Policy, Planning, and Remath WQRK ll Education and Employment Population and Human Resources Department The World Bank July 1988 WPS 27 Family Background and Student Achievement Marlaine E. Lockheed, Bruce Fuller, and Ronald Nyirongo Prior research has underestimated the influence of family back- ground on student achievement in developing countries. The Poi;cy. Plannin& and Research Cmpex distributesPPR Woding Papers to di_ninate the finding of wock in pogn and to encourge the cxchange of ideas among Bank staff and all others intested in developent issucs. Thee papas carry the names of the authors, eflect only their views. and should be used and cited accordingly. The findings. intsprutioUs. and canchisiom are the cuthors own. They should not be attributedto the World Bank, ita Board ofDirectg. its managancnt, or any of itsmanbercaiunmis. t Policy, PlannIng, and Re_arch Eduoatlon and Employm nt| Past research in developing countries has shown These variables were more consistently related that school-related influences have a greater to pupil achievement than were the conventional effect on student achievement than does family indicators, parental education and occupation. background, a finding that contrasts sharply with research in industrialized countries. This has led If, as these two studies indicate, family to the conclusion that schools in developing background is as important to students in countries are more effective than schools in developing countries as in industrial ones, two industrialized countries. types of action are suggested. First, education programs could be designed to take into account But the earlier work suffers from conceptual family background characteristics of students. flaws. It has defined family background in They might include early intervention programs, material terms and failed to consider other such as preschool or a change in school sched- motivational factors. Earlier research has also ules to better meet patterns of child labor. used measurements (such as level of parental Second, education systems could work to education and occupational status) more appro- improve student motivation and parental support priate to the industrialized world than to the directly by promotg the importance of educa- class structure of the country being studied. tion. Two studies of student motivational behav- In sum: Researchers should be more careful ior in Thailand and Malawi address these in their modeling of family and school character- shortcomings. In the Thailand study, conven- istics in the developing world. Failure to recog- tional measures of family background (parental nize the family's early and apparently lasting education and occupation) were kept constant. influence is a failure to accommodate education Student achievement in both urban and rural programs to indigenous realities. settings was related to such motivational vari- ables as educational expectations, attitudes and This paper is a product of the Population and effort. Human Resources Department, Education and Employment Division. Copies are available free The Malawi study employed definitions of from the World Bank, 1818 H Street NW, family background more relevant to a develop- Washington, DC 20433. Please contact Rose- ing country: labor demands placed on children, marie Rinaldi, room S-6027, extension 33278. basic attributes of houses, and mother tongue. The PPR Working Paper Series disseminates the findings of work under way in the Bank's Policy, Plaming, and Research Complex. An objective of the series is to get these findings ou. quickly, even if presentations are less than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official policy of the Bank. Copyrignt 0 1988 by the International Bank for Reconstruction and Development/he World Bank Table of Contents Family Background and Student Achievement 3 A Broader Conception of Pupil Background 5 Study 1: Thailand 8 Study 2: Malawi 17 Conclusions 21 family Effects on Stndent Achievement In Thailand and Malawi INTRODUCTION Past research from Third World countries shows that school-related factors have stronger effects on student achievement than do family background factors. This finding contrasts sharply with evidence from industrialized countries, where family background characteristics explain substantially larger proportiors of variation in educational. This has led to the optimistic inference that Third World schools &re more effective than schools in industrialized countries, both in raising achievement and in providing routes for social mobility. However, prior work suffers from conceptual and methodological flaws. Two important conceptual shortcomings are: (a) limiting the concept of family background to material aspect of class, and failing to consider other aspects of family background, such as motivational variables, that influence student performance, and (b) using measures of family background (such as parental educational attainment and occupational status) useful in for industrialized settings rather than indicators of class that are more culturally valid for the country in which the study is conducted. Methodologically, prior research suffers from: (a) being limited to cross-sectional data that are incapable of distinguishing initial level of achievement from achievement gain over time and that tend to confound the effects of school and family characteristics, and (b) seldom analyzing school and family effects separately for different regions within countries; specifically, past research has rarely included non-urban samples. We suggest that once these conceptual and methodological shortcomings are addressed, the influence of pupil background may b˘ greater in -1 - developing countries than the modest effects suggested by earlier findings. This paper presents the results of two studies of family background offacts on student achievement that, taken together, address the shortcomings of past research. We argue, first, that attention ahould be given to a greater variety of student background variables that may affect achievement by mediating, supplemknting or substituting for conventional background variables. In particular, we are interested in perceptions held by children that may motivate higher school performance. The first study, an analysis of longitudinal data from Thailand, explores the effects of several such motivational variables in explaining how family background influences student achievement over time. Second, we raise the question of whether the influence zfi stadent background on achievement can be adequately tested in the Third World by relying on conventional measures of class, borrowed from industrialized societies. Culturally specific indicators may show distinctly greater influence over student achievement than such imported indicators as parents' occupational status and formal educational attainment. The second study, an analysis of crossectional data from Malawi, examines how estimates of student background effects on learning can be improved by utilizing material indicators of class that are more relevant to local conditions. The paper is organized as follows. In this section, we review the literature on family effects on student achievement in developing countries. In the second section, we present Study 1 on Thailand, and in the third section we present Study 2 on Malawi. In the final section, we discuss our findings in relation to educational policy. - 2- FI dly background and student achiev_nt Education advocates in Third World governments and international agencies have been buoyed in recent years by evidence that the offect of school characteristics on student achievement, when compared with the effect of student family background, appears to be greater in Third World countries than within industrialized countries. For example in their widely cited reanalysis of International Education Association (IEA) and other data from 29 countries, Heyneman and Loxley (1983) report that--taken altogether--characteristics of schooLs in developing countries explain a greater share of the variance in student achievement than is explsined by student background characteristics. In addition, a significant negative correlation was found between the percentage of variance explained by school characteristics and the nation's level of wealth. As a case in point, they note that 27 percent of the variance in achievement among children in India was attributable to school factors while only 3 percent -as explained by variation in background characteristics. In contrast, achievement levels within industrialized countries were explained mostly by student background. Similarly, a recent review of approximately 60 empirical studies of determinants of achievement in developing countries found that school characteristics were related to student achievement in the vast majority of studies, after holding constant student social class background (Fuller, 1986). Yet there is strong evidence that student family background contributes significantly to both educational attainment and achievement in developing countries. Family background affects the probability that children enroll in, attend, and complete various levels of education. For example, analyzing historical data from the -3- philippines, Smith and Choung (1986) fou,nd that parental occupational and educational level shaped children's school attainment, with the some level of magnitude, since the early 20th century. Similarly, using household income as a rough proxy for student social class, Chernichovsky and Meesook (1985) found significant educational attainment effects in Indonosia. In Nepal, Jamison and Lockheed (1987) found that parental and grandparental social class (landholding, ca-te, schooling) strongly determined child school participation. Adult literacy was *lso measured, and found to be strongly associated with years of schooling, a finding recently repeated in Brazil (Psacharopoulos and Arriagada, 1987). This study underscores tho importance of separating school attainment (enrollment and persistence) from level of academic achievement while in school as two distinct outcomes. Students' family background also affects lkarning while in school. For example, Schiefelbein and Simmons (1981) found that social class significantly helped predict achievement in 28 of 37 Third World studies reviewed. Moreover, family background may be quite salient for familiar subjects, such as reading, and less so for subjects relatively less familiar to many Third World communities, such as science. On this point, Schiefelbein and Simmons noted that student family characteristics accounted for a higher proportion of variance in reading achievement than in science achievement in multivariate stu .ies from India, Peru, and Malaysia. And Heyneman and Loxley's findings, while widely celebrated, applied only to achievement in science. Third, family background appears more strongly to deternine student achievement within urban schools, whereas school factors have - 4 - greater influence on the achievement of rural and va-y poor children (Brazil, Wolff 1970; Iran, Ryan 19731 Indonosia, Haron 1977). This pattern may result from less variation in social class within rural azdas or to the fact that Western measures of class are even less valid in rural areas (compared to Third World urban areas). In sum, family backgro appears to be more influential in developing countries when (&, the formal school is highly institutionalized, (b) class structures are more defined, and (c) in subjects that are familiar or linked to parentO' own knowledge. In contrast, when the school is foreign to the setting and teaches unfamiliar areas of knowledge (such as, a rural school providing science instruction), school characteristics appear to be more efficacious than family background. A Broader Conception of Pupil Background Studies of family background effects on student achievement in Third World settings would be improved by the inclusion of two types of variables: (a) a broader range of family background indicators, and (b) more culturally valid indicators of social and material background. Broader range of indicators. Most fundamentally, the construct of "pupil background" should be broadened to include other aspects of parents and the child that may operate independently of material aspects of class. We focus on two sets of factors in this regard. First, the value placed on staying in school and achieving well has been found to vary among parents within several societies, including the United States, Japan (Hess and Holloway 1984), Mexico (holloway et al. 1986), and Taiwan (Stevenson et al. 1985). Parents' own reports -5- of value placed on schooling and the frequency with which they communicate this priority is *s*ociat*d with higher achievement in reading and mathematics independent of parents' occupational and educational own schooling (at least among young children in Japan and the U.S.; Hess et al. 1980). This communication of value may take the form of expectations for higher levels of school attainment, greater support (both material and non-material) for learning, or communication regarding the relevance of education for adult occupational attainment. Yet these factors have received slight attention from researchers working within developing countries. A notable exception is Jamison and Lockheed's (1987) research on determinants of child schooling in Nepal, which examined a number of such variables, including p_..atal attitudinal modernity and demand for child education; both were found to have strong and posit..ve effects on child school participation. In a parallel study undertaken in Thailand, Cochrsne and Jamison (1982) found that parental aspiration for educa,-ion was the most important predictor of male educational attainment, while parental landholding (an indicator of wealth) was the most important predictor of female educational attainment. Another example is provided by a study of urban Grade 10 students from Sri Lanka, (Niles, 1981) which found that family social class (parental education and occupation) and parental interest in the child's progress and parental aspirations were strongly associated with student achievement, and taken together with material possessions in the home (non-significant) and school factors (only modestly significant) accounted for 49% of the variance in student achievement. Second, pupils vary in the effort they expend on school work and the extent to which they feel efficacious in performing well. This is -6- an old idea in the school effects literature. Coleman et al (1966), for instance, found that students who perceived that their achievement was due to their own efforts, end not to their teachers or school, achieved at higher levels (controlling for their class background). More recently, Brookov-r *t al. (1979) found that pupils achieved at higher levels when they felt that their effort on school work was recognized and rewarded by their teachers (again, independent of pupils' social class). Natriello (1987) also found that whera students' work was more frequently and carefully evaluated, achievement levels were higher. For a recent review of factors related to mathematics and science achievement, s*e Lockheed et al. (1985). Relevant social class indicators. Most studies of family background characteristics in developing countries have employed two indicators of social class: parental educational attainment and occupational status. As H-yneman has noted, in Third World settings there may be little variation in terms of education or occupation; educational levels of adults are uniformly low, and occupations are preponderantly related to rural agrici lture (at least in the lowest income countries). Yet social class differences are quite pronounced, and have been found to strongly effect child school participation and performance. For example, in the Jamison and Lockheed study, which employed indicators of parents' social class that were relevant to the Nepalese setting -- amount of land owned by the father, his level of literacy, caste membership, and district of residence, capturing inter-regional differences in wealth -- child school participation was strongly determined by these social class elements. - 7- In sum, the true influence of pupil background on achievement can be better estirated if we take into account: (a) those family background characteristics -- garental values and press to achievement and the child's own effort and perceived efficacy -- that appear to operate independently of conventional aspects of class (for the few countries in which theme factors have been studiedi, and (b) better indicators of material and *octal aspects of class. STUDY 1: THAILAND Our first analysis examines family effects on Grade 8 mother s achievement gain in Thailand, and focuses on a broader range of background indicators, specifically motivational variables. Method Sample. Data were drawn from the Second Interntional Mathematics Study, conducted under the auspices of the International Association for the Evaluation of Educational Achievement. The IEA SIMS sample comprised 99 mathemathics teachers and their 4030 eighth-grade students and was derived from a two-stage, stratified random sample of classrooms. The primary sampling units were the twelve national educational regions of Thailand plus Bangkok. Within each region, a random sample of lower-secondary schools was selected, with replacements. At the second stage, a random sample of one class per school was selected from a list of all eighth grade mathematics classes within the school. The esulting sample represented a 12 sample of eighth grade mathematics classrooms within each region. At both the beginning and end of the school year, students were administered a mathematics test covering five curriculum content areas -8- (agrithmetic, algebra, geometry, statistics and measurement). ftudents also completed a short background questionnaire at the pretest and a longer ort at the poattest administration. In the following sections, a description of each of the variables analyzed in this paper is provided; definitions of variables and summary statistics are presented in Table 1. Mathematicn achievement measures. The IE developed five mathematics testi for is in SIMS, one forty-item instrument (the core test) and four thirty-five item instruments (rotated forms A through D). The five test instrtments contained roughly equal proportions of items from each of five mathematics curriculum content areas. In Thailand, students were pretested using the core test and or. rotated form. At posttest, students again took the core test and one rotated form, but were prevented from repeating the rotated form taken at pretest. In this analysis, we created formula scores from the four rotated forms, after they were equated (through the core form) to adjust for differences in test length and difficultys scores were adjusted for guessing. A complete description of the equating and score development procedure is provided in Lockheed, Vail and Fuller (1987). Family backzround indicators. Basic information about each student included his or her sex, age, number of older siblings, paternal and maternal education, paternal and maternal occupational status, and home language. Parental occupation was classified into four international categories: (a) unskilled or semi-skilled worker, (b) skilled worker, (c) clerical or sales worker, and (d) professional or managerial worker. Because paternal and maternal occupational status were highly correlated (r - .39) we analyzed the effects of paternal -9- occupational status only in this paper. Highest parental education was also classified into four categories: (a) vary little or no schooling, (b) primary school, (c) secondary school, and (d) college, university or some form of tertiary education. Because paternal and maternal educational attainment were also highly correlated (r - .58), we aaalyzed the *affects of maternal educational attainment only. District level per capita income was derived from World Bank estimates. Student educational expectations were measured by a single item that asked about the number or year more of full-time education the student expected to complete. Parental encouragement was measured by a four-item index composed of Likert-type statements a-king students to describe their parent's intarest in and encourageme for mathematics achievement, for example "My parents encourage me to learn as much mAthematics as possible"; response alternatives ranged from "Exactly like" the student's parent (- 1) to "Not at all like" the student's parent (- 5). The four items comprised a single factor, with principal component factor loadings ranging from .72 to .83 and communality of 2.43. A low score represented greater perceived parental support. Student attitudes. Three indices of student attitudes were analyzed: (a) perceived mathematics ability, (b) perceived usefulness of mathematics, and (c) motivation toward mathematics achievement. All were developed from a factor analysis of the student attitude survey, which contained Likert-type items having response alternatives ranging from "Strongly disagree (- 1)" to "Strongly agree ( 5)." Factors were initially identified through varimax factor analyses, and then confirmed through principal component analyses, from which factor scores were constructed. Perceived mathematics ability (or mathematics - 10 - self-efficacy) was formed from five items (e.g.: "I am not so good at mathamatics") having principal component factor loadings ranging from .63 to .79 and communality of 2.55; a low value represtnted a positive attitude, due to the reverse wording of the items. Perceived usefulness of mathematics was formed from eight items (e.g.: "Mathematics is important to get a good job") having principal component factor loadings ranging from .37 to .61 and communality of 2.26. Motivation was measured by three items (e.g.: "I want to do well in mathematics") having principal component factor loadings ranging from .71 to .81 and communality of 1.77. Stu.dent effort was measured by the number of hours the student reported spending on homework during the previous week. Results This section is divided into four sections. First, we report the effects of family background on pretest mathematics achievement, using a conventional cross-sectional approach, and on achievement gain, taking advantage of the longitudinal design of the study. Second, we examine family background effects on the mediating variables of student educational expectations, perceived parental encouragement, attitudes and effort; this analysis is conducted first without statistically controlling for student pretest achievement and then second with such statistical controls. Third, we report the effects of family background and mediating variables on achievement gain. Finally, we report differences for rural and urban schools. Family back2round effects on student achievement. Cross-sectional analyses of family background effects on student achievement typically find statistically significant, albeit modest, effects. Our cross-sectional analyses of determinants of mathematics achievement in Thailand are consistent with previous findings. We found statistically significant effects on both pretest and posttest scores for paternal occupation, maternal education and district-level per capita income and less strong but statistically significant effects for use of language of instruction at home and for its interaction with district-level income (Table 2, columns 1 and 2). Students having fathers with more professional occupations and mothers with higher levels of education had higher levels of eighth grade mathematics achievement. Students in wealthier districts and those in families using the language of instruction at home also scored higher on both tests. The interaction between district wealth and use of language of instruction at home was statistically significant, with the negative sign on the coefficient indicating that these two variables substitute for one another. The total variance in achievement explained by family background variables, however, was modest, with 7Z explained at both pretest and posttest. Our longitudinal analysis, moreover, showed little effect of family background effects on posttest, once pretest achievement was statistically controlled. Of the family background factors examined, only district-level per capita income exerted further influence on student achievement gain (Table 2, column 3). In the next section, we examine why this apparent lack of effect is misleading. Family backiround effects on parental support, student attitudes and effort. In the previous section we noted that conventional measures of family social class background, although correlated with initial and final levels of eighth grade mathematics achievement, contributed little to student achievement Rain over time. Presumably, - 12 - this was because by the eighth grade, family characteristics had completed their influence over achievement. However, families continue to influence student achievement by providing material and non-material support for learning activities, by raising children's educational expectations, and by reinforcing student motivation and effort. In this section, we examine the extent to which such important motivation variables are affected by family background. Family background effects on six motivation variables are reported in Table 3. Overall, little variance in parental support, student attitudes or effort was explained by conventional social class background variables (variance explained ranged from 0 to 62). However, several effects were statistically significant, and are reported here. First, we found significant gender-related. effects. Boys reported lower educational aspirations, less parental encouragement, lower valuing of mat!hematics and less motivation in mathematics than did girls. However, they reported higher perceived mathematics ability than did girls, although in actuality, there were no sex differences in achievement. Second, conventional social class measures were significantly related to the mediating variables. Higher paternal occupational status and maternal educational attainment levels were positively related to higher levels of student educational aspirations, perceived parental encouragement, and. motivation. In addition, paternal occupational status was positively related to student perceptions of mathematics as useful, and maternal education was positively related to greater student effort. Finally, district per-capita income was positively related to educational aspirations. - 13 - Effects of backaround factors on motivation variables, with initial achievement held constant. After the effect of initial mathemtics achievement was statistically controlled, family background variables continued to affect five of the six student attitudes and perceptions; they had no effect on student effort (Table 5). With protest achievement held constant, boys still reported higher level of perceived ability than girls, and lower levels of perceived parental encouragement, educational expectations, motivation and p-rce-ved utility. With initial achievement held constant, paternal occupation and maternal education were still significantly related to student *ducational expectations, perceived parental support and motivation. Surprisingly, the higher the district-level per-capita income, the less students reported mathematics as being relevant to their future employment. Effects of motivation variables on student l1arnint zain. As we noted previously, the main impact of family social status and student background characteristic was through initial performance levell once established, background characteristics had little additional effect on achievement gain (Table 2, column 3). By comparison, student educational expectations, student attitudes and student effort were positively related to learning zain (Table 4). Controlling for the effects of pretest achievemont and family characteristics, five of the six motivation variables--the exception was perceived parental support--were positively and significantly related to posttest student achievement. The fact that perceived parental support is unrelated to achievement gain is surprising. It may be the students are unaware of parental support or that parents - 14 - may not have specific opinions about mathematics (as compared with reading, for example). These findings have important implications for understanding achievement among students in developing countries, where many students come from rural low-income families lacking material supports for learning. Most research has attended to differences in material support between low-incomee and higher income children. The argument is made that greater material support is available to students from wealthier, better educsted families and less to students from poorer, less educated families. This difference in material wealth has been claimed to explain the clear differences in achievement between the two groups. The evidence of this study, however, suggests that there may be a different explanation. Students from wealthier and more educated families, who are largely urban residents, may perceive that stchooling is more important for thu5r futures than do students from rural settings. Schooling is more compatible with urban students' goals, and therefore parental support and student attitudes have more positive effects for them. To explore this question further, we analyzed the effects of family background on student achievement gain, separately for urban and rural students. Urban rural differences. To test the hypothesis that parental support and student attitudes would have stronger effects for urban students than for rural students, we separatesd the sample into two groups, those from urban districts and those from rural districts. All analyses were reconducted, but only selected ones are reported in Tables 6 and 7. - 15 - Family background accounted for relatively little protest achievement variance in either type of school (62 for students in urban schools and 7X for students in rural schools). The effects of maternal education, paternal occupation, sex and district per-capita income were statistically significant, although not consistent for the two groups (Table 6, columns 1 and 2). For both urban and rural children, maternal education was strongly associated with pretest achievement. For urban children only, paternal occupational status was associated with pretest achievement, while for rural children only, district-1. el per capita income was positively associated with pret it achievement. The failure to find significant sex differences in achievement for the total sample may be explained by the rural-urban differences. In urban schools, girls outperformed boys by 1.2 points, but in rural schools, boys outperformed girls by 1.0 points. Since the urban and rural samples were approximately equal in size, these effects would cancel each other out. Holding constant stadent background characteristics, we found that student educational expectations, attitudes, motivation and effort were positively related to student achievement gain in both urban and rural settings; also, in both settings, perceived parental encouragement was unrelated to learning gain (Table 7). In both urban and rural settings, family background characteristics that were significantly related to initial levels of achievement were unrelated to gain, except insofar as they operated through other variables. - 16 - STUDY 2: MALAWI Our second analysis examines social class effects on achievement among a modest sample of 105 primary school students in the east African nation of Malawi. Here we focus on material facets of class that are relevant to a particular Third World setting, moving beyond Western indicators. Method Samle. Fourth- and seventh-grade students from 11 urban and 10 rural primary schools (most Malawi primary schools include 8 grades) were randomly selected and interviewed, as part of a project evaluation conducted by World Bank staff. Interviews were conducted in the local tribal language, although government staff doing the interviews did not necessarily come from the local tribe. Schools selected were spread across the country's three administrative regions. Given resource constraints, selected schools tended to be close to paved roads, which biased the sample, and under-represented children of remote subsistence farmers. Nevertheless, variation in class background was great, as detailed below. Achievement measures. Students were tested in both mathematics and in Chichewa, the state-sanctioned national language. Each exam consisted of ten items, developed by a group of primary schools teachers and reflecting the curriculum they covered in their own classes. Items from the pool developed by the teachers were field tested and a final set of items, with a reasonable range of difficulty were selected for inclusion in the final instrument. The Chichewa exam included prima-ily vocabulary items. The mathematics exam included seven one-step computational problems and three, more - 17 - difficult, word problms. Since students from grades 4 and 7 were included, raw test scores were converted to standardized scores, and grade-level was included as a control variable in the regression models. Setting-specific indicators of social class. In the Malawi study, we focused on material aspects and measures of class that are more relevant to Third World settings, moving beyond the global constructs historically used within industrialized nations. For instance, we asked pLpils about seven different work tasks they might engage in after school. Four dealt with farm-related activities (feeding livestock, carrying fodder or water). Other tasks wore more general, taking care of siblings or preparing meals, for instance. We also asked pupils and their headmasters about housing conditions, particularly whether families lived in thatched-roof huts or in Western structures. Students were asked whether their houses had electricity and whether their parents owned a radio. Students also identified their mother tongue. We were especially interested in whether their parents spoke Chichewa, the dominant tribal language, or one of the more than 35 other tribal languages spoken in Malawi. In addition to these situation-specific measures of class, conventional indicators were also used: parents' occupation and schooling levels. Analysis. Our analysis examined the influence of these indigenous measures of class on math and language achievement. OLS estimates were run for all students, then separately for students in rural and urban schools. Regression models for all students included a control on whether he or she attended an urban or a rural school. In some instances, attendance at an urban school was neaatively related to achievement. Thus, while urban children generally come from families - 18 - of higher social class, the over-crowded and under-staffed nature of urban primary school may offset the (urban) advantages of family background. A full analysfi of this issue is beyond the scope of the present paper. But this effect did necessitate at least controlling on the type of school attended. Results Table 8 reports on definitions and mean levels of all variables included in the analysis. The sample included an over-representation of families in skilled, modern-sector jobs. Nationwide, only 32 percent of all adults are engaged in non-farm occupations nationwide, and this includes household heads that have small plots but commute" to the city to work as street vendors or in semi-skilled wage jobs. Yet a disproportionate number of sampled students reported that their fathers were employed in skilled occupations, a mean of 2.7 on the occupational scale that ranged from white-collar professional (-4) to subsistence farmer (-1). Headmasters estimated that 42.7 percent of all families came from farming households and that 56.2 percent lived in thatched-roof houses. The strength of this sample is that variation is substantial. A representative sample would have yielded a distribution heavily skewed toward subsistence farmers, constraining the normality of distributions. Nevertheless, caution is warranted in generalizing broadly from our results. Note the mean scores on the mathematics and Chichewa exams (our dependent measures). For Chichewa, the mean was 5.5 with a standard deviation of 2.3; for mathematics, the mean was 3.8 with a standard deviation of 2.1. Below you will see that our findings are - 19 - more robust in explaining variation in Chichews achievement. This may be due, in part, to the better distribution of , it scores. Table 9 reports alternative specifications for estimating pupil achiovement from our various indicators of social class. The first column reports a regression of the total test score on *11 measures of class. Pupils coming from houses with electricity did better on the achievment exams. The interaction of having more work tasks after school and living in a thatched-roof structure negatively influenced achievement. Curiously, th poercent of families coming from farming backgrounds (as estimated by the headmaster) was positively related to achievement. The direction of this effect is difficult to explain. It may indicate that parents who own their land, compared to farmhands on estates or unskilled urban workers, provide stronger encouragement of school achievement. This variable might also be a stronger proxy for attendance at a rural school, which are typically smaller and display a lover student:te*cher ratio than urban schools. In sum, the full model explains 21 percent of the variance in achievement with moderate derees of freedom. Statistically insignificant (p).10) exogenous variables were then allowed to step out of the model, to test for stability in the coefficients held by the significant predictors. As expected the r-squares dropped following this procedure. Similar models were run for the Chichewa and mathematics scores separately. For Chichewa, family background explained a larger share of the variance. The full model explained 27 percent of the variance. The strong negative effect from urban-school attendance arises here. The region's wealth and housing with electricity exerted a positive influence on Chicheva achievement. The interaction of after-school - 20 - work tasks and residing in a thatched-roof hut again was negatively related to achievement. Models estimating mathematics achievement were less efficacious in explaining variation. Table 10 reports reduced models, splitting the sample between students attending urban and rural schools. In general, social class factors more strongly explain achievement in rural schools. This is especially true with regard to performance on the Chichewa exam. Rural pupils whose mother tongue was Chichewa, not surprisingly, did better on the language test. Rural pupils from homes with electricity and radios also performed at higher levels. R-squares reached 35 percent in explaining variation in rural pupils' performance, although th, degrees of freedom are quite low for the split sample. In sum, indicators of class more relevant to the Third World setting -- labor demands placed on children, basic attributes of houses, and mother-tongue -- were more consistently related to pupil achievement than were global, Western proxies. This study does not inform us as to how these largely-material facets of class operate on the child to lower school achievement. However, the consistent effects of situationally-relevant measures of class suggest that researchers working in Third World settings have inadequately specified family background factors in the past. Studies that employ only constructs of class from Western industrialized settings may retsult in underestimating the actual achievement effects of class. CONCLUSIONS This paper has examined the effects of two types of social class background factors on enhancing the academic achievement of children: (a) motivational characteristics of families that add to, mediate or - 21 - substitute for material background character±sties, and (b) more valid and culturally relevant indicators of social class and mAterial family background characteristics of children. The Thailand study provide strong evidence that motivational variables influence achievement. Holding constant conventional indicators of student backgtound (gender, parental education and occupation) we found that student educational expectations, attitudes, motivation and effort were positively related to student achievemsnt gain in both urban and rural settings. In both urban and rural settings, conventional family background characteristics that were significantly related to initial levels of achievement were unrelated to gain, except insofar as they operated through the motivational variables. The Malawi study demonstrates that indicators of family background more relevant to the Third World setting, such as labor demands placed on children, basic attributes of houses, and mother-tongue, were more consistently related to pupil achievement than were conventional indicators of parental eduzation and occupation. If family background is as salient in developing countries as it is in industrialized countries, what implications does this have for educational policy makers or development advocates in planning aducation programs in the Third World? These two studies suggest at least two important avenues for action. First, education programs could be designed that specifically take into account the family background characteristics of students. This might involve early intervention programs, such as preschools; in-school intervention programs, such as changing school hours and calendars to better meet local patterns of child labor; or - 22 - alternative school programs, such as distance education. Second, education systems could undertake to improve student motivation and parental aupport directly, through local informational meetings and other mechanism to promote awareness of the importance and benefits of education. In general, these types of interventions have been tried, with various degree of success, in developing countries, particularly with respect to the education of girls (Lycette, 1986). The specifics of appropriate interventions will undoubtedly be t_ermined by local conditions, but a clearer understanding of how families affect achievement in Third World settings will better inform appropriate educational decisions. Finally, researchers should more carefully model the influence of family and school characteristics of achievement in the Third World. Misspecification of family background antecedents will lead to over-estimates of the school's real effect, especially when levels of class and school quality are colinear. Optimism about the school's potential influence in the Third World need not be dampened. But if we fail to recognize the family's early and apparently lasting influence, we will unknowingly fail to accommodate educational programs to indigenous family and class realities. - 23 - Tabim 1: Variable Namms. oniaMrtiam. mmm andi Standard Owatiamlow Thluiland. 1291-2 Samie Variable Oescriptiona Totalb RuralC Urband Iban S.D. Mean S.O. Mean S.D. YROT Posttest mth score 11.89 4.46 0.52 4.32 13.09 4.50 XROT Pretest mth scare 8.58 3.77 7.38 3.57 9.71 3.85 XSEX Sex (lufemale; 2Ie) 1.53 .24 1.54 .25 1.52 .24 XAGE Age (In amnths) 171.13 4.48 172.15 4.31 170.20 4.53 YFOCCI Father's occupatlonal status (11low; 4hil0g) 2.39 .45 2.25 .40 2.52 .48 YMEDUC Mother's educatlonal attalrunt (1.10w; 4h10Ih) 1.95 .38 1.80 .33 2.09 .41 HCALC Calculator at hore (lbyes) .23 .22 .21 .20 .38 .23 SPCI81 01st, Iet Per Capita lnOr (In bhats) 13495.08 2547.38 9208.07 1247.U 17363.0 1839.15 YHLANG Lse langme of Instructlon at hame (1 yes) .52 .24 .47 .25 .56 .24 YMOREED Educatloai elaectatlos (1-low) 3.71 .46 3.59 .50 3.82 .42 YPARENC Parental mncouragnt (1uhlgi) 2.14 .4 2.11 .47 2.16 .46 YPERCEV Perceived math ability (1uhi0g) 4.06 .43 4.08 .42 4.04 .44 YFUTUR Percolved future importance of math (1ulo) 2.08 .42 2.08 .43 2.05 .42 YDESIRE Motivatlon to suceed In math (1low) 5.60 .35 5.53 .37 5.61 .34 Notes: (a) For a complete descriptl')n of variables, coult text. (b) MlninA sample size for total sample is 3747. (c) MIninA sample size for rual saiple Is 1893. (d) MinInu sample size for urban sample Is 2054. - 24 - Table 2: The effect of student background characteristics on Grade 8 protest and posttest mathematics achievement Thailand, 1981-82 Independent variables Protest Poattest Poattest Posttest Pretest score -- -- .82*** 80*** (59.04) (54.05) Sex (2 - boy) -.17 .18 .32 (.68) (.62) (1.45) Age .26 .36 .14 (1.01) (1.14) (.61) Age squared .00 -.00 .00 (1.38) (1.59) (.89) Father's occupation .69*** .59*** .03 (4.82) (3.41) (.24) Mothers education .89*** *97*** .26 (5.30) (4.79) (1.70) Calculator at home .46 .62 .25 (1.63) (1.84) (1.00) District per capita ircome (1000 bh) .28*** *35*** 12** (7.11) (7.36) (3.55) Lang. of inst. at home 1.57* 2.33** 1.08 (2.18) (2.70) (1.67) Dist. PCI x Lang of inst. -.13* -.00** -.09* (2.48) (3.10) (2.01) Constant -12.57 -14.87 4.71 4.05 Adjusted R-squared .07 .07 .48 .49 N 3642 3627 3801 3625 ***p < .001, **p < .01, * p < .05. Note: Unstandardized regression coefficients; t-statistics in parentheses. - 25 - Table 3t The effect of student background characteristic* on student expectations, perceived parental support, perceived ability, attitudes and effort in Thailand, 1981-82 Independent YMOREED PARENC YPERCEV YFUTURE YDESIRE YMWHKL variab les Sex (2 - boy) -.14*** .27*** -.10** -.09** -.21*** .04 (4.35) (8.54) (3.19) (2.99) (8.78) (.31) Age -.00 -.02 .03 .02 .03 .00 (.08) (.66) (.88) (.55) (1.36) (.03) Age squared -.00 .00 -.00 -.00 -.00 -.00 (.34) (.84) (.8_) (.66) (1.51) (.15) Father's occupation .12*** -.06** .00 .04* .05** -.09 (6.66) (3.30) (.24) (2.39) (3.28) (1.20) Mothers education .12*** -.07** -.03 .02 .06*** .23* (5.91) (3.30) (1.50) (1.14) (3.68) (2.54) Calculator at home .05 .00 -.02 .03 .01 .13 (1.31) (.10) (.48) (.84) (.21) (.85) District PCI (1000 bh) .01* .01 -.00 -.00 .01 .02 (2.07) (1.44) (1.94) (1.26) (1.37) (.85) Lang. of inst. at home .00 -.16 -.21* -.07 .03 -.52 (.01) (1.77) (2.39) (.85) (.39) (1.35) Dist. PCI x Lang of inst. .00 .00 .00* .00 -.00 .00 (.01) (1.95) (2.50) (.42) (1.46) (.83) Constant 3.72 3.24 1.83 .89 2.99 4.61 Adjusted R-squared .06 .04 .00 .01 .04 .00 N 3628 3597 3599 3582 3617 3574 ***p < .001, ** p < .01, * p <.05. Note: Unstandardized regression coefficients; t-statisties in parentheses. - 26 - Table 4: The effect of student background characteristics on student expectations, perceived parental support, perceived ability, attitudes and effort, with pretest achievement controlled, in Thailand, 1981-82 Independent YHOREED PARENC YPERCEV YFUTURE YDESIRE YMWHKL variables Pretest .02*** -.01*** -.04*** .03*** 01*** .06*** (11.89) (3.40) (18.60) (14.04) (9.35) (6.40) Sex (2 - boy) -.13*** .27*** -.10*** -.08** -.21*** .05 (4.28) (8.50) (3.50) (2.90) (8.77) (.40) Age -.00 -.02 .04 .01 .03 -.01 (.13) (.59) (1.24) (.34) (1.22) (.07) Age squared -.00 .00 -.00 -.00 -.00 -.00 (.06) (.76) (1.28) (.37) (1.31) (.02) Father's occupation 10*** -.05** .03 .02 .04** -.13 (5.84) (3.03) (1.82) (1.30) (2.58) (1.71) Mothers education .10*** -.06** .00 -.00 .05** .18* (4.96) (3.01) (.05) (.04) (2.88) (1.99) Calculator at home .04 .01 .00 .02 -.00 .10 (1.01) (.19) (.09) (.52) (.07) (.67) District PCI (1000 bh) .00 .01 .00 -.01** .00 -.03 (.68) (1.84) (.15) (2.97) (.26) (1.60) Lang. of inst. at home -.04 -.15 -.15 -.12 .00 -.61 (.44) (1.65) (1.86) (1.42) (.03) (1.59) Dist. PCI x Lang of inst. .00 .00 .01 .01 -.01 .03 (.53) (1.80) (1.90) (1.06) (1.05) (1.10) Constant 4.05 3.12 1.35 1.22 3.19 5.11 Adjusted R-squared .10 .04 .09 .06 .06 .01 N 3626 3595 3597 3580 3615 3572 ***p < .001, ** p < .01, * p <.05. Note: Unstandardized regression coefficients; t-statistics in parentheses. - 27 - Table 5: The effect of student background characteristics, pretest achievement, student expectations, perceived ^rental support, perceived ability, attitudes and effort on Grade 8 mathemat..cs poattest achievement in Thailand, 1981-82 Alternative specifications dependent (1) (2) (3) (4) (5) (6) ariables etest math achievement .78*** .80*** .75*** .77*** .78*** .79* (51.94) (53.51) (49.06) (50.93) (52.52) (53.04) x (2 - boy) .43 .35 .14 .42 .61** .36 (1.93) (1.56) (.63) (1.89) (2.72) (1.61) e .16 .10 .23 .03 .10 .16 (.69) (.44) (1.01) (.14) (.45) (.68) e squared -.00 -.00 -.00 -.00 --.00 -.00 (.95) (.74) (1.31) (.42) (.73) (.96) ather's occupation -.05 .04 .08 .03 -.01 .09 (.41) (.29) (.64) (.27) (.09) (.71) thers education .19 .24 .23 .23 .19 .21 (1.28) (1.59) (1.60) (1.54) (1.29) (1.3, Calculator at home .22 .23 .26 .21 .26 .27 (.89) (.89) (1.03) (.84) (1.04) (1.05 District PCI (1000 bh) .12*** .13*** .13*** .14*** .12*** .14* (3.45) (3.61) (3.63) (4.13) (3.57) (3.89) .ang. of inst. at home 1.09 1.13 .96 1.27* 1.10. 1.31* (1.70) (1.74) (1.51) (1.99) (1.72) (2.02) Dist. PCI x Lang of inst. -.00* -.00* -.00 -.00* -.00 _.00* (2.06) (2.10) (1.77) (2.26) (1.91) (2.35) Educational expectations .80*** -- -- -- -- -- (6.58) arental encouragement -.02 -- -- -- -- (.83) 'erceived math ability -1.47*** __ (11.47) Future importance of math 1.22*** -- -- (9.45) Motivation 1.10*** -- (7.06) Student effort .13*** (4.74) Constant -8.78 -.71 -5.37 2.41 -7.53 -6.53 Adjusted R-squared .49 .49 .51 .50 .49 .49 3609 3579 3581 3564 3599 3555 ***p < .001, ** p < .01, * p <.05. Note: Unstandardized regression coefficients; t-statistics in parentheses. - 28 - Table 6s Effect of hom background, parental support, Student educational expectations, student attitudes, and student offort on Grade 8 mathematics pretest achievement in urban and rural schools, Thailand, 1981-82 Ind-pondent School Location Variables Urban Rural Sex (2 - boy) -1.22*** 1.03** (3.43) (2.94) Age .19 .46 (.59) (1 00) Age squared .00 .00 (.92) (1.18) Father's occupation .84*** .36 (4.47) (1.62) Mothers education .77*** 98*** (3.52) (3.65) Calculator at home .39 70 (1.03) (1.62) District PCI (1000 bh) .00 27* (.05) (2.23) Lang. of inst. at home -3.07 .68 (1.64) (.41) Dist. PCI x Lang of inst. .15 .27 (1.34) (.04) Constant .55 -32.57 (.02) (.81) Adjusted R-squared .06 .05 N 1995 1646 *** p < .001, ** p < .01, * p <.05. Note: Unstandardized regression coefficients; t statistics in parentheses. - 29 - Table 7s Family background effects on Grade 8 mathematics achievement gain in Thailand, 1981-82a Independent School Location Variables Rural Urban YMOREEDC 71*** 92*** (4.31) (5.14) YPARENC .01 - 05 ( .04) ( .30) YPERCEV -1.30*** -1.66*** (6.76) (9.62) YFUTURE 1.17*** 1.27*** (6.21) (7.21) YDESIRE *75*** 1.47*** (3.38) (6.76) YMHWXL .11** 16*** (2.74) (3.92) Notes (a) This table summarizes the results of 12 regression analyses, identical in design to those reported in Table 2. (b) Numbers are parameter estimates with t-statistics in parenthesis. (c) Effects controlling for student age, sex, paternal occupation, maternal education, household calculator ownership, household language use, and district per-capita income. - 30 - Table 8s Malawi Studys Variable Definitions and Descriptive Statistics Variable Moan SD Raw Pupil Exam Scores Chichewa (10 items) 5.5 2.3 Maths (10 items) 3.8 2.1 Pupil Exam Z-Scores Chichewa 0.00 0.99 Maths 0.00 1.00 Exogenous Predictors Urban school (2-urban) 1.54 __ Region's wealth, Z labor force 32.0 4.4 employed in non-agricultural jobs Pupil sex (2-girl) 1.49 __ Pupil's grade-level 5.5 1.5 Pupil's work tasks, five possible 3.1 0.5 farm related jobs (e.g. fted animals) Pupil's mother tongue is Chichewa, 1.6 -- language of high-status tribe (2-yes) Pupil's father's work 2.7 1.1 (4-point scale by skill level) Pupil's mother's schooling level 1.7 0.5 (2-yes, completed standard 4) Pupil's mother's reading practices 6.6 1.2 (10-point scale, e.g. reads books) Pupil's house has electricity 1.4 -- (2-yes) Pupil's parents own radio 1.8 -- (2-yes) Headmaster's estimate I pupils from 42.7 32.9 farming families Headmaster's estimate I pupil living 56.2 37.0 in thatch-roof huts Note: Standard deviations not reported for dichotomous variables. - 31 - Table 9: lnfluhI of Couitry-Specific Social Class Mseures on PUplI Achleveuent In Malawi ToTAL TiiT SCE CHINCHE1A ST SARE MATH ITST SCaE (1)", (2)" (3)0 (1)a (2)" (3) (1)a (2) Jrban school -.33 -.10 - -0.81 -.88 - .49 .64 _ control) (10.54)d ( 0.03)e (-2.31)' (-11.14)" (1.28) ( 3.38) on's wealth -.005 .03 .04 _.03 (-0.15) (1.52) ( 3.85)0 (-1.18) (girl.l) -.28 -.07 -.22 (-0.87) (-0.37) (-0.88) rade level -.05 -.02 a.02 (-0.48) (-0.37) (-0.39) other tongue .51 .21 .30 hichewa ( 1.56) ( 1.07) (1.38) er's work .17 .15 .02 ( 0.95) ( 1.3) ( 0.18) er's schoolIng .18 .41 .35 -.23 ( 0.46) ( 1.72) ( 2.62) (-0.88) er's readIng -.15 -.10 -.03 tices (-0.82) (-1.05) (-0.11) I's work tasks -.005 -.005 -.002 -.003 -.001 -.001 -.003 -.003 -.003 thatch roof (-2.24)' (-5.68)' (-4.92)* (-1.84)+ (4.38)" (-5.71)' (-1.69) (-4.92)' (-5.46)* Nuse w/electricity 1.01 1.08 1.13 .59 .42 .78 .43 ( 2.16)* ( 8.07)** ( 9.39)"s ( 2.05)' C 2.85) (10.50)" ( 1.38) idio In house .37 .26 .12 ( 0.89) ( 1.02) ( 0.42) s estimate fathers .02 .02 .02 .01 .01 .02 .01 .01 Ing ( 2.75)" ( 7.42)"0 (12.39)*' ( 1.52) (11.46)08' ( 2.78)" ( 7.09)" ( 3.62 s estimate homes .02 .02 .006 .02 .02 .02 'thatch roofs (1.53) ( 2.93) ( 0.70) (1.67) ( 5.17)' 4.50 cept -2.13 -1.69 -1.93 -1.36 -.67 -1.25 -.77 -1.58 -.20 13,88 5,96 3,98 13,88 5,98 3,100 13,88 4,99 3,100 *square .21 .16 .14 .27 .20 .14 .17 .10 .07 :e: Lhnstandardlzed regression coefficlients reported. Full model Redjced model with urban control Reduced model without urban control T-values are reported for full models F-values repwrted for reduced models where backward elimination of non-slIgnlflcant exogenous variables was run. + pc.07 * pc.05 **p.01 "** p<.001 x.10 used as criterion for variable to remain In the equatlon. - 32 - Table 10: lnfluence Of CoUntry-Specif ic Social Clans ksasures on PLPII Achlevewet In Urban and Rural Schools In alawl TOTAL TEST SCORE CHICHENA TEST SCORE MATH TEST SCORE Urban Rural Urban Rural Urban Rural RegIon's wealth .09 (8.00)" Sex (girl-1) -.59 (-5.37)0 Grade level Mother tongue 1.22 .56 .65 Chichewa (7.90)* (5.19)' - (6.06)- father's work Mother's school Ing .55 (4.30)' Mother's readIng practloes Pupil's work tasks -.004 x thatch roof ( 5.57)* House w/electrIcity 1.21 .52 1.06 (8.28)" (4.16)' (5.74)' Radio In house .52 (4.23)* H's estlaato fathers .02 .013 farmlng (6.15)' (5.21)' H's estIlate ho es -.01 .04 w/thatch roofs (5.95)' ( 5.49)* Intercept -2.40 1.80 -3.54 -2.82 -0.23 -2.03 Of 2,55 1,42 3,58 5,38 1,56 2,42 R-sqyare .16 .16 .26 .35 .09 .22 Note: Lhstandardized regresslon coeffIcIents and f-values reported. PN.05 used as crIterIon for varlable to remain In equatlon. - p<.05 " pc.01 *- P<.C.O1 - 33 - References Armitage, J., Batista, J., Harbison, R., Holsinger, D. & Leite, R. (1986). School quality and achievement in rural Brazil. (Discussion Paper No. EDT25) Washington, D.C: The World Bank, Education and Training Department. Brookover, W., Beady, C., Flood, P., Schweitzer, J. and Wisenbaker, J. (1979). School social systems and student achievement: Schools can make a difference. New York: Praeger. Chernichovsky, D. and Meesook, 0. (1985). School enrollment in Indonesia. World Bank Staff Working Paper no. 746. Washington, D.C.: The World Bank. Cochrane, S. and Jamison, D. (1982). Educational attainment and achievement in rural Thailand. In A. Summers (Ed.), New directions for testing and measurement: Productivity assessment in education. No. 15. San Francisco: Jossey-Bass. Coleman, J., Campbell, E., Hobson, C., Mood, A., Weinfeld, F. and York, R. (1966). Equality of educational opportunity. U. S. Department of Health, Education and Welfare, U. S. Office of Education. Washington, D.C.: Government Printing Office. Fuller, B. (1986). Raising school quality in developing countries: What investments boost learning? Discussion Paper No. 2. Washington, D.C. The World Bank. Haron, I. (1977). Social class and educational achievement in a plural society: Peninsular Malaysia. Unpublished doctoral dissertation, University of Chicago. Hess, R. D., Kashiwagi, K., Azuma, H., Price, G. G. and Dickson, W. P. (1980). Maternal expectations for mastery of developmental tasks in Japan and the United States. International Journal of Psychology, 15: 259-271. Hess, R. D. and Holloway, S. D. (1984). Families and schools as educational institutions. In R. Parke (Ed.) Review of Child Development Research. Volume 6, Chicago: University of Chicago Press. Holloway, S. D., Kashiwagi, K., Hess, R. D., and Azuma, H. (1986). Mothers' causal attributions about mathematics: A comparison of Japan and the United States. Internation Journal of Psychology. Heyneman, S., Jamison, D. and Montenegro, X. (1983). Textbooks in the Philippines: Evaluation of the pedagogical impact of a nationwide investment. Educational Evaluation and Policy Analysis, 6 (2), 139-150. - 34 - Heyneman, S. & Loxiey, W. (1983). The effect of primary school quality on academic achi-vement across twenty-nine high and low-income countries. American Journal of Sociologv, 88 (6), 1162-1194. Jamison, D. and Lockheed, M. (1987). Participation in schooling: Determinants and learning outcomes Economic Development and Cultural Chante, 35, (2), 279-306. Jamison, D., Searle, B., Galda, K. and Heyneman, S. (1981). Improving elementary mathomatics education in Nicaragua: An experimental study of the impact of textbooks and radio on achievement. Journal of Educational Psychologv, 73 (4), 556-567. Lockheed, M., Thorpe, M., Brooks-Gunn, J., Casserly, P. and McAloon, A. (1985). Sex and ethnic differences in middle school mathematics.science and computer science: What do we know?. - Princeton, N.J.: Educational Testing Service. Lockheed, M., Vail, S. and Fuller, B. (1986). How textbooks affect achievement in developing countries: Evidence from Thailand. Education Evaluation and Policy Analysis, 8(4), 379-392. Natriello, G. (1987). The impact of evaluation processes on students. Educational Psychologist, Niles F. (1981) Social class and academic achievement: A third world reinterpretation. Comparative Education Review, 25, 419-430. Psacharopoulos, G., and Arriagada, A.M. School participation, trade attainment, and literacv in Brazil: A 1980 census analysis. (Discussion Paper EDT _). Washington, D.C.: The World Bank, Education and Training Department Ryan J. (1973). Educational resources and scholastic outcomes: A study of rural -rimary schoolini in Iran. Unpublished doctoral dissertation, Stanford University. Schiefelbein, E. and Simmons, J. (1981). Determinants of school achievement: A review of research for developing countries (mimeo). Ottawa: International Development Research Centre. Smith H. and Cheung, P. (1986). Trends in the effects of family background on educational attainment in the Philippines. American Journal of Sociology, 91, (6), 1387-1408. Stevenson, H., Stigler, J., Lee, S. and Lucker, G. (1981). Cognitive performance and academic achievement of Japanese, Chinese and American children. Child Development, 56, (3), 718-734. WattanL&waha, N. (1986). A study of equity in mathematics teaching and learning in lower secondary school in Thailand. Unpublished doctoral dissertation, University of Illinois, Urbana-Champaign. - 35 - Wolfe, B. and Behuman, J. (1984). Who ia schooled in developing countries? The rolez of income, parental schooling, sex, residence and family size. Economics of Education Review, s, (3), 231-224. Wolff, L. (1970). Why childron fail In the first -rade in Rio Grande Do Suls Imolications for policy ond research. Report propared for the U. S. Agency for International Development, Human Resources Office, Rio de Janeiro, Brazil. - 36 - PPR Working Paper Series Title Author Date Contact WPS7 Institutional Analysis of Credit Cooperatives Avishay Braverman April 1988 C. Spooner J. Luis Guasch 37570 WPS8 Prospects for Equitable Growth in Rural Sub-Saharan Africa Steven Haggbiade April 1988 C. Spooner Peter B. Hazell 37570 WPS9 Can We Return to Rapid Growth? Andrea Boltho June 1988 J. Israel 31285 WPSIO Optimal Export Taxes for Exporters of Perennial Crops Mudassar lmran June 1988 A. Kltson-Walters Ron Duncan 33712 WPSI1 The Selection and Use of Pesticides In Bank Financed Public Health Projects Bernhard Liese June 1988 C. Knorr Norman Gratz 33611 WPS12 Teacher-NonTeacher Pay Differences in Cote d'lvoire Andre Komenan June 1988 R. Vartanian Christiaan Grootaert 34678 WPS13 Objectives and Methods of a World Health Survey Trudy Harpham June 1988 A. Menciano lan Timaeus 33612 WPS14 The Optimal Currency Composition of External Debt Stljn Claessens June 1988 S. Bertelsmeier 33768 WPS15 Stimulating Agricultural Growth and Rural Development in Sub-Saharan Africa Vijay S. Vyas June 1988 H. Vallanasco Dennis Casley 37591 WPS16 Antidumping Laws and Developing Countries Patrick Messerlin June 1988 S. Torrijos 33709 WPS17 Economic Development and the Debt Crisis Stanley Fischer June 1988 C. Papik 33396 PPR Working Paper Series Title Author Date Contact WPS18 China's Vocational and Technical Training Harold Noah June 1988 W. Ketema John Middleton 33651 WPS19 Cote d'lvoire's Vocational and Technical Education Christiaan Grootaert June 1988 R. Vartanian 34678 WPS20 Imports and Growth in Africa Ramon Lopez June 1988 61679 Vinod Thomas WPS21 Effects of European VERs on Japanese Autos Jaime de Melo June 1988 S. Fallon Patrick Messerlin 61680 WPS22 Methodological Problems in Cross- Country Analyses of Economic Growth Jean-Paul Azam June 1988 E. Zamora Sylviane Guillaumont 33706 WPS23 Cost-Effective Integration of Immunization and Basic Health Se-vices in Developing Co'n,ries: The Problem of Joint Costs A. Mead Over, Jr. WPS24 World Bank Investments in Vocational Education and Training John Middleton July 1988 W. Ketema Terri Demsky 33651 WPS25 A Comparison of Alternative Training Modes for Youth in Israel: Results from Longitudinal Data Adrian Ziderman WPS26 Changing Patterns in Vocational Eaucation John Middleton July 1988 W. Ketema 33651 WPS27 Family Background and Student Achievement Marlaine E. Lockheed July 1988 R. Rinaldi Bruce Fuller 33278 Ronald Nyirongo WPS28 Temporary Windfalls and Compensation Arrangements Bela Balassa June 1988 N. Campbell 33769