseignement servicio; est dispensé, de sorte que des · Dado que las escuelas de las efforts devraient être déployée comunidades pobres cuentan con pour garder en poste les institu- muy pocos recursos, se deberían teurs formés, améliorer le moral destinar medios a las escuelas des enseignants et développer la más necesitadas a fin de eliminar formation en cours d'emploi ; las diferencias derivadas del · Les écoles situées dans les com- financiamiento basado en la munautés les plus pauvres ont comunidad, y accès à des ressources très insuf- · Se ha dejado de lado al sector fisantes ; il faudrait donc achemi- privado, a pesar de su importancia ner des ressources vers les éta- creciente; se debe prestar atención blissements les plus nécessiteux x i i F O R E W O R D a este sector tanto en la pour éliminer les distorsions estrategia del gobierno como associées au mode de fi- en el apoyo que proporciona nancement communautaire ; ~NOL el Banco. et A ·Le secteur privé n'a guère été pris en compte bien qu'il ESP FRANÇAIS joue un rôle croissant ; il importe donc qu'il reçoive l'attention voulue dans la stratégie de l'État et dans le cadre de l'appui de la Banque. Gregory K. Ingram Director-General Operations Evaluation The World Bank Group x i i i E X E C U T I V E S U M M A R Y Executive Resumen Résumé Summary analytique ~NOL A ENGLISH ESP FRANÇAIS The Millennium Development Los objetivos de desarrollo Les objectifs de développe- Goals aim for universal primary edu- del milenio (ODM) se proponen alcan- ment pour le Millénaire visent à cation by 2015 and gender equality zar la enseñanza primaria universal universaliser l'enseignement pri- in enrolments at all levels of educa- para el año 2015 y la igualdad en la maire à l'horizon 2015 et à réaliser tion. The Education for All (EFA) ini- tasa de matrícula en todos los nive- l'égalité des taux de scolarisation tiative lays out a strategy for les de enseñanza. La iniciativa Edu- des filles et des garçons à tous les achieving these goals. The Bank's cación para Todos (EPT) representa niveaux de l'enseignement. L'initia- own strategy stresses the school una estrategia para el logro de dichos tive « Éducation pour tous » (EPT) quality aspects of EFA, emphasizing objetivos. La propia estrategia del définit la stratégie à mettre en oeuvre the need to focus on preserving learn- Banco insiste en aspectos de la EPT pour atteindre ces objectifs. Dans sa ing outcomes while access to edu- relacionados con la calidad de la propre stratégie, la Banque met cation is expanded. This report enseñanza, con especial hincapié en l'accent sur les aspects qualitatifs assesses the impact to date of the la necesidad de conservar los de l'enseignement, en soulignant la efforts over the past 15 years toward resultados del aprendizaje al mismo nécessité de privilégier la préserva- increasing the quantity and quality tiempo que se amplía el acceso a la tion des acquis scolaires tout en élar- of basic education in one African educación. En este informe se eva- gissant l'accès à l'éducation. Le country, Ghana. lúan los efectos conseguidos hasta la rapport fait le bilan de l'impact à ce Ghana typifies many of the chal- fecha con los esfuerzos de los últimos jour des actions menées pendant les lenges faced by African countries as 15 años por aumentar la cantidad y 15 dernières années en vue du they strive to meet the MDGs. Hav- calidad de la educación básica en renforcement quantitatif et qualitatif ing established one of the best edu- un país africano, Ghana. de l'éducation de base dans un pays cation systems in Africa, the number Ghana es un ejemplo represen- africain, le Ghana. of children attending primary school tativo de muchos de los desafíos L'exemple du Ghana illustre un began to fall in the mid-70s. School con que se encuentran los países grand nombre des défis que les quality was falling with non-salary africanos que tratan de alcanzar los pays africains s'efforcent de relever recurrent expenditures being ODM. Tras haber establecido uno pour atteindre les ODM. Le Ghana squeezed out. Many schools had de los mejores sistemas educativos avait mis en place un des meilleurs no more than one textbook to a de África, el número de alumnos de systèmes d'éducation d'Afrique, class and the majority of primary las escuelas de enseñanza primaria avant que les effectifs de l'ensei- school graduates were illiterate. In comenzó a descender a mediados gnement primaire ne commencent 1986, the government embarked on del decenio de 1970. La calidad de à régresser au milieu des années an ambitious reform program to los servicios escolares disminuyó al 70. La compression des dépenses de increase efficiency by restructuring contraerse los gastos recurrentes no fonctionnement non salariales a pre-university education and salariales. Muchas escuelas no entraîné la dégradation de la qualité increasing cost recovery among sen- tenían más que un libro de texto por de l'enseignement. De nombreuses ior secondary and tertiary students, clase, y la mayoría de los alumnos écoles n'avaient pas plus d'un enabling resources to be re-allo- que terminaban la enseñanza prima- manuel scolaire par classe et la cated to basic education. In the mid- ria eran analfabetos. En 1986, el majorité des élèves terminant l'ensei- 1990s a policy of free, compulsory Gobierno emprendió un ambicioso gnement primaire restaient anal- universal basic education (FCUBE) programa de reforma para aumen- phabètes. En 1986, le gouverne- was launched. Since 1997, the edu- tar la eficacia reestructurando la ment a lancé un programme de cation sector has been decentral- educación preuniversitaria y aumen- réforme ambitieux afin d'améliorer x v B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S ized with increased commu- tando la recuperación de l'efficacité en restructurant nity management and the costos entre los alumnos del l'enseignement pré-universi- introduction of School Man- sector secundario y tercia- taire, en faisant participer agement Committees and ~NOL rio, lo que permitiría reasig- davantage aux coûts les A School Performance Assess- nar recursos a la educación élèves du deuxième cycle de ment Meetings. básica. A mediados del dece- l'enseignement secondaire et ENGLISH ESP The government's efforts nio de 1990 se puso en FRANÇAIS ceux de l'enseignement to improve education have marcha una política de edu- supérieur et en libérant ainsi been supported by the World Bank cación básica gratuita, obligatoria des ressources au profit de l'éduca- and other donors. The Bank's assis- y universal. Desde 1997, el sector de tion de base. Au milieu des années tance began with the Health and la educación se ha descentralizado 90, le Ghana a adopté la politique Education Rehabilitation Project gracias a una creciente gestión consistant à rendre l'éducation de (HERP), which supplied school comunitaria y a la introducción de base universelle, gratuite et obliga- learning materials. The reform pro- comités de gestión escolar y reunio- toire. Depuis 1997, le secteur de gram was supported by two edu- nes de evaluación del desempeño l'éducation a été décentralisé, la cation sector adjustment credits escolar. gestion étant de plus en plus assurée (EdSAC I and II). These adjustment Los esfuerzos del Gobierno por au niveau local avec la création des credits were followed by two invest- mejorar la enseñanza han contado comités de gestion des établisse- ment projects: the Primary School con el apoyo del Banco Mundial y ments scolaires et l'organisation de Development Project and the Basic otros donantes. La asistencia del réunions d'évaluation de l'efficacité Education Sector Investment Credit Banco comenzó con el proyecto de de l'enseignement. (BESIP). The resources provided by rehabilitación de los sectores de la Les efforts déployés par le gou- the Bank have been predominately salud y la educación, que suminis- vernement pour améliorer le sys- used for school building and reha- tró material escolar. El programa de tème d'éducation sont appuyés par bilitation, and textbook supply. reforma contó con el apoyo de dos la Banque mondiale et d'autres Through these five projects close to créditos de ajuste para el sector de bailleurs de fonds. La première 35 million textbooks have been pro- la educación. Estos créditos fueron opération de la Banque a été son vided and 8,000 school pavilions seguidos de dos proyectos de inver- projet de réhabilitation des secteurs constructed. Despite the emphasis sión: el Proyecto de desarrollo de la de la santé et de l'éducation (HERP), on the importance of "software" in enseñanza primaria y el Crédito de qui a fourni des matériels didacti- contemporary education strategies, inversión para el sector de la edu- ques. La Banque a appuyé le pro- the Bank's lending has focused cación básica. Los recursos sumi- gramme de réforme par deux crédits almost exclusively on "hardware" nistrados por el Banco se han utili- à l'ajustement du secteur de l'éduca- and instructional materials (text- zado predominantemente para la tion (EdSAC I et II). Ces crédits ont books and teachers' guides). construcción y rehabilitación de été suivis de deux projets d'investis- The Bank's support helped the escuelas y para la distribución de sement : le projet de développe- government carry out a reform pro- libros de texto. A través de esos ment des écoles primaires et le crédit gram that was resisted by the teach- cinco proyectos se han suministrado d'investissement pour l'éducation ing profession and some segments casi 35 millones de libros de texto, fondamentale (BESIP). Les res- of the population. The strong com- y se han construido 8.000 pabello- sources accordées par la Banque mitment shown by government, and nes escolares. A pesar de la insisten- ont été avant tout utilisées pour la firm actions it took to implement cia en la importancia del compo- construction et la rénovation d'écoles the reforms, demonstrate the high nente "inmaterial" en las actuales ainsi que pour la fourniture de degree of ownership. In that favor- estrategias de enseñanza, el financia- manuels scolaires. Ces cinq projets able context, the Bank's financing miento del Banco se ha orientado ont permis d'acquérir près de 35 reinforced the government's posi- casi exclusivamente a los aspectos millions de manuels et de construire tion, allowed textbooks to be printed "físicos" y al suministro de materia- 8 000 structures temporaires. Malgré for the new syllabus in Junior Sec- les educativos (libros de texto y l'importance accordée aux aspects ondary Schools (JSS) and training guías para el personal docente). non matériels dans les stratégies x v i E X E C U T I V E S U M M A R Y for 40,000 JSS teachers to pre- El apoyo del Banco ayudó d'éducation en vigueur, la pare them for the new sys- al Gobierno a realizar un pro- Banque a centré ses prêts tem. While there was grama de reforma que encon- presque exclusivement sur la opposition to the cost recov- ~NOL tró resistencia entre el perso- construction et l'équipement A ery measures in second-cycle nal docente y algunos ainsi que sur le matériel and tertiary institutions, the segmentos de la población. didactique (manuels scolaires ENGLISH ESP majority of parents were El fuerte compromiso demos- FRANÇAIS et livres du maître) more concerned with the trado por el Gobierno y las Grâce à ces concours, la quality of basic infrastructure, the firmes medidas que adoptó para Banque a aidé le gouvernement à improvement of which--with sub- aplicar las reformas demuestran su mettre en oeuvre un programme de stantial Bank support--helped main- gran identificación con los proyec- réforme auquel s'opposaient le tain the momentum behind the tos. En ese contexto favorable, el corps enseignant et certaines reforms. Alongside formal condi- financiamiento del Banco reforzó la couches de la population. La solide tionality for restructuring the edu- posición del Gobierno, hizo posible détermination du gouvernement et cation system and introducing cost que se imprimieran libros de texto la fermeté des mesures adoptées recovery measures, informal policy para el nuevo programa de escue- pour appliquer les réformes dialogue was greatly facilitated by las secundarias y se impartieran acti- montrent à quel point il s'est engagé the Bank's senior education spe- vidades de capacitación para 40.000 à l'égard du programme. Dans ce cialist being resident in Accra and profesores de ese nivel, como pre- contexte favorable, le financement developing a close working rela- paración para el nuevo sistema. Si de la Banque a renforcé la position tionship with senior ministry offi- bien hubo oposición a las medidas du gouvernement, a permis d'im- cials. The Bank operated behind the de recuperación de costos en las primer des manuels pour le nou- scenes in facilitating donor co-ordi- instituciones del ciclo secundario y veau programme des établissements nation for the education sector, terciario, la mayoría de los padres d'enseignement secondaire du although donor competition meant mostraron más preocupación por la premier cycle ainsi que de former that the anticipated sector-wide ap- calidad de la infraestructura básica, 40 000 enseignants de ces établisse- proach to education in the latter cuya mejora ­con notable apoyo ments pour les préparer au nouveau part of the 1990s failed to materialize. del Banco­ ayudó a mantener el système. Bien que les mesures de This study conducted a survey in impulso en favor de las reformas. recouvrement des coûts dans les collaboration with Ghana Statistical Junto con la condicionalidad formal établissements d'enseignement Service and the Ministry of Educa- para la reestructuración del sistema secondaire du deuxième cycle et tion, Youth and Sports covering de educación y la introducción de d'enseignement supérieur aient 1,740 households, 706 basic schools, medidas de recuperación de costos, suscité une certaine opposition, la and 3,129 teachers. This nationally el diálogo informal sobre políticas se majorité des parents étaient davan- representative survey was carried vio enormemente facilitado por el tage préoccupés par la qualité des out in the same 85 areas of the hecho de que el especialista supe- infrastructures de base dont la réno- country as the education module rior de educación residiera en Accra vation ­ financée en grande partie of the second round of the Ghana y entablara una estrecha relación par la Banque ­ a permis de main- Living Standards Survey in 1988/89, de trabajo con los altos cargos minis- tenir le cap des réformes. Parallèle- enabling a unique and detailed pic- teriales. El Banco contribuyó, entre ment aux discussions formelles sur ture of changes in schools over the bastidores, a facilitar la coordina- les conditions applicables à la 15-year period. These data show ción de los donantes para el sector restructuration du système d'éduca- large improvements in school qual- de la educación, aunque la compe- tion et l'instauration des mesures ity, especially with respect to mate- tencia entre éstos impidió que se de recouvrement des coûts, le rial inputs. For example: hiciera realidad el planteamiento dialogue informel de politique géné- · In 1988, less than half of schools sectorial amplio previsto para la rale avec les autorités a été consi- could use all their classrooms segunda mitad de los años noventa. dérablement facilité par la présence when it was raining, but in 2003 En el marco del presente estudio d'un spécialiste chevronné de over two-thirds can do so. se llevó a cabo una encuesta en l'éducation résidant à Accra, qui a x v i i B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S ·Fifteen years ago over two- colaboración con el Servicio noué d'étroites relations de thirds of primary schools de Estadística de Ghana y el travail avec les hauts fonc- reported occasional short- Ministerio de Educación, tionnaires du ministère. La ages of chalk, only one in 20 ~NOL Juventud y Deportes, en la Banque a oeuvré en coulisse A do so today, with 86 percent que se incluyeron 1.740 pour faciliter la coordination saying there is always hogares, 706 escuelas prima- entre bailleurs de fonds dans ENGLISH ESP enough. rias y 3.129 maestros. Esta FRANÇAIS le secteur de l'éducation. ·The percentage of primary encuesta nacionalmente Cependant, à la fin des an- schools having at least one Eng- representativa se llevó a cabo en las nées 90, l'approche sectorielle envi- lish textbook per pupil has risen mismas 85 áreas del país que el sagée pour l'éducation n'a pas été from 21 percent in 1988 to 72 módulo de educación de la segunda mise en pratique du fait de la con- percent today and for math ronda de la encuesta sobre niveles currence entre bailleurs de fonds. books in Junior Secondary de vida de Ghana de 1988/89, lo Une enquête a été effectuée dans School (JSS) these figures are 13 que permitió obtener una perspec- le cadre de l'étude, en collaboration and 71 percent, respectively. tiva singular y detallada de los cam- avec le Service statistique du Ghana bios registrados en las escuelas en et le ministère de l'Éducation, de la School quality has improved across un período de 15 años. Dichos datos Jeunesse et des Sports, et a porté the country: in poor and non-poor revelan grandes mejoras en la cali- sur 1 740 ménages, 706 écoles de communities alike. But there is a dad de la enseñanza, sobre todo base et 3 129 enseignants. Cette growing disparity within the public en lo que respecta a los insumos enquête, représentative de l'en- school sector. Increased reliance on materiales. Por ejemplo: semble du pays, intéressait les 85 community and district financing · En 1988, menos de la mitad de régions qu'avait couvertes le means that schools in relatively las escuelas podían utilizar todas module éducation du deuxième prosperous areas continue to enjoy sus aulas cuando llovía; en 2003, cycle de l'enquête sur les niveaux better facilities than do those in less la proporción era superior a los de vie du Ghana en 1988­1989. Il well off communities. Future dos tercios. s'en est dégagé un tableau à la fois investments in school quality can- · Hace 15 años, más de dos terce- unique en son genre et détaillé de not be solely demand driven, which ras partes de las escuelas de l'évolution des établissements will tend to favor the better off. enseñanza primaria menciona- scolaires sur une période de 15 ans. Demand-driven programs should ban situaciones ocasionales de Les données recueillies font appa- be complemented by interventions escasez de tiza; hoy, eso sólo raître une amélioration sensible de in disadvantaged schools, which ocurre en una de cada 20, y el la qualité de l'enseignement, notam- can be identified through the annual 86% declara que tienen siempre ment en ce qui concerne les school census conducted as part of un suministro suficiente. moyens matériels disponibles. Par the Education Management Infor- · El porcentaje de escuelas de exemple : mation System (EMIS). enseñanza primaria que tienen al · En 1988, moins de la moitié des The importance of the private menos un libro de texto de inglés écoles pouvaient utiliser toutes sector has increased greatly in the por alumno ha subido del 21% les salles de classe en cas de last 15 years. Close to 20 percent of en 1988 al 72% en la actualidad; pluie alors qu'en 2003 plus des the schools in the survey areas are en cuanto a los libros de mate- deux tiers d'entre elles étaient private, compared to fewer than 5 máticas, en las escuelas de ense- en mesure de le faire. percent five years ago. Private ñanza secundaria los porcentajes · Il y a 15 ans, plus des deux tiers schools are not all elite schools. son el 13% y el 71%, respectiva- des écoles primaires signalaient Many are in relatively poor areas mente. qu'elles manquaient parfois de and many do not perform well on craie alors qu'aujourd'hui ce n'est quality measures. La calidad de la enseñanza ha mejo- le cas que pour une école sur 20 ; Improving school quality has rado en todo el país: en las comu- 86 % des établissements indi- been accompanied by increased nidades pobres y en las menos quent qu'ils en ont toujours suffi- enrolments, which have grown by necesitadas. Pero hay una creciente samment. x v i i i E X E C U T I V E S U M M A R Y over 10 percent over the 15 divergencia dentro del sector · Le pourcentage d'écoles years. By 2000, over 90 per- de la enseñanza pública. La primaires disposant au moins cent of Ghanaians aged 15 mayor dependencia del d'un manuel d'anglais par and above had attended ~ NOL financiamiento comunitario élève est passé de 21 % en A school compared to 75 per- y de los distritos significa 1988 à 72 % aujourd'hui. cent 20 years earlier. In addi- que las escuelas de las zonas Pour les manuels de mathé- ENGLISH ESP tion, drop-out rates have relativamente prósperas con- FRANÇAIS matiques au niveau du pre- fallen, so completion rates tinúan contando con mejores mier cycle du secondaire, les have risen: by 2003, 92 percent of instalaciones que las que se encuen- chiffres sont respectivement de those entering grade 1 complete tran en comunidades pobres. Las 13 % et de 71 %. Junior Secondary School (grade 9). inversiones futuras en calidad de Gender disparities have been virtu- la enseñanza no pueden estar La qualité de l'enseignement s'est ally eliminated in basic enrolments. impulsadas únicamente por la améliorée dans l'ensemble du pays Primary enrolments have risen in demanda, ya que ello tendería a dans les localités pauvres comme both disadvantaged areas and favorecer a quienes se encuentran dans les plus riches. On constate amongst the lowest income groups. en posición más desahogada. Los cependant une disparité croissante The differential between both the programas impulsados por la des établissements publics. En effet, poorest areas and other parts of the demanda deberían complementarse les établissements sont de plus en country, and between enrolments of con intervenciones en las escuelas plus tributaires du financement des the poor and non-poor, have been desfavorecidas, que se pueden iden- communautés locales ou des narrowed but are still present. tificar mediante el censo escolar communes, si bien que ceux qui Rising attainment has been anual realizado a través del Sistema sont situés dans des zones relative- accompanied by higher achievement. de información sobre la gestión de ment prospères continuent à It is no longer the case that most pri- la enseñanza (EMIS). disposer de plus de moyens que mary graduates are illiterate. The La importancia del sector pri- ceux que l'on trouve dans des loca- survey undertaken for this study con- vado ha aumentado enormemente lités plus démunies. À l'avenir, les ducted math and English tests among en los 15 últimos años. Casi el 20% investissements consacrés à l'amélio- 9-55 year olds identical to the tests de las escuelas de las zonas inclui- ration de la qualité de l'enseigne- carried out 15 years ago, enabling a das en la encuesta son privadas, ment ne pourront pas être unique- direct comparison of learning out- mientras que hace cinco años el ment induits par la demande pour comes. Today, less than a fifth of porcentaje era de menos del 5%. No éviter de favoriser les plus nantis. those who have completed grades 3- todas las escuelas privadas son cen- Les programmes induits par la 6 scored two or less out of eight on tros para minorías selectas. Muchas demande devraient être complétés the short English multiple choice se encuentran en zonas relativa- par des interventions au profit des test--the same as guessing--com- mente pobres y muchas tienen nive- écoles défavorisées, que l'on peut pared to nearly two-thirds in 1988. les de calidad más bien bajos. identifier à la faveur du recense- Test scores are significantly higher La mejora de la calidad de la ment annuel des écoles organisé today for both math and English. enseñanza ha ido acompañada de dans le cadre du système d'informa- Children completing the nine years un aumento de la tasa de matrí- tion sur la gestion de l'éducation. of basic education in 2003 scored cula, que ha crecido más del 10% L'importance du secteur privé higher that those with ten years of en 15 años. En el año 2000, más del s'est considérablement accrue au basic education under the old system 90% de la población de Ghana de cours des 15 dernières années. Près 15 years ago. But the shortening of 15 o más años de edad había ido a de 20 % des établissements scolaires post-basic education from seven to la escuela, mientras que 20 años des régions étudiées sont privés three years has had a small adverse antes la proporción era del 75%. contre moins de 5 % il y a cinq ans. impact on learning outcomes Además, las tasas de abandono han Ces établissements privés ne sont amongst secondary graduates. disminuido, por lo que han aumen- pas tous prestigieux. Bon nombre Using the English test results to tado las tasas de terminación de los d'entre eux sont situés dans des measure literacy shows that the lit- estudios: en 2003, el 92% de los régions relativement pauvres et pour x i x B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S eracy rate among those aged alumnos que comienzan el beaucoup d'autres, la qualité 15-24 (one of the MDG indi- primer grado terminan las de l'enseignement laisse à cators) has risen from 49 per- escuelas secundarias (nove- désirer. cent to 68 percent between ~NOL no grado). Las disparidades L'amélioration de la A 1988 and 2003. The increase de género se han eliminado qualité est allée de pair avec in school quality (higher prácticamente en las tasas l'accroissement des taux de ENGLISH ESP scores achieved by those de matrícula en la enseñanza FRANÇAIS scolarisation, qui a atteint enrolled in school) accounts básica. Éstas han aumentado plus de 10 % sur les 15 an- for over half (57 percent) the en las zonas desfavorecidas y entre nées. En 2000, plus de 90 % des increase in literacy, with the remain- los grupos de ingreso más bajo. La Ghanéens âgés de 15 ans et plus der coming from increased quantity diferencia entre las zonas más étaient scolarisés contre 75 % 20 (higher enrolments). pobres y otros lugares del país, y ans plus tôt. En outre, les taux Statistical analysis of the survey entre la tasa de matrícula de la d'abandon ont diminué, ce qui a fait results shows the importance of población pobre y la de mejor posi- croître les taux de réussite. En 2003, school infrastructure on enrolments. ción económica, se ha reducido, 92 % des élèves entrant en première Building a school, and so reducing pero no ha desaparecido todavía. année du primaire achevaient le children's travel time, has a major El mayor índice de terminación premier cycle de l'enseignement impact on enrolments. While the de los estudios ha ido acompañado secondaire (neuf années d'études). majority of children live within 20 de un mayor nivel de instrucción. Ya Il n'y a pratiquement plus de dispa- minutes of school, some 20 percent no puede decirse que la mayoría de rités des taux d'inscription des filles do not and school building has los alumnos que terminan la ense- et des garçons dans l'enseignement increased enrolments among these ñanza primaria sean analfabetos. de base. Les taux de scolarisation au groups. In one area surveyed, aver- En la encuesta llevada a cabo para niveau primaire se sont accrus à la age travel time to the nearest school este estudio se efectuaron pruebas fois dans les régions défavorisées et was cut by 45 minutes with enrol- de matemáticas y de inglés entre parmi les groupes aux revenus les ments increasing from 10 to 80 per- alumnos de 9 a 55 años de edad, plus faibles. L'écart entre les régions cent. In two other areas average idénticas a las realizadas 15 años les plus démunies et le reste du travel time was reduced by nearly antes, lo que hizo posible la com- pays ainsi qu'entre les taux de scola- 30 minutes and enrolments paración directa de los resultados. risation des pauvres et ceux du reste increased by over 20 percent. Reha- Hoy, menos de una quinta parte de la population s'est atténué mais bilitating classrooms so that they de quienes han terminado los n'a pas disparu. can be used when it is raining also grados tres a seis consiguieron La progression de la scolarisa- positively affects enrolments. Com- resultados inferiores a dos sobre un tion s'accompagne d'une amélio- plete rehabilitation can increase máximo de ocho en la prueba breve ration des résultats scolaires. Dé- enrolments by as much as one third. de elección múltiple de inglés ­equi- sormais, la plupart des élèves qui Across the country as a whole, the valente a una adivinanza­, en com- terminent le cycle primaire ne sont changes in infrastructure quantity paración con casi dos tercios en plus analphabètes. Lors de l'enquête and quality have accounted for a 4 1988. Los resultados de las pruebas effectuée dans le cadre de l'étude, percent increase in enrolments son significativamente más eleva- les participants âgés de 9 à 55 ans between 1988 and 2003, about one dos hoy tanto en matemáticas como ont subi des tests en mathémati- third of the increase over that en inglés. Los niños que terminaron ques et en anglais identiques à ceux period. A large part of this improve- los nueve cursos de enseñanza qui avaient été administrés il y a 15 ment can be attributed to the World básica en 2003 consiguieron mejo- ans, ce qui a permis d'établir des Bank, which has been overwhelm- res resultados que los que habían comparaisons directes entre les ingly the main funder of better infra- cursado 10 años de enseñanza acquis scolaires. À présent, moins structure in this period. básica en el antiguo sistema hace 15 d'un cinquième des élèves ayant Learning outcomes depend sig- años. Pero la reducción de la edu- terminé trois à six années d'étude nificantly on school quality, includ- cación postbásica de siete a tres obtient au test abrégé d'anglais à ing textbook supply. Bank-financed años ha tenido un pequeño efecto choix multiple une note égale ou x x E X E C U T I V E S U M M A R Y textbook provision accounts negativo en el nivel de ins- inférieure à deux sur huit ­ for around one quarter of trucción entre los alumnos même résultat que s'ils devi- the observed improvement que terminan la enseñanza naient ­ contre près de deux in test scores. But other major ~NOL secundaria. tiers des participants en 1988. A school-level determinants of La utilización de los Les notes obtenues aux tests achievement such as teach- resultados de las pruebas de sont sensiblement plus ENGLISH ESP ing methods and supervision inglés para medir el nivel de FRANÇAIS élevées aujourd'hui en of teachers by the head alfabetización en las perso- mathématiques et en anglais. teacher and circuit supervisor have nas de 15 a 24 años revelan que éste Les notes des élèves qui avaient not been affected by the Bank's (uno de los indicadores de los terminé le cycle de neuf années interventions. The Bank has not ODM) ha subido del 49% al 68% d'éducation de base en 2003 étaient been heavily involved in teacher entre 1988 y 2003. El aumento de la supérieures à celles des élèves qui training and plans to extend in-serv- calidad de la enseñanza (resultados avaient accompli dix années d'édu- ice training have not been realized. más altos conseguidos por los alum- cation de base dans l'ancien système Support to "hardware" has been nos matriculados) explica más de la 15 ans auparavant. En revanche, le shown to have made a substantial mitad (57%) del incremento del raccourcissement, de sept à trois positive contribution to both attain- nivel de alfabetización, y el resto se ans, du cycle post-éducation de ment and achievement. But when debería al aumento de la cantidad base a eu une légère incidence satisfactory levels of inputs are (mayores tasas de matrícula). négative sur les résultats scolaires reached -- which is still far from the El análisis estadístico de los des diplômés de l'enseignement case for the many relatively resultados de la encuesta demues- secondaire. deprived schools -- future improve- tra la importancia de la infraestruc- Mesuré sur la base des résultats ments could come from focusing tura escolar y su influencia en la tasa des tests d'anglais, le taux d'alpha- on what happens in the classroom. de matrícula. La construcción de bétisation des élèves de 15 à 24 ans However, the Bank's one main effort una escuela, y por lo tanto la reduc- ­ l'un des indicateurs des ODM ­ est to change incentives -- providing ción del tiempo de desplazamiento passé de 49 % à 68 % entre 1988 head teacher housing under the Pri- de los niños, tiene una importante et 2003. Plus de la moitié (57 %) de mary School Development Project in repercusión en el número de alum- l'augmentation de ce taux est attri- return for the head teacher signing nos matriculados. La mayoría de los buable à l'amélioration qualitative a contract on school management niños viven a menos de 20 minutos de l'enseignement (les personnes practices--was not a great success. de la escuela, pero aproximada- scolarisées ont obtenu de meilleures Others, notably DFID and USAID, mente el 20% se encuentra a mayor notes), le reste découlant de l'aug- have made better progress in this distancia, y la construcción de mentation quantitative (un plus direction but with limited coverage. escuelas ha aumentado el nivel de grand nombre d'élèves inscrits). School building and rehabilita- matrícula en esos grupos. En una de L'analyse statistique des résultats tion has been a cost effective means las zonas incluidas en la encuesta, de l'enquête met en évidence l'in- of increasing enrolments. Other el tiempo medio de desplazamiento fluence des infrastructures scolaires activities are most cost effective in a la escuela más próxima se redujo sur le niveau des effectifs. La cons- improving test scores, with textbook 45 minutos, y la matrícula pasó del truction d'une école a pour effet de supply being one of the most effec- 10% al 80%. En otras dos zonas, el réduire la durée du trajet pour les tive. The question for the Bank is the tiempo medio disminuyó casi 30 élèves et a donc un impact impor- balance to be maintained between minutos, y la matrícula aumentó tant sur le nombre d'élèves inscrits. these traditional, proven and still más del 20%. La rehabilitación de las La majorité des enfants habitent à necessary activities and other activ- aulas para que puedan utilizarse une vingtaine de minutes de l'école, ities such as promoting community también en días de lluvia influye mais ce n'est pas le cas pour 20 % management and enhancing the también positivamente en ese des élèves et c'est parmi ceux-ci effectiveness of teaching. mismo sentido. La rehabilitación qu'augmenteront les inscriptions Better education leads to better completa puede aumentar la matrí- suite à la construction d'écoles. Dans welfare outcomes. Existing studies cula hasta un tercio. En el conjunto une des régions couvertes par x x i B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S on Ghana show how edu- del país, los cambios en la l'enquête, la durée moyenne cation reduces fertility and cantidad y calidad de la du trajet jusqu'à l'école la mortality. Analysis of the sur- infraestructura han explicado plus proche a été écourtée vey data shows that educa- ~NOL un aumento del 4% de la de 45 minutes et le taux A tion improves nutritional matrícula entre 1988 y 2003, d'inscription est passé de outcomes, with this effect aproximadamente un tercio 10 % à 80 %. Dans deux ENGLISH ESP being particularly strong for del crecimiento total durante FRANÇAIS autres régions, la durée children of women living in ese período. Una gran parte moyenne du trajet a été poorer households. Regression de esta mejora puede atribuirse al réduite de près de 30 minutes et les analysis shows there is no economic Banco Mundial, que ha sido con effectifs ont augmenté de plus de return to primary and JSS education, gran diferencia la principal fuente de 20 %. La rénovation des salles de but there is a return to cognitive financiamiento de la infraestructura classe pour qu'elles puissent être achievement. Children who attain en ese período. utilisées même en cas de pluie a higher test scores as a result of El nivel de instrucción alcanzado également un effet positif sur le attending school can expect to enjoy depende significativamente de la nombre d'élèves. Une rénovation higher income; but children who calidad de la enseñanza, incluido el complète peut contribuer à une learn little in school will not reap suministro de libros de texto. La augmentation des effectifs pouvant any economic benefit. disponibilidad de libros de texto, atteindre un tiers. Dans l'ensemble The lessons of the Ghana edu- financiada por el Banco, representa du pays, l'amélioration quantitative cation experience are: aproximadamente una cuarta parte et qualitative des infrastructures a · Increasing the availability and de la mejora observada en las pun- permis d'accroître les effectifs de quality of classrooms and instruc- tuaciones de las pruebas. Otros 4 % entre 1988 et 2003, ce qui repré- tional materials directly con- grandes determinantes del nivel de sente un tiers de l'augmentation tributes to both educational instrucción alcanzado, como los totale pendant cette période. Ces attainment and achievement. métodos docentes y la supervisión progrès peuvent en grande partie However, such a "hardware" de los maestros por el director de être mis à l'actif de la Banque approach will become less rele- la escuela y el supervisor de la zona mondiale, qui a été de loin la prin- vant as all schools attain the no se han visto afectados por las cipale source de financement des desired level of quality. Ghana is intervenciones del Banco. Éste no infrastructures. not yet in that position: substan- ha intervenido intensamente en la Les acquis scolaires dépendent tial inputs are still required for the capacitación del personal docente, dans une large mesure de la qualité most disadvantaged schools. y los planes de ampliar la capacita- de l'enseignement, y compris la Even where good school quality ción en los servicios no se han disponibilité de manuels scolaires. is achieved, educational out- hecho realidad. Se ha comprobado Environ un quart de l'amélioration comes, while improved, are still que el apoyo a los aspectos "mate- des notes obtenues aux tests est far from satisfactory. riales" ha contribuido en forma con- attribuable à la fourniture de · The evidence is clear that super- siderable y positiva tanto a la termi- manuels financée par la Banque. vision of teachers by the head nación de los estudios como al nivel Les interventions de celle-ci ne teacher and circuit supervisor de instrucción. Pero cuando se han portent cependant pas sur d'autres matter, as do the teaching meth- alcanzado niveles satisfactorios de facteurs importants indépendants ods adopted by the teacher, insumos ­lo que dista todavía (par exemple les méthodes d'ensei- including the language used as mucho de ser la realidad en muchas gnement et la supervision des ensei- the medium of instruction. Since escuelas relativamente desabasteci- gnants par le directeur d'école et attempts to remove untrained das­ las mejoras futuras podrían ser l'inspecteur) qui influencent les teachers have been unsuccessful, resultado de una mayor atención a résultats scolaires. La Banque n'a and since not all trained teachers lo que ocurre en las aulas. No obs- pas été associée de près à la forma- appear familiar with improved tante, el principal esfuerzo del tion des enseignants et les projets de methods anyway, there is a Banco por cambiar los incentivos formation en cours d'emploi ne se strong case for pushing forward ­ofrecer vivienda a los directores de sont pas réalisés. Il est établi que x x i i E X E C U T I V E S U M M A R Y with in-service training. escuela en el marco del Pro- l'appui dans le domaine de Efforts should also be made grama de desarrollo de las la construction et du matériel to retain trained teachers and escuelas primarias a cambio a eu un effet positif important to improve teacher morale. ~NOL de la firma de un contrato sur la scolarisation et les A Achieving both of these sobre prácticas de gestión résultats scolaires. Toutefois, means better teaching con- escolar­ no fue un gran lorsque l'enseignement dis- ENGLISH ESP ditions, including paying éxito. Otras instituciones, FRANÇAIS posera de tous les moyens teachers on time. sobre todo el DFID y USAID, nécessaires ­ ce qui est loin · The downside of community and han realizado mayores progresos d'être le cas pour de nombreuses district financing of schools is en esta dirección, pero con cober- écoles relativement démunies ­ il that it leads to disparities in tura limitada. faudra peut-être mettre l'accent sur resource availability. There La construcción y la rehabilita- ce qui se passe dans la salle de remains a class of schools in ción de escuelas han sido un pro- classe pour progresser encore poorer communities--particu- cedimiento eficaz en función de los davantage. L'initiative prise par la larly but not only in rural areas-- costos de aumentar la tasa de matrí- Banque pour trouver d'autres inci- which are very poorly resourced. cula. Otras actividades son más efi- tations ­ fournir un logement au Resources should be directed to caces para mejorar las puntuaciones, directeur d'école au titre du projet the most needy schools to over- destacando entre ellas el suministro de développement des écoles pri- come the bias that results from de libros de texto. El problema que maires en échange de quoi celui-ci community-based financing. se plantea al Banco es cómo con- signerait un contrat concernant les School mapping continues to seguir el debido equilibrio entre méthodes de gestion des écoles ­ play an important role, which estas actividades tradicionales, com- n'a pas particulièrement réussi. means that support to EMIS is probadas y todavía necesarias y D'autres bailleurs de fonds, en parti- important. otras iniciativas, como la promo- culier le DFID et l'USAID des États- ción de la gestión comunitaria y el Unis ont eu, dans ce domaine, de While not a major part of this study, logro de una mayor eficacia de la meilleurs résultats, quoique de it is clear that the private sector has enseñanza. portée limitée. been neglected. But it is of growing Una mejor educación da lugar a La construction et la rénovation importance so attention needs to un mayor bienestar. Los estudios rea- d'écoles ont été des moyens effi- be paid to it in both government lizados sobre Ghana revelan de qué caces d'accroître le nombre d'en- strategy and Bank support. manera la educación reduce la fecun- fants scolarisés. D'autres activités didad y la mortalidad. El análisis de sont très efficaces pour relever le los datos de encuestas demuestra niveau des résultats des tests, notam- que la educación mejora los ment la fourniture de manuels sco- resultados nutricionales, siendo este laires. Il s'agit pour la Banque de efecto especialmente notable en el maintenir l'équilibre entre ces acti- caso de los hijos de mujeres que vités classiques, qui ont fait leurs viven en los hogares más pobres. El preuves et sont encore nécessaires análisis de regresión revela que la et d'autres interventions consistant, enseñanza primaria y media no tiene par exemple, à promouvoir la ges- ninguna rentabilidad económica, tion communautaire et à renforcer pero sí la tiene el progreso cognitivo. l'efficacité de l'enseignement. Los niños que alcanzan puntuaciones L'amélioration de l'éducation se más elevadas en las pruebas debido traduit par celle des acquis au plan a la mayor asistencia escolar pueden du bien-être. Les études sur le prever niveles de ingresos más altos; Ghana montrent comment l'éduca- en cambio, los niños que aprenden tion a pour effet de réduire la fécon- poco en la escuela no conseguirán dité et la mortalité. L'analyse des ningún beneficio económico. données de l'enquête indique que x x i i i B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S Las enseñanzas de las ini- l'éducation permet d'amé- ciativas en el terreno de la liorer l'état nutritionnel et educación en Ghana son las que son influence dans ce ~NOL siguientes: domaine est particulièrement A · La mayor disponibilidad y forte sur les enfants dont les calidad de aulas y material mères appartiennent à des ESP docente contribuye directa- FRANÇAIS ménages pauvres. L'analyse mente tanto a prolongar los de régression montre que la años de estudio como a mejorar rentabilité économique de l'ensei- el rendimiento escolar. No obs- gnement primaire et de l'enseigne- tante, este planteamiento resultará ment secondaire du premier cycle menos relevante a medida que las est nulle, alors que les connais- escuelas vayan alcanzando el sances acquises ont une rentabilité. nivel de calidad deseado. Ghana En effet, les enfants qui ont eu de no se encuentra todavía en esa meilleures notes aux tests parce situación: se necesitan todavía qu'ils ont fait des études peuvent considerables insumos para las espérer un niveau de revenu plus escuelas más desfavorecidas. élevé. Par contre, ceux qui ont peu Incluso donde se consigue una appris à l'école n'en tireront aucun buena calidad escolar, los resul- avantage. tados educativos, si bien han Les enseignements suivants mejorado, distan todavía mucho peuvent être tirés des résultats de ser satisfactorios. obtenus par le Ghana au plan de · Las pruebas demuestran clara- l'éducation : mente que la supervisión del per- · L'augmentation quantitative et sonal docente por el director del qualitative des salles de classe centro y el supervisor de la zona et des matériels didactiques a son importantes, y que también une incidence directe sur la lo son los métodos de enseñanza scolarisation et les résultats adoptados por el maestro, inclu- scolaires. Toutefois, cette ap- ida la lengua utilizada como proche centrée sur les aspects medio de instrucción. Como los matériels sera moins pertinente à intentos de prescindir del per- mesure que l'ensemble des sonal docente no capacitado no écoles atteint le niveau de qualité han dado fruto y como no todos souhaité. Le Ghana n'en est pas los maestros capacitados pare- encore là. Les écoles les plus cen estar familiarizados con défavorisées ont encore d'impor- métodos mejorados, hay fuertes tants besoins matériels. Même razones para insistir en la capa- dans les cas où l'enseignement citación en el servicio. Debería est de bonne qualité, les résultats asimismo hacerse lo posible por scolaires, bien qu'améliorés, sont mantener el personal capacitado loin d'être satisfaisants. y mejorar su moral. Para conse- · Il est clair que la supervision des guir esos dos objetivos hay que enseignants par le directeur mejorar las condiciones de ense- d'école et l'inspecteur ainsi que ñanza, incluida la remuneración les méthodes pédagogiques em- puntual del personal docente. ployées, y compris la langue · El aspecto negativo del financia- d'enseignement, ont toutes leur miento escolar por las comunida- importance. Puisqu'il n'a pas été x x i v E X E C U T I V E S U M M A R Y des y los distritos es que da possible d'obtenir le départ lugar a divergencias en la des enseignants non quali- disponibilidad de recursos. fiés, et que de toutes les ~NOL Parece que hay escuelas en façons tous les enseignants A las comunidades más pobres qualifiés ne semblent pas ­sobre todo pero no exclu- maîtriser les nouvelles ESP sivamente en las zonas rura- FRANÇAIS méthodes, l'argument en les­ que cuentan con muy faveur de la formation en pocos recursos. Estos deberían cours d'emploi n'en devient que destinarse a las escuelas más plus convaincant. Il faudrait aussi necesitadas, con el fin de supe- déployer des efforts pour main- rar el sesgo resultante del finan- tenir les enseignants qualifiés en ciamiento de base comunitaria. poste et rehausser le moral du La distribución geográfica de las corps enseignant. La réalisation escuelas continúa desempeñando de ces deux objectifs revient à un papel importante, lo que sig- améliorer les conditions de travail nifica que es preciso mantener el des enseignants et à verser leurs apoyo al Sistema de información salaires à temps. sobre la gestión de la enseñanza · Le financement de l'éducation (EMIS). par les communautés et les muni- cipalités présente un inconvé- Aunque ello no constituye un nient : il crée des disparités de aspecto fundamental del presente ressources. On trouve encore des estudio, es claro que se ha descui- écoles particulièrement démunies dado el sector privado. Pero su dans les zones pauvres ­ surtout, importancia es cada vez mayor, por mais pas uniquement en milieu lo que habrá que prestarle atención rural. Les ressources devraient tanto en la estrategia gubernamen- être affectées aux écoles qui en tal como en el apoyo del Banco. ont le plus besoin pour com- penser le déséquilibre imputable au financement de l'éducation par les communautés locales. L'établissement de la carte scolaire continue à jouer un rôle impor- tant, d'où la nécessité d'appuyer le système d'information sur la gestion de l'éducation. Bien qu'il ne représente pas une partie importante de la présente étude, le secteur privé a manifeste- ment été négligé. Étant donné son importance croissante, il convient de lui accorder la place qui lui revient, tant dans la stratégie du gouverne- Gregory K. Ingram ment que dans le programme d'aide Director-General de la Banque. Operations Evaluation The World Bank Group x x v ABBREVIATIONS AND ACRONYMS AR Attendance rate BE/SIP Basic Education School Improvement Program CIDA Canadian International Development Agency CPI Consumer price index CWIQ Core Welfare Indicators Questionnaire DACF District Assemblies Common Fund DEOC District Education Oversight Committee DFID Department for International Development EdSAC Education Sector Adjustment Credit EFA Education For All EMIS Education management information system EU European Union FCUBE Free, compulsory universal basic education GER Gross enrollment ratio GES Ghana Education Service GLSS Ghana Living Standards Survey GSS Ghana Statistical Service ICR Implementation Completion Report IDA International Development Association JSS Junior Secondary School MDG Millennium Development Goal MOEYS Ministry of Education, Youth and Sports NDC National Democratic Congress NER Net enrollment ratio NGO Non-governmental organization NPP New Patriotic Party PCR Project Completion Report PE Personal emoluments PNDC Provisional National Defence Council PSD Primary School Development PTA Parent Teacher Association PTR Pupil-teacher ratio QUIPS Quality Improvement in Schools OED Operations Evaluation Department SAR Staff Appraisal Report SMC School Management Committee SPAM School Performance Assessment Meeting SSS Senior Secondary School SUR(E) Seemingly Unrelated Regression (Equations) TTC Teacher Training Center UK ODA United Kingdom Overseas Development Administration UNDP United Nations Development Program UNICEF United Nations International Children's Fund URS University Rationalization Study USAID United States Agency for International Development WSD Whole School Development x x v i i 1 Introduction Scope and Purpose of the Study Education and the International Development Agenda "All agree that the single most important key to development and to poverty alleviation is education. This must start with universal primary education for girls and boys equally..." James D. Wolfensohn, January 19991 E ducation is central to international poverty are increasingly incorporated into project design. reduction goals, as reflected in it inclusion This study examines the impact of external sup- in two of the Millennium Development port provided by the World Bank on the achieve- Goals (MDGs): universal primary education and ment of education goals in the case of one gender equality in school enrollments. Support African country, Ghana. for education has also manifested itself in the Education for All (EFA) initiative. Launched at Education in Ghana Jomtien (Thailand) in 1990, the movement gained Ghana's education sector, once one of the most international support through a partnership of respected in Africa, has faced difficult challenges UNESCO, UNICEF, UNDP, and the World Bank, in the past two decades. Basic education was and was given a further boost by the Dakar expanded following independence, as was the World Education Forum in April 2000.2 case in neighboring countries. But by the mid- The World Bank's Education Sector Strategy seventies the number of children attending pri- (World Bank 1999) is complementary to the mary school in Ghana started to fall (Figure framework of action adopted at Dakar, with a 1.1). In 1975 there were over 2.3 million children stress on quality. The goal is to "ensure that, by in primary school: this figure had fallen by over 2015, every boy and girl in the developing world one million by the early eighties. Quality as has access to and completes a free and com- well as quantity suffered. Non-salary recurrent pulsory primary education of good quality"3 expenditures were squeezed out; falling real The emphasis on quality has led to a focus on wages and frequent late payments demoralized issues such as parental and community partici- the teaching force. The majority of primary pation and improved teaching methods, which school graduates were illiterate. Meanwhile, 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S G h a n a ' s e d u c a t i o n s y s t e m w e n t i n t o d e c l i n e i n t h e m i d - 1 9 7 0 s ... F i g u r e 1 . 1 s t a r t i n g a s l o w b u t s t e a d y r e c o v e r y s i n c e t h e m i d - 1 9 8 0 s ( i n d e x o f t o t a l p r i m a r y e n r o l l m e n t s ) Source: World Bank World Development Indicators, 2003 government spending was excessively oriented tion Sector Improvement Credit. Since 1986 the toward the tertiary sector. Bank has lent close to $260 million in support The government embarked on an ambitious of education in Ghana, accounting for close to reform program in 1986 to restructure pre-uni- half of all external assistance to the sector. versity education and introduce greater cost recovery at secondary and tertiary levels. These Evaluation Questions: What Explains changes, together with the higher economic Educational Performance? growth resulting from the economic reform pro- Many factors contribute to educational outcomes. gram, led to a steady recovery in the number of Access to, and quality of, school facilities are children attending school (Figure 1.1). While in important. But so is the home environment, principle there has always been free universal including the importance parents put on their primary education in Ghana, fees charged at the child's education and the time the child has to local level have been one factor in restraining spend working in household or other enter- enrollments. Free compulsory universal basic prises. To what extent can improved educational education (FCUBE), introduced in 1996, aimed outputs and the resulting welfare outcomes be at eliminating these fees.4 Since 1997 education attributed to the changes in school inputs and services have been decentralized, including the management and the support the Bank has pro- introduction of School Management Commit- vided to these? The challenge for this report is tees and School Performance Assessment Meet- to answer the following five questions: ings for increased community management and · What changes have occurred to school attain- accountability. ment and achievement (education outputs),5 The World Bank has supported these devel- including the MDG indicators of completion opments through 10 projects, of which 5 have and gender equality in enrollments, in Ghana assisted basic education: the Health and Edu- since the start of reforms in 1986? cation Rehabilitation Project, the Education Sec- · What are the determinants of changes in toral Adjustment Credits I and II, the Primary basic educational outputs for children of basic School Development Project, and Basic Educa- school age in Ghana? 2 I N T R O D U C T I O N · Which education interventions have the great- determinants of school attainment and achieve- est impact on the determinants of educa- ment, a review of the Bank's portfolio of edu- tional outputs? cation investments and an inception visit to · What has been the role of the Bank and Ghana.6 Data collection focused on a household other external donors in promoting education and school survey replicating the data collected interventions that result in improved school in the second round of the Ghana Living Stan- attainment and achievement? dards Survey (GLSS2) in 1988/89. Interviews · Do improved school attainment and achieve- were carried out in 84 of the 85 clusters covered ment support better welfare outcomes as by the 1988 survey, including 1,740 households, captured in the MDGs, such as lower child 704 schools and 3,129 teachers (Table 1.1 and mortality, better nutrition, and reductions in Box 1.1). Achievement tests were taken by over income-poverty? 3,500 people. Quantitative data collection was supple- This report is thus primarily concerned with mented by fieldwork in Ghana interviewing key determining changes in education outputs and informants, visits to district offices and to schools outcomes and attributing, or not as the case in urban and rural areas. Existing reports on edu- may be, any improvements to activities sup- cation in Ghana and other donor projects were ported by the Bank and other agencies. The collected and a review carried out of the rele- study does not therefore cover the same ground vant World Bank project files. as a country sector study, and is less concerned with topics of relevance, efficiency, and efficacy, Outline which are usually central to OED's approach. Chapters 2 and 3 describe the inputs into The framework for this analysis is provided Ghana's education system. The former reviews by Figure 1.2. The ultimate concern is improved the changes that have taken place in basic welfare, which is the outcome of, among other education since reforms were initiated in 1986 things, the higher level of education outputs. and chapter 3 reviews the Bank's education Cognitive development is an output of the edu- portfolio together with that of other donors. cation system. Producing this output requires that Chapter 4 brings these two strands together, students attend and stay at school, with the identifying the impact of the Bank and other quality of the output depending on the quality external agencies on education policies and of the various inputs, both hard (physical infra- basic education outputs. The analysis of the structure) and soft (learning environment and determinants of educational attainment and methods). The World Bank has supported the achievement in Ghana is presented in Chapter inputs into the educational process both directly 5, linking these determinants to the interven- (e.g., financing school building) and indirectly tions supported by the Bank and others. Chap- (support to policy reform). ter 6 goes on to examine the relationship between education outputs and welfare out- Overview of the Study comes. Chapter 7 concludes with lessons learned and implications for future support to Approach education. The technical annexes present more The evaluation framework for this study was detailed analysis to substantiate the arguments developed through a literature review of the made in the report. 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S H o w e d u c a t i o n a l i n p u t s a f f e c t F i g u r e 1 . 2 w e l f a r e o u t c o m e s Education budget Education budget Policy reform Policy reform Househol d Household Community Community allocation plus allocation plus (influenced by (influenced by characteristic characteristicss characteristic characteristicss donor programs donor programs policy dialogue) policy dialogue) INPUTS Community Community involvement involvement School School Learning Learning Teacher Teacher Teacher Teacher School School rehabilitation, rehabilitation, material s materials training training pay pay supervision supervision construction, facilitie construction, facilities s Teacher Teacher Teacher Teacher moral e morale quality quality INTER- MEDIATE OUTPUTS Teaching and learning environment Teaching and learning environment (school quality) Education Output Education Outputss Attainment/Quantity(completion) Attainment (completion) Attainment (completion) OUTPUTS Achievement/Quality Achievement Achievement (learning gains, testtest scores) (learning gains, scores) Welfare outcomes Welfare outcomes OUTCOMES 4 I N T R O D U C T I O N C o v e r a g e o f d a t a c o l l e c t i o n T a b l e 1 . 1 i n s t r u m e n t s 1988 2003 Clusters 170 whole survey 84a 85 education module Household survey Householdsb 3,190 1,740 Individualsb 14,924 7,191 Testsc 3,718 3,582 School survey Primary 286 417 Middle/JSS 233 289 Teachers 0 3,129 a. One cluster was no longer inhabited in 2003 b. In 1988 approximately half of these numbers were in clusters covered by the education module c. Number of people taking the Raven's test. E v a l u a t i o n d e s i g n : c o s t s a n d B o x 1 . 1 b e n e f i t s The main data collection instrument for the impact test scores, allowing analysis of the factors behind evaluation was the re-surveying of households and changes in school attainment and achievement. The schools in the 85 communities covered in the educa- data also allow analysis of changes in school-level tion module of the 1988/89 Ghana Living Standards Sur- inputs over the period of the study. vey (GLSS2). The total cost of this survey, from the The quantitative data were supplemented by qual- household and school listing through to data entry and itative information from fieldwork and a review of the cleaning, was US$263,000. Household surveys typi- literature. Two trips were undertaken during which cally cost US$100 per household, suggesting that the key informant interviews were carried out with gov- survey of 1,740 households accounted for just less than ernment officials at central and district level, repre- one half of the total survey budget. The school and sentatives of the teachers' union and NGOs. Schools teacher questionnaires (the latter including application were visited outside of Accra, meeting with teachers, of the English, math and local language tests) cost just parents and pupils in different parts of the country. under US$50 each. The data were collected by Ghana Statistical Ser- The unique feature of the study design was the vice, working in collaboration with the Ministry of application of the same English and math tests used 15 Education, who advised on the design of the school and years earlier. The nationally representative random teacher questionnaires and provided enumerators for sample of people taking the same test over this period the school survey. Data analysis undertaken by both gives a firm basis for mapping progress in learning these organizations has been incorporated into the outcomes. The study is unusual in linking data on both report. school and household characteristics with student 5 2 Changes in Basic Education Since the 1980s In 1986 the Government of Ghana embarked on an ambitious program of education reform. The main element of this reform -- the restructuring of the education system -- was successfully carried out. A second stage of reforms to decentralize the school system is still underway. There have been substantial improvements in school-level inputs to the education system over the past 15 years. The availability of material inputs -- chalk, textbooks,and desks -- has risen markedly. The development of school infrastructure has kept pace with growing enrollments and has improved in quality. Some neg- ative aspects can be noted. First, the percentage of trained teachers has fallen and in-service training remains scant. Second, teacher absenteeism has risen and the quality of teaching and supervision of teachers by head teachers and circuit supervisors is uneven. Hence, while physical and mate- rial inputs have improved, there is less strong evidence of improved teach- ing within schools. Finally, the reliance on community financing widens the gap between well-resourced schools in affluent communities and badly resourced ones in the poorest areas. If education targets are to be met, attention necessarily needs to be paid to the latter group where enrollments, attainment, and achievement are lowest. The Education System Before 1986 cent by 1983. This spending was skewed in two From a position of having been one of the best ways: (i) large subsidies to secondary and terti- in Africa, Ghana's education system was by the ary levels, meaning that only one-third of edu- early eighties in the throes of a crisis with sev- cation expenditure went to the primary sector, eral underlying sources. Prolonged economic and (ii) recurrent expenditure was almost entirely decline prior to the introduction of reforms had absorbed by wages of teaching and non-teach- led to a compression of educational expenditure ing staff, a problem exacerbated by the large from 6.4 percent of GDP in 1976 to just 1.5 per- number of "ghost workers."7 The physical qual- 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S ity of basic education facilities was very poor; system for all children, and O and A-levels schools structures were dilapidated and many replaced with the secondary certificate. lacked chairs, desks, and even chalk. The struc- · Improve the teaching/learning process by ture of the system was inefficient, the school year increasing school hours and the quality of was short, as was the school day at just four teachers, including the phasing out of hours. However, pre-university education could untrained teachers (i.e., those with no formal extend to a staggering 17 years. teaching qualification, often called "pupil At independence in 1957 Ghana's education teachers"). system consisted of six years of primary educa- · Increase cost recovery at the secondary and tion, followed by five years in secondary leading tertiary levels. to O-levels, and a further two years ("sixth form") · Make educational planning and management to the A-levels required for university admission. more effective. Entrance to secondary was by means of a com- mon entrance exam. However, the majority of stu- All four elements of the reform program were dents went from primary to middle school for up implemented and most sustained. to four years.8 Many children from better-off The restructuring was phased as shown in Fig- homes attended private primary schools and ure 2.1. The last cohort of middle school students were able to skip the middle school stage: in 1985, was admitted in 1986/87; when they graduated 30 percent of secondary entrants were from pri- in 1989/90 middle schools ceased to exist. Mean- vate primary schools, most of the rest coming from while, the first JSS cohort was admitted in the fourth year of middle school. Thus the major- 1987/88, so that schools simultaneously con- ity went through a 6,4,7 system, totaling 17 years tained both JSS and middle school students for of pre-University education. three years. The first JSS students took the new Between Independence in 1957 and the mid- ninth grade Basic Education exam at the end of 1980s there were nine attempts at educational the 1989/90 academic year, the successful can- reform, starting with the Botsio Commission in didates forming the first cohort to enter the new 1960.9 Most important was the 1972 Dzobo Com- SSS system in January 1991, completing in mission whose report, "The New Structure and December 1993. Content of Education," formed the basis for the From 1987 to the mid-90s there was a sub- 1986 reforms. The Dzobo Commission recom- stantial drop in the percentage of untrained mended that middle schools be replaced with teachers from 50 to 20 percent in primary Junior Secondary Schools (JSS), with a stronger schools, and 35 to 14 percent in JSSs (see below). vocational orientation, following which 118 JSSs This decline has been reversed in recent years, were created on an experimental basis. However, particularly in primary schools. The reversal is opposition from the middle classes and the partly because of the growth of the private teaching profession, including the Ghana Edu- school sector, in which most teachers are cation Service (GES) created in 1974, forestalled untrained. extension of the reforms. But 14 years after the The reform also included three forms of cost Dzobo Commission the PNDC government recovery: (1) increased charges for textbooks, finally implemented the proposed changes. (2) removing boarding and feeding subsidies for secondary and tertiary institutions, and (3) removal The 1986 Reform Program of student subsidies for tertiary education. Charges The education reform program adopted in 1986 for textbooks were raised to cost-recovery levels, sought to: with the intention of setting up a revolving fund. · Change the structure of the school system by However, the fund was not well managed (e.g., replacing the 6,4,7 system with 6,3,3, short- BESIP SAR: 12) and did not become a basis for ening pre-university education from 17 to 12 sustainable textbook supply, which has contin- years. Middle schools were to be replaced by ued to be supported by external donors. More- JSSs, which would be an integral part of the over, textbook charges were abolished for primary 8 C H A N G E S I N B A S I C E D U C AT I O N S I N C E T H E 1 9 8 0 S F i g u r e 2 . 1 R e s t r u c t u r i n g o f e d u c a t i o n s y s t e m 86/87 87/88 88/89 89/90 90/91 91/92 92/93 93/94 94/95 Last Last cohort Middle cohort graduate Middle schools closed school admitted First cohort First JSS not take BE JSS cohort yet begun exam (end admitted of year) SSS Last Form Last cohort Old (old 1 cohort complete system system) admitted Form 5 finished SSS First cohort First cohort (new New SSS system not yet begun admitted complete System) (Jan 1991) (Dec 1993) students in 1995. Boarding and feeding subsidies FCUBE and Decentralization were removed first through an increase in the Once the new structure was in place, sector parental contribution followed by the removal of policy was outlined in 1996 in the strategy doc- the government's contribution. Removal of sub- ument "Free Compulsory Universal Basic Edu- sidies for university students was delayed for cation (FCUBE)," which stated the government's some time on account of its political unpopular- commitment "to making schooling from Basic ity manifested in frequent protests but eventually Stage 1 through 9 free and compulsory for all proceeded with some modifications. The Uni- school-age children by the year 2005... [and] to versity Rationalisation Study was completed in improving the quality of the education services March 1988 and in September of that year the gov- offered" (GoG [MoE], FCUBE, April 1996: 1). In ernment announced its intention of removing principle, this statement did not signal any subsidies from the tertiary sector. However, two change in policy, but was one of the periodic months later, the government proposed a loan attempts by government to abolish unsanctioned scheme for tertiary students that contained an ele- fees that proliferate at the local level.10 ment of subsidy. While cost recovery has not been The significance of FCUBE was twofold: (1) as extensive as at first envisaged, parental con- it provided a basis for a coordinated sector pro- tribution to costs for senior secondary and terti- gram providing a framework for donor support ary education have become an established part to education; and (2) it laid out the institutional of the education system in Ghana. and other measures to support the nascent Improvements to educational planning cen- decentralization program, including increased tered around strengthening the Ministry of Edu- community participation in school manage- cation. The Policy, Budgeting, Monitoring, and ment. Evaluation division was created (with UNDP FCUBE had three costed components: technical support partly financed by the Bank). · Improving quality of teaching and learning, The divisions of Curriculum Design and Devel- consisting of (1) the review and revision of opment, and Supplies, were both relocated in teaching materials in line with a revised, more the Ministry from GES. A school mapping was focused, syllabus, (2) new measures on carried out in 1987 (under the project prepara- teacher incentives, including teacher prizes tion facility from the Bank) and educational sta- and teacher housing in rural areas, and (3) a tistics began to be collated on a systematic basis shift to in-service teacher training using dis- since 1988 resulting in the later establishment of tance learning materials. the Education Management Information System · Strengthening management at both central (EMIS) with World Bank and USAID support. and district level; and 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S · Improving access and participation, though, school boards, which already existed at second- inter alia, facility construction and rehabil- ary level. Annual School Performance Assess- itation and pilot scholarship schemes to ment Meetings (SPAMs) were to be key events at encourage girls' participation at primary which the SMC, teachers, and the rest of the level. community could meet together. Armed with data from the most recent Performance Monitor- In addition to the above, measures were to be ing Test (PMT), which ranks each school in the undertaken to ensure the financial sustainabil- district based on test results, they are to prepare ity of the education sector. a plan to improve school performance. There has been progress regarding the first two elements of the first component, but the shift to Budget in-service teacher training has not really taken off. In the early 1980s government expenditure fell The GSS/OED survey data show that less than 5 below 10 percent of GDP. At around one fifth percent of basic school teachers receive such of total spending, education spending was just training on a regular basis. The largest changes 1.5 percent of GDP. From 1984­87 education have taken place with respect to decentralization. expenditure grew rapidly for three reasons: The Local Government Acts of 1988 and 1993 education claimed a growing share of a budget shifted responsibility for the administration of that was a growing share of a growing GDP education to the districts, and the 1995 Ghana (Figure 2.2).11 Real expenditure grew at an Education Service Act created District Education average rate of 35 percent a year over this Oversight Committees (DEOCs) as well as com- period, and the share of education spending in munity-level School Management Committees GDP more than doubled (see Annex B). The (SMCs). Whereas Parent Teacher Associations growth in real spending exceeded the growth (PTAs) had been expected to play a largely rev- in student numbers so real spending per stu- enue raising function, the SMCs were to act like dent also increased. G o v e r n m e n t s p e n d i n g o n e d u c a t i o n F i g u r e 2 . 2 h a s r i s e n : c e n t r a l g o v e r n m e n t e d u c a t i o n e x p e n d i t u r e 70 6. 0 60 Real expenditure 5. 0 (left scale) 50 4. 0 (billions) 40 Share of GDP GDP cedis (right scale) of 3. 0 Real 30 Percent 2. 0 Percent; 20 Share of government expenditure 1. 0 10 (left scale) 0 0. 0 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Source: MoE and GSS data 1 0 C H A N G E S I N B A S I C E D U C AT I O N S I N C E T H E 1 9 8 0 S These increases were sustained into the early every classroom today compared to 78 per- 1990s. Real spending and education's share of cent 15 years ago. GDP continued to rise, passing 5 percent in · Fifteen years ago over two-thirds of primary 2001. Total spending on education rose faster still schools reported occasional shortages of as a result of (1) increased parental contributions, chalk, but today 86 percent say there is always (2) the growth of the private sector in the 1990s, enough. (3) substantial donor support to the sector since · The percentage of primary schools having at 1990, and (4) the introduction of GETFund in least one English textbook per pupil has risen 2001 (see footnote 11). However, the share of from 21 percent in 1988 to 72 percent today; education in central government spending has and the percentage of JSS having at least one fallen, though partly mitigated by the one-third math book per pupil has risen from 13 to 71 of Common Fund resources that are spent by percent.14 District Assemblies on schools.12 The share of basic education in total educa- Despite this overwhelmingly positive message tion spending has fluctuated around an average there remain some schools, most typically in poor of 67 percent over the period 1989­2001, being rural areas, in which conditions, while improved, above this average in the early 1990s and again remain poor (see below). in the most recent years. No substantial reori- entation of the education budget appears to have Material Inputs taken place in the period since 1989.13 However, The four materials inputs for which data can be at an average for the period of 42 percent, the compared between 1988 and 2003 -- availabil- share going to primary education is above the ity of chalk, math and English book availability, one-third reported for the early 1980s, showing and desks15 -- were combined into an index of that the shift took place during the major expan- material inputs.16 For each of these four variables sion in funding in the mid-1980s. there has been a highly significant improve- ment in the level of inputs at both primary and School-Level Inputs JSS level, and the index shows an improvement School quality can be measured by four differ- in nearly every area surveyed (Figure 2.3). ent types of inputs: Figure 2.3 shows the cluster level average of · Material inputs, such as chalk and textbooks the material input index for 1988 and 2003, cal- · Physical inputs, such as classrooms and black- culated separately for primary and middle/JSS. boards In each graph the clusters have been ranked · Teachers according to the value of the index in 1988, so · School management. that the clusters with the schools with the fewest material inputs in that year appear to the left of Data were collected on each of these aspects in the scale. Where the line for 2003 lies above that both 1988 and 2003 school surveys and are for 1988 there has been an increase in the mate- used here to show how the situation in schools rial input index for that cluster. Two points jump has changed over time (Annex D provides a out from these graphs: more detailed analysis). · There has been a substantial increase in the level of material inputs across the country, Physical and Material Inputs especially in primary schools. In only two The main message from the GSS/OED school clusters (which had the maximum value of 1 survey is the overwhelming improvement in in 1988) has the level of material inputs physical and material inputs. For example: declined at primary level. For middle/JSS · In 1988 less than half of schools could use all there have been an improvement in all but 9 their classrooms when it was raining, but in of the 76 clusters 2003 over two-thirds can do so. · The improvement has been greatest the lower · 94 percent of schools have a blackboard in the initial level of the index, meaning that the 1 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S c h o o l s i n n e a r l y a l l a r e a s h a v e m o r e m a t e r i a l i n p u t s t h a n b e f o r e : F i g u r e 2 . 3 c l u s t e r - l e v e l m a t e r i a l i n p u t s t o s c h o o l q u a l i t y (a) Material Primary (a) Material Middle/JSS Source: GLSS2 and GSS/OED school survey clusters in which schools that were the most proportion with a blackboard and the quality of deprived in 1988 have seen the largest those boards, the presence of a library and own improvements in material inputs. water supply. Two of these have not improved (number of classrooms and library) for either type The share of private schools in the sample of school, one (library) has not for primary increased from 5 to 20 percent between 1988 schools, and another (classrooms that can be used and 2003. But the increase in school quality does when raining) for middle/JSS. The lack of change not result from the better quality of private of there being sufficient classrooms shows that schools. Figure 2.3 also shows the material classroom building has kept pace with growing input index for 2003 calculated for public student numbers. The number of classrooms has schools alone. In general this line is not far increased, but been matched by more students. removed from the overall cluster average. Overall, there has been a significant increase in Indeed it is above it, indicating that public the index of physical inputs (Figure 2.4). schools have a higher level of material inputs Figure 2.4 shows the change in physical inputs than do private ones, in 22 of the 41 clusters that in the same way as Figure 2.3 showed material have private schools. When the changes in the inputs. Well over half of the clusters have expe- index and its components are calculated for rienced an overall improvement in physical inputs. public schools only these changes all remain Once again, although private schools per- significant at the 1 percent level (Annex E). form better in some respects, their increase does not account for the improvement in school qual- Physical (building) Inputs ity that has taken place. In 2003, private schools Physical inputs have also increased, though to a had superior inputs with respect to the per- lesser extent. The indicators used are the ade- centage of classrooms that could be used when quacy of the number of classrooms, the pro- raining and having their own water supply. portion that can be used when raining, the They also had slightly better average quality 1 2 C H A N G E S I N B A S I C E D U C AT I O N S I N C E T H E 1 9 8 0 S T h e q u a l i t y o f s c h o o l i n f r a s t r u c t u r e h a s i m p r o v e d i n F i g u r e 2 . 4 m o s t a r e a s : c l u s t e r - l e v e l p h y s i c a l i n p u t s t o s c h o o l q u a l i t y (a) Physical Primary (a) Physical Middle/JSS 1. 0 1.0 0. 8 0.8 0. 6 0.6 index index input input 0. 4 0.4 Physical Physical 0. 2 0.2 0.0 0. 0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Cluster Cluster 1988 2003 2003 (public only) 1988 2003 2003 (public only) Source: GLSS2 and GSS/OED school survey chalkboards, although the difference is not quite schools in off-road rural communities. Such statistically significant. There is no difference schools have difficulty in attracting teachers18 and with respect to having sufficient classrooms, parents who can ill afford any cash contributions. chalkboards, or a library. There is growing dichotomy within the public sector between these schools and those of rel- Allocation of Material and Physical Inputs atively more affluent parents in urban areas.19 There were biases in the allocation of material inputs in 1988. By 2003 these had been elimi- Teachers nated, with the exception of desks. But there has The number of primary teachers rose from 47,900 been a continued bias against poorer areas in in 1980 to 84,400 in 2001. For JSS these num- the distribution of physical inputs. The source bers are 22,500 and 43,000 respectively. In line of these differences is the basic school financ- with the reform program, the proportion of ing and distribution system. Chalk and text- teachers who are trained rose, particularly in pri- books are supplied centrally through GES to their mary school reaching nearly 80 percent from a district offices, which distribute them to schools. low of just 50 percent (Figure 2.5). But this This system was not functioning in 1988 owing trend was reversed in the mid-1990s, so that to lack of materials and transport. But today it today only 60 percent of primary teachers are works so as to ensure sufficient supplies in the trained. This is partly because of the growth of majority of schools.17 However, infrastructure private schools, which typically do not require is the responsibility of districts, which may also their teachers to be trained. In the 2003 GSS/OED supply desks, with additional support from the school survey 87 percent of public basic school PTAs. Schools in wealthier districts will benefit teachers were trained, whereas just 12 percent from both higher levels of district support and of teachers in private schools had teacher train- higher parental contributions, resulting in dis- ing. A second explanation is that trained teach- crepancies in resource availability. The worst- ers are taking study leave and not returning to resourced schools are "bush schools" that is basic education -- either joining the adminis- 1 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T h e p r o p o r t i o n o f t e a c h e r s w h o a r e F i g u r e 2 . 5 t r a i n e d r o s e ... a n d t h e n f e l l a g a i n 100 90 80 70 60 50 trained 40 30 Percent 20 10 0 84/85 88/89 92/93 96/97 00/01 Primary JSS Source: MoE data tration, teaching in secondary school, or leaving (defined as less than 20) compared to 18 per- education altogether.20 cent 15 years ago (Figure 2.6(a)). But more Ghana has a low pupil­teacher ratio (PTR) schools suffer from having too few teachers, compared to other countries. Official policy is defined as a PTR of 50 or more, especially in to raise the PTR in the interests of efficiency. The northern regions where 54 percent of primary increase in the average ratio for primary schools schools had a high PTR (Figure 2.6(b)). from 30.6 to 36.0 between 1988 and 2003 there- The quality of teachers is measured by fore represents greater efficiency. Only 13 per- teacher training and the methods they employ, cent of primary schools now have a low PTR including student supervision and time on task. E f f i c i e n c y g a i n s h a v e b e e n F i g u r e 2 . 6 ( a ) r e a l i z e d b y i n c r e a s i n g t h e p u p i l ­ t e a c h e r r a t i o F i g u r e 80 70 1988 2003 60 50 schools 40 of 30 Percent 20 10 0 Low Medium High Source: GLSS2 and GSS/OED school survey 1 4 C H A N G E S I N B A S I C E D U C AT I O N S I N C E T H E 1 9 8 0 S B u t t h e P T R i s t o o h i g h i n t h e F i g u r e 2 . 6 ( b ) S a v a n n a h r e g i o n ( P T R b y z o n e , 2 0 0 3 ) 80 70 Low Medium s 60 High school 50 of 40 Percent 30 20 10 0 Coastal Forest Savannah Source: GLSS2 and GSS/OED school survey The proportion of trained teachers has fallen and (Annex D). In-service training also helps. Head the provision of in-service training is unsatis- teacher supervision of teachers has a signifi- factory: 35 percent of the 3,129 teachers inter- cantly positive impact on the use of improved viewed in the GSS/OED survey stated that they methods, as does the teacher having direct con- received no teacher in-service training at all in tact with the circuit supervisor. the past year. Of those who have received such In 2003, data were also collected on the fre- training, 70 percent have received it three times quency with which teachers set homework, a year or less. Less than 3 percent of teachers look at and assess students' work for both math benefit from in-service training once a month and English. Homework is set at least once a or more. week by over 95 percent of teachers for math Teaching methods can be broken down into and English, and work assessed with the same use of improved methods, the frequency with frequency by half the teachers surveyed. Less which teachers set homework and time on task. attention is paid to local languages with home- In 2003, teachers were asked three questions to work set frequently by 80 percent of teachers and test their familiarity with improved teaching far fewer assessing work on a regular basis. On methods. About a third of teachers use a student- average one-third of the time in the classroom centered learning approach and use simulations is spent on task based on a narrow definition, (role play) on a regular basis, though about a but 72 percent using a broader definition. There fifth of the latter could not explain them prop- is considerable variation around these averages. erly. About one-fifth use cues to help explain dif- Teacher absenteeism has increased over the ficult words. In summary, improved teaching past 15 years. In 2003 nearly 13 percent of teach- methods are far from unknown, but not wide- ers had been absent in the past month for rea- spread, being utilized by a minority of teachers. sons other than sickness,21 compared to just Trained teachers are significantly more likely to over 4 percent in 1988. Correspondingly, more use improved methods than untrained ones, schools are affected by absenteeism today than although there is not a significant difference in 1988. Fifteen years ago 85 percent of schools between teachers who have received university- did not suffer at all, whereas this figure has level teacher training and those trained by TTCs now fallen to 61 percent. There is a substantial 1 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S difference between public and private schools: vate schools receive infrequent supervision visits, 80 percent of private schools have no problem compared to only 7 percent of public schools. with absenteeism, compared to not much more Virtually all schools have a PTA. Over 99 per- than half of public schools. Absenteeism is cent of public basic schools had them in 2003, greater in rural areas, probably for the follow- as did over 95 percent of private schools. How- ing reasons: (1) teachers may live in town some ever, it is not the mere presence of a PTA that distance from the school and suffer transport will make the difference, but the extent to which problems, (2) they have to travel to town once it provides support to the school. There is con- a month to collect their pay, which they may find siderable variation in the extent to which PTAs is not yet there, and (3) rural teachers attend to have provided support to schools and in the their farming activities.22 More generally, absen- value of parents' monthly contributions. Econo- teeism is linked to low teacher morale and poor metric analysis shows that the level of commu- working conditions, in particular not receiving nity support to the school through the PTA is pay on time (see Annex D). closely related to the community's economic well-being. On average, schools in the better-off School Management areas among the survey areas can expect to The focus on software rather than hardware receive 10 times as much in PTA contributions means an increased focus on issues of school as can schools in the least well off areas.23 The management. At the school level the majority of actual range is far higher, with several schools head teachers are actively involved in the dif- not requesting a PTA contribution compared to ferent types of supervision. Notably, less than 5 the maximum of 150,000 cedis per child ($20) percent of teachers say that the head teacher (Annex C). does not look at their lesson plans on a regular School Management Committees are also basis. However, fewer than half say that the widespread, being present in over 80 percent of head actually discusses the lesson plan with the schools surveyed.24 However, in only half of them. And, while the large majority of schools schools had SMCs met in the preceding month have visits from the circuit supervisor, nearly half or provided support in the past year, and in even (44 percent) of teachers have no direct contact fewer helped the school in dealings with out- with him or her. This latter finding confirms side agencies. The lower prevalence of SMCs that from the evaluation of the Primary School than PTAs is largely explained by the fact that Development project in the late 1990s, which they are not required at private schools: over 90 found that many circuit supervisors merely percent of primary schools have SMCs. checked staffing numbers and enrollments rather Virtually all public primary schools (92 per- than observing teachers in the classroom or cent) have had a School Performance Assess- other activities that might positively affect learn- ment Meeting, at 98 percent of which an action ing (Fobih et al. 1999: p.33). plan was agreed. However, knowledge of SMCs In 1988, circuit supervisors visited schools just and the SPAM among households is far less over once every two months on average. By 2003 common than the school-level data suggests it the mean number of visits rose from 6 to 9 a year should be and participation rates correspond- for primary schools and a bit less for JSS. There ingly low. Only 6 percent of households say that is little variation between areas of the country as someone attended a SPAM at their child's to the frequency of visits, but 45 percent of pri- school. 1 6 3 The Bank's Education Portfolio in Ghana Five Bank projects have provided support to basic education: the education component of HERP, EdSAC I and II, PSD, and BESIP. The money from all of these projects has been largely devoted to hardware and instructional mate- rials, mainly school building and rehabilitation, and textbooks and school furniture. The Bank's contribution to more recent changes in school man- agement, such as school management committees and increased emphasis on in-service training, has been rather more limited. Overview complemented by two investment projects: (1) Following the engagement of the World Bank Community Secondary School Construction with Ghana's Economic Recovery Program in (CSSC, 1991­95) to create the extra capacity at 1983 the Bank undertook initial education sec- the secondary level, especially in under-served tor analytical work in 1984. Further discussions areas, expected to be generated by the reforms; the following year resulted in a preparation mis- and (2) Tertiary Education (1992­98), which sion in September 1985, which proposed a sec- was left out of EdSAC II because of its political tor approach.25 Prior to the sectoral adjustment sensitivity. credit a $0.3 million project preparation facility Two further projects also supported formal enabled some planning activities, such as a basic education: Primary School Development school mapping exercise as well as purchase of (PSD, 1993­98) and the Basic Education Sector essential school materials such as pens and pen- Improvement Project (BESIP, 1996­02). Basic cils, and further emergency support was provided education can include adult education programs, under the Health and Education Rehabilitation and these have been supported by two additional Project (1986­91, see Table 3.1). The first two projects: Literacy and Functional Skills (1991­95) education projects were sector adjustment cred- and National Functional Literacy Program its EdSAC I (1986­91) and II (1990­94), the first (1992­98) and National Functional Literacy. A of which was the first SECAL for education by final project focused on vocational skills train- the Bank and foreshadowed the later adoption ing as part of a broader informal sector project. of the sector approach more generally.26 These Over the period 1986­2002 the Bank dis- two projects, which were directly linked to the bursed $260 million to projects supporting reforms described in the previous chapter, were Ghana's education sector, an average of $17 1 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S W o r l d B a n k s u p p o r t t o G h a n a ' s T a b l e 3 . 1 e d u c a t i o n s e c t o r , 1 9 8 6 - 2 0 0 3 Budget Outcome rating Project IDA Total Bank rating OED Approved Closed Health & education rehabilitation 18.0 18.1 Not S 1/86 12/91 (HERP) o/w education component 6.1 rated Education sector adjustment 38.3 45.5 S S 12/86 12/91 Education sector adjustment II 53.2 S MU 5/90 12/94 Community secondary school construction 14.7 19.6 S MS 6/91 6/95 Literacy and functional skills 27.8 S S 3/92 12/97 Tertiary education 44.8 51.0 MS1 MS 10/92 9/98 Primary school development (PSD) 53.2 56.6 U MU 6/93 12/98 Basic education (BESIP) 47.9 241.6 S S 6/96 12/02 Vocational skills and informal sector 5.8 U U 3/95 6/01 National functional literacy program 23.7 S n.a. To close 12/04 Key: S = Satisfactory, MS = Moderately/marginally satisfactory; MU = Moderately/marginally unsatisfactory; U = Unsatisfactory. 1/ The rating system used by the Bank's operational staff allow for only satisfactory or unsatisfactory ratings, but in this case it was stated that the project outcome was "barely satisfactory". Source: World Bank project documents. million a year, peaking at nearly $40 million in els; and (d) ensure the financial sustainabil- 1995 when five projects were disbursing simul- ity of the new system. taneously (Figure 3.1). This study focuses on the five credits that have Use of Funds supported formal basic education: HERP, EdSAC Since both EdSAC I and II were budget support, I and II, Primary School Development, and the it may seem that the attribution of the funds to Basic Education Sector Investment Project. specific expenditure items is not worthwhile. However, although the funds were budget sup- The Sector Adjustment Credits: port, they were disbursed against expenditures EdSAC I and II on a schedule prepared against a positive list and agreed on a tranche-by-tranche basis with the Objectives Ministry of Education,27 with agreement on the The objectives of the two EdSACs were linked overall education budget as one of the condi- to the reforms begun in 1986: tions for tranche release. Procurement was car- · EdSAC I (1986­91): (a) Change the structure ried out by a Project Management Unit (PMU) of the education system; (b) improve peda- located in the ministry. The Bank's task manager, gogic efficiency and increase access; (c) who was based in Accra from late 1987, was improve budgeting procedures and effect involved in monitoring procurement decisions cost savings and cost recovery measures. and procedures. Hence it makes more sense to · EdSAC II (1990­94): (a) Complete the restruc- say that the Bank financed these items than turing of the school system to a 12-year cycle; would usually be the case with budget support (b) extend the reform to senior secondary (with the partial exception of an early dis- education; (c) consolidate the basic education bursement, which was retroactive finance for reforms so that primary and JSS leavers textbooks the government had already printed).28 acquire the cognitive skills needed to take Under the two EdSACs just under 30 percent advantage of education offered at higher lev- of IDA funds were used for school building and 1 8 T H E B A N K ' S E D U C AT I O N P O R T F O L I O I N G H A N A B a n k d i s b u r s e m e n t s o n a n a n n u a l F i g u r e 3 . 1 a n d p r o j e c t b a s i s , F Y 8 6 ­ 0 2 40 35 30 National functional literacy* 25 BESIP* Vocational skills * 20 PSD Tertiary education project Literacy and functional skills 15 CSSC EdSAC II 10 EdSAC I HERP 5 0 1987 1989 1991 1993 1995 1997 1999 2001 Source: World Bank World Development Indicators, 2003 rehabilitation, and a similar amount for school ment's resource allocation procedures. Such furniture and equipment (Table 3.2). Other effects are frequently negative as donor systems expenses includes items such as vehicles, so can impose large transaction costs on the bor- about two-thirds of total funds went on "hard- rower (see White and Dijkstra 2003: Chapter ware." The next largest item was teaching mate- 12). However, in the case of EdSAC they appear rials, which includes both the development and to have been positive, with the Ministry of printing of teacher materials and textbooks. A Education requesting more frequent supervi- relatively small amount (only 2 percent under sion. Procurement procedures are prone to EdSAC II) was spent on teacher training. How- corruption and bureaucratic delay. The presence ever, under each of EdSAC I and II over 20,000 in the field of the Bank's task manager facili- JSS teachers received teacher training to orient tated timely and detailed comments on bidding them to the new system. Under EdSAC II 24 per- procedures and familiarized ministry staff with cent of the funds were allocated to primary competitive tendering procedures, which were education and another 15 percent to JSS.29 Nearly adopted for all ministry procurements in the 95 percent of the funds spent in the primary sec- early 1990s (an EdSAC II condition). These tor were used for civil works (mainly the con- interventions would have been unnecessary if struction of school pavilions), as were just under procurement had been problem free, but it one-third of the funds benefiting JSSs with most had not. Or they could have been costly if of the remainder (61 percent) for textbooks. there were many donors imposing different The system of administering budget sup- procedures. But the Bank was the only donor port has systemic effects, that is, the way in of substance prior to 1990 (other donors sup- which the donor's system for managing the ported education by co-financing EdSAC I). aid inflow (i.e., procurement, monitoring, and The first significant bilateral support was reporting requirements) affects the govern- USAID's $35 million Primary Education Pro- 1 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S A l l o c a t i o n o f r e s o u r c e s u n d e r T a b l e 3 . 2 E d S A C I a n d I I EdSAC I EdSAC II US$ US$ millions Percent millions Percent IDA Total IDA Total IDA School building and rehabilitation 11.3 17.5 29.4 38.3 15.2 28.5 Teacher training 3.4 3.4 8.8 7.5 1.1 2.1 Teaching materials 8.1 8.1 21.0 17.8 12.0 22.5 School furniture and equipment 9.4 10.3 24.4 22.5 18.2 34.1 Other expenses 6.3 6.3 16.4 13.9 6.8 12.8 Total 38.6 45.7 100.0 100.0 53.3 100.0 Source: calculated from project data in Bank implementation completion reports. ject (PREP, 1990­05) of which $32 million was was met each time. This condition was kept budget support channeled through the PMU for EdSAC II with the added requirement that responsible for EdSAC procurement. actual expenditure should be in line with the budget and that the share of basic education Conditionality and Reforms should stay at least its 1989 level (62 percent). Both credits consisted of three tranches. EdSAC The second of these targets was met, but the II conditionality was simplified to the same set first was not with larger amounts going to ter- of six conditions for each tranche release, but tiary and vocational training, resulting in ten- complicated by the introduction of perform- sions between the Bank and government, ance indicators that were used to judge progress the latter accusing the Bank of bringing up but which did not have legal status. This ambi- arbitrary conditionalities.30 guity might explain some of the tensions that · Cost recovery: (1) Boarding and feeding sub- emerged between GoG and the Bank in the sidies: Government subsidy to feeding and later period (see below). boarding costs for secondary school students Although the government implemented an were reduced -- although not to the level impressive range of reforms, this does not mean required by the Bank. This partial slippage that the conditionalities attached to the two was allowed to pass, and the subsidies were EdSACs were problem free. Far from it. The later removed altogether. The condition to policy conditions under the two credits can be eliminate the feeding and boarding subsidy divided into four areas (the conditions are listed at the tertiary level was postponed and the in full in Annex J): Bank accepted the introduction of a subsi- · Restructuring: these conditions matched the dized student loan scheme. Delays in com- government's own timetable for the intro- pletion of the University Rationalization Study duction of the new system and were met (URS) and its implementation were major accordingly. Bank reviews of the adjustment factors behind the delayed release of the sec- credits noted an initial lack of trained teach- ond and third tranches of EdSAC I. (2) Book ers for all subjects and teaching materials for charges were introduced and increased at a JSS. However by June 1987, 7,000 JSS teach- rate to ensure full cost recovery, with the ers had been trained and the army mobi- proceeds paid into a revolving fund. In Jan- lized to distribute textbooks, indicating that uary 1995, the charge was abolished at pri- whatever shortcomings there were did not mary level. The revolving fund did not arise from government complacency. become a basis for sustainable textbook pur- · Budget: EdSAC I required that there should chases with textbook supply continuing to be agreement on the education budget, which depend on external finance. 2 0 T H E B A N K ' S E D U C AT I O N P O R T F O L I O I N G H A N A · Staffing: (1) A payroll audit was undertaken and improving the quality of education: to eliminate ghost workers, with 5,722 ghosts · PSD: The overall goal of this project was to removed from second-cycle institutions by increase learning achievements and enroll- January 1987. A freeze on new posts was ments in primary schools throughout the breached in 1988 with GES employment country. In order to accomplish this, the proj- increasing by close to 7,000 (to a total of ect had the specific objective of increasing the 158,102) as a result of the hiring of untrained amount and improving the quality of instruc- middle school leavers, in contravention of tional and learning time in primary schools, both the condition and government's own particularly as far as 1,983 of the least well- policy that no new untrained teachers should endowed primary schools are concerned. be hired. Extensive staff cuts brought GES · BESIP was intended to help the Government employment down to 146,000 by mid 1990. of Ghana to implement FCUBE, specifically During EdSAC II negotiations a ceiling of aiming to (a) improve the teaching process 153,000 was agreed,31 which was as good as and learning outcomes; (b) strengthen man- kept until the third tranche32 was approved, agement of the basic education system but breached shortly thereafter rising to through better planning, monitoring and eval- nearly 155,000 in 1994. The Bank wrote to uation by MOE/GES at central, regional and the government asking it to keep to "its district levels, and by promoting active ceiling" of 153,000 but had no leverage since involvement of communities in the manage- the funds were disbursed. (2) Freezing the ment of schools; (c) improve access to basic size of GES at a time of growing enroll- education, especially of girls, the poor and ments had the desired effect of increasing the other disadvantaged segments of the popu- pupil­teacher ratio. At senior secondary level lation; and (d) ensure financial sustainability the condition that class sizes for optional of the Government program for basic edu- subjects be at least 20 was not met (the cation over the longer term. Bank had originally proposed 25), the Bank responding merely by requesting gov- Project Components and Use of Funds ernment to send a further instruction to For PSD two main areas of activity were iden- schools to reduce the number of options tified: taught in schools missing the target. The · Policy and management changes: (1) in- letter was sent but not complied with by all creased instructional time, (2) reducing stu- schools. dent fees and levies, (3) improve skills and motivation of head teachers, (4) community The EdSAC II targets not contained in the legal involvement in selection of head teachers, covenant included designing and implementing (5) orientation of district officials and com- a plan for in-service training, the introduction of munity leaders, (6) support to school super- the new circuit supervisor system and the intro- vision, and (7) school mapping. duction of performance testing (the Criterion Ref- · Investment in physical infrastructure: (1) con- erence Test). Each of these things was done, struction of classrooms, (2) construction of though the funding and technical support to do head teachers' housing, (3) provision of so was provided by USAID rather than the Bank. roofing sheets. Communities were to be responsible for building the external walls Investing in Basic Education: ("cladding") for pavilions constructed by the PSD and BESIP project. Objectives These activities were to be carried out in the Both the Primary School Development Project 1,983 most deprived schools. This number of (PSD) and Basic Education Sector Improvement schools covered by the project was later Program (BESIP) emphasized increasing access increased to 2,178 in response to pressure from 2 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S MPs. In the mid to late 1990s there were approx- implemented. For example, schools did not pro- imately 11,200 public primary schools, mean- vide the required length of instructional time, ing that about 20 percent of all schools received community involvement was negligible other support from the Primary School Development than in some SIF schools, and there was little project. Eighty-five percent of PSD funds were impact from orientation and training of officials, spent on civil works (Table 3.3), constructing community leaders, and teachers. PSD's main a school pavilion (a cement floor and roof with achievement was the provision of physical infra- girder supports as shown in the cover photo of structure. this report) and house for the head teacher in The BESIP SAR stated that "despite increased each beneficiary school. In return for the accom- resource inputs and enrollments, the reform modation the head teacher was to sign an movement has had very limited success so far agreement with PTA and DEOC on holding in improving the quality of teaching and learn- meetings out of school time, providing teacher ing outcomes" (p.5) so that "more attention has training, community relations, and attending to be paid to software" (SAR: 14). However, training. The communities were to sign contracts most Bank resources for the project were to clad the pavilion (i.e., construct external devoted to hardware and instructional materials, walls) within six months of completion. especially following the Mid-term Review when The project was restructured at the mid-term the project was restructured to focus on three review to better support the FCUBE, incorpo- components: (1) civil works, (2) textbook sup- rating the Education Management Information ply, and (3) EMIS. As a result, the allocation to System (EMIS), provision of teaching materials, civil works and goods increased by about $19 a Schooling Improvement Fund (SIF), and an million, giving rise to the large share devoted to information, education, communication (IEC) hardware and materials in project expenses: program. These changes made little difference $15.4 million of the total budget of $47.9 were to the allocation of funds. Less than 4 percent spent on civil works and a further $25.8 million was spent on training materials and training. on goods (presumably mostly textbooks and While components may be important even if furniture, though also including vehicles and they do not have much money spent on them, other equipment). Approximately one-third was the Bank's implementation completion report spent on school building and rehabilitation, just rated the project as unsatisfactory noting that over another third on textbook supply and just many required reforms had been only partially under 10 percent on school furniture: in total 77 A l l o c a t i o n o f r e s o u r c e s u n d e r P S D T a b l e 3 . 3 p r o j e c t a n d B E S I P Primary School Development BESIP US$ Percent US$ Percent School building and rehabilitation 38.0 67.1 16.3 34.2 Head Teachers' Housing 10.5 18.6 0.0 0.0 Training materials 2.1 3.7 1.3 2.7 Training 1.3 2.7 Teaching materials 0.0 0.0 2.0 4.3 School furniture 0.0 0.0 4.2 8.9 Textbook supply 0.0 0.0 16.4 34.3 Other expenses 6.0 10.6 6.2 13.0 Total 56.6 100.0 47.8 100.0 Sources: World Bank project documents. 2 2 T H E B A N K ' S E D U C AT I O N P O R T F O L I O I N G H A N A percent of the project budget was spent on each district (totaling 330 schools) by the end hardware and materials inputs. of 2004. Of more significance is the EU MPP, which has financed some 1,500 classroom The Role of Other Donors blocks (a block usually contains three class- The Bank's role should be put in perspective rooms) around the country. Regarding text- against the contribution of other donors. The books, the main input was $10 million from main agencies active in basic education are USAID under PERP in 1991. To some extent USAID and DFID, and some support to school these books would have replaced those sup- building through the EU's Micro-projects Pro- plied by HERP, which had become worn out, gram (Annex B). The largest contributions -- though USAID also supplied books for social USAID through QUIPS and DFID through science and sciences, which had been largely Whole School Development (WSD) -- have neglected in HERP procurements. In summary, complemented rather than competed with the the other donors active in basic education have Bank's inputs since there has been a focus on by and large not overlapped in supplying the software (district management, community par- items provided by the Bank. Where they have ticipation, teacher training, etc.). QUIPS contains overlapped the contribution of other donors are small grants to beneficiary schools that have not insubstantial, but are on a smaller scale than been used for construction in many cases, but those of the Bank. the program will only cover three schools in 2 3 4 The Bank's Impact on Education Policies and Outputs The Bank has provided both finance and policy support to the education sec- tor over the past 15 years. Despite the clear government ownership of the edu- cation reform program, the Bank can be argued to have played an important role in its implementation. While critics argue that the reforms were carried out too quickly, it is at least as plausible that delays would have resulted in failure. The Bank's policy conditions underpinned the reforms, its finance helped them be realized, assisted by technical support. Over the past 15 years the Bank has provided close to 35 million textbooks and financed the con- struction of 8,000 school pavilions, being the main provider of both these types of support. The Bank and Educational Reform books would be available, and that schools Critics of the World Bank argue that it forces would not be closed because there was no reform on unwilling countries.33 The evidence food. in this case suggests a contrary position. Here · Aspects of the reforms were not favored by was a set of reforms the government wished to the Bank, notably the increased vocational- undertake, which it used Bank assistance to ization of the curriculum. The Bank accepted carry out. The strong domestic ownership was this policy in order to retain its position sup- shown by: porting the education sector, staff saying that · The strong domestic dynamic to education they saw no alternative at the time in view policy issues.34 The reforms were not of the strong position taken by the senior designed by World Bank staff, but based on MOE official. The government wished to the recommendations of the 1972 Dzobo expand senior secondary education more Commission, restated by the Education Com- rapidly than the Bank thought wise. In the end mission of 1985. In the view of the Bank task the Bank supported the Community Sec- manager of the time the reforms were ondary Schools Project, for which it had tried accepted by the government in 1973, and the unsuccessfully to find another donor. As a final Bank merely helped bring them back to life example, the Bank quickly accepted the gov- and simplify the curriculum, ensure that ernment's view that it was politic to bring in 2 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S a subsidized student loan scheme at tertiary were enjoying the bulk of government spend- level once feeding subsidies were eliminated. ing. Moreover, the benefits of economic reform · The government, including the President, would take some time to reach rural residents publicly reaffirmed their commitment to the outside of the cocoa producing region, so reforms and made the case for them to the expanding educational provision and improving public. The reform program was first quality would build support for reform more announced in national radio and TV broad- generally.36 casts in October 1986. The more difficult The political commitment of PNDC is clear "second phase" dealing with second-cycle from the decisive manner in which reform was and tertiary reforms beginning in 1990 handled. As the reforms got underway, key civil received renewed support from the Presi- servants were replaced and a new PNDC Sec- dent. Furthermore, the Minister of Finance fre- retary for Education appointed. She was joined quently spoke of the need for cost recovery by another prominent PNDC member as Deputy in health and education, this case being Minister who was to remain in the post for nine repeated in the 1987 National Program for years and is widely recognized to have been the Economic Development. central figure in steering through the reforms. A · The government took several decisive steps second Deputy Minister, responsible for higher in support of the reform prior to it being education, was in place for seven years. This launched and to ensure it was followed team moved to end corruption, weeding out through, including substantial increases in ghost workers (by the end of 1986, 5,722 ghost education spending. workers had been removed from second-cycle · Finally, the reforms made sense given the institutions alone) and regaining control of edu- political position of the ruling Provisional cational policy from GES by relocating three National Defense Council (PNDC) at the time. divisions (Supplies, Curriculum Research and Development, and PBME) within the ministry. Why did PNDC embrace reforms that had proved To circumvent possible delays from GES oppo- politically difficult for well over a decade, and sition the army was mobilized to distribute text- how was it able to successfully implement them? books to the new JSS schools. Student unrest was The opposition to the reforms came from the also tackled with a firm hand, with arrests and middle class elite, which were not PNDC's polit- closure of the universities -- these strong moves ical base. During Rawling's first year in power did not threaten the government's popularity he directly attacked wealth and implemented since the universities were widely regarded as stringent anti-corruption measures. His subse- elitist (Tsikata, 2001: 73 and Nugent, 1996: 118). quent adoption of the liberalization agenda can Three roles can be identified for the Bank in be attributed to the fact that it would undermine supporting the reforms: (1) money, (2) techni- rent seekers to the benefit of the wider popu- cal assistance, and (3) donor mobilization. lation.35 PNDC was not overly concerned about · The role of money. Unlike some macroeco- middle-class opposition. Students were a special nomic adjustment programs in which there case, since Rawlings did have support in the stu- may be nothing obvious to finance, the edu- dent-based June the Fourth Movement (JFM). But cational reforms in Ghana required financial JFM was on the left wing of the party, which was support. The main requirements were teacher alienated by the adoption of an IMF program in training in the new curriculum, textbooks and 1983. The loss of this support base, and Rawl- other teaching materials for that curriculum, ings' populist inclination, implied a need to and school building and rehabilitation for the broaden PNDC's appeal. Reform of the educa- expansion of enrollments. The Bank sup- tion sector was an obvious candidate. The chil- ported each of these activities. Even with the dren of the rural poor were either not attending growth in spending on education, the gov- school at all, or attending second or third-rate ernment was covering not much more than facilities, whereas the children of the better off salaries, so the Bank funding paid for many 2 6 T H E B A N K ' S I M PA C T O N E D U C AT I O N P O L I C I E S A N D O U T P U T S of these requirements. Bank assessments of the Sustainability of Reform impact of the EdSACs argued that the local cost The restructuring of education is well entrenched. financing provided by these credits was cen- The 1996 manifesto of the main opposition tral to the implementation and sustainability party, the National Patriotic Party (NPP),39 crit- of the reforms, allowing them to be com- icized the restructuring on the grounds that it was pleted before opposition could mount. Pay- rushed and done with inadequate consultation.40 ing for activities that facilitated growing No intention was announced to reverse to the enrollments helped build support for the gov- reforms. To the contrary the origin of the JSS sys- ernment's educational policy. Money also tem was traced to Busia's Progress Party41 with contributed to the restructuring of the ministry, a commitment to ensure access up to JSS 3 for which helped government to increase its con- all Ghanaians. While cost recovery measures trol over GES. It used the reform program were criticized, the proposed policies put the implementation and the EdSAC credit to carry state's role as paying teachers' salaries, with out this agenda, such as the relocation of key communities responsible for much else. Since activities within the Ministry supported by NPP came to power in 2000 there have been no Bank technical assistance. signs of a policy reversal. The recent education · Technical assistance. The Bank financed tech- sector strategy makes no reference to changing nical assistance for studies that played a role the structure of the system and reaffirms the in planning, policy, and implementation. The decentralization measures introduced in the sec- project preparation facility financed both the ond half of the 1990s. school mapping exercise and the University Rationalization Study (URS). Technical inputs Donor Coordination on textbook design were provided, as well as Donor support for EdSAC I had the features of more day-to-day support on managing pro- a sector-wide approach (SWAp) a decade before curement. There was also informal influence the term came into usage and the Bank can on these various aspects, in particular from claim some credit for achieving this degree of the Bank's education specialist resident in donor coordination. A pre-condition for a SWAp Accra.37 For example, he commented on is a clearly defined, government-owned sector drafts of the URS before it was officially sub- strategy, which the first wave of educational mitted to the Bank. His role in budget mon- reforms clearly were.42 The Bank was active in itoring was mentioned in Chapter 3. promoting donor coordination by facilitating · Mobilizing donor support. The Bank helped donor discussion on the sector and mobilizing present the government's case to outsiders. co-financing for EdSAC I. An important stage in This was the first sectoral adjustment credit this process was a donor meeting held in Vienna in education and the Bank was undoubtedly in September 1987.43 Having a clear lead donor instrumental in coordinating donors in a way helps take a sector program forward and the so as to support the reforms. Bank occupied this position, though it proba- bly helped that the meeting was jointly spon- While the reforms were government-driven, the sored by UNICEF, since donors may have Bank did have some influence on the shape of resisted being directed solely by the Bank. the program. For example, the government was Donor coordination continued in the early persuaded to restrict vocational training at JSS 1990s, helped by the fact that the PMU for level to an introduction to tools. But there were EdSAC, which received technical assistance other areas where the Bank was the one to give from the Bank, assumed responsibility for the ground. For example, the Bank went ahead and management of all external projects,44 though supported senior secondary schools in a more not their policy functions, which rested in the full-fledged program that proved a costly failure, ministry. In 1990 the Bank shared with USAID with $18 million wasted on workshop equipment consultants who designed the latter's new proj- that was not used.38 ect for basic education. The Bank went on to 2 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S play a role in setting up a donor forum for the strategy: despite the fact that a strategy already education sector in August 1994. existed and that no other donors were invited Similar efforts were made to ensure a coordi- to the meeting.46 From this time onward first nated approach to the Basic Education Sector DFID and then USAID went their own way with Investment Project (BESIP), but these were far less programs to finance basic education. Only dur- successful. Having been at the forefront of donor ing 2003, with the new government strategy coordination in the early 1990s, Ghana has had providing a basis, has a stronger degree of donor no education sector program in recent years. coordination emerged. Rather it has had three large donors (World Bank, USAID, and DFID) with remarkably similar proj- Education Sector Outputs ects under different management systems with an undoubted increased transaction costs for gov- Budget ernment.45 How did this situation arise? The struc- The Bank review of BESIP estimated that the IDA ture appeared to be in place for a sector program. credit represented about 8 percent of the annual There was already a donor coordination group MOE expenditures (recurrent + investment) on and a government strategy (FCUBE). BESIP, the basic education. This estimate under-states the Bank's project, was meant to be synonymous importance of the Bank's resources since over with FCUBE -- the FCUBE document is headed 95 percent of the government's basic education "the Basic Education Sector Investment Program" spending is for wages and salaries (see Annex on the cover page. Bank documentation during B). Over the period 1989­2001, the value of the preparation of BESIP frequently refers to the World Bank disbursements was one-third of fact that a sector approach is to be adopted; the total government non-wage spending in edu- project budget -- of $250 million -- covered the cation, but much higher for the basic sub-sec- whole FCUBE program of which $50 million was tor. This picture is little changed by taking into to come from the Bank. account the support schools receive from the dis- The sector approach appears to have tricts, whose total spending is less than 5 per- foundered on donor competition, despite the cent of government spending with about efforts of the Bank to encourage a government- one-third going to education. While schools led process. In July 1994, the Bank's education have benefited from these resources, they do not specialist in Accra wrote to the Minister of Edu- match the scale of World Bank financing. cation referring to discussions they had had on The scale of the Bank's operations has also developing a new approach to donor financing matched that of other donors. Bilateral aid to of basic education in Ghana, proposing to invite education totaled $350 million over the period donors to a preliminary assessment of the sec- 1989­2001, compared to the Bank's $260 million. tor later that month. The letter stressed the Within basic education the main players have importance of Government being seen to be been USAID ($88 million in the Primary Edu- firmly in charge of developing the comprehen- cation Project and QUIPS), DFID (£50 million in sive basic education program. It was suggested Whole School Development) and school build- that the Minister formally write to other donors ing by the EU under its Micro-projects pro- informing them of the Ministry's plans, and to gram.47 The value of these bilateral programs undertake a Ghanaian-led analysis to develop approximately equals that of the Bank. strategy with a meeting in November to get donors on board. But, whereas in 1987 other Activities donors had no experience in education and Table 4.1 summarizes the physical activities were willing to follow the Bank's lead, this was financed by World Bank resources.48 As shown not the case nine years later. The crucial episode in the previous chapter, the bulk of financing has appears to have been a workshop in London been directed to civil works and textbooks. (supported by the Overseas Development HERP began with the distribution of 6.1 mil- Administration, now DFID) to develop a sector lion textbooks to basic schools, which were 2 8 T H E B A N K ' S I M PA C T O N E D U C AT I O N P O L I C I E S A N D O U T P U T S A c t i v i t i e s i n W o r l d B a n k b a s i c T a b l e 4 . 1 e d u c a t i o n p r o j e c t s HERP EdSAC Ia EdSAC IIa PSD BESIP Textbooks Primary 6.1 million 1.8 million 1.5 million 6.6 million JSS 5.6 million 8.4 million 4.0 million Stationary Exercise booksb 7.5 million $1.9 million $0.6 million Pens and pencilsb 5.6 million unspecified school unspecified school supplies and supplies to JSS equipment School furniture Primary 13,800 dual desks JSS $ 1.1 million 3,450 dual desk Both 3,750 library tables and chairs; 893 teachers tables and chairs Technical equipment for JSS $ 3.4 million Classroom rehabilitation 2,300 New classroom blocks Primary 2,000 school 2,000 school 3,727 school 101 pavilions pavilions pavilions (2,908 clad) JSS 350 school 50 pavilions JSS workshops 67 workshops Other civil works Teacher accommodation 2,178 head 344 four-unit teacher's houses blocks Toilet facilities 151 pit latrines Roofing of schools 500 classrooms 999 classrooms Other activities In-service training >20,000 JSS 22,000 JSS teachers teachers Approx. 2,000 primary head teachers a. Most quantities are imputed from dollar expenditure using an assumed unit cost, based on unit cost from another project adjusted for inflation. b. Beneficiaries not stated, so that some may have gone to non-basic students. Source: World Bank project documents 2 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S mostly for math, English, and science.49 In 1990, (Table 4.1) -- econometric analysis shows that there were 2.8 million children in public basic PSD made a significant contribution to schools schools, so they would have received, on aver- having a greater proportion of classrooms that age, two textbooks each. Data collected toward can be used when it is raining (Annex D) -- the end of HERP showed 100 percent coverage pavilions were constructed under all four of the for 7 of the 20 textbook titles printed, with an main Bank projects. Many pavilions remain average of 82 percent and a minimum of 73 per- unclad, frequently with low internal walls. PSD cent. For the 15 teacher guides printed average alone reached 25 percent of primary schools, coverage was 78 percent, with complete cover- overall close to one-fifth of public basic schools age for 5 titles. HERP therefore turned the situ- have benefited from World Bank civil works in ation around from one of practically no textbooks the past 15 years. in most classrooms to having one book per stu- Other civil works include head teachers' hous- dent in most schools for the three core subjects. ing under PSD, JSS workshops, and improved This book supply supported the reform process toilet facilities. There has been some provision by putting in place textbooks adapted to the new of school furniture, notably under BESIP, which syllabus.50 rehabilitated 2,300 primary classrooms and pro- However, intended textbook lifespan is only vided furniture for them. three years, the revolving fund for textbook The Bank has been less active in other areas. procurement did not become well established, A notable exception is the teacher training pro- and fees for primary texts were dropped in vided at the time of the reforms to both junior 1995. Hence the bank has continued to supply and senior secondary school teachers to ready textbooks, most recently 11 million books under them for the new syllabus. These were short one- BESIP. Under all projects combined the Bank has off courses. As such they gave teachers some financed the provision of close to 35 million text- familiarity with the new JSS system and the new books. syllabus, hence supporting the reform program, School pavilions have been the main type of but cannot be expected to have had a signifi- civil works, with over 8,000 of these constructed cant effect on teaching methods. But the further under the various projects, for which the com- development of in-service training has been munity was expected to provide the external much more the provenance of other donor proj- walls. As noted in various World Bank reports, ects, notably QUIPS and WSD, as have support this was frequently not done. Although the to SMCs, provision of other teaching materials majority of PSD-constructed schools were clad and encouraging improved teaching methods. 3 0 5 Educational Performance Has Improved Both educational attainment and achievement have risen in Ghana over the past 15 years. The enrollment rate has risen and dropouts reduced, so that completion has risen from 60 to 73 percent. The gender gap in primary enrollments has been virtually eliminated and the gap in enrollments between children from poor and non-poor households narrowed. At the same time, test scores have improved. Children completing JSS today with nine years of basic education perform better in the math and English tests than did chil- dren leaving middle school after ten years of schooling in the 1980s. School Attainment: Higher Enrollments Growing enrollments have narrowed enroll- and Better Completion Rates ment differentials. The gap between male and The school system expanded throughout the female enrollments has been virtually elimi- reform period. The number of basic schools nated (Figure 5.2).54 Closing of the gender gap increased by 50 percent from 12,997 in 1980 to is in part a function of growing enrollments: 18,374 in 2000. This expansion has enabled ris- when enrollments are 100 percent then there can ing enrollments. By 2000, over 90 percent of be no gaps. The gender gap remains greatest Ghanaians aged 15 and above had attended where enrollments are lowest: notably in the school compared to 75 percent 20 years earlier Northern region.55 Enrollments have expanded (Annex H, Figure 3). The downturn in enroll- most rapidly in the savannah (Northern and the ments that had begun in the mid-70s was two Upper regions), where the attendance rate reversed. The basic school enrollment rate has for 7­12 year olds was just 52 percent in 1988. risen steadily since the start of the reforms, accu- On the other hand, although rural enrollments mulating an increase of over 10 percentage have risen, they have not done so more quickly points between 1988 and 2001.51 than those in urban areas so that the differen- GLSS data show continuously improving tial has remained. Finally, primary enrollments school attendance rates52 among children of pri- have risen more rapidly among the poor than mary and junior secondary school age (Figure the non-poor, although a substantial gap remains. 5.1).53 On the other hand, attendance rates at the In 1988 only 60 percent of the poorest quintile secondary level showed a large initial increase attended school compared to 80 percent of the but have since leveled off. top quintile. By 2003 these figures were 77 and 3 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S M o r e c h i l d r e n a r e a t t e n d i n g s c h o o l F i g u r e 5 . 1 ( a t t e n d a n c e r a t e s b y a g e g r o u p ) 100 group 80 age 60 of 40 enrolled 20 Percentage 0 7 to 12 13 to 15 16 to 18 Age group 1988 1992/93 1998/97 2003 Source: GLSS 2, 3, and 4 and GSS/OED household survey 94 percent, respectively. The narrowing of the Ghana, compared to close to 30 percent in gap in enrollments between the poor and non- neighboring Togo.57 Students may drop out poor means that support of the expansion of pri- before completing their education, though the mary education has been pro-poor. But for data show this to be relatively rare, and declin- junior and senior secondary schools enrollments ing. In 2003, 95 percent of those aged 15 or less have grown more rapidly among the less poor who began primary school reached Grade 6, and (Annex H paragraph H.1.20). Spending for these 92 percent went on to complete JSS.58 Fifteen sub-sectors also benefits the poor, but by less years ago these figures were 86 and 73 percent, than the non-poor have benefited. respectively, showing that retention, and so also Repetition is not a large problem in Ghana completion has much improved. But variations since there is automatic progression from one remain. Completion rates are lower in rural grade to the next provided minimal attendance areas, particularly in the savannah zone, where requirements are satisfied.56 World Bank data 9 percent of students do not complete Grade 6 show a repetition rate of less than 5 percent in (Table 5.1). Male and female retention rates are A n d g e n d e r a n d r e g i o n a l g a p s a r e F i g u r e 5 . 2 c l o s i n g ( e n r o l l m e n t d i f f e r e n t i a l s , 6 ­ 1 1 y e a r o l d s ) 1.00 rate 0.90 0.80 national to 0.70 Ratio 0.60 1988 1992/93 1998/97 2003 Female Rural Savannah Source: GLSS 2, 3, and 4 and GSS/OED household survey 3 2 E D U C AT I O N A L P E R F O R M A N C E H A S I M P R O V E D D r o p - o u t r a t e s a r e l o w ( r e t e n t i o n T a b l e 5 . 1 r a t e , 2 0 0 3 ) Percentage of those enrolling in grade 1 who complete Grade 4 Grade 6 JSS Region Coastal 97.4 97.4 92.5 Forest 97.0 95.0 91.4 Savannah 93.2 91.5 n.a. Rural/urban Urban 97.8 96.6 92.7 Rural 94.9 93.8 90.4 Sex Male 96.1 95.0 90.2 Female 96.9 95.5 93.1 Source: GSS/OED household survey. comparable until grade 6, and a slightly higher age cohorts are less likely to have completed proportion of girls complete JSS than do boys. than younger ones. Second, the line for 2003 lies The poor remain more likely to drop out than above that for 1988. People aged 18­24 today the non-poor (Annex H, Table H.13). are more likely to have completed primary than Figure 5.3, which presents data from GLSS2 the same age group 15 years earlier, both as and the GSS/OED survey, on the percent of dif- enrollments have risen and drop-outs fallen. ferent age groups (all children) which have The figure also shows completion rates for completed primary. These data confirm the ris- females from the 2003 data. The female com- ing completion rate in two ways. First, the line pletion rate has converged on that for male over drawn from each survey is downward sloping time, although a gap remains. Completion rates -- within each survey the data show that older have improved for all income groups. In 1988 M o r e c h i l d r e n f i n i s h s c h o o l F i g u r e 5 . 3 ( p r i m a r y c o m p l e t i o n r a t e s ) 80 y 70 60 primar 50 40 completed 30 % 20 10 0 18-24 25-34 35-49 >= 50 Age range 1988 2003 2003 (female) Source: World Bank World Development Indicators, 2003 3 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S only 65 percent of children entering P1 from Fifteen years later the GSS/OED survey re-vis- households in the bottom quintile completed ited the same 85 communities and carried out basic education; by 2003, 74 percent do so exactly the same tests in 1,740 households. The (some way below the figure of 91 percent for results clearly show that children are better edu- the top quintile). cated today than they were 15 years ago. Primary graduates scored an average of 5.6 on the short Improved Test Scores English test and 5.7 on the math test. These In 1988, Ghana Statistical Service (GSS) visited higher scores have been achieved in the context 1,524 households in 85 different areas of the of growing enrollments, so that a greater pro- country.59 Each person aged between 9 and 55 portion of those aged 9­55 took the tests in 2003 years and with at least three years of schooling than in 1988. was asked to take a short English reading test The improvement in the output of the basic of eight multiple choice questions and a math education system in Ghana is shown by Figure test of eight sums (two addition, two subtraction, 5.4, which plots the regression-based mean test two multiplication, and two division). Those score against years of education for 1988 and scoring five or more on either test took a longer, 2003.62 The test score shown is the combined test more advanced test.60 The results revealed the score, which is the sum of simple and advanced poor quality of education being received by test scores, with a resulting maximum of 37 for Ghanaian children. Children who had com- English and 44 for math.63 Several points emerge pleted three years of primary education scored from these graphs. First, children at all levels of an average of 0.8 on the short English test -- basic education (grades 1­10 in 1988 and 1­9 worse than if they had simply guessed all the today) score higher marks today than did their answers.61 Children who had completed all six counterparts 15 years ago. Second, Junior Sec- years of primary did not do much better, with ondary School graduates score higher than did an average mark of only 3.1. In the simple math Middle School graduates, despite the latter receiv- test the average score for primary graduates ing 10 rather than 9 years of education.64 Third, was 4.9. the gain is larger at lower grades and for Eng- S c h o o l o u t p u t s h a v e i m p r o v e d : F i g u r e 5 . 4 t e s t s c o r e s p l o t t e d a g a i n s t y e a r s o f s c h o o l i n g Math English test 40 4 0 35 3 5 Senior Middle/ Senior Primary Middle/ Primary Secondary JSS S econdary JS S 30 3 0 score score 25 2 5 2003 th 2003 ma English 20 2 0 Combined 1988 15 Combined1 5 1988 10 1 0 5 5 0 0 0 2 4 6 8 10 12 14 16 18 2 0 0 2 4 6 8 10 12 1 4 16 18 20 Years of schooling Years of schooling Source: GLSS2 and GSS/OED household survey 3 4 E D U C AT I O N A L P E R F O R M A N C E H A S I M P R O V E D lish is reversed for secondary school graduates cent) of children who have completed Grades who score worse today than did their counter- 3­6 scored 5 or less on the simple English test, parts 15 years ago. Although not the subject of meaning they are barely literate and one-fifth (19 this report, which focuses on basic education, percent) scored 2 or less, i.e., the same as guess- the shortening of pre-university education from ing, and so are illiterate. But 15 years ago these 17 to 12 years may have been at the expense of figures were 78 and 62 percent, respectively. the quality of senior secondary school gradu- Negative perceptions of the state of education ates.65 More specifically, the data show that it is arise from comparing the system today with the compression of the previous seven years of that pre-crisis, some 30 years ago rather than 15. secondary into three of senior secondary that has Such views also arise from continued middle- caused this deterioration.66 class discontent regarding the reforms and their While children of better-off households on impact on senior secondary education. average score higher, scores have improved for There is corroborating evidence of improved children of households across the income dis- educational outcomes from the Criterion Refer- tribution. There is greater uniformity in per- ence Test (CRT) carried out since 1992, the formance across income groups for primary mean English score rising from 29.9 to 36.9 school children today than 15 years ago. Nev- between 1992 and 2000 and math by a similar ertheless for math the improvement has been amount (Figure 5.5). While covering a shorter greatest for the children of the relatively better time period than the two GSS surveys, the tests off (Annex G, Table G.4). show the same clear improvement in test scores. A longer-term perspective is provided by Whilst the CRT confirms the improvement which looking at the scores of those leaving school after has taken place it also confirms that standards completing Grades 5 or 6 across all age groups. are still very low: the most recent CRT shows that Table 5.2 shows the decade average scores of in 2000 less than 10 percent of children reached primary school leavers. Both the English and mastery level in math, and less than 5 percent math scores show a U shape, declining into the did so in English. 1980s but then picking up in the past decade. The CRT scores also show the better per- The finding that educational outcomes are formance of children in private schools. In 2000 improving appears to run contrary to general the mean CRT English score was 70 in private concerns about the poor state of Ghana's basic schools compared to 39 in public schools. It education.67 There is no contradiction here. might be thought that the improvement found While things have got better there is still ample in the GSS/OED survey data can be attributed room for improvement. Nearly one-half (46 per- to increasing enrollments in private schools. S t u d e n t p e r f o r m a n c e i s r e t u r n i n g t o t h e l e v e l s a t t a i n e d 4 0 y e a r s T a b l e 5 . 2 a g o ( a v e r a g e t e s t s c o r e s o f p r i m a r y s c h o o l l e a v e r s b y d e c a d e ) English Math 1950s 5.4 4.2 1960s 2.5 4.1 1970s 3.8 4.1 1980s 0.9 3.2 1990s 3.1 4.6 Notes: calculated for those leaving school after completing Grades 5 and 6. 1990s is 1990-2002. Source: GSS/OED survey. 3 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S C r i t e r i o n R e f e r e n c e T e s t s c o r e s i n F i g u r e 5 . 5 p u b l i c s c h o o l s h a v e r i s e n e a c h y e a r , 1 9 9 2 ­ 2 0 0 0 38 36 English 34 score 32 CRT Math 30 Mean 28 26 24 1992 1993 1994 1995 1996 1997 1999 2000 Source: MoE This is not the case. The CRT data reported in which is usually measured indirectly as the per- Figure 5.5 are for public schools only and they centage of those aged 15 and above who have show an improvement in the 1990s. Similarly in completed grade 5 (Annex H). However, the the GSS/OED survey data, whilst students from test score data allow calculation of actual lit- private schools do better, it does not explain eracy measures, defining a person as literate away the improvement in scores. For example, in English if he or she scored 5 or more on the the average score in the simple English test for simple English test. Using this definition, the primary graduates has risen from 3.2 in 1988 to literacy rate among those aged 15­24 (which 4.9 in 2003. Considering public schools alone is the age group for the MDG indicator) has these figures are 2.8 and 4.6. The scores for pub- risen from 49 percent to 68 percent between lic schools alone are lower, but the improvement 1988 and 2003, and in the population of the is in fact a bit larger. whole from 37 to 45 percent. A decomposition Since a larger proportion of the population analysis shows that the increase in school qual- are now attending school, the improved test ity (higher scores achieved by those enrolled results mean that the average, quality-adjusted, in school) accounts for over half (57 percent) level of education (what economists call the the increase in literacy, with the remainder stock of human capital) is rising. The most com- coming from increased quantity (higher enroll- mon education stock measure is adult literacy, ments) (Annex H).68 3 6 6 Causes and Consequences of Improved Educational Performance Both educational attainment and achievement have risen in Ghana over the past 15 years. The World Bank, through its support of the policy reform process and financing school-level improvements in quality (principally classroom con- struction and rehabilitation) and the availability of teaching materials, has contributed to these improvements. Improved educational performance has been one factor behind better social indicators such as lower fertility and mortal- ity and improved nutrition, as well as sustaining economic growth. Better Inputs the reform process, documented in Chapter 4, and its pressure to restrain growth in GES employment. Improving School Efficiency School efficiency has increased over the past 15 Improvements in School Facilities years in the following three ways: The funds provided by the Bank have largely · The reduction in pre-university education been used to improve school facilities, notably from 17 to 12 years: the reduction of basic the supply of nearly 35 million textbooks and education from 10 to 9 years was achieved construction of close to 8,000 school pavilions. with no cost in terms of children's cognitive The contribution of these to the level of outputs performance, indeed basic school graduates was discussed in the preceding chapter. Prior to get better test scores today than they did 15 the Bank's involvement, communities were years ago. required to construct their own schools, which · Increasing instruction time: the school day was were usually mud-walled structures of limited increased from four to five hours. Although suitability and lifespan. Neither other donors this reform took some time to take hold it has nor the government have been as active in the now become well-established (Annex D). fields focused on by the Bank so that it has been · The increase in the pupil­teacher ratio: the the major player in improving these aspects of PTR in primary schools increased from 30 to school quality. 36 over the period. Improving Teacher Conditions The Bank contributed to these efficiency One channel through which school inputs affect improvements through its role in underpinning student learning outcomes is through their impact 3 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S on teacher performance. A number of teacher are significant determinant of whether or not a level variables, such as time on task, the use of child attends and stays in school:69 improved teaching methods and their monitor- · Child characteristics: children with more sib- ing of student performance, have an effect on lings are less likely to attend school, especially test scores. These teacher variables depend, in those of lower birth order. This finding fits turn on school facilities and management. The with the common observation that older chil- GSS/OED teacher survey asked respondents of dren work to pay for the education of their their perception of their living and working con- younger sibling, being themselves deprived ditions. Unsurprisingly, multiple regression analy- of education. The female dummy was sig- sis shows the former to be most strongly related nificantly negative in 1988 but not in 2003, to the availability of water and electricity in the indicating the closing of the gender gap in home and if lodging is provided. Perceptions of enrollments. working conditions are related to classroom · Household characteristics: children of higher quality (if disrupted by noise, the presence of income households are more likely to go to internal wall, and quality of the chalkboard) school, as are children of more educated par- and the availability of teaching materials. School ents. The latter results means there is an management (an active PTA and contact with the "inter-generational multiplier effect" as chil- circuit supervisor) also has a positive impact on dren sent to school 15 years ago as a result perceived working conditions. of improvements in the education system are Teachers' attitudes to both their working and more likely to send their own children to living conditions are strongly influenced by school today. whether or not they receive their pay on time. · Proximity of school facilities: the greater the The problem of late pay, which is more com- distance to the nearest primary school the less mon for new teachers, has been considerably likely is a child to be enrolled. This effect was alleviated over the years but is still an issue, with stronger in 1988 than 2003, as school build- 28 percent of teachers often not receiving their ing means that the vast majority of children pay on time. now live close to a primary school so there How teachers feel about their living and is little variation in the independent variable. working conditions matters as it affects teacher In 2003 children were significantly more likely morale, which affect both absenteeism and the to attend school if there was a private school likelihood of remaining a teacher. Teachers were in the vicinity. asked if they enjoy being a teacher and if they · Quality of school facilities: the school having intend to remain one or not. These two variables an adequate number of classrooms to cater were combined to construct a measure of teacher for all grades significantly affects the likelihood morale. Both living and working conditions are of a child going to school, as does the avail- significant determinants of teacher morale. And ability of other materials such as chalk and teacher morale is in turn a significant determi- desks.70 nant of the likelihood of a teacher being guilty · Staffing: In 2003 parents were more likely to of absenteeism (Table D.43). Both teacher absen- send their children to schools that had a low teeism and the outflow of trained teachers con- pupil teacher ratio and less likely to send them stitute considerable sources of inefficiency in the to schools with a high ratio. These results are system. picking up two phenomena. One is that over- crowding deters parents from sending children From Inputs to Outputs (Attainment and to school. But a high pupil teacher ratio also Achievement) results from having insufficient teachers -- one or two to cover four or even all six Increasing Enrollments grades, which is not unknown in rural areas Multivariate analysis of the child-level data from -- which makes parents doubt that their child 1988 and 2003 shows that the following factors will receive an education. 3 8 C A U S E S A N D C O N S E Q U E N C E S O F I M P R 0 O V E D E D U C AT I O N A L P E R F O R M A N C E Building a school, and so reducing children's and household characteristics. These regressions travel time, has a major impact on enrollments. are reported in Annex G. Schooling improves test While the majority of children live within 20 min- scores, each additional year of schooling increas- utes of school, some 20 percent do not and ing the combined English score by 3.6 points and school building has increased enrollments among math by 4.9 points.72 The 10 percent of the age these groups. In one area surveyed, average group attending school who would not have travel time to the nearest school was cut by 45 done so 15 years ago can be expected to increase minutes with enrollments increasing from 10 to their combined English score by 20 points if 80 percent. In two other areas average travel time they complete primary (as 95 percent do) and 27 was reduced by nearly 30 minutes and enroll- if they go on to complete JSS (as do 86 percent). ments increased by over 20 percent. Calculation For math these figures are 16 and 21, respectively. using the regression estimates suggest that on The direct impacts of the recorded increase average building a school in a community in in material and physical items between 1988 and which the children previously had to walk an 2003 increased math scores by 1.6 and English hour to school will increase enrollments in that by 2.0 points. These figures understate the gains community by around 5 percent (Annex I). in the most deprived areas. Ensuring that a Building a new classroom block to replace an school has one math and English book per child unusable one will increase enrollments in the compared to the situation in the mid-1980s of school's catchment area by 7.5 percent. one text per classroom will increase average Eng- The regression estimates can be used to lish scores of children in that school by 6 points examine which have been the most important and math scores by close to 10 points. factors behind enrollment growth:71 The 2003 school and teacher surveys col- · The largest single effect comes from the elim- lected data not collected in 1988, allowing a more ination of gender bias, accounting for a 4 per- detailed analysis of test score determinants. Mea- cent increase in enrollments. This autonomous sures of the quality of school infrastructure, in effect partly reflects the success of efforts to particular if classes are disrupted by noise, the get girls into school, though these are not presence of internal walls and chalkboard qual- something the World Bank has supported ity, all have a significant impact on for test directly. scores. The combined effect of these three infra- · Higher household incomes have accounted structure variables can improve English scores for enrollment growth of about 2.5 percent by 11.3 points and math by 10.1. over the period. Increased parental education The regressions using the 2003 data also accounts for close to another 2 percent. show that process matters. The most important · Improved school facilities, including reduced single variable in determining test score out- distance to school, have accounted for about comes is teaching methods. If all teachers in the a 4 percent increase in enrollments between school used improved methods then, compared 1988 and 2003, about one third of the increase to a situation in which none do so, children's over that period. A large part of this improve- English scores would be 6.2 points higher and ment can be attributed to the World Bank, their math score 8.8 higher. Important determi- which has been overwhelmingly the main fun- nants of use of improved methods are teacher der of better infrastructure in this period. How- training (notably for teachers in the coastal ever, this attribution must be seen in the context region), including in-service training. Supervision of a functioning education system in which gov- by the head-teacher and contact with the circuit ernment ensures a supply of trained teachers. supervisor also increase the likelihood that improved methods will be adopted. None of Determinants of Test Scores these are areas to which the Bank can be said Linking children to the school they have attended to have contributed. Efforts to improve head- allows a regression analysis of the determinants teacher performance through the provision of of test score outcomes incorporating both school housing were judged to be unsuccessful.73 3 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S Questions to teachers regarding the use of Textbook provision is a very cost-effective classroom time allowed the construction of a time means of improving learning outcomes, with on task variable, and this too was found to sig- teacher training being the next most cost effective nificant affect test scores. Time on task itself is (Annex G). School infrastructure also has a ben- a function of teacher training and contact with eficial effect on learning outcomes, but its largest the circuit supervisor. In addition, the ability of benefit is from enabling higher enrollments. teachers in the school to speak the local language improves student math scores, presumably since From Outputs to Outcomes they do not have to rely on English, of which students may have a poor grasp, to explain dif- Education and Social Outcomes ficult concepts. There is a well-established literature linking The results reported in the preceding para- educational outputs to welfare outcomes, both graphs pose something of a puzzle. Better economic and social. Studies of education and teaching methods and time on task improve test social outcomes tend to focus on the effect of scores, and teacher training enhances both of female education (Table 6.1). Where both male these things. But private schools largely recruit and female education is included, then the lat- untrained teachers and it is well established ter is shown to be more important.74 The most that, on average, private schools get better commonly studied outcomes have been fertility test results (chapter 5). There are two answers and child nutrition. All studies from Ghana find to this puzzle. The first is the finding that the that the higher levels of education reduce fer- private school dummy is significant for English tility, normally measured as the number of births. scores but not for math. So, once factors relat- The findings from studies of child nutrition ing to both the student's background and are more ambiguous. An early study found no school facilities are controlled for, there is no significant impact, but more recent studies find pure "private school effect" for math. There is that education does improve nutritional status. however one for English, perhaps reflecting the Both Ruel et al. (1999) and Maxwell et al. (2000) enforcement of English as the language of presenting different analyses of the same data tuition in these schools. The second answer to from Accra find that mother's education is sig- the puzzle is that there are indeed aspects of nificantly associated with better child nutrition. private schools, such as teacher discipline, In addition, there is a considerable indirect effect which are conducive to good learning out- from education on improved childcare prac- comes. This fact does not contradict the find- tices, which also improve nutrition. Good care ing that children will learn better still, even in practices, supported by education, can com- private schools, if improved teaching methods pensate for lower income. Hence the nutritional are employed. status of children of educated mothers at lower Home factors also matters to student per- income levels can equal that of children in formance. The two measures of parental involve- higher-income families. Children of mothers ment in a child's education (membership of PTA with little education living in low-income house- and meeting with a teacher) give a combined holds have the worst nutritional status. impact of 3.5 and 3.9 points on math and Eng- Analysis carried out using the GSS/OED lish scores respectively. Income also matters; eco- data supports the view that education can sub- nomic growth (the between sample rise in stitute for income in achieving better nutritional incomes) has increased average English scores outcomes. These data show that maternal edu- by 2.2 points and math scores by 1.2 points. As cation has a significant impact on child nutri- in the case of enrollments, to the extent that edu- tion (Annex K).75 This impact is higher for cation affects these household characteristics women in poorer households. In the poorest there is a multiplier effect whereby the educa- households a woman completing JSS increases tional performance of children of educated par- the expected nutritional status (measured by ents improves. height for age) by 4.5 points, sufficient to 4 0 C A U S E S A N D C O N S E Q U E N C E S O F I M P R 0 O V E D E D U C AT I O N A L P E R F O R M A N C E R e s u l t s f r o m s t u d i e s o f e d u c a t i o n T a b l e 6 . 1 a n d s o c i a l o u t c o m e s i n G h a n a Education Contraceptive measure Child survival Fertility prevalence Nutrition Alderman (1990) Female schooling Insignificant Male schooling Insignificant Benefo and Schultz Mother's Positive Positive (1996) education Glewwe and Desai Test scores Mostly (1999) insignificant (mother's math positive) Gyimah (2002) Secondary or Positive Positive higher Gupta and Mahy (2003) Maternal education: Positive None, 1-7 years, 8 or more Oliver (1999) Mother's years of Positive schooling, test scores Ruel et al. (1999) None, primary, Positive secondary Maxwell et al. (2000) Mother's education Positive level Sackey (2003) Mother's years Positive Positive Positive Positive of schooling Father's years Positive Positive Insignificant Positive of schooling OED analysis Maternal education Positive Paternal education Positive (indirect through income) move the child out of the category of being nels through which education operates are also malnourished. shown by studies showing education to affect A smaller number of studies establish the link both income and child care practices. between education and lower child mortality. The results from the studies mentioned above For example, Benefo and Schultz (1996) find can be used to investigate the scale of educa- a weak impact of mother's education on child tion's impact on welfare outcomes. Infrastruc- mortality (but a strong one on fertility). They tural improvements will result in an increase in show these effects to be stronger when women enrollments of around 10 percent and reduce live in a community with good access to water the dropout rate (Annex I). Hence, the average and weaker when there is poor access to health schooling of mothers will rise. This rise in facilities, thus supporting the view that educa- schooling leads to relatively small, though not tion facilitates the better use of other amenities negligible, changes in fertility and child mortality so as to improve welfare outcomes. The chan- (Table 6.2).76 The study of Accra shows the 4 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P e r c e n t a g e c h a n g e s i n w e l f a r e T a b l e 6 . 2 o u t c o m e s c a u s e d b y h i g h e r s c h o o l a t t a i n m e n t Fertility Mortality Nutrition Benefo and Schultz ­2.4 ­3.6 -- Ruel et al. -- -- 10.3-20.6 Sackey ­4.8 ­0.7 -- OED analysis 4.8-27.2 Source: Annex K. impact on nutrition (stunting, i.e., the height for Analysis of the GSS/OED presents a clear age z-score) a more substantial impact of an message: education matters only to the extent improvement of between 10 and 20 percent. The that it results in higher cognitive achievement channels for this increase are both improved (Annex K). Education can affect earnings both childcare practices and higher income. OED's directly -- more educated people earn more, own analysis supports these substantial effects. which may result simply from a screening func- The lower increases from education are realized tion -- and indirectly through raising their cog- among wealthier families, with the largest nitive skills, which are rewarded with higher absolute gain to poorer households. earnings. OED's analysis shows that there was a direct return to education in 1988 but this is Education and Economic Outcomes no longer so for primary and JSS in 2003, for The most comprehensive analysis of the eco- which the return in fact appears to be negative. nomic returns to education in Ghana is that by But schooling raises test scores and those with Teal (2001), which brings together data from four higher test scores earn more. Those who get rounds of the GLSS (1987/88­1998). He finds that higher test scores as a result of schooling do there is a positive return to all levels of educa- enjoy higher earnings. To the extent that poorer tion, but that it is higher for higher levels.77 The children in less well resourced schools are not rise in the average level of education accounts reaping educational benefits from school atten- for about one-third of growth in per capita dance, they will not enjoy economic gains, gen- income that has taken place over the decade erating a vicious circle of poverty. 4 2 7 Lessons Learned and Progress Toward the MDGs Progress Toward the Millennium Development Goals T he education MDG is to "ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling." In addition, the third MDG of gender equality has two edu- cation-related targets (progress toward gender equality in enrollments and literacy). Table 7.1 shows the progress made in the period 1988­2003 and extrapolates these trends to 2015. Ghana has made considerable progress toward both these goals, in particular the gen- turned toward software. This focus turns out der equality in education has been achieved to have been beneficial. The inputs the Bank with respect to primary enrollments and is likely has provided (books and buildings) have to be achieved, or close to achieved, for the other been shown to have made a significant con- gender indicators. On the other hand, at current tribution to both educational attainment and rates of progress enrollments will fall short of the achievement. Two caveats are perhaps in target of UPE and up to nearly a quarter of order: (1) the focus on hardware and mate- those aged 15­24 will be illiterate.78 Closing rials took place within the context of an these gaps will require, among other things, agreed program of educational reform with focusing attention on difficult-to-reach areas a government that has been committed, espe- and ensuring that all schools receive the required cially in the early years, to improving the level of inputs.79 This is not simply a matter of quantity and quality of education; and (2) the regional disparities: there are deprived schools projects of the other major donors (DFID in even the better off districts. and USAID) have focused on software, though the extent of their impact is not yet wide- Lessons Learned spread. The main conclusions from this report, and the · The lesson to be drawn is that getting enough lessons to be learned from them, are as fol- classrooms and those classrooms being in lows: decent shape is a necessary ingredient of · The Bank focused its spending on hardware educational strategy. But it cannot be the and instructional materials, even when the sole ingredient. Indeed, it will become less rhetoric of strategy and project documents important as all schools attain the desired 4 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P r o g r e s s t o w a r d t h e e d u c a t i o n T a b l e 7 . 1 r e l a t e d M D G s Predicted for 2015a 1988 2003 Low High Goal 2: Universal Primary Education Complete Grade Six Total 65.4 76.9 82.1 87.5 Girls 55.6 72.1 80.1 88.8 Net enrollment ratio in primary education 72.1 84.1 89.5 95.1 Proportion of pupils starting grade 1 who reach grade 5 88.5 95.8 98.9 100.0 Literacy rate of 15 to 24-year-oldsb 49.0 68.0 77.6 88.4 Goal 3: Promote gender equality and empower women Ratio of girls to boys in primary and secondary education Primary 95.0 100.0 100.0 100.0 Secondary 78.0 86.0 88.8 91.7 Ratio of literate females to malesb 83.0 92.0 95.0 98.0 a. "Naïve" predictions based on simple extrapolation, assuming 1988-2003 growth rate is sustained for high scenario and that it is halved for low scenario. b. Defined as scoring 5 or more on English reading test. Source: data from GLSS2 and GSS/OED survey. level of physical and material inputs. But be made to retain trained teachers, which Ghana is not yet in that position: substantial may suggest some reconsideration of the cur- inputs are still required for the most disad- rent policy regarding study leave. Finally, vantaged schools. Even where good school teachers being able to speak the local language quality is achieved, educational outcomes, helps student math learning. while improved, are still far from satisfac- · The downside of community and district tory. Improving them will indeed require financing of schools is that it leads to dis- attention to software. parities in resource availability. There remain · The evidence in this report of beneficial effects a class of schools in poorer communities -- from community management is not strong. particularly but not only in rural areas -- However, it can be argued that these changes which are very poorly resourced and the have yet to really take hold so that their effect community can do nothing about it and often can be felt. It is shown that parental involve- lacks the political connections to attract dis- ment does matter, but this could well be prox- trict finance. ying for parental interest in child's education. But the evidence is clear that supervision of Some immediate implications of the analysis teachers by the head teacher and circuit super- are: visor matter, as do the teaching methods · It is necessary to focus resources on the most adopted by the teacher. Since attempts to needy schools. The bias that results from remove untrained teachers have been unsuc- community-based financing needs to be over- cessful, and since not all trained teachers come. School mapping continues to play an appear familiar with student-centered important role, so support to EMIS is impor- approaches anyway, there is a strong case for tant. pushing forward with efforts to emphasize the · The private sector has been neglected, but it role of in-service training. Efforts should also is important, so attention needs to be paid to 4 4 L E S S O N S L E A R N E D A N D P R O G R E S S T O WA R D T H E M D G S it in both government strategy and Bank sup- · Teacher morale is reasonable, but is affected port. Possible areas of attention are enforc- by living and working conditions and espe- ing registration to avoid very poor schools, cially if teachers get their pay on time. Resolv- but taking care to not be too restrictive. ing payment problems will raise morale and Teacher certification could be required, and reduce absenteeism. while there does not seem a need to require · Inputs matter, textbooks in particular. While formal teacher training, the provision of in- sustainable textbook financing is a desirable service training would help promote better goal, donors should not be averse to large- learning outcomes. scale textbook provision, such as the World · Teaching methods matter a lot for test out- Bank has done. Thought might also be given comes. Teacher training seems to affect this, to providing exercise books and pencils to the but differentially. This supports the idea that most needy basic schools (perhaps by geo- teacher training should look at method as well graphical targeting to bottom 20 districts, but as content. In-service training matters, but probably not given the substantial intra-dis- there is not much of it at present. trict variation in school quality). 4 5 ANNEXES 49 Annex A: Test Examples 53 Annex B: Budget Analysis 63 Annex C: School Costs 71 Annex D: School-Level Changes in Inputs, Management, and Methods 107 Annex E: Tables Of School Quality Variables 125 Annex F: Variable Definition 129 Annex G: Analysis of Test Scores 145 Annex H: Data on Educational Performance 165 Annex I: School Attainment 177 Annex J: Conditions Attached to Bank Credits 191 Annex K: Education and Welfare Outcomes 197 Annex L: Evaluation Approach Paper 4 7 ANNEX A: TEST EXAMPLES Short math test 1. 1 + 2 = 5. 24 + 17 = 2. 5 ­ 2 = 6. 33 ­ 19 = 3. 2 3 = 7. 17 3 = 4. 10 ÷ 5 = 8. 41 ÷ 7 = 4 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S Short English Reading Test John is a small boy. He lives in a village with his brothers and sisters. He goes to school every week. In his school there are five teachers. John is learning to read at school. He likes to read very much. His father is a teacher, and his parents want him to become a school teacher too. 1. Who is John? 5. What is John doing at school? (A) An old man (A) Helping the teacher (B) A small boy (B) Talking with his friends (C) A school teacher (C) Learning to read (D) A school (D) Teaching the class 2. Where does John live? 6. Who is a school teacher? (A) In a village (A) John (B) In a city (B) John's father (C) In a school (C) John's brother (D) In a forest (D) John's mother 3. What does John do every week? 7. What do John's parents want him to do? (A) Works with his father (A) Go to school (B) Plays with his friends (B) Learn to read (C) Helps his brothers and sisters (C) Obey his teachers (D) Goes to school (D) Become a teacher 4. How many teachers are there at John's school? 8. The best title for this story is (A) One (A) John Learns to Read (B) Three (B) Why Reading is Important (C) Five (C) John's Village (D) Six (D) Schools in Ghana 5 0 A N N E X E S A sample of questions from the Advanced Mathematics test 3. There are 4 rows of chairs and 12 chairs A in each row. How do you find out the total number of chairs? 12 cm (A) 12 + 4 (B) 12 ­ 4 (C) 12 4 (D) 12 4 B 12 cm. C Note: figure not drawn to scale 8. 1 + 1 = 14. If the perimeter of the triangle ABC is 30 2 3 centimetres, what is the length, in cen- timetres of side AB? (A) 1 5 (A) 1 2 (B) 2 6 (B) 3 (C) 2 5 (C) 6 (D) 5 (D) 18 6 36. Which CANNOT be the intersection of 3 13. 1% of 400 is planes? (A) 1 (A) 1 point (B) 4 (B) 1 line (C) 40 (C) 3 concurrent lines (D) 400 (D) 3 parallel lines 5 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S A sample page from the Advanced English test (...) The cat brushed against the old man. He did not move. He only stood, staring in the win- dow of the house. The party inside looked warm and friendly, but no one noticed him. The old man walked sadly on, followed by the cat 8. What was inside the house? (A) A party (B) Some dogs (C) An old lady (D) A meeting 9. The man is described as being (A) Old (B) Young (C) Thin (D) Small Directions: For questions 10­15, read the passage below. Each line of the passage has a number. In each line, there is a box with four possible choices. Pick the choice that best completes the sentence in each numbered line. Mark the letter (A,B,C, or D) of the choice on your answer sheet. (A) hears. (B) hearing. 10. Sound is something we It comes to your (C) heard. (D) hear. (A) eyes (B) nose 11. in different ways. It might be pleasant, (C) ears (D) mouth (A) when 12. like the voice of a friend, (B) as unpleasant, like the yelp (...) (C) or (D) since 5 2 ANNEX B: BUDGET ANALYSIS E ducation expenditure data were obtained these budget lines are for existing institutions they from the Ministry of Education (MoEYS), have been excluded from the totals shown here.80 which compiles annual data as provi- The resulting data show a substantial rise in sional (budgeted) and actual, broken down to education's share of GDP over the period 1984­87 budget lines and functional classifications. These and a continued, but erratic, rise since then. data were provided to the study team by the min- Data on education's share of government istry for the period FY 1989­2001, although the spending are available from the Quarterly Digest data for 1993 could not be located. of Statistics published by Ghana Statistical Ser- This annex is primarily a technical note that vice. The tables provide the total of recurrent and explains how the analysis was performed. Some capital (development) expenditure of educa- observations are made as to the results insofar tion. The resulting percentage is shown in Fig- as are required to support the argument in the ure B.1. The share of education rose from main report. 1982­87 and then leveled off before falling in the 1990s. This decline has been slightly miti- Trends in Education Spending gated by spending through the districts, which is discussed below. As a percent of government expenditure In August 2001 the Ghana Education Trust and GDP Fund (GETFund) was established by act of par- Data on expenditure as a percent of GDP were liament to provide additional resources to all lev- collected from both the Ministry of Finance and els of education financed out of an equivalent Ghana Statistical Services. However, the educa- to two and one half percent out of the prevail- tion expenditure data from these sources did ing rate of the Value Added Tax and any other not correspond to those provided by the Ministry contributions. In practice the bulk of the funds of Education, although the discrepancy was rel- have been used for Tertiary education: in 2002 atively minor (in most years it was around 5 per- of the 140 billion cedis disbursed, 125 billion (89 cent, which was the median, in only one year did percent) were allocated to the tertiary level, the it exceed 7 percent, reaching 17 percent). Since bulk of it (90 billion) being a contribution to the the MoF/GSS series were only available until student loan scheme. This use of general taxa- 1997, the presentation here is based on OED's tion to finance tertiary education is a regressive own calculations, using the MoEYS expenditure fiscal policy. The value of GETFund disburse- data and GDP data from World Development ments is around 10 percent of government Indicators. The latter series is identical to those spending on education, thus boosting overall from MoF/GSS other than 1990­92. For the years education spending but reducing the share of 1982­88, the MoF/GSS data were used. The dis- basic education in that spending. crepancy in 1988 (the year in which the series are joined) is just 0.2 percent of GDP, so no adjust- Real education expenditure ments were made in linking the data. However, Analysis of real expenditure trends often deflates five new budget lines were added under the Min- expenditure by the consumer price index (CPI) istry of Education budget heading in 1999. Since or sometimes the GDP deflator. However, this pro- 5 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S F i g u r e B . 1 T r e n d s i n e d u c a t i o n e x p e n d i t u r e 70 6.0 60 5.0 50 4.0 (billions) P 40 GD cedis 3.0 of Real 30 Percent 2.0 Percent; 20 1.0 10 0 0.0 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Share of government expenditure Real expenditure Share of GDP (right axis) Source: MoE and GSS data cedure is inappropriate if there has been a change The same calculation was performed for junior in real wages, especially if wages are a substan- secondary. A weighted average was taken of the tial component of total expenditure. The correct two series using the respective weights in approach is to separate out wage and non-wage employment over the whole period (66:34 for elements of expenditure and deflate the former primary:JSS). This wage series was re-based at by a wage index and the latter by a price index. 1989=100. The latter approach was followed here, using The wage and price series were used to the non-food component of the CPI as the price deflate personal emoluments and the non-salary index, re-based for 1989=100. There is a break component of expenditures respectively. The in Ghana's CPI with a new series reported since resulting series were summed to give total real 1998 and no GSS data provide an overlap expenditure. between the two series. However, the IMF Sta- The result is shown in Figure B.1. There has tistical Annex reported an annual inflation rate been a continual, if slightly erratic, increase in for the non-food CPI for 1998. This figure was real expenditure throughout the period 1989­99, used to link the two series. which has been reversed in the last two years The wage index was constructed as follows. for which data are available, though GETFund Data are available on the personal emoluments began disbursements in 2001. for all budget headings. However, we do not have employment data corresponding to each Basic Education Share of Education of these budget heads. But data are available on Spending the number of teachers employed in primary and JSS. The implicit salary per person was calculated Data and method by dividing personal emoluments for primary by The education budget is divided into the fol- the number of teachers in public primary schools. lowing headings: 140, covering MoEYS head- 5 4 A N N E X E S quarters, 141 for GES, 142 for regional services, experience and are adopted here. which includes both the cost of district offices The basic education share shown in Figure B.2 and the funds flowing to school facilities them- was calculated from the MoEYS spreadsheets selves, 143 for special education for the handi- using the MoEYS's own methodology. The results capped, 144 for national culture (including were checked against the MoEYS's own reported archives), and 145 for tertiary education. Each figures for the basic share and discrepancies rec- of these codes is divided into several lines cor- onciled.81 MoEYS estimates of the basic share responding to the departments and units within were not available for 1999­2001, but replication the various organizations. of MoEYS's figures for 1989­98 is a safeguard that The Ministry of Education spreadsheets show the figures reported are accurate. provisional (budgeted) and annual expenditure The MoEYS spreadsheets break down the by each budget line, separated into personal basic share into primary and JSS by using a set emoluments and non-salary items. Recent ratio of 0.7:0.3. Following this procedure results spreadsheets provide a different breakdown of in the same trends being observed as can be non-salary items, including investment costs, seen in the basic education data. However, the whereas data for earlier years refer solely to procedure loses the information contained in recurrent costs. The share to basic education is the fact that the two largest single line items calculated by applying a coefficient to each line -- primary and middle/JSS, which together item. For items solely dedicated to basic edu- account for over half of total spending and over cation (that is, the basic education staff within 80 percent of basic education expenditure -- GES, and the primary school and middle/JSS can be allocated to these two categories. The budget lines) a coefficient of one is assigned. The primary school share reported here was cal- coefficient for other budget lines varies, the culated by (1) deducting the sum of the pri- most common value being 0.6. All central MoEYS mary and middle/JSS budget lines from the budget lines have a weight of zero. These coef- basic school total; (2) adding the primary ficients are set by the ministry based on their school line to the pro-rated (0.7) residual; and B a s i c e d u c a t i o n s h a r e o f c e n t r a l F i g u r e B . 2 g o v e r n m e n t s p e n d i n g 80 70 60 spending 50 education 40 of Share 30 20 1989 1991 1993 1995 1997 1999 2001 Provisional Actual Primary (actual) JSS (actual) Source: MoE and GSS data 5 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S (3) calculating the resulting share. The equiv- district income on average and more in poorer alent procedure was followed for middle/JSS, and rural districts.83 using a share of 0.3. While the constitution allowed for the hand- This methodology implicitly distributes admin- ing over of all responsibility for local services to istrative costs, other than those of MoEYS HQ, the district this is not what has happened in prac- across the sectors proportional to their share of tice. District-level offices of the various min- spending. Hence "basic share" is not a measure istries and agencies have been created, such as of funds flowing to basic schools. An alternative the Ghana Education Service (GES), which are approach is to attempt to separate out adminis- accountable to their central body rather than the trative costs into another line item. However, District Assemblies. Hence GES is responsible for according to an academic researcher who has the distribution of recurrent inputs to schools, analyzed these data,82 the school budget lines and teachers' salaries are paid centrally,84 with (primary, etc.) also contain some administrative the districts being responsible for infrastructure costs of an unknown, and varying, amount, such as school rehabilitation, although they also making an accurate figure for the administrative provide school furniture. overhead difficult to obtain. Given these uncer- Expenditure data show that total district expen- tainties, the appropriate procedure was to adopt diture is just over 4 percent of central government the same approach as MoEYS. expenditure (Appiah et al. 2000), but between The basic share has fluctuated, but with no 20­25 percent of the government's development trend. budget (Boko 2002). Of this amount, on average a little over one-third is spent on education. District-Level Expenditure Hence the creation of the DACF has increased The PNDC government expressed its commit- the share of government spending on education ment to decentralization in 1983, but the first con- by just over 1 percent over and above the increase crete steps were taken five years later in 1988 seen in Figure B.1. The district spending may when the current structure of 110 districts was safely be assumed to be virtually entirely devoted set up (an increase from 65) and the District to basic schools, so that the share received by Assemblies created. Article 35 of the 1992 con- basic education has also been increased by the stitution embodies the principle of decentral- introduction of the DACF. However, the magni- ization, with the parameters elaborated in tudes involved are not sufficient to make any sub- Chapter 20. Article 252, states that through the stantial difference to the trends noted above. District Assemblies Common Fund (DACF) not This does not mean that the spending does not less than 5 percent of government revenue matter at school level. Since central resources are should be provided to districts. The Common largely financing salaries (see next section) addi- Fund, formally created by the DACF Act in 1993, tional financing at local level has the potential to is allocated by a formula including a needs- create notable variations in school quality. based component to ensure poverty targeting. Of the seven indicators used in the district allo- Functional Distribution of Expenditure and cation rule, two relate to education. A parlia- Relative Importance of Donor Finance mentary-approved formula also provides The MoEYS spreadsheets show the breakdown guidelines on the use of funds, all of which of spending for each budget line under the fol- should be used for investment purposes. lowing headings: (i) personal emoluments (PE), In addition to the Common Fund, districts can (ii) traveling and transport, (iii) general expen- raise their own funds from taxes, fees, and ditures, (iv) maintenance and repairs, (v) sup- levies, as well as receiving ceded revenue from plies and stores, (vi) investment costs, and central government by which the income from (vii) subventions. Since 1999, headings (ii) to (v) certain taxes is meant to be given to the districts. have been replaced by two headings: (i) admin- In practice, the Common Fund is the main source istrative costs and (ii) service costs, which are of district revenue, accounting for two-thirds of listed under items (iii) and (iv) in Table B.1. 5 6 A N N E X E S F u n c t i o n a l c l a s s i f i c a t i o n o f T a b l e B . 1 e d u c a t i o n s p e n d i n g ( p e r c e n t ) Mainte- nance Personal Traveling General repairs Total emolu- and expen- and Supplies Investment Sub- expenditure ments transport ditures renewals and stores costs vention Primary 1989 100.0 99.3 0.5 0.2 0.0 0.0 0.0 0.0 1990 100.0 99.3 0.5 0.2 0.0 0.1 0.0 0.0 1991 100.0 99.4 0.3 0.2 0.0 0.0 0.0 0.0 1992 100.0 99.6 0.2 0.2 0.0 0.0 0.0 0.0 1994 100.0 98.8 0.5 0.6 0.0 0.1 0.0 0.0 1995 100.0 98.8 0.5 0.6 0.0 0.1 0.0 0.0 1996 100.0 99.3 0.3 0.2 0.0 0.2 0.0 0.0 1997 100.0 98.4 0.3 0.2 0.0 1.1 0.0 0.0 1998 100.0 97.4 0.7 0.5 0.5 0.8 0.0 0.0 1999 100.0 95.9 3.8 0.2 0.0 0.0 0.0 0.0 2000 100.0 97.5 0.0 2.5 0.0 0.0 0.0 0.0 2001 100.0 98.6 0.0 1.4 0.0 0.0 0.0 0.0 JSS 1989 100.0 97.5 1.5 0.5 0.5 0.0 0.0 0.0 1990 100.0 98.0 1.0 0.4 0.6 0.1 0.0 0.0 1991 100.0 98.7 0.6 0.3 0.4 0.1 0.0 0.0 1992 100.0 99.2 0.3 0.2 0.3 0.1 0.0 0.0 1994 100.0 98.9 0.4 0.3 0.3 0.1 0.0 0.0 1995 100.0 98.9 0.5 0.5 0.1 0.1 0.0 0.0 1996 100.0 99.3 0.3 0.2 0.1 0.1 0.0 0.0 1997 100.0 97.8 0.3 0.2 0.1 1.7 0.0 0.0 1998 100.0 98.1 0.5 0.6 0.3 0.6 0.0 0.0 1999 100.0 97.0 2.8 0.2 0.0 0.0 0.0 0.0 2000 100.0 97.6 0.0 2.4 0.0 0.0 0.0 0.0 2001 100.0 98.7 0.0 1.3 0.0 0.0 0.0 0.0 Secondary 1989 100.0 90.2 5.3 3.5 1.0 0.0 0.0 0.0 1990 100.0 88.9 4.3 3.2 1.2 2.4 0.0 0.0 1991 100.0 92.1 3.1 2.3 0.8 1.8 0.0 0.0 1992 100.0 94.4 1.8 1.6 0.5 1.7 0.0 0.0 1994 100.0 93.1 2.8 2.4 0.8 0.9 0.0 0.0 1995 100.0 85.3 3.1 8.4 0.9 2.3 0.0 0.0 1996 100.0 84.4 3.8 8.0 0.8 3.0 0.0 0.0 1997 100.0 85.5 4.1 7.8 0.1 2.4 0.0 0.0 1998 100.0 85.6 3.2 6.0 1.5 3.8 0.0 0.0 1999 100.0 84.4 6.7 0.2 8.7 0.0 0.0 0.0 2000 100.0 79.5 0.0 5.3 0.0 0.0 15.2 0.0 2001 100.0 95.8 0.0 2.7 0.5 0.0 1.0 0.0 5 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e B . 1 F u n c t i o n a l c l a s s i f i c a t i o n o f ( c o n t i n u e d ) e d u c a t i o n s p e n d i n g Mainte- nance Personal Traveling General repairs Total emolu- and expen- and Supplies Investment Sub- expenditure ments transport ditures renewals and stores costs vention Teacher training 1989 100.0 83.2 9.9 3.7 1.2 2.0 0.0 0.0 1990 100.0 84.6 8.5 3.5 0.9 2.5 0.0 0.0 1991 100.0 86.1 7.0 3.5 0.9 2.5 0.0 0.0 1992 100.0 93.7 2.6 1.8 0.5 1.5 0.0 0.0 1994 100.0 94.1 2.5 1.7 0.5 1.2 0.0 0.0 1995 100.0 92.5 2.0 4.2 0.2 1.1 0.0 0.0 1996 100.0 93.6 2.3 2.7 0.5 1.0 0.0 0.0 1997 100.0 94.5 2.0 2.3 0.3 0.9 0.0 0.0 1998 100.0 95.3 1.3 1.5 0.6 1.3 0.0 0.0 1999 100.0 97.7 1.4 0.1 0.8 0.0 0.0 0.0 2000 100.0 95.5 0.0 3.9 0.0 0.0 0.7 0.0 2001 100.0 96.8 0.0 2.6 0.2 0.0 0.4 0.0 Total expenditure 1989 100.0 74.8 1.7 1.3 0.5 5.0 0.0 16.8 1990 100.0 71.0 1.4 1.5 0.8 7.6 0.0 17.7 1991 100.0 76.8 1.1 1.0 0.7 3.8 0.0 16.5 1992 100.0 78.9 0.7 1.1 1.0 5.2 0.0 13.1 1994 100.0 75.3 1.1 1.5 2.2 2.8 0.0 17.2 1995 100.0 78.2 1.3 2.4 1.2 3.4 0.0 13.4 1996 100.0 77.4 1.4 1.9 1.0 3.0 0.0 15.2 1997 100.0 77.4 1.5 2.1 1.1 2.7 0.0 15.2 1998 100.0 76.1 2.0 2.7 1.5 3.0 0.0 14.6 1999 100.0 85.5 0.0 8.3 4.4 0.0 1.8 0.0 2000 100.0 88.7 0.0 5.7 3.9 0.0 1.7 0.0 2001 100.0 95.0 0.0 3.7 0.8 0.0 0.5 0.0 Source: MoEYS expenditure spreadsheets. From 1989­98, PE accounted for between 70 subventions now appear as PE. To ensure com- and 80 percent of total expenditure. The share parability, in the data shown here spending on jumped after that because of a change in the sys- tertiary education in the period since 1999 has tem of financial administration. Support to terti- been treated as though it were a subvention. ary institutions was made entirely through In the basic education sectors that are the subventions until 1998,85 so that tertiary spend- focus of this study, an even greater proportion ing on other categories is not reflected in the of spending has been for PE -- staying at 99 per- table. But from 1999 the subventions stopped, so cent in primary until 1997 and never falling that wages and salaries previously covered by the below 96 percent. In primary, virtually nothing 5 8 A N N E X E S was spent on the three areas that have benefited are entirely outside of the budget then they most from Bank support -- maintenance, repairs have provided up to an additional 40 percent of and renewals, supplies and stores, and invest- resources compared to government's non-wage ment costs. The figures are little different for JSS, spending. But if Bank funds pass through the though with a small (but still negligible) amount budget they have accounted for up to 70 per- of spending on maintenance. The PE share is cent of non-wage expenditure. It is likely to be lower in teacher training and secondary, but in the norm that the funds are off-budget. It is the former case, the difference is largely con- most likely that some EdSAC resources passed sumed by traveling and transport. A slightly through the budget. The figures are serious larger amount is spent on supplies and stores for under-estimates for the Bank's contribution to the secondary, but this presumably includes the non-wage component of basic education spend- costs related to boarders. From this discussion ing, since it has been shown that government it is clear that the larger percentage of supplies finances little other salaries in that sub-sector. The and stores in total expenditure reflects supplies value of the Bank's resources for non-salary and stores purchased for administrators in MoEYS spending in basic education is many times that HQ and GES, not for schools. of the government. Data on bilateral flows are more difficult to The role of donor finance86 come by. The Development Assistance Com- As a first approximation, it can be said that mittee (DAC) online database provides annual in basic education government is financing commitment data to each recipient by country "nothing but salaries." There is a small amount and sector. The coverage of these data is some- for supplies such as chalk and some infrastruc- what uneven, so there is a danger of under- ture improvements and school furniture are reporting. In addition, the data refer to financed by the districts. Communities also commitments rather than disbursements. For finance some inputs, but these amounts will be what they are worth, the data show that total small compared to official funding, especially bilateral support to education has been about since responsibility for construction was taken $318 million, compared to $257 million from the away from communities in the 1990s. It is clear Bank. There was very little bilateral funding that most of what has been done in upgrading before 1990, when the first large USAID project school infrastructure as well as textbook supply began, which was in fact budget support, as was has come from elsewhere. The relative impor- EdSAC. In the early 1990s, Bank and bilateral tance of donor finance illustrates this point. funding were on a par. In the later part of the Data are available for Bank disbursements on 1990s, bilateral funds have exceeded those from an annual basis from Implementation or Project the Bank. Over the period as whole, the Bank Completion Reports (ICRs and PCRs). These has provided about 45 percent of external sup- data are reported in Table B.2(a). These totals port to the education sector.87 were converted to cedis using the average The main donors involved in basic education exchange rate. The disbursement figures can have been the World Bank, USAID, DFID (each then be expressed as a percentage of total of which has put a comparable amount into the expenditure and of total non-PE expenditure, see sector), and the EC through the micro-projects Table B.2(b). Figure B.3 shows total Bank dis- program. These activities are summarized in bursements on an annual basis and the ratio of Table B.3. these to non-PE expenditure. If the Bank funds 5 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S 2002 4.00 14.7 n.a. 10.70 2001 1.00 1.00 14.4 n.a. 12.37 2000 1.30 18.6 14.2 17.30 3.9 1999 1.00 0.10 1.50 1.28 37.3 1998 4.20 5.90 0.40 2.57 13.1 104.4 1997 2.07 0.80 1.28 34.7 56.2 15.40 15.20 disbursements 1996 2.64 0.80 35.9 24.4 12.20 20.30 1995 2.20 2.51 8.20 9.20 38.1 13.4 16.00 3.3 (annual 1994 4.60 3.26 1.39 0.60 2.50 12.4 5.0 1993 8.70 4.31 2.02 3.20 18.2 1992 1.02 5.41 0.00 17.0 19.2 10.60 education 1991 0.62 23.5 18.6 10.21 13.30 to 9.9 1990 0.62 9.90 12.3 4.7 6.9 1989 0.93 4.67 3.2 0.9 1988 0.68 3.16 support million) 9.4 1.2 1987 0.62 9.36 Donor US$ otalT 3.47 5.80 38.32 53.20 15.18 10.63 44.80 53.20 256.9 317.6 (a) skills literacy* project B.2 (DAC) functional skills* Bank I II and estimate. education functional orld W bilateral *OED able T HERP EdSAC EdSAC CSSC Literacy rtiaryeT PSD ocationalV BESIP* National otalT otalT Note: 6 0 A N N E X E S .. 0.0 5.5 90.7 40.0 14.8 2001 1664.2 226.6 1437.6 6.8 1.3 75.6 57.5 34.2 26.0 2000 1110.9 889.9 221.0 of 9.7 1.1 4.5 4.1 1999 93.0 914.6 43.1 698.7 215.9 4.6 1998 28.5 227.9 614.9 19.4 17.0 468.2 146.7 155.3 percent as 1997 64.1 103.6 534.4 12.0 52.9 10.5 85.6 413.4 121.0 and 5.5 1996 51.0 34.7 445.8 11.4 50.6 34.4 345.0 100.8 4.9 1995 41.1 14.5 271.9 59.2 15.1 69.4 24.4 (total 212.7 9.9 2.7 5.9 2.0 6.5 1994 166.9 41.3 24.0 125.6 1993 9.9 2.7 .. .. .. .. ,, .. .. education 6.9 7.7 6.1 to 1992 112.9 89.1 23.8 28.8 17.0 32.6 expenditure) 8.2 6.5 1991 76.5 58.8 17.7 10.7 46.0 24.3 36.5 support 3.0 3.7 1990 59.2 5.0 42.0 17.2 17.2 20.8 21.3 1.1 1.6 1989 43.9 2.5 32.9 11.1 10.0 15.8 14.8 Donor government (b) expenditure B.2 (billions) salaries wage Bank and total non total non-wage government Bank orld ages able W Bilateral otalT W Non-wage orld Percent Percent Percent Percent T Expenditures W Bilateral 6 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S F i g u r e B . 3 T o t a l B a n k d i s b u r s e m e n t s 45.0 80 40.0 70 35.0 60 30.0 50 25.0 40 millions 20.0 Percent US$ 30 15.0 20 10.0 5.0 10 0.0 0 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Education credit disbursements (left axis) As a percent of non-salary education expenditure (right axis) M a i n b i l a t e r a l ( a n d E C ) s u p p o r t e d T a b l e B . 3 a c t i v i t i e s i n b a s i c e d u c a t i o n Project/Program Period Budget Activities DFID Whole School 1988­2005 UK£50 million Support to 2 pilot schools in each district. Development Construction of 125 classroom blocks. EC Micro-projects 1990­91 ECU1.1 million District allocated fund for community 1991­94 ECU6.0 million activities with allocation guideline of 20% 1994­96 ECU7.0 million for education. In practice about 30% used 1996­ ? ECU9.0 million for education (classroom rehabilitation, VIP construction etc.). 1,855 projects financed over period covered by these data. Japan School block construction 2000­2002 US$0.5 million Approximately 80 classroom blocks. through Grant Assistance for Grassroots Projects USAID Primary Education Project 1990­1995 US$ 35 million US$32 million budget support plus (PREP) US$3 million TA Quality Improvements in 1997­2004 US$ 53 million US$14 million budget support Primary Schools (QUIPS) US$39 million for improvements in 330 schools (includes demand-driven infrastructure component for program schools only). 6 2 ANNEX C: SCHOOL COSTS G LSS2 and the GSS/OED collected data Presentation of Tables on school costs from three separate Tables C.1 and C.2 show the costs of fees and survey instruments: materials for primary and junior secondary 1. The school survey contained a question on schools, respectively, from the school survey. school expenses (see Box C.1). These costs are calculated as averages for urban 2. The household survey collected data on var- and rural areas, and for three ecological zones ious categories of educational expenditure for separately. The information on the amount in each child. cedis paid by households for pupils' education 3. The price questionnaire collected data on is obtained through interviews with the head- the prices of school supplies. teachers of each of the surveyed schools. These figures are not the amounts actually paid by This annex presents a summary of these data for households, but averages estimated by the head- 2003. teachers interviewed. Since fees can vary across S c h o o l q u e s t i o n n a i r e q u e s t i o n s B o x C . 1 o n s c h o o l e x p e n s e s Please tell me the amount in cedis that students have to pay for the following items. If the amount varies by grade, please tell me the average for all grades. Amount Comments 1. Enrollment fee 2. School fee 3. Sports and culture fee 4. PTA levy 5. Other fees (e.g., District Assembly levy) 6. Value of materials for practicals 6 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e C . 1 S c h o o l c o s t s : P r i m a r y s c h o o l s ( c e d i s ) Total fees Enrollment Sports & (sum of fee School fee culture fee PTA levy Other fees Materials Total cols. 1, 2, (1) (2) (3) (4) (5) (6) (7) 3, and 5) Coast Urban 13341 42813 5019 8374 4629 1319 75494 65802 Rural 1400 6640 6436 2112 3240 2120 21948 17716 Forest Urban 6801 33820 4998 9050 8633 3748 67051 54252 Rural 1835 15608 3912 3519 8589 2330 35793 29944 Savannah Urban 1615 16308 3258 5538 4850 0 31569 26031 Rural 343 5829 3834 2040 1100 3 13149 11106 Average 5884 26476 4630 6393 6558 2242 52182 43548 S c h o o l c o s t s : J u n i o r s e c o n d a r y s c h o o l s T a b l e C . 2 ( c e d i s ) Total fees Enrollment Sports & (sum of fee School fee culture fee PTA levy Other fees Materials Total cols. 1, 2, (1) (2) (3) (4) (5) (6) (7) 3, and 5) Coast Urban 10071 39869 7316 9136 6971 3107 76470 64227 Rural 0 0 7574 2553 5095 6526 21747 12669 Forest Urban 7361 29458 5244 8760 10638 6369 67830 52701 Rural 850 7150 4378 3917 4247 12608 33150 16625 Savannah Urban 1588 12882 3994 4388 8000 0 30853 26464 Rural 0 0 4126 2079 527 3211 9942 4653 Average 5358 22500 5572 6741 7239 6302 53712 40669 different grades within the same school, the average expenses per pupil calculated across head-teachers were asked to provide a figure that ecological zones and for urban and rural areas was the most representative of what was paid separately. by children of all grades attending the school. Comparing table C.1 with C.3, and C.2 with Tables C.3 and C.4 show household educa- C.4, shows some differences between reported tional expenses for children attending primary school and household costs. In the case of pri- and junior secondary schools. The information mary, the total fees of cedis 43,500 is close to on these expenses was obtained from the house- the tuition and registration fees households say hold questionnaire. Parents were asked how they pay (cedis 46,200). In the case of JSS the much they had spent during the 12 months pre- households claim to pay rather more than the ceding the interview on the items presented in schools report. This may reflect the differing the table. The expenses were reported in the composition of attendance, but may also reflect questionnaire for each child separately, thus under-reporting of fees, especially unofficial allowing the distinction between primary and JSS ones, by schools. Government policy is that (and higher levels of education). The figures are there are no fees, such as enrollment fees, in 6 4 A N N E X E S H o u s e h o l d s ' c o s t s : P r i m a r y s c h o o l s T a b l e C . 3 ( c e d i s ) Tuition & Uniforms Books & Cafeteria registration PTA levy & clothes supplies Transport & lodging fees Other Total (1) (2) (3) (4) (5) (6) (7) (8) Coast Urban 8074 42946 53320 59641 244890 89565 45653 909793 Rural 2594 23308 15364 10573 121094 23099 14549 314399 Forest Urban 10195 39693 42839 33733 237249 66674 45695 591333 Rural 2833 30845 18640 12090 146202 29468 18337 305793 Savannah Urban 9658 26216 9868 180 64216 24641 22018 207774 Rural 2641 22122 6891 618 44066 7519 6786 112898 Average 6008 33079 28729 24311 164012 46272 28400 462024 H o u s e h o l d s ' c o s t s : J u n i o r s e c o n d a r y T a b l e C . 4 s c h o o l s ( c e d i s ) Tuition & Uniforms Books & Cafeteria registration PTA levy & clothes supplies Transport & lodging fees Other Total (1) (2) (3) (4) (5) (6) (7) (8) Coast Urban 10378 64920 88480 116227 246908 75239 54513 1418049 Rural 4386 47500 88273 101409 157500 38068 40955 785227 Forest Urban 24496 53708 79375 103843 339843 74915 87407 891814 Rural 4774 31870 47496 44521 223432 40449 30615 531570 Savannah Urban 4938 19438 29250 0 82813 41813 25188 245475 Rural 4557 43273 25309 2045 64977 18109 21259 224961 Average 12055 50033 68305 80149 235719 58378 53407 876883 public basic schools. Hence, in the govern- and JSS) that were covered by the survey. The ment's view the non-zero values shown in the table displays their geographical distribution. public school section of tables C.7 and C.8 Table C.7 and C.8 are disaggregations of reflect either respondent or data entry error. Tables C.1 and C.2 for primary and private Table C.5 shows the costs of selected school schools. Similarly, tables C.9 and C.10 are dis- items. These costs were obtained through mar- aggregations of tables C.3 and C.4. ket surveys carried out in each of the selected clusters. In every locality, three different shops Discussion selling stationery and clothes were interviewed, and the average prices of each item per local- Regional Disparities ity were calculated. The figures shown are again The regional disparities in the above data are averages across ecological region and rural/urban very marked. Education costs are usually high- areas. est in the urban coastal areas and lowest in the Table C.6 shows the number of school of each rural savannah. The second most expensive type (private and public) and level (primary area is urban forest. Next is urban savannah. That 6 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e C . 5 P r i c e s o f e d u c a t i o n i t e m s School School Exercise book Pencil Eraser uniform (boys) uniform (girls) (1) (2) (3) (4) (5) Coast Urban 1111 270 253 41674 41493 Rural 1033 280 217 26875 27167 Forest Urban 977 232 239 22077 22205 Rural 1072 255 228 22800 21909 Savannah Urban 1050 267 242 24250 23417 Rural 1153 331 220 24909 24091 Average 1069 267 236 29848 29429 D i s t r i b u t i o n o f s c h o o l s b y l e v e l , T a b l e C . 6 t y p e , a n d l o c a t i o n Primary schools Junior Secondary schools Public Private Total Public Private Total Coast Urban 62 29 91 57 13 70 Rural 22 3 25 18 1 19 Forest Urban 108 35 143 86 18 104 Rural 75 22 97 54 6 60 Savannah Urban 21 5 26 14 3 17 Rural 32 3 35 19 0 19 Average 320 97 417 248 41 289 is, costs are higher in all three urban areas than Levels of Schooling all three rural areas. This result partly reflects the As is to be expected, JSS costs more than pri- concentration of private schools in urban areas. mary, though these differences are not particu- Looking at public schools only, the same pattern larly fee related (Figure C.1). The aggregate data holds for JSS, but not primary, for which costs from the school survey show similar levels of are higher in rural forest than either savannah both fees and total costs (Tables C.1 and C.2). region and for JSS quite above those of rural But the household data show that parents spend coast (Tables C.7­C.10). nearly twice as much to send a child to JSS as For primary schools in aggregate, total fees they do to primary. Expenditures are greater for in the coastal region are six times those in the JSS for all the expenses shown, but the differ- savannah (Table C.1). The difference is even ence is greatest for transport and books and sup- greater for JSS (Table C.2). The household data plies. Comparing public schools only, JSS appears show an even greater divergence, with costs more expensive according to both school-level being eight times as much for primary and nearly data and household data, with the gap being seven for JSS (Tables C.3. and C.4). large for the latter (Tables C.7­C.10). These disparities are not much explained by differences in the cost of school supplies, which Public versus private do not vary too greatly other than the greater cost Private schooling of course costs more than of uniforms in urban coastal (Table C.5). public (Tables C.7­C.10), with the differential 6 6 A N N E X E S S c h o o l c o s t s i n p u b l i c a n d p r i v a t e T a b l e C . 7 s c h o o l s : P r i m a r y ( c e d i s ) Total fees Enrollment Sports & (sum of fee School fee culture fee PTA levy Other fees Materials Total cols. 1, 2, (1) (2) (3) (4) (5) (6) (7) 3, and 5) Public schools Coast Urban 262 246 6504 10459 6675 1869 26016 13688 Rural 0 500 7223 2082 3682 2045 15532 11405 Forest Urban 792 3190 5856 9178 4055 2426 25495 13892 Rural 107 2547 4727 4124 3388 2887 17779 10768 Savannah Urban 0 952 3581 6143 52 0 10729 4586 Rural 94 0 4100 1763 1203 3 7163 5397 Average 353 1823 5483 6802 3824 1999 20282 11482 Private schools Coast Urban 39933 129367 1998 4467 468 200 176433 171767 Rural 11667 51667 667 2333 0 2667 69000 64000 Forest Urban 25343 136909 2351 8657 22761 7829 203850 187364 Rural 7727 60136 1136 1455 26318 432 97205 95318 Savannah Urban 8400 80800 1900 3000 25000 0 119100 116100 Rural 3000 68000 1000 5000 0 0 77000 72000 Average 23888 109784 1855 5163 15456 3036 159181 150982 S c h o o l c o s t s i n p u b l i c a n d p r i v a t e T a b l e C . 8 s c h o o l s : J S S ( c e d i s ) Total fees Enrollment Sports & (sum of fee School fee culture fee PTA levy Other fees Materials Total cols. 1, 2, (1) (2) (3) (4) (5) (6) (7) 3, and 5) Public schools Coast Urban 0 687 5283 10736 8791 3355 28852 14761 Rural 0 0 7550 2583 5378 6889 22400 12928 Forest Urban 4541 5905 5920 9581 5031 7057 38035 21397 Rural 37 2194 4559 4278 4459 6972 22500 11250 Savannah Urban 0 0 4171 4829 786 0 9786 4957 Rural 0 0 4126 2079 526 3211 9942 4653 Average 1596 2700 5360 7313 5182 5500 27650 14837 Private schools Coast Urban 47000 212867 2500 3267 300 2200 268133 262667 Rural 0 0 8000 2000 0 0 10000 8000 Forest Urban 20833 202544 2794 7833 4094 3083 241183 230267 Rural 8333 51750 2750 667 2333 0 65833 65167 Savannah Urban 9000 122333 3167 3667 1667 0 139833 136167 Rural ... ... ... ... ... ... ... ... Average 26907 174798 2833 4814 2260 2058 213670 206798 6 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S H o u s e h o l d s ' c o s t s i n p u b l i c a n d p r i v a t e T a b l e C . 9 s c h o o l s : P r i m a r y ( c e d i s ) Tuition & Uniforms Books & Cafeteria registration PTA levy & clothes supplies Transport & lodging fees Other Total (1) (2) (3) (4) (5) (6) (7) (8) Public schools Coast Urban 4401 20664 23627 13503 132379 13222 26460 681220 Rural 2260 21122 13450 7481 109012 13055 10302 283901 Forest Urban 6557 33019 26926 12573 156941 16479 21793 356998 Rural 2344 27054 17656 4079 139512 13095 13658 261147 Savannah Urban 4914 16978 8175 0 46860 9054 16032 160170 Rural 2425 21840 6733 656 34627 3707 7203 100814 Average 3638 24893 16987 6419 111212 11617 15588 302521 Private schools Coast Urban 11810 65612 83526 106575 359342 167224 65177 1142306 Rural 6250 47167 36250 44333 253000 132750 60917 647333 Forest Urban 16563 51371 70688 70763 377787 154515 87522 1001419 Rural 6023 55554 25052 64304 189804 136189 48830 596791 Savannah Urban 34167 73944 18611 1111 153889 105172 52944 453728 Rural 6167 26700 9467 0 197600 69533 0 309467 Average 13205 57936 64384 78640 324339 151501 67305 946354 H o u s e h o l d s ' c o s t s i n p u b l i c a n d p r i v a t e T a b l e C . 1 0 s c h o o l s : J S S ( c e d i s ) Tuition & Uniforms Books & Cafeteria registration PTA levy & clothes supplies Transport & lodging fees Other Total (1) (2) (3) (4) (5) (6) (7) (8) Public schools Coast Urban 4173 29407 47619 67272 187247 22805 39420 1452985 Rural 4325 50000 92950 93550 135750 32125 39650 754450 Forest Urban 19554 47873 66889 88542 329446 33936 65229 724693 Rural 3562 29315 41557 26215 221315 23111 33122 485705 Savannah Urban 3933 20733 28200 0 52333 14467 23867 188373 Rural 4663 39628 23828 2093 66488 8088 21753 213030 Average 8269 36730 50217 53625 206044 24018 41845 770316 Private schools Coast Urban 23605 140618 175579 220579 374079 187008 86684 1343579 Rural 5000 22500 41500 180000 375000 97500 54000 1093000 Forest Urban 40904 73080 120826 154640 374360 210965 161040 1446655 Rural 14625 52625 95750 193250 240625 181325 10250 904225 Savannah Urban 20000 0 45000 0 540000 452000 45000 1102000 Rural 0 200000 89000 0 0 449000 0 738000 Average 27555 104487 142342 188720 357187 199027 100733 1313096 6 8 A N N E X E S Z e r o a n d m a x i m u m r e s p o n s e s i n T a b l e C . 1 1 p r i m a r y s c h o o l c o s t d a t a No. of schools Maximum value for which zero (cedis) 1. Enrollment fee 233 (253) 40,000 2. School fee 223 (254) 190,000 3. Sports and culture fee 9 (317) 40,000 4. PTA levy 57 (312) 150,000 5. Other fees 79 (306) 120,000 (e.g., District Assembly levy) appearing larger from the school data than the lic primary schools 233 reported charging no household data. This is since items not included enrollment fee compared to a maximum charged in the school data -- uniforms, transport, and in one school of 40,000 cedis (Table C.11). cafeteria/lodging -- have a much lower differ- There are 57 schools not even charging a PTA ential. levy, compared to a maximum of 150,000. The table shows in brackets the number of responses Funds available to schools to each question. It cannot be assumed that The sports and culture levy is the only official non-responses are zero, since zero was an school fee set by government. In addition, the accepted response. It is just as likely that respon- PTA levy is intended to benefit the school. Fig- dents were reluctant to provide the information ure C.1 shows the average of fees other than the since such fees are not meant to be charged. This sports and culture levy, these being amounts would mean that the school costs under-estimate likely to be retained by the school. From this fig- charges. The household data show higher fees ure, the differentials seem quite large. In fact, than the school data for JSS, though they are sim- they are even larger. For example, of 317 pub- ilar for primary. 6 9 ANNEX D: SCHOOL-LEVEL CHANGES IN INPUTS, MANAGEMENT, AND METHODS Introduction lected from the teacher roster in the school This annex reports the data from the school questionnaire. and teacher surveys on school quality. Tabula- The next section discusses the variables to be tions are presented on variations of school qual- used in more detail, with subsequent sections dis- ity by zone (coastal, forest, and savannah), cussing in turn monetary (capital and recurrent) rural/urban, and the economic well-being of inputs, teacher quality, methods, morale, and the community in which the school is situated. school management. For some variables regression results are pre- sented to explore the determinants of different Four Dimensions of School Quality: aspects of school quality. Insofar as the data per- Variable Selection mit, comparisons are made with 1988. These The school survey from 1988 and the school and comparisons are reported in two ways: (1) com- teacher surveys provide a number of variables parisons of the whole sample of 519 schools in that can be used to measure school-level inputs 1988 with the 706 surveyed in 2003, and (2) com- to the education process. Four dimensions of parison based only on the panel of 196 schools school-level inputs can be identified: that could be matched between the two survey · Physical rounds.88 A teacher survey was not carried out · Material in 1988, although limited information was col- · Teachers · School management Table D.2 identifies the variables that can be S a m p l e T a b l e D . 1 used to measure the quantity and, where pos- s i z e s sible, quality, of these various inputs. Few data on the last dimension, school management, 1988 were collected in 1988, so that comparisons Primary schools 286 across time are largely restricted to the first three Middle/JSS 233 dimensions. The subsequent sections of this Total 519 annex present the main findings with respect to 2003 each dimension. This section provides an Primary schools 417 overview of the variables to be used. JSS 289 Most of the variables are self-explanatory. Total 706 Explanation is provided here for those that are not (Annex F provides a variable list with defi- Teachers 3,129 nitions): Memo item: 1. Adequate classrooms and adequate number matched schools of teachers are both based on dividing the Primary schools 128 actual number available by the required num- JSS 68 ber. The required number is the number of Total 196 classes taught in the school, taking into 7 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S u r v e y - b a s e d m e a s u r e s o f s c h o o l T a b l e D . 2 q u a l i t y Quantity Quality Physical Classrooms Adequate number of classrooms % of classrooms that can be used when raining Classes held in shared classrooms* Height of internal walls* Noise disruption* Chalkboard % of classrooms with chalkboard Board quality Water % of schools with own water supply Type of Water/Storage Library % of schools with library Material Chalk Availability English textbooks English Textbooks -- Pupil Ratio Textbook usage* Mathematics textbooks Mathematics Textbooks -- Pupil Ratio Textbook usage* Desks Writing Places -- Pupil Ratio Chairs Seating places -- Pupil ratio* Teachers and teaching methods Teachers Adequate number of teachers Teacher morale (subjective)* % trained teachers Absenteeism Teacher test scores Head's assessment* Teaching methods* Frequency of homework* School management Community involvement Existence SMC and PTA* Active SMC and PTA* SPAM* Role of SPAM* Circuit supervisor Frequency of visits Activities of circuit supervisor* Head teacher Activities of head teacher* Note: * not collected in 1988. account multiple streams and shifts. If a pri- Detached head teachers are excluded from the mary school teaches grades 1 to 6 and has two numerator. streams without shifts it needs 12 classrooms 2. The percentage of classrooms that can be and 12 teachers.89 But if that school operated used when it is raining is 100 less those that a shift system it would need only 6 classrooms cannot be used when it is raining plus those but still require 12 teachers. Classrooms in that cannot be used at all divided by total such poor state of repair that they cannot be number of classrooms (including those that used are excluded from the number available cannot be used at all). (the numerator). The 2003 data contain addi- 3. Board quality is a subjective assessment. In tional variables on classroom quality. 1988 this assessment was made by the 7 2 A N N E X E S respondent (usually the head) for the whole · Today 94 percent of schools have a black- school. The same question was asked in 2003 board in every classroom compared to 78 per- and is used for the purposes of comparisons. cent 15 years ago However, in 2003 teacher-level data on board · Fifteen years ago over two-thirds of primary quality (and more specific questions on that schools reported occasional shortages of quality) are also available and used for analy- chalk, only one in twenty do so today, with sis specific to 2003. 86 percent saying there is always enough 4. Textbook availability. The number of books · The percentage of primary schools having at at each grade is summed across grades and least one English textbook per pupil has risen divided by total enrollments, which is equiv- from 21 percent in 1988 to 72 percent today alent to an enrollment-weighted average for and for math books in JSS these figures are the school.90 13 and 71 percent, respectively. 5. Writing places. For 2003 desks per pupil is calculated in an analogous way to textbook Despite the greatly improved school quality, availability, but using adjusted enrollments variation remains across the country, with some where the adjustment takes account of a "biases" in the allocation of school resources. split shift (i.e., the same desk can be used Analysis reported below shows that the strongest by different pupils in morning and after- bias comes from the ability of better-off com- noon). However, for 1988 there is a cate- munities to better support schools in their local- gorical variable on whether there are enough, ity. It is also shown that the Bank's Primary some, or no desks. The 2003 data are cate- School Development Project made a significant gorized for comparability (see Annex F for contribution to aspects of school quality. cut-offs). 6. Teacher morale is a subjective measure based Material Inputs on two questions, whether the respondent The material inputs for which data can be com- enjoys being a teacher, and if he or she plans pared between 1988 and 2003 are the availabil- to remain a teacher for their whole career. The ity of chalk, math and English books, and desks. head's assessment is a categorical classifica- For each of these four variables there has been tion of all teachers as very good, good, aver- a strongly significant (all significant at 1 percent age, poor, or very poor. None of these data level) improvement in the level of inputs at both were collected in 1988. primary and JSS level (see Table D.3). This state- 7. Teaching methods were assessed through ment is also true for the panel of 196 schools. three questions in the teacher questionnaire The four variables were combined into a sim- designed to assess the extent to which the ple index of material inputs. The resulting fig- teacher claims to use "improved methods," ures are shown in Figure D.1.91 including a check on their knowledge of The figures show the cluster level average of these methods (more detail is provided the material input index for 1988 and 2003, cal- below). culated separately for primary and middle/JSS. In each graph, the clusters have been ranked Physical and Material School-Level Inputs according to the value of the index in 1988, so that the clusters with the schools with the fewest The Main Message: School Quality Has material inputs appear to the left of the scale. Improved Where the line for 2003 lies above that for 1988 The main message from the school survey is the there has been an increase in the material input overwhelming improvement in school quality. index for that cluster. Two points jump out from For example: these graphs: · In 1988 less than half of schools could use all · There has been a substantial increase in the their classrooms when it was raining, but in level of material inputs across the country, 2003 over two-thirds can do so. especially in primary schools. In only two 7 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S i g n i f i c a n t c h a n g e s i n t h e T a b l e D . 3 a v a i l a b i l i t y o f m a t e r i a l i n p u t s Primary Middle/JSS Total Chalk *** *** *** English books *** *** *** Math books *** *** *** Desks *** *** *** Note: *** significant at 1%, ** significant at 5%, * significant at 10%. Significance is based on chi-squared statistic based on cross-tabulation of categorical version of variable against year (1988 and 2003). See Annex E for cross-tabulations and Annex F for variable definitions. clusters (which had the maximum value of 1 improvement of school quality has been con- in 1988) has the level of material inputs centrated in the most disadvantaged schools. In declined at the primary level (Table D.4). the general sample it has been seen that 86 per- For middle/JSS there have been, mostly small, cent of schools today say there is enough chalk declines in nine of the 76 clusters.92 compared to 67 percent 15 years ago. In Table · The improvement has been greatest the lower D.6 most the observations lie on the upward the initial level of the index, meaning that the sloping diagonal from the bottom left. All schools clusters in which schools that were the most that said there was never enough chalk in 1988 deprived have seen the largest improvements today have enough. Of the 102 suffering occa- in material inputs (Table D.5).93 sional shortages 15 years ago, 88 (86 percent) now always have sufficient. And all but 3 of those The panel data for chalk availability in primary which had sufficient in 1988 still do so today. schools provides a clear example of how the Simply put, the 28 schools already having suf- S c h o o l s i n n e a r l y a l l a r e a s h a v e m o r e m a t e r i a l i n p u t s t h a n b e f o r e : F i g u r e D . 1 c l u s t e r - l e v e l m a t e r i a l i n p u t s t o s c h o o l q u a l i t y (a) Material Primary (b) Material Middle/JSS 1.0 1.0 0.8 0.8 0.6 0.6 index index input input 0.4 0.4 Material Material 0.2 0.2 0.0 0.0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Cluster Cluster 1988 2003 2003 (public only) 1988 2003 2003 (public only) Source: GLSS2 and GSS/OED school survey 7 4 A N N E X E S S u m m a r y o f o b s e r v a t i o n s i n T a b l e D . 4 F i g u r e s D . 1 a n d D . 2 Primary Middle/JSS Number of clusters experiencing deterioration Material 2 9 Physical 22 31 Number of clusters in which public schools have lower quality than private schools Material 19 13 Physical 29 17 Memo: Number of clusters with private schools in 2003 41 24 C o r r e l a t i o n c o e f f i c i e n t b e t w e e n T a b l e D . 5 c l u s t e r - l e v e l c h a n g e i n i n d e x a n d t h e i n i t i a l ( 1 9 8 8 ) v a l u e Material Physical Primary ­0.88 ­0.65 Middle/JSS ­0.71 ­0.65 C h a n g e i n c h a l k a v a i l a b i l i t y T a b l e D . 6 a g a i n s t i n i t i a l a v a i l a b i l i t y f o r p r i m a r y s c h o o l p a n e l Change in rating Never Occasional Always (2003­1988) enough (=1) shortages (=2) enough (=3) Total ­2 0 0 2 2 ­1 0 7 1 8 0 0 7 28 35 1 0 88 0 88 2 16 0 0 16 Total 16 102 31 149 ficient chalk had no room for improvement. 2003. Hence it might be argued that the observed But the 16 who never had enough could at increase in school quality simply results from the worst stay the same -- though in the event all better quality of private schools. This is not so. those 16 now report having sufficient supply. Figure D.1 also shows the material input index The share of private schools in the sample for 2003 calculated for public schools alone. In increased from 5 to 20 percent between 1988 and general this line is not far removed from the over- 7 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S all cluster average. Indeed it is below it, indi- primary schools, and another (classrooms that cating that public schools have a superior level cannot be used when raining) for middle/JSS of material inputs than do private ones in 22 of (Table D.7). The lack of change with respect to the 41 clusters that have public schools (Table there being sufficient classrooms shows that D.4). When the changes in the index and its com- classroom building has kept pace with growing ponents are calculated for public schools only student numbers. So the number of classrooms these changes all remain significant at the 1 has increased, but been matched by more stu- percent level (Annex E, Tables E.6­E10). dents. For many schools there has been no Private schools do have a higher level of shortage of physical facilities (although their material inputs in some respects. Although there quality is a different matter), so that no improve- is no overall significant difference in the mate- ment in this measure is expected or required. rial input index for public and private schools Overall, there has been a significant increase in in 2003 this result conceals that private schools the index of physical inputs. do significantly better on two of the four com- Figure D.2 shows the change in physical ponents of the index (English books and desks), inputs in the same way as Figure D.1 showed but significantly worse on two (chalk and math material inputs.94 There are now many more books) (Annex E, Tables E.15­E.19). Private clusters, though still well less than half (Table schools on average have higher levels of some D.4), which have not experienced an improve- inputs than do public ones, but this is not the ment. Once again, although private schools do main reason for the observed improvement in perform better in some respects, their increase school quality between 1988 and 2003. does not account for the improvement in school quality that has taken place. Physical Inputs In 2003, private schools had superior inputs The index for physical inputs comprises the with respect to the percentage of classrooms that adequacy of the number of classrooms, the pro- could be used when raining and having their portion that can be used when raining, the pro- own water supply. They also had slightly bet- portion with a blackboard and the quality of ter average quality chalkboards, although the dif- those boards, the presence of a library and own ference is not quite statistically significant (prob water supply. Two of these have not improved value=0.11). There is no difference with respect (sufficient number of classrooms and library) for to having sufficient classrooms, chalkboards, or either type of school, one (library) has not for a library. S i g n i f i c a n t c h a n g e s i n p h y s i c a l T a b l e D . 7 i n p u t s t o s c h o o l q u a l i t y Primary Middle/JSS Total Adequate classrooms -- -- -- Classrooms which cannot be used when raining *** -- *** Percentage of classrooms with a chalkboard *** *** *** Chalkboard quality *** *** *** Own water supply *** *** *** Library -- ** ** Note: *** significant at 1%, ** significant at 5%, * significant at 10%, and -- indicates no significant change. Significance is based on chi-squared statistic based on cross-tabulation of categorical version of variable against year (1988 and 2003). See Appendix E for cross-tabulations and appendix F for variable def- initions. 7 6 A N N E X E S T h e q u a l i t y o f s c h o o l i n f r a s t r u c t u r e h a s i m p r o v e d i n F i g u r e D . 2 m o s t a r e a s : c l u s t e r - l e v e l p h y s i c a l i n p u t s t o s c h o o l q u a l i t y (a) Physical Primary (b) Physical Middle/JSS 1. 0 1.0 0. 8 0.8 0. 6 0.6 index index input input 0. 4 0.4 Physical Physical 0. 2 0.2 0.0 0. 0 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 Cluster Cluster 1988 2003 2003 (public only) 1988 2003 2003 (public only) Source: GLSS2 and GSS/OED school survey Allocation of School Inputs in water supply, which is clearly linked to their Tables D.8 and D.9 summarize some distortions location, and their relatively lower likelihood of in the allocation process. Looking first at income having a library. Previous biases, notably in (Table D.8), a difference can be seen with respect school infrastructure, have been rectified. to material and physical inputs. In the former case, However, in the case of regional allocation the allocation biases that existed in 1988 have been opposite appears to be the case, biases having reduced or eliminated except for the number of appeared in the allocation of both physical and desks per student. However, in the case of phys- material resources. In general, where such bias ical inputs the bias has continued for three of the exists, then the coastal region is most well pro- six measures, and emerged in one case where it vided for. The exception is for math books, for was not previously present. Only in one case has which the forest region has the greatest avail- it been reduced and in one eliminated. Based on ability. In all cases, the savannah zone is the least the chi-squared statistic, the bias in allocation well resourced, except for the mild exception of has increased for physical inputs, while it has the presence of a library. But in general, the for- declined for material ones. While these biases est and coastal regions are relatively close in should be a source of concern to policymakers, resource availability, with schools in the savan- they must be seen in the context of the overall nah region trailing behind. rise in the level of inputs that has taken place across virtually the whole country. More on variations in monetary school inputs Table D.9 incorporates table D.8 and presents According to broad aggregate measures, biases biases in allocation by zone that may underlie in the allocation of resources have lessened. these income biases (since non-coastal zones, par- Yet Figures D.1 and D.2 show considerable vari- ticular savannah, are poorer, as are rural ones). ation in the community-level averages. This sug- Considering first biases against rural areas, these gests that the differences in the levels of school are much less today than they were 15 years ago. inputs are not well explained by broad aggre- The only significant biases against rural areas are gate categories such as rural/urban or zone. 7 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S A l l o c a t i o n o f e d u c a t i o n r e s o u r c e s T a b l e D . 8 b y e x p e n d i t u r e q u i n t i l e ( p u b l i c s c h o o l s o n l y ) 1988 2003 Comment Physical Adequate number of Classrooms No bias Significant (5%)1 bias Allocation bias emerged against poorer clusters Classrooms cannot be used Significant (1%) bias Significant (10%) bias Reduced bias against when raining against poorer clusters against poorer clusters poorer clusters Chalkboard Significant (1%) bias No bias Allocation bias eliminated against poorer clusters Board quality Significant (10%) bias Significant (10%) bias Continued bias against against poorer clusters against poorer clusters poorer clusters Library Significant (10%) bias Significant (5%) bias Continued bias against against poorer clusters against poorer clusters poorer clusters Water Significant (1%) bias Significant (1%) bias Continued bias against against poorer clusters against poorer clusters poorer clusters Material Chalk Significant (1%) bias Significant bias (10%) in Allocation bias eliminated against poorer clusters favor of poorer clusters English books Significant (5%) bias in Significant (10%) bias in (Progressive) allocation favor of poorer clusters favor of poorer clusters bias reduced Math books No bias No bias (Progressive) allocation bias reduced2 Desks Significant (1%) bias Significant (1%) bias Continued bias against against poorer clusters against poorer clusters poorer clusters Adequate number of Teachers Significant (5%) bias Significant (5%) bias Continued bias against against poorer clusters against poorer clusters poorer clusters Notes: 1. The percentage in each cell, where shown, is the level of significance of the chi-squared statistic in the cross-tabulation against income quintile. 2. Although there is no significant relationship in either period the prob value is 0.102 in 1988, with the allocation favoring poorer clusters, and 0.654 in 2003. This view is supported by an analysis of variance, opment (WSD) program supported by DFID. which finds that the variation within zones is sig- Data on support from WSD and QUIPS, which nificantly greater than that between them for both are ongoing, were collected in the school ques- indices, and all their component parts with just tionnaire. PSD beneficiary schools within the two exceptions (English books and classrooms GSS/OED sample were identified from a list of that can be used when it is raining). Two pos- all beneficiary schools. There are 19 of the lat- sible explanations are pursued here for the large ter in the panel of 196 schools, 14 WSD schools within-area variations: the role of projects and and just 5 who have benefited from QUIPS. community support. Bivariate analysis of the panel data shows that There have been three major project initiatives PSD (Table D.10) is associated with significantly providing direct support to primary schools for larger improvements in the proportion of class- infrastructure and supplies: the World Bank's Pri- rooms that can be used when it is raining, avail- mary School Development project (PSD), ability of desks, and quality of blackboards -- USAID's QUIPS, and the Whole School Devel- as well as with the physical input index. Since 7 8 A N N E X E S B i a s e s i n t h e a l l o c a t i o n o f e d u c a t i o n a l r e s o u r c e s , 1 9 8 8 a n d T a b l e D . 9 2 0 0 3 ( b i a s i s a g a i n s t p o o r e r h o u s e h o l d s a n d r u r a l a r e a s u n l e s s o t h e r w i s e n o t e d ) Income Region Rural/urban 1988 2003 1988 2003 1988 2003 Physical Adequate number of Classrooms -- ** -- *** *** -- Classrooms cannot be used when raining *** * -- -- *** -- Chalkboard *** -- -- ** *** -- Board quality * * -- *** * -- Library * ** *** *** -- * Water *** *** *** *** *** *** Material Chalk *** *1 -- *** -- **2 English books ** * -- -- ***2 -- Math books -- -- -- * *2 -- Desks *** *** *** *** * -- Adequate number of Teachers ** ** -- *** *** *** Explanatory note: The more stars the more significant the bias in resource allocation. A -- indicates no significant bias. The bias is against poorer, non-coastal and rural communities unless otherwise stated. Significance is based on the chi-squared statistic calculated from the bivariate cross-tabulation of the school measure against each characteristic (income quintile, region, rural/urban) in turn. Notes: 1. in favor of poorer communities; 2/ in favor of rural communities. PSD's main activity was the financing of new A second approach to the analysis is to esti- classroom blocks (which don't leak and have mate regressions for the school input variables good blackboards), usually to replace old ones, using the 2003 data only. In addition to the and to provide desks for these blocks, these project dummy variables, measures are also results make sense. Many classrooms in Ghana entered of community well-being (per capita are in pavilions, that is, a roof on supports but expenditure), the level of fees, if support has with low or no wall. These cannot be used dur- been provided by the PTA or SMC, and the ing heavy rain. In the 1980s, these were often value of the PTA levy. Dummies are entered for community-made structures from tree trunks/ zone, rural/urban, and private schools. The large branches and thatch. PSD replaced these results are summarized in Table D.11. with concrete and corrugated iron pavilions. The main points to emerge are as follows: The construction of walls (cladding) was left to · The regression models do not explain the dis- the community as their contribution, though tribution of textbooks; none of the variables this was not always done. are significant in the model for math books By contrast, QUIPS appears to have had a per- and only the primary dummy in that for Eng- verse impact on the change in some school lish books. On the other hand, financial inputs, though the sample size is very small resources do matter for desks and, to a lesser and endogeneity is the likely cause of these extent, chalk availability. results.95 No impact appears from WSD. These · Financial variables matter for physical inputs, findings are not surprising in light of the fact that being highly significant for the index as a neither project has focused on hardware and are whole and some of them at least being sig- both of more recent vintage and still ongoing. nificant for classrooms that can be used when 7 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S I m p a c t o f p r o j e c t s o n c h a n g e i n T a b l e D . 1 0 s c h o o l i n p u t s ( p a n e l d a t a ) PSD QUIPS WSD Material inputs Chalk .. .. .. English books .. .. .. Math books .. .. .. Desks +ve (***) ­ve (*) .. Index .. .. .. Physical inputs Sufficient classrooms .. .. .. Use classrooms when raining +ve (***) ­ve (*) .. Chalkboard .. .. .. Chalkboard quality +ve (***) ­ve (**) .. Library .. .. .. Water .. .. .. Index +ve(**) ­ve (**) .. Note: *** significant at 1% level, ** 5% level and * 10 % level. it is raining, chalkboard quality, library, and School Management water supply. The PTA levy is never signifi- cant, but is correlated with other financial vari- Supervision by Head Teacher and Circuit ables. Moreover, the school-level data reflect Supervisor the levy set, not the amount actually col- The focus on software means an increased focus lected in additional contributions.96 on issues of school management. Data on two · The positive impact of the PSD project on aspects of school management are available for classrooms that can be used when it is rain- both 1988 and 2003: the frequency of visits of ing is supported by this analysis, as is its the circuit supervisor (formerly school inspec- impact on the number of classrooms. tor) and the presence of the PTA.97 The former, which has increased over time, is shown later These findings indicate that the level of inputs to be a significant determinant of the quality of to schools reflects the economic well-being of teaching methods. The latter variable is not very the surrounding community -- directly through revealing, since virtually all schools have a PTA, the level of fees they can afford, the level of the although some private ones do not. What mat- PTA levy, and the likelihood of help from the ters is the extent to which the PTA supports the PTA or SMC. On top of these, the community's school. In 2003, considerably more data were col- level of expenditure matters, presumably pick- lected on the activities of the circuit supervisor ing up other channels through which support is and the head teacher in supervising the work of provided. However, the wealth of the commu- teachers. Data were also collected on the sup- nity does not matter for textbooks, and matters port provided by the PTA and the workings of less for chalk, since these are things provided the SMC and SPAM. centrally through GES. It does matter for desks, which are increasingly likely to be provided by Frequency of Supervision Visits the district, whose resource availability depends In 1988, circuit supervisors visited schools just on that of the population's income. over once every two months on average (Table 8 0 A N N E X E S sig- *** *** ** * .. .. ** .. .. .. .. .. .. Index individually is inputs, neither ater *** *** ** .. .. .. * .. .. .. *** W together entered * .. .. school ** .. .. .. .. .. .. .. .. .. Library when, separately *** ** .. .. .. .. .. .. .. * .. .. *** Physical quality entered Chalkboard were help monetary .. .. .. .. .. .. ** .. .. .. (­ve) .. .. SMC *** of and Chalkboard A PT (3) be when .. * .. .. .. .. .. .. .. .. ** (­ve) .. 10%; * can raining *** level and Classrooms used 5% ** of .. .. .. .. .. .. 1%, ** .. .. .. ** .. .. at Sufficient classrooms significant *** .. .. .. .. * * .. .. .. .. .. .. *** (2) Index negative, are * ** .. .. .. .. .. .. .. .. .. .. .. which Desks determinants dummies, .. .. area .. .. .. .. .. .. .. .. .. .. .. Math the books logged. Material is except variable stated, .. .. .. .. .. .. .. .. .. ** .. .. .. English books Significant 2003 otherwise expenditure the (4) unless * .. .. .. * ** .. .. (­ve) .. .. .. and *** Chalk *** positive jointly; are so D.11 A are PT they fees coefficients from from levy type capita all able A dummies (1) though T expenditure SMC School Per Help Help PT Forest Savannah Rural Private Primary PSD QUIPS WSD Financial resources Area School dummies Project dummies Notes: nificant 8 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S F r e q u e n c y o f s c h o o l v i s i t s b y T a b l e D . 1 2 c i r c u i t s u p e r v i s o r / s c h o o l i n s p e c t o r , 1 9 8 8 a n d 2 0 0 3 ( p e r c e n t ) 1988 2003 Primary Middle Primary School school/JSS School JSS Twice a year or less 25.2 21.0 16.3 11.1 Between three and six times a year 37.4 44.2 39.6 41.2 Between every one to two months 21.3 18.9 21.6 31.8 Once a month or more 16.1 15.9 22.5 15.9 Total 100.0 100.0 100.0 100.0 Memo items Sample size 282 230 417 289 Mean number of visits 6.40 6.30 9.07 7.89 D.12). But there was considerable variation allow for sample selection (Heckman).98 The around this average, with over a fifth of middle selection equation is a probit model of the fac- schools/JSS and a quarter of primary schools tors affecting whether a school is visited at all. receiving only two visits or less a year. By 2003 Private schools are less likely to receive any the mean number of visits rose, from 6 to 9 a visits at all, and those that do have fewer of them. year for primary schools and a bit less for JSS. The same is true of rural schools. Conversely, And the proportion of schools receiving infre- large schools are more likely to be visited and quent visits fell to 11 percent for JSS and 16 for to receive more visits. Finally, there has been an primary. "autonomous shift" with schools more likely to Perhaps surprisingly, bivariate analysis does be visited, and to have more visits, in 2003 than not reveal any significant difference between in 1988. The zone dummies are not significant. rural and urban schools (Table D.13). Rural schools were less likely to be among either the Activities of the Head Teacher and Circuit least visited or the most visited, more of them Supervisor falling into the category "between three and six Table D.15 reports results on the engagement of times a year" than is the case for urban schools. the head teacher and circuit supervisor with However, there is a pattern regarding ecologi- teachers in five supervision activities. The major- cal zones. Schools in the forest zone are more ity of head teachers are, according to the teach- likely to be visited by circuit supervisors. This ers in their schools, actively involved in the was the case in both 1988 and 2003. In addition, different types of supervision. Notably, less than in 1988 schools in the coastal zone were visited 5 percent of teachers say that the head teacher less frequently than those in the savanna area, does not look at their lesson plans on a regular whereas in 2003 the situation was reversed. basis (meaning at least once a week, which is Finally, there were few private schools in 1988 the frequency with which it is expected to be and the difference in supervision rates was not done). However, less than half say that the head significant. But by 2003 there is a large gap, with actually discusses the lesson plan with them. 45 percent of private schools receiving infrequent Somewhat higher percentages look at samples supervision visits. of students' work and sits in class at least once These differences are partly supported by a week. multivariate analysis (Table D.14). Since there are While the large majority of schools have vis- 49 schools receiving no visits at all the esti- its from the circuit supervisor, 44 percent of mated model is a two-part estimation model to teachers have no direct contact with him or her 8 2 A N N E X E S B i v a r i a t e a n a l y s i s o f f r e q u e n c y o f T a b l e D . 1 3 c i r c u i t s u p e r v i s o r / s c h o o l i n s p e c t o r v i s i t s , 1 9 8 8 a n d 2 0 0 3 ( p e r c e n t ) Urban Rural Coastal Forest Savannah Public Private All 1988 Twice a year or less 27.4 19.9 32.7 13.4 26.9 23.0 29.6 23.4 Between 3­6 times a year 37.1 43.3 39.6 43.8 34.4 40.3 44.4 40.4 Between 7­11 times a year 18.1 22.0 14.4 21.9 29.0 20.2 18.5 20.0 Once a month or more 17.3 14.9 13.4 21.0 9.7 16.5 7.4 16.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 237 282 202 224 93 491 27 519 2003 Twice a year or less 14.4 13.7 14.6 14.4 12.4 6.7 44.9 14.0 Between 3­6 times a year 36.4 47.1 35.1 39.6 53.6 41.2 36.2 41.5 Between 7­11 times a year 27.9 22.0 32.7 23.0 22.7 29.2 11.6 24.8 Once a month or more 21.3 17.3 17.6 23.0 11.3 22.9 7.2 19.7 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 451 255 205 404 97 568 138 706 (Table D.15). The reported figures suggest that Involvement of the Community: the PTA, SMC, those that do have contact with teachers carry and SPAM out the full range of activities shown in the The Parent Teacher Association (PTA) provides table, though not much more than half discuss a means by which parents can support the career development. schools attended by their children usually finan- Table D.15 also reports tests of significance cially but also by providing help in kind. Virtu- for these variables for teachers in schools sup- ally all schools have a PTA. Over 99 percent of ported by QUIPS and WSD (the full results are public basic schools did so in 2003, as do 94 per- given in Annex E, Tables E.47 and E.48). There cent of private schools (Table D.16). Since PTAs are only 2, out of a possible 10, significant are so widespread, statistical analysis will not be results for schoolteachers. However, it is worth able to pick up any effect they may have on remarking that no teachers at all in QUIPS and school processes and outputs. However, it is not WSD schools reported that head teachers do not the mere presence of a PTA that will make the look at their lesson plans. By contrast with the difference, but the extent to which it provides results for head teachers, there are seven sig- support to the school. There is considerable nificant results for circuit supervisors -- four out variation in the extent to which PTAs have pro- of five for WSD, and the case that is not sig- vided support to schools and in the value of par- nificant is only marginally not so. However, ents' monthly contributions (see below). legitimate questions can be raised about the SMCs are also widespread, being present in direction of causation since both projects are over 80 percent of the schools surveyed (Table have begun work first in districts considered to D.17). However, in only half of schools had have the necessary capacity to administer the SMCs met in the preceding month or provided project. support in the past year, and in even fewer 8 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S M u l t i v a r i a t e a n a l y s i s o f t h e T a b l e D . 1 4 n u m b e r o f v i s i t s f r o m t h e c i r c u i t s u p e r v i s o r Coefficient z-stat Number of visits Rural ­0.95 ­1.86 ** Primary 0.64 1.33 Private ­5.65 ­7.75 *** 2003 3.01 6.09 *** Number of teachers 0.13 2.36 ** Intercept 8.07 5.94 *** Selection equation Rural ­1.86 * Forest 0.00 0.00 Savannah 0.00 0.00 Number of teachers 0.02 2.36 ** Private ­0.70 ­7.55 *** 2003 0.37 6.00 *** Intercept 0.96 5.55 *** /athrho 13.38 0.89 /lnsigma 2.09 101.82 Number of obs 1218 Censored obs 49 Uncensored obs 1169 Log likelihood = ­4144 Note: *** significant at 1%, ** 5% and * 10%. helped the school in dealings with outside agen- ents should ensure children attend school (41 per- cies. Where the school had not asked for sup- cent) and parents should provide pencils and port from the SMC this was often because it was exercise books (38 percent). The most common felt that the SMC would be either unwilling or actions for teachers were to provide extra classes unable to help (48 percent of cases for primary (33 percent) and to be punctual (17 percent). and 39 percent for JSS). The lower prevalence Problems of absenteeism were mentioned in of SMCs than PTAs is largely explained by the less than 10 percent of cases.99 Responsibility for fact that they are not required at private schools: implementation of the action plan was seen to over 90 percent of public schools have SMCs. But rest with the head teacher (47 percent of cases) the facts remain that these SMCs are not active or the circuit supervisor (24 percent). In only 20 in a large number of schools. For most of the percent of cases were the PTA or SMC said to be questions asked, the PTA was seen as a more responsible. Finally, in only 6 percent of cases supportive organization. was it said that the planned actions were not Virtually all public primary schools (92 percent) being carried out at all, and in 42 percent they have had a SPAM, at 98 percent of which an were claimed to be being carried out completely. action plan was agreed. The most common The school survey suggests there is little dif- actions agreed at the SPAM were (remembering ference between rural and urban areas with that it is a teacher replying to the survey) that par- respect to any of the variables shown in Table 8 4 A N N E X E S H e a d t e a c h e r a n d c i r c u i t T a b l e D . 1 5 s u p e r v i s o r s u p e r v i s i o n s Head teacher Circuit supervisor % in all schools doing so % in all on a schools regular doing so QUIPS WSD basis1 QUIPS WSD at all Sits in on class .. .. 61.3 * ** 41.4 Looks at a sample of students' work ** .. 52.4 .. ** 43.6 Looks at lesson plans .. ** 95.4 *** .. 51.3 Discusses lesson plans .. .. 45.7 ** ** 41.7 Discusses career development .. .. 57.2 .. *** 27.4 Memo item: Percent of teachers responding "Yes" to question whether they had had direct contact with the circuit supervisor n.a. n.a. 70.3 62.5 55.7 Notes: *** significant at 1% level, ** 5% level and * 10 % level. 1/ Regular is at least once a week for all categories other than career development, which is if head teacher ever does so. D.17. If anything, school respondents reported mon that the school-level data suggests it should PTAs to be more active in urban areas, though be, and participation rates correspondingly low. there was no difference for SMCs. But the data Only 6 percent of households say that someone collected from households give a different pic- attended a SPAM at their child's school. ture. Households with children in basic school But while rural communities may be easier to were asked if there was a PTA and SMC at the mobilize in support of schools, they also tend child's school and if any household member was to be less well off, reducing their ability to pro- a member of the organization. They were also vide financial support. Table D.19 reports regres- asked if there had been a SPAM at the child's sion results from the analysis of average PTA school and if any household member had contributions per pupil at the community-level.100 attended. The rates for rural households are sig- The elasticity of PTA contributions with respect nificantly higher than those for urban households to community income is close to two. That for all six questions (Table D.18). Knowledge and means that doubling community income participation in PTAs is widespread. However, increases the value of contributions to schools knowledge of SMCs and the SPAM is far less com- nearly threefold. In 2003, the richest community P r e s e n c e o f P T A a t p u b l i c a n d T a b l e D . 1 6 p r i v a t e s c h o o l s , 1 9 8 8 a n d 2 0 0 3 1988 2003 Public Private Total Public Private Total Yes 96.3 96.2 96.3 99.1 95.7 98.4 No 3.7 3.8 3.7 0.9 4.3 1.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 488 26 514 568 138 706 8 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S c h o o l m a n a g e m e n t o r g a n i z a t i o n s , T a b l e D . 1 7 2 0 0 3 PTA SMC Primary JSS Primary JSS Organization associated with school 97.8 99.3 81.06 85.8 Organization met in the last month 53.4 51.1 50.3 42.3 School asked for or got support from organization in the last year 87.3 88.9 38.1 61.6 Organization provided support in the last year 63.7 66.6 50.0 46.8 Organization helped with dealings with district or outside agencies 40.7 45.3 44.1 44.0 H o u s e h o l d k n o w l e d g e o f a n d p a r t i c i p a t i o n i n s c h o o l m a n a g e m e n t T a b l e D . 1 8 o r g a n i z a t i o n s ( h o u s e h o l d s w i t h c h i l d r e n i n p u b l i c b a s i c s c h o o l o n l y , p e r c e n t ) Urban Rural Total Member/ Member/ Member/ Exists Attend Exists Attend Exists Attend PTA 92.7 96.4 97.5 *** 96.7 *** 95.1 96.5 SMC 42.5 5.0 59.3 *** 11.3 *** 51.0 8.2 SPAM 19.5 3.2 27.6 *** 9.1 *** 23.6 6.2 Note: *** significant difference between rural and urban at 1% level. in the sample was more than five times richer ents appreciate a good headmaster; (2) a good than the poorest, suggesting that schools in the headmaster is also one who is active in solicit- former will receive 15 times as much money ing support from the community (i.e., the vari- through PTA contributions than schools in the able proxies for an unobserved variable of head latter. The actual range is far higher, since some teacher-community interaction); or (3) it is indi- schools collect no contribution compared with rectly picking up the effect of teacher quality, a maximum of 150,000 cedis (see Annex C). On which is improved by a good head teacher. top of that, the rural dummy is significantly neg- The cluster of school management variables is ative: rural communities give less cash support highly inter-correlated. In the full model the two to schools through PTA contributions, most SMC variables are both positive but insignificant probably reflecting the fact that there is less (Model 1) -- but either one alone is significant, cash around in rural areas (which rely far more with SMC participation being the stronger of the on own production and barter than do urban two. The SPAM variables are particularly highly areas). correlated with the SMC variables. Including School-level variables also affect the level of just one SPAM variable, and no SMC variable, PTA contributions. Contributions are higher makes the former significantly positive. Given when head teachers are active in monitoring the extent of this inter-correlation no weight teachers' work (this variable is discussed more should be given to the negative coefficient on below). This result may be picking up one of the SPAM variable in Model 1. While the PTA three things, or a combination of them: (1) par- variables have negative coefficients, neither is 8 6 A N N E X E S C o m m u n i t y - l e v e l r e g r e s s i o n T a b l e D . 1 9 a n a l y s i s o f d e t e r m i n a n t s o f ( l o g g e d ) P T A e x p e n d i t u r e p e r p u p i l Model 1 Model 2 Coeff. t-stat Coeff. t-stat Community variables Average community income (logged) 1.86 3.43 *** 1.95 3.98 *** Forest ­0.17 ­0.45 .. .. Savannah ­0.14 ­0.23 .. .. Rural ­0.85 ­2.16 ** ­0.86 ­2.40 ** School variables Head teacher supervision Activities 4.37 2.39 ** 3.59 2.61 ** Teacher social relations ­0.01 ­0.38 .. .. PTA ­1.15 ­0.57 .. .. SMC 0.80 0.78 .. .. SPAM ­2.13 ­1.71 * .. .. SMC participation 2.57 1.66 2.97 2.30 ** PTA participation ­0.26 ­0.25 .. .. SPAM participation 0.77 0.40 .. .. Dummy (cluster 40)1 2.34 1.75 * 2.40 1.90 * Intercept ­19.78 ­2.06 ** ­22.46 ­2.72 *** R2 0.44 0.43 N 80 80 Notes: *** significant at 1%, ** significant at 5% and * significant at 10%. 1. Exceptionally high PTA contributions are probably explained by some PTA-managed investment, such as constructing a classroom block. significant (and do not become so in any model holds that are contributing. The lower part of specification). Teacher social relations with the Table D.20 reports the results of the selection community were not significant.101 equation. An obvious omission is whether the Similar results were found from a household- child is at a private school with no PTA (recall level analysis of the determinant of PTA contri- that applies to 4 percent of private schools). butions, but also some differences (Table D.20). However, the variable of whether the respondent The dependent variable in this case is the log states that there is a PTA at the child's school is of PTA contributions with respect to each child a good proxy for this (as well as picking up the in basic school (so the model is only estimated small number of public schools with no func- for households that have a child currently tioning PTA), and appears as the most significant enrolled in basic school). Many such house- determinant of whether a PTA contribution is holds make a zero contribution, so that OLS esti- paid or not. mation would be biased. Instead the Heckman Both community and household income model is used, which is a two-part estimation (expenditure) matter for how much is paid to procedure. First a probit model is estimated of the PTA. The higher a household's income the whether or not the household will pay any con- more likely it is to make a PTA contribution, tribution and second the determinants of the level although the average community income does of that contribution estimated for those house- not matter for this decision, which is a sensi- 8 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S D e t e r m i n a n t s o f h o u s e h o l d P T A T a b l e D . 2 0 c o n t r i b u t i o n s ( H e c k m a n m a x i m u m l i k e l i h o o d e s t i m a t i o n ) Full model Parsimonious model Coeff. z-stat Coeff. z-stat Dependent variable: PTA expenditure (logged) Household expenditure (logged) 0.29 2.70 *** 0.25 2.69 *** Average community expenditure (logged) 1.01 6.73 *** 0.99 6.74 *** Education of household head 0.12 3.95 *** 0.13 4.19 *** Average community knowledge of SMC 0.51 2.24 ** 0.54 2.37 *** Household participation in PTA 0.71 3.24 *** 0.74 3.48 *** Average community participation in PTA 0.23 0.78 .. .. Knowledge of SPAM 0.12 1.40 0.13 1.43 Participation in SPAM 0.05 0.32 .. .. Teacher social index 0.00 ­0.56 .. .. Supervision activities of head teacher 1.47 3.08 *** 1.48 3.28 *** Dummy for high observations 2.20 4.93 *** 2.19 4.94 *** Rural ­0.61 ­5.53 *** ­0.58 ­5.61 *** Forest ­0.17 ­1.68 * ­0.20 ­2.07 ** Savannah ­0.38 ­2.52 ** ­0.41 ­2.89 *** Intercept ­12.17 ­4.40 *** ­11.49 ­4.47 *** Selection equation Household expenditure (logged) 0.44 5.04 *** 0.44 5.59 *** Average community expenditure (logged) 0.02 0.13 .. .. Education of household head ­0.02 ­0.49 .. .. Average community knowledge of PTA 0.73 2.23 ** 0.62 2.01 ** Household knowledge of PTA 2.33 8.57 *** 2.35 8.80 *** Average community participation in PTA ­0.20 ­0.64 .. .. Participation in SMC 0.57 3.29 *** 0.59 3.47 *** Supervision activities of head teacher 1.11 2.37 ** 1.07 2.35 ** Teacher social index ­0.02 ­2.69 *** ­0.01 ­2.67 *** Rural ­0.40 ­3.65 *** ­0.40 ­4.33 *** Forest 0.19 1.68 * 0.20 1.76 * Savannah ­0.10 ­0.57 ­0.09 ­0.57 * Intercept ­8.38 ­3.08 *** ­8.09 ­5.68 *** /athrho 0.33 0.86 0.17 0.61 * /lnsigma 0.08 1.62 0.07 2.31 ** No of observations 1348 1348 Of which censored 365 365 Log likelihood ­2116 ­2117 ble result. These results make sense since the actually pays depends on its own resources, not PTA/school set the PTA levy as a minimum those of the surrounding community. Average amount, which will be done with reference to community income is the stronger (larger and community income. But whether a household more significant) determinant of the level of PTA 8 8 A N N E X E S contribution, although household income also implication is that where these school man- matters. The elasticity of PTA expenditure with agement organizations are active in the com- respect to community expenditure appears as munity that each household feels more inclined 1 in this model (ranging from 0.90 to 1.05 in to make a contribution and that contribution is the various model specifications estimated). larger. This is lower than that estimated in the com- As with the community results, the presence munity-level model, but is an under-estimate of of a head teacher who is actively involved in the effect of income. Doubling community supervising teachers increases both the likelihood income means doubling the income of every that a payment is made and the level of that pay- household, so the income elasticity is the sum ment. The only possibly perverse result in the of the coefficients on the two income terms, regression is that the better teacher-community which is 1.3. This is still an underestimate, relations then the less likely are households to since the doubling of income will, through the pay PTA fees. This finding may reflect either that selection equation, make households more teachers with good social relations are less well likely to contribute at all. The other house- placed to enforce payment, or that relations are hold characteristic included -- education of good precisely because they do not do so. the household head -- matters for the level of Finally, and unlike in the community results, the contribution but not whether it is made or the location dummies are significant. Forest not. region residents are more likely, and savanna School management organization variables ones less likely, to pay PTA fees than those in are important for PTA contributions. For reasons coastal region. But both regions pay a smaller already given, the household stating that there amount than do coastal regions. Rural residents is a PTA matters for if a contribution is made. are both less likely to pay, and to pay less if they But average community knowledge of the PTA pay at all, than urban residents. This finding is matters as does whether the household has consistent with the community-level results. been actively involved with the SMC. Commu- These location variables may reflect variations nity knowledge of the PTA and SMC both mat- in the availability of cash in the local economy, ter for the level of the contribution. The which is required if fees are to be paid. T a b l e D . 2 1 T e a c h e r t e s t s c o r e s , 1 9 8 8 a n d 2 0 0 3 1988 2003 t-stat Raven's test 29.6 29.0 ­3.23 *** Standard error (3.2) (6.5) No. of observations 430 3,061 English 22.6 22.5 ­0.35 Standard error (2.5) (4.0) No. of observations 436 3,051 Maths 19.9 21.4 8.01 *** Standard error (3.4) (5.5) No. of observations 435 3,050 Local language n.a. 24.9 Standard error 3.9 No. of observations 1,793 Note: *** significant at 1%. Figures for 1988 are mean of school-level average. 8 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S Teaching Conditions and Methods changes in the education levels of teachers. One is the rise of private schools. These schools Teacher Training and Test Scores typically do no require recruits to have teacher Table D.21 shows the teacher test scores for 1988 training but use secondary school graduates. In and 2003. No change is expected in the Raven's 2003, 94 percent of private school teachers have test unless teachers are now being drawn from at least secondary education, compared to 72 a different segment of the population. In fact, percent of teachers in public schools. But less there is a small but significant drop. However, than 15 percent of private school teachers had mirroring progress in the rest of the population, teacher training, compared to 88 percent in pub- the math score has risen significantly, though lic schools. English has not. These results thus do not give The more general trend in trained teachers is any clear message regarding the academic abil- shown in Figure D.3.102 In public schools ity of teachers. untrained teachers are called "pupil-teachers." The level of schooling among teachers has These are teachers on the teaching staff, and risen. In 1988, only 40 percent were secondary should be distinguished from parental volunteers school graduates, compared to three-quarters who may also help out, especially when class today (Table D.22). Two factors lie beyond sizes are large or a school short of teachers. Offi- T a b l e D . 2 2 T e a c h e r s c h o o l i n g ( p e r c e n t ) 1988 2003 Primary 1.7 0.0 Middle/JSS 57.2 19.5 Senior Secondary 40.8 75.6 Tertiary 0.4 5.0 Total 100.0 100.0 No. of observations 4547 3129 T e a c h e r e d u c a t i o n a n d t r a i n i n g b y T a b l e D . 2 3 p u b l i c v s . p r i v a t e , 2 0 0 3 Public Private Total Schooling Middle/JSS 22.4 6.2 19.5 Senior Secondary 72.3 90.1 75.6 Tertiary 5.3 3.7 5.0 Total 100.0 100.0 100.0 Teacher training Yes 87.5 14.7 74.4 No 12.5 85.3 25.6 Total 100.0 100.0 100.0 No. of observations 2564 565 3129 9 0 A N N E X E S F r a c t i o n o f t r a i n e d t e a c h e r s , F i g u r e D . 3 p r i m a r y a n d J S S 100 90 80 70 60 50 trained 40 30 Percent 20 10 0 84/85 88/89 92/93 96/97 00/01 Primary JSS Source: MoE data cial policy has been to eliminate pupil-teachers, a student-centered learning approach and use and as part of the reforms in the 1990s they were simulations on a regular basis, though about a given a period to acquire training. If they failed fifth of the latter could not explain them prop- to do so they would lose their jobs. This policy erly. And about one-fifth use cues to help explain resulted in the rise in the proportion of teach- difficult words. In summary, improved methods ers who are trained, which is clearly shown in are far from unknown, but their use cannot be the figure. Since the late 1990s the upward trend described as widespread, being used by a minor- has reversed itself especially in primary schools, ity of teachers. though this is partly the result of the rise in pri- The data from these questions were used to vate schools. construct a single composite variable on teach- The government's policy is to increase reliance ing methods (TMETHODS). This variable is a on in-service training (INSET) to develop teach- simple average of the three responses (multiplied ing skills. This has not happened. Only 3 percent by 100/4), each response re-scaled as necessary of teachers receive such training on a regular basis. to range from 1 to 4. Teachers unable to correctly describe simulations were re-coded as 1 (`Never' Teaching Methods use) for question two. The average value of this Teachers were asked three questions to test variable for all 3,129 teachers is 62.5 (Table their familiarity with improved teaching meth- D.27), but with a reasonable degree of variation ods, and the extent to which they claim to use (the coefficient of variation is equal to 0.28). them in the classroom. The questions, described Trained teachers are significantly more likely to in the titles to Tables D.24­D.26, asked if chil- use improved methods than untrained ones, dren are encouraged to explore material by although there is not a significant difference themselves, the use of simulations (role play), between teachers who have received university- and the use of cues in explaining a word. In the level teacher training and those trained by TTCs. second case, in which they were asked directly There is also significant variation across the if they use simulations, those claiming to use country, with both the forest and savannah them were asked to explain the approach. The zones showing significantly less use of improved results show that about a third of teachers use methods than the coastal region. 9 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S W h i c h o f t h e f o l l o w i n g d e s c r i b e s T a b l e D . 2 4 y o u r a p p r o a c h t o t e a c h i n g ? Number Percent Allow children to explore material on their own 1,141 36 Present material to children which you have prepared in advance 1,988 64 Total 3,129 100 H o w o f t e n d o y o u u s e s i m u l a t i o n s T a b l e D . 2 5 a s a n i n s t r u c t i o n a l a p p r o a c h ? Of which percentage able to give a Number Percent correct explanation Often 1,007 32 80 Sometimes 1,229 39 76 Rarely 357 11 44 Never 536 17 n.a. Total 3,129 100 73 O n e o f y o u r p u p i l s h a s d i f f i c u l t y i n p r o n o u n c i n g a w o r d i n a g r o u p T a b l e D . 2 6 r e a d i n g . H o w w i l l y o u g o a b o u t h e l p i n g t h e o t h e r p u p i l s t o u n d e r s t a n d t h e w o r d ? Number Percent Would not do so 18 1 Tell the student to read the word again 661 21 Explain or define the word 1,286 41 Use cues in the story to explain the word 653 21 Other 511 16 Total 3,129 100 These findings remain valid for multivariate However, neither of the teacher living and work- analysis, shown in Table D.28. The table reports ing conditions are significant and neither is results both a full model (Model 1) and a more morale. None of these three variables are indi- parsimonious one (Model 2) including only vari- vidually significant if the other two are dropped. ables which are significant in the full model. Both Head teacher supervision of teachers has a sig- teacher training and in-service training are sig- nificantly positive impact. Introducing the vari- nificant, the former markedly so. Also significant ables that make up this composite separately into are the teacher's level of schooling and his or the regression shows the strongest effect to her ability as measured by the Raven's test. come from the head teacher "sitting in" on the 9 2 A N N E X E S B i v a r i a t e t a b u l a t i o n o f t e a c h i n g T a b l e D . 2 7 m e t h o d s a g a i n s t t e a c h e r t r a i n i n g a n d g e o g r a p h i c a l z o n e ( p e r c e n t ) Teaching Standard Number of methods deviation observations t-stat* Teacher training None 57.3 16.6 803 Certificate 64.1 17.7 2247 9.84 Tertiary 66.3 18.5 80 4.16 Geographical zone Coastal 65.2 18.3 933 Forest 61.2 17.6 1835 ­5.49 Savannah 61.4 15.9 361 ­3.71 Total 62.5 17.7 3129 Note: *for teacher training the t-stat compares with the row above, for zone both t-stats compare with the coastal region. class (this question was interpreted as appear- The primary school dummy is not signifi- ing in the class rather than necessarily sitting cant, so primary teachers are neither more nor down for any length of time). The teacher hav- less likely to use improved methods than those ing had direct contact with the circuit supervi- in JSS. However, the combined school dummy sor also has a positive effect, although the is significant at 10 percent. This finding may variables capturing the activities of the circuit reflect a spillover effect, which is more likely in supervisor are insignificant. larger schools (although the number of teach- Two teacher perception variables play a sig- ers is not significant when included in the nificant role. One is that teaching methods are model), or may reflect the concentration of bet- better the worse the teacher perceives the sup- ter teachers in those schools. ply of materials to be. There are two possible Finally, the zonal dummies are significant, explanations for this finding. The first is that the implying that there are factors common to non- teacher compensates for inadequate materials coastal areas that are not included in the model, by using more innovative methods. A second which encourage the lesser use of improved explanation is that teachers wishing to use teaching methods in these areas. One explana- more innovative methods are more likely to per- tion may be that teacher training in the coastal ceive materials as being inadequate as those region introduces teachers to these methods teachers who invest less heavily in method. more successfully than does teacher training Support for the second explanation is given by elsewhere. To test this hypothesis interactive the fact that teachers at the same school (i.e., terms were introduced of the coastal dummy with objectively the same materials available) multiplied by each of the teacher training vari- are more likely to perceive materials as inad- ables.103 The results (Model 3) strongly support equate the higher their value of TMETHODS. the hypothesis: the two interactive terms are Second, the variable DISCIPLINE, which cap- significantly positive (with the result that coef- tures the severity with which a teacher believes ficients of the training variables and their sig- various offences committed by teachers should nificance is reduced, becoming insignificant in be punished, is significantly positive. This vari- the case of in-service training),104 and the zonal able should probably not be interpreted as an dummies become insignificant. explanatory variable but rather as the correla- Dummies were not included in the regressions tion in different aspects of professionalism for the WSD and QUIPS projects since these proj- among teachers. ects are restricted to primary schools and so their 9 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S R e g r e s s i o n a n a l y s i s o f d e t e r m i n a n t s o f T a b l e D . 2 8 t e a c h i n g m e t h o d s Model 1 Model 2 Model 3 Coeff. t-stat Coeff. t-stat Coeff. t-stat Teacher training and ability Teacher training 6.31 8.77 *** 6.31 8.93 *** 5.24 6.31 *** In-service training 3.36 2.60 *** 3.21 2.50 ** 1.74 1.16 Teacher schooling 2.58 3.37 *** 2.67 3.81 *** 2.59 3.70 *** Teacher Raven's score 0.10 2.10 ** 0.11 2.24 ** 0.10 2.15 ** Teaching conditions Teacher morale 0.90 0.94 .. .. .. .. Teacher working conditions ­0.15 ­0.31 .. .. .. .. Teacher living conditions 0.03 0.11 .. .. .. .. Supervision Head teacher supervision 2.76 2.19 ** 2.33 1.88 * 2.33 1.89 * Visit by circuit supervisor 4.32 5.19 *** 3.60 5.51 *** 3.53 5.40 *** Circuit supervisor supervision ­4.31 ­1.41 .. .. .. .. Other variables Teacher's discipline 1.18 3.30 *** 1.15 3.23 *** 1.15 3.23 *** Teacher's perception of materials 0.77 3.03 *** 0.75 2.97 *** 0.70 2.76 *** Primary teacher ­0.27 ­0.38 .. .. .. .. Combined school 1.37 1.79 * 1.41 1.85 * 1.44 1.90 * Forest ­3.55 ­4.73 *** ­3.52 ­4.75 *** ­0.49 ­0.37 Savannah ­5.34 ­4.50 *** ­5.49 ­4.69 *** ­2.45 ­1.54 Constant 47.38 11.34 *** 45.60 13.79 *** 43.87 13.01 *** Interactive terms Teacher training in coastal region .. .. .. .. 3.49 2.46 *** In-service training in coastal region .. .. .. .. 5.55 1.92 ** N 2,939 2,953 2,953 R2 0.068 0.067 0.070 Note: *** significant at 1%, ** significant at 5% and * significant at 10%. inclusion would have reduced the sample size. D.29 summarizes results for two of these. Home- When they were introduced into the regression work is set at least once a week by over 95 per- neither variable was significant, suggesting the cent of teachers for math and English, and work projects have no direct effect over and above the assessed with the same frequency by half. The indirect effect through their influence on any of most striking result is the lesser attention paid the variables already included in the model. to local languages, for which homework is set However, bivariate analysis of the difference in infrequently by nearly 20 percent of teachers and means of teachers in project schools and other work rarely assessed in over half the total num- teachers showed no significant impact of either ber of cases. project on teaching methods as captured in our These data were used to construct an index data, suggesting that they have no indirect effects. of student monitoring by teachers.105 A regres- Data were collected on the frequency with sion model was used to examine the determi- which teachers set homework, look at and assess nants of this behavior (Table D.30). Both students' work for both math and English. Table in-service training and visits by the circuit super- 9 4 A N N E X E S M o n i t o r i n g o f s t u d e n t p e r f o r m a n c e T a b l e D . 2 9 b y t e a c h e r s Homework Assess work Local Local English Math language English Math language Once a month or less 3.8 3.0 17.3 39.5 38.6 52.1 Once a week 78.8 76.7 79.2 50.2 51.2 44.9 Daily 17.4 20.2 3.5 10.3 10.2 2.9 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 2055 2041 1552 2054 2041 1561 visor were found to exert a significant positive · Disciplining students impact on student monitoring. School quality · All students engaged in copying, reading, or variables appear not to matter, although the other forms of exercise availability of materials has a positive effect, · Dealing with students on a one-to-one basis which may be picking up particularly good · Addressing the whole class schools. Teachers who think that school man- Enumerators reported difficulties in adminis- agement is a problem are more likely to under- tering this question. The total implied class time take monitoring (perhaps because better teachers from summing the answers to the above ques- more readily perceive such problems), whereas tions ranged from 5 minutes to four hours. How- teachers who think morale is a problem are less ever, these extremes were limited to a handful likely to monitor student performance (perhaps of observations. Of the 3,128 observations with as bad teachers are more likely to complain complete data, 3,107 are retained in the sample about morale). if those with a total class time of less than 25 min- utes or more than two hours are dropped. The Textbook Use results, shown in Table D.32, show that on aver- The data show a substantial rise in textbook age 28 percent of time is spent on getting ready availability. However, it is not the availability of for class or disciplining students. By a broad def- textbooks that matters but whether they are inition of time on task therefore 72 percent of used or not. The teacher questionnaire asked class time is spent on task, though this varies whether textbooks had been used in the last class from 29 to 100 percent. A narrow definition of taught on each subject. For math and English, time on task takes into account only the activi- nearly all teachers had textbooks available and ties that actively engage all students, in which over 90 percent used them (Table D.31). For local case the average is 33 percent, ranging from 0 language nearly half did not have textbooks, and to 73 percent. a lower percentage of those who have them used The robustness of these results was checked them. These findings are consistent with those by considering the means only for those report- of Okyere et al. (1997) who found that textbooks ing a total class time of one hour or less, a sam- were used when there are sufficient of them. ple of 2,215 teachers (Table D.33). The percentage distribution of activities is hardly Table Time on Task changed. The teacher questionnaire included questions Bivariate analysis shows no relationship about classroom activity, specifically the amount between the sex or location of the teacher but of time in a typical class spent: a positive association with teacher training. These · Preparing for class, handing out materials, results are supported by multivariate analysis writing on the board material for exercises or for the determinants of the broad definition of copying time on task (the narrow definition regression 9 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S D e t e r m i n a n t s o f s t u d e n t T a b l e D . 3 0 m o n i t o r i n g b y t e a c h e r s Full model Parsimonious model Coeff. t-stat Coeff. t-stat Teacher training, ability and attitude Teacher training ­4.12 ­0.75 ­6.99 ­1.38 In-service training 15.61 2.01 ** 15.58 2.03 ** Teacher schooling 3.24 0.65 .. .. Teacher's Ravens score 0.04 0.11 .. .. Teacher's English score ­0.14 ­0.22 .. .. Teacher's Math score 0.22 0.49 .. .. Teaching method 0.56 4.52 *** 0.56 4.72 *** Discipline 3.72 1.54 3.73 1.6 Teaching conditions Teacher's morale ­6.28 ­0.98 .. .. Teacher working conditions 5.20 1.55 2.92 0.93 Board quality: size ­1.80 ­0.25 .. .. Board quality: easy to clean ­0.65 ­0.10 .. .. Materials to display 6.86 2.14 ** 6.89 2.33 *** Shared classroom 11.56 1.52 10.41 1.42 Full-sized internal walls ­8.63 ­1.17 .. .. Class disturbed by noise ­5.48 ­1.19 .. .. Living conditions 2.71 1.30 2.83 1.4 Supervision Head teacher supervision 2.55 0.30 0.48 0.06 Circuit supervisor supervision 6.42 0.31 3.64 0.18 Visit by circuit supervisor 13.11 2.32 ** 12.29 2.23 ** Teacher perceptions of Student discipline 0.28 0.14 .. .. Student ability ­3.32 ­1.63 ­3.77 ­1.98 ** Availability of materials ­0.24 ­0.14 .. .. Teacher morale ­4.04 ­1.85 * ­3.51 ­1.67 * School management 4.09 1.76 * 4.26 1.9 * Other variables Primary school 28.70 5.37 *** 27.03 5.64 *** Combined School ­0.33 ­0.06 .. .. Urban/rural ­0.83 ­0.18 .. .. Forest ­7.19 ­1.40 .. .. Savannah ­8.40 ­1.00 .. .. Constant 57.42 1.58 50.04 1.83 * N 2323 2381 R2 0.042 0.040 Note: *** significant at 1%, ** significant at 5% and * significant at 10%. 9 6 A N N E X E S M o s t t e a c h e r s u s e t e x t b o o k s w h e n T a b l e D . 3 1 t h e y a r e a v a i l a b l e English Math Local language No textbooks available 1.9 2.4 44.2 Yes 94.5 90.7 47.6 No textbooks available 3.6 6.9 8.2 Total 100.0 100.0 100.0 Memo: % of those having textbooks who use them 96.3 93.0 85.3 T a b l e D . 3 2 T i m e o n t a s k : c l a s s r o o m a c t i v i t i e s Coefficient Mean Minimum Maximum of variation Total class time (minutes) 51 25 120 Of which (percent) Preparing for class 21 0 68 0.55 Disciplining students 7 0 48 0.69 All students engaged copying etc 33 0 73 0.30 Dealing with student 1-to-1 13 0 68 0.62 Addressing whole class 26 0 80 0.51 Memo: time on task Broad definition 72 29 100 0.17 Narrow definition 33 0 73 0.30 T i m e o n t a s k : c l a s s r o o m a c t i v i t i e s T a b l e D . 3 3 ( o n l y t h o s e w i t h c l a s s t i m e o f a n h o u r o r l e s s ) Coefficient Mean Minimum Maximum of variation Preparing for class 22 0 65 0.51 Disciplining students 7 0 33 0.65 All students engaged copying etc 34 0 73 0.30 Dealing with student 1-to-1 12 0 63 0.62 Addressing whole class 25 0 80 0.54 9 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e D . 3 4 D e t e r m i n a n t s o f t i m e o n t a s k Broad definition Narrow definition Coeff. t-stat Coeff. t-stat Teacher characteristics Male ­0.34 ­0.67 ­0.23 ­0.55 Teacher training 2.96 4.63 *** ­0.02 ­0.03 In-service training ­0.37 ­0.40 0.31 0.41 Old dummy 0.26 0.34 0.97 1.55 Teaching methods ­1.04 ­0.80 ­1.31 ­1.23 School characteristics Primary ­0.29 ­0.55 0.95 2.20 ** Private 0.42 0.52 ­0.70 ­1.06 Board quality 1.41 0.33 2.31 0.65 Class have internal walls 0.23 0.30 0.35 0.56 Class disturbed by noise ­1.00 ­2.08 ** 0.27 0.69 Display material available 1.83 5.36 *** 0.29 1.05 Head teacher supervision 3.70 4.11 *** 0.92 1.25 Circuit supervisor supervision 5.42 3.12 *** 1.11 0.78 Textbook availability 1.11 3.45 *** 0.48 1.83 Chalk availability 1.33 3.44 *** 0.02 0.07 School mean teacher morale 2.03 2.84 *** 2.22 3.79 *** School mean discipline 1.84 4.02 *** 0.51 1.35 School mean teacher perception of management ­1.30 ­3.40 *** ­0.37 ­1.19 Community characteristics Savannah ­0.03 ­0.04 ­1.93 ­2.58 *** Forest 2.76 5.23 *** 0.38 0.88 Rural ­0.06 ­0.12 0.31 0.74 Intercept 44.65 7.95 **** 23.52 5.12 *** R squared 0.08 0.02 No. of observations 2,919 2,919 does not yield good results); see Table D.34. The quality of infrastructure (class disturbed by exter- regression shows also that teacher training mat- nal noise). School management also matters: ters, as does overall teacher morale at the school supervision of the teacher by both the head (the individual teacher's morale is significant if teacher and the circuit supervisor improve time mean morale is not included). In-service train- on task, as does teachers' perception of school ing is not significant, but it should be recalled that management (the worse is the perception then less than 5 percent of teachers receive such train- the less the time on task). ing on a frequent basis. School quality matters to the time spent on Teacher Conditions, Motivation, and task: materials matter (display material, text- Morale books, and chalk, and the average of teachers' This section considers three related variables: perceptions of material availability), as does the teacher morale, teacher working conditions, and 9 8 A N N E X E S T a b l e D . 3 5 M e a s u r e s o f t e a c h e r m o r a l e Enjoy Remain Morale 0 13.3 31.0 10.2 1 86.7 69.0 24.0 2 .. .. 65.9 Total 100.0 100.0 100.0 No. of observations 3,129 3,129 3,129 teacher living conditions. Few data were col- diality of relations with the local community to lected on these issues in 1988 -- though those make a subjective livings conditions index (Table available mostly show a clear improvement -- D.36). so the discussion is mostly restricted to an analy- Teachers appear more satisfied with their sis of the 2003 data. working conditions (60 percent replying they are good or very good) compared to their living con- Descriptive Analysis of Teacher Morale and ditions (46 percent describing as good or very Conditions good). Urban teachers are less satisfied with The teacher morale variable is constructed as the their working conditions than those in rural simple sum to two questions: do you enjoy areas, though there is no difference in the per- being a teacher (no=0, yes=1) and do you intend ception of living conditions (although the objec- to remain as a teacher (no=0, yes=1)? The result- tive measures given below suggest they are ing variable, shown in Table D.35, is categori- worse in rural areas). cal from 0 to 2.106 The variable suggests Teacher morale is related to all three of the reasonably high morale, with two-thirds of teach- subjective perception variables mentioned above, ers being in the top category. in particular working conditions (Table D.37) Questions were also asked on teachers' sub- Objective data were also collected on teacher jective perceptions of living and working con- conditions. Regarding living conditions data are ditions, based on a 5- and 4-point scale, available on whether pay is received on time, if respectively. The former of these was combined housing is provided for the teacher, the type of with the results of a question regarding the cor- water supply at the teachers' housing, if teacher S u b j e c t i v e p e r c e p t i o n s o f w o r k i n g T a b l e D . 3 6 a n d l i v i n g c o n d i t i o n s How would you describe How would you describe the your working conditions? conditions of your accommodation? Urban Rural Total Urban Rural Total Very poor 8.0 4.9 7.0 Very poor 6.3 5.6 6.1 Poor 33.7 30.6 32.7 Poor 19.4 20.0 19.6 Good 54.2 59.2 55.8 Adequate 27.2 29.5 28.0 Very good 4.2 5.2 4.5 Good 38.5 38.4 38.5 Total 100.0 100.0 100.0 Very good 8.5 6.4 7.8 No. of obs. 2096 1033 3129 Total 100.0 100.0 100.0 Chi-squared Chi-squared (Prob.) 16.39 (0.001) (Prob.) 6.05 (0.196) 9 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S R e l a t i o n s h i p b e t w e e n t e a c h e r T a b l e D . 3 7 m o r a l e a n d t e a c h e r c o n d i t i o n s Teacher morale Chi- Low Medium High Total squared Subjective living conditions Very poor 10.1 6.0 5.5 6.1 15.6 Poor 21.7 20.4 19.0 19.6 (0.049) Adequate 27.4 26.4 28.7 28.0 Good 32.4 39.2 39.2 38.5 Very good 8.5 8.0 7.7 7.8 Total 100.0 100.0 100.0 100.0 Subjective living index Low 30.8 25.2 21.9 23.6 16.1 Medium 25.2 22.8 25.7 24.9 (0.003) High 44.0 52.0 52.5 51.5 Total 100.0 100.0 100.0 100.0 Subjective working conditions Very poor 15.4 9.6 4.8 7.0 152.2 Poor 49.1 37.6 28.3 32.7 (0.000) Good 33.3 48.7 61.9 55.8 Very good 2.2 4.1 5.0 4.5 Total 100.0 100.0 100.0 100.0 No. of observations 318 750 2061 3129 housing has electricity, and if the teacher is a were almost never late in 1988. The problem of member of a group within the community (e.g., late pay is greater for new teachers. Teachers church, cultural organization, or sports club). with less than a year's experience are significantly Data on the first of these were also collected in more likely to say that they never receive their 1988, although at the school rather than teacher pay on time. But there remains a not insignifi- level and for a smaller number on water and cant group of older teachers who get their pay electricity.107 All variables that can be compared late, and this is found to be a critical factor in across time show an improvement in the living teacher morale (see below). conditions of teachers. Of the five objective living conditions four In 1988, 95 percent of schools did not pro- show a bias in favor of teachers living in urban vide lodging for any of their teachers. By 2003 areas, two strongly so (water and electricity). Just this figure fell to 70 percent. Today, 60 percent one indicator is more favorable in rural areas. of teachers have access to pipe-borne water That is having lodging provided, though this is compared to only 20 percent 15 years ago. In not common anywhere (21 percent of teachers only 13 percent of schools do no teachers have in rural areas and 6 percent in urban). In con- electricity in their home in 2003 compared to half sequence, the objective index of living condi- in 1988. In 2003, 72 percent of teachers reported tions, calculated as the average of the scaled that they always or in most months received their values of these five variables, is better in urban salary on time, compared to the only 25 percent areas than rural (Table D.38) -- posing a puz- of schools for which it was reported that salaries zle as to why subjective perceptions of living 1 0 0 A N N E X E S O b j e c t i v e i n d e x o f l i v i n g T a b l e D . 3 8 c o n d i t i o n s Urban Rural Total Low 22.9 53.1 32.9 Medium 42.9 31.9 39.3 High 34.2 14.9 27.8 Total 100.0 100.0 100.0 No. of observations 2096 1033 3129 Chi-squared (Prob.) 306.2 (0.000) conditions do not vary between rural and urban pay on time, which has a significantly posi- areas. tive effect on all three measures. Having good Data on working conditions covered both social relations with the community are also teacher-level data on the various dimensions of important. Other aspects of living conditions school quality and school-level data. The analy- affect the subjective perception of living con- sis of material and physical inputs earlier has ditions, but not the other two variables. already shown how these have improved in · Teacher-level school variables: several of nearly all schools between 1988 and 2003. these are significant, virtually all with the expected sign. Both board (easy to clean) and Determinants of Teacher Morale and classroom (not disrupted by noise) quality Conditions affect teacher perceptions of both working Each of teacher morale, and teacher percep- and living conditions. While some aspects of tions of living and working conditions was mod- school quality thus seem to spill over into per- eled as an ordered probit. In the first and most ceptions of living conditions (but not vice general equation, both subjective and objective versa, see previous bullet), two perverse measures were included. However, it seems results appear with respect to in-service train- likely that the three dependent variables are ing and visits of the circuit supervisor. Also determined simultaneously so that there is a surprising is the negative impact of most vari- problem in including the subjective perceptions ables measuring "quality" of other teachers at as regressors. Hence model 2 in each case drops the school. these variables, which may be considered as · School management and projects: the school reduced form estimate. Since a more parsimo- management variables send mixed signals. nious model is identified in each case with some This result partly follows from their inter-cor- differing regressors the model can be considered relation. Both PTA variables have a positive to be identified, with model structure deter- effect on subjective working conditions, and mined by a data analytic approach. PTA is positive in the subjective living con- Table D.39 summarizes the results, from ditions equation. Despite the small number which the following main points emerge: of observations, the WSD dummy is signifi- · Teacher characteristics: young teachers and cant in one case (living conditions). especially males, especially those in rural · Community characteristics: taking into account areas, are more generally dissatisfied. Better- all these factors, the indices are systematically qualified and -educated teachers also tend to higher in forest and savannah zones and in be less satisfied. Living in the home district rural areas. However, there is a negative effect and with one's spouse both have a positive from community income and education effect. (which are positively correlated with the · Teacher living conditions: a resoundingly objective measure of living conditions). The robust result is the importance of receiving likely explanation (arising from field obser- 1 0 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S D e t e r m i n a n t s o f m o r a l e a n d T a b l e D . 3 9 c o n d i t i o n s ( r e s u l t s f r o m r e g r e s s i o n a n a l y s i s ) Subjective Subjective working conditions living conditions Teacher morale Teacher characteristics Male ­ve (***) ­ve (***) ­ve (***) Young (<30) .. .. ­ve (***) Young in rural area ­ve (*) .. ­ve (*) Young male in rural savannah .. +ve (**) Old (>50) .. .. +ve (***) Living with spouse .. .. +ve (***) Living in home district .. +ve (**) .. Raven's score .. +ve (**) .. English score ­ve (**) +ve (**) .. Years of schooling ­ve (*) .. ­ve (***) Level of teacher training ­ve (**) .. .. Subjective indices .. .. Morale +ve (***) .. n.a. Subjective living conditions +ve (***) n.a. .. Subjective working conditions n.a. +ve (***) +ve (***) Teacher living conditions .. .. .. Pay received on time +ve (***) +ve (***) +ve (***) Water in residence .. +ve (***) .. Electricity in residence .. +ve (***) .. Lodging provided .. +ve (***) .. Member of social organization .. .. +ve (***) Perception of social relations +ve (***) .. +ve (*) Teacher-level school data .. .. .. Frequency of in-service training .. ­ve (*) .. Teach extra classes +ve (**) +ve (**) .. Have to share classroom ­ve (*) .. .. Noise disrupts classes ­ve (***) ­ve (*) .. Head teacher visits classes .. +ve (*) .. Display material available +ve (***) .. .. Board easy to clean +ve (***) +ve (***) .. Teacher meets with circuit supervisor +ve (**) .. .. Circuit supervisor monitoring of teacher activities +ve (*) ­ve (*) .. School-level data .. .. .. Own water supply +ve (**) .. .. Average level of teacher training ­ve (***) +ve (*) .. (continued on next page) 1 0 2 A N N E X E S D e t e r m i n a n t s o f m o r a l e a n d T a b l e D . 3 9 c o n d i t i o n s ( r e s u l t s f r o m ( c o n t i n u e d ) r e g r e s s i o n a n a l y s i s ) Subjective Subjective working conditions living conditions Teacher morale Average level of teacher schooling ­ve (*) .. .. Average teacher test score .. ­ve (*) .. School management .. .. .. PTA +ve (*) +ve (*) .. SMC ­ve (*) .. .. PTA met in last month +ve (***) .. .. SMC helped in the school in past year ­ve (*) .. .. Plan from SPAM being implemented .. +ve (***) .. Project dummies .. .. .. Whole School Development .. +ve (**) .. QUIPS .. .. .. Primary School Development .. .. .. Community variables .. .. .. Average income .. .. ­ve (***) Average education of household heads .. .. ­ve (***) Savannah +ve (***) .. +ve (***) Forest +ve (***) .. +ve (***) Rural +ve (***) +ve (***) .. vation) is that teachers compare themselves toward finding a greater degree of absenteeism with their peers in the neighboring commu- in that year. nity -- they are much lower down the scale However, despite this bias, the data clearly in well off communities than poor ones, and show that absenteeism has increased over the so will be less satisfied with their lot. past 15 years. In 2003, nearly 13 percent of teachers had been absent in the past month, Absenteeism compared to just over 4 percent in 1988 (Table In 1988, data were collected on absenteeism at D.40). the school level, asking how many teachers had Correspondingly, more schools are affected been absent for reasons other than sickness by absenteeism. In 1988, 85 percent of schools during the last 12 months. In 2003, this question did not suffer at all; whereas this figure has was included in the teacher roster of the school now fallen to 61 percent, with 13 percent of questionnaire, asking if the teacher had been schools have over one-third of the teachers absent in the past four weeks for reasons other being absent for reasons other than sickness in than sickness. The results of the two surveys the past month (Table D.41). should not be comparable because the longer Table D.42 reports bivariate analysis of school- reference period used in 1988 will bias results level absenteeism rates using 2003 data. The 1 0 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P e r c e n t t e a c h e r s a b s e n t b y y e a r T a b l e D . 4 0 a n d t y p e o f s c h o o l Middle/ Primary JSS Total 1988 4.7 3.7 4.3 2003 12.8 12.8 12.8 A b s e n t e e i s m r a t e s b y y e a r ( p e r c e n t T a b l e D . 4 1 o f s c h o o l s i n e a c h c a t e g o r y ) 1988 2003 None 85 61 Up to a third 11 26 Between one to two thirds 3 9 More than two thirds 1 4 Total 100 100 No. of observations 518 706 C r o s s - t a b u l a t i o n s f o r a b s e n t e e i s m T a b l e D . 4 2 r a t e s , 2 0 0 3 Coastal Forest Savannah Urban Rural Public Private None 66 56 69 61 61 56 80 Up to a third 25 28 19 27 23 29 14 Between one to two thirds 5 12 10 9 10 11 4 More than two thirds 4 4 2 2 7 5 1 Total 100 100 100 100 100 100 100 No. of observations 205 404 97 451 255 568 138 Chi-squared 13.9** 9.9** 28.3*** Note: ** significant at 5% level, *** significant at 1% level. strongest difference is between public and pri- distance from the school and suffer transport vate schools: 80 percent of private schools have problems, (2) they have to travel to town once no problem with absenteeism, compared to not a month to collect their pay, which they may find much more than half of public schools. There is not yet there, and (3) rural teachers attend to is also a significant relationship with rural ver- their farming activities.108 Finally, absenteeism is sus urban schools, 7 percent of rural schools suf- worst in the forest zone perhaps because of the fering absenteeism rates of over two-thirds. greater scope for profitable farming in the zone. Likely reasons for greater absenteeism in rural Multivariate analysis of teacher-level data also areas are that: (1) teachers may live in town some shows that private schools are less likely to suf- 1 0 4 A N N E X E S D e t e r m i n a n t s o f t e a c h e r T a b l e D . 4 3 a b s e n t e e i s m Coefficient z-stat Teacher characteristics Male 0.110 1.15 Age ­0.006 ­1.24 In-service training 0.030 0.22 Teacher training 0.189 1.28 Teacher perception of morale 0.068 1.69 * Teacher conditions Teacher morale ­0.133 ­1.93 * Subjective working conditions ­0.112 ­1.69 * Pay on time ­0.088 ­2.05 ** Subjective living conditions 0.073 1.69 * Social relations 0.124 1.52 Home district 0.122 1.31 Living with spouse 0.094 1.27 Objective living conditions ­0.515 ­1.73 * School characteristics Primary ­1.505 ­1.67 * Private school ­0.299 ­1.75 * PTA helped in last month ­0.113 ­1.21 QUIPS ­0.359 ­1.37 Desks ­0.179 ­2.10 ** Pupil teacher ratio 0.011 3.45 *** Head teacher discuss lesson plans 0.495 2.18 ** Community characteristics Per capita expenditure 0.130 1.00 Forest 0.240 2.55 ** Intercept 0.725 Number of obs 1606 Pseudo R2 0.074 Note: *** significant at 1%, ** 5% and * 10%. fer from absenteeism (Table D.43). It also shows when entered alongside the index. There may that poor working conditions are associated be a direct effect of time taken by teachers in with a greater likelihood of absenteeism. The going to inquire about their pay. Low morale is subjective working condition index is signifi- also associated with absenteeism. cantly negative. The most important compo- Some direct measures of school conditions also nent of that index -- receiving pay on time -- matter. A high pupil-teacher ratio encourages is so important it is also significant in its own right absenteeism, as does poor facilities as measured 1 0 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S by lack of desks. The effect of the head teacher spouse, being in the home district and having discussing lesson plans is a perverse result. good social relations, that appear conducive to On the other hand there are a cluster of liv- absenteeism.109 Presumably such circumstances ing condition variables, such as living with provide distractions from work. 1 0 6 ANNEX E: TABLES OF SCHOOL QUALITY VARIABLES Recurrent Inputs (a) Full sample T a b l e E . 1 C h a l k ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Never enough 12.7 8.4 8.7 9.3 10.9 8.8 Occasional shortages 66.8 5.5 72.7 10.7 69.5 7.6 Always enough 20.5 86.1 18.6 79.9 19.6 83.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 283 417 231 289 514 706 Chi-squared (Prob) 388.9 (0.000) 220.6 (0.000) 553.2 (0.000) T a b l e E . 2 E n g l i s h b o o k s ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less than one book between two 58.4 11.3 42.1 22.1 51.1 15.7 At least one book between two 20.6 16.3 37.3 42.6 28.1 27.1 At least one book per student 21.0 72.4 20.6 35.3 20.8 57.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 286 417 233 289 519 706 Chi-squared (Prob) 214.13 (0.000) 27.4 (0.000) 218.1 (0.000) 1 0 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e E . 3 M a t h b o o k s ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less than one book between two 35.7 11.0 34.3 5.5 35.1 8.8 At least one book between two 32.9 35.7 52.4 23.2 41.6 30.6 At least one book per student 31.5 53.2 13.3 71.3 23.3 60.6 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 286 417 233 289 519 706 Chi-squared (Prob) 69.5 (0.000) 184.8 (0.000) 211.0 (0.000) T a b l e E . 4 D e s k s ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 No desks or tables at all 18.2 1.2 9.0 1.0 14.1 1.1 Some, but not enough 70.3 31.4 77.7 31.1 73.6 31.3 Enough for everyone 11.5 67.4 13.3 67.8 12.3 67.6 Total 100 100 100 100 100 100 No. of observations 286 417 233 289 519 706 Chi-squared (Prob) 288.1 (0.000) 159.8 (0.000) 391.0 (0.000) T a b l e E . 5 R e c u r r e n t i n p u t i n d e x ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Low (<0.5) 325.7 1.0 27.7 0.7 32.1 0.8 Medium (0.5­0.75) 49.8 19.2 60.2 28.4 54.5 22.9 High (>0.75) 14.5 79.9 12.1 70.9 13.4 76.2 Total 100 100 100 100 100 100 No. of observations 283 417 231 289 514 706 Chi-squared (Prob) 320.5 (0.000) 203.5 (0.000) 524.5 (0.000) 1 0 8 A N N E X E S (b) Public schools only T a b l e E . 6 C h a l k ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Never enough 13.4 7.2 8.8 10.5 11.3 8.6 Occasional shortages 70.6 3.4 72.6 9.7 71.5 6.2 Always enough 16.0 89.4 18.6 79.8 17.2 85.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 262 320 226 248 488 568 Chi-squared (Prob) 336.7 (0.000) 205.9 (0.000) 536.7 (0.000) E n g l i s h b o o k s ( p u b l i c s c h o o l s T a b l e E . 7 o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less than one book between two 61.0 13.1 42.1 21.8 52.2 16.9 At least one book between two 22.0 17.2 37.3 44.4 29.1 29.0 At least one book per student 17.0 69.7 20.6 33.9 18.7 54.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 264 320 228 248 492 568 Chi-squared (Prob) 185.6 (0.000) 24.8 (0.000) 187.3 (0.000) T a b l e E . 8 M a t h b o o k s ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less than one book between two 37.1 8.8 34.6 4.0 36.0 6.7 At least one book between two 34.8 40.6 52.2 23.0 42.9 32.9 At least one book per student 28.0 50.6 13.2 73.0 21.1 60.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 264 320 228 248 492 568 Chi-squared (Prob) 74.3 (0.000) 183.0 (0.000) 216.0 (0.000) 1 0 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e E . 9 D e s k s ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 No desks or tables at all 18.9 1.6 9.2 0.8 14.4 1.2 Some, but not enough 73.1 35.3 78.5 34.3 75.6 34.9 Enough for everyone 8.0 63.1 12.3 64.9 10.0 63.9 Total 100 100 100 100 100 100 No. of observations 264 320 228 248 492 568 Chi-squared (Prob) 201.1 (0.000) 142.2 (0.000) 341.2 (0.000) R e c u r r e n t i n p u t i n d e x ( p u b l i c T a b l e E . 1 0 s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Low (<0.5) 33.6 1.6 27.4 4.0 30.7 2.6 Medium (0.5­0.75) 51.1 17.8 55.3 25.0 53.1 21.0 High (>0.75) 15.3 80.6 17.3 71.0 16.2 76.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 262 320 226 248 488 568 Chi-squared (Prob) 261.4 (0.000) 145.4 (0.000) 404.2 (0.000) (c) Panel data T a b l e E . 1 1 C h a l k 1988 2003 Never enough 10.8 7.2 Occasional shortages 69.6 8.2 Always enough 19.6 84.5 Total 100 100 Observations 194 194 1 1 0 A N N E X E S T a b l e E . 1 2 E n g l i s h b o o k s 1988 2003 Less than one book between two 46.9 17.3 At least one book between two 30.6 28.1 At least one book per student 22.4 54.6 Total 100 100 Observations 196 196 T a b l e E . 1 3 M a t h b o o k s 1988 2003 Less than one book between two 30.6 7.1 At least one book between two 40.3 30.1 At least one book per student 29.1 62.8 Total 100 100 Observations 196 196 T a b l e E . 1 4 D e s k s 1988 2003 No desks or tables at all 15.8 1.5 Some, but not enough 76.5 34.7 Enough for everyone 7.7 63.8 Total 100 100 Observations 196 196 1 1 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S (d) Public versus private, 2003 T a b l e E . 1 5 C h a l k Public Private Total Never enough 8.6 9.4 8.8 Occasional shortages 6.2 13.8 7.6 Always enough 85.2 76.8 83.6 Total 100.0 100.0 100.0 Total 568 138 706 Chi-squared (Prob) 9.4 (0.009) T a b l e E . 1 6 E n g l i s h b o o k s Public Private Total Less than one book between two 16.9 10.9 15.7 At least one book between two 29.0 18.8 27.1 At least one book per student 54.0 70.3 57.2 Total 100.0 100.0 100.0 Total 568 138 706 Chi-squared (Prob) 12.2 (0.007) T a b l e E . 1 7 M a t h b o o k s Public Private Total Less than one book between two 6.7 17.4 8.8 At least one book between two 32.9 21.0 30.6 At least one book per student 60.4 61.6 60.6 Total 100.0 100.0 100.0 Total 568 138 706 Chi-squared (Prob) 20.6 (0.000) 1 1 2 A N N E X E S T a b l e E . 1 8 D e s k s Public Private Total No desks or tables at all 1.2 0.7 1.1 Some, but not enough 34.9 16.7 31.3 Enough for everyone 63.9 82.6 67.6 Total 100.0 100.0 100.0 Total 568 138 706 Chi-squared (Prob) 17.7378 (0.000) T a b l e E . 1 9 R e c u r r e n t i n d e x Public Private Total Low (<0.5) 2.6 3.6 2.8 Medium (0.5­0.75) 21.0 21.7 21.1 High (>0.75) 76.4 74.6 76.1 Total 100.0 100.0 100.0 Total 568 138 706 Chi-squared (Prob) 0.4567 (0.796) Physical Inputs (a) Full sample T a b l e E . 2 0 A d e q u a t e c l a s s r o o m s ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less then half necessary 0.0 0.0 0.0 0.0 0.0 0.0 More than half necessary 22.7 21.3 19.7 17.0 21.4 19.5 Required amount 77.3 78.7 80.3 83.0 78.6 80.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 Total 286 417 233 289 519 706 Chi-squared (prob) 0.2 (0.663) 0.7 (0.412) 0.6 (0.429) 1 1 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S C l a s s r o o m s t h a t c a n n o t b e u s e d T a b l e E . 2 1 w h e n r a i n i n g ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 More than half 28.7 19.9 20.2 19.7 24.9 19.8 Less than half 23.8 12.0 18.9 13.8 21.6 12.7 None 47.6 68.1 60.9 66.4 53.6 67.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 286 417 233 289 519 706 Chi-squared (prob) 31.6 (0.000) 2.7 (0.264) 26.9 (0.000) P e r c e n t a g e o f c l a s s r o o m s w i t h a T a b l e E . 2 2 c h a l k b o a r d ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 None 3.1 0.5 2.1 0.7 2.7 0.6 Less than half 6.6 1.9 4.3 0.7 5.6 1.4 More than half 20.6 3.6 25.8 3.8 22.9 3.7 All 69.6 94.0 67.8 94.8 68.8 94.3 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 286 417 233 289 519 706 Chi-squared (prob) 76.4 (0.000) 66.3 (0.000) 142.6 (0.000) T a b l e E . 2 3 B o a r d q u a l i t y ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Poor 9.6 8.2 5.3 9.3 7.7 8.7 Fair 33.0 17.4 38.5 10.7 35.4 14.7 Good 52.8 69.1 49.6 79.9 51.4 73.5 Excellent 4.6 5.3 6.6 0.0 5.5 3.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 282 414 226 289 508 703 Chi-squared (prob) 24.8 (0.000) 82.2 (0.000) 81.5 (0.000) 1 1 4 A N N E X E S T a b l e E . 2 4 O w n w a t e r s u p p l y ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Yes 16.1 28.3 17.2 31.1 16.6 29.5 No 83.9 71.7 82.8 68.9 83.4 70.5 Total 100 100 100 100 100 100 No. of observations 286 417 233 289 519 706 Chi-squared (prob) 14.1 (0.000) 13.5 (0.000) 26.4 (0.000) T a b l e E . 2 5 L i b r a r y ( f u l l s a m p l e ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Yes 7.7 9.8 9.9 16.6 8.7 12.6 No 92.3 90.2 90.1 83.4 91.3 87.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 286 417 233 289 519 706 Chi-squared (prob) 1.0 (0.329) 5.0 (0.026) 4.8 (0.029) P h y s i c a l i n d e x ( p u b l i c s c h o o l s T a b l e E . 2 6 o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Low (<0.5) 44.7 26.3 35.4 27.0 40.6 26.6 Medium (0.5­0.75) 46.5 53.9 57.5 47.4 51.4 51.2 High (>0.75) 8.9 19.8 7.1 25.6 8.1 22.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 Total 282 414 226 289 508 703 Chi-squared (prob) 31.6 (0.000) 30.3 (0.000) 53.8 (0.000) 1 1 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S (b) Public schools only A d e q u a t e c l a s s r o o m s ( p u b l i c T a b l e E . 2 7 s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Less then half necessary 0.0 0.0 0.0 0.0 0.0 0.0 More than half necessary 22.7 21.6 19.7 18.5 21.3 20.2 Required amount 77.3 78.4 80.3 81.5 78.7 79.8 Total 100.0 100.0 100.0 100.0 100.0 0.0 No. of observations 264 320 228 248 492 568 Chi-squared (prob) 0.1 (0.736) 0.1 (0.742) 0.2 (0.661) C l a s s r o o m s t h a t c a n n o t b e u s e d T a b l e E . 2 8 w h e n r a i n i n g ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 More than half 29.9 22.5 20.2 21.0 25.4 21.8 Less than half 22.3 11.6 18.4 15.3 20.5 13.2 None 47.7 65.9 61.4 63.7 54.1 65.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 264 320 228 248 492 568 Chi-squared (prob) 21.6 (0.000) 0.8 (0.665) 15.2 (0.001) P e r c e n t a g e o f c l a s s r o o m s w i t h a T a b l e E . 2 9 c h a l k b o a r d ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 None 2.7 0.6 1.8 0.4 2.2 0.5 Less than half 7.2 2.2 4.4 0.4 5.9 1.4 More than half 20.8 3.4 26.3 4.0 23.4 3.7 All 69.3 93.8 67.5 95.2 68.5 94.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 264 320 228 248 492 568 Chi-squared (prob) 61.2 (0.000) 61.4 (0.000) 122.0 (0.000) 1 1 6 A N N E X E S B o a r d q u a l i t y ( p u b l i c s c h o o l s T a b l e E . 3 0 o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Poor 10.0 9.1 5.4 10.5 7.9 9.7 Fair 34.9 18.3 38.9 9.7 36.7 14.5 Good 51.3 67.8 49.3 79.8 50.4 73.1 Excellent 3.8 4.7 6.3 0.0 5.0 2.7 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 261 317 221 248 482 565 Chi-squared (prob) 22.1 (0.000) 78.6 (0.000) 78.0 (0.000) O w n w a t e r s u p p l y ( p u b l i c s c h o o l s T a b l e E . 3 1 o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Yes 13.3 21.3 16.7 26.6 14.8 23.6 No 86.7 78.8 83.3 73.4 85.2 76.4 Total 100 100 100 100 100 100 No. of observations 264 320 228 248 492 568 Chi-squared (prob) 6.4 (0.012) 6.0 (0.009) 12.9 (0.000) T a b l e E . 3 2 L i b r a r y ( p u b l i c s c h o o l s o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Yes 5.7 9.7 10.1 16.1 7.7 12.5 No 94.3 90.3 89.9 83.9 92.3 87.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 No. of observations 264 320 228 248 492 568 Chi-squared (prob) 3.2 (0.074) 3.8 (0.052) 6.5 (0.011) 1 1 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P h y s i c a l i n d e x ( p u b l i c s c h o o l s T a b l e E . 3 3 o n l y ) Primary Middle/JSS Total 1988 2003 1988 2003 1988 2003 Low (<0.5) 46.7 30.0 35.3 28.6 41.5 29.4 Medium (0.5­0.75) 46.4 53.9 58.4 50.0 51.9 52.2 High (>0.75) 6.9 16.1 6.3 21.4 6.6 18.4 Total 100 100 100 100 100 100 Total 261 317 221 248 482 565 Chi-squared (prob) 31.6 (0.000) 30.3 (0.000) 53.8 (0.000) (c) Panel data T a b l e E . 3 4 A d e q u a t e n u m b e r o f c l a s s r o o m s 1988 2003 Less then half necessary 0.0 0.0 More than half necessary 14.8 18.4 Required amount 85.2 81.6 Total 100.0 100.0 No. of observations 196 196 C l a s s r o o m s t h a t c a n n o t b e u s e d T a b l e E . 3 5 w h e n r a i n i n g 1988 2003 More than half 29.6 22.4 Less than half 24.5 19.4 None 45.9 58.2 Total 100.0 100.0 No. of observations 196 196 1 1 8 A N N E X E S P e r c e n t a g e o f c l a s s r o o m s w i t h a T a b l e E . 3 6 c h a l k b o a r d 1988 2003 None 1.0 0.5 Less than half 4.1 2.0 More than half 23.5 3.6 All 71.4 93.9 Total 100.0 100.0 No. of observations 196 196 T a b l e E . 3 7 C h a l k b o a r d q u a l i t y 1988 2003 Poor 8.9 9.9 Fair 35.9 19.3 Good 47.9 66.7 Excellent 7.3 4.2 Total 100.0 100.0 No. of observations 192 192 T a b l e E . 3 8 W a t e r 1988 2003 No 85.7 81.1 Yes 14.3 18.9 Total 100.0 100.0 No. of observations 196 196 T a b l e E . 3 9 L i b r a r y 1988 2003 No 92.9 86.7 Yes 7.1 13.3 Total 100.0 100.0 No. of observations 196 196 1 1 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S (d) Public versus private, 2003 T a b l e E . 4 0 A d e q u a t e c l a s s r o o m s Public Private Total Less then half necessary 0.0 0.0 0.0 More than half necessary 20.2 18.1 19.8 Required amount 79.8 81.9 80.2 Total 100.0 100.0 100.0 No. of observations 568 138 706 Chi-squared (prob) 0.3 (0.573) C l a s s r o o m s t h a t c a n n o t b e u s e d T a b l e E . 4 1 w h e n r a i n i n g Public Private Total More than half 21.8 11.6 19.8 Less than half 13.2 10.9 12.7 None 65.0 77.5 67.4 Total 100.0 100.0 100.0 No. of observations 568 138 706 Chi-squared (prob) 8.9 (0.011) S h a r e o f c l a s s r o o m s w i t h a T a b l e E . 4 2 c h a l k b o a r d Public Private Total None 0.5 0.7 0.6 Less than half 1.4 1.4 1.4 More than half 3.7 3.6 3.7 All 94.4 94.2 94.3 Total 100.0 100.0 100.0 No. of observations 568 138 706 Chi-squared (prob) 0.1 (0.994) 1 2 0 A N N E X E S T a b l e E . 4 3 B o a r d q u a l i t y Public Private Total Poor 9.7 4.3 8.7 Fair 14.5 15.2 14.7 Good 73.1 75.4 73.5 Excellent 2.7 5.1 3.1 Total 100.0 100.0 100.0 No. of observations 565 138 703 Chi-squared (prob) 5.9 (0.117) T a b l e E . 4 4 O w n w a t e r s u p p l y Public Private Total Yes 76.4 46.4 70.5 No 23.6 53.6 29.5 Total 100.0 100.0 100.0 No. of observations 568 138 706 Chi-squared (prob) 48.2 (0.000) T a b l e E . 4 5 L i b r a r y Public Private Total Yes 87.5 87.0 87.4 No 12.5 13.0 12.6 Total 100.0 100.0 100.0 No. of observations 568 138 706 Chi-squared (prob) 0.0298 (0.863) T a b l e E . 4 6 P h y s i c a l i n d e x Public Private Total Low (<0.5) 29.4 15.2 26.6 Medium (0.5­0.75) 52.2 47.1 51.2 High (>0.75) 18.4 37.7 22.2 Total 100.0 100.0 100.0 Total 565 138 703 Chi-squared (prob) 27.5 (0.000) 1 2 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S F r e q u e n c y o f h e a d - t e a c h e r a n d c i r c u i t T a b l e E . 4 7 s u p e r v i s o r a c t i v i t i e s Head-teacher Circuit supervisor Other Other QUIPS WSD schools Total QUIPS WSD schools Total Sits in on class Never 4 12 303 319 30 50 1749 1829 Less than once a week 25 20 802 847 40 65 1094 1199 At least once a week 35 65 1388 1488 4 5 70 79 Daily 8 18 333 359 0 0 13 13 Total 72 115 2826 3013 74 120 2926 3120 Looks at a sample of students' work Never 3 11 233 247 34 53 1673 1760 Less than once a week 37 33 1116 1186 38 62 1199 1299 At least once a week 31 62 1303 1396 2 4 47 53 Daily 1 8 171 180 0 0 7 7 Total 72 114 2823 3009 74 119 2926 3119 Looks at lesson plans Never 0 0 68 68 22 48 1449 1519 Less than weekly 0 0 72 72 47 69 1418 1534 At least once a week 72 115 2684 2871 5 3 60 68 Total 72 115 2824 3011 74 120 2927 3121 Discusses lesson plans Never 9 10 310 329 31 57 1717 1805 Less than weekly 36 54 1214 1304 38 60 1149 1247 At least once a week 27 51 1296 1374 0 3 42 45 Total 72 115 2820 3007 69 120 2908 3097 Discusses career development Never 26 52 1207 1285 48 78 2134 2260 Less than once a month 30 36 1089 1464 23 28 633 684 At least once a month 16 27 522 256 3 14 153 170 Total 72 115 2818 3005 74 120 2920 3114 1 2 2 A N N E X E S F r e q u e n c y o f h e a d - t e a c h e r a n d c i r c u i t T a b l e E . 4 8 s u p e r v i s o r a c t i v i t i e s ( p e r c e n t ) Head-teacher Circuit supervisor Other Other QUIPS WSD schools Total QUIPS WSD schools Total Sits in on class Never 5.6 10.4 10.7 10.6 40.5 41.7 59.8 58.6 Less than once a week 34.7 17.4 28.4 28.1 54.1 54.2 37.4 38.4 At least once a week 48.6 56.5 49.1 49.4 5.4 4.2 2.4 2.5 Daily 11.1 15.7 11.8 11.9 0.0 0.0 0.4 0.4 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Memo: absolute total 72 115 2826 3013 74 120 2926 3120 Prob. value 0.251 0.147 0.090 0.020 Looks at a sample of students' work Never 4.2 9.6 8.3 8.2 45.9 44.5 57.2 56.4 Less than once a week 51.4 28.9 39.5 39.4 51.4 52.1 41.0 41.6 At least once a week 43.1 54.4 46.2 46.4 2.7 3.4 1.6 1.7 Daily 1.4 7.0 6.1 6.0 0.0 0.0 0.2 0.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Memo: absolute total 72 114 2823 3009 74 119 2926 3119 Prob. value 0.029 0.255 0.214 0.038 Looks at lesson plans Never 0.0 0.0 2.4 2.3 29.7 40.0 49.5 48.7 Less than weekly 0.0 0.0 2.5 2.4 63.5 57.5 48.4 49.2 At least once a week 100.0 100.0 95.0 95.4 6.8 2.5 2.0 2.2 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Memo: absolute total 72 115 2824 187 74 120 2927 3121 Prob value 0.113 0.036 0.000 0.109 Discusses lesson plans Never 12.5 8.7 11.0 10.9 44.9 47.5 59.0 58.3 Less than weekly 50.0 47.0 43.0 43.4 55.1 50.0 39.5 40.3 At least once a week 37.5 44.3 46.0 45.7 0.0 2.5 1.4 1.5 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Memo: absolute total 72 115 2820 3007 69 120 2908 3097 Prob. value 0.114 0.376 0.014 0.039 Discusses career development Never 36.1 45.2 42.8 42.8 64.9 65.0 73.1 72.6 Less than once a month 41.7 31.3 38.6 48.7 31.1 23.3 21.7 22.0 At least once a month 22.2 23.5 18.5 8.5 4.1 11.7 5.2 5.5 Total 100.1 100.4 100.0 100.0 100.0 100.0 100.0 100.0 Memo: absolute total 72 115 2818 3005 74 120 2920 3114 Prob. value 0.565 0.331 0.139 0.008 1 2 3 ANNEX F: VARIABLE DEFINITION Variable Name Data and construction Physical: quantity Total number of classrooms CROOMS All classrooms at the school (including unusable) Required classrooms REQCROOM Sum of number of classes in each grade, where the number is divided by 2 if that grade is split shift Adequate number of classrooms ADQCROOM CROOMS (less those which cannot be used at any time) divided by REQCROOM ADQCROOMC Categorical version of ADQCROOM: 1 (0­.49), 2 (0.50­.89), 3 (>= 0.9) % of classrooms with chalkboard BOARD No. of classrooms with chalkboard/CROOMS Own water supply WATER If the school has own water supply No=1, Yes=1 Library LIB If the school has a library No=1, Yes=1 Physical: quality % of classrooms that cannot RCROOMS Classrooms that cannot be used at all plus those that cannot be used when raining be used when raining divided by CROOMS RCROOMSC Categorical version of RCROOM (reversed): 1 (>= 0.50), 2 (0.01­.049), 3 (0) Classes held in shared classrooms* CLASSSHA School average (from teacher questionnaire) of if have to teach in a shared classroom Size of internal walls* CLASSWLA School average (from teacher questionnaire) of if have fall size internal walls Noise disruption* CLASSNSA School average (from teacher questionnaire) of if class is disturbed by external noise Board quality BOARDQUAL School survey respondent assessment of overall chalkboard quality School average of teacher BQUALTA Average of teacher responses as to quality of chalkboard in assessment of board quality* the class in which they teach School average of teacher BQUALSZA Average of teacher responses as to the size of chalkboard assessment of if adequate in the class in which they teach board size School average of teacher BQUALCLA Average of teacher responses as to how easily the assessment of if board can chalkboard in the class in which they teach can be cleaned be cleaned Type of Water/Storage WATTYPE Type of water supply from piped water to reservoir/other. Recurrent: quantity (continued on next page) 1 2 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S Variable Name Data and construction Chalk availability CHALK School survey assessment of chalk availability English textbooks­pupil ratio EBOOKR Sum of English books available in each grade divided by total enrollments (ENR) EBOOKC Categorical version of EBOOKR: 1 (0­0.49), 2 (0.50­0.99), 3 (>=1) Mathematics textbooks­pupil ratio MBOOKR Sum of math books available in each grade divided by total enrollments (ENR) MBOOKC Categorical version of MBOOKR: 1 (0­0.49), 2 (0.50­0.99), 3 (>=1) Writing places­pupil ratio DESKS Total writing places (sum across ps6q12b1-b6) divided to adjusted enrollments (ENRA) DESKSC Categorical version of DESKS: 1 (0­.09), 2 (0.10­0.89), 3 (>= 0.90) Seating places­pupil ratio* SEATS Total writing places (sum across ps6q11a1-b6) divided to adjusted enrollments (ENRA) SEATSC Categorical version of SEATS: 1 (0­.09), 2 (0.10­0.89), 3 (>= 0.90) Teachers and teaching methods Number of teachers NOTEACH This variable is problematic since we do not want to include detached heads. They should not be included in the roster, so the variable is calculated by summing the number of teachers in the roster. However, in some cases it appears the head was included though detached. These cases have been adjusted by inspection. Required number of teachers REQTEACH Calculated in the same way as REQCROOMS but without the adjustment for split shift. Adequate number of teachers ADQTEACH NOTEACH/REQTEACH % trained teachers TEATRAIN Proportion of teachers who have teacher training. Absenteeism ABSENT Proportion of teachers absent for reasons other than sickness (note reference period longer in 1988). Teacher test scores TSCORE Simple average of teacher's scores on English, math and Raven's tests. Head's assessment of if a GOODTEACH Head's subjective assessment good teacher* GOODTEACHA School average of GOODTEACH Teaching methods* Teacher morale (subjective)* TMORALE If teacher enjoys being a teacher (No=1, Yes=1); ts6q5 TMORALEA School average of TMORALE TWCOND Teacher assessment of working conditions (1 Very Poor, 2 Poor, 3 Good, 4 Very Good) TWCONDA School average of TWCOND Frequency of homework* AVEHOME How often homework is set per week, averaged over all three subjects (or number for which data available) Frequency student work inspected* AVEINSPECT How often students' work is inspected per week, averaged over all three subjects (or number for which data available) Frequency student work assessed* AVEASSESS How often students' work is assessed per week, averaged over all three subjects (or number for which data available) Monitoring of student performance* STMONITOR Simple average of AVEHOME, AVEINSPECT and AVEASSESS STMONITOR School average of STMONITOR (continued on next page) 1 2 6 A N N E X E S Variable Name Data and construction School management SMC* SMC If there is a SMC at the school SMCMEET If the SMC has met in last month (No: 0; Yes: 1) SMCHELP If the SMC has provided help to the school during the last year (No: 0; Yes: 1) PTA PTA If there is a PTA at the school PTAMEET If the PTA has met in last month (No: 0; Yes: 1) SMCHELP If the PTA has provided help to the school during the last year (No: 0; Yes: 1) SPAM* SPAM If the school has had a SPAM in the previous year (No: 0; Yes: 1) SPAMPLAN Frequency of visits by CSVISIT How often the circuit supervisor (school inspector in 1988) circuit supervisor has visited the school in the last 12 months CSVISITC Categorical version of CSVIST: 0 (Never), 1 (1­5 times); 2 (6­11 times); 3 (12 or more times) Activities of head teacher HTCLASS and circuit supervisor* CSCLASS HTCLASSA CSCLASSA Frequency with which head teacher (circuit supervisor) sat in on class (per day). HTCLASSA and CSCLASSA are school averages. HTSMP, CSSMP HTSMPA, CSSMPA Frequency with which head teacher (circuit supervisor) inspected students' work (per day). HTSMPA and CSAMPA are school averages. HTLLESS, CSLLESS HTLLESSA, Frequency with which head teacher (circuit supervisor) looked CSLLESSA at lesson plans (per day). HTLLESSA and CSLLESSA are school averages. HTDLESS, CSDLESS HTDLESS, CSDLESS Frequency with which head teacher (circuit supervisor) discussed lesson plans. HTDLESSA and CSDLESSA are school averages. HTDCD, CSDCD HTDCDA, CSDCDA Frequency with which head teacher (circuit supervisor) discussed career development. HTDCDA and CSDCDA are school averages. Pupil data Total enrollments ENR Total enrollments Adjusted enrollments ENRA Enrollments adjusted for split shifts. Note: * not available from 1988 data 1 2 7 ANNEX G: ANALYSIS OF TEST SCORES Introduction of the tests score analysis, and the same proce- I n 1988/89 cognitive tests were administered dure was followed for this study. The short Eng- to a sub-sample (1,594 households) of the lish reading test consists of a few English 3,200 households interviewed by the second sentences that make a short story.112 The person round of the Ghana Living Standard Survey is required to read the sentences and then (GLSS2).110 The 2003 survey used exactly the answer eight multiple-choice questions to meas- same tests as those used in 1988/89, and included ure the person's understanding of the story. The also a local language test. The second part of this Short Mathematics test consists of eight arithmetic annex describes the tests, discusses their limi- operations (addition, subtraction, multiplication, tations, and evaluates their reliability. The third and division), of increasing difficulty. The part presents a descriptive analysis of the tests advanced English and mathematics test are the results and the fourth part builds an economet- same tests used by a study on educational ric model of the determinants test outcomes. achievements conducted in Kenya and Tanza- nia in 1980. These tests were designed by the Description of the Cognitive Tests Used Educational Testing Service, based in Prince- A total of seven tests were administered to mem- ton, New Jersey, based on school-leaving exam- bers of the 1,740 households interviewed in inations of primary and secondary Kenyan and 2003. These tests are: Tanzanian students (Knight and Sabot 1990). · Raven's Colored Progressive Matrices test The advanced English test contains 29 multiple- · Short local language test choice questions. Some of the questions, as in · Short English test the short English test, are based on the under- · Short math test standing of a short story. In other cases, the per- · Advanced local language test son taking the test has to select the correct word · Advanced English test from a choice of four in a sequence of sen- · Advanced math test tences.113 The advanced mathematics test con- sists of 36 questions of increasing difficulty. The The Raven's Progressive Matrices test is a meas- questions cover the knowledge of all basic math- ure intellectual ability intended to be inde- ematics, including arithmetic, fractions and pro- pendent of education and experience.111 The test portions, decimal numbers, real problem solving, consists of a puzzle with a missing piece that the geometry, equations, and algebra. The local lan- person taking the test has to find among a guage tests, short and advanced, are a new fea- choice of 6 possible pieces. The test comprises ture of the 2003 survey. They were designed by 36 questions divided into three parts of 12 ques- the Department of Linguistics and Ghanaian tions each. The short English and math test Languages at the University of Ghana in asso- were originally designed as a screening device ciation with the Department of Linguistics at to prevent people with very low skills from the University College of Education at Winneba. attempting the more advanced tests (Glewwe They consist of a translation/adaptation of the 1999). In 1988, given the low performance in the English tests to the most commonly spoken advanced test, the short test results became part Ghanaian languages.114 1 2 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S All household members aged between 9 and local languages was rejected for the same rea- 55 were eligible to take the tests.115 A maximum son that a broader range of test instruments was time for the completion of the tests, ranging from not used. That is that comparability was needed 10 minutes for the short English to 30 minutes with the 1988 tests. This comparability is cen- for the advanced mathematics test, was applied. tral to the study design, and was preserved at Despite the large number of tests and the length the expense of using test instruments known to of time required for completion, the rate of be imperfect. reported refusal on the short tests was only 5 per- In order to assess the reliability of the tests cent. On average, however, only 50 percent of "Cronbach's alpha" was calculated. Cronbach's individuals taking the short test were able to take alpha is a commonly used measure of the qual- the advanced ones, and many eligible individ- ity of test instruments, which captures their uals did not take the short tests because they internal consistency. The statistic is calculated found them too difficult. The 2003 survey using the responses to individual questions (i) screened eligible individuals with the use of for different individuals (j). If a test is internally flash cards, thus reducing the number of peo- consistent, then the scores across questions of ple taking the test.116 This procedure has some two individuals should be reasonably corre- implications for the measurement of mean test lated. People doing well will do well on the scores that are discussed below. Details of the same questions and badly on the same ones. procedure adopted for the administration of the People who do badly should get the same ques- tests can be found in the Test Administrator's tions wrong as other people doing badly. Cron- Instruction Manual designed for the survey which bach's alpha is based on the correlation is available on the study website. coefficient between the test scores of all possi- ble pairs of test takers. To calculate the statis- Limitations and Reliability of the Tests tic a random sample of 100 people was used for There are three limitations to the use of the set each test. A value of the alpha statistic of 0.7 and of tests used in this study. First, the use of above is considered to be an indication of a rea- numeracy and literacy tests focuses on a narrow sonable degree of consistency. The Raven's test range of child development, ignoring other scores turned to be very high, and all other tests aspects covered in the school curriculum.117 A also have coefficients equal to or in excess of second limitation was the use of a language 0.7 other than the short local language test, (English) for the literacy test, which is the first which at 0.64 falls not far below the threshold language for only a very small minority of Ghana- (Table G.1). ian children. This problem has been partially cor- rected by the introduction of a test in Ghanaian Descriptive Analysis of the Tests languages, that allowed children to be tested in Before describing test performance and com- both languages. This fact will particularly affect paring scores of 1988 with those of 2003, the the test scores of younger children since until computation of test scores used for the analysis recently local language has been the medium of is discussed. Short tests were composed of eight instruction until the end of third grade. A third questions, and only individuals scoring five or problem is that the mathematics test was more were allowed to take the advanced tests designed in English, thus giving an advantage of any type. Additionally, a screening mechanism to those individuals that are proficient in this lan- was used, whereby the person was invited to guage. Test administrators in the field were take the short test only if able to read very short advised to translate the mathematics questions sentences or able to solve a simple arithmetic whenever necessary. However, we performed operation displayed on a flash card. As a result, regression analysis, not reported here, which not all individuals were administered the entire shows that mathematics test scores were to set of tests. This procedure results in a problem some extent dependent on English proficiency. in the computation of the tests scores that is The possibility of re-designing the math test in resolved in the following section. 1 3 0 A N N E X E S V a l u e s o f t h e C r o n b a c h ' s a l p h a T a b l e G . 1 t e s t Short Advanced English 0.72 0.79 Mathematics 0.75 0.82 Local language 0.64 0.70 Raven 0.94 n.a. Censoring and screening 2003. The data suggest that this problem is more As in 1988, the advanced tests proved to be very serious for the short English scores than for the difficult, and the majority of the persons involved short math. In order not to lose information, the in the exercise did not score more than five on option of dropping all zero scores from the the short tests. On the other hand, for people 1988 or both data sets was rejected (a solution with higher level of education, the short tests that would have introduced other problems of were very easy, and large number of persons sample selection). Instead, we assigned a score scored eight, which is the maximum score. The of zero to those individuals who did not take the latter can be seen as a problem of censoring from English test because it was too difficult, and a above. To clarify this point, assume that the random score between 0 and 4 to those who English test scores are a measure of the latent were not able to take the mathematics test.119 variable English language ability. However, the test has eight questions so that there is a maxi- Combined test scores mum score of eight. If the data had not been cen- In the next section a multivariate model is pre- sored at eight, then those with higher ability sented explaining test scores by individual, would have scored more than eight. But they household, and school characteristics. The test could not do so, so that scores are bunched (cen- scores used in the regressions are a combination sored) at eight. This bunching constitutes a of the short and the advanced tests scores. Since problem when we want to compare mean scores short and advanced data were missing for many of different groups or surveys, because simple people it was necessary to impute the missing means of test scores will under-estimate the real test scores. difference in ability between the groups. A for- The short test scores, ranging from 0 to 8, mula can be applied that adjusts the simple show little variability across the sample, and an means for censoring.118 obvious solution is to add the short test scores As a result of applying the screening process to the advanced ones to get a combined test described above, many individuals were reported score. However, not all the children took the as not having taken the test in 2003 because it advanced test, because they scored less than 5 was "too difficult." On the other hand, in 1988, or found the test too difficult. Also, some peo- when no screening was applied, few people said ple who took the advanced test skipped the short the test was too difficult, but many scored zero test as being too easy. Various options were on the short tests (especially in English when 14 examined in order to impute scores, and the fol- percent scored zero compared to only 1.3 per- lowing were adopted. To impute advanced test cent in 2003; for math these figures are 2.5 and scores, the advanced scores were regressed on 0.3 percent, respectively) -- these people would the short scores of reading, math, and the Raven's have been screened out using the 2003 proce- test. The resulting predicted values were used dure. Ignoring them will overestimate mean to calculate scores for who did not take the scores of 2003 respect to those of 1988, since advanced test.120 Similarly, regressions were esti- those who were scoring zero in 1988 are likely mated of short tests scores on the advanced test to having been screened out of the sample in scores of reading and math and we calculated 1 3 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S A v e r a g e t e s t s s c o r e s : w h o l e T a b l e G . 2 s a m p l e 1988 2003 t-stat p-value Raven's 19.4 19.4 0.11 0.914 Short English* 6.2 6.6 3.75 0.000 Short math* 5.5 5.9 8.16 0.000 Short local* ... 6.4 Advanced English 12.3 13.2 4.16 0.000 Advanced math 8.7 10.1 6.93 0.000 Advanced local ... 15.5 Combined English 17.7 19.2 5.28 0.000 Combined math 14.5 16.2 6.26 0.000 Combined local ... 21.1 * Corrected for right censoring. the predicted values in order to impute values show a significant improvement in all test scores for the missing short tests scores.121 For the lat- for both groups other than the Raven's test, ter model a tobit specification was used cor- which has increased for primary graduates only. recting for both right and left censoring. Test score improvement by income group Overview of test score outcomes The improvement in test scores can be observed Tables G.2 and G.3 show the average test scores for children from all income groups (Table G.4). for all sample individuals and for primary grad- For primary students the improvement has lev- uates of 2003 and 1988. Tests marked by a star eled the performance between children from dif- have been corrected for right censoring as ferent backgrounds in English, but with the described above. The last columns of the table opposite effect for math. The latter is also true report t-statistics and p-values of the difference for JSS, whereas for English scores for JSS stu- in the means between the two surveys. The data dents the benefit has been uniform. A v e r a g e t e s t s s c o r e s : p r i m a r y T a b l e G . 3 g r a d u a t e s u n d e r 1 5 1988 2003 t-stat p-value Raven's 20.2 22.3 2.87 0.004 Short English* 4.7 7.0 5.87 0.000 Short math* 5.2 6.3 3.42 0.000 Short local* ... 6.2 Advanced English 10.2 14.5 6.88 0.000 Advanced math 7.2 9.7 5.04 0.000 Advanced local ... 14.1 Combined English 14.5 20.8 7.72 0.000 Combined math 12.7 15.9 5.44 0.000 Combined local ... 19.8 * Corrected for right censoring. 1 3 2 A N N E X E S T e s t s c o r e s b y s c h o o l i n g a n d T a b l e G . 4 i n c o m e t e r c i l e 1988 2003 I II III I II III Primary school students Raven's 15.7 15.8 16.9 15.9 15.6 17.4 Short English 0.8 1.7 3.7 4.3 4.8 5.5 Short math 3.4 3.7 4.0 4.0 3.8 4.5 Advanced English 3.4 8.6 10.2 12.3 12.3 14.0 Advanced math 5.2 4.9 6.0 14.0 11.7 15.3 JSS students Raven's 20.2 19.6 21.8 20.1 21.1 24.3 Short English 4.0 4.0 4.8 6.1 6.3 6.6 Short math 5.1 4.9 5.3 5.8 5.7 6.0 Advanced English 10.5 12.0 12.3 15.0 15.4 17.2 Advanced math 7.2 7.5 8.4 13.9 16.5 20.0 Modeling Test Scores Children that were purposely or incidentally In this section a model is defined and estimated excluded from the sample can be grouped into of children's test scores. Information from both the following three categories. surveys is used in order to detect determinants of changes in test scores over time. The inter- Eligible Children Not Tested. Some 25 percent of est is to assess the impact of school quality on the eligible children did not take the tests. About students' achievements using household survey one-fifth of these did not do so because they data, in order to control for the effects of indi- found the test too difficult (mostly in 2003). vidual and household characteristics. The This problem has been corrected by randomly dependent variables used in the model are the assigning a low grade (math) or a zero grade combined results of the mathematics, English, (English) to those children with at least three and local language tests described in the previ- years schooling. Another 20 percent of these chil- ous section. The following section describes the dren were in school at the time of the exercise, sample and discusses the selectivity problem 20 percent were traveling, 5 percent were ill, 5 from the choice of the sample. A listing of vari- percent refused to take the test, and 30 percent ables determining test scores follows, and finally did not take the test for other unspecified rea- the regression results are presented. sons. In general, the exclusion of eligible chil- dren from the exercise seems to have operated Sample selection randomly. There might be some concern, how- The sample used in the regressions consists of ever, that children "traveling" and those not tak- all children aged between 9 and 15 who have ing the test for "other" reasons could share recently attended or are currently attending common and distinctive characteristics. For school in the locality of residence. Only children example, they could be workers, and thus have with at least three years of schooling are less education. included. This sample is not a random sample, because only children either with at least three Children Living Elsewhere. Child fostering is very years of schooling (1988) or able to read the flash common in Western Africa. Nearly 50 percent of cards (2003) took the tests. Other children did the sample children are "foster" children, in the not take the test as a result of refusal or absence. sense they are not living in their household of 1 3 3 BOOKS, BUILDINGS, AND LEARNING OUTCOMES origin. Parents can use fostering for reasons as have common characteristics that prevented different as sending children to better schools, them form attending or completing school. How- reduce the burden on household resources, ever, the quality of the schools available and the exploit opportunities in more developed areas, job opportunities also matter. In Appendix I, and strengthening kinship ties (Lloyd and Gage- using a complete set of household and locality Brandon 1994). Though households of origin of explanatory variables, a model of school achieve- foster children may well be poorer on average ment is estimated whose results are used here than other households, foster children are not to correct for the selectivity bias caused by the necessarily sharing common characteristics, pre- sample being dependent upon school attainment. cisely because the reasons at the origin of the "fos- tering" choice are so different. Additionally, while Selectivity Adjustments 25 percent of children could not be interviewed The selection of a non-random sample can bias because they were "fostered out," another 25 per- the regression results. The presence of this bias cent were interviewed because they were "fos- can be tested for when necessary by including a tered in." Foster children are therefore largely sample selection term in the test score regression. represented in our sample. To clarify the problem, suppose that cognitive tests A more serious problem is whether it is pos- are administered to the sample of children grad- sible to relate foster children behavior to the char- uated from primary school. In order to assess the acteristics of the fostering household. It could be importance of school quality, a regression is esti- argued that the household of origin makes mated of test scores using a school quality index choices for these children, rather than the house- as a regressor, resulting in a positive coefficient hold of residence (Glewwe 1999) so that the rel- on this variable. But suppose that only wealthier evant household characteristics are missing for families can afford for their children to complete children who are fostered in. A priori this hypoth- primary school. Also suppose that children from esis appears most appropriate for the schooling wealthier families are better nourished, do not decision, whereas for test scores the household work, and that this improves their mental devel- of residence may be the most relevant. The opment and their performance at cognitive tests. hypothesis that it is true parental characteristics As a result, we might erroneously attribute good can be tested by interacting a dummy for foster test scores to school characteristics, while they are, children (F=1 if fostered in, 0 otherwise) with at least in part, determined by factors related to household characteristics (X). The coefficient household wealth. on FX will be insignificant if household char- Since selection bias in the description given acteristics affect fostered children in the same above can be seen as an omitted variable prob- way as they do non-fostered children. If it is the lem, the presence of the bias should be detected characteristics of the "true household" that mat- by a test on omitted variables such as the Ram- ter then the coefficient will be equal and oppo- sey test. If a selectivity bias is found a correc- site to that on X. tion can be made by modeling the selection of children to be in the sample. Children Attending School. The sample chosen Algebraically, the equation of interest is: for the estimation implies that children who never attended school, or who left school before yi = xi + (1) i completing primary (or the third year of primary depending on the sample), are not considered. where yi is the test score for child i, xi is the qua- In 2003, 11 percent of children aged between 10 lity index of the school attended by the child and and 20 had never attended schools, and 10 per- is the parameter to be estimated. But only a se- cent of children aged over 20 starting primary lected number of children enter equation (1), since school never achieved grade six. The same per- many children never go to school or drop out be- centages are considerably higher for the children fore reaching a given grade. We can use the entire surveyed in 1988. These children are likely to sample of children and model school attendance 134 A N N E X E S based on a series of child and household charac- Schooling: (continuous) number of completed teristics. This is called the `selection equation': years of schooling122 Order: (continuous) child's birth order zi = jw + ui * (2) Siblings: (continuous) number of alive siblings ji j Ability: (continuous) estimated innate ability where zi is a variable defining whether the * Ability missing: (dummy) children whose innate child is attending school or not and the wji are ability is missing a set of explanatory variables. The variable yi is observed only when zi is larger than zero. The ability variable was estimated using the * Equation (1) corrected for selectivity is thus the model formulated by Glewwe (1999). This model expectations of yi conditional on zi being larger regresses Raven's test scores on the age, sex, * than zero: and years of schooling of the child and of the par- ents. The model predicts Raven's scores for chil- E[yi|zi>0] = Xi + i(u) + vi (3) dren and their parents at the same time, calculating household fixed effects. The ability variable is where i(u) is the inverse Mill's ratio obtained obtained as the sum of the estimated household from the selection equation (Greene 2000). This fixed effect and the error term, thus assuming that ratio is usually derived after running a probit of children inherit their innate ability from their par- the selection equation and takes the form ents. Tables G.5 and G.6 present the results of the regression run on the 1988 and 2003 samples. The (X) = (4) results are very similar, and the only differences (X) are the significance of the age square term and where and are the density and the cumulated the coefficient estimates of years of schooling and density of the normal distribution, and the X are gender. It seems that gender and older age have the predicted values of the selection equation. A become less important in explaining poor per- significant is a test for the presence of selec- formance, while the effect of additional years of tion bias, and at the same time a correction of the schooling has increased. estimates of the equation (1) for that bias. In the case of the ordered probit achievement regression Household variables (used in Annex I)the inverse Mill's ratio for the Coast Forest and Savannah: (dummies) three children having attained at least a given grade is: main agro-ecological zones of the country 1 ­ (cutoff2 ­ X) excluding Accra = 1 ­ (cutoff2 ­ X) (5) Rural: (dummy) residence in rural areas as defined by the 1984 and 2000 demographic cen- where the cutoff is the threshold used for the sus respectively sample selection, for example the achievement Mother's education: (continuous) completed of at least grade three. years of schooling of the mother Father's education: (continuous) completed years Explanatory Variables of schooling of the father The explanatory variables used in the model are Per capita expenditure: (continuous) logarithm the characteristics of the child, of the household, of household per-capita expenditure. Expendi- and of the school attended. The variables are ture values of 1988 were actualised to 2003 listed in Annex F with their description (variable using the consumer price index with an * are not available for 1988). *Parents meeting the teacher: (dummy) the par- ents are regularly meeting the child's teacher in Individual variables order to discuss progress in school Sex: (dummy) child sex *PTA: (dummy) membership of any of the house- Age: (continuous) age of the child in completed hold member of local PTA (Parents and Teach- years ers Association) 1 3 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T a b l e G . 5 I n n a t e a b i l i t y r e g r e s s i o n ( 1 9 8 8 ) Coefficient t-statistic Age 0.36 6.19*** Age squared 0.00 ­4.81*** Education 0.69 5.19*** Education squared 0.01 2.19** Father's education ­0.01 ­1.21 Mother's education ­0.12 ­0.98 Sex 0.09 0.78 Sex*age 0.31 0.44 Sex*education ­0.06 ­2.49** Constant ­0.29 ­3.64*** Observations 1732 F-statistic 58.41 R square 0.32 T a b l e G . 6 I n n a t e a b i l i t y r e g r e s s i o n ( 2 0 0 3 ) Coefficient t-statistic Age 0.25 3.84*** Age squared 0.00 ­2.52** Education 1.14 7.54*** Education squared ­0.01 ­1.06 Father's education ­0.01 ­1.85* Mother's education ­0.02 ­0.14 Sex ­0.17 ­1.22 Sex*age 0.21 0.28 Sex*education ­0.01 ­0.49 Constant ­0.21 ­2.23*** Observations 1716 F-statistic 35.45 R square 0.23 School variables quantity and quality of classrooms and black- Index of recurrent inputs: this an index of recur- boards, availability of water and library. rent inputs described in Annex D. It includes *Classrooms with internal walls: (continuous) school availability of books, writing places and school average of classrooms with full size inter- chalk. nal walls Index of physical inputs: this an index of phys- *Noise disruption: (continuous) school average ical inputs described in Annex D. It includes of outside noise disturbing classes 1 3 6 A N N E X E S *Private: (dummy) whether the school attended eficiary of the WSD program improves them.124 is private Private school students perform significantly Teacher education: (continuous/categorical) better in English though not in math, which is average number of teacher s years of schooling an unsurprising result as English is the medium of instruction in private schools. Having own School management water supply is robustly significant. *SPAM plan: (dummy) school had a SPAM in the The data for 2003 allow the inclusion of last year whose plan was actually carried out rather more school characteristics and doing so Visits of circuit supervisor: (continuous/cate- leads to more satisfactory results, especially for gorical) number of inspection visits by the cir- JSS, as shown in Table G.8. Years of schooling cuit supervisor are significant and the coefficient rather higher. Textbook availability has the right sign in all four Other cases for JSS and is always significant. This is the Lambda: (continuous) is the inverse Mill's ratio case only for math books at primary level (Eng- obtained form the attainment regression reported lish books are negative but insignificant). Teach- in Annex I. ing methods significantly improve test scores for The set of test score equations -- English, math in JSS but have a perverse effect in primary. math, and local language -- might appear to Time on task is always positive and significantly comprise a system of seemingly unrelated regres- so in two cases. And there is substantial evidence sion equations (SURE), meaning that the error that the quality of infrastructure matters: test terms between each equation are correlated scores are significantly higher in schools in with one another. Using SUR estimation rather which classrooms have full-sized internal walls than OLS improves the efficiency of the estimates. and lower in ones where noise disrupts teach- However, SUR estimation requires the sample to ing. Participation in WSD positively affects both be the same for each equation, so including the English and math scores in JSS, and having been local language test reduces same size quite con- a PSD beneficiary is good for math scores. As siderably. Moreover, there is no gain if the before, English test scores are significantly bet- regressors are the same in each case. Whilst it ter in private schools, but this is not so for math might be thought that there could be some vari- (primary schools only, there is no difference in ation in the regressors (math books for math either for JSS). For both primary and JSS teach- scores etc.), these distinctions did not prove ers being able to speak the local language good ones to keep. Results are first presented improves student math scores.125 Finally, teach- (Table G.7) for English and math. The local lan- ers' perceptions of student discipline show that guage estimates, which gave rather different indiscipline significantly worsens English scores. results, are then discussed. Turning to other characteristics: rural children It proves quiet difficult to get good results perform worse, as do girls in math. The dummy from the pooled data. It is always the case that for fostered-in children has a staggeringly large schooling is positively and significantly related coefficient. The parental education interactive with higher test scores -- and this remains so dummies are not significant, suggesting that it is even if just one or two years of schooling are the actual residence that matters -- but in the case included in the model. However, interacting of income the term is significant, more than off- school quality variables with years of schooling setting the beneficial impact of household income does not yield good results. A few of the school on test scores. Variables measuring parental qualities have a "shift effect" on test score out- involvement also matter: having met with the stu- comes, notably math textbook availability is sig- dent's teacher improves math scores and being nificant in some, but by no means all, model in the PTA improves English scores. specifications.123 A high pupil-teacher ratio is Finally, when local language test scores are detrimental to English test scores (though appears included the most striking result is the lower good for maths scores in JSS), and being a ben- explanatory power of most the variables (results 1 3 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T e s t s c o r e d e t e r m i n a n t s : p o o l e d T a b l e G . 7 ( a ) d a t a c h i l d r e n i n p r i m a r y s c h o o l ( O L S ) Math English Coefficient z-stat Coefficient z-stat School variables Years schooling 0.92 2.39 ** 1.21 2.67 *** Math books 0.46 1.22 0.11 0.24 Classrooms can be used when raining 0.01 0.66 ­0.01 ­0.84 Water supply 1.28 1.44 2.46 2.48 ** Board quality 0.59 1.30 ­0.67 ­1.28 Teacher test score ­0.15 ­1.26 ­0.10 ­0.78 Private school 0.42 0.43 2.89 2.62 *** High PTR ­1.00 ­0.95 ­4.37 ­3.68 *** PSD dummy ­0.62 ­0.46 2.89 1.98 ** WSD dummy 0.83 0.50 3.14 1.67 * QUIPS dummy ­2.17 ­1.48 ­0.98 ­0.61 Community variables Forest ­1.27 ­1.90 * ­1.73 ­2.21 ** Savannah ­0.29 ­0.25 ­2.30 ­1.73 * Rural ­0.21 ­0.34 ­1.00 ­1.39 Child characteristics Age 0.29 1.44 0.26 1.10 Sex ­0.29 ­0.52 ­0.20 ­0.30 Ability 0.35 5.95 *** 0.49 7.34 *** Ability missing 4.09 4.45 *** 7.96 7.50 *** Household variables Mother's education 0.13 2.00 ** 0.23 3.26 *** Income 0.07 0.12 1.41 2.11 ** Other Survey dummy 3.56 4.12 *** 8.32 8.08 *** Selectivity correction ­1.25 ­1.02 0.41 0.28 Number of obs. 331 298 R2 0.29 0.57 not shown). Schooling still matters, as does Interpretation parental education. But virtually none of the Interpretation of the results is made by school variables contribute: none of the project analysing the change in test scores attributable dummies are significant, nor are any of the to the change in each of the determinants. The school input variables. However, a variable more complete model estimated using the 2003 measuring if teachers took the local language test data is used for this purpose. For variables col- is significant. lected in each of the two rounds the compari- 1 3 8 A N N E X E S T e s t s c o r e d e t e r m i n a n t s : p o o l e d T a b l e G . 7 ( b ) d a t a c h i l d r e n i n m i d d l e / J S s c h o o l ( O L S ) Math English Coefficient z-stat Coefficient z-stat School variables Years schooling 2.13 5.29 *** 3.25 5.95 *** English books 0.16 0.32 0.29 0.43 Math books 0.55 1.06 ­0.47 ­0.67 Classrooms can be used when raining ­0.01 ­0.91 ­0.01 ­0.61 Water supply 0.39 0.48 1.03 0.97 Board quality ­0.21 ­0.44 0.51 0.81 Teacher test score 0.10 0.82 0.08 0.48 Private school ­0.83 ­0.60 1.29 0.69 High PTR 2.99 1.90 * ­0.33 ­0.16 WSD dummy ­1.81 ­0.99 3.57 1.50 QUIPS dummy 2.41 1.06 2.86 0.97 Community variables Forest ­1.26 ­1.83 * ­0.54 ­0.59 Savannah ­0.28 ­0.23 ­0.32 ­0.20 Rural ­1.41 ­2.07 ** ­2.17 ­2.32 ** Child characteristics Age ­0.99 ­3.46 *** ­1.16 ­3.02 *** Sex ­2.39 ­4.08 *** ­1.31 ­1.64 * Ability 0.24 4.20 *** 0.42 5.55 *** Ability missing 2.39 2.62 *** 4.21 3.36 *** Household variables Mother's education 0.17 2.73 *** 0.16 1.95 ** Income ­0.72 ­1.23 ­0.18 ­0.22 Other Survey dummy 3.82 4.29 *** 8.44 6.99 *** Selectivity correction 1.30 0.60 0.68 0.19 Number of obs. 272 250 R2 0.45 0.58 son is made using the change in sample means Schooling improves test scores, each addi- (Table G.9a). For other variables (which are tional year increasing the combined English mostly 0­1 dummies) the minimum and maxi- score by 3.6 points and math by 4.9 points. mum values are used (Table G.9b). The excep- Enrollments have risen and dropout rates are tion is schooling, which saw a small drop in the low: in 2003, 95 percent of those beginning pri- sample, but the sample is known not to be rep- mary complete it and 86 of them complete JSS resentative. (Annex H). The 10 percent of the age group 1 3 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S T e s t s c o r e d e t e r m i n a n t s , p r i m a r y T a b l e G . 8 ( a ) s c h o o l s 2 0 0 3 ( O L S ) Math score English score Coeff z Coeff z School characteristics English books ­0.35 ­0.72 ­0.67 ­1.04 Math books 1.04 1.81 * 1.27 1.78 * Physical Index 4.96 1.92 * 4.79 1.61 Low PTR 2.22 2.05 * ­0.57 ­0.39 High PTR ­0.51 ­0.45 ­4.15 ­3.05 *** Teachers speak local language 2.65 1.90 * ­0.21 ­0.13 Teachers' discipline 1.27 2.11 ** 0.56 0.91 Teaching methods ­5.77 ­1.73 * ­4.54 ­1.13 Time on task 0.07 1.66 * 0.03 0.60 Board easy to clean ­0.14 ­0.09 ­2.34 ­0.93 Classrooms have internal walls 1.94 1.45 1.22 0.61 Class disrupted by noise ­1.22 ­0.77 ­1.60 ­1.00 Private school 0.53 0.33 4.05 1.90 * PSD dummy ­1.23 ­1.11 2.95 2.52 ** WSD dummy 0.61 0.38 3.51 1.95 * Circuit supervisor discuss lesson plans ­3.17 ­0.17 15.16 0.68 Head teacher sits in on lessons 3.29 1.87 * 3.92 1.69 * Student monitoring 0.41 0.58 ­0.89 ­1.20 School had a SPAM 0.00 0.00 0.52 0.78 Teachers morale 1.26 1.10 0.38 0.25 Students indiscipline ­0.80 ­1.55 ­0.91 ­1.85 * Community characteristics Forest ­2.49 ­2.78 *** ­1.64 ­1.45 Savannah ­0.70 ­0.59 ­1.97 ­1.50 Rural ­1.96 ­2.52 ** ­2.70 ­2.79 *** Child characteristics Age 0.52 1.87 * 0.26 0.77 Sex 0.05 0.08 0.10 0.12 Years of schooling 1.19 2.56 ** 1.58 2.78 *** Ability 0.38 4.40 *** 0.50 6.95 *** Ability missing 5.97 3.34 *** 7.22 3.34 *** Fostered in 14.89 0.49 23.30 0.73 Fostered*income ­1.12 ­0.56 ­1.50 ­0.70 Household characteristics Mother's education 0.17 1.94 * 0.31 2.67 ** Income 0.29 0.36 2.03 2.24 ** Parent in PTA 1.33 1.13 ­0.88 ­0.64 Parent met with teacher ­1.80 ­2.40 ** ­0.72 ­0.64 Other variables Selectivity correction 0.97 0.54 3.48 1.74 * Constant ­18.33 ­1.27 ­37.97 ­2.24 ** No. of obs. 206 204 R2 0.38 0.45 1 4 0 A N N E X E S T e s t s c o r e d e t e r m i n a n t s , j u n i o r T a b l e G . 8 ( b ) s e c o n d a r y s c h o o l s 2 0 0 3 ( O L S ) Math score English score Coeff. z Coeff. z School characteristics English books 1.44 2.34 ** 0.04 0.05 Math books 2.38 3.52 *** 2.20 3.20 *** Physical Index 0.19 0.08 4.48 1.57 Low PTR ­2.62 ­3.18 *** ­0.81 ­0.73 High PTR 3.19 1.84 * 2.65 1.56 Teachers speak local language 2.26 1.96 * 0.34 0.17 Display material available 0.86 1.01 1.44 1.03 Teachers' discipline 0.85 1.29 ­0.11 ­0.12 Teaching methods 14.25 2.93 ** 9.24 1.37 Time on task ­0.03 ­0.38 0.19 3.13 *** Board easy to clean 3.44 1.67 * 1.72 0.50 Classrooms have internal walls 3.42 1.90 * 3.77 1.76 * Class disrupted by noise ­3.95 ­3.61 *** ­3.69 ­2.36 ** Private school 1.14 0.69 8.45 4.18 *** WSD dummy 3.55 2.48 ** 3.67 1.67 * QUIPS dummy 1.69 0.89 2.94 2.34 ** Circuit supervisor discuss lesson plans 22.43 1.66 * 63.34 2.55 ** Head teacher sits in on lessons ­3.58 ­2.09 ** ­5.35 ­1.97 * Student monitoring 0.11 0.21 ­0.11 ­0.19 Community characteristics Forest 0.51 0.53 2.90 2.23 ** Savannah 4.83 3.49 *** 4.16 2.61 ** Rural ­3.12 ­4.30 *** ­3.05 ­2.88 *** Child characteristics Age ­1.22 ­2.91 *** ­1.61 ­3.65 *** Sex ­2.32 ­2.81 *** ­0.06 ­0.06 Years of schooling 2.93 5.25 *** 3.39 4.63 *** Ability 0.13 1.82 * 0.23 2.11 ** Ability missing ­0.02 ­0.01 1.45 0.42 Fostered in 53.25 2.80 *** 62.79 2.00 ** Fostered*income ­3.42 ­2.68 *** ­4.15 ­2.00 ** Total hours worked 0.01 2.17 ** 0.02 2.11 ** Household characteristics Father's education 0.25 3.67 *** ­0.02 ­0.18 Mother's education 0.04 0.47 0.23 1.88 * Income 1.92 2.01 ** 1.73 1.17 Parent in PTA 3.01 1.89 * 4.07 2.24 ** Parent met with teacher 2.75 3.28 *** 0.47 0.49 Other variables Selectivity correction ­19.30 ­3.77 *** ­12.63 ­1.47 Constant ­35.36 ­1.91 * ­31.89 ­1.24 No. of observations 137 137 R2 0.67 0.59 1 4 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S C h a n g e i n t e s t s c o r e s a t t r i b u t a b l e T a b l e G . 9 ( a ) t o b e t w e e n s a m p l e c h a n g e s i n e x p l a n a t o r y v a r i a b l e s Sample mean Coefficient Change 1988 2003 English Math English Math Forest 0.60 0.40 3.51 0.86 ­0.70 ­0.17 Savannah 0.11 0.17 4.76 4.95 0.27 0.29 Rural 0.44 0.56 ­3.45 ­2.81 ­0.40 ­0.33 Age 12.48 12.54 ­1.70 ­1.30 ­0.09 ­0.07 Sex 0.45 0.47 0.13 ­1.60 0.00 ­0.04 Years of schooling 5.04 4.90 3.59 2.76 ­0.51 ­0.39 Ability 5.93 7.74 0.17 0.12 0.31 0.22 Missing ability 0.48 0.33 ­0.91 ­3.20 0.13 0.46 Father's education 6.24 7.81 ­0.04 0.12 ­0.06 0.19 Mother's education 3.81 5.14 0.16 0.02 0.21 0.03 Household expenditure 13.96 14.78 2.70 1.43 2.23 1.18 Fostered in 0.23 0.20 74.64 49.92 ­2.23 ­1.49 Fostered*expenditure 3.27 3.00 ­4.82 ­3.02 1.33 0.83 Hours worked 136.86 30.58 0.01 0.01 ­1.06 ­1.06 English books 1.69 2.47 0.17 1.70 0.13 1.32 Math books 2.00 2.44 1.86 1.53 0.80 0.66 Physical index 0.50 0.59 4.03 0.60 0.38 0.06 Low PTR 0.30 0.18 ­1.83 ­2.16 0.23 0.28 High PTR 0.03 0.11 0.88 1.04 0.08 0.09 attending school who would not have done so structure matter. The most important single vari- 15 years group can expect to have an increase able is teaching methods. If all teachers in the in their English score of 20 if they complete pri- school used modern methods then, compared mary and 27 if they go on to complete JSS. For to a situation in which none do so, children's math these figures are 16 and 21, respectively. English scores would be 6.2 higher and their The increase in recurrent and physical items math score 8.8. The three infrastructure variables between the two rounds increased math scores combined can improve English scores by 11.3 by 1.6 and English by 2.0 points. This under- points and math by 10.1. states the gains in the most deprived areas. Home factors also matters to student per- Ensuring that a school has one math and Eng- formance. The two measures of parental involve- lish book per child compared to the situation ment in a child's education (membership of PTA in the mid-1980s of one text per classroom will and meeting with a teacher) give a combined increase average English scores in children from impact of 3.5 and 3.9 on math and English that school by 6 points and math scores by scores respectively. Income also matters; eco- close to 10 points. nomic growth (the between sample rise in Turning to the variables collected only in incomes) has increased average English scores 2003, it is shown that both process and infra- by 2.2 and math scores by 1.2 points. 1 4 2 A N N E X E S C h a n g e i n t e s t s c o r e s a t t r i b u t a b l e T a b l e G . 9 ( b ) t o m a x i m u m p o s s i b l e c h a n g e s i n e x p l a n a t o r y v a r i a b l e s Assumed values Coefficient Change Low High English Math English Math Display material available 0.00 1.00 1.32 1.03 1.32 1.03 Discipline 0.00 6.00 0.70 1.14 4.20 6.84 Teaching methods 0.00 1.00 6.20 8.80 6.20 8.80 Board easy to clean 0.00 1.00 1.94 2.38 1.94 2.38 Internal class walls 0.00 1.00 5.08 3.87 5.08 3.87 Class not disrupted by noise 0.00 1.00 4.28 3.81 4.28 3.81 Private 0.00 1.00 7.61 0.67 7.61 0.67 PSD 0.00 0.07 3.11 5.40 0.22 0.38 WSD 0.00 0.05 6.62 3.80 0.36 0.21 QUIPS 0.00 0.06 3.40 2.42 0.21 0.15 Circuit supervisor discuss lesson plan 0.00 1.00 22.40 ­1.12 22.40 ­1.12 Student monitoring (homework etc) 0.00 5.00 0.36 0.50 1.80 2.50 1 4 3 ANNEX H: DATA ON EDUCATIONAL PERFORMANCE What Has Happened to Enrollments? Some Definitions Table H.1 is reproduced from the 2002 Educa- Enrollment data may be either net or gross, for tion Sector Strategy. It shows primary enrollments which the definitions are as follows, given here in 2000 down slightly from those in 1990, hav- for the case of primary education: ing behaved erratically during the course of the decade. However, the report of the Ghana Sta- Net enrollment rate, NER = tistical Service (GSS) on the Ghana Living Stan- (No of children of primary school dards Survey reported a school attendance rate age in primary school) for 6­11 year olds of 73.2 percent for GLSS2 (No. of children of primary school age) (1988/89), rising a full 10 percent to 83.1 per- cent for GLSS4 (1998/99), see Table H.2. Gross enrollment rate, GER = To understand what is going on here and get an accurate picture of what is happening to (No of children in primary school) enrollments it is helpful to first clarify definitions. (No. of children of primary school age) O f f i c i a l d a t a o n p r i m a r y T a b l e H . 1 e n r o l l m e n t s Primary school Gross Proportion School-age enrollment Enrollment enrolled in Gender Year Population (Public& private) Ratio private schools parity 1986 2,173,089 1,679,072 77.3 4.1 0.81 1990 2,453,146 1,945,422 79.3 7.3 0.82 1991 2,544,676 2,011,062 79.0 10.2 0.84 1992 2,638,831 2,047,293 77.6 9.7 0.85 1993 2,736,919 2,138,635 78.1 10.7 0.85 1994 2,838,678 2,154,676 75.9 10.9 0.87 1995 2,944,253 2,197,172 74.6 11.0 0.87 1996 3,048,161 2,333,347 76.5 13.1 0.88 1997 3,155,758 2,445,353 77.5 13.1 0.89 1998 3,267,002 2,562,229 78.4 13.1 0.90 1999 3,382,649 2,684,689 79.4 13.1 0.91 2000 3,154,152** 2,477,990 78.6** Note: ** Data from the 2000/2001 MOEYS annual school census and population data from the 2000 national population census conducted by the Ghana Sta- tistical Service. All other population figures are based on projections from the 1984 Population Census. Source: SRIMPR Division MOEYS. 1 4 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S A t t e n d a n c e r a t e s r e p o r t e d f r o m T a b l e H . 2 G h a n a L i v i n g S t a n d a r d s S u r v e y GLSS2 GLSS4 Age range (1988/89) (1998/99) 6­11 73.2 83.1 12­15 71.8 80.4 16­18 54.1 47.0 19­25 14.1 13.5 However, using household survey data that ask cance of this category is slight compared to over- for the age of each of each household member age children still in primary school. Hence the NER and whether they are currently in school or not, will be close to the AR, but the gap between these it is most straightforward to work out an atten- two and GER may be quite substantial. dance rate (AR is the term used by GSS, it does Similar reasoning suggests a different order- not refer to the proportion of children enrolled ing for children in secondary school. Consider actually attending school, which is another mean- the case of SSS, for which the expected age is ing of the term), given here for children aged 16­18. In this case NER= 50 Age range 1988 2003 2003 (female) Source: GLSS2 and GSS/OED household survey Completion is correlated with income. Chil- data are not reported in the MOEYS publication dren from poorer households are less likely to Education Indicators at a Glance or in the Edu- complete their education than the children of the cation Sector Review. The 2003 GSS/OED sur- better off (Table H.13). However, completion vey asked how many pupils were in school on rates have improved for all income groups, so the day of the survey. The responses indicated that the poorest third today are more likely to attendance rates of around 80 percent (Table complete primary than were the richest third 15 H.14). Since non-attendance may be seasonal, years ago (Table H.13). at least in rural areas, only detailed data collec- tion, possibly of administrative sources, can Attendance really get at this issue. The school census collects data on attendance, In addition, the GLSS household questionnaire asking the average attendance for each grade for asks how many hours a child spent in school the a month preceding the census. However, these previous week. These data, shown in Table R e t e n t i o n r a t e b y e x p e n d i t u r e T a b l e H . 1 3 t e r c i l e 1988 2003 Tercile I Tercile II Tercile III Tercile I Tercile II Tercile III P1 0.98 0.99 0.99 P1 0.99 1.00 1.00 P2 0.96 0.97 0.98 P2 0.99 0.99 1.00 P3 0.93 0.93 0.96 P3 0.97 0.98 0.99 P4 0.90 0.91 0.93 P4 0.96 0.97 0.98 P5 0.86 0.88 0.92 P5 0.93 0.94 0.97 P6 0.83 0.86 0.89 P6 0.90 0.93 0.96 M1 0.79 0.81 0.86 JSS1 0.84 0.89 0.93 M2 0.74 0.77 0.82 JSS2 0.8 0.88 0.91 M3 0.69 0.72 0.78 JSS3 0.76 0.86 0.88 M4 0.65 0.68 0.74 SS1 0.35 0.52 0.58 SS1 0.18 0.26 0.33 1 5 6 A N N E X E S T a b l e H . 1 4 A t t e n d a n c e b y g r a d e Primary JSS Grade 1 82.0 JSS1 79.7 Grade 2 83.8 JSS2 79.9 Grade 3 81.9 JSS3 76.5 Grade 4 81.5 Grade 5 83.2 Grade 6 82.4 Primary 82.5 JSS 78.7 H.15, mainly capture the rise in school hours age cohorts based on the year of birth. Seven resulting from the official policy of lengthening years were added to the year of birth to give the the school day from 4 to 5 hours, the rise of extra age at initial enrollment and decade averages cal- classes and the longer hours in the increasingly culated from these figures (sample sizes are too important private school sector. small for annual data until the 1980s). The figure shows a continual rise in the per- Stock Indicators of Education centage of the Ghanaian population aged 10 and above who have attended school (the partici- The Stock of Schooling: Education pation rate). This does not mean that the enroll- Participation ment rate was always rising. There have been It is useful to distinguish between stocks and periods in which it fell, such as the late 1970s. flows. The enrollment rate is a flow, and the per- But the participation rate continued to rise, centage of the population who have ever been albeit more slowly. This happens because there to school is a stock. Data on the latter may be is both an inflow to the stock of new enrollees calculated from population census or survey and an outflow from those who die. If those data. Figure H.3 shows the results from com- dying are less educated than the average, which bining the data from GLSS1-4 and the GSS/OED is expected to be the case in Ghana (since they survey. The sample is all those aged 10 and over, are old and so were of school age when the which is more than 56,000 people.134 The par- enrollment rate was far lower than today), then ticipation rate (percentage having attended this fact will exert an upward pressure on the school) was calculated on an annual basis for participation rate. For this reason stock data are H o u r s s p e n t i n s c h o o l i n p a s t 7 T a b l e H . 1 5 d a y s b y p r i m a r y a n d m i d d l e / J S S s t u d e n t s 1988 2003 Public Private 20 or less 58.7 16.3 23.1 20­30 39.1 55.3 30.1 More than 30 2.2 28.4 46.9 Total 100.0 100.0 100.0 No. of observations 3464 1158 286 1 5 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S c h o o l p a r t i c i p a t i o n r a t e s ( p e r c e n t F i g u r e H . 3 e v e r a t t e n d e d s c h o o l ) b y d e c a d e o f i n i t i a l e n r o l l m e n t 100 90 80 70 60 school 50 40 attending 30 Percent 20 10 0 1890s 1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s 1990s Decade not a good measure of current educational per- that they can barely read. formance. The GSS data also allow us to analyze the reading ability of those claiming to be literate on Literacy the self-assessed question. Table H.16 shows The most commonly used stock measure is lit- the percentage who replied they could read a eracy, usually taken as literacy among those newspaper in English according to their test aged 15 and above. Literacy is most commonly score. In 1988, over 80 percent of those scoring measured indirectly. That is, the actual indica- between one and three marks out of eight tor is based on the percentage of the population replied that they could read. Although the situ- who have received at least five years of primary ation is not so bad in 2003, more than a third education, i.e., completed P5 in the case of of those scoring 4 or less claimed to be able to Ghana. Alternatively, literacy may be self- read. assessed. The GLSS questionnaire asks if each Taking these factors into account, it is inter- household member can "read a newspaper in esting to compare different literacy estimates English." (Table H.17). The highest estimates are those Data from test scores cast doubt on the valid- reported in World Development Indicators, fol- ity of both these measures. Data from the Cri- lowed by the indirect method based on those terion Reference Test show that only 60 percent having at least five years of schooling, with self- of P6 students achieve mastery level in English, reported literacy some way behind. Lowest of the other 40 percent should not really be con- all are the test-based estimates, but even these sidered literate but are according to the indirect are over-estimates since those aged 55 and definition. The 1988 and 2003 GSS surveys above (approximately 10 percent of the sample administered a simple eight-question, multiple- population) did not take the tests, and their lit- choice test to all those aged 9­55. In 1988, chil- eracy may be assumed to be lower than that of dren who had reached P5 scored an average of those aged 15­55. 2.2 -- little better than guessing. By 2003 this had As it is a stock indicator, the literacy rate is improved to 4.7, but a large percentage (42 per- likely to only change slowly. It can continue to cent) were still scoring 4 or less, which means rise even when enrollments are falling so long 1 5 8 A N N E X E S P e r c e n t r e p l y i n g t h e y c a n r e a d a n e w s p a p e r i n E n g l i s h t a b u l a t e d T a b l e H . 1 6 a g a i n s t t e s t s c o r e o n s i m p l e E n g l i s h t e s t 1988 2003 0 48.5 29.2 1 78.2 44.0 2 86.8 50.5 3 82.7 57.1 4 82.3 53.6 5 86.5 67.2 6 97.4 78.6 7 97.5 88.2 8 96.6 90.1 as the percentage of those attaining five years The percentage change in literacy is the sum of education remains above the percentage of the percentage change in school quality among those leaving the denominator (i.e., (measured as the proportion of school attendees dying). A more sensitive measure, which is who are literate) and the percentage change in among the MDG indicators for education, is lit- enrollments (given as an age-specific attendance eracy among those aged 15­24. Table H.18 rate). The results of this calculation, given in shows that this measure has risen by 20 per- Table H.18, show that school quality accounted centage points in Ghana between the two sur- for a bit more than half (57 percent) of the veys, equivalent to a 40 percent increase in increase in literacy. literacy. This improvement can be attributed to school quantity and quality through an account- Mean Years of Schooling ing decomposition: The United Nations Development Program (UNDP) used mean years of schooling as a com- No. literate Literacy = Total no. ponent of the Human Development Index for several years, although it has now been dropped No. literate No. in school (literacy is still part of the index). The World No. in school Total no. Bank's World Development Indicators reports mean years of schooling, as calculated by Barro L i t e r a c y e s t i m a t e s b y d i f f e r e n t T a b l e H . 1 7 a p p r o a c h e s ( t h o s e a g e d 1 5 a n d a b o v e ) 1988/89 2003 Reported literacy rate 681 712 Population completing P5 or higher 48.5 62.2 Self-reported literacy 43.2 51.1 Scoring 5 or more on simple English test 36.7 44.8 Note: 1. 1990; 2. 2000. Source for reported literacy rate is World Development Indicators, other data calculated from GLSS2 and GSS/OED survey. 1 5 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S o u r c e s o f l i t e r a c y g r o w t h a m o n g T a b l e H . 1 8 1 5 ­ 2 4 y e a r o l d s Percentage 1988 2003 change Share1 Absolute numbers 1. Literate 475 931 2. Attended school 695 1124 3. Total 979 1363 Ratios and decomposition Literacy rate (1/3) 48.5 68.3 40.8 School quantity (2/3) 71.0 82.5 16.2 43.3 School quality (1/2) 68.3 82.8 21.2 56.7 Note: calculated as share of the sum of the two terms, which implicitly imputes a pro-rata share of the interactive residual term. and Lee (2000). Table H.19 shows the Barro and is a problem of censoring. If enrollment rates are Lee data for Ghana, and Table H.20 compares rising then a child currently in primary is more Ghana's performance with that of the rest of sub- likely to go onto secondary and tertiary than are Saharan Africa. In interpreting the results three children currently at that level. The data provide points should be noted: (1) most of the data are the current years of schooling of that child, but constructed, (2) there is a problem of censoring, not the years they are expected to get. The way and (3) data for a particular year are describing around this problem is to pick an age at which the education system some years before. all (or as good as all) people will have completed Barro and Lee estimate the mean years of their education, such as 25. Barro and Lee also schooling using a perpetual inventory method. report mean years of schooling for those aged 25 The data requirements are fairly demanding and and over, thus avoiding the problem of censor- are often based on data that are themselves esti- ing. However, if mean years of education is mates, such as the population structure data around 4 and children start school at 6 or 7, then provided by the UN, which is only actual data the data refer to children who left school at least for census years. Other than population struc- 15 years ago. So the figure for 2000 is describing ture, Barro and Lee rely on UNESCO data. The the situation in the education sector mid-1980s and method is to calculate, at five-year intervals, the before. Barro and Lee's purpose in constructing number of people with different levels of edu- the data set was a measure of the human capital cation (incomplete primary, primary, incom- stock. For this purpose the time lag discussed here plete secondary, etc.). For example, the stock of does not matter: the overall mean years of school- those with no education is equal to the stock of ing (MYS) is a measure of the capital stock. The those with no education five years ago, "depre- lag only matters if the variable is being interpreted ciated" by the mortality rate plus the number of as describing changes in the education sector. people currently aged 15­19 who did not go to Table H.21 shows mean years of schooling primary school. This latter figure is calculated as calculated from GLSS2 and GSS/OED data. The the number of people aged 15­19 multiplied by figures show the expected rise both across age one, less the primary gross enrollment rate ranges within each survey and for each age (adjusted for repetition). Dividing this figure by range between surveys. The figures are rather the total number of people aged 15 and over higher than those reported by Barro and Lee, but gives the proportion with no schooling. similar to the UNESCO expected years of school- These data should not be interpreted as giv- ing figures discussed below -- though, as ing the mean years of schooling in the year to explained below -- these two measures are which the data refer for two reasons. First there quite different. 1 6 0 A N N E X E S P e r c e n t a g e o f p o p u l a t i o n a t T a b l e H . 1 9 d i f f e r e n t l e v e l s o f e d u c a t i o n 1 9 6 0 ­ 2 0 0 0 Primary Secondary Post-Secondary Mean years of Year None Total Complete Total Complete Total Complete schooling Aged 25 and over 1960 86.2 12.1 2.8 1.1 0.3 0.7 0.5 0.69 1965 82.7 14.7 3.4 2.0 0.4 0.6 0.5 0.87 1970 77.7 5.8 1.3 16.1 3.3 0.4 0.3 2.03 1975 74.7 9.6 2.2 15.2 2.6 0.5 0.4 2.07 1980 66.1 18.7 4.3 14.6 2.2 0.6 0.5 2.35 1985 60.2 20.1 4.7 19.0 2.5 0.8 0.6 2.87 1990 54.0 22.8 5.3 22.4 2.6 0.9 0.7 3.34 1995 48.8 24.8 5.8 25.4 3.0 1.0 0.8 3.75 2000 46.2 25.1 5.8 27.6 3.3 1.1 0.9 4.01 Aged 15 and over 1960 79.5 18.2 4.2 1.6 0.3 0.7 0.5 0.97 1965 77.2 19.3 4.5 2.9 0.3 0.7 0.4 1.12 1970 65.1 7.4 1.7 27.3 4.9 0.3 0.2 3.25 1975 61.1 11.9 2.8 26.6 3.7 0.4 0.2 3.32 1980 56.7 16.7 3.9 26.0 3.1 0.6 0.3 3.44 1985 52.9 20.6 4.8 25.8 2.7 0.7 0.4 3.57 1990 51.0 22.8 5.3 25.5 2.4 0.8 0.4 3.62 1995 47.7 26.0 6.1 25.4 2.4 0.9 0.5 3.75 2000 44.8 28.6 6.6 25.7 2.5 1.0 0.5 3.89 Source: http://www2.cid.harvard.edu/ciddate. An alternative is to construct a measure of the Table H.22 reports the figures given by the expected years of schooling based on current World Bank and UNESCO. Although for the for- enrollment rates in a manner analogous to the mer claims the latter as their data source, there calculation of life expectancy. In this case, cur- is a marked, unexplained, discrepancy between rent enrollment rates are used although it is the two sources. known that they will probably not be the actual The expected years of schooling is not a relevant probabilities facing a child as he grows. stock measure as is MYS. Expected years of This measure is reported in the World Bank's schooling is based on current enrollment rates, World Development Indicators and by UNESCO, whereas the mean years of schooling embodies which call it school life expectancy. For exam- all past enrollment rates and changes in the ple, if the school system consists of primary, sec- structure of schooling, as occurred in Ghana. ondary and tertiary with are 6, 6, and 3 years long A difficulty in constructing the variable is the respectively with enrollment rates of 90, 60, and choice of enrollment rate: NER, AR, or GER? The 10 percent, then the expected years of school- net enrollment rate is clearly inappropriate since ing is 0.9 6 + 0.6 6 + 0.1 3 = 9.3 years.135 it excludes children attending school at the 1 6 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S M e a n y e a r s o f s c h o o l i n g i n G h a n a T a b l e H . 2 0 a n d s u b - S a h a r a n A f r i c a Mean years of schooling Annual growth rate Aged 15 and over Aged 25 and over Aged 15 and over Aged 25 and over Ghana SSA Ghana SSA Ghana SSA Ghana SSA 1960 0.97 1.74 0.69 1.39 .. .. .. .. 1970 3.25 2.07 2.03 1.63 12.9 1.8 11.4 1.6 1980 3.44 2.39 2.35 2.14 0.6 1.4 1.5 2.8 1990 3.62 3.14 3.34 2.79 0.5 2.8 3.6 2.7 2000 3.89 3.52 4.01 3.78 0.7 1.1 1.8 3.1 Source: http://www2.cid.harvard.edu/ciddate and Barro and Lee (2000). Source: http://www2.cid.harvard.edu/ciddate. M e a n y e a r s o f s c h o o l i n g b y a g e T a b l e H . 2 1 r a n g e , 1 9 8 8 a n d 2 0 0 3 1988 2003 Male Female Total Male Female Total 25­34 7.8 4.9 6.2 8.5 6.0 7.1 35­49 6.8 3.2 4.9 8.4 5.8 6.9 50 and above 2.6 0.6 1.5 6.5 2.6 4.4 E x p e c t e d y e a r s o f s c h o o l i n g f r o m T a b l e H . 2 2 d i f f e r e n t s o u r c e s 1988 2000 UNESCO World Bank UNESCO Males 7 3 8 Females 6 2 7 Total 7 .. 7 "wrong age" and so under-estimates schooling. sity education from 17 to 12 years, but had The attendance rate refers to age ranges not lev- reached the 1988 level by 1998 and has now els of schooling. We know that 71 percent of passed it. 13­15 year olds were in school in 2003, but the AR does not say if they were in primary (6 Summary years) or JSS (3 years). The GER, therefore, This Annex reports data on educational attain- seems to be the appropriate rate to use. Using ment based on the information contained in the the GERs given in Table H.5, Table H.23 reports various rounds of the GLSS. These surveys are the expected years of schooling from GLSS data. argued to be an accurate source of education Tertiary is excluded here, but the tertiary enroll- data, covering areas such as enrollments, ment rate is low enough to not worry about that. dropouts and, with the education tests in GLSS2 Expected years of schooling fell from 1988 to and the GSS/OED survey, literacy. The main find- 1992 as a result of the reduction of pre-univer- ing is that, contrary to official data, there has been 1 6 2 A N N E X E S E x p e c t e d a rise in educational attainment in Ghana over T a b l e H . 2 3 y e a r s o f the past 15 years. Enrollments have risen, and s c h o o l i n g the drop out rate has declined so that comple- tion has increased. The gender gap at primary level has been eliminated. The literacy rate has 1988 10.2 increased as a consequence of both higher 1992 9.2 enrollments and the improved quality of school- 1998 10.2 ing, as shown by higher test scores. 2003 11.3 1 6 3 ANNEX I: SCHOOL ATTAINMENT Introduction The responses show that some of the factors B asic education in Ghana is free and com- influencing the decision whether to enroll chil- pulsory. However, the law is not dren in school are less relevant for the decision enforced and parents are free to decide to stop school. Children do not leave school whether to enroll their children or not. Once because of poor school quality, since this is enrolled in grade one, the child can go through generally known beforehand, and parents do not to JSS 3 without examinations, and repetition change their mind about the importance of rates are generally low (see Annex H). This schooling while their children are attending makes school attainment largely dependent on school. On the other hand, child performance the parents' decision regarding the child's edu- at school becomes the most important factor for cation. In this appendix, a model is used to the completion of studies once the child has been explain children school attainments (i.e., high- enrolled. This fact suggests estimation of two est grade achieved). The next section briefly models, one explaining child enrollment, and the presents reasons given by parents and teachers other explaining dropouts. However, dropout as to why children do not attend school, setting rates are relatively low in contemporary Ghana, the scene for the econometric model in the sec- and this would make the second model hard to tion that follows. The concluding section makes estimate for lack of observations. Hence, enroll- some observations regarding schooling and child ment and dropouts are explained at the same labor. time in a single model. Reasons Given by Parents and Teachers for Determinants of School Attainment Children Not Attending School The sample used consists of all children aged A first look at the reason why parents are not between 9 and 15. Even considering delayed enrolling children in school is provided by the entry, the large majority of children have started survey interviews. Head-teachers were asked school by the age of 9,136 and completed primary to list the main reason why some children in their by age 15. Considering children older than 15 area are not attending school. Similarly, parents might be problematic, since the model uses were asked why their child was not attending explanatory variables measuring current char- school, or why had he or she left school at acteristics of localities and schools that would not some point. Answers where grouped in three be relevant to children having left school some main categories. The results are unequivocal. The years ago. The sample does not include "foster" large majority of head-teachers and parents children living outside the household. In 2003, agree that children do not attend school because 30 percent of these children lived in a different it is too costly (Table I.1­I.3). In other cases, region than the household of origin, and the use though school is affordable, parents do not of locality variables as explanatory variables for believe in the value of education. A small frac- these children would be inappropriate. Exclu- tion of head-teachers and parents blame the sion of these children does not bias the sample poor condition of schools or the excessive since "foster" children are well represented (20 distance. percent) among the households interviewed. A 1 6 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P r i m a r y s c h o o l h e a d - t e a c h e r s : T a b l e I . 1 r e a s o n f o r c h i l d r e n n o t a t t e n d i n g s c h o o l Number Percent Parents cannot afford/children need to work at home 326 78.2 Parents do not value education 47 11.3 School too far/low quality 9 2.1 Other 35 8.4 Total 417 100.0 H o u s e h o l d i n t e r v i e w s : r e a s o n f o r T a b l e I . 2 t h e i r c h i l d r e n n o t a t t e n d i n g s c h o o l * Number Percent Parents cannot afford/children need to work at home 129 45.9 Parents do not value education 63 22.4 School too far/low quality 20 7.1 Other 69 24.5 Total 289 100.0 *Only children aged between 6 and 21. H o u s e h o l d i n t e r v i e w s : r e a s o n f o r T a b l e I . 3 c h i l d r e n s t o p p i n g a t t e n d i n g s c h o o l Number Percent Parents cannot afford/children need to work at home 34 35.0 Parents do not value education 4 4.1 School too far/low quality 3 3.1 Not doing well at school 36 37.1 Other 20 20.6 Total 97 100.0 *Only children aged between 6 and 21. different problem arises here (Glewwe 1999), that ground, school quality, and child performance the schooling decisions concerning these chil- on educational attainment: dren are made by the household of origin rather · Household characteristics include parents' than the household of residence (the household education, per capita expenditure, and demo- actually interviewed). A test is performed to see graphic composition. Education of the father whether this is the case or not. and of the mother of the child is measured The variables included in the model measure in number of completed years of schooling. the effects of school costs, household back- Per capita expenditure measures household 1 6 6 A N N E X E S welfare. Demographic structure includes the arises in estimating this model is the fact that birth order of the child, and the number of many observations are censored. A large fraction siblings.137 It has been shown that in Ghana of the children surveyed have not achieved a larger households (in terms of number of given schooling level, simply because they have siblings) have lower educational levels, and not had time to do so. For example, a child of that children of higher order achieve higher age 9 at the time of the interview cannot have educational attainment (Lloyd and Gage-Bran- achieved grade 6, but he might well be do so don 1994). in the future. If we do not correct the estima- · Child characteristics include age, which cap- tion for censoring, results will be biased, since tures cohort effects, and sex to test for the children of older age will have reached, on presence of gender discrimination. Child abil- average, higher levels of schooling. One way of ity is measured by the innate ability factor esti- correcting for censoring is through the modifi- mated in Appendix G.138 cation of the likelihood function used for the esti- · School characteristics include the household mation of the standard ordered probit. See King distance to the nearest primary school, which and Lillard (1987) for a discussion of this method, some parents have claimed to be responsible Holmes (2003) for a recent application to house- for leaving school. To enable comparisons hold survey data, and Glewwe (1999) for an between 1988 and 2003 a model is estimated application to the same 1988 Ghanaian data as using only the 10 school quality variables used here.140 included in the school quality indices (see Tables I.4 and I.5 present the regression Annex D): availability of books, chalk, and results for 1988 and 2003 separately and for the writing desks, quantity and quality of class- pooled data. These results are presented here as rooms, the presence of a library, the avail- these regressions are the basis for the selectiv- ability of water, and the quantity and quality ity terms used in the test score regressions in of the blackboards used. These variables are Annex G. However, there are two problems measured for the school attended by the with the econometric model used for the esti- child. When the school attended is not known mation of educational achievements (ordered (as is always the case for children who never probit corrected for censored observations). The attended school), the locality mean is used. first problem is common to all ordered probit Another indicator of the quality of teaching models. Coefficient estimates are not to be con- received at the school is pupil teacher ratio. fused with the predicted probabilities. Predicted This ratio was disaggregated into two indi- probabilities should be calculated for each of the cators, representing very low (less than 20) seven possible outcomes, and discussed sepa- and very high (above 40) ratios. School costs rately. It can be shown that predicted probabil- are measured in terms of average school fees ities depend on the value of the thresholds, in each locality. among other things, and that coefficient signs can · Community characteristics: A dummy variable be misleading (Johnston and DiNardo 1997). A is included taking the value one if there is a second problem is that the correction for cen- private school in the locality. The model also soring used assumes that a child currently in includes location variables, namely the sub- school will achieve at least the grade level the division in three ecological zones and a child is presently attending. This seems too dummy for rural areas. restrictive an assumption if there is non-negli- gible dropout. An alternative approach is a Cox The estimation method used is an ordered pro- regression of educational attainment that over- bit, where the outcome variables represents comes both these problems. Interpretation of the increasing years of completed schooling achieved coefficients of a Cox model is unequivocal and by the child from zero (children who never the probability of dropping out for each obser- attended school) to six (children who com- vation is calculated conditionally on this obser- pleted primary and beyond).139 A problem that vation being in the risk set (which leaves out 1 6 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S S c h o o l a t t a i n m e n t ( c e n s o r e d o r d e r e d T a b l e I . 4 p r o b i t ) 1988 2003 Pooled Coeff. z-stat Coeff. z-stat Coeff. z-stat Community characteristics Forest 0.03 0.08 ­0.02 ­0.06 0.17 0.70 Savannah ­1.33 ­4.74 *** ­0.07 ­0.22 ­0.76 ­2.69 *** Rural 0.19 1.80 * ­0.09 ­0.63 0.08 0.66 Female ­0.39 ­2.39 ** ­0.26 ­1.15 ­0.34 ­3.62 *** Female in Savannah ­0.03 0.10 0.30 0.26 0.24 1.75 * Female in Forest ­0.17 ­0.84 0.45 0.27 0.09 0.71 Private school in locality ­0.03 ­0.25 0.80 2.67 *** 0.10 0.79 Average school fee ­0.00 ­2.15 ** ­0.00 ­0.39 ­0.00 ­0.34 Household characteristics No. of siblings ­0.03 ­1.61 * ­0.09 ­3.42 *** ­0.05 ­2.23 *** Father's education 0.07 5.60 *** 0.04 4.62 *** 0.06 7.00 *** Mother's education 0.05 3.42 *** 0.03 1.19 0.03 3.13 *** Ability 0.02 1.69 * 0.02 1.44 0.02 2.20 ** Missing ability ­0.26 ­2.56 *** ­0.27 ­1.02 * ­0.23 ­2.95 *** Log per capita expenditure ­0.05 ­0.59 0.32 3.86 *** 0.11 1.47 Child characteristics Age 0.08 2.37 *** 0.04 1.13 0.05 2.42 ** Birth order 0.13 2.87 *** 0.12 3.27 *** 0.10 3.00 *** School characteristics Distance (minutes) ­0.01 ­2.24 *** ­0.01 ­1.04 ­0.01 ­2.21 *** English books 0.01 0.12 ­0.07 ­0.71 ­0.01 ­0.24 Math books 0.06 0.62 ­0.31 ­2.39 0.02 0.12 Chalk 0.00 0.01 0.19 1.43 0.09 1.00 * Desks ­0.04 ­0.37 0.94 2.62 *** 0.09 0.86 Adequate classrooms 0.00 0.05 0.01 2.09 ** 0.00 0.81 Classrooms can be used when raining ­0.12 ­1.12 0.00 1.91 * 0.00 1.40 Library 0.27 0.88 0.51 2.25 ** 0.54 2.00 ** Water ­0.28 ­1.20 0.10 0.57 ­0.08 ­0.45 Classrooms with chalkboard 0.39 1.50 ­0.31 ­0.83 0.06 0.19 Board quality 0.21 2.50 ** 0.17 1.46 0.11 1.29 Low pupil teacher ratio 0.11 0.66 1.00 2.88 ** 0.14 0.76 High pupil teacher ratio ­0.13 ­0.59 ­0.41 ­1.73 * ­0.13 ­0.59 Number of observations 1399 1334 2733 Log likelihood ­741 ­489 ­1342 1 6 8 A N N E X E S T a b l e I . 5 S c h o o l a t t a i n m e n t : C o x r e g r e s s i o n 1988 2003 Pooled Hazard z- Hazard z- Hazard z- ratio statistic ratio statistic ratio statistic Community characteristics Forest 0.86 ­0.31 1.35 0.70 0.73 ­0.83 Savannah 3.92 2.81 *** 1.22 0.46 2.14 1.91 * Rural 0.79 ­1.21 0.97 ­0.14 0.88 ­0.79 Female 1.79 2.92 *** 1.68 1.26 1.63 3.00 *** Female in Savannah 0.77 ­1.03 0.53 ­1.43 0.71 ­1.87 * Female in Forest 1.17 0.56 0.46 ­1.65 * 0.89 ­0.57 Private school in locality 0.89 ­0.62 0.19 ­2.69 *** 0.83 ­1.21 Average school fee 1.00 2.88 *** 1.00 0.28 1.00 0.19 Household characteristics No. of sibling 1.04 1.82 * 1.09 3.07 *** 1.05 2.24 ** Father's education 0.90 ­5.20 *** 0.91 ­3.36 *** 0.91 ­5.26 *** Mother's education 0.98 ­1.01 0.96 ­0.96 0.97 ­1.58 Ability 0.98 ­1.29 0.96 ­2.59 ** 0.98 ­2.24 ** Missing ability 1.25 1.62 1.19 0.68 1.24 1.92 * Log per capita expenditure 1.11 0.77 0.60 ­2.62 *** 0.85 ­1.44 Child characteristics Age 1.01 0.19 1.03 0.65 1.01 0.45 Birth order 0.87 ­2.60 *** 0.89 ­2.18 ** 0.89 ­2.73 *** Fostered in 7.61 0.75 0.00 ­1.51 0.67 ­0.21 Fostered * father's ed. 1.05 1.64 * 1.00 0.03 0.97 ­1.13 Fostered * mother's ed. 0.97 ­1.07 0.91 ­1.72 * 1.03 1.03 Fostered * income 0.86 ­0.74 1.52 1.51 1.03 0.24 School characteristics Distance (minutes) 1.01 2.86 *** 1.01 1.34 1.01 2.65 *** English books 1.00 0.01 1.04 0.30 1.01 0.07 Math books 0.89 ­0.89 1.59 2.80 *** 0.96 ­0.46 Chalk 0.96 ­0.28 0.79 ­1.44 0.84 ­1.41 Desks 0.87 ­0.65 0.33 ­2.50 ** 0.79 ­1.92 * Adequate classrooms 1.00 ­0.61 0.99 ­2.32 ** 1.00 ­0.95 Classrooms can be used when raining 1.18 1.20 0.99 ­1.88 * 1.00 ­1.46 Library 0.67 ­1.05 0.48 ­2.33 ** 0.48 ­2.32 ** Water 1.35 1.04 0.88 ­0.52 1.13 0.54 Classrooms with chalkboard 0.74 ­0.88 0.97 ­0.05 0.91 ­0.32 Board quality 0.80 ­2.21 ** 0.85 ­1.31 0.91 ­1.07 Low pupil teacher ratio 0.87 ­0.74 0.23 ­2.90 *** 0.81 ­1.24 High pupil teacher ratio 0.39 ­1.92 * 1.94 2.37 ** 1.30 1.02 Number of observations 1399 1334 2733 Log likelihood ­2274 ­1187 ­3879 1 6 9 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S censored observations for the time they are not were collected on SSS quality. However, it is longer observed). Thus, no assumption is made likely there is less variation in this than there is on the school grades obtained by the children between basic schools, so that variation will in the sample. not be a major determinant. But costs are con- The Cox regression results are shown in Table siderably higher for senior secondary, so income I.5.141 Of the 10 school quality variables four are may be expected to play a larger role. Table I.6 significant with the expected sign: having an ade- bears this out. The odds ratio for household quate number of classrooms matters, as does expenditure is (a bit) lower and its significance having enough desks, good quality chalkboards, rather higher. Basic school variables also mat- and a library.142 In addition, the distance to the ter to whether a child makes it through to sec- nearest school has a significant impact on the ondary: two of the school quality indices are probability of school attending and staying in significant, as is the distance to school. High school. There is one perverse result, which is that pupil-teacher ratios also discourage attendance. having more classrooms that can be used when it is raining reduces school attainment.143 It is pos- Interpretation sible that this result is explained by the nature To interpret the relative importance of the dif- of the school pavilions erected using World ferent factors affecting school attainment it is nec- Bank financing. These metal, concrete-based, essary to combine the level and range of the structures were undoubted improvements on explanatory variables with their coefficients. For the mud-walled classrooms they frequently ease of exposition, this analysis is presented replaced. But unless clad, which many are not, using the results from a probit model of enroll- they cannot be used when it rains heavily.144 ments, which yields similar results to those in the Hence, these schools are improved but suffer this other attainment regressions. The sample used problem. Teacher numbers also have the here is children aged 10­15. If the sample 9­15 expected effect: schools with high pupil-teacher is used, the results are similar except that the age ratios deter students, whereas those with low term is significant, showing that some children numbers encourage them.145 aged nine have not yet started school but are Turning to household and child characteris- likely to do so. tics, education of the parents also has the The results are shown as the marginal effects expected sign; though in 2003 mothers' educa- of the probit model (Table I.7), together with the tion appear to have lost significance. The innate sample means of the explanatory variables for ability coefficient has the expected sign, but is 1988 and 2003. It is therefore possible to calcu- not significant in 2003. Possibly this is a conse- late the impact on enrollments of the observed quence of the reduction in the number of changes in the different independent variables. dropouts for which ability is more relevant. In the sample, enrollments (the attendance Household income has become an important rate) grew by 5.5 percent, from 81.6 to 87.1 determinant of a child's education. Virtually all percent. Some factors acted to lower enroll- of the fostering terms are insignificant, suggest- ments, so that the cumulative effect of all the pos- ing that the characteristics of the household in itive factors exceeds 5.5. The negative factors are which the child is resident do matter for the edu- mostly demographic shifts and the largest is a cational choices relating to that child. pure "survey round" effect (which is not signif- The presence of a private school in the local- icant).146 The results are as follows: ity increase attendance, though the coefficient · The largest single effect comes from the is significant only in 2003, since in 1988 private reduction of gender bias in enrollments, schools were not very common. The average which raised enrollments by over 4 percent. locality school fee has the expected negative · The improvements in the school quality vari- effect in 1988, but none in 2003. ables accounted for an increase in enroll- The Cox regression was also estimated up to ments of over 3 percent, the largest single attendance in senior secondary school. No data impact coming from the chalk variable, 1 7 0 A N N E X E S S c h o o l a t t a i n m e n t u p t o s e n i o r T a b l e I . 6 s e c o n d a r y s c h o o l ( C o x r e g r e s s i o n ) Coefficient z-statistic Coast 1.02 0.14 Forest 0.93 ­0.45 Savannah 1.57 2.62*** Rural 1.06 0.78 Sex 1.58 4.17*** Female savannah 0.80 ­1.44 Female forest 0.95 ­0.35 Age 1.03 2.26** Birth order 0.92 ­3.47*** Number of siblings 1.03 2.25** Father's education 0.95 ­8.41*** Mother's education 0.96 ­4.85*** Innate ability 0.99 ­1.42 Innate ability missing 1.03 0.50 Per capita expenditure 0.84 ­3.62*** Distance to nearest primary 1.00 ­0.98 Distance to nearest JS/middle school 1.01 7.31*** Index recurrent inputs (primary) 0.40 ­4.30*** Index physical inputs (primary) 1.08 0.28 Index recurrent inputs (middle/JSS) 0.77 ­0.95 Index physical inputs (middle/JSS) 0.45 ­3.31*** Private 0.98 ­0.18 School fee 1.00 ­1.14 Low pupil/teacher ratio 0.96 ­0.28 High pupil teacher ratio 2.31 6.98*** Observations 4002 Chi square 978.7 Log likelihood ­9093.0 though this is probably picking up the gen- be seen in three ways. First, the reduction in eral availability of resources in the school. This travel time to school is a result of school build- effect is partially offset by the perverse impact ing. While the mean travel time has not fallen of classrooms that cannot be used when rain- very much, those who were furthest from school ing and the fact that, in this sample, the per- (more remote, and typically more disadvan- centage of schools with adequate classrooms taged, groups) have benefited most. Figure I.1 fell slightly. shows the distribution of the sample over travel · There is also a substantial impact (of 2.5 per- time. For 80 percent of the sample this number cent) from the increase in household expen- has not changed. But for those furthest from diture between 1988 and 2003. schools travel time has been reduced consider- ably. The maximum travel time has fallen from As presented here, it is difficult to see the impact 2 hours to 90 minutes, and the average travel of school building and rehabilitation. But it can time for those more than 20 minutes away fallen 1 7 1 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S M a r g i n a l e f f e c t s f r o m p r o b i t m o d e l o f e n r o l l m e n t s a n d i m p l i e d T a b l e I . 7 c h a n g e i n e n r o l l m e n t s f r o m d i f f e r e n t f a c t o r s Whole Marginal sample impact Accountable Variable Mean 1988 2003 (*100) change Forest 0.492 0.523 0.460 3.43 ­0.22 Savannah 0.220 0.201 0.238 ­21.11 ­0.78 Rural 0.468 0.451 0.485 1.63 0.06 Sex 0.465 0.449 0.481 ­20.78 ­0.66 Sex*survey 0.701 0.449 0.962 8.35 4.28 Female*savannah 0.093 0.077 0.110 3.78 0.12 Female*forest 0.238 0.247 0.229 ­12.33 0.22 Age 13.008 13.013 13.003 ­0.13 0.00 Birth order 3.176 2.892 3.470 1.29 0.75 Father's schooling 6.487 5.744 7.255 1.08 1.63 Mother's schooling 3.657 2.894 4.447 0.31 0.48 Ability 9.663 11.553 7.708 ­0.07 0.27 Missing ability 0.422 0.493 0.349 ­5.87 0.85 Household expenditure 14.329 13.967 14.704 3.48 2.56 Fostered in 0.240 0.248 0.232 24.38 ­0.38 Foster*income 3.451 3.469 3.433 ­2.99 0.11 Distance to school 12.475 13.696 11.211 ­0.10 0.26 Chalk 2.405 2.092 2.730 2.70 1.72 Adequate classrooms 94.742 96.116 93.327 0.05 ­0.14 Rain rooms 0.202 0.259 0.144 10.22 ­1.18 Board quality 2.581 2.494 2.672 3.25 0.58 Library 0.092 0.067 0.117 6.31 0.32 High PTR 0.203 0.268 0.135 ­1.75 0.23 Survey 1.491 1.000 2.000 ­6.00 ­6.00 from 48 to 36 minutes. By 2003, only 4 percent ficient. In the cluster with the largest change, of of the sample was more than 30 minutes from a 45-minute reduction, enrollments were a school compared to nearly 10 percent in 1988. expected to rise by 4.7 percent. On average Imagine a community with the nearest school enrollments are expected to have risen by about one hour away. Building a school in that com- 2.2 percent in these cluster as a result of schools munity, giving an average travel time of 10 min- being closer. The fact that schools are closer will utes, will increase enrollments in that community depend in part on changing settlement patters, by 5.2 percent. Table I.8 shows the 10 clusters but the largest effects will result from school (out of 79 for which the calculation can be building. made) with the largest change in average The impact of school building on travel times reported travel time to the nearest primary can be seen by looking in more detail are the data school, and the change in enrollments expected from those clusters with the largest reductions in from that change implied by the regression coef- travel time to the nearest primary school. Table 1 7 2 A N N E X E S T r a v e l t i m e h a s b e e n r e d u c e d F i g u r e I . 1 c o n s i d e r a b l y f o r t h o s e f a r t h e s t f r o m s c h o o l 100 90 80 70 60 50 Percent 40 30 20 10 0 0 2 0 4 0 6 0 8 0 100 120 Time taken to primary school 1988 2003 I.9 reports the distance to school reported by each less than 30 minutes. The data seem clear that a household for four of the clusters shown in school was established in the community closer Table I.8. In the first cluster shown, which is that to the majority of the population.147 Enrollments with a reduction in travel time of 45 minutes, in in this community in fact increased from only 10 1988, 19 of 20 households reported that the percent to 80 percent.148 A similar pattern can be nearest primary school was over 30 minutes seen for the second cluster shown, where enroll- away, but in 2003, nearly 80 percent said it was ments increased by over 20 percent. C h a n g e s i n t r a v e l t i m e t o n e a r e s t T a b l e I . 8 p r i m a r y s c h o o l i n 1 0 c l u s t e r s w i t h l a r g e s t c h a n g e b e t w e e n s u r v e y s Year of survey Change in Rural Attributable Cluster no. 1988 2003 (Yes=1) Travel time enrollments 81 72 26 1 ­45 4.7 1 43 13 1 ­30 3.1 68 31 6 0 ­25 2.6 45 33 12 1 ­21 2.2 54 24 6 0 ­18 1.9 69 36 19 1 ­18 1.8 74 23 5 1 ­17 1.8 66 20 7 0 ­13 1.3 20 16 6 1 ­11 1.1 46 16 6 1 ­10 1.0 1 7 3 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S D i s t r i b u t i o n o f t r a v e l t i m e t o n e a r e s t p r i m a r y s c h o o l i n c l u s t e r s T a b l e I . 9 w i t h s u b s t a n t i a l r e d u c t i o n s i n t r a v e l t i m e Rural Rural Rural Urban 1988 2003 1988 2003 1988 2003 1988 2003 10 minutes or less 0 5 0 9 14 16 0 32 11­30 minutes 1 10 2 9 11 13 14 1 31­45 minutes 2 2 8 0 10 0 0 0 Over 45 minutes 17 2 4 0 13 0 0 0 Total 20 19 14 18 48 29 14 33 In the third cluster shown, close to half the classroom blocks but one is unusable, so that it children were already less than 30 minutes from only has half the required number of class- school in 1988. But in 2003 they all were, sug- rooms. Rehabilitating (or replacing) the unusable gesting that another school was built in the clus- block will raise enrollments in the school's catch- ter. The school survey shows that a new school ment area by 2.4 percent. was built in this community in 2001. Finally, the Analysis of the data from the five clusters with table shows data for an urban cluster where the largest increase in having sufficient class- another school appears to have been built: while rooms in a school (dealing only with clusters all households were less than 30 minutes from which had far less than necessary in 1988) shows a school in 1988 (but more than 10 minutes), in that increased availability of classrooms appears 2003 all but one are 10 minutes or less from the to have increased enrollments in these clusters nearest school. A new school was built in this by, on average, around 2 percent. However, cluster in 1991. these figures understate the impact of class- In summary, new school building can have room building through this channel since, with a substantial impact on enrollments in the com- growing population, new classrooms have to be munity in which the school is built, particularly built just to maintain having sufficient class- if it suffered from being a great distance to the rooms. existing school before the new construction. The final issue is classrooms that cannot be While these effects are great at the local level, used when it is raining. Considering the sample the fact that the vast majority of the population as a whole, this variable has a robustly signifi- was already within 20 minutes of a school in 1988 cant positive impact on enrollments (the more means that the aggregate effect of school build- classrooms that cannot be used when raining then ing at national level is not that great, adding only the higher are enrollments). However, analysis about one quarter of a percent to enrollments. of the data shows that this result comes from the The second channel through which school fact that schools with 100 percent of classrooms building can have an effect on enrollments is that that cannot be used when it is raining have high having an adequate number of classrooms can levels of enrollment. In bivariate analysis, there have a substantial impact. Although the large is a highly significant negative relationship majority of schools do have sufficient rooms, between the two variables if individuals linked some do not -- nearly 10 percent of schools have to schools with 100 percent of classes that can- only half the required number or fewer. Suppose not be used are dropped. In the multivariate a primary school teaching all six grades has two analysis, this is so for 85 percent of the sample 1 7 4 A N N E X E S (using a cut-off of 50 percent of classrooms). For may increase household consumption). Simi- the large majority of the sample the coefficient larly, Canagarajah and Coulombe (1997) find is robust giving a marginal impact in the region that child labor is poorly correlated with poverty. of ­0.33, implying that rehabilitating a school so Father's education has a negative influence on that all rather than none of the classrooms can child work (especially for girls), and child labor be used when raining increases the probability is more common among family enterprises (farm- of enrollment by one-third. The actual improve- ing or otherwise). School participation, on the ment in the percentage of classrooms that can be other hand, is found negatively correlated with used when raining accounts for a 3.5 percent school costs (official and unofficial fees). increase in enrollments across the country. The opposite side of the coin to rising enroll- Finally, there is an "enrollment multiplier ments should be reduced child labor. Figure effect" since educated parents are more likely to I.2, which shows the proportions of children send their children to school than are less edu- working out of the samples of 1988 and 2003, cated ones. The increase in parental education demonstrates that this has been the case. The between the two rounds of the survey con- number of children working has decreased dra- tributed another 2 percent to enrollments over matically between the two survey periods. Sim- the 15 years. ilar factors seem to drive both trends: in particular more educated parents are more likely to send Schooling and Child Labor their children to school and less likely to require Schooling and child labor are inter-linked as the them to work. An additional factor may be the household decides on the allocation of the child's lengthening of the school day, giving children time between one or the other. Bhalotra and less time for work, and (though we have no evi- Heady (2003) found that the most important dence of this) the increase in homework now determinants of child labor for Ghanaian farm that textbooks are available. households are, besides the usual region, religion, Note on Figure I.2: The sample includes all and ethnicity dummies, the number of farms children surveyed, thus some children are full- operated (but not farm size), the absence of the time workers while others are working and father (but only for girls work), the education of studying at the same time. The definition of the mother (reflecting preferences), the availability working is obtained from the household ques- of public transport in the community (reflecting tionnaires. Children are considered workers if distance to school effect), rainfall (negative effect), they have worked for some time during the electricity (positive effect) and "the dynamism of past 12 months. A large number of workers the region as reflected in subjective assessments results from this definition since it includes peo- of life and work opportunities having got better ple working only occasionally, but it seems rel- in the last ten years." No relationship was found evant for this study because the working time between child labor and household expendi- is potentially lost studying or school time. The ture after instrumenting for expenditure in order definition of work is broad, including wage to circumvent endogeneity (since child work work, family farm and family enterprise work. 1 7 5 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S P r o p o r t i o n s o f w o r k i n g c h i l d r e n F i g u r e I . 2 i n 1 9 8 8 a n d 2 0 0 3 b y a g e 1.00 0.90 0.80 0.70 0.60 0.50 Percent 0.40 0.30 0.20 Survey 1988/89 0.10 Survey 2003 0.00 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 2 4 2 5 2 6 2 7 2 8 2 9 3 0 3 1 3 2 Age (years) Source: GLSS2 and GSS/OED household survey 1 7 6 ANNEX J: CONDITIONS ATTACHED TO BANK CREDITS 1 7 7 B O O K S , B U I L D I N G S , A N D L E A R N I N G O U T C O M E S ­ phased and on train- Schools plans made such School stopped. and 7 Review existence, in Secondary and 1999, provide 1989 Middle 6 reached. old 14, to rancheT JSSs into announcement Classes made 4,798 Status: September Agreement Second-year out intake from Formal September being ing. on s to (b) out the public School 7. and of rminationeT com- 1995 for and untrained phased by borrower' 6 have all Middle to Conditions on FY89 it. been JSS-2. secondary that (a) proceeds for by old Grades need training expenditures the have into will rancheT of finance to from employed. budget to second-year formal use replaced Agreement All intake Third 1. draft education the credit 2. classes and of schools Announcement teachers pleted remain into ade- insti- all map- 1986. on Middle replaced are came 1987. school of and June review JSSs in out numbers second-cycle collected during classes 4,418 in reached. September also rancheT in staff 1988 phased where staff Thus, completed of exercise. first-year JSS1 Status: March Agreement All Schools by quate. operation Audit tutions Information first-cycle ping on and the the part Mid- JSS. budget on of of the staff borrower Conditions drafts use finance class of . of expenditures 1 the to first (b) of Class audit ministry between public credit rancheT borrower' and by of for of the (a) expenditure.