Policy Research WORkING PAPERS L Country Operatlons Technical Department Latin America and the Caribbean The World Bank November 1992 WPS 1028 Socioeconomic and Ethnic Determinants of Grade Repetition in Bolivia and Guatemala Harry Anthony Patrinos and George Psacharopoulos Children from less wealthy households and children of indig- enous origins are more likely to repeat a grade, so targeted interventions could be directed at the poor and could have an indigenous component, such as bilingual education. Policy Reearch WorkngPaper disseninate the findings of work in progress and encourage the exchange of ideas among Bank staff and al others interegdh development issues.Thesepapems, distributed by theReseafch Advisory Staff, carry thenames of the authors, reflect ordy tibrviews, and should be used and ited accordongly.Te findings, interpretations, and conclusions ametheauthors-own.Tey should not be attributed to the World Bank, its Board of Directors, its management, or any of its member countries. Policy Research Country Operallons WPS 1028 This paper-a product of the Latin America and the Caribbean Technical Department-is part of a larger effort to document poverty conditions in the region. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Liliana Longo, room 14-187, extension 39244 (November 1992, 26 pages). After reviewing the literature on repetition Their empirical analysis shows that certain (students repeating grades in school) in develop- populations are more likely to repeat a grade: ing countries, Patrinos and Psacharopoulos children from less wealthy households and examine factors related to repetition in Bolivia children of indigenous origins. This suggests that and Guatemala. They develop a model to esti- any targeting activities could be directed to the mate the incidence and determinants of repeti- poor and could have an indigenous component, tion. They use multivariate logistic regression such as bilingual education. analysis to estimate the determinants of repeti- tion, using the results in simulations to determine probabilities of who is more likely to repeat. The Policy Research Working PaperoSeries disse oinates the frdings of work mnder way in theBark An objectiveof the series is to get these rindings out quickly, even if presentations are less than fuilly polished. The findings, interprctations, and conclusions in these papers do not necessarily represent of ficial Bank policy. Froduced by the Policy Research Dissemination Center * SOCIOECONOMIC AND ETNINC DETERMINANTS OF GRADE REPETITION IN BOLIVIA AND GUATEMALA by Harry Anthony Patrinos and George Psacharopoulos The World Bank Washington, DC 20433 Table of Contents I. Introduction 1 II. Literature Review 1 m. A Bird's Eye View of Repetition in Latin America 3 IV. An Empirical Investigation 5 Descriptive Statistics 9 Multivariate Analysis 13 V. Conclusion 18 References 19 Annex 1 Mean Sample Characteristics, Bolivia 1989 22 Annex 2 Zero-Order Correlation Matrix 24 Annex 3 Logistic Regression Results: The Determinants of Repetition in Bolivia 25 * The authors wish to thank Haeduck Lee and Hongyu Yang for providing assistance during the various stages of preparing this paper. I. Introduction School dropout and grade repetition are among the most serious problems plaguing educational systems around the world today. This problem is especially pronounced in Latin America and the Caribbean, the region with the highest incidence of repetition and dropout in the world (see Unesco 1991). Nearly half the children who enroll in the first grade never complete the fourth grade. Latin America leads the developing world in the percentage of repeaters in primary education. On average, about 29 percent of all primary students are repeaters each year; and 42 percent are repeating the first grade (Schiefelbein 1992). Clearly, repetition is the key issue in Latin American primary education, both for improving overall efficiency and reaching suitable levels of quality (Schiefelbein 1989). Repetition also acts as a signal that something is wrong, that achievement is inadequate, that the quality of schooling is low, or that learning disabilities brought from a deprived learning environment are present. School repetition and school dropout are closely linked, the one often leading to the other. They can be attributed to the supply side, i.e. the offering of low quality instruction, or to the demand side, i.e. families needing child labor for agricultural production and hence withdrawing their children from school, or forcing them to repeat because they have not learned the material during the school year because of work-related distractions. It is important to note that the reduction of repetition and dropout are not ends in themselves. Regarding repetition, leaming is the issue. If a student learns more by repeating a grade, repetition is not a problem (see Psacharopoulos, Rojas and Velez 1992), except in terms of efficiency costs if alternatives exist. However, it is a problem if repetition becomes perennial, leading to dropout before a child attains literacy. At some stage, a child must "drop out' of school, e.g. at the end of the primary, secondary or higher education cycle. But if dropout occurs before a child finishes ihe compulsory level of education, presumably also missing literacy for life, then dropping out of school is a very serious problem. In this paper we first review the literature on repetition in developing countries and then examine a series of factors that relate to repetition based on household survey data for both Bolivia and Guatemala. We then proceed to estimate the determinants of repetition using a series of household demographic and ethnic variables. Results from the analysis are used to simulate the probabilities of repetition against particular characteristics. II. Literature Review The high levels of repetition and drop out at the primary schooling levels prevent the realization of universal primary education in many developing countries (Davico 1990; Schiefelbein 1989). Repeaters reduce the enrolment capacity of their class, thus preventing other children from being admitted. Repeaters also caatse overcrowding, which raises school costs (Unesco 1984). Many authors have found that the major obstacle is the first grade (see, for example, Davico 1990; Schiefelbein 1992). Repeating the first grade often leads to more repetition, which is followed by eventual drop out, often resulting in illiteracy for life. 2 Davico (1590) found that poor teachers and iniadequate support to poor children are to blame for the high rates of repetition and drop out in Brazil. It is often the case that the poorest teachers are to be found in the first grade. There is also evidence of discrimination against students from disadvantaged backgrounds, and a belief among teachers that poor students cannot learn. The teaching method reinforces such attitudes, leading to high rates of repetition and drop out for those from disadvantaged backgrounds. Schiefelbein also finds that repetition is especially severe among the most deprived; that is, those whose parents have the lowest levels of education and income are more likely to repeat a grade (Schiefelbein 1992: 7). Recent research for Uruguay has found that repetition decreases significantly as the level of the parents' education income increases (Schiefelbein 1992: 6). Repetition rates are also high among those living in rural, isolated areas, and among indigenous peoples and non-whites. It has been found that the repetition rate is twice the national average in Malleco province in Chile, which has the highest proportion of indigenous people in the country (Schiefelbein 1992). Significant differences by gender have not been detected, and although girls in general do- better than boys, gender differences have been diminishing over time (Schiefelbein 1992: 8). Schiefelbein (1992, 1989, 1975) adds some further explanatory factors, such as children with learning disabilities, "age heterogeneity' in the classroom (see below), language and, in some cases, the lack of bilingual education. Psacharopoulos and Yang (1991), in their study of educational attainment in Venezuela, also examined the problem of repetition. Using a 1987 sample of school children aged 10 to 18 years, they found that grade repetition was associated with urban areas (negative) and with being male (positive). They also established that family background (father's education) has a significant effect on lowering the repetition rate. Regarding the determinants of grade repetition, Schwille et al. (1991) presented a detailed causal model to explain grade repetition. Three main categories were put forward, including: student characteristics, such as gender, ethnicity, family influences, place of residence, age, prior repetition experience, aptitudes, and perceived opportunities for work/employment; national policies regarding language, repetition, secondary school entrance, other access and selection policies, and the quality of instruction; and school characteristics, such as school management practices, coverage of syllabus, quality of instruction, compliance with national policies, and the capacity of the school relative to the number of non-repeaters. All this is hypothesized to cause repetition, which in turn affects student leaming, student motivation, student self-esteem, the examination success rate, the enrolment rate, the dropout rate, and the mean time required to produce graduates. The demand for child labor, whether paid or unpaid, affects schooling, often causing grade repetition. This relationship was examined by Bowman and Goldblatt (1984), who found that the demand for the productive contributions of children to the household economy are closely associated with "overage" and the proportions of youth who do not complete even four years of primary schooling (Bowman and Goldblatt 1984). 3 Using a unique follow-up data set for Northeast Brazil, Gomes-Neto and Hanushek (1991) examine the causes and effects of grade repetiton. They find that student achievement levels are very important, as is the availability of advanced grades. Jamison (1978) examined the determinants, among other things, of student failure and repetition in Nicaragua. Through an evaluation of the Radio Mathematics Project, and using maximum likelihood methods, he found that males are less likely to fail, that the number of times repeated has negative impact on failing during the current year, and that student achievement on test score are significant factors. The costs of repetition are very high for a developing country. Estimates for Brpzil show that the cost of repetition represents an amount equivalent to the entire federal govemment contribution to first level schooling (World Bank 1986). Policies to reduce repetition and dropout are advocated for the more efficient use of education resources. Cuadra and Fredriksen (1992) estimated efficiency gains as3ociated with reducing repetition and dropout through a simulation exercise. They found that reductions in repetition and dropout rates would result in more students completing the basic cycle of education at a lower input per graduate. Less repetition frees places and lowers the cost of graduating a cohort, while reduced dropout results in more students completing the cycle. Furthermore, less repetition may mean less dropout as the foregone earnings associated with schooling, which are greater when repetition is higher, decrease as the rate of repetition falls. High levels of grade repetition do not only have cost and efficiency implications, but also affect the foregone earnings associated with attending school, outcomes, and the returns to education. It has been found that repetition rates affect the rate of return to education. Assuming it takzs on average two extra years to complete the basic education cycle, and more for those from.- lower socioeconomic backgrounds, the returns to schooling are found to drop significaitly in Brazil, from 13 percent to 8 percent, after repetition is factored into the calcw.ations (Jallade 1977). Also, Behrman and Deolalikar (1991), using Indonesian data, find that failure to account for repetition and drop out leads one to overestimate the returns to schooling. m. A Bird's Eye View of Repetition in Latin America Table 1 presents key information on primary schooling in Latin America, as well as conservative estimates of the indigenous population in some countries. While many children do not start school on time - according to the official primary school entry age -, most do, however, eventually enroll. Primary school coverage is complete in the Anglophone Caribbean region and Mexico, as well as for a number of Central and South American nations. The lowest levels of primary school coverage occur in Haiti, Guatemala, El Salvador, the Dominican Republic, and Nicaragua. 4 Table 1. School Access, Repetition, Grades Attained, and Indigenous Population in Latin America and the Caribbean Sub-region/ School Accass (%. Repetition Yea in Grades Indigenous Country On time Ever (%) Primary Attained (%) South America .. 93 45 7.1 4.2 Argentina .. .. 26 .. .. 0.1 Bolivia 61 90 38 4.8 2.2 66 Brazil 69 92 54 7.7 3.9 1 Chile 44 98 11 6.4 5.5 9 Colombia 43 83 44 6.2 4.5 1 Ecuador 82 100 28 6.2 5.2 21 Paraguay 100 100 28 6.3 4.8 3 Peru 66 97 29 6.5 5.1 40 Uruguay 43 100 18 6.3 5.7 Venezuela 76 95 28 6.4 4.5 1 Central America .. 81 47 6.7 4.1 Costa Rica 43 99 23 6.4 5.2 1 El Salvador 65 73 51 7.0 4.0 21 Guatemala 52 72 47 6.4 4.0 50 Honduras 83 93 49 6.9 4.5 7 Nicaragua 73 78 60 6.8 2.7 3 Panama 82 99 26 6.5 5.4 6 Gulf of Meico .. 96 33 5.8 4.1 Haiti 1 44 54 4.8 1.3 Mexico 96 100 32 5.7 4.3 11 Dominican Republic 76 74 55 6.7 3.4 Anglophone Caribbean .. 100 7 5.6 4.9 Bahamas 2 100 0 6.0 5.7 Barbados 87 99 18 5.7 5.0 Dominica 95 98 26 6.3 5.2 Grenada 88 100 15 6.0 4.4 Guyana 73 100 8 6.0 5.4 3 Jamaica 82 100 4 5.3 4.7 St. Kitts and Nevis 21 100 2 6.0 5.9 St. Vincent 87 99 23 6.3 5.2 St. Lucia 96 100 4 5.8 5.4 Trinidad and Tobago 100 100 8 5.7 4.8 Overall 41 93 42 6.8 4.2 Sour=cs: Sie (1992); nigos pouliom euimuw: Ber (1990). Note: .. available. 5 The rate of primary school coverage, however, has little relation to the rate of grade repetition in a country. Repeiition rates at the primary level are very high, especially in South and Cent-al America. The rate of repetition, for instance, is over 50 percent in Brazil, and 60 percent in Nicaragua. The lowest rates of repeC4;on occur in the Anglophone Caribbean. In addition to very high repetition rates, primary schooling in most Latin American countries is inefficient. It talk, on average, seven years to complete four grades of primary schooling in South America. This means that one-and-one-half years of schooling are required to complete one grade at the primary level. In Brazi and Haiti, the differential between years of schooling and grades attained is even greater. This inefficiency occurs at a high level in Nicaragua, El Salvador, the Dominican Republic, Bolivia, and Guatemala. The difference between years of schooling and grades attained is negligible in most Caribbean countries. Another feature of Table 1 is that it provides estimates of the indigenous population expressed as a percent of the total population for countries where this is relevant and the data was available. Several Latin American countries have sizable indigenous populations, the biggest being in Bolivia and Guatemala, where, according to most estimates, they constitute the majority of the population. In many countries, due to a variety of factors, including language, lack of provision of social services, geographical location, and discrimination (see Kelley 1988; Stephen and Wearne 1984), being an indigenous person is associated with, among other things, extreme poverty and illiteracy. This is especially the case in rural, isolated areas, which places children in an extreme disadvantage. In rural Peru, for instance, where the majority of the population is indigenous, it has been found that 70 percent of Quechua-speaking people over the age of five have never received any schooling, relative to only 40 percent of rurat non-indigenous Peruvians (Hernandez 1988: 126). IV. An Empirical Investigation In this paper we focus on children attending the primary level, aged less than 15 years. The reason for concentrating on primary schooling is that repetition is less of a problem at the secondary or tertiary levels. Also, since we do not have matching data on learning, we focus on household demographic, wealth and other demand factors in determining repetition. We also control for ethnicity by including indigenous identity as an independent variable in the analysis. For this reason we concentrate on the two Latin American countries with the highest proportion of indigenous peoples in their populations, Bolivia and Guatemala. The data from Bolivia used in this analysis come from the 1989 Encuesta Integrada de Hogares (EIH conducted by the Insdtuto Nacional de Estad(sdca. The survey covers only urban areas with populations over 10,000. Our analysis is limited to 5,614 enrolled primary school children between the ages of 7 and 14 who are sfill in school. For Guatemala, the data come from the Encuesta Nacional SociDemogrofica (ENSD), conducted in 1989 by the 6 Insdtwo Nacloial de Estdsdca. The survey covers about 11,000 households in all parts of the country. Our analysis is limited to 2,974 enrolled primary school children between the ages of 10 and 14. The key dependent vaiablc aong the two groups of primary school children is REPEATER, or movrae', or age/grade distorton." A repeater is a student whose age is above the normal for fte grade he/she just completed or is currently attending. Using a variable called schoolng-for-age (SAGE) (see Pcharopouos and Yang 1991), we assess the progress of a child in the school system and esimate rpetition, or 'ovenge'. Using the formula: SAGE( Yea of Schoolinz Age - E where E represents the usual school entry age in the country, age six in Bolivia, seven in Guatemala. We consider all those with a score under 100 as being below normal progress in the school system because of grade repition or lte entry. However, as ths measure of repeition may be captuing late entrants, it can be described more as a measure of "age-gade distortion,' or *overage.' ScFtefelbein (1992: 31) noted that 60 pent of children in Bolivia enter school on time, that is, at age six; first grade repetition is 38.4 percent. The sample we are using, however, is urban only; therefore, !ate entrants may be fewer. Moreover, our method does not overmate repetition - it underestimates it. According to our modd, only 16.8 percent of Bolivian school children are estimated to be first ,rade repeaters. The Guatemalan educaton sysem, however, is caczed by high rates of failure at all grade levels, accompad by repeating. While 52 percent of children enroll on time, more than 70 percent of rural stutens are older than the expected age for their grade, compared to S9 perct in urban areas. Sill, only 72 percent of Guatemalans ever enroll; and the rate of repetition is 47 percent (Schiefelbein 1992), although higher in rural areas. Our method gives a much higher rate, but it should be noted that this has much to do with the fact that the Guatemalan sample consists of children are eady oveaged; the sample is limited to those ten years of age and older (the Guatemalan survey did not collect information on those under ten years of age). Also, our sample, however, includes students between the ages of 7 and 14 (10 and 14 in Guatmala), in al prmary school grades, so that a number of 'late entrants" will have dropped out, and will consequendy not be included in our sdy. An umpoant charactec of the Latin American primary school, as well as a factor detemining repetition, is te pmrence of childrn of various ages in the same grade, some of whom are themselves repeaters (Schiefebin 1992: 10). Estimates of the uoveragedw in primary schools in Lain Ameca and the Caribbean range from 43 percent in Brazil, 34 percent in the ret of South Amaica, and 32 perent in Centrl Ameica (32 percent), to just 11 percent in the Anglophone Caribbean (OAS 1992). In general, age/grade distortion is much greater il. rural 7 areas. In Peru, over 90 percent of urban children are in school by age six, while the same can be said for only 57 percent of rural children (nlon and Moock 1991). Bowman and Goldblatt (1984) examined the phenomenon of "overage" in Mexican primary schools. They found that overage and low family incomes are strongly related, and that "low incomes (low ability to pay) had more serious negative effects on the schooling of girls than of boys.' A particularly interesting amd useful finding was that overage in the first grde, and overage in primary school in general, are not the same thing; in fact, they are negatively correlated. Early grade repetition As strongly associated with drop out, while being overage in primary school (but not first grade) indicates school progress. On the question of whether "overage" more heavily represents repetition or late entry, the answer is unclear and although there are competing theories, the evidence to date has not been conclusive. Enrolling in primary school on time has been found to be positively associated with family wealth and parents' schooling, and negatively associated with school costs in Pet l (Ilon and Moock 1991). This means that for a variety of reasons it is the poor who are more likely to enroll late, repeat grades and eventually drop out of school, given the relationship between overage, repetition, and dropout. Other reasons put forward for delayed enrollment in primary school include borrowing constraints, the opportunity cost of child labour, a lack of school places, and malnutrition (Glewwe and Jacoby 1992). Some researchers, however, argue that the principal cause of age/grade distortion is grade repetition, and not late entrance (Verhine and de Melo 1988: 564). Schiefelbein (1992, 1975) hypothesizes that "age heterogeneity" is also a cause of repetition. This is because of the effect on learning and teaching methods in the classroom when there are children of various ages in the same grade. It is "difficult. for the teacher to generate learning experiences of interest for the whole range of ages" (Schiefelbein 1992: 10). He goes on to report that 'for the most part, age heterogeneity corresponds to overage resulting from repetition." Repetition generates age heterogeneity, which causes problems in the classroom for non-repeaters as well, as teaching methods change in an attempt to accommodate children of various agei in the same grade. Repetition is expected to be greater, the higher the level of age heterogeneity in a given classroom. In other words, present repetition leads to a cycle of increased repetition. All this shows that even the late entrants in our sample of "repeaters" has consequences, for themselves, as well as for the students in a normal schooLig trajectory. The evidence and theory, therefore, tentatively point to the conclusion that overage and repetition may be the same thing in Latin America. The independent variables to be considered in the empirical analysis ame as follows: AGE, the student's age: ranging from 7 to .4 in Bolivia, and 10 to 14 in Guatemala MALE, the gender variable (0-1 dummy) INDIG, whether the student is indigenous (0-1 dummy) 8 REGION, residence in one of Bolivia's eight Departamentos, or urban/rural in Guatemala PRIVATE, attendance at a private school (0-1 dummy), Bolivia only SM, mother's years of schooling YFAM, family income MALEHHD, male household head (0-1 dummy) OCCHHD, occupation of household head (5 major categories in Bolivia, 4 in Guatemala) NSIBS, the number of brothers and sisters NROOMS, number of rooms in the household KITCHEN, the presence of a kdtchen in the houseihold (0-1 dummy) RUNWATER, the presence of running water in the household (0-1 dummy) The means and standard deviations of the sample variables are presented in Annex 1. We are interested in the effect of the above independent variables on repetition. The model to be estimated is a regression with a dichotomous dependent variable, REPEATER, defined as above. For this reason, logistic regression analysis will be used. The model is as follows: REPEATER =f(AGE, MALE, INDIG, REGION, (PRIVATE,) SM, YFAM, MALEHHD, OCCHHD, SIBS, NROOMS, KITCHEN, RUNWATER) As mentioned above, the aim of this empirical study is to explain the determinants of repetition with household charazteristics, controlling for ethnicity. It is expected that family income and indicators of relative wealth (such as KITCHEN, RUNWATER, NROOMS) will be negatively associated with repetition. Positive associations with repetition are expected for the variables controlling for indigenous origins and numbers of brothers and sisters. As parentai schooling is considered important, it is expected that mother's schooling will be negatively associated with repetition. Mother's schooling was chosen over father's schooling as a control variable because at early ages it is the mother who spends more time with the child. Nevertheless, the presence of a male household head is also controlled for, the hypothesis being that female-headed households are likely to be less wealthy, consequently less able to provide the proper home environment for the child to ensure success at school. It is, therefore, texpected that the presence of a male household head will be negatively as ociated with repetition. We also control for the houseiiold head's occupation, regardless of gender. Region of residence is expected to show great differences in repetition rates, reflecting the quality and accessibility of 9 schooling in different areas, as well as relative wealth. For Bolivia, being an urban-only sample, we control for Departamento of residence. For Guatemala, we use the rural/urban divide, ano judging from previous analyses, we expect the rate of repetition to be considerably higher in rural areas, for reasons such as inferior quality schools. Age, another control variable, is expected to be positively ass.ociated with repetition because the longer one remains in primary school, the more likely it is they have repeated a grade. Descriptive Statistics The mean sample characteristics for repetition by discrete independent variables are presented in Table 2. The mean repetition rate increases with age, as those students that continue in the system accumulate repeated grades. Girls have a slightly higher rate of repetition than boys. This aggregate result is somewhat surprising since, accord..,g to other accounts, boys have higher rates of repetition in most Latin American countries (Schiefelbein 1992). Indigenous ethnicity and repetition are highly correlated. A primary school student of indigenous origin is almost twice as likely to be a repeater relative to a non-indigenous child in Bolivia, while the rate of repetition for indigenous children in Guatema;a is almost 90 percent. The rate of repetition varies by region, or Departamento in Bolivia, of residence. It ranges from only 16 percent in Chuquisaca in southern Bolivia, to 37 percent in Beni, in the north. The Bolivian sample refers only to the urban sector. In Guatemala, however, where we possess information on both the rural and the urban sectors, great diversities are made apparent. While the overall rate of repetition is only 55 percent in urban areas, it is over 80 percent in rral areas. For Bolivia, the type of school attended is known. For those attending public schools, the rate of repetition is over 29 percent; it is a mere 8.7 percent for those in private schools. Parent's schooling and family income also affect repetition rates. In Bolivia, if the father has less than a primary education, the aggregate repetition r- e is over 34 percent; however, for those whose mothers have a complete secondary education or more, the repetition rate on average is less than 9 percent. In Guatemala, for those whose mother (father) has incomplete primary schooling, the rate of repetition is 74 percent; for those whose mother (father) has a complete secondary education, the rate of repetition is only around 30 percent. The rate of repetition also varies according to family income qsuintile, ranging from almost 29 percent for those in the bottom 20 percent, to less than 17 percent for those in the top 20 percent in Bolivia. In Guatemala, repetition varies from 84 percent at the bottom of the income distribution, to 50 percent at the top. The occupation of the household head also matters. In Bolivia, the rate of repetition is lower for children from households where the household head is an employer (12 percent), as 10 opposed to a laborer (35 percent). In Guatemala, the lowest rate of repetition registered is for those where the household head is a public sector employee (49 percent), and the highest is for self-employed household heads (78 percent). Repetition is found to be less for school children in Bolivia in households where the household head is male, while in Guatemala, surprisingly, repetition is lower if the household head is female. The re...ons for this anomaly are unclear. Several other household variables are used in the analysis. In general, the fewer brothers and sisters one has, the more rooms in the house, the presence of a kitchen and running water are all associated with lower rates of repetition in both countries. il Table 2. Mean Repetition Rates by Sdoeted Sample aerlic (%) Characteristic Bolivia Guatemala Age 8 14.8 9 22.8 10 25.6 60.7 11 28.2 65.2 12 30.6 70.2 13 37.4 78.9 14 49.7 100.0 Gender Female 25.5 71.8 Male 24.1 70.6 Ethnicity non-Indigenous 22.6 64.3 Indigenous 40.8 87.6 Region/Depanamento Chuquisaca 15.8 Cochabamba 17.8 Oruro 22.4 Sante Cruz 22.4 Tarija 23.5 Potosi 26.9 La Paz 29.5 Beni 36.8 Rural 83.0 Urban 55.0 Type of School Public 29.4 Private 8.7 Mother's Schoolig Incomplete Primary 29.9 74.3 Complete Primary + 18.5 46.0 Complete Secondary + 8.6 31.4 (Caurd) 12 Table 2 (Cont'd) Characteristic Bolivia Guatemala Father's Schooling Incomplete Primary 34.4 74.5 Complete Primary + 23.2 55.2 Complete Secondary + 11.5 29.8 Family Income Quintile Bottom 20 28.5 84.1 Next 20 28.8 78.8 Middle 20 26.5 74.0 Next 20 23.4 66.3 Top 20 16.6 49.9 Household Head Gender Female 32.7 63.0 Male 24.2 72.4 Household Head Occupation Employer 12.2 61.3 Employee 18.6 Self Employed 27.6 77.8 Laborer 35.1 Public Sector Employee 49.2 Private Sector Employee 72.5 Siblings I to2 18.0 65.1 3 to4 21.2 75.6 5 + 33.3 80.5 Number of Rooms 1 36.6 86.7 2 27.3 78.7 3 25.1 71.9 4 18.4 64.0 S or more 13.1 42.3 Kitchen No 31.6 76.1 Yes 23.0 64.6 Running Water No 29.9 80.7 Yes 18.7 62.7 Overall 24.8 71.2 13 Multivariate analysis The results of a multivariate model controlling for the factors that simultaneously determine grade repetition appear in Annex 3. Since the "primary school repeater" is a 0-1 limited dependent variable, a logit model was fitted. The model expresses the probability (P) of someone being a primary school repeater, as a function of various characteristics (X), such as age, indigenous origin, gender, and so on. p z 1 1+ e-EPIX1 The reported coefficients in the last column of Annex 3 are partial derivatives indicating the change in the probability of repeating a grade in primary education relative to a unit change in one of the independent variables, a p P (1 - P) where P refers to the dependent variable-probability of the event, beta to the logit coefficient and X to the string of independent variables used in the regression. For example, every extra year of mother's education decreases the probability of the child repeating a school grade by more than six percentage points in Guatemala, relative to a mean repetition rate of 77.9 percent. The results of the logit analysis are used to estimate probabilities of repetition against selected sample characteristics. Probabilities of repetition are simulated by varying one characteristic at a time, while holding other variables constant at their mean levels. The results of the simulations are presented in Table 3. Highlights are given in Figure 1. As expected, the variables age and indigenous origin have a large, positive, and significant effect on being a repeater. The older the child is, the greater the probability of being a repeater; from 12 percent at age 8, to over 52 percent at age 14 in Bolivia, and from 55 percent at age 10, to 94 percent at age 14 in Guatemala. A child of indigenous origin has at least a five percent greater probability of being a repeater than does a non-indigenous child in Bolivia, and a more than nine percent chance of being a repeater in Guatemala. 14 Table 3. Predicted Probability of Repetition by Selected Sample Characteristics (%) Characteristic Bolivia Guatemala Age 8 11.6 10 21.0 55.3 12 35.0 81.3 14 52.2 93.9 Gender Male 21.8 78.2 Female 23.0 77.5 Ethnicity Indigenous 28.5 83.5 non-Indigenous 21.8 75.1 Type of School Private School 14.9 Public School 25.3 Mother's Schooling (Years) 0 33.5 90.0 6 24.2 60.7 12 16.8 21.0 Family Income Quintile Bottom 20 22.3 79.0 Next 20 22.3 78.7 Middle 20 22.4 78.3 Next 20 22.4 77.8 Top 20 22.6 76.6 Household Head Gender Male 21.7 78.6 Female 30.2 73.0 Running Water Yes 19.8 75.1 No 24.9 80.8 Kitchen Yes 22.0 78.4 No 24.2 77.5 (CoUT 15 Table 3 (Cont'd) Characteristic Bolivia Guatemala Household Head Occupation Laborer 24.3 Employee 18.8 Self Employed 25.1 78.3 Employer 13.1 78.0 Private Sector Employee 79.3 Public Sector Employee 70.2 Reglon/Departamento Chuquisaca 12.8 Cochabamba 14.9 Oruro 16.0 Potosi 18.3 Tarija 18.5 Santa Cruz 18.1 Beni 31.4 Urban 70.1 Rural 82.6 Number of Siblings 0 16.0 62.4 2 19.0 74.0 4 22.4 83.0 6 26.2 89.3 Number of Rooms 1 27.7 84.2 2 25.1 81.1 3 22.4 77.5 4 20.4 73.6 5 18.4 69.2 6 16.5 64.4 Repeater (overalU probability) 24.2 77.9 16 The findings associated with indigenous people have been verified in previous analyses, especially for Guatemala. Rojas (1991) found that being "non-white" significantly affects educaticnal attainment in Guatemala. Carvajal and Morris (1989/1990), analyzing 1986 data from Guatemala's Programa Nacional de Educacion Bilingue (PRONEBI) from 297 communities and from a questionnaire administered to the same communities, found sizable differences anong indigenous groups' with respect to grade repetition and dropout, ranging from 30 to 46 percent repetition, and 6 to 16 percent drop out rates. The authors attempted to explain the differences with the use of community socioeconomic characteristics and differences among indigenous groups. It was found that bilingualism improves grade repetition and drop out rates, and that the opportunity cost of child labour should be taken into account when devising school schedules. In a country where more than 40 percent of the population begins school without knowledge of Spanish, there is a serious need for teachers with indigenous language knowledge. The promotion standards in Guatemala have been found to be unrealistic. It is usually necessary for first grade students to learn to read and write Spanish or else they repeat the grade; this, no doubt, is very difficult for those who do not know Spanish when they enter first grade. The Bilingual Education Project had a significant impact on promotion rates, more than nine percent higher for bilingual students relative to the control group in the first grade in 1983 (Townsend and Newman 1985). Region of residence also has a significant and large impact on repetition rates. Departamento of residence in urban Bolivia showed great differences, as did rural/urban sector in Guatemala. The probability of repetition is considerably lower in urban Guatemala. The number of brothers and sisters one has also has a positive and significant impact on being a repeater. The probability of being a repeater in Bolivia for those with no siblings is only 16 percent, while for those with 6 siblings the probability of repetition is over 26 percent. In Guatemala, this ranges from 62 to 89 percent. Attendance in a private school has a very large, negative and significant impact on being a repeater in Bolivia. The difference between public and private school repetition probabilities is great, at only 14.9 percent for private school students, and over 25 percent for public school students. We do not have information on type of school attended for Guatemalan children. In Bolivia, female headed households are associated with higher probabilities of repetition, compared with male headed households. The opposite relationship holds true in Guatemala, however, and the reasons are unclear at present. In addition, the probability of repetition was less for those Bolivian children where the household head was an employer, compared with the other occupational categories controlled for in the analysis. Mother's years of schooling also had a negative and significant impact on repetition, the probability of repetition in Bolivia declining from over 33 percent when the mother had no schooling, to less than 17 percent when the mother had 12 years of schooling. In Guatemala, the probability of repetition There are 25 different indigenous languages spoken in Guatemala (Carvajal and Morris 1989/1990). 17 Repetition probability (W) Bolivia 60 40.8 40 Non-lndig n g Pubilo Privat Bottom 20 TOD 20 Prim- Prim Sea. =thnio group 80htool type Family Inbomo. quintie Mlothor'o eduostion Repetition probability (%) Guatemala lOOr~ ~ ~ ~~~~~~~~~~2. 300 875. = ~~~~84.1 80 I60 f zf. Ryg ms 49.9 4 40 No .ind:ndi uPi vae Bottom2 20 Too 20 r m i _ Ethnic group SFam'i inyo F. quln tIne Mot,eS , Moher oidus NRepetition Probabilitie byut omlo eonomi 2 0Toprlmm Etnooop00,yIomquni ohr' doto Flue .Reeilo roailte b 0coecnmc hratrltc 18 drops dramatically for those whose mother has 12 years of schooling (21 percent), from the high registered for those whose mother has no schooling (90 percent). Parental education was found to have a positive association with school attendance in Guatemala (Balderston 1989), where it was also found that the opportunity cost for child labour was negatively associated with school attendance. Child malnutrition, as measured by height, was also found to be significantly related to schooling (see Glewwe and Jacoby 1992 for a similar study on Ghana). Interestingly, family income did not have much of an effect on repetition rates. The probability of repetition in both countries did not vary much among the five income categories presented in Table 3. This, however, shows that income is not a very good indicator of household wealth, this being reflected in the fact that other household wealth indicators did show significant and predicted differences in relation to repetition rates. For example, the presence of running water in the household was associated with a lower rate of repetition. Also, the more rooms in a house, the lower the rate of repetition, from 28 percent in a one room home, to 17 percent in a six room household in Bolia, and from 84 percent (one room) to 64 percent (six rooms) in Guatemala. V. Conclusion An empirical investigation into the causes of grade repetition (or overage) in Bolivia and Guatemala, while limited to socioeconomic and household factors, led to the findings that certain populations are more likely to be affected. Children from less wealthy households are more likely to repeat grades. Students of indigenous origins also have higher probabilities of grade repetition, suggesting that any targeting activities should have an indigenous component. A very large difference in repetiton rates was found between ypes of schools (in Bolivia), private versus public, favoring thse students attending the private schools. While this may have much to do with the so mic background of the students attending private schools, an investigation into the teaching and promotion practices of these schools is warranted. One simple method of reducing repetition is to introduce mandatory promotion. This, however, ignores the fact that learning occr tirough repetition, as is discussed by Gomes-Neto and Hanushek (1991). Continually high repetition rates, on the other hand, may also not be very efficient as they are associated with dropping out of school. The solution, therefore, rests in policies designed to get at retio and overage by targeting the right population: the poor. Moreover, in countres such as Bolivia and Guatemala, with large indigenous populations, the provision of bilingual educaon coUld be implemented quickly and extensively.2 20n bilingual eduation Latin America, see: Amadio (1990), Lopez (1990), Lopez and d'Emilio (1992), Albo and d'Emilio (1990), Moya (1990), Varese (1990), Gurdian and Salamanca (1990), Zuniga (1990), Martinez (1990), and Cossio (1991). 19 REEERENCE Albo, X. and L. d'Emilio. 1990. "Indigenous Languages and Intercultural Bilingual Education in Bolivia." Pro=ts 20,3: 321-30. Amadio, M. 1990. "Two Decades of Bilingual Education in Latin America (1970-90)." Pros=ts 20,3: 305-10. Balderston, J.B. 1984. "Determinants of Children's School Participation.' In J.B. Balderston et al. Malnourished Children of the Rural Poor. Boston: Aubum House Publishing Company. Behrman, J.R. and A.B. Deolalikr. 1991. "School Repetition, Dropouts, and the Rates of Return to Schooling: The Case of Indonesia." Oxford Bulletin of Economics and Statistics 53,4: 467-480. Berger, J. 1990. The Gaia Alas of First Peoples. New York: Doubleday. Bowman, M.J. and P. Goldblatt. 1984. "School Attainments in a Development Perspective: Transition Patterns in Mexico." Comparative Education Center, Department of Education, University of Chicago, Working Paper 84-6. Carvajal, M.J. and F.K. Morris. 1989/1990. 'Educacion Formal: tVehiculo de Integracion Economica y Social del Indio Guatemalteco." Revista de la Integracion y el Desarrollo de Croamnerica 45-46: 95-110. Cossio, C.Y. 1991. "The Implementation of Language Policy: The Case of Ecuador.' International Review of Education 37,1: 53-66. Cuadra, E. and B. Fredriksen. 1992. "Scope for Efficiency Gains Resulting from Reduction in Repetition and Dropout: A Simulation Exercise." World Bank, Population and Human Resources Department, Education and Employment Division, Document No. PHREE/92/55. Davico, M.I. 1990. "The Repeat and Drop-out Problem: A Study in Brazil on thu Role of the Teacher." ProaMts 20,1: 7-11. Glewwe, P. and H. Jacoby. 1992. "An Economic Analysis of Delayed Primary School Enrollment and Childhood Malnutrition in a Low Income Country.' Washington, D.C.: Population and Human Resources Department, The World Bank. Gomes-Neto, J.B. and E.A. Hanushek. 1991. 'The Causes and Effects of Grade Repetition: Evidence from Brazil.' Rochester Center for Economic Research, Worldng Paper No. 295. 20 Gurdian, G. and D. Salamnanca. 1990. "Bilingual Education in Nicaragua." Prospects 20,3: 357-364. Hernandez, I. 1988. "Identidad Indigena y Educacion." Desarrollo Economico 28,109: 121- 137. Ilon, L. and P. Moock. 1991. "School Attributes, Household Characteristics, and Demand for Schooling: A Case Study of Rural Peru." International Review of Education 37,4: 429-452. Jallade, J.-P. 1977. "Basic Education and Income Inequality in Brazil: The Long-Term View." World Bank Staff Worldkg Paper No. 268. Jamison, D.T. 1978. "Radio Education and Student Repetition in Nicaragua." In P. Suppes, B. Searle, and J. Friend, eds., The Radio Mathematics Project: Nicaragua 1976-1977. Stanford: Institute for Mathematical Studies in the Social Sciences, Stanford University. Kelley, J. 1988. "Class Conflict or Ethnic Oppression? The Cost of Being Indian in Rural Bolivia." Rural Sociolo 53,4: 399-420. Lopez, L.E. 1990. "Development of Human Resources in and for Bilingual Intercultural Education in Latin America." Proscts 20,3: 311-320. Lopez, L.E. and L. d'Emiio. 1992. "Bilingual Education Beyond National Frontiers. Bolivian-Peruvian Cooperation." The Maor Project of Education in Latin America and the Caribbean Bulletin 27: 41-56. Martinez, P.P. 1990. "Towards Standardization of T inguage for Teaching in the Andean Countries." Prospects 20,3: 377-386. Moya, R. 1990. "A Decade of Bilingual Education and Indigenous Participation in Ecuador." Proscts 20,3: 331-344. Organizacion de los Estados Americanos. 1992. Niveles Educativos Basicos en America Latina y el Caribe: Situacion Actual y Perspectivas. Documentos de Trabajo del Proyecto Multinacional de Educacion Basica (PRODEBAS), 1. Psacharopoulos, G. and H. Yang. 1991. "Educational Attainment among Venezuelan Youth: An Analysis of Its Determinants." International Journal of Educational Development 11,4: 289-294. Psacharopoulos, G., C. Rojas, and E. Velez. 1992. "Achievement Evaluation of Colombia's Escuela Nueva: Is Multigrade the Answer?", World Bank, Policy Working Papers No. 896. 21 Rojas, E. 1991. "Factores que explican los niveles de escolaridad media de los hijos en Guatemala." Santiago: UNESCO-OREALC. Cited in Schiefelbein 1992. Schiefelbein, E. 1992. "Repetition in Latin America's Primary Schools: Magnitudes, Causes, and Possible Solutions", The World Bank, LATHR (mimeo). Schiefelbein, E. 1989. "Repetition: The Constraint for Reaching Universal Primary Education in Latin America." Bulletin: The Major Project in the Field of Education in the Latin American and Caribbean Region 18: 7-28. Schiefelbein, E. 1975. "Repeating: An Overlooked Problem of Latin American Education." Comparative Education Review 19,3: 468-487. Schwille, J. et al. 1991. "Is Grade Repetition Always Wasteful? New Data and Unanswered Questions." Bridges Research Report Series 7 (January). Stephen, D. and P. Wearne. 1984. Centrl America's ndians with a summary by R. Stavenhagen. London: Minority Rights Group. Townsend, J. and B. Newman. 1985. "Bilingual Education Project in Guatemala: Preliminary Results on the Test of Principal Hypothesis during 1983." INCAP. Unesco, Office of Statistics. 1984. "Wastage in Primary Education from 1970 to 1980." ProV= 14, 3: 347-367. Varese, S. 1990. "Challenges and Prospects for Indian Education in Mexico." Prospcts 20,3: 345-356. Verhine, R.E. and A.M.P de Melo. 1988. "Causes of School Failure: The Case of the State of Bahia in Brazil." Pro[pects 18,4: 557-568. World Bank. 1986. Brazil: Finance of Primary Education. A World Bank Country Study. Washington, D.C.: The World Bank. Zuniga, M. 1990. "Educational Policies and Experiments among Indigenous Populations in Peru." Pros=ts 20,3: 365-376. 22 Annex 1. Mean Sampk Charaotristics, Bofvla 1989 Variable Mean s.d. N Individual Charaeterhtks Repeater .25 .43 6402 Age 10.28 2.18 6402 Years of Schooling 4.25 2.10 6402 Male .51 .50 6402 Indian .12 .33 6402 Private School .22 .42 6402 Region (Deparmeato) Chuquisaca .04 .18 6402 La Paz .38 .48 6402 Cochabamba .16 .37 6402 Oruro .10 .30 6402 Potosi .05 .22 6402 Tarija .02 .15 6402 Santa Cnuz .23 .42 6402 Beni .03 .17 6402 Household Characteristics Mother's SchooHng 7.33 4.30 5616 Father's SchooHng 8.54 4.39 5956 Family Income 675.75 1196.13 6402 Male Household Head .92 .27 6402 Siblings 4.06 1.72 6402 Rooms (A) 3.07 1.99 6402 Kitchen .79 .41 6402 Running Water .45 .50 6402 Household Head Occupation Laborer .14 .35 6402 Employee .36 .48 6402 Self Employed .33 .47 6402 Employer .04 .20 6402 Other .12 .32 6402 (Coul'd) 23 Annex 1 (Cont'd) Mean Sample Carcter, Guatemala 1989 Variable Mean s.d. N Individual Characteristics Repeater .71 .45 2974 Age 11.66 1.27 2974 Years of Schooling 3.22 1.54 2974 Male .54 .50 2974 Indian .29 .46 2974 Rural .58 .49 2974 Household Characteristics Mother's Schooling 3.20 2.70 2974 Father's Schooling 3.55 2.70 2974 Family Income 400.57 522.29 2974 Male Household Head .86 .34 2974 Siblings 2.78 1.37 2974 Rooms (#) 2.91 1.47 2974 Kitchen .43 .49 2974 Running Water .53 .50 2974 Household Head Occupation Private Sector Employee .37 .48 2974 Public Sector Employee .11 .31 2974 Self Employed .41 .49 2974 Employer .03 .17 2974 24 Annex 2 Zero-order Correlaion abtrix (Selected Variables) Variable SF SM YFAM Bolivia S .063 .080 .056 REPEAT -.243 -.231 -.057 SAGE .215 .230 .061 OVERAGE -.202 -.195 -.053 Guatemala s .429 .475 .249 REPEAT -.337 -.393 -.214 SAGE .437 .504 .255 OVERAGE .392 .458 .227 Note: All comlai oefficmn so udcally significant at tdo 1% probability level or boUr. 25 Annex 3 Lagisti Regresson Results: he Dr o Rhpeto Iin Dollvb (Dependent Variable: REPEAT) Loit Vuuiab Variable Coofficient M6m Moe (U) AGE 0.354 10.240 6.15 (20.3) MALE 4.071 .50 -1.23 (1.0) INDIG 0.356 .101 6.21 (3.3) C7quLsa1ca b-0701 .032 -12.19 (2.9) CiK9abaUnba 4.603 .166 -10.48 (5.3) Chqo -4.460 .099 4.00 (3.6) Potosi 4.26S .046 -4.61 (1.5) Tar(a 4.245 .024 -4.26 (1.0) Sama Cruz 4.352 .238 -6.14 (3.6) BanS 0.473 .032 8.24 (2.5) PRIVATE -0.661 .245 -11.49 (5.4) SM -0.075 7.328 -1.32 (7.0) YFAM(xlOO) 0.009 .708 .16 (0.3) MALEHHD -0.443 .913 -7.72 (3.7) LABORER 0.125 .138 2.17 (0.9) EMEPLOYEE -0.3SS .380 6.17 (3.0) SELFEMPLOYED 4.217 .323 -3.77 (1.8) EMPLOYER 40.680 .042 -11.82 (2.5) NSIBS 0.104 4.003 1.81 (4.9) NROOMS -0.132 3.125 -2.31 (5.5) KITC:HE3N 4.125 .796 -2.17- (1.4) RUNWATER 0.293 .467 -5.09 (3.7) Constant -3.425 Chi-!squa 1058 N N614 s Note. Numexns in panwldmYs _tvo t-valeK (CoWLdd 26 Anex 3 (Cont'd) Lisc Regresdon Re1ut: The Deteinauts of Repetition in Guatemala (Dependent Variable: REPEAl) Logit Variable Margsia Vaiable Coefficient Mean Effect (%) AGE 0.629 11.660 14.31 (14.3) MALE 0.037 .538 0.84 (0.4) INDIO 0.MIS .29S 11.72 (3.9) RURAL 0.708 .S77 16.11 (6.7) SM -0.293 3.199 -6.67 (12.3) YFAM(xlOOO) 40.186 .401 -4.23 (1.8) MALEHHD 0.304 .863 6.92 (2.0) PRIVATE SECrOR 0.074 .369 1.68 (0.4) PUBLIC SECTOR -0.412 .105 -9.38 (2.0) SELFEMPLOYED 0.017 .414 0.39 (0.1) NSEBS 0.037 2.782 0.84 (7.1) NROOMS -0.21S 2.912 -4.89 (5.2) KITCHEN 0.0S2 .429 1.18 (0.S) RUNWATER -0.33S .S32 -7.62 (3.1) Consdant -S.859 Chi-squaw 971 N 2974 Note: Numbers in p_mumu _puut t-valuus. Policy Research Working Paper Series Contact Tite Author Date for paper WPS1 006 Preparing Muftiyear Railway Jorge M. Rebelo October 1992 A. Turner Investment Plans: A Market-Oriented 30933 Approach WPS1007 Global Estimates and Projections Rodolfo A. Bulatao October 1992 0. Nadora of Mortality by Cause, 1970-2015 Patience W. Stephens 31091 WPS1 008 Do the Poor Insure? A Synthesis of Harold Alderman October 1992 C. Spooner the Literature on Risk and Christina H. Paxson 32116 Consumption in Developing Countries WPS1009 Labor and Women's Nutrition: Paul A. Higgins October 1992 C. Spooner A Study of Energy Expenditure, Harold Alderman 32116 Fertility, and Nutritional Status in Ghana WPS1010 Competition and Efficiency in Dimitri Vittas October 1992 W. Pitayatonakarn Hungarian Banking Craig Neal 37664 WPS1 011 How Tax Incentives Affect Decisions Robin Boadway November 1992 C. Jones to Invest in Developing Countries Anwar Shah 37754 WPS1012 The Brady Plan, the 1989 Mexican Haluk Unal November 1992 W. Patrawimolpon Debt Reduction Agreement, and Bank Asli Demirguc-Kunt 37664 Stock Returns in the United States Kwok-Wai Leung and Japan WPS1 013 The lmpact of Mexico's Retr;.ining Ana Revenga November 1992 D. Young Program on Employment and Wages Michelle Riboud 30932 Hong Tan WPs1014 Ethnicity, Education, and Earnings George Psacharopoulos November 1992 L. Longo in Bolivia and Guatemala 39244 WPS1015 Benefit Incidence Analysis in Thomas M. Selden November 1992 C. Jones Developing Countries Michael J. Wasylenko 37754 WPS1 016 Europe and Central Asia Region, My T. Vu November 1992 0. Nadora Middle East and North Africa Region Eduard Bos 31091 Population Projections, 1992-93 Ann Levin Edition WPS1 017 Rural Poverty, Migration, and the Richard E. Bilsborrow November 1992 WDR Environment in Developing Countries: 31091 Three Case Studies WPS1018 Tariff and Tax Reform: Do World Anand Rajaram November 1992 P. Pender Bank Recommendations Integrate 37851 Revenue and Protection Objectives? Policy Research Working Paper Series Contact Title Author Date for paper WPS1019 How Effective are Directed Credit Anita M. Schwarz November 1992 M. Raggambi Policies in the United States? 37664 A Literature Survey WPS1020 Another Look at Population and Nancy Birdsall November 1992 S. Rothschild and Global Warming 37460 WPS1 021 Measuring Welfare Changes from Jonathan R. Coleman November 1992 D. Gustafson Commodity Price Stabilization in Chris Jones 33714 Small Open Economies WPS1022 A New Approach to Evaluating Trade James E. Anderson November 1992 M. T. Sanchez Policy J. Peter Neary 33731 WPS1023 Tarfff Index Theory James E. Anderson November 1992 M. T. Sanchez 33731 WPS1024 An Exact Approach for Evaluating Will Martin November 1992 D. Gustafson the Benefits from Technological Julian M. Alston 33714 Change WPS1025 Openness and Economic Kazi M. Matin November 1992 D. Ballantyne Performance in Sub-Saharan Africa: 38004 Evidence from Time-Series Cross- Country Analysis WPS1026 Financial Liberalization and Paul D. McNelis November 1992 A. Maranon Adjustment in Chile and New Zealand Klaus Schmidt-Hebbel 31450 WPS1027 Lessons from Bank Privatization Guillermo Barnes November 1992 W. Pitayatonakarn in Mexico 37664 WPS1028 Socioeconomic and Ethnic Harry Anthony Patrinos November 1992 L. Longo Determinants of Grade Repetition George Psacharopoulos 39244 in Bolivia and Guatemala