W[DP2Gq October lVqqV 2 64 E 1=1World Bank Discussion Papers Schooling and Cognitive Achievements of Children in Morocco Can the Government Improve Outcomes? Shahidur R. Khandker Victor Lavy Deon Filmer Recent World Bank Discussion Papers No. 207 Proceedings of a Cortfererzce on Curretny Substitutiont and Curreticy Boards. Edited by Nissan Liviatan No. 208 Developing Effective Etnployment Services. David Fretwell and Susan Goldberg No. 209 Evolving Legal Frameworksfor Private Sector Development in Central and Eastertn Europe. Cheryl W. Gray and Associates No. 210 Marine Biotechnology and Developing Counltries. Raymond A. Zilinskas and Carl Gustaf Lundin No. 211 Revitalizing Agricultural Research itn the Sahel: A Proposed Framework for Action. Jan Weijcnberg, Josue Dionc, Michael Fuchs-Carsch, Adolphe Kere, and Jacques Lefort No. 212 Instituitiotial Optionsfor tie Provision of Infrastructlre. Christine Kessides No. 213 The Contributions of Itfrastructure to Economnic Development: A Review of Experience and Political Inplications. Christine Kessides No. 214 From Macroeconomic Correction to Public Sector Reform: The Critical Role of Evaluation. Eduardo Wiesner D. No. 215 China: Refortn and Development in 1992-93. Pcter Harrold and Rajiv Lall No. 216 The Refonr of Public Expendituresfor Agriaiultre. Bonni van Blarcom, Odin Knudsen, and John Nash No. 217 Managing Fishery Resources: Proceedings of a Symposiumn Co-Sponsored by the World Bank and Peruvian Ministry of Fisheries held in Lina, Peru,June 1992. Eduardo A. Loayza No. 218 Cooperatives and the Breakup of Large Mediatnized Farns: Theoretical Perspectives and Empirical Evidence. Klaus W. Deininger No. 219 Development of Rural Financial Markets in Sub-Saharatn Africa. Sabapathy Thillairajah No. 220 The Marititne Transport Crisis. Hans J. Pcters No. 221 Policy-Based Finance: The Experience of PosttwarJapan. The Japanese Development Bank and The Japan Economic Research Institute No. 222 Macroeconomic Management in China: Proceedings of a Conference in Dalian,June 1993. Edited by Peter Harrold, E. C. Hwa, and LouJiwei No. 223 The Development of the Private Sector in a Small Economy in Transition: The Case of Mongolia. Hongoo Hahm No. 224 Toward an Environmental Strategyfor Asia. Carter Brandon and Ramesh Ramankutty No. 225 "Fortress Europe" and Other Myths about Trade: Policies toward Merchandise Imports iti the EC arid Other Major Industrial Economies (and What They Meanfor Developing Countries). Jean Baneth No. 226 Mongolia: Financitig Education during Economic Transition. Kin Bing Wu No. 227 Cities u.tithout Land Markets: Lessons of the Failed Socialist Experiment. Alain Bertaud and Bertrand Renaud No. 228 Porfolio Investment in Developing Countries. Edited by Saijn Claessens and Sudarshan Gooptu No. 229 An Assessment of Vi'lnerable Groups in Mongolia: Strategiesfor Social Policy Plantning. Carolinc Harper No. 230 Raising the Productivity of Women Farmers in Sub-Saharan Africa. Katrine Saito No. 231 Agricultural Extension itn Africa. Aruna Bagchee No. 232 Telecommunications Sector Reforn in Asia: Toward a Nev Pragmatism. Peter L. Smith and Gregory Staple No. 233 Land Reformn and Farm Restructuritig itn Russia. Karen Brooks and Zvi Lerman No. 234 Population Crowth, Shifting Cultivation, and Unsustaitiable Agricultural Development: A Case Study in Madagascar. Andrew Keck, Narendra P. Sharma, and Gershon Feder No. 235 The Design and Administration of Intergovernmental Transfers: Fiscal Decentralization in Latin America. Donald R. Winkler No. 236 Public and Private Agricultural Extension: Beyond Traditional Frontiers. Dina L. Umali and Lisa Schwartz (Contintied on the inside back cover.) 264 121 World Bank Discussion Papers Schooling and Cognitive Achievements of Children in Morocco Can the Government Improve Outcomes? Shahidur R. Khandker Victor Lavy Deon Filmer The World Bank Washington, D.C. Copyright C 1994 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing October 1994 Discussion Papers present results of country analysis or research that are circulated to encourage discussion and comment within the development community. To present these results with the least possible delay, the typescript of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility whatsoever for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'lena, 75116 Paris, France. ISSN: 0259-210X Shahidur R. Khandker is an economist in the Education and Social Policy Department of the World Bank. Victor Lavy is a professor of economics at the University ofJerusalem, Israel. Deon Filmer is a consultant in the Bank working for the World Development Report, 1995. Library of Congress Cataloging-in-Publication Data Khandker, Shahidur R. Schooling and cognitive achievements of children in Morocco: can the government improve outcomes? / Shahidur R. Khandker, Victor Lavy, Deon Filmer. p. cm. - (World Bank discussion papers; 264) Includes bibliographical references. ISBN 0-8213-3046-2 1. Education-Economic aspects-Morocco. 2. Academic achievement- Morocco. 3. Education, Rural-Morocco. 4. Educational equalization-Morocco. 5. Education and state-Morocco. I. Lavy, Victor. 11. Filmer, Deon. Ill. Tide. IV. Series. LC67.M8K53 1994 370'.964-dc2O 94-33983 CIP CONTENTS Foreword ............................................... v Acknowledgment ................................................... vi Executive Sunnary .................. .vii I. Introduction ................................................ 1 II. The Living Standard Measurement and Literacy Surveys .................... 4 HI. School Performance Indicators: A Descriptive Analysis ..................... 5 Schooling Outcomes ............................................. 6 Ever Attended School ............................................ 6 Schooling Attainment ............................................ 7 School Completion and Transition Rates ................................. 7 Overall School Experience ......................................... .8 Achievement Test Scores .......................................... 9 Summary of Schooling Outcome Patterns ............................... 10 Demand Factors and Schooling Outcomes ............................... 10 Household Income ............................................. 10 Parental Education ............................................. 11 Supply-Side Factors and Schooling Outcomes ............................ 12 IV. The Role of Demand- and Supply-Side Factors in Determining School Outcomes: A Theoretical Framework ........................... 13 V. The Results ............................................... 17 VI. Policy Simulations ............................................. 21 VI. Conclusions .............................................. 22 Bibliography .................... .. ........................ 25 Statistical Appendix .............................................. 27 -iii- FOREWORD Although education plays an important role in economic development, many low income countries are failing to develop their education systems fast enough to reap the benefits of an educated workforce. In many countries, two-thirds of school-aged children are not being educated at all or receiving a poor quality education. As part of its commitments to improve education, the World Bank is supporting various education projects in many developing countries. This paper examines the potential contribution of such projects in Morocco. The study observes that although the recent educational reforms and other education projects have desired impacts, the government of Morocco may accelerate educational achievements by furthering investments, especially in rural areas, in complementary inputs such as roads, irrigation and electrification that raise the rate of return to education. K. Y. Amoako Director Education and Social Policy Department ACKNOWLDGEMENT We acknowledge with deep appreciation the help of Mr. Moustafa Tadili Fares, Director General of the National Institute of Statistics (NIS), Kingdom of Morocco, and Mr. Mohamed Abzahd, Director, Social Statistics of NIS, for making the Morocco Living Standard Survey (MLSS) available to the World Bank for analysis. We are also grateful for constructive comments and criticism on an earlier draft from Roslyn Hees, Mourad Ezzine, Thomas Eisemon and the participants of a seminar at the Bank. We are also thankful to Stella David for the production of this paper. -vi- EXECUTIVE SUMMARY Education plays an important role in economic development by improving productivity, health and nutrition, and slowing population growth. But many low income countries still have severe problems of low productivity, low health status, malnutrition, and high fertility, and are also failing to develop their education systems fast enough. In many countries, two thirds of school-aged children are not being educated or are receiving a poor quality education. Educational planners blame poor educational achievements on inadequate and inefficient resource allocation. On average low income countries spend some 3 percent of GNP on education as compared to 6 percent in high income countries. Moreover, low income countries tend to spend more on teachers salaries and less on teaching materials and teachers' training: outlays that tend to be more effective in positively influencing educational outcomes. Morocco, however, is a middle-income country which spends 7.3 percent of GNP on education. Yet its attainments in education and other social development indicators are similar to those of low income countries. It spends about 85 percent of its education budget on teachers' salaries. Moreover, although 52 percent of the population lives in rural areas, Morocco allocated only 10 percent of its total educational investment to rural education in the eighties. Thus, in the education sector it seems that Morocco's problem is not one of inadequate outlays but inefficient resource allocation. The Government of Morocco (GOM) with the help of the World Bank has recently implemented major educational reforms and has increased investment in th. country's rural educational system. The GOM effort includes, among others, building new schools in rural areas, more training to teachers, and better teaching materials. However, given Morocco's dual economy this supply-based approach to educational development may be necessary but not sufficient to improve schooling outcomes. Demand-side factors such as irrigation and electricity which increase the rate of return to education and hence educational attainment are also likely to be important. Recognizing that both supply-and demand-side factors play important roles and are often complements in schooling production, this paper, using data from the Morocco Living Standard Survey (MLSS), explores the relative effectiveness of both supply- and demand-side factors in determining educational outcomes. The school outcome variables include school enrollment, educational attainment, drop-out rate, and the cognitive achievement of students by gender across different regions. The supply- side explanatory variables include both school access (such as the accessibility of primary and secondary schools, the presence of a paved road and the condition of the roads) and quality (such as teachers' education and the highest class offered by the school) variables. On the demand-side both household (parents' education and assets, and availability of clean drinking water) and market demand variables (such as availability of electricity, irrigation and new crop varieties) are included. A descriptive analysis of MLSS data reveals the following stylized facts. The school enrollment rate for the upper income group is almost twice that of the low income group. Urban children and rural boys consistently outperform rural children and rural girls in both school attendance and attainment. The school attendance rate is 90 percent in urban areas as compared to 48 percent in rural areas and 64 percent for rural boys compared to 32 percent for rural girls. Similarly, the completion rate for rural primary school pupils is 60 percent as compared to 87 percent for urban pupils and it is 63 percent for rural boys compared to 56 percent for rural girls. There are also sharp differences in cognitive achievement between male and female and between -vii- rural and urban children in Morocco. Urban children outperform their rural counterparts and boys outperform girls. For example, the mean test score in math is 11 for rural boys and only 4 for rural girls, while it is 19 for urban boys and 16 for urban girls. However, with additional schooling rural girls improve significantly on the math test. Thus, the mean math score of rural girls with no schooling is only 40 percent that of nrral boys and 47 percent that of urban girls. In contrast, at the middle secondary level, the math score of rural girls is 90 percent that of rural boys and 86 percent that of urban girls. Under-investment in education and gender differences are more pronounced in rural than in urban areas, especially with regard to school enrollment and attainment. Poor access to, and low quality of, schools are perhaps the most important sources of rural-urban and gender differences in schooling attainments. About 44 percent of rural children do not have a primary school and some 97 percent do not have a secondary school in their locality. Some 30 percent of primary schools are satellite schools and 71 percent are co-educational. The average class size in primary schools is 21 students. On average primary teachers have nine years of education and 93 percent of them have a school diploma. Only 20 percent of primary teachers are female. Econometric analysis confirms that both supply- and demand-side factors are important determinants of rural-urban and gender differences in schooling outcomes. The results clearly indicate that improved school quality and access have significant positive effects on the school enrollment and attainments of both boys and girls. The presence of a primary school, or a paved road increases the school enrollment. On the other hand, the school enrollment increases with the number of grades offered in primary school. Teachers' years of education is also positively related to the school attainment of children, especially of girls. However, educational resources are inefficiently allocated in rural Morocco. For example, low quality schools not only have teachers with below average years of education but are also found to have a higher proportion of trained female teachers and lower size of class. The average class size is small because demand for schooling is low as a result of low school quality. In contrast, more trained teachers are located in low quality schools because of recent recruitment policy that requires that new recruits are placed in rural schools. Thus, their effects on improving school quality are not yet realized. The demand-side factors also play an important role. Investment in electricity, irrigation, and new crops increases the returns to education and hence the demand for education. Rural electrification increases the school attendance of both boys and girls, while investment in irrigation and advanced crop technology increase the school attendance of boys. In contrast, public investment in clean water improves the efficiency of time-use, especially of girls, at home and thus reduces the transactions cost of schooling leading to higher school attainment. These findings indicate the need for more public investment in rural infrastructure in order to improve overall educational attainment. Note, however, that although public investments in water, roads, electricity, and new crops are gender-neutral, the benefits of these public investments are not. For example, investments in roads, electrification, and safe drinking water improve girls' school outcomes more than boys', while the opposite is true for investment in irrigation and new crop varieties. Evidence suggests that parents seem to bias private investment toward boys rather than girls. The presence of an additional male member in the family reduces the education girls receive. Moreover, as the household expenditure data show, when gender bias is substantially low, educating girls is more expensive than educating boys. Thus, if male- female cost differences are a partial source of gender bias in educational outcomes and if there are higher social returns to female education, it may be justifiable for the government to introduce stipend for girls -viii- to encourage parents to send more girls to school. The Bank-financed female secondary school stipend project in Bangladesh shows how the government can help improve female education in a country that favors boys over girls. -ix- I. Introduction The benefits of education are well known and well documented (e.g., Haddad and others 1990; Psacharopoulos 1984; Schultz 1961). Yet many developing countries lag behind in developing their education system. Education improves productivity, directly contributing to economic development. It also helps promote development by slowing population growth and improving health and nutrition. In many developing countries, two-thirds of the school-aged children remain outside the education system. Among those who register for school, approximately three-fourths do not even complete primary school. In many countries children either do not have access to education or are enrolled in schools of inferior quality. Educational planners blame poor educational achievements on inadequate and inefficient resource allocation. Low income countries spend about 3 percent of GNP on education as compared to almost 6 percent in high income countries. Furthermore, the resources allocated to education tend to be spent on salaries rather than on inputs such as text bookm and reading materials or on teacher training. These factors have a much stronger influence on educational achievement (Verspoor 1990). Morocco is a middle income country and spends 7.3 percent of GNP on education. However, Morocco's educational attainments are below the levels expected for a country at this level of per capita income. In fact, Morocco's social and educational indicators are, in some cases, below those of low income countries (see table 1). Although Morocco spends a larger part of its GNP on education as compared to many low income countries, it inefficiently allocates these resources. For example, Morocco spends about 85 percent of its education budget on teachers' salaries (UNESCO 1991). It also allocates a disproportionate amount of resources to the urban and wealthy classes. Morocco is a highly dualistic society with a large gap in social and educational indicators between rural and urban areas. Some 52 percent of the country's population live in rural areas, 45 percent of whom live below the poverty line. Rural areas do not, however, receive their proportionate share of public resources. For example, in 1989, about 22 percent of villages with a population of less than 1000 had virtually no access to primary education, and only 100 of the 682 middle schools in Morocco were located in rural areas. Worse still is the fact that none of the 340 high schools in the country were located in rural areas. According to a World Bank estimate, rural areas received only 10 percent of the total investment in education during the 80's (World Bank 1991). To reduce rural-urban differentials in school access and achievements, the Moroccan government, with the help of the World Bank, has implemented major educational reforms and has increased investment in the country's rural school systems in recent years.' The objectives were to improve access to primary and middle schools by investing in the rural educational infrastructure as well as optimizing the resources allocated to the education sector. Inadequate resource allocation to the rural education sector and inefficient overall resource allocation to education are no doubt important factors in explaining low educational attainment in Morocco. Improving school accessibility by building additional schools in rural areas or by improving ' The Rural Primary Education Project of 1989 and the Rurl Basic Education Development Project of 1991 are two examples of the programm which the governmmat has undertaken in recent yea to improve the perfornance of the educational sector. -1- school quality through the training of teachers or through better reading material will certainly affect schooling outcomes. However, the role of demand-side factors in educational outcomes should not be underestimated. Education is an important form of human capital and the demand for human capital is perceived as an input in production. Therefore, the demand for education, which is influenced by the rate of return to education, is often related to the availability of complementary inputs such as electricity, irrigation, advanced technology (such as the development of new breeds of crops) and so on. These complementary inputs increase the marginal productivity of human capital and hence the return to education and thus help promote educational attainment. Believing that there is (latent) demand for education, policy makers often ignore the demand aspects of education and emphasize only supply-side factors such as access to and quality of schools. In Morocco the demand-side factors may play a more important role due to the dualistic nature of its economic development. A little over a decade ago, the rate of return to education was estimated at 16 percent (Psacharopoulos 1981), yet Morocco's school enrollment remains remarkably low compared to even low income countries. However, judging from the national rate of return to education, one cannot draw the conclusion that the demand for education is equally high in all areas. In fact, educational attainment is low in rural areas perhaps because of low returns to education. Although it is possible that rural educated people migrate to urban centers to receive higher returns, it is difficult to judge the extent of this phenomenon. Moreover, massive rural out-migration is perhaps undesirable from the perspective of overall economic development. Thus, investing in complementary inputs such as electrification and irrigation is perhaps as important as building a school or improving the quality of existing schools in rural areas. Note that the supply-side and demand-side factors often reinforce each other in influencing schooling outcomes. For example, better access to and better quality of schools reduces the cost of schooling and hence increases demand. This increase in better quality schooling should improve returns to education which in turn will generate further demand for schooling. Similarly, investments in complementary inputs such as electrification or irrigation that increase returns to education and hence demand for education also promote the demand for better quality schools, thereby generating further demand for education. Thus, poor access to or low quality of schools is often linked to low returns to education and hence to inadequate demand for schooling that jointly produces lower educational attainments. The policy issue is whether supply-oriented educational investments in school quality or accessibility are adequate to promote schooling in rural Morocco. Policy makers thus need to assess the relative effectiveness of various supply- and demand-oriented public investment policies. An evaluation of government programs, therefore, can shed light on how to improve educational attainments in a country that lacks in human resources. The recent 'Morocco Living Standard Survey (MLSS)" and the accompanying literacy survey carried out by the Directorate of Statistics of the Ministry of Planning with help from the World Bank and UNDP, provide an opportunity to examine the supply-demand nexus of schooling outcomes. The aim of this report is to use these surveys in assessing the school performance of male and female students from various regions in Morocco and the role played by demand- and supply-side factors in various educational outcomes. The objective is to disaggregate patterns of demand for education into various categories such as primary and secondary levels and identify factors that impede children's participation in schools, especially that of female students in rural areas. We are interested in school attendance, educational attainment, drop-out rates and the cognitive achievements of students by gender -2- across different regions. The policy variables of particular interest to us are access variables such as the accessibility of primary and secondary schools, the presence of a paved road and the condition of the roads. The other supply-side factors include school quality variables such as teachers' average years of education and the highest grade offered in a given school. Improved school access and quality will reduce the cost of education and hence increase demand. These improvements will also help increase returns to education and hence promote demand. The demand-side factors include both household- and community-level variables. Household- level variables include assets and parental education, while community-level variables include local health conditions and productive potential. Local health conditions are measured by availability of clean drinking water, while productive potential is proxied by the availability of electricity, irrigation and new crop varieties. Given household demand for education, the productive potential of a region will increase the rate of return to education and hence influence children's schooling outcomes. In contrast, local health conditions influence children's health outcomes which in turn affect children's performance in school. The data analysis confirms that educational and cognitive achievements in Morocco are higher for wealthy and urban households than for poor and rural households. The school enrollment rate for the upper income group is almost twice that of the low income group. Urban children and rural boys consistently outperform rural children and rural girls, respectively, in school attendance and attainment. The school attendance rate is 90 percent in urban areas as compared to 48 percent in rural areas and 64 percent for rural boys as compared to 32 percent for rural girls. Similarly, the completion rate for rural primary school pupils is 60 percent as compared to 87 percent for urban pupils and it is 63 percent for rural boys as compared to 56 percent for rural girls. There are also sharp differences in cognitive achievement between male and female and between rural and urban children in Morocco. Urban children outperform their rural counterparts and boys outperform girls. For example, the mean test score in math is 11 for rural boys and only 4 for rural girls, while it is 19 for urban boys and 16 for urban girls. However, with additional schooling rural girls improve significantiy on the math test. Thus, the mean math score of rural girls with no schooling is only 40 percent that of rural boys and 47 percent that of urban girls. In contrast, at the middle secondary level, the math score of rural girls is 90 percent that of rural boys and 86 percent that of urban girls. Under-investment in education and gender differences are more pronounced in rural than in urban areas, especially with regard to school enrollment and attainment. Econometric analysis of the MLSS data confirms that poor access to and low quality of schools are important factors in rural-urban and gender differences in schooling attainments. Rural Morocco has poor access to primary and secondary schools and the quality of existing rural schools is low though not as low as one might expect. There is, in fact, counter-intuitive evidence that suggests that educational resources are inefficiently allocated in rural Morocco. More importantly, rural Morocco lacks complementary investments in roads, irrigation and safe drinking water, all of which are found to significantly contribute to schooling outcomes, especially among girls. Increasing the returns to education and improving the efficiency of the rural school system would improve both school attendance and attainment. These findings call for increased government efforts to improve the efficiency of resource allocation in education. The findings indicate the need for more public investment in rural infrastructure in order to improve overall educational attainment. In particular, investment in roads, electricity, irrigation, new high-yielding crop varieties and clean drinking water, would effectively complement -3- investment in education infrastructure and quality. Furthermore, although public investments are gender neutral, parents seem to bias private investment toward boys rather than girls. Yet parents in Morocco who enroll both boys and girls in school may need to spend on average more for girls than for boys at each level of schooling. If the cost differences are partial source of gender bias in parental investment, and if there are large social gains in female education, then the government may provide stipend to girls as an incentive to parents for promoting equal investment in boys and girls. The report is organized as follows: Section II discusses the MLSS and literacy survey data. Section m presents the stylized facts about Morocco's educational outcomes. Section IV outlines a theoretical framework for estimating the causal impacts of both supply- and demand-side factors on educational outcomes. Section V presents the results. Section VI discusses some policy simulations and their implications. Section Vn concludes the report. H. The Iiving Standard Measurement and Literacy Surveys The data used in this report is taken from the 1990-91 Morocco Living Standards Survey (MLSS), a comprehensive household survey implemented by the Morocco Statistical Department on random samples of households throughout Morocco. The MLSS consists of a broad household questionnaire that includes information about schooling, employment, labor and non-labor income, consumption and expenditures, health, fertility and savings. The MLSS sample, stratified across the seven economic regions of Morocco, contained 20 primary clusters of 24 households within each region. Each primary cluster was further divided into 3 secondary clusters of 8 households each. The sample included 3360 households with almost 20,000 individuals. From this sample, 2 of the 3 secondary clusters in each primary cluster were targeted for inclusion in a Literacy Survey, for a total expected sample size of 16 households (including all eligible household members) per primary cluster, or 2224 households nationally.2 The literacy survey was administered to household members aged 9 to 69. Information collected included self-reported mother tongue, educational experience, self-judgments on language-specific reading and writing abilities, math abilities and reading and writing practice at home and at the work place. Those with formal education were also given direct assessment tests of document familiarity, basic reading and writing skills in Arabic and French (when some French ability was reported) and mental and written math.3 The full literacy survey consisted of the following nine sections: (1) general self-reported questions on literacy skills and behavior; (2) questions concerning basic health care behaviors (addressed to heads of households and their spouses, and other persons aged 20 to 50 only); (3) assessment of 2 When a target household or a majority of eligible individuals wihin a household could not be surveyed for any reason, a replacement household of similar size from the third secondary cluster in a given primary cluster was selected and surveyed. For greater detail on overall sample structure and representativity, see Lavy and others (1992). 3 High school grduates with a baccalaureate or its equivalent were exempted from the tests and later assigned the highest score possible. The concept of baccalaureate 'equivalency' and specific cases encountered in pretesting led to considerable discussion regarding the appropriate criteria for respondent seUction and assessment. A summary of the final system of assessment decision criteria is available in Lavy and others (1992), and in the Literacy Survey Instruction Manual. -4- information location skills, using common literacy artifacts; (4) mental arithmetic assessment (two levels); (5) Arabic reading assessment (two levels), including word decoding, word-picture matching Oevel 1), and text comprehension (evel 2); (6) Arabic writing assessment (two levels), including signature and word dictation (level 1) and sentence dictation Oevel 2); (7) French reading assessment (two levels, equivalent to Arabic reading assessment); (8) French writing assessment (two levels; equivalent to Arabic writing assessment); and (9) written numeracy assessments (two levels), including number recognition and writing Oevel 1) and solving of written equations Oevel 2). The results of the direct assessments were used to produce separate scores for each skill or knowledge area. Words used in Arabic and French level-l reading and writing tasks were selected from primary school textbooks currently in use in Morocco. Texts used in text-comprehension tasks were adapted from primary, secondary and adult literacy textbooks and from local Moroccan newspaper articles. The information location task, sample knowledge of literacy artifacts (envelope, newspaper, electricity bill, medicine label, national identity card) commonly found in Moroccan homes. Math problems, both oral and written, sample the basic arithmetic functions and concepts of measurement, percentages, fractions, decimals and plane geometry. Within each task, efforts were made to order items by increasing level of difficulty (for those tasks using school text content, according to their point of introduction in the curriculum); these orders were adjusted on the basis of pilot testing results.' Supplementary surveys of communities and schools were carried out in all the rural clusters. The community questionnaire was administered in 324 rural communities (i.e. douars), while the school survey was conducted in 186 schools (both primary and secondary). However, these 186 schools were found to serve only 240 out of 324 rural communities. In total there are 240 rural communities for whom we have both community- and school-level information. The community questionnaire was designed to provide information on the availability of social and infrastructure services such as transportation, health, financial and agricultural extension services, as well as general information on local labor markets conditions. The school survey includes detailed information on all the primary and secondary schools attended by children from the community. The school questionnaire includes variables describing the physical condition of the school structure, availability of teaching materials, number and level of trained teachers and their gender composition, as well as summary statistics of schooling outcomes at the school level, such as repetition and drop-out rates by grade level. The school data provides a fair characterization of access to and quality of schooling services available to households in the community. m. School Performance Indicators: A Descriptive Analysis In this report, we discuss the school performance of children aged 7 to 20. The term performance is used here to denote school enrollment rates and achievement scores in math and languages. The data are broken down by gender, location, and region in order to highlight differences in school outcomes among various groups of interest. Thus, four location-gender groups - rural male, ' The survey draft was subjected to pretesting, revision, full pilot testing, and a second revision prior to full-scale application in the field. Pretesting of the literacy survey was undertaken as a means of testing the adequacy of interviewer training and the efficiency of survey format and administration procedures. As such, it was intended to provide qualitaive information for the finalization of methology, rather than to allow detailed psychometric analyses (carried out in the ater piloting phase). Psychometric analyses of all assessment modules were performed in order to examine the internal consistency of the modules, the reliability and difficulty level of individual items and the equivalence of the two fonns of each module. -5- rural female, urban male and urban female - and seven regions - South, Tensift, Center, North-West (which includes the capital city - Rabat), North-Central, Oriental and South-Central provinces of Morocco - are used to investigate the patterns of male-female, rural-urban and regional differences in various school outcomes. The sample contains 6801 children aged 7 to 20 from 2397 households taken from seven provinces (see table 2). 42.2 percent of these children are urban and 57.8 percent are rural. 49.6 percent are male and 50.4 percent are female. There is little variation in the male-female ratio across the seven regions. The North-Western region has, however, the lowest percentage of rural children (52.5 percent), while the Southern region has the lowest male rural population (26.4 percent against 31.2 percent rural females). Schooling Outcomes We examined three schooling outcomes: (a) never attended school; (b) attended and completed school; (c) currently enrolled in school. Two more outcomes can be defined using the sample of current or ex-students. These are the completion rates of primary level (up to grade 5), secondary I level (up to grade 9) and secondary 2 level (up to grade 13) and the transition rate from primary to secondary 1 and from secondary I to secondary 2 level. The achievement scores in mathematics, Arabic and French are indicators of schooling outcomes as well. Ever Attended School As table 3 shows, about 66 percent of all children have attended school at one time or another. There is a marked difference between the school enrollment rate in urban and rural areas. Children in rural areas are far less likely to attend school - the participation rate is only 48 percent compared to 90 percent in urban areas. The difference in male-female ratios in school attendance is not as conspicuous (only 10 percent) in urban Morocco. In contrast, the school enrollment of rural female children is half that of rural male children (32 percent compared to 64 percent). There has been a 50 percent increase (from 60 to 90 percent) in the school attendance rate in urban Morocco since the introduction of several World Bank-financed school projects in 1989. In rural areas there has been a 20 percent increase, (from 40 percent to 48 percent), with an overall national increase of 30 percent (see table 3). This is a significant increase by any measure. In order to further examine the impact of program placement among the younger population, the sample is disaggregated into various age-cohort groups. The school enrollment rate for both males and females in urban and rural areas has recently increased dramatically especially for urban female children (from 61 percent for 20-30 year olds to 84 percent for 10-20 year olds (see figure 1)). Although enrollment is lowest among rural women in all age-cohort groups, they have, nonetheless, experienced the largest relative increase in school enrollment in recent years. The enrollment rate is 11 percent for the 20-30 year-old age group, and 30 percent for the younger group (10-20 years old), an increase of 173 percent as compared to 48 percent for rural males of the same age groups. These figures indicate that public investment in education has contributed to an increase in school attendance, but more incentives are still needed in order to improve school attendance rates in rural areas, especially among female children. -6- Schooling Attainment The male-female differences in mean grade attainment in both urban and rural areas are small, the mean grade being slightly lower for girls (see figure 2). The mean grade in rural areas is 4.93 for girls and 5.43 for boys. It is 6.96 for girls and 7.14 for boys in urban areas. A significant difference, however, exists between rural and urban areas. The mean grade for urban children is 7.05 compared to 5.18 for rural children. Primary schools go up to grade 6 which means that many rural children drop out of the school system even before completing the primary cycle. Table 4 shows that, contrary to expectation, girls are doing better than boys in some regions, especially in Tensift where the mean grade in urban areas is 7.16 for girls compared to 6.65 for boys and in rural areas 5.61 for girls compared to 5.38 for boys in rural areas. It also shows, however, that in some rural areas girls are doing much worse than boys. The male-female gap is the highest in the Central-South region where boys attain a full grade more than girls (5.04 vs. 4.03). Moving beyond simple means, figure 3 presents the number of children in each age group still in school, conditional on having ever attended school. The age-schooling profile confirms that at each age level, both girls and boys in urban areas tend to spend the same amount of time in school, whereas boys stay in school longer than girls in rural areas. In each age-group the percentage of males and females attending school is higher in urban than in rural areas. This difference is particularly noticeable in the over 11 age group. A regional breakdown (not shown here) also confirms these findings. An alternative approach to school attainment is to calculate survival probabilities by grade, i.e. the probability that an individual will "survive" until a particular grade (and then either remain in school or drop out). Figure 4a shows the Kaplan-Meier survival curves (taking into account right censoring) for grade attainment among four location-gender groups. The results are consistent with the overall patterns that emerged earlier, i.e., a similar age-schooling profile for girls and boys in urban areas, lower probabilities of survival for rural girls relative to rural boys and markedly different patterns in rural and urban areas. In fact, the probability of survival in rural areas is about half that for urban areas. The survival pattern reveals an additional dimension of male-female differences in rural areas - - rural girls have a much lower probability of surviving grade 6 (the end of primary school) than their male counterparts. The chance of surviving grade 6 in rural areas is 43 percent for girls and 56 percent for boys. The difference in survival rates for boys and girls persists from grade 6 through 12 after which they tend to equalize. A regional breakdown of survival patterns (figures 4b and 4c) identifies three regions where male-female differences are particularly large (the remaining four regions show remarkably similar patterns for boys and girls): the Southern, Oriental and South-Central. In these regions, the survival rates at grade 6 are lower for girls than for boys and do not converge in subsequent grades. School Completion and Transition Rates The above discussion points to the fact that there may be important thresholds in the Moroccan educational system which have different impacts on individuals depending on location and gender. Up to university level, the schooling system can be divided into three major components: primary school, -7- secondary level 1 and secondary level 2. Primary school lasts for six years, secondary level I lasts four years and secondary level 2 lasts three years. The completion rate for primary school is similar for boys and girls in both rural and urban areas (see figure 5). This rate is 63 percent for boys and 56 percent for girls in rural areas and 88 percent and 85 percent, respectively, in urban areas. The primary completion rate is even higher for girls than for boys in regions such as Tensift, Center and North-West when the data is broken down by region (not shown here).5 However, in some rural areas, the completion rate for primary school is substantially lower for girls than for boys. In the Southern, Oriental and South-Central regions, for example, the primary completion rate for boys is 71, 73 and 57 percent, respectively; for girls, the rate is 55, 53 and 36 percent, respectively. The transition rate to secondary school after completing the primary cycle is in general much lower in rural areas than in urban areas (see figure 6). Although the rate is similar for male and female children in urban areas (86 percent for boys and 85 percent for girls), it is lower for girls than for boys in rural areas (57 percent for boys and 45 percent for girls). The male-female differences in rural areas are significant in all regions (table 5). The male-female difference is highest in the Oriental and Central- South regions. Thus, the transition rate from primary to secondary level I is 64 percent for boys and 34 percent for girls. In the South-Central region it is 43 percent for boys and only 13 percent for girls. Once a child is in secondary level 1, the probability of completing this cycle does not vary significantly by gender. It does, however, vary substantially according to location. Figure 5 shows that the completion rate is approximately 80 percent in urban areas but only 35 percent in rural areas. Although on average the completion rate is lower at the secondary level than at the primary level in all areas, it declines more in rural than in urban areas: the completion rate declines by 17 percent in urban areas as compared to 25 percent in rural areas.' The transition rate from secondary level 1 to secondary level 2 also follows the previously identified pattern. Thus, while we find that transition rates are similar for boys and girls, they differ between rural and urban areas. Unlike the transition pattern of the first cycle (i.e. from primary to secondary level 1), the rural-urban gap during the transition of the second cycle (i.e. from secondary level 1 to level 2) is narrower. In fact, there is almost a 10 percent increase in the transition rate for rural children, whereas the rate declines slightly for urban children. Overall School Experience Table 5 shows the typical overall school experience of Moroccan children at various levels. In this table, we follow a fictitious group of 100 children of each gender in each location, and compute, based on the above attendance, attainment and completion rates, the number of children remaining in the school system at each level. This clearly shows the sharp rural-urban differences in the school performance of Moroccan children: of 200 children (100 of each gender) in each location, some 60 5 The difference is the highest in the Tensift region where some 64 percent of rural girls complete primary school compared to 57 percent for rural boys. ' Brealing the sample down by region reveals a very similar pattern. However, interpreting the results of regional differenes may be misleading because of the small sample problem, especially in rural areas. -8- children complete secondary level 2 in urban areas compared to only 4 children in rural areas. Approximately 107 children complete secondary level 1 in urban areas compared to only 11 in rural areas. In contrast, some 156 children complete primary level in urban areas compared to only 58 in rural areas. Of 200 children aged 7-20, some 104 children never attend school in rural areas compared to only 19 children in urban areas. Gender differences are quite substantial in rural areas, while they are only marginal in urban areas. However, the gender gap in rural areas matters most in terms of school attendance. Thus, out of 100 children of each gender, some 64 male children attend school in rural areas compared to only 32 females. Once girls are enrolled in primary school, their completion rates at each school level are similar to (albeit slightly lower than) those of their male counterparts. Achievement Test Scores A series of tests covering mathematical, Arabic and French language skills was administered to a subsample of individuals from the MLSS. Of the 6801 children in the sample, 3785 were tested. Of these, 1089 (29 percent) were rural males, 1124 (30 percent) were rural females, 758 (20 percent) were urban males and 813 (21 percent) were urban females. The children came from 1477 households, of which 815 (55 percent) were rural and 662 (45 percent) are urban households. Since persons with a baccalaureate diploma were not tested, only the test scores at the primary and secondary 1 schooling levels are considered in this report.7 The test score results, stratified by schooling level, are presented in figures 7a, 7b and 7c. There is clearly an upward trend in achievement scores for all three tests as educational attainment increases. Math and Arabic scores are most influenced by the transition from no schooling to primary schooling whereas the largest change in French scores occurs with the transition from primary to secondary level 1 . Rural-urban differences are most pronounced on the French test in which the scores of those with primary schooling in urban areas is about double that of their counterparts in rural areas. The score is 7.6 for urban boys compared to 3.8 for rural boys and 7.2 for urban girls compared to 4.0 for rural girls. T he rural-urban difference is also present at the secondary level 1, although it is not as pronounced as at the primary level. On the math and Arabic tests, the rural scores range between 70 and 80 percent of those in urban areas. Given location, there is a strong similarity between the scores achieved by male and female children. The only test with a substantial male-female difference is the math test on which boys outperform girls. The largest difference is in rural areas and among those with no schooling where girls score only about 40 percent as high as boys. At the primary and secondary level 1, girls score about 85 percent as high as boys. On the Arabic and French tests the mean scores of girls range from 85 to 107 percent of those of boys. In Arabic and French, at both secondary level I and 2, girls in urban areas outperform boys. Breaking these scores down by region reveals very little regional variation (not shown in any of the tables or figures). In certain regions rural female children are scoring lower than their male 7 For more on the test scores, see Lavy ct al. (1993). -9- counterparts. This is especially the case for math scores at the primary school level in the North-West, Center-North and Center-South regions. The same holds true for secondary level 1 in the Southern and South Central regions. On the Arabic and French tests, rural girls score lower than rural boys in the North-West, Center-North, Center-South and Southern regions. Summary of Schooling Outcome Patterns The main stylized facts emerging from the analysis can be stated as follows: (i) Urban children consistently outperform rural children according to all schooling outcome measures. (ii) Gender differences are particularly pronounced in rural areas. (iii) Gender differences are most pronounced in the rate of ever attending school. (iv) Given location and ever attending school, gender differences are low in school attainment, completion and transition from one level to another as well as in scores achieved in Arabic and French skill tests. (v) The completion rates at primary and secondary level 1 do not vary by gender but are higher in rural than in urban areas. (vi) There are regional differences in the schooling outcomes of rural girls. Rural girls in the Southern, Tensift and North-Central regions, for example, do not attend as frequently or stay in school as long as their counterparts in other regions. Demand Factors and Schooling Outcomes Schooling outcomes are influenced by a variety of individual, household, community- and school- level factors. These factors can be classified into two major groups - demand-side and supply-side. We will consider first the demand-side, focusing on the association of schooling outcomes with household income and parental education. Household Income Household per capita expenditure is used as a proxy for household income. The distribution of children by household expenditure quintiles is presented in table 6a. In the poorest (i.e. lowest) quintile, children of age 7-20 are predominantly (79 percent) from rural areas. The proportion of rural children declines steadily as we get into higher quintiles. In fact, urban children account for 83 percent of all children in quintile 5. Figure 8 shows the association between household per capita expenditure and the proportion of children who ever attend school. This figure clearly shows an upward trend in the proportion of children who ever attend school, for all location-gender groups, as household per capita expenditure increases. The proportionate increase in the rate of ever attending school is correlated with changes in per capita expenditure and varies with location and gender. Within each quintile, the proportionate increase is lowest for rural girls and highest for urban boys. The rate for rural girls ranges from 25 percent in the lowest quintile to 46 percent in the highest quintile. In contrast, the rate for urban boys varies from 89 percent in the poorest quintile to 100 percent in the richest quintile. The rural-urban gap in the school attendance rate narrows as income increases. For example, the school participation rate for boys in the poorest quintile is 59 percent in rural areas and 89 percent in urban areas. In contrast, the attendance rate for boys in the richest quintile is 67 percent in rural areas and 100 percent in urban areas. The male-female gap in school participation also narrows as income increases. For example, in rural areas the school attendance rate for girls in the poorest quintile is only 32 percent of the rate for -10- boys in the same quintile, while the proportion is 68 percent in the richest quintile. In fact, the school participation of rural girls increases by more than 84 percent from the lowest to the highest quintile. In contrast, the corresponding increase for rural and urban boys is 12 and 13 percent, respectively. However, the school participation of urban girls falls by 11 percent from the poorest to the richest quintile. This is perhaps due to the early age of marriage for girls from the wealthier classes in Morocco. Figure 9 shows the mean grades attained by children of various expenditure quintiles, conditional on ever attending school. It shows that rural-urban differences in the mean grade completed increase with income. The mean grades completed by rural boys and girls are, 81 and 77 percent respectively, of those for urban boys and girls in the lowest quintile, while in the wealthiest quintile the percentages are 77 and 72. Thus, the income-induced increase in the rural-urban gap in school attainment is similar for males and for females. The percentage of those still in school by age shows that girls have consistently lower attainment rates than boys in almost all the expenditure quintiles, especially among children aged 12 to 16 in rural areas. Survival rates by grade confirm this result which is consistent for all expenditure quintiles (see figure 10). Although girls in rural areas have lower probabilities of surviving grade six than boys, they tend to outperform boys in the higher grades. There is also a large difference between rural and urban areas in terms of the percentage of children who stay in school once they are enrolled. The rural-urban gap tends to increase with income. The same pattern is also seen in the completion and transition rates (figures 1la, 1 lb and 12). TMe primary completion rates increase with income, the largest increase being for rural girls; the completion rate for girls increases from 48.6 percent in the lowest quintile to 88.9 percent in the highest quintile. For rural boys the primary completion rate only increases from 58.8 to 70.0 percent. The transition rate from primary to secondary level 1 does not follow a uniform pattern for either boys or girls. The transition rate for rural girls in the lowest quintile is almost double that for rural girls in the highest quintile (45.2 percent versus 25 percent) and follows a vertical u-shape as income increases. Results of the achievement tests do not indicate a wide variation in scores between expenditure groups. The gender gap on the language tests is a small one. The largest difference between girls and boys is to be found on the math test for secondary 1 in rural areas among the middle expenditure group. Girls from the second, third and fourth quintiles have math scores 60 to 80 percent lower than hose of boys. Parental Education We turn next to the association between schooling outcomes of children and their parents' education. The majority of children whose parents have no education are from rural areas. Some 67 percent of children from families where the head or the spouse has no education are from rural areas. However, children's educational attainments improve with parents' schooling (dramatically in the case of the spouse's education). As figure 13 shows, the rate of ever attending school for children from households where the head has no education is 62 percent for rural boys, 29 percent for rural girls, 94 percent for urban boys and 84 percent for urban girls. The percentages in households where the head has an education of secondary level 1 are 82, 62, 100 and 94 percent, respectively. A similar pattern emerges for the association of spouse's education with children's schooling (with the exception of rural areas where the percentage of boys who ever attend school declines from 64 to 50 percent when the spouse education level changes from no education to secondary level 1). -11- There is, however, a sharp difference in the association between the education level of the household head and the spouse and the girl's education in rural areas. The school participation rate of rural girls increases by 55 percent when the household head's education increases from none to prinary level. In contrast, the participation rate increases by 135 percent when the spouse's education changes from none to primary level. The spouse's education also has a stronger impact than the head's education on boys' and girls' schooling in urban areas. Once children are in school, the parents' education does not appear to play a role in determining the length of time a child spends in school. Figure 14 shows that the mean grade achieved hardly varies with parent's education. Although there is a positive association between parent's education and children's mean grade attained, this association does not vary significantly with gender or location. Supply-Side Factors and Schooling Outcomes The supply factors include variables that measure access to and quality of schools. Access is measured by whether there is a school in the community, whether the community has a paved road and how many months of the year the road is unusable. These access variables were measured and collected at the community level. In contrast, school quality is measured by variables such as teacher-student ratios, male-female teacher ratios, school supplies, textbook supplies, etc. which are collected at the school level. About 44 percent of rural children do not have a primary school located in their community, while some 97 percent do not have a secondary school in their community (table 7).' Thus, the presence of a school substantially improves a child's schooling outcomes. For example, as table 7 shows, 56 percent of rural boys and 24 percent of rural girls ever attend school when there is no primary school in the community. In contrast, the school attendance rate increases to 72 percent for boys and 37 percent for girls if there is a primary school in the community. Similarly, 56 percent of rural boys and 44 percent of rural girls go beyond primary school if there is no secondary school in the community. However, 83 percent of rural boys and 75 percent of rural girls go beyond primary school if there is a secondary school in the community. The presence of a secondary school in the community thus increases substantially the mean grade attained by girls who go beyond primary school (table 9). Thus, when girls go beyond primary school they attain a mean grade of 11 if there is a secondary school in the community as compared to a mean grade of 9 if there isn't. Figure 15 shows the survival rate of boys and girls at various grades in the school system when a primary and secondary school are present in the community. When there is neither a primary nor secondary school in the community, the survival rate declines for both boys and girls with virtually no gender difference in school attainment. However, when there is a primary school but no secondary school, the survival rate beyond the primary level is higher for boys than for girls. When there is both a primary and a secondary school located in a community, the survival rate increases for both boys and girls. However, the rate is lower for girls than for boys. This indicates that providing a secondary school in a community does not automatically reduce the gender gap that exists in school attainment beyond the primary level. More information on school quality and parental preferences are 'Similar conimunity-lcvcl infornation is not available for urban aLeas. Therefore, a similar analysis was not performed for urban children. -12- needed in order to understand why girls respond differently than boys in deciding whether to continue beyond primary school level. The presence of a paved road in the community especially influences the schooling outcomes of rural children. Thus, in the absence of a paved road, 21 percent of rural girls as compared to 58 percent of rural boys ever attend school (table 10). If a paved road exists, the school participation rate increases to 48 percent for girls and 76 percent for boys. The survival rate by grade also varies by gender in the presence of a paved road: girls drop out in larger proportions than boys even before completing the primary cycle in the absence of a paved road in the community. However, the presence of a paved road increases girls' survival rate only at the primary level. The survey data on primary schools is summarized in table 11. The table presents the salient characteristics of the rural school system in Morocco in terms of access and quality. As table 11 indicates, most primary schools are located near the community - the average distance is 50 meters. About 42 percent of primary students go to school on foot with the trip lasting 49 minutes on average. 30 percent of primary schools are satellite schools. The highest grade offered in the primary system is fifth. 71 percent of these primary schools are co-educational. Only 3 percent of students purchase books from school and only 1 percent get their books free from school. The average size of a primary school is 21 students. The mean years of education for primary teachers is 9; 93 percent of them have capes (i.e. a school diploma certificate). Only 20 percent of primary teachers are females; 56 percent of them have capes. These schools report that only 14 percent of students complete primary school, 83 percent of whom go on to secondary school. This descriptive analysis shows the association, though not the causality, between various supply- and demand-side factors and schooling outcomes. For policy formulation, however, it is essential to identify the causal impact of various supply- and demand-side factors on schooling demand. This requires the estimation of an econometric model of schooling demand using the MLSS data. In the following sections, we first describe an economic model that explains the relationships between potential explanatory variables and schooling outcomes in an estimable equation form and then estimate the model using the MLSS data and analyze the results. IV. The Role of Demand- and Supply-Side Factors in Determining School Outcomes: A Theoretical Framework The purpose of this theoretical framework is to justify the inclusion of certain variables in an estimated model and to provide an economic interpretation of their coefficients. Variables are often included in estimated models that are difficult to justify. To avoid this kind of problem, a household model is developed.9 A schooling outcome such as school enrollment is a home-based non-market production that utilizes home-, market- and community-supplied inputs. Parents are key players in schooling outcomes since they initiate the demand for schooling. The price structure facing parents in demanding schooling for children also influences schooling outcomes. Parental influence is direct via preferences (proxied by education) and income. School quality, on the other hand, which is characterized by such factors as 9 A similar model is developed in Khandker (1993); for general discussion of the household model, am Singh, Squire and Stauss (1986). -13- school location and the number of classes and teachers (male and female), exerts both a direct and indirect impact on schooling. Similarly, the community environment, as proxied by variables such as roads and electrification, can also have both a direct and indirect effect on schooling. Many community variables also represent the role of both demand-and supply-side factors in influencing schooling. Given the high incidence of ill-health among school children in developing countries, public investment in, for example, private health care can substantially improve schooling outcomes by (a) directly affecting the health status of children and hence their cognitive achievements and (b) increasing the efficiency of time devoted by children to schooling and other activities. Investment in public health that improves the health conditions of other members of a family such as the mother or sibling can also affect the schooling outcomes of children, especially girls who are responsible for the care of sick household members. It seems, therefore, that the government can play an active and important role in schooling outcomes through various interventions that affect both the school and community environment as well as the price structures of various goods and services which the household consumes. Identifying the precise influences of all these factors on schooling outcomes is difficult due to the complex relationship between the variables. Nevertheless, estimating the impact of these variables on schooling outcomes is not impossible. In order to facilitate the understanding of the interactions among the variables and their estimation, we first model the underlying household decision-making process that is based on children's health, education and expected earnings. Following this, we discuss the estimation strategy. Assume that parents, consisting of a mother (m) and a father (f) with preferences rm and rf and a combined income (I) net of consumption, decide how much health care and schooling will be allocated to their children; this, in turn, determines the children's expected earnings over the life cycle.'0 Parents care about the investment in their children because it directly yields utility (U) to them. The utility function of the parents can be written as: (1.1) U = U(S, H, R, L; rI, r) where S, H, R and L are, respectively, the schooling, health, expected earnings and leisure of their children. As shown in (1.1), the curvature of the utility function is influenced by the parents' preferences, rm and r'. Both schooling and health are outputs of home-based non-market production that use both market-purchased and home-provided inputs, given the technology of production and the environment in which the household production takes place. The production function of schooling is given by: (1.2) S = S(M, H, K; a) where M is a vector of market-purchased inputs such as books, papers, pencils, etc., H is the children's health status, K is the "effective" time children devote to schooling and a represents both individual endowment and the school environment. Equation (1.2) states that given the technology and environment, " For simplicity, we assume that the number of children, male and female, is given. In other words, the number of children is not a choice variable. Doing otherwise would only complicate the model and its eatimation. In particular, family planning programs that affect ferility outcomes will aso influence the schooling and health outcomcs of children. -14- the health status of children directly affects their schooling outcomes. However, children's health (H) is in turn produced with market-purchased inputs (Z) in a health environment summarized by r, given the household production technology. The production function for health is given by: (1.3) H = H(Z; r) The ultimate objective of parents in giving their children better schooling and health is to help children maximize their life-time expected earnings (R). The expected earnings of children are, therefore, assumed to be directly influenced by schooling (S) and indirecdy by health status (H), given the productive environment as summarized by a. By productive environment we mean the productive potential of the local area. The productive potential in a rural setting may be represented by the extent of irrigation, the presence of new high-yielding varieties of crops and the quality of the physical infrastructure. The children's expected earnings function is given by: (1.4) R = R(S; H, a) In the earning equation (1.4), we assume that children's health status (H) and productive environment (a) affect the earning impact of schooling, but do not directly affect earnings. That is, better health and a productive environment only influence the productivity effect of schooling. The health status of children also affects the quality of time (N) devoted to schooling which is given by: (1.5) K = K(N, H) The 'effective' time (K) devoted to schooling is thus directly affected by children's time allocation to schooling and their health status. Because improved health status can improve an individual's total time available E (see Grossman 1972), and since children in a rural setting where the labor market is not perfectly competitive can alternatively allocate their time for the income generation of parents at some remuneration x, the children's time constraint can be written as: (1.6) Y + L + N = E(H) where Y is time spent on current income generation for parents. The parents' budget constraint is then given by (1.7) PMM + PzZ = I + TrY The first-order conditions, derived from the maximization of (1.1) subject to the structural conditions (1.2)-(1.7), for the optimal amounts of market-purchased inputs, M and Z, used in the production of schooling and health, respectively, and the time spent on schooling (N) and leisure (L) are given by: (2.1) (US + URRS)SM = 4Pm -15- (2.2) [(UH + US + U RRX)(SH + SKKKI)]HZ = §[Pz - dENHzj (2.3) (US + URRS)SKKN = §$ (2.4) UL = T where v is the marginal utility of income which is positive. Equations (2.1) and (2.3) state that in order to obtain the optimal utilization of inputs in school production, the relative productivities of inputs, M and N, in school production must be equal to the ratio of their respective prices, i.e. (3.1) SMI(SKSN) = PM/'r Equations (2.1) and (2.2) also state that the use of market-purchased inputs (M) and children's time spent on schooling (N) depends not only on how parents value children's schooling and expected earnings, but also on the expected return to children's schooling. In equation form, this may be written as: (3.2) US = 4(PMlSM) - UR& = §'(T/SKKN) - URRS Similarly, equation (2.2) implies that parental investment in children's health is in part influenced by the reurn to children's education and the alternative use of time for current income generation. In particular, improved health can reduce the shadow prices of health inputs (Z) since health (a) directly increases current income, (b) increases the total healthy days available that have alternative use in augmenting parental full income and (c) reduces the cost of producing schooling via improved efficiency. In equational form, this can be written as: (3.3) UH = §A/(SH + SKKH)HZ where A = IPZ E T{EH + 1 + (SH/SM)PM)Hz] is the shadow price of producing health for children. On the other hand, equations (2.3) and (2.4) imply that the parents' marginal utility from children's leisure must be equal to their marginal valuation of children's time in alternative uses such as income generation that augments parental full income. The reduced-form demand equations for schooling and health of children are given by: (4) S*, H- = D1(PM,, PZ , I, 7, X, a, , r, rf), i=S, H The explanatory variables in the schooling and health outcomes equation are prices, household income, parental preferences, current opportunity cost of children's time in schooling, health and school environment and productive potential which is considered exogenous to the parents' investment decisions. Substituting (4) into (1.4), we derive the optimal earnings function for children: (5) R' = R(S5; H', a) -16- Equation (5) states that improved market opportunities have both a direct and indirect effect on children's expected earnings. An improvement in productive potential increases the demand for human capital, which in turn increases the returns to schooling and hence affects both the schooling and health status of children. Similarly, an improvement in the health environment also improves children's schooling and expected earnings via the direct impact on schooling and indirect impact on the efficiency of time use in schooling production. For similar reasons, changes in school access and quality that affect school outcomes may also affect returns to schooling and hence the demand for human capital. The causal impact of many policy variables are a priori difficult to quantify due to their complex interactions with household decision-making and observed behavior. An empirical implementation of this theoretical framework can, however,shed light on the magnitudes and significance of the effects of supply- and demand-side factors. We will estimate the reduced-form schooling outcomes equation (4) that incorporates the impact of all variables considered exogenous to household decision-making, including market opportunities that influence returns to education. The earnings equation (5) is also estimated in order to calculate the return to education for men and women. Assuming that these estimates are the expected returns to children's schooling, they are then evaluated in terms of community and school quality as well as children's school achievements. This exercise will help identify the policy matrix that the government could pursue in order to promote schooling among children, especially in rural areas. V. The Results As section m suggests, the problems that Morocco faces in terms of schooling outcomes are far more critical in rural areas. Therefore, the empirical implementation of the model in section IV is carried out only for rural Morocco."1 A total sample of 3652 children aged 7-20, equally divided between boys and girls, is used to estimate the model (table 11). The school participation rate among these children is 48 percent; they have attained an average of 3 years of education. The household demand for schooling is influenced by household income. However, since household income is endogenous, we use the value of land and farm assets as instruments. Since these variables are highly correlated with household income in a rural setting, they should capture the impact of household income on schooling outcomes. Parents' education may influence preferences and hence may affect the technology of schooling production as an input. Male heads of households have an average of two years of education while the female spouse has an average of one. The local demand for educational skills is proxied by the extent of electrification, irrigation and new improved crop varieties at the community level. About 16 percent of rural communities under study have electricity, 42 percent have some irrigated land and 35 percent grow new varieties of crops. These variables measure rural productive potentials and hence determine the impact of returns to education on the demand for human capital. The local health conditions are measured by the sources of water from pipelines and wells as opposed to rivers and ponds. Only 11 percent of these rural households have access to tap water, while 50 percent use well water. The access to a school is measured by the presence of a school and a paved road; it is also measured by how many months of the year the road is passable. Some 56 percent of the communities have a primary school, while only 3 percent have a secondary school. Only 3 percent of the communities have access to a paved road. The roads are blocked due to bad weather and other conditions an average " Anodwff reason for swlcting rural areas for analysis is that community- and school-lcvel variables were collected only from rural areas. -17- of one month a year. The school quality variables included in the regression include the highest grade in a primary school (grade 5), the proportion of female teachers (18 percent) and the mean education level of teachers (9 years). Although data on other indicators is available (see table 1 1), they are correlated with each other and often correlated with community characteristics (for example, a wealthier community has a better quality school). Therefore, an aggregate measure of school quality is desirable and was estimated. This is a latent school quality indicator which was estimated using the fixed-effects method. The math scores of children who ever attend a particular community school were regressed against primary school dummies, among others."2 The fixed-effects coefficients of school dummies are then treated as latent school quality variables in the schooling outcomes regression. Since community characteristics may influence school quality, three alternative models were specified and estimated: Model I includes, among others, community-level variables, such as pure school access variables, but no school quality variables; models II and III use the alternative measure of school quality along with community characteristics. Table 12 presents the probit regression of school enrollment by gender. Table 13 presents the Cox regression of school attainment (measured by last class attended) that takes into account right-hand censoring. Right-hand censoring is relevant since the highest grade attained is not observed for currently enrolled children. The school attainment regression measures the probability of failure or drop-out at each grade. Thus, the negative coefficient of an explanatory variable means that it tends to reduce the probability of failure or drop-out and hence increase school attaimnent. Ifrrespective of model specification, the sex-composition of a family does not influence the school enrollment rate of either boys or girls, but does influence a girl's school attainment. A 10 percent increase in the proportion of male members in the family increases the probability of a girl dropping out of school by 20 percent. This may indicate that male children get priority in schooling over female children. A male head's education affects both boys' and girls' school participation. A female spouse's education, on the other hand, influences only the girl's school enrollment. However, the female's education rather than the male's is more important for a girl's school enrollment. A 10 percent increase in female education increases the probability of a girl's school enrollment by 3 percent, while a similar percentage increase in the male's education increases a girl's participation by only half of one percent. Although the male's education does not affect a girl's school attainment, it decreases the probability of a boy's dropping out of school and hence increases his school attainment. For example, a 10 percent increase in the male's education increases the probability of a boy's school enrollment by 1 percent and reduces the probability of his dropping out by 7 percent. Note that female education only influences a girl's enrollment but not her attainment. Land ownership increases the probability of a girl's school participation, while the value of farm assets increases it for a boy. However, none of these household income-proxy variables have any impact on the school attainment of any gender. The presence of a paved road increases a girl's probability of ever attending a school by 40 percent and reduces the probability of dropping out by 5 percent. Improved road conditions increase the probability of school participation of both boys and girls by 20 and 32 percent, respectively. An improved road reduces a boy's probability of dropping out by 36 percent. 2 The ertimts are not shown here. The reason for using nmth rather tn Arabic scores is that schooling has a greater influence on math skills. -18- The presence of a primary school in a community increases the probability of a boy's school attendance by 28 percent. Surprisingly, the school's presence does not influence a girl's school participation or the schooling attainment of either gender. once children are enrolled in school. Rural electrification increases the school attendance of both boys and girls, but its impact is more pronounced (by about 30 percent) for girls than for boys. Access to tap water and well water relative to pond or river water improves the school enrollment of both boys and girls quite substantially and, similarly to electrification, its impact is much larger for girls than for boys. Given school participation, sources of clean water do not affect girls' school attainments but do increase that of boys. Investments in irrigation and advanced technology such as new varieties of crops increase the school attendance of boys, but not that of girls. This may suggest that boys are more often employed in farming than girls and hence benefit more from these investments. Boy's enrollment increases with the number of grades offered in primary school. The proportion of female primary teachers is negatively correlated with the school attainment of boys.'3 Teachers' years of education is positively correlated with the school attainment of girls. The effects of a more aggregate measure of Oatent) school quality confirm that school quality increases the school attainments of both boys and girls and the school enrollment of boys. An F-test is conducted for the joint significance of supply-side and demand-side variables. Table 15 presents the results for models I and I."4 The test suggests that the school access variables such as the presence of different types of schools and a paved road as well as road conditions influence the school participation of both boys and girls and the school attainment of boys. Unobserved school quality makes its largest impact on the school attainment of boys and girls. Even when controlling for their role in latent school quality, the demand-side factors (electrification, irrigation, new crop varieties, tap and well water) influence the school participation of both boys and girls although not to the extent they influence school attainment, given that children are already enrolled in school. Nevertheless, the findings suggest that once children are enrolled in school, it is school quality that most significantly influences their school attainments. Since Oatent) school quality is found to increase both school attainment and participation, it is important for policy formulation to relate latent school quality to observed school and community characteristics.15 Table 16 shows the distribution of community and school characteristics in the highest and lowest similarly of the distribution of latent school quality. The table clearly shows that there are substantial differences in the attributes of a high quality school community with those of a low quality school community. In particular, a low quality school community has a lower percentage of paved roads and higher incidence of blocked roads. It also has a lower percentage of electrification, irrigation and availability of clean water. The community has fewer schools with a complete primary cycle, fewer '3 This may reflect some quality of female primary teachers which is unobserved. In fact, as table 11 suggests, female teachers are less trained than their male counterpart and fewer of them have capes. 4An F-test for the joint significance of three school quality variabla (according to model 11) showed no significace. '5The fixed-effects coefficients of the math teat (lawte school quality) is regressed on observed primary wchool and community characteristics (estimates are not shown here). This shows, for exampb, that better roads and irrigation and the preaene of a more complete primary cycle significantly improve the quality of a school. However, due to the small ample (less thn 100), we cannot precisely estimate these relationships. -19- primary schools in general and no secondary school. A low quality school also means teachers with less years of education, especially males, but a higher proportion of trained female teachers.16 Thus, the school system in rural Morocco suffers not only from poor access, but also poor quality. Because demand-side variables influence the school participation of rural children and since school participation is a serious problem in rural Morocco, these findings have implications for government investment policy. As noted earlier, the demand-side factors influence schooling outcomes through their effect on the marginal productivity of education and hence on the returns to education. Thus, we will now examine the returns to education and how they correlate with observed community and school characteristics. The rates of return to education depend on how labor markets reward the educated labor force. Table 17 presents the estimates of rates of return to various levels of schooling by gender and region. A semi-log wage regression with correction for the wage-labor market participation decision is estimated for salaried workers above the age of 20.17 The average rate of return to a year of education in Morocco is 11 percent. The returns to a year of education at the primary, secondary 1 and secondary 2 levels of schooling are 6, 13 and 22 percent, respectively. The return to education is higher for men than for women at all levels except the secondary 1 level. Returns to education at all levels are higher in urban than in rural areas since there are more opportunities in urban areas. For example, the returns to primary and higher secondary education are almost zero in rural areas as compared to 9 and 23 percent, respectively, in urban areas. There also exist sharp regional differences in rates of return to education in Morocco. On average, the rate of return to education is lowest in the South (9 percent) and highest in the Center-North region (14 percent). The returns to education are again estimated at the community level (not shown). The highest and lowest similarly of community-level rates of return to education are then compared with the distribution of observed community and school characteristics (table 18). In particular, this table confirms that a community with low rates of return to education is characterized by fewer paved roads and primary schools and the absence of electricity, clean water and a secondary school. A low return community has a primary school that is relatively new, and has a higher proportion of trained female teachers and teachers with more years of education.1" Consequently, the low return area is also characterized by low school enrollment and attainment among children aged 7-20. 16 The presence of more trained female teachers, relatively older schools and lower average class size in a low quality school requires further investigation. We were informed that female recruitment is a recent phenomenon in Morocco's school system. Female teachers are young and better trained and are located in rural areas. This explains why we see more tained female teachers in a low quality school community. Because they are relatively new recruits, their impact has not yet been felt in improving school quality. However, the quality of relatively older schools has been improved. On the other hand, it is likely that a low quality school is located in a remote area where the population is smafler, resulting in lower student-teacher ratios in low quality schools. Moreover, better quality schools attract more students. " The estimates are not shown here. A multinomial logit function is first estimated with a choice of whether to work, whether to be self-employed and whether to work for a wage. The instruments for identifying the wage decision from the participation decision are various categories of household assets and unesaned income. The estimates are then used to calculate the participation bias factor-lambda-which is then used in the log-wage regression. i Perhaps this suggests that educational resources are not effectively utilized in rural areas. It also confirms our earlier conjecture that young teachers (male and female) with higher education are being placed in rural areas, but that they have not yet been effective in improving school outcomes. -20- VI. Policy Simulations The results confirm that rural Morocco suffers from poor access to and low quality of schools, inefficient allocation of educational resources as well as underinvestment in physical infrastructure. Thus, improving access and school quality will help improve schooling outcomes. Similarly, investment in rural infrastructure will also improve opportunities for the productive employment of educated rural people. In the same way, improved efficiency in educational resource allocation will help improve both the quality of the school system and the education the students receive. Based on the estimates of school participation and attainment, we performed policy simulations in order to suggest ways in which the government of Morocco could improve children's school performance, especially in rural areas. Table 19 presents three sets of policy simulations based on the three models estimated and presented in tables 13 and 14. The first set simulates the impact of ensuring that every community has a primary school, access to a paved road and so on. The overall enrollment rate in the sample is 0.49. This rate increases by more than 10 percent to 0.56 if all communities have access to a paved road. At the means of all variables (according to model 1), 68 percent of boys and 29 percent of girls ever attend school. This percentage jumps to 71 for boys and 39 for girls if every community has access to a paved road. The proportionate gain from investment in paved roads is higher for girls than for boys. Similarly, boys would gain more than girls in school enrollment if every community had a primary or secondary school although the increase is not large. The gains are more significant for girls if every community had electricity and clean water. The enrollment rate of girls almost doubles, from 0.29 to 0.51, following a complete electrification of rural communities. The gain is almost as dramatic if running water becomes available to all. In contrast, the gains are larger for boys if every rural community has irrigation and new crop varieties. The gains are relatively larger for changes in demand-side policy variables than for changes in supply-side variables. Repeating the same simulations with the school attainment regressions suggest that not all the policy changes that will induce higher enrollment rates will also lead to more schooling attainment. For example easier access (in terms of better roads or shorter distances) to primary schools induces higher enrollment rates, but does not induce children to remain in school longer. Conversely, access to a local middle school has a modest impact on school enrollment, but has a great effect on school attainment: it increases the mean schooling by almost 20 percent; more for boys (from 5.1 to 5.8 years) than for girls (from 4.9 to 5.4 years). Another variable that has a large effect on attainment is school quality as measured by our latent school fixed effect variable. Raising the quality of rural schools to that of the best school in the sample leads to an increase of one full year of schooling completed. This gain does not vary by gender, being almost as high for girls as for boys. On the other hand, the effect of school quality on school enrollment is only evident for boys: the simulation in table 19 suggests that enrollment of boys would be up by more than 10 percent if the quality of all schools were equal to the top school in the sample. Thus, the government could potentially raise attendance rates to 97 percent for boys and 82 percent for girls and increase years of schooling by 5 years for boys and 7 years for girls by increasing investment in rural Morocco. Even with all possible policy interventions considered in this paper, we find that gender bias still persists, especially, in the school enrollment of rural Morocco. This is because, even if government policies are gender neutral, parents seem to direct private investment toward more for boys than for girls, -21- at least in enrolling boys and girls in a school. Parents may have different reasons for allocating resources in this fashion (see Herz and others, 1991 for details). However, our findings suggest that parents incur higher schooling costs for girls than for boys. As table 20 shows, parents in Morocco need to spend more for girls than for boys at each level of schooling. The figures are for the country-level average of schooling costs collected as part of household expenditure survey. Note, however, that rural parents spend less on girls education than on boys. In contrast, in urban areas where gender bias is substantially low, parents spend more on girls than on boys. Therefore, if there is no gender bias, girls are more cosdy to educate than boys. Therefore, if a family has both a boy and a girl and if the schooling costs are substantial relative to family income, parents may decide to send the boy rather than the girl to school. It follows that if social returns are high for female education, as many studies (e.g., Subbarao and Raney 1993) suggest, then the government may find it justified to subsidize female education. VII. Conclusions Unlike many developing countries which do not allocate sufficient resources to education due to budgetary constraints, Morocco spends about 7 percent of its GNP on education. Yet its educational attainments in terms of school enrollment and attainment are low compared to an average middle-income country or even an average low-income country. The problem is partly due to Morocco's inadequate resource allocation to education in rural areas and partly to an inefficient resource allocation to education. To taclde these problems, Morocco has invested in educational reforms in recent years that were geared to improve both the access to and quality of primary and middle schools, especially in rural areas. This report has examined whether this supply-oriented approach is adequate to improve school attainment in Morocco. The report's premise is that both demand- and supply-side factors are critical to school attainment of Morocco. In other words, demand-side factors such as investment in rural infrastructure are as important as supply-side factors such as investment in school infrastructure. Given the household demand determined by parental education and income, the local market demand for skills generated through public investments in electrification, roads, irrigation and high-yielding varieties of crops, influences the demand for schooling. Public investment in the provision of pure drinking water influences schooling outcomes by improving the efficiency of families' resource allocation to education. Investment in safe drinking water supplies is especially important in rural areas, where lack of pure drinking water constitutes a health hazard and forces rural families to cut back resource allocation to education. This reduction in resources may disproportionately affect female children who are expected to care for sick members of the family. It is hypothesized, therefore, that investment in drinking water provision affects the school enrollment and attainment of girls to a greater extent than that of boys. The importance of public investment derives from its influence on returns to education, school quality and efficiency in the resource allocation to education. Improved school quality in turn increases returns to education. Thus, returns to education are not fixed and investments directed at promoting the access to and quality of schools and the infrastructure jointly determine both the levels of school attainment and returns to education. The descriptive analysis of MLSS data confirms that educational investment in Morocco is biased toward the wealthier urban population. Neither are educational outcomes gender-neutral, especially in rural areas. The school enrollment rate for the upper income group is almost twice that of the low income group. Urban children and rural boys consistently outperform their respective rural and female -22- counterparts in both school attendance and attainment. The school attendance rate is 90 percent in urban areas compared to only 48 percent in rural areas. The rate is 64 percent for rural boys compared to 32 percent for rural girls. Among primary school students the completion rate is 60 percent for rural children compared to 87 percent for urban children. In contrast, the rate is 56 percent for rural girls compared to 63 percent for rural boys. Rural boys and urban girls outperform rural girls in math and other cognitive tests. The performance of rural girls in cognitive achievements, however, improves significantly as their educational attainment increases. Insufficient investment in education and its effect on the gender gap is thus more pronounced in rural than in urban areas. In terms of supply-side factors, we find that the presence of a school or a paved road plays an important role. For example, only 56 percent of rural boys and 24 percent of rural girls ever attend school if there is no primary school in the community. The school attendance rate increases to 72 percent for boys and 37 percent for girls if there is a primary school in the community. Similarly, 56 percent of rural boys and 44 percent of rural girls go beyond primary school when there is no secondary school in the community. In contrast, this rate increases to 83 percent for boys and 75 percent for girls if there is a secondary school in the community. The presence of a paved road increases school participation for both boys and girls. When there is no paved road in a community, the school attendance rate is 21 percent for rural girls and 58 percent for rural boys. In contrast, the rate increases to 48 percent for rural girls and 76 percent for rural boys if there is a paved road in the community. The presence of a secondary school or paved road also decreases the drop-out rate and thus increases school attainment for both boys and girls. However, the survival rate is lower for girls than for boys. This finding suggests that public investment in additional secondary schools or roads in rural areas may help reduce overall rural-urban differences in school attainment but not necessarily gender differences in schooling outcomes. However, the major gender difference in rural Morocco is in school enrollment where the rate for girls is only half that for boys. Evidently, when girls do enroll in school, they perform well and, as a result, school attainment hardly varies by gender. Our econometric analysis suggests that poor access to and quality of schools as well as inefficient resource allocation are important factors in the rural-urban and gender differences in school outcomes. The results also indicate that public investment in roads, irrigation, new varieties of crops and water supply improve schooling outcomes. This is because public investment increases the return to education as well as improving school access and quality and hence school attendance and attainment. The benefits of these public investments are, however, not gender-neutral. For example, investment in roads, electrification and safe drinking water improve girls' outcomes more than boys', while the opposite is true for investment in irrigation and new crop varieties. Since public investment is gender-neutral, parents seem to bias private investment toward male children. The Moroccan family clearly displays a preference for the education of boys over girls. The more education the father has, the more the boy will receive; the more education the mother has, the more the girl will receive. Overall the presence of an additional male member in the family reduces the education girls receive. The findings call for more government effort to improve efficiency in resource allocation to education. They also call for more public investment in rural education and infrastructure that are required to improve overall educational attainments. Government has an important role to play in influencing parental investment in education. Although building more higher quality schools in rural areas will help promote schooling outcomes, this is not sufficient. Public investment in roads, electricity, irrigation, new high-yielding crop varieties and pure drinking water supplies are the types of investment in rural infrastructure that clearly increase the demand for education. However, as our findings suggest, -23- public investments do not necessarily reduce the gender bias in school enrolment; parents seem to bias private investment toward boys rather than girls. In such a case, if, as different studies show, the social returns to female education are high then the government of Morocco may introduce stipend for girls to offset parental bias against female education. The Bank-financed female scholarship project in Bangladesh shows how the government can help improve female education in a country that strongly favors boys over girls. -24- BIBLIOGRAPHY Grossman, M. 1972. 'On the Concept of Health Capital and the Demand for Health." Joumnalof Political Economy, vol. 80 (March-April), pp. 223-55. Haddad, W., M. Carnoy, R. Rinaldi and 0. Regel. 1990. "Education and Development: Evidence for New Priorities." World Bank Discussion Papers 95. Herz, Barbara, Subbarao, Kalanidhi, Habib, Masooma and Raney, Laura. 1991. 'Letting Girls Learn: Promising Approaches in Primary and Secondary Education." World Bank Discussion Papers 133. Khandker, Shahidur R. 1993. "Schooling Outcomes and School Efficiency in Rural Bangkadesh". (Mimeo). World Bank, Washington, D.C. Lavy, Victor, J. Spratt and N. Leboucher. 1993. 'Incidence, Patterns of Change and Correlates of Illiteracy in Morocco, 1991." Mimeo, World Bank, Washington, D.C. Psacharopoulos, George. 1981. "Returns to Education: An Updated International Comparison." Comparative Education, 17(3) (1981): pp. 321-41. Psacharopoulos, George. 1984. "The Contribution of Education to Economic Growth: International Comparisons." International Comparisons of Productivity and Causes of the Slowdown (ed. John Kendrik) (Cambridge: American Enterprise Institute/Ballinger Publishing Company, 1984), pp.335-60. Schultz, T. W. 1961. 'Investment in Human Capital." American Economic Review. March 1961. Singh, I., L. Squire and J. Strauss. 1986. Agricultural Household Models: Extensions. Apolications. and Policy, Baltimore, Johns Hopkins University Press. Subbarao, Kalanidhi and Laura Raney. 1993. "Social Gains from Female Education: A Cross- National Study." World Bank Discussion Paper 194. Washington, D.C. UNESCO. 1991. World Education Report 1991. UNESCO. Verspoor, Adriaan. 1990. "Educational Development: Priorities for the Nineties." Finance and Development. vol. 27 (March), pp. 20-23. World Bank, 1989. Rural Primary Education Project, Report 7434-MOR. World Bank, 1991. Rural Basic Education Development Project, Report 8936-MOR. -25- STATISTICAL APPENDIX -27- Table 1: Major Indicators of Economic and Social Development in Morocco, 1990 Average for lower middle Average for Indicators Morocco income countries low income countries Per capita income US$ 950 USS 1,530 $ 360 Life expectancy at birth (years) 62 65 63 Population growth rate 2.6 2.2 2.0 Total fertility rate 4.5 4.0 3.8 Infant mortality rate (per 1000 live births) 67 51 69 Adult literacy rate (percent) 49 75 60 Adult female literacy (percent) 38 68 48 Primary school enrollment (net) 55 86 68 Female enrollment as percentage of male primary (gross) enrollment 65 89 78 Secondary school (gross) enrollment 36 54 36 Female enrollment as percentage of male secondary (gross) enrollment 68 109 64 Source: World Development Report, 1992 Note: The school enrollment figures are for 1989. -28- Table 2 Population Breakdown in LSMS Sample (Children 7-20) Rural Urban Regions Male Female Male Female Total South 236 279 191 187 893 26.43 31.24 21.39 20.94 100.00 Tensift 284 295 177 215 971 29.25 30.38 18.23 22.14 100.00 Center 1 298 296 155 170 919 32.43 32.21 16.87 18.50 100.00 North-West 250 274 242 232 998 25.05 27.45 24.25 23.25 100.00 Central-North 317 300 205 223 1045 30.33 28.71 19.62 21.34 100.00 Oriental 284 276 251 224 1035 27.44 26.67 24.25 21.64 100.00 Central-South 285 258 199 198 940 30.32 27.45 21.17 21.06 100.00 All Regions 1954 1978 1420 1449 6801 28.73 29.08 20.88 21.31 100.00 -29- Table 3 Proportion Ever Went to School (Children Ages 7-20) Rund Urban Region Male Female Male Female Total South 167 70 180 156 573 70.76 25.09 94.24 83.42 64.17 Tensift 142 84 164 183 573 50.00 28.47 92.66 85.12 59.01 Center 201 92 152 148 593 67.45 31.08 98.06 87.06 64.53 North-Wet 163 99 226 195 683 65.20 36.13 93.39 84.05 68.44 Centrl-North 186 69 199 183 637 58.68 23.00 97.07 82.06 60.96 Oriental 187 122 235 202 746 65.85 44.20 93.63 90.18 72.08 Centnl-South 212 97 192 176 677 74.39 37.60 96.48 88.89 72.02 Ever attended 1258 633 1348 1243 4482 al regions 64.38 32.00 94.93 85.78 65.90 Never Attended 696 1345 72 206 2319 35.62 68.00 5.07 14.22 34.10 TOTAL 1954 1978 1429 1449 6801 100.00 100.00 100.00 100.00 100.00 -30- Table 4 Mean Last Grade Attended For Those Who Ever Went to School Rural Male Rural Female Urban Male Urban Female Variable Mean St. Dev. Mean St. Dev. Mean St. Dev. Mean St. Dev. South 5.62 (2.85) 4.71 (2.29) 7.08 (3.49) 6.28 (3.0) Tensift 5.38 (2.77) 5.61 (2.76) 6.65 (3.34) 7.16 (3.3) Center 5.57 (2.52) 5.41 (2.65) 7.13 (3.35) 6.66 (3.5) North-West 5.09 (2.45) 4.83 (2.33) 7.27 (3.58) 7.38 (3.8) Central-North 5.62 (2.66) 5.13 (2.91) 7.26 (3.58) 7.30 (3.6) Oriental 5.70 (2.49) 4.89 (2.23) 7.29 (3.63) 6.57 (3.5) Central-South 5.04 (2.18) 4.03 (1.64) 7.15 (3.36) 7.27 (3.5) All Regions 5.43 (2.56) 4.93 (2.44) 7.14 (3.49) 6.96 (3.53) -31- Table 5 Profile of School Experience of Moroccan Children at Different Levels Rural Boys Rural Girls Urban Boys Urban Girls No. Percent No. Percent No. Percent No. Percent 100 100 100 100 Did not go to primary 36 36 68 68 5 5 14 14 Went to primary school 64 64 32 32 95 95 85 85 Did not complete primary 24 37 14 44 11 12 13 15 Completed primary school 40 63 18 56 84 88 72 85 Did not go to secondary 1 17 43 10 55 12 14 11 15 Went to secondary 1 23 57 8 45 72 86 61 85 Did not complete secondary 1 15 64 5 66 15 21 11 18 Completed secondary 1 8 36 3 34 57 79 50 82 Did not go to secondary 2 3 33 1 47 10 17 11 22 Went to secondary 2 5 67 2 53 47 83 39 78 Did not complete secondary 2 4 86 0 0 8 18 15 38 Completed secondary 2 1 14 2 100 39 82 24 62 -32- Table 6a Distribution of Children by Household Per Capita Expenditure Quintiles Rura Urban Per Capita expenditure Male Female Male Female Total Observations Quintiles Percent Quint 1 50.20 49.39 17.89 18.63 36.49 2482 Quint 2 27.89 27.45 26.13 24.22 26.61 1810 Quint 3 14.07 15.32 27.96 27.88 20.28 1379 Quint 4 6.60 6.17 19.65 18.98 11.84 805 Quint 5 1.24 1.67 8.37 10.28 4.78 325 TOTAL (Percent) 100.00 100.00 100.00 100.00 100.00 No. of observation 1954 1978 1420 1449 6801 Table 6b Distribution of Children Ever Beeu to School by Expenditure Quintiles Rural Urban All Regions Male Female Male Female Total Quint 1 981 977 254 270 2482 59.33 24.67 88.58 81.85 51.13 Quint 2 545 543 371 351 1801 66.97 36.65 93.53 88.87 67.57 Quint 3 275 303 397 404 1379 71.63 42.57 97.22 88.61 77.59 Quint 4 129 122 279 275 805 75.97 40.16 97.13 88.36 82.11 Quint S 24 33 119 149 325 66.67 45.45 100 73.15 79.69 -33- Table 7 Distribution of Children by Presence of a School Presence of a School Rural Male Rural Female Totad A. Presence of a primary school No 818 782 1600 (45.3) (42.4) (43.8) Yes 988 1064 2052 (54.7) (57.6) (56.2) B. Presence of a secondary school No 1752 1787 3539 (97.0) (96.8) (96.9) Yes 54 59 113 (3.0) (3.2) (3.1) Total number of children 1806 1846 3652 (100.0) (100.0) (100.0) Note: Percentages are in parentheses. Table 8 Percentage of Children Ever Attend School by Presence of a School Ever Attended a School Presence of a School Rural Male Rural Female Total A. Presence of a primary school No 56.1 24.3 40.6 (818) (782) (1600) Yes 71.9 36.7 53.6 (988) (1064) (2052) Total 64.7 31.4 47.9 (1806) (1846) (3652) B. Presence of a secondary school Ever Went Beyond Primary School No 55.8 43.9 52.1 (344) 9157) (501) Yes 83.3 75.0 80.8 (18) (8) (26) Total 57.2 46.5 53.5 (362) (165) (527) Note: Percentages are in parentheses -34- Table 9 Mean Grade Attained by Presence of a School Mean Grade Atained by Pfeaice of a School Presaee of a School Rund Male Rural Female Total Obs. Mean (S.D.) Obs. Mean (S.D.) Obs. Mean (S.D.) A. Primary school No 454 5.14 (2.39) 190 5.00 (2.57) 644 5.10 (2.45) Yes 676 5.54 (2.61) 367 4.89 (2.34) 1093 5.31 (2.53) Tota 1130 5.38 (2.53) 557 4.93 (2.42) 1687 5.23 (2.50) on PYesmece of primary school Mean grade aftained by those who went to beyond pnrmary school but no scondary school No 5 6.40 (3.05) 0 0 0 5 6.40 (3.05) yes 34 6.65 (3.06) 23 5.78 (3.44) 57 6.30 (3.22) Total 39 6.62 (3.02) 23 5.78 (3.44) 62 6.31 (3.18) PreYaue of a scondary school No 255 9.13 (1.78) 93 9.11 (1.89) 348 9.12 (1.80) Yea 19 9.21 (1.62) 6 11.17 (0.41) 25 9.68 (1.65) Totl 274 9.14 (1.76) 99 9.23 (1.89) 373 9.16 (1.80) Table 10 School Performances of Children by Whether Paved Road is Accessible to Community Ever been in school Paved road to the community Rural Male Rural Female Total No 58.27 21.25 39.64 (1136) (1157) (2293) Yes 75.67 48.33 61.81 (670) (689) (1359) Total 64.73 31.42 47.89 (1806) (1846) (3652) Mean grade attained No 58.27 21.35 39.64 (662) (247) (909) Yes 5.59 5.18 5.43 (507) (333) (840) Total 5.42 4.96 5.27 (1169) (580) (1749) Note: Number of observations are in parentheses. -36- Table 11 Mean School Characteristics Characteristics Observation Mean Standard Deviation Distance to primary school (Km) 240 0.05 0.29 Mode of transport to primary school (= 1 if foot) 240 0.42 0.49 Time spent to go to school 240 49.10 70.59 Type of school (1 = mother, 0 = satellite) 240 0.71 1.73 Age of primary school 218 24.05 17.97 Is primary school co-educated? 240 0.71 0.46 Books bought by students from primary school 240 0.77 0.42 Books bought from primary school 240 0.03 0.17 Books free from primary school 240 0.01 0.11 Highest class offered in primary school 218 5.27 1.44 Do all children get in if applied 218 0.89 0.31 Proportion of female teacher in primary school 218 0.19 0.22 Proportion of female primary teachers with capes 240 0.56 0.49 Proportion of male primary teachers with capes 240 0.93 0.23 Proportion of all teachers with capes 240 0.93 0.20 Proportion of male teachers with training 214 0.87 0.23 Proportion of female teachers with training 214 0.51 0.49 Mean education of primary teachers 216 9.37 4.25 Average size of classes in primary school 218 20.72 8.91 Failure rate in primary school 215 0.13 0.07 Proportion of primary students who completed last class 196 0.14 0.07 Proportion of students who completed last class of primary go to secondary level 177 0.67 0.36 -37- Table 12 Descriptive Statistics (Sample of children aged 7 - 20) Variable No. of Mean Standard observations Deviation Proportion of female children 3652 .495 .500 School participation 3652 .479 .499 School attainment 3652 2.922 2.621 Age of the child 3652 13.091 3.907 Proportion of male members in the household 3652 .487 .160 Education (in years) of the head 3652 1.514 2.040 Education (in years) of the spouse 3188 1.092 .713 Land under cultivation 3652 5.188 22.275 Value of farm assets x 1000 3652 2.091 10.900 South 3652 .135 .342 Tensift 3652 .152 .359 Center 1 3652 .158 .365 North-West 3652 .124 .330 Central-North 3652 .153 .360 Oriental 3652 .142 .349 Central-South 3652 .136 .343 Paved road in the community 3652 .372 .483 No. of months road is blocked 3652 .096 .473 Is there any mAsid or Koranic school 3652 .789 .408 Is there any primary school? 3652 .562 .496 Is there any secondary school? 3652 .031 .173 Is there electricity? 3644 .163 .370 Source of drinking water - pipe? 3652 .108 .310 Source of drinking water - well? 3652 .494 .500 Some land is irrigated? 3490 .423 .494 New crops in the past 10 years? 3565 .347 .476 Highest class in community primary school 2615 5.309 1.352 Proportion of female teachers 2615 .182 .207 Mean years of education of teachers 2610 9.093 4.224 -38- Table 13 Determinants of School Participation of Rural Children in Morocco Model I Model II Model Im Explanatory Variables Boys Girls Boys Girls Boys Girls Proportion of male members -0.093 -0.428 -0.069 -0.379 -0.225 -0.049 (0.352) (1.577) (0.223) (1.212) (0.677) (0.149) Education (in years) of the male head .078 .048 .065 .052 .084 .037 (3.582) (2.546) (2.559) (2.465) (2.977) (1.701) Education (in years) of the female spouse .054 .297 .101 .284 .135 .306 (0.731) (4.623) (0.995) (3.726) (1.258) (4.087) Land under cultivation .001 .003 .002 .002 .002 .003 (0.373) (1.952) (1.084) (0.616) (1.099) (1.946) Value of farm assets x 1000 .006 .002 .005 .002 .017 .005 (1.989) (0.651) (1.261) (0.536) (2.389) (1.098) Paved road in the community .134 .444 .032 .407 .020 .351 (1.542) (4.971) (0.308) (3.831) (0.174) (3.026) No. of months road is blocked -0.168 -0.340 -0.227 -0.301 -0.198 -0.327 (2.137) (3.330) (1.946) (2.009) (1.781) (2.193) Is there any mAsid or Koranic school? .029 -0.070 .092 -0.028 .241 .099 (0.326) (0.753) (0.816) (0.247) (2.078) (0.815) Is there any primary school? .284 .006 .106 -0.185 .191 .054 (3.734) (0.076) (0.977) (1.583) (1.790) (0.467) Is there any secondary school? .211 .007 .621 .200 .312 .165 (0.894) (0.032) (1.974) (0.699) (1.024) (0.619) Is there electricity? .453 .698 .494 .701 .480 .824 (3.664) (6.340) (3.522) (5.475) (2.739) (5.388) Source of drinking water - pipe? .414 .555 .365 .607 .393 .418 (2.163) (3.500) (1.652) (3.242) (1.647) (2.140) Source of drinking water - well .078 .443 .048 .528 .010 .445 (1.013) (5.137) (0.479) (4.904) (0.090) (3.813) Some land is irrigated? .181 .047 .154 .034 .337 .183 (2.276) (0.556) (1.511) (0.325) (3.172) (1.742) -39- Table 13 (continued) Model I Model II Model m Explanatory Variables Boys Girls Boys Girls Boys Girls New crops in the past 10 years? .325 .085 .237 .069 .101 -0.045 (3.154) (0.844) (1.940) (0.594) (0.771) (0.361) Highest class in primary school .074 -0.024 (2.075) (0.645) Proportion of female teachers .104 .067 (0.450) (0.282) Mean education (years) of teachers -0.001 -0.011 (0.077) (0.501) Latent school quality 0.025 .0.005 (2.101) (0.426) Chi-square 184.94 335.03 131.60 235.38 154.71 238.07 Number of observations 1490 1517 1089 1094 969 963 Note: t-statistics are in puetheses. -40- Table 14 Detenninants of School Attainment of Rural Children in Morocco Model I Model 11 Model Im Boys Girls Boys Girls Boys Girls Proportion of male members .602 1.183 .623 1.264 .571 2.004 (1.428) (2.175) (1.288) (1.913) (1.175) (2.916) Education (in years) of the male head -0.735 -0.018 -0.065 -0.030 -0.071 -0.040 (2.303) (0.536) (1.817) (0.713) (1.898) (0.985) Education (in years) of the female spouse -0.031 -0.006 -0.053 -0.022 -0.021 .030 (0.353) (0.091) (0.511) (0.263) (0.222) (0.412) Land under cultivation -0.001 -0.001 -0.001 .001 -0.001 -0.002 (0.437) (0.426) (0.155) (0.197) (0.379) (0.839) Value of farm assets x 1000 .005 .008 -0.009 -0.000 -0.004 .004 (0.987) (0.118) (1.380) (1.079) (0.825) (0.506) Paved road in the community -0.089 -0.178 -0.147 -0.480 .168 -0.149 (0.691) (0.968) (0.964) (2.031) (1.005) (0.574) No. of months road is blocked .432 -0.199 .313 -11.676 -0.365 -0.343 (3.919) (0.543) (1.835) (0.000) (2.357) (0.858) Is there any m'sid or Koranic school? .122 .002 .125 .273 .230 .179 (0.896) (0.008) (0.726) (0.992) (1.355) (0.658) Is there any primary school? -0.148 -0.122 .089 .064 -0.172 .209 (1.248) (0.645) (0.542) (0.234) (1.051) (0.719) Is there any secondary school? -0.239 .146 -0.107 .655 -0.062 .373 (0.664) (0.318) (0.256) (1.036) (0.163) (0.742) Is there electricity? -0.234 .093 -0.093 .068 -0.121 .068 (1.417) (0.427) (0.502) (0.255) (0.588) (0.243) Source of drinking water - pipe? -0.280 .163 -0.719 .245 -0.639 -0.025 (1.095) (0.527) (2.265) (0.641) (2.094) (0.062) Source of drinking water - well -0.260 .036 -0.248 .209 -0.253 .026 (2.105) (0.178) (1.570) (0.840) (1.656) (0.101) Some land is irrigated? -0.079 -0.060 -0.057 -0.274 -0.047 -0.005 (0.649) (0.332) (0.376) (1.267) (0.311) (0.024) -41- Table 14 (continued) Model I Model I Model HI Boys Girls Boys Girls Boys Girls New crops in the past 10 years? .018 .180 .090 .138 -0.127 .111 (0.125) (0.840) (0.525) (0.567) (0.683) (0.410) Highest class in primary school -0.025 .088 (0.420) (0.991) Proportion of female teachers .675 -0.313 (1.972) (0.643) Mean education (years) of teachers -0.019 -0.086 (0.635) (1.722) Latent school quality -0.097 -0.125 (4.784) (4.425) Chi-square 119.89 39.00 93.03 37.39 131.99 56.36 No. of observations 968 479 750 368 671 347 Note: t-statistics are in parenileses -42- Table 15 F-test for Joint Siginifcance of School Access, Quality and Demand-Side Variables Model I Model III Policy Variable Male Female Male Female School access Participation 5.16 7.52 3.0 3.37 Attainment 3.76 0.42 2.04 0.47 School access + latent school quality Participation 4.26 2.82 Attainment 7.54 3.34 Demand-side factors Participation 8.17 18.60 5.65 12.68 Attainment 1.89 0.26 1.67 0.05 Note: School access includes the availability of paved road and different types of schools as well as road conditions, while the demand-side factors include electrification, irrigation, new crop varieties, piped and well water. 43- Table 16 Correlates of Latent School Quality Selected community and school U Lower docile characteristics sciool qualitv school qualitv Highest class 5.36 (1.50) 3.78 (2.11) Proportion of female teachers 0.09 (0.13) 0.24 (0.21) Mean years of teachers' education 9.79 (4.40) 9.37 (3.69) Distance to primary school (km) 0 0.33 (0.71) Is primary school co-educated? 0.87 (0.35) 0.67 (0.50) Do all children get in if applied 0.93 (0.27) 0.78 (0.44) Average size of the class in the primary school 23.07 14.92 (7.37) Average student-teacher ratio 25.60 (8.97) 15.39 (7.08) Failure rate in primary school 0.11 (0.05) 0.17 (0.12) Proportion of female teachers with training 0.36 (0.49) 0.57 (0.53) Proportion of male teachers with training 0.91 (0.23) 0.83 (0.22) Proportion of female teachers with capes 0.47 (0.52) 0.67 (0.50) Proportion of male teachers with capes 0.93 (0.26) 0.98 (0.07) Age of primary school 22.79 (22.62) 27.89 (16.65) Paved road 0.53 (0.52) 0.22 (0.41) Road blocked in months 0 0.47 (1.06) Primary school 0.60 (0.31) 0.40 (0.31) Secondary school 0.06 (0.24) 0 Electricity 0.13 (0.35) 0.07 (0.26) Water source - pipe 0.07 (0.26) 0 Some land irrigated 0.36 (0.49) 0.36 (0.50) No. of observations 15 9 Note: Standard deviations are in parentheses. -44- Table 17 Estimates of Returns to Education by Gender and Location Regions South Tensift Center North- Center- Center- West North Oriental South Country Primary All 0.04 0.07 0.09 0.06 0.11 0.01 0.06 0.06 (0.02) (0.02) (0.03) (0.02) (0.03) (0.03) (0.03) (0.03) Male 0.03 0.06 0.09 0.06 0.08 0.02 0.05 0.06 (0.02) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) Female 0.07 0.09 0.04 0.08 0.20 -0.29 0.12 0.04 (0.07) (0.05) (0.07) (0.07) (0.06) (0.11) (0.07) (0-07) Rural -0.01 0.03 0.03 -0.01 0.18 0.03 0.01 0.04 (0.04) (0.04) (0.04) (0.04) (0.05) (0.04) (0.04) (0.04) Urban 0.08 0.11 0.12 0.10 0.08 0.03 0.11 0.09 (0.03) (0.03) (0.03) (0.03) (0.03) (0.04) (0.04) (0.03) Secondary 1 level All 0.08 0.16 0.14 0.16 0.13 0.10 0.15 0.13 (0.05) (0.05) (0.05) (0.05) (0.05) (0.06) (0.05) (0.05) Male 0.08 0.11 0.11 0.13 0.09 0.05 0.15 0.10 (0.05) (0.06) (0.06) (0.06) (0.06) (0.06) (0.06) (0.06) Female -0.03 0.21 0.11 0.13 0.20 0.61 0.05 0.18 (0.15) (0.10) (0.13) (0.11) (0.13) (0.20) (0.13) (0.00) Rural 0.13 0.00 0.21 0.11 -0.19 0.11 0.12 0.07 (0.10) (0.10) (0.08) (0.08) (0.20) (0.09) (0.14) (0.00) Urban 0.08 0.18 0.11 0.12 0.17 0.11 0.12 0.13 (0.06) (0.06) (0.07) (0.07) (0.07) (0.08) (0.07) (0.07) Secondary 2 level All 0.17 0.20 0.19 0.17 0.27 0.27 0.25 0.22 (0.08) (0.08) (0.08) (0.08) (0.08) (0.09) (0.09) (0.08) Male 0.13 0.24 0.14 0.16 0.30 0.30 0.22 0.21 (0.08) (0.09) (0.09) (0.09) (0.09) (0.10) (0.09) (0.09) Female 0.30 0.07 0.32 0.15 0.10 -0.13 0.37 0.17 (0.22) (0.15) (0.19) (0.14) (0.17) (0.23) (0.21) (0.19) Rural 0.09 0.01 -0.19 0.14 0.24 0.26 0.43 0.14 (0.19) (0.21) (0.17) (0.16) (0.28) (0.20) (0.24) (0.21) Urban 0.18 0.21 0.21 0.15 0.29 0.32 0.25 0.23 (0.09) (0.10) (0.11) (0.11) (0.10) (0.13) (0.12) (0.11) All schooling level 0.09 0.12 0.12 0.11 0.14 0.10 0.13 0.11 (0.008) (0.007) (0.007) (0.007) (0.008) (0.009) (0.009) (0.003) Note: Standard errors are in parentheses -45- Table 18 Correlates of Retumns to Education (At the Community Level) Community and school characteristics Upper decile Lower decile of retuns of returns Highest class in primary cycle 5.45 (1.51) 5.75 (0.71) Proportion of primary female teachers 0.28 (0.22) 0.30 (0.09) Mean year of primary teachers' education 8.74 (4.39) 10.23 (3.89) Do all children get in primary school if applied 0.82 (0.40) 0.63 (0.52) Age of primary school 31.0 (16.0) 28.25 (19.24) Proportion of female primary teachers with capes 0.71 (0.46) 0.98 (0.05) Proportion of male primary teachers with capes 0.92 (0.25) 1.0 (0.) Proportion of female teachers with training 0.82 (0.33) 0.98 (0.05) Proportion of male teachers with training 0.93 (0.15) 0.87 (0.15) Paved road 0.64 (0.50) 0.40 (0.52) Presence of a primary school 0.73 (0.47) 0.20 (0.42) Presence of a secondary school 0.09 (0.30) 0 Electricity 0.27 (0.47) 0 Piped water 0.18 (.040) 0 Some land irrigated 0.30 (0.48) 0.30 (0.48) New crops 0.46 (0.52) 0.44 (0.53) School participation rate 0.61 (.30) 0.41 (0.28) School attainment 3.17 (1.69) 2.42 (0.95) No. of observations 11 10 Note: Standard deviations are in parentheses. Table 19 Improved School Performance for Changes in Selected Policy Variables (Simulation) Enrollment School Attainment Boys Girls All Boys Girls All At means of all variables (Model 1) 0.68 0.29 0.49 5.15 4.94 5.08 Paved road for all communities 0.71 0.39 0.56 5.20 5.04 5.15 Roads are not blocked 0.68 0.30 0.50 5.20 4.91 5.12 Koranic school for all comnmunities 0.68 0.28 0.48 5.14 4.94 5.06 Primary school for aU communities 0.72 0.29 0.50 5.24 5.00 5.15 Secondary school for all communities 0.75 0.29 0.52 5.81 5.42 5.68 Electrification for all communities 0.80 0.51 0.66 5.53 4.77 5.19 Piped water for all communities 0.80 0.47 0.62 5.21 4.82 5.00 Well water for all communities 0.69 0.37 0.53 5.27 4.96 5.15 Irrigation for all communities 0.71 0.30 0.51 5.23 5.04 5.19 New crops made available for all communities 0.75 0.31 0.53 5.24 4.89 5.11 Changes in all except Koranic school and well water 0.97 0.82 0.89 6.58 5.33 6.03 Changes in all policy variables 0.97 0.87 0.91 6.70 5.35 6.08 At means of all variables (Model 1I) 0.71 0.29 0.52 5.17 4.87 5.08 All communities with 6-years primary school 0.72 0.29 0.52 5.23 4.90 5.12 For changes in all variables 0.99 0.90 0.95 6.56 4.98 5.93 At means with latent school quality (Model III) 0.71 0.33 0.52 5.20 5.04 5.14 With highest quality for all schools 0.79 0.31 0.58 6.16 5.91 6.08 Changes in all variables 0.99 0.89 0.94 7.49 6.09 6.86 47- Figure 1 Percentage Ever Attended School 100 - 00 75! 75- Ca 54X In U V E a: - 50 co - - 50~~~~~~~~~~~~~~~~~~~~~~~~s 4-.... 433 40X- i 25- 203 0-c _c_ .--. 03 . FWI r iX~n ...OX. 0% 0X 0X ON iOC.<=20 30C.<=40 10<,c=60 70<,<-eo ħ0''-20 50<.~~~~~~~~~~~~~4 9 . ........ .. , rur. mal rur. fem urb. mal urb. fem Figure 3 Percent Still In School by Age o rural male + rural female urban male a urban female I00 - A 75- 02 Ca 0 U)~5 25 8 ~10 12 14 16 18 20 age Figure 4a Kaplan-Meier Survival Curves: All Regions Group 1 rural male Group 2 rural female Group 3 urban male Group 4 urban female 1.00 - >. 0.75- 41-f 0 cm 0.50- X 0.25 - 0.00 - 0 4 8 12 16 Last Class Attended Figure 4b Kaplan-Meier Survival Curves by Region Group 1: rural male Group 2: rural female Group 3: urban male Group 4: urban female r 00 -3 £00 - G.50 L 0.50- a.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~5 L 0.21 - 2 0.25- 22 0.o0 - 0t00 - Lant Closm Attended Last ClaIs Attended u' South Tmnsift 1.00 -1.00- D0OD - O4Do- 0 t0.50- I 05 0.21- 0.21- Last Class Attended List ClIse Attendad Center North-Wast Figure 4b (Cont'd.) Kaplan-Meier Survival Curves by Region Group i: rural male Group 2: rural female Group 3: urban male Group 4: urban female 0.7r X 7r 0.50 - 2 .D0 -^tjt 0.28- 0.28- 2 0.00 I 0.00 I Lost Class Attsndsd Lost Cliis Attended Center-North Oriental 1.00- 0.73 kL 0.50- 2 Lost Class Attended Center-South Figure 5 Completion Rate by Schooling Level ioo BB.OX 1OO K 85.3X 82.01; 78.5% 7! 7! 75 - W62.9% Ul~~~ ru. ol ru, 1n ub u r,fnrr ul rr r r.uluU u U L... _ .24 gg 0 .... _ soU- : , g ,.g.g ,g., g~~~~~~~~~~~~. . . . . . . ., N 2! 2!-1_[1 51 Ut r~ur'. rnul rur. fern urr, nm urD, f.m rur'. mel rur'. tim u... ... ... Seconary 2cDlScodr School .. .. .. . 100~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~........ M.~~~~~~~~~8 AK 4 2 5 53271...... - 50~~~~~~~~~~~~~~~~~~~......... ..... ...............x . smm~.. U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..... ... .... r'ur. inal Iur. loin urS. mul urb, f min rr a u.fm ub e .fr SPcinary 2SchDooleodryiSho Figure 6 Transition Rates by Location-Gender Group 65.0% 83.:X 7B.1% 777~~~~~~~~~~~~~~~7A 7 66.7% m ~~~~~~~~~~~~~~~~~~~~~~~53.3% C.. A~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~AKi C.~~~~~~~~~~~~~~~~~~~~~~~.. ...... rur. n.l rur. {un urt. rul urb. fur rur, rut rur. {r urb. nd urD. fun T s RL .... .. 0. C.~~~~~~~~~~~~~~~~~~~~...... .. 25 ~~~~~~~~~~.......2 5 ...... rur. mal nur. fRim Urt. inul urb, fain rur, Mal rur. lin ur'b. Nil urn. IBM~ Transition Rate Primary --> Secondar'y i Transition Pate Secondary i --> Secondary 2 BuIEOOLJOS E 366 BunIoowos 36De w2i-jn lofl Lua.UfnU l~iwwiJ.fJ LUBqjfl Juw'qun w6e.UfnU tewdfnJ .. . ... . ..... .... C ..... ..... ...... .....~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ... ... ... .. . .. .. . . . . ... . . . . ...................~~o : O ..... ................ ........ ..... ..~ ~ ~ ~ ~ ~CJ ... ... .. ... ..... . . ... .. .. . . . .. . . . . . . . ........................ .... .........~ ~ ~ ~ ~ ~ ~ ~ ...... .. ... ...... ...... ...... ........ ....... ~ ~ ~ ~ ~ .C....... or o... X.: : ... ::..L J ... ... .. ......... .. ..1. . .. mm maaaa~~~= ...... ..... ...oa ..... ........ .......~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~, E E ' D ... ... .. ... .. ... .. . .. .. . .. . . : . . . . . . . .. . . . . .... .. . . . . . o;frT sas;~~~~~~~~~~~~~~~~~~~~........ . ... ... ... . ..... ..... .....~~~~~~~~~p..-C ..... . .... ..... ...... qC E :-mr:...OJOOS T.. :1....I42H. UOGH. 2/ GJnbTJ~~~~.... .. bu7jooU os Z oes Bu;ToOLOS T r as wai qjn leu-wqun uiaeunj jew unu waeyqjn jewuqjn wuse-nj le uun . ... .... .... ... .. ... _ ., .. - ..... . _ . . ,, ,,, _ . ,,,,,, .-- ........ . _ - o ..... -.. . ....... , , , , . , , - _ : . . .. . . - -.:: ........ .. .. .......... '.'.',''.''''''' ... ... ..... ...... .. ,,,,,,,.,,, .. ............, ...., .., . .. .o .. 0:.'':- . . -. ': .: .... . 0 0 ' E § - 5i n :: ~~~~~~~~~~~~..:.::: .: . . ..:. . .. ..: .. .... -:--..-...... ..:.. .-.-..: .-. .... ... . . ........... .. . T t' S TO D E X; 5 5 9 ; E 5 5-9 _OR - OE . ... ... ... ... ... ... , ,.,. _ ,.,,.,,-,,,.,.,-,,,,,,,,,,,,,,,. _ .......... = _ -o ............... ... ............ .... . .. .... . .. . . . . -.- : . -:- :-.-: -. :-:: . :- ... - .. -.:. -. .- - -- .... - : : -: .......... . ... ..... ... ... .... .... ... .... ... .. - ... - ..-:..- ... . . . .-.. . . . - - - : :- ... ,: -:- :-.-: . : : ........ ... - -.-.:.- .... ,,,,,:,,,,,,,,,,,,:,,,-,, .~~~. . . . . . . . . . . . . . . . . . . . . . . . . : . -. .-- .-.: :-:- ....... - : ,,,,,,::,, ., .- -.-,,, .................... .. . . .. ... .... .... .. . .... .... ... .... ... . . . . .... . . : , . , :,... .. : , ... .. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.,:. --:. .: .-::.... . :.- ........ .. .............. : ..-... : :.. : :..:.............. ... ........ , - - ................ ... . - - - ~~~~~~~~~~~~~~~~~.. .. . .. .. . .. .. . ... .. . .. .. . .. .. . X X .. .. ... . .. .. . - . ... .. ...... .. ......... ...... . . .... ,~~~~~~~~~.. .. ....... . . .... ........ ................ ' . - ' : ':: . :'. . : : . . ' -' - .'.' :'- ''- .- - .- .' -.':: :''' ... :: :.: : : m u~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... . . . . . . .. . . . . . . : -: -:- :- : ..... ......... .. . . ....... . ...... .. : .......................... .. ... ... . n .. ... .. ... ... .. ...... .. ...... .. ... ... n -:.:.::.'...:.. '' '' ..''.-...... . . . . ' . : : - , , - .:. :-:..:::,''..'''.:.,,,.,.,. ._S..._S .~ ~ ~ ~ ~ ~~~~~~~~~0 ::::::: :::: ..---- 6P 5U S -Jdd .... ..§.u:::::: . :- - -.- -:--- - ;~ ON . , . : - -- , :,,,,,,,,~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~L .....---.- -. -.-.-.-.....-. Z -. __ __ ____ a ~ OEE .... .... C,~ ~~99 oz ~~~~~~~~~~~~~~~~~~~~~or zOJOOS 4301 o5qezv UPGH qL euntgj ulutooLos Z oas DU;tcowos; Das wu' qun le wu qn waG unu .ew.unu waI qun leiqun wUJS[nJ lfwlJnu . . . .. .. . . .. . . . . . . . . .i60 m C BuyisouoS~~~~ SFd6soWS° :-::::::::-::::::: : -:: .... : --:::- : -: ::::: : ::::::-:..-: : :. ............. . .............. : : ::: :::::::: ::::. : ::.-:. :-:1........ . .. .... .... ... .. ....... ....... . :-::: :-:: :-: ..:: ......., ,., . :: . :. . : :: . E.':, :::.... ........_ _ ~~~~~~~~~~~~~~~~~.. .... . .. .. . LE~OZ fl a° ii cLan6 j .. . . . . . . . . . ... . . . . . . . . . . . . .. . . . .. .. .... .. .. . . :: . . .. a. .... .... .... ........ ............5UHOCL4..S.W VJd ..........ON .............. .........................~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~~~~~~~~U- we.. ..... ......u~ tewuweqnui~n rwn *0 C~~~~~~~~~~~~~~~~~~~~~............... ..........................00 0.... .. .. .. . .. . ..as.. .. .. . .. .. . .. .. . . . . .. . .. . .. .. . .. . .... . . . . . . . . . . ............ ~~ ....... fl L ESL a; ci~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...... ............. ... .. .. . .. . . .. . . .. . . .. .. .. .. .. . .. .... . . . .. . .. . . . .. . . .. . . .. . . . 9W ........ ........M ............................... ........... ............. ........... ~ ~ ~ .......... .......... . .........-..C E... ...... . . . .... 0~~~~~~~~~~~~~~~~~~~~~C ...... . .. ..... .... .....S E... .... .... .... .... ... .... .... .....C. .8J003.. .I...OUGJ... :m U7~~~~~0 jn~ Figure 8 Percentage Ever Attended by Expenditure Quintile too BH.6 LOG 8B.9 Ot 97.2 88.5 75 67.0 7 709.... 7:- - l ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~... ... 5.7 -..' rur. *rl rur. m urb. *^1 urD. m rur. erl rur. de urD. *rl urD, rar~~~~~~~~~~~~~~~~~~~~~......... .. ... .. Ouintile 4 . Ouintile 5, Riche3t Guintile~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~... .. ... go go~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. .. . 42.6 24.7 20 20 . ... U' f~~~~~~up. all Fur, fter Lrb, sel urVb. fop Pup. Mal Fur, femn LIb. em2 u'b, fers Fur. gol Pup. foe utb. miii u"b fee 0Quirnti2e 1, Poorlost Quintils 0uintile 2 Ouinvtħl0 3 97.1 LOi00 76.0 7. 40.2~~~~~~~~~~~~~~~~~~~7. pu. . . . ..fr....O sb enPU.mO p.m tbh a04 Guirvtoim 4 ~ ~ ~ uinti1e 5, Ric.h........ .... Figure 9 Mean Last Grade Attended by Expenditure Quintile 12 12~~~~~~~~~~~~~~~~~~~~~~~~~~ 12- 10 10~~~~~~~~~~~~~~~~~~~~~~~~~~~~i to 6.10 ~~~~a .2 7036.77 4 4~~~~~~~GA 0~ Guitile 1 ~ ~urlt~10~ur. il rw faa rb. ni jib.faa ~aa~ Sr.10 rb i ub fa...nl u~ ............. .i. ..... . 0 0U~~~~~~~~~~flt~~~~~~~1O 1, Poorest Du~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~nt..............2. 12 12~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.5......... .0........ 10 21 10 B 7~~~~~~~~~~~~~~~~~~.........7 6.44 569~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~......... a * I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~...... . C-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~... . Pup. ell Nur. fin irb, oil trb. fNa Nu. alil rNu. fen O,. Bill jtb. fNau.ol ru,fa r.cl ub g 0 Guintile 1, Pooretil A Giħntie OuInties 2 Ounttl1e Figure 10 Kaplan-Meler Survival Curves by Expenditure Ouintiles LO- CO.t- O,X - 0.71 ~~~~~~~~~~~~~~~~~~~~0,71- D 4 z D | tZ D | l2 2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2~~~~~~~~~~~~~~~~~ U t C1 Attended Lilt Coin Attendlo Lilt CIM Attended Ouintil3 1: Pooroat Quintile Quinotle 2 Gulntili 3 a, 0'~~~~~10 4 rup9. 1: Rural Mis. Group 2: Rural eenal t g roup 3: Urban Hele Group 4: Urban Fenale * ~~~~~~~~~~~~~3 i~~~~~~~~~ 0.00 - 0.00 - LUt Clrn Attnded Lilt ClM Attinded Qu1nt1le 4 Guintile !: Richest GuintiA Figure Ila Primary School Completion Rates by Quintile 87. ,51 89.2 88.9 75~~~~~~~~~~~~~~~~~~~50. ... LID 96.9~~~~~~~7. 7575 a 75 .... SD SD~~~~~~~~~~~~~~~~~~~~~~~~........... ......... ..... ........... .... .. ...... ruP. all NJl'. ii ftc PuP. 3*1 PuP. eel... ................. .........Dunl1 5 ices u.it1 Figure lb Secondary I School Completion Rates by Quintile LOO 100 LOO- 84.9 72.7 74 73. 75 75 75 * ~~~~~~~~~~~50.0 ...... . ........ rur. crl rur. n. urr,. Cal~~~~~~~~ r..... rur.0" r.Nl.c 32.5 4 38 S 25.3 23 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~23 77 23 ol r~~~iue. ,.l eu ferI r b. so I Ur, SEagr a Ie nu1. ferI ħr b I5 I wrb f a P'up. Wi Mru. fogn Lr b Il aI ul, flua 0uintile 1, Poonost Quintħle Quin~tile 2 Guħrnti1B 3 ħ00 toO g~~~~~~~~~~~~~~~~7. 2.0 83.374 4..~~~~~~~~~~~~~~~~~~~~7. a~~~~~~~~~~~~~~~~~~~~~~~~~~~~......... 50.0 35.3 lIf id 11ff.2ru fle a Ur b . Mal Jrb .fea !Lir. u uI~.M jS l GUin~tilS 4 GUintils 5, RiCheSt Guħritils Figure 12 Pr)imary to Secondary I. Transition Pates by Ouintile S 75 7 a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....... ....... L~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.... ....... r 80~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..... ....... rr.8 54.pu. ur urb. ...... ur........ r.2u. .ar..ur . ..l ur...n.up........n.ib. i rb fr Ouintia ~~~~~~~, Poorest Guintile Duirtils 2 0u5.r~~~~~~~~~~~~~~~~~~~~6.B483t.i...... LCD too~~~~~~~~~~4. ............ ....... ..... .. 8 7..3.. .. . ... . .. . .. . .. .. .. .. . .. .. . .. .... 4.. 4,~~~~~~~~~~~~~~~~~~~~~~~~~..... . a~~~~~~~~~~~~~~~~~~....... I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.......... L~~~~~~~~~~~~~~. ..... - 50 50~~~~~~~~~~~~~~~~~~~~...... .............. ... 25 25 25.0~~~~~~~~~~~~~~~... .... ...... .. 0 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.......... . MaP. 331 PUP. fern tab Eli tat fEb PuP~~~~~~~~~~~~~ ~~~~~~. ..... P.. E ......t0fE GuIrtile 4 Guvitale 5, Richest~~~~~~~~~~~~~........ .. .....t.... Figure 13 Percent Ever Attended by H-ead and Spouse Schooling rural male a rural fHIBLI rural male [J rural laminla Iurban nalne 5 urban f male urbanl mile. urbln lnanls 100 ioo 95.1 BLO~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~10 0 13.6 ~~14.0 u.s 17.1 I 10,0~~~~~~~~~~~~~~~~a. 11.0 B40 4, 1, . .......... B~~D.0 77.3 72.1 ... 75 ... ..75 12.462 2 ....... . ... . I0.D 53.0 .... ....~~~~~~~~~~~~~~~~~~~3 †.~~~~~~~~..... 31.0 50 .. 45 0 none primary mao I ace~~~~~~~~~~~~~~~~~. ...... . .... ... e...... a.y.s... . Percent Ever' Atten~~~~~e~ Dy HesO scr~~~~oo2ing Level Percent Ever Attended Dy 5pcL~~~~~~~~se Scroclin.....Level Figure 1.4 Mean Grade Attended by Head and Spouse Schooling I rural male H rural female a rural male 0 rural lemele 12 I urban male urban female 12 urban riale 0 urban female 10 10 5.2 8.0 5 13 7.3 Gr GA 53 ~~~~~~~~~~~~~~~~~6.55. 43 44~~~~~~~~~~~~~~. 6 ...~~~2 22 22~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~............ 2~~~~~~~~~~~~~5 .......2.. 0~~~~~~~43.. . none primary eec i.. e ...2 0....r.ay ecIee Mean Grade Atten ...e . by4edSho n Lee4enGaeAteddb pue coln ee Figure 1.5 Kaplan-Meier Survival Curves by Schools Present in Douar i.00 - i.00 - _ i 0.75- E 0.7S- 0.50- 0'50 - 0.2!- 2 8 0.23- 2 0.00 - 0.00 - __________________________________________________ X o.oo i t i s.do o.8o | 8 X {4.do Loot Clan, Attended Lilt ClinI Attended -a No Primary, No Secondary Primary, No Secondary 1.0 - > 0.75- X. 0.50- Z 2 Group 1: Rural Male a 0.25- Group 2: Rural Female 0.00 - 1 2.00 Lnt Clnas Attnlided Primary, Secondary Figure 16 Kaplan-Meier Sunvival Curves by Presence of Road in Douar Group i: Rural Male 5roup 2: Rura2 Female 6rup 1: Rural Mals Group 2: Rural Fsmo; i .00 I 00 0.75 I 0.75 0.25 0 25 2 0.00 0.00 0 7 14 0 X 16 Last Class Attendst Last Class Attended Road Present No Road Present Distributors of World Bank Publications ARGENTINA The Middle East Observer ITALY PORTUGAL Carlo. Hirsch, SRL 41. Sherif Street Licosa Comminaonarna Sansoni SPA Livrana Portugal Galeria Gueme Cairo Via Duca Di Calabria 1/ I Rua Do Carmo 70-74 Florida 165. 4th Floor-Ofc. 453/465 Casella Postaie 552 1200 Lisbon 1333 Buenos Aires FINLAND 50125 Firenze Akateeminen Kir1akauppa SAUDI ARABIA, QATAR AUSTRALIA, PAPUA NEW GUINEA. P.o. Box 128 JAPAN (arir Book Store FIJI. SOLOMON ISLANDS, SF00101 HelInki 10 Eastern Book Service P.O. Box 3196 VANUATU, AND WESTERN SAMOA Hongo 3-Chome, Bunkyo-ku 113 Riyadh 11471 D.A. Information Services FRANCE Tokyo 648 Whitehors Road World Bank Publcaons SINGAPORE, TAIWAN Mit harri3132 66, avenue d'lena KENYA MYANMAR,BRUNEI Victora 75116 Par Africa Book Servce (E.A.) Ltd. Gower Asa Pacific Pte Ltd. Quaran House. Mfangano Street Golden Wheel Building AUSTRIA GERMANY P Q Box 45245 41, Kallang Pudding, #04-03 Gerold and Co. UNO-Verlag Nairobi Singapore 1334 Graben 31 Poppelsdorfer Allee 55 A-1011 Wien D-530 Bonn I KOREA, REPUBLIC OF SOUTH AFRICA, BOTSWANA Pan Korea Book Corporation Forsingic htif: BANGLADESH HONG KONG, MACAO P.O. Box 101. Kwangwharnun Oxford University Pres Micro Industites Development Asia 2000 Ltd. Seoul Southern Afica Assistance Society (MIDAS) 4648 Wyndham Seet P.O. Box 1141 Houe S, Road 16 Winnung Centre Korean Stock Book Centre Cape Town 8O0 Dhanmondi R/Area 2nd Floor P.O. Box 34 Dhaka 1209 Central Hong Kong Yeoetdo For sutecnptirn order: Seoul Intenatbonal Subscription Service Bnna office: HUNGARY P.O. Box 41095 Pine View, Ist Floor Foundation for Market Economy MALAYSIA Craighafl I 00 Agrabad Commercial Area Dombovan Ut 17-19 Unversity of Malaya Cooperative Johanneburg 2024 Chittagong4100 H-1117 Budapest Bookshop, Liurted P.O. Box 1127. Jalan Pantai Bar SPAIN BELGIUM INDIA 59700 Kuala Lumpur Mundi-Prensa Libros, SA. lean De [annoy Allied Publishers Private Ltd. Castelo 37 Av. du Roi 202 751 Mount Road MEXICO 28001 Madrid 1060 Bruseis Madras -600002 INFOTEC Apartado Postal 224860 Llbrerta internucwnai AEDOS CANADA Bnch ofice: 14060 Tlalpan. Mexico DYF Conseil de Cent 391 LeDiffuseur 1SJN. HeredtaMwrg 08009 Barceiona iSIA Boul. de Mortagne Ballard Estate NETHERLANDS Boucherville Quebec Bombay - 400 038 De Lindeboom/inOr-Publikaties SRI LANKA AND THE VfALDiVES J4B SE6 P.O. Box 202 Lake House B/ookhop 13/14AsafAL Road 74B0AE Haaksbergen P.O. Box 244 Renouf Publishing Co. New Delhi - 110 002 ltOO Sir Chittampalam A- 1294 Algoma Road NEW ZEALAND Cardiner Mawatha Ottawa, Ontario 17 Chittaranian Avenue EBSCO NZ Ltd. Colombo 2 KIB 3W8 Calcutta - 700 072 Private Mati Bag 99914 New Market SWEDEN CHILE Jayedeva Hostel Building Auckland Forosingle kiL Invertec IGT SA. 5th Main Road, Gandhinagar Fritnzes Fackboksloretaget Av Sinta Ntria 6400 Bangalore -560009 NIGERIA Regeringsgatan 12. Box 16356 Edif ido iNlEC Of. 201 Univerty Press Limitd 5-103 27 Stockholm Santiago 3-5-1129 Kactuguda Three Crowns Building Jericho Cross Road Prlvate Mal Bag 5095 For mitcertion ord, CH]NA Hyderabad-500 027 ibadan Wennergren-WilliamsAB China FLnanciati & Economki' P. O Box 1305 Publishing House Prarthana Flats. 2nd Floor NORWAY S-171 25 Solna 8, Da Fo Si Dong Jie Near Thakore Baug, Navrangpura Narvesen informabon Center Beijing Ahmedabad - 380 009 Book Department SWrMERLAND P.O. Box 6125 Elterstad For sing tile: COLOMBIA Patiala House N-0602 Oslo 6 Llbrazie Payot Infoenlace Ltda. 16-A Ashok Marg GIY postale 3212 Apartado Aereo 34270 Lucknow - 226 001 PAKISTAN CH 1002 Lausanne Bogota D.E Mirza Book Agency Central Bazaar Road 65 Shahrih-e-Quaid-e-Azam For smwtin onrder: COTE D'IVOIRE 60 Bajaj Nagar P Q Box No 729 Libraine Payot Centre d'Edition et de Diffusion Nagpur 440 010 Lahome 54000 Service des Abonnements Africaines (CEDA) Case postale 3312 04 B.P. 541 INDONESIA PERU CII 1002 Lausanne Abidjan 04 Plateau Pt. Indira Limited Editorial Desarroilo SA Jalan Borobudur 20 Apartado 3824 THAILAND CYPRUS p.O Box 181 Liua I Central Department Store Center of Applied Research jakarta 10320 306 Siom Road Cyprus College PHILIPPINES Bangkok 6. Diogena Street Engomu IRAN International Book Center P.O Box 2006 Kowkab Publishers Suite 1703, Cityand 10 TRINIDAD & TOBAGO. ANTIGUA Nicosia P.O. Box 19575-511 Condominium Tower I BARBUDA, BARBADOS, Tehran Ayala Avenue, H.V. dela DOMINICA. GRENADA, GUYANA, DENMARK Costa Extension JAMAICA, MONTSERRAT, ST. SamfundsUtteratur IRELAND Makati MetoMarula KITS& NEVIS ST LUCIA, Rosenoerns A)lt 11 Government Supplies Agency ST. VINCENT &GRENADINES DK-1970 FrederiksbergC 4- Harcourt Road POLAND Systematics Studies Unit Dublin 2 International Publishing Service #9 Watts Street DOMINICAN REPUBLiC Ul. Piekna 31/37 Curepe Editor Taller. C. por A. I SRAEL 00-677 Warzawa Tnnidad, West India Restauracion e Isbel Is Cet6lica 309 Yoznot Literature Ltd. Apartado de Correos 2190 Z-1 PO Box56055 For rascrriron orders: UNIMD KINGDOM Santo Domingo Tel Aviv 61560 IPS lounals Microinfo Ltd. Ul. Okrezna 3 P.O. Box 3 EGYPT. ARAB REPUBLIC OF 02-916 Warszawa Alton. HAmpsture GU34 2PG Al Ahram England Al Galaa Street Cairo Recent World Bank Discussion Papers (continued) No. 237 Indonesian Experience with Financial Sector Reform. Donald P. Hanna No. 238 Pesticide Policies in Developing Countries: Do They Encourage Excessive Use?Jumanah Farah No. 239 Intergovernment Fiscal Relations in Indonesia: Issues and Reform Options. Anwar Shah and Zia Qureshi No. 240 Managing Redundancy itn Overexploited Fisheries. Joshua John No. 241 Institutional Change and the Public Sector in Transitional Economies. Salvatore Schiavo-Campo No. 242 Africa Can Compete!: Export Opportunities and Challengesfor Garments and Home Products in the U.S. Market. Tyler Biggs, Gail R. Moody, Jan-Hendrik van Leeuwen, and E. Diane White No. 243 Liberalizing Trade in Services. Bemard Hoekman and Pierre Sauvc No. 244 Women in Higher Educatiotn: Progress, Constraints, and Promising Initiatives. K. Subbarao, Laura Raney, Halil Dundar, andJennifer Haworth No. 245 What We Know About Acquisition of Adult Literacy: Is There Hope? Helen Abadzi No. 246 Formulating a National Strategy on Inrfonnation Technology: A Case Study of India. Nagy Hanna No. 247 Improving the Transfer and Use of Agricultural Information: A Guide to Information Technology. Willem Zijp No. 248 Outreach and Sustainability of Six Rural Finance Institutions in Sub-Saharan Africa. Marc Gurgand, Glenn Pederson, and Jacob Yaron No. 249 Populatiorn and Income Change: Recent Evidence. Allen C. Kelley and Robert M. Schmidt No. 250 Submission and Evaliation of Proposalsfor Private Power Generation Projects in Developing Coutitries. Edited by Peter A. Cordukes No. 251 Supply and Demandfor Finance of Small Enterprises in Ghana. Ernest Aryeetey, Amoah Baah-Nuakoh, Tamara Duggleby, Hemamala Hettige, and William F. Steel No. 252 Projectizing the Governanice Approach to Civil Service Reform: An Institutional Environment Assessmentfor Preparing a Sectoral Adjuistment Loan in the Gambia. Rogerio F. Pinto with assistance from Angelous J. Mrope No. 253 Small Firms Informally Financed: Studiesfrom Bangladesh. Edited by Reazul Islam, J. D. Von Pischke, and J. M. de Waard No. 254 Indicatorsfor Monitoring Poverty Redjuction. Soniya Carvalho and Howard White No. 255 Violence Against Women: The Hiddern Health Burden. Lori L. Heise with Jacquelinc Pitanguy and Adrienne Germain No. 256 Womet 's Health and Nutrition: Making a Difference. Anne Tinker, Patricia Daly, Cynthia Green, Helen Saxenian, Rama Lakshminarayanan, and Kirrin Gill No. 257 Improving the Quality of Primary Education int Latin America: Towards the 21st Century. Lawrence Wolff, Emesto Schiefelbein, andJorge Valenzuela No. 258 How Fast is Fertility Declining in Botswana and Zimbabwe? Duncan Thomas and Ityai Muvandi No. 259 Policies Affecting Fertility and Contraceptive Use: Ani Assessment of Twvelve Sub-Saharan Countries. Susan Scribner No. 260 Financial Systerns in Sub-Saharan Africa: A Comparative Study. Paul A. Popiel No. 261 Poverty Alleviation and Social Investment Funds: The Latin American Experience. Philip J. Glaessner, Kye Woo Lee, Anna Maria Sant'Anna, andJean-Jacques de St. Antoine No. 262 Public Policyfor the Promotion of Family Farns in Italy: The Experience of the Funddfor the Formnation of Peasant Property. Eric B. Shearer and Giuseppe Barbero No. 263 Sef-Employmentfor the Unemployed: Experience in OECD and Transitional Economies. Sandra Wilson and Arvil V. Adams The World Bank Headquarters European Office Tokyo Office H 1818 H Street, N.W. 66, avenue d'lena Kokusai Building Washington, D.C. 20433, U.S.A. 75116 Paris, France 1-1 Marunouchi 3-chome Chiyoda-ku, Tokyo 100japan Telephone: (202) 477-1234 Telephone: (1) 40.69.30.00 Facsimile: (202) 477-6391 Facsimile: (1) 40.69.30.66 Telephone: (3) 3214-5001 Telex: MCI 64145 WORLDBANK Telex: 640651 Facsimile: (3) 3214-3657 MCI 248423 WORLDBANK Telex: 26838 Cable Address: INTBAFRAD WASHINGTONDC 13046 EDC 100 0-8213-3046-2 SCHOOLING & COGNITIVE 111111I111 111111111II 400000014548 $7.95 ISBN 0-8213-3046-2