WAorld Bank Reprink Series: Number 408 Dean T. Jamison and Marlaine E. Lockheed Participation in SchoolLng Determnants and Learning Outcomes in Nepal Reprinted with permission from Econtioic Development and Cultural Change, vol. 35, no. 2 (anuary 1987), pp. 279-306, published by tIie University of Chicago Press. Participation in Schooling: Determinants and Learning Outcomes in Nepal* Dean T. Jamison World Bank Marlaine E. Lockheed World Bank I. Introduction Substantial evidence suggests that the level of adult education attained is related to such dimensions of development as individual earnings and employment in the urban labor markets, agricultural productivity, human fertility, and health and nutritional status.' Nonetheless, there are relatively few empirical studies of the mechanisms underlying these effects, In order to understand better such mechanisms on develop- ment outcomes, this paper uses data from rural Nepal to examine the effects of education on such potentially mediating variables as adult cognitive competencies and attitudes.2 A further potentially important consequence of schooling and improved cognitive competence of adults is an increase in the desire for schooling for their children. This paper reports on this and other determinants of school participation, as well as its cognitive consequences. The article is organized as follows: Section II presents background material concerning the determinants of adult cognitive cornpetencies and attitudes and of child school participation; Section III describes the data; Section IV examines the determinants of adult attitudes and cognitive competencies; and Section V examines the determinants of child school participation. The sixth section summarizes our conclu- sions. HI. Background DETERMINANTS OF ADULT COGNITIVE COMPETENCIES There have been numerous and diverse studies attempting to relate school outcomes to variations in the amount or quality of school in- @ 1987 by The University of Chicago. All rights reserved. 0013-0079/87/3502-0004$01 .00 280 Economic Development and Cultural Change puts. The findings of these studies have sometimes been interpreted to imply that schooling does not make a difference. Yet there is growing evidence that the number of years spent in school does affect both adult cognitive competencies and adult attitudes, which, in turn, are determinants of child school participation. Although these studies pro- vide little guidance concerning how to improve the quality of school- ing, they do conclusively show the enduring impact of schooling on various dimensions of cognitive capacity. Relatively few of the avail- able empirical studies concerning the effects of years of education are from less developed countries, and one purpose of this paper is to use data from Nepal to help fill this gap. Two recent reviews have summarized most of the relevant litera- ture. Harnqvist has summarized studies from the United States and Sweden, and Sheffield has summarized studies from a number of low- income countries.3 The overwhelmingly conisistent finding of these studies was that years of schooling affect literacy (reading and writing), numeracy, and modernity.4 One important shortcoming of the available literature is that, with data gathered only on adults, it is difficult to ascertain the extent to which an observed correlation between completed schooling and cog- nitive competence may result, in part, from the plausible hypothesis that more innately able individuals would both attend school longer and, in any case, perform better on tests of adult competence. Although it is difficult to assess with complete adequacy the extent to which this hypothesis is correct, by including a test of "ability" along with our other, more achievement-oriented measures, we are able, in part, to control for this effect. DETERMINANTS OF SCHOOL PARTICIPATION School participation in Nepal is low for all but primary school males, and total e.nrollment rates for children 9-15 are well below 20% (table 1). In general, little empirical work on the determinants of school en- rollments in less developed countries has been conducted; Birdsall provides a valuable overview of available evidence.5 Since male pri- mary school enrollment is now common (though far from universal), much work has recently been directed toward explaining female enroll- ments. Reviewing these studies, Bowman and Anderson cite five major determinants of female enrollments-ethnic and regional differences, caste differences, paternal occupation, paternal education and attitu- dinal modernity, and such miscellaneous factors as foreign travel or language usage.6 Such factors have also been shown to be related to positive parental attitudes toward schooling for children. Research on determinants of school participation in Nepal in par- ticular is very limited. The research that does exist suggests that-in Dean T. Jamison and Marlaine E. Lockheed 281 TABLE I SCHOOL ATTENDANCE IN NEPAL BY SEX AND SCHOOL LEVEL SCHOOL LEVEL Lower Primary Secondary Secondary Age of students (years) 6-8 9-12 13-15 Total population: 1,096,548 1,331,035 771,698 Estimated female 548,274 (50%) 665,512 (50%) 385,849 (50%) Estimated male 548,274 (50o) 665,512 (50%) 385,849 (50%) Total enrolled: 769,049 226,639 82,158 Female enrolled 169,640 (22%) 41,788 (18%) 13,598 (17%) Male enrolled 599,409 (78%) 184,851 (82%) 68,560 (83%) Total enrollment rate (%): 70 17 10 Female enrollment rate 31 6 4 Male enrollment rate 108 28 17 SouRCE.-Computed from Nepal: Primary Education-a Subsector Study, Report no. 135 (Paris: Unesco, October 1978), annexes 1, 2. addition to the factors cited by Bowman and Anderson-household economic factors, student health and nutritional status,7 school avail- ability, and school quality and relevance are all important. In the next subsections we discuss the factors commonly believed to affect school participation, with particular reference to conditions in Nepal. School Availability Lack of school supply could arise because there is no school within walking distance, because the available schools are overcrowded, or because the quality of available schooling is low. Walking distance, School availability in Nepal has increased dramatically in the past 30 years. In 1951, there were 321 primary schools, enrolling less than 1% of eligible children; in 1975, there were 8,708 schools, enrolling 59% of the children;8 by 1982, the number of primary schools had risen to 9,404.9 Since there are large numbers of primary schools in Nepal, it would seem that distance to school is not a major problem on the average. There probably is, however, a substan- tial minority of primary school-age children in rural areas for whom distance to primary school is a problem. At the lower-secondary level the number of schools drops by more than 75%, and distance to school may therefore be more of a problem. Shrestha reports, however, that, although distance to school was related to school participation in Gorkha District, it was reported by parents in Jumia, Dhankuta, and Chitawan districts to be one of the least important reasons for not sending children to school.10 Crowding. The average student-to-teacher ratio is low for both 282 Economic Development and Cultural Chainge primary and secondary levels (nationally only about 32; in the districts examined in this paper [Bara and Rautahat], 28 and 30). Thus, over- crowding probably poses no problem for most students at either level. School quality and relevance. Another important problem for primary and lower-secondary schooling, particularly in rural areas, is the students' lack of access to schooling that is relevant to their lives and of a quality sufficient to make attendance worthwhile. In the rela- tively few studies of determinants of school-leaving behavior in devel- oping countries, the poor quality of teachers has emerged as a principal reason for sludents' dropping out." School quality affects student dropout propensity both directly and indirectly through increasing the incidence of repetition, which itself causes higher dropout rates. Direct evidence about the quality of learning in Nepal's primary schools is limited. One recent study completed by the Institute of Education,"2 however, suggests that the quality of learning is quite low and reflects the very poor conditions for learning: untrained teach- ers, insufficient materials, and monotonous teaching methods. Many teachers lack the school-leaving certificate eamed after tenth-grade graduation, and perhaps only one-third of all teachers have training beyond the tenth grade. Teaching materials are scarce, although the impressive effort now under way to prepare and distribute textbooks and teachers' guides for all subjects at all grade levels may help to improve the situation. Teaching methods are traditional, with classes dominated by teacher lectures and unembellished readings from text- books. Students are expected to respond in unison to the teacher's questions, to recite aloud from their texts, and, in general, to mem- orize. Students are rarely asked to relate the content of their les- sons to their own experiences and perceptions and are rarely expected to analyze, synthesize, or consider questions of cause rather than ques- tions of fact. Students are not encouraged to ask questions.'3 Demandfor Schooling Lack of parental demand for education could arise from out-of-pocket costs for schooling, the opportunity cost of student's time, and/or pa- rental attitudes regarding schooling. Out-of-pocket costs. These include direct tuition expenses and the cost of books. Under Nepal's New Education System Plan, the government finances all of a primary teacher's salary and three-fourths of that of a lower-secondary teacher, except in remote areas, where all the costs of lower-secondary teachers are borne by the government. It has been estimated that costs to the student for primary education are about Rs 15 per year for copybooks, pencils, and examination fees. The addition of school uniforms, estimated to cost Rs 70 each, in- creases the direct cost of primary school attendance to Rs 90-Rs 300 annually. The costs to a lower-secondary student are about Rs 78 for Dean T. Jamison and Marlaine E. Lockheed 283 copybooks, pencils, examination fees, and books. Although in remote areas teacher and book costs will be somewhat more subsidized than these numbers would indicate, students in these areas are far less able to afford cash expenditures, as total household cash income in rural Nepal is only Rs 1,500-Rs 2,000 annually. The out-of-pocket cash expenditures can be expected to be a significant barrier to students' access to schooling, particularly at the lower-secondary level. Shrestha found that low family income was the most important reason reported for parents not educating children in the Chitawan District and the next most important reason in the Dhankuta District.'4 Cost of student's time. At both the primary and lower-secondary levels, the time that a rural student spends in school must frequently be at the expense of his or her doing useful work at home, 5 Studies show that the demand for female child labor is higher than the demand for male child labor. Acharya and Bennett report that, in eight representa- tive districts in Nepal, girls aged 5-9 work 3.4 hours per day and that girls aged 10-14 work 7.3 hours per day, which is 50% more than the number of hours that boys in the same age groups work. '6 A study of farm households In Kabre Palanchowk found that 25% of girls and 13% of boys aged 6-14 were employed in farm labor.'7 A study undertaken in Pokhara, Nepal, and reported in Kasaju found that Youngsters who do not come to school need to work for their parents, . . . need to support their family's economy and . .. cannot complete the school because the number of hours they are required to stay at school does not match with the number of hours they can afford to spend. In a study, "A survey of parental reasons for not sending their children to primary school in Pokhara Town Panchayat Area," Mr. Upadhya has pointed out the need for a child to work in support of the family as one of the most frequently listed reasons for non-enrollment. The most fre- quently stated reason, states Mr. Upadhya, was that the parents did not see any value in educating girls.'8 This finding is confirmed by Shrestha, who reports that the primary reason parents in Jumla, Gorkha, and Dhankuta do not wish to educate their children is that there would be a "lack of supporting members for household."'9 The second most frequently mentioned reason for not educating girls in Jumla, Gorkha, and Chitawan is that "girls are to be given away in marriage." Attitudes. According to Ministry of Education data, in 1977, only approximately 30% of the age cohort of female children were involved in primary schools. Female school enrollment in rural areas has been shown to be related to parental attitudes, particularly parental conservatism.20 The evidence from Shrestha and Upadhya suggests that parental conservatism may be one factor that restricts girls' educa- tion in Nepal. A recent study of parental attitudes in Dohkha, Bank, 284 Economic Development and Cultural Change and Kathmandu districts reported that "the general notion of these parents about girls was that girls would not remain an asset to them all along, which implies that higher investment in their education was not warranted from the practical point of view.,"21 THE CAUSAL MODEL In the present paper, we examine the effects of these fa ;.ors across three generations. Figure I illustrates the causal structure. fhe vari- ables in boxes A-D are entirely exogenous and assumed to affect second- and third-generation variables in boxes F-H; the variables in box E-first generation landholding and literacy-are also exogenous, but they are separated to clarify the intergenerational structure of this model. The variables in box A-district, caste, and age-and in box E EXOGENOUS FACTORS CROSS-GENERATION FACTORS J'ERACY & -ANDHOLLDNG FIRST GENEPAr;ON ------------------------- ----- SE-COND GENERAIION ZPM AN 1 ,E117. ~c ASXA AJTUDE 4,AILABIL '_1AP7_ SC --XQL SECOND GENERATION - -- - -- -- -- -- ---- -- -- THIRD GENERATION .N ~AMIL, '.I4 $CHOULt I'AJ1TICtPAtlIqP FIG. I.-A schematic model of the determinants and outcomes of schooling Dean T. Jamison and Marlaine E. Lockheed 285 are assumed to influence "innate" ability as measured by the Raven's Progressive Matrices (RPM) test; this variable, combined with the ex- ogenous variables, is assumed to influence the amount of schooling an individual receives, the amount of land currently owned by the individ- ual, and the adult cognitive competencies of that individual. All the preceding variables taken together, plus the availability of schools (box B), are assumed to influence the adult attitude toward school (box G). Finally, all these variables, taken in combination with the number of children in the household (box C) and the child's sex (box D), are assumed to affect the third-generation's school participation. III. Data Data for this analysis were obtained from a survey of 795 households studied as part of a World Bank research project that examines the effects of education and agricultural extension on rural development. The site chosen for the project was determined by the major research objectives concerning the effects of education on agricultural produc- tivity. The households to be interviewed were chosen randomly within each of six panchayats in two districts-Bara and Rautahat. A list of owners of all dwellings in each panchayat was obtained from the local rural health workers (these lists had been prepared in connection with the health programs). Then, using a three-digit, random number table obtained from the Nepal Central Bureau of Statistics, households to be interviewed were selected randomly within a panchayat until a 15% sample of households in each panchayat had been chosen (10 extra households per panchayat were chosen for replacement). Thv study households were visited three times, first in October- November 1977, next in January-March 1978, and last in April-May 1978. Field-workers for the interviewing were recruited by the mem- bers of the New ERA research team in Birgunj, Kalaiya (district head- quarters of Bara), and Guar (district headquarters of Rautahat). The final field team comprised seven males and seven females, who worked in pairs. All the field-workers spoke both Nepali and Bhojpuri; they translated the questions from Nepali (in which the instruments were written) to Bjohpuri in the field. Table 2 shows the calendar of data collection and indicates the overall nature of the data collected. Information obtained from the study households at one or more of the three field visits includes data on household characteristics, ag- ricultural productivity, fertility, nutritional status, and various educa- tion measures of household members. The availability of primary and secondary schools in the sample villages, as well as the distance to primary and second schools, was obtained in May 1979. This informa- tion was available for 23 of the 28 villages studied. In this paper, data from three generations are analyzed, with first- TABLE 2 DATA COLLECTION INFORMATION Round I (October-November 1977) Round 2 (January-March 1978) Round 3 (April-May 1978) 1. Household roster 1. Household roster (married) 1. Children health and nutrition status re- vised 2. Household information (including assets, 2. Education information for members of health, and nonfarm income) household only (literacy, numeracy, mo- dernity, Raven's Progressive Matrices) 3. Children's nuiritional and health status 3. Background, employment, and marital status of members of household over age 14 4. Fertility questions for married women, 4. Health, nutrition, and family planning part of Rautahat knowledge and attitudes 5. Fertility question for narried women, part of Rautahat and all of Bara Dean T. Jamison and Marlaine E. Lockheed 287 generation data on literacy and landholdings used to predict sec- ond-generation cognitive competencies and with first- and second- generation data used to predict third-generation school participation. In the former analysis, the record for the head of the household pro- vides the unit of analysis, whereas, in the latter analysis, both house- hold and child records serve as units of analysis. A child was defined as a son or daughter of the household head, a nephew or niece of the household head, a grandchild or great-grandchild of the household head, or an adopted child. These children are the third generation. Younger siblings of the head of the household, aunts or uncles, sisters- in-law, daughters-in-law, servants, and unrelated children were ex- cluded. Third-generation data include a measure of school attendance as a deviation from the average attendance of the child's age cohort and an indicator of the child's sex. The second generation is the child's parent who is the head of the household. Second-generation data in- clhide information about this household head: the total number of years of schooling completed, a measure of literacy, a measure of numeracy, a measure of basic ability (Raven's Progressive Matrices Test [RPM], a measure of attitudinal modernity (a modified version of the short form of the Inkeles Overall Modernity Scale [OM]), a measure of de- sire for schooling of male children, and the household landholdings. The first generation is the father of the head of the household. First- generation data include the literacy and landholdings of the child's paternal grandfather, as attributed to him by the head of the household. District of residence and caste status are considered exogenous vari- ables. Those variables, analyzed in this paper, are described in table 3; the mean, standard deviation, and sample size for each variable are given in the sections of this paper in which they are analyzed. The analyses in this paper are restricted to (a) households with children ages 6-16 and for which there are complete data for the heads of the households (N = 282); (b) households with children ages 6-16 for which there are complete data sets for the heads of the households who reside in villages for which school availability data were obtained (N = 213); (c) all heads of households with complete data sets (N = 369); (d) all heads of households with complete data sets who reside in villages for which school availability data were obtained (N = 285); (e) children ages 6-16 who reside in households for which there are com- plete data sets and who reside in villages for which school availability data were obtained (N = 225). The determinants of second-generation (adult) cognitive competencies and attitudes will be reported in Section IV of this paper. Section V will present the results of two analyses of the determinants of third-generation (child) school participation: one conducted at the household level and one at the individual child level of analysis. 288 Economic Development and Cultural Change TABLE 3 DESCRIPTION OF VARIABLES Variable Name Variable Description PlST District indicator: I - Bara; 0 = Rautahat FLAND Total land held by head of household's father, in bighas: I bigha - 0.676 ha FLIT Indicator of head of household's father's literacy: I = literate; 0 = not literate Cl Caste indicator: I = head of household is a member of one of the upper castes of Terai origin (e.g., Brahmin, Rajput, Kayastha); 0 = head of household is a member of another caste AGE Age of head of household, in years RPM Head of household's score on Raven's Progressive MatriPes Test (range = 0 to 36) SYRS Schooling of head of household, in years of schooling completed HHLIT Indicator of literacy of head of household: I = literate; 0 = not literate HHNUM Numeracy score of head of household, as proportion correct of 14 items HHCOMP Reading comprehension score of head of household (range = 0-3) MOD Head of household's modernity score (range = 9-18) SD Amount of schooling desired for boys in household: 0 = will not educate child; I = as long as child wants; 2 = up to high school; 3 = up to intermediate; 4 = up to graduation; 5 = beyond graduation LAND Total land owned, in bighas (l bigha = 0.676 ha) PG616 Percentage female school-aged (6-16 years) children in household C06 Number of children 0-6 years in household PGC Interaction termn: PG616 = C06 SEX Sex indicator: I male, 0 = female SEXO Sex indicator: I female children in households with no other children ages 0-6; 0 = other SEXC Sex indicator: I = female children in households with other children ages 0-6; 0 = other SAP School indicator: I = primary school available in village; 0 = no primary school in village SAS School indicator: I = lower secondary school available in village; 0 = no lower secondary school in village DISP Distance to primary school (0 kosh; 0.5 kosh; 1.0 kosh) DISS Distance to secondary school (0 kosh; 0.5 kosh; 1.0 kosh) PCEl Percentage of school-aged children (6-16) in household having completed at least 1 year of school PCE2 Child's school participation as a deviation from the mean years of school completed for his or her age cohort CS Schooling indicator: I = child completed at least I year of school; 0 = other IV. Determinants of Adult Cognitive Competencies and Attitudes METHODS OF ANALYSIS We have ordered our variable recursively, starting with background characteristics of the individual and moving sequentially through the variables that are explained. In this section we discuss only boxes A, B, E, F, and G of figure I and analyze data from the sample of all heads of households, whether or not school-age children were in the house- hold. Dean T. Jamison and Marlaine E. Lockheed 289 TABLE 4 MEANS, STANDARD DEVIATIONS, AND SAMPLE SIZES OF VARIABLES ANALYZED IN SECTION III HEADS OF fHOUSEHOLDS ALL HEADS OF IN VILLAGES HOUSEHOLDS WITH SCHOOLS (N = 369) (N = 285) VARIABLES Mean SD Mean SD District (DIST) .523 .500 .632 .483 Father literate (FLIT) .117 .321 .112 .316 Father land (FLAND) 4.414 8.191 4.016 6.186 Caste (C 1) .022 .146 .021 .144 Schooling (SYRS) 1.390 2.990 1.435 3.122 Numeracy (HHNUM) .672 .226 .673 .229 Reading comp (HHCOMP) .724 1.565 .765 1.605 Raven's score (RPM) 13.453 4.307 13.491 4.233 Modernity (MOD) 13.621 1.774 13.520 1.728 Land (LAND) 1.913 2.977 1.841 2.491 Age (AGE) 41.713 12.342 41.663 12.139 Attitude toward school (SD) 2.328 1.090 2.291 1.073 Primary school (SAP) ... ... .653 .477 Secondary school (SAS) .. ... 1.790 .408 Distance to primary school (DISP) ... ... .279 .409 Distance to secondary school (DISS) ... ... .798 .568 We analyzed the data using the ordinary least squares linear re- gression of the Statistical Package for the Social Sciences (SPSS).22 Our tables show the estimated regression coefficients below which, in parentheses, are the F-values indicating the statistical significance of the estimated coefficient. The means and standard deviations of the variables used in these analyses are reported in table 4. RESULTS The results of some of these analyses are presented in tables 5-7, which show the estimated determinants of the second generation RPM score, schooling, literacy, reading comprehension, numeracy, attitu- dinal modernity, and attitude toward school. In this section each table will be discussed separately. RPM Score The first column of table 5 presents the results of a single multiple regression analysis of the effects of first-generation landholding, first-generation literacy, caste status, and district of residence on second-generation adult RPM score. This regression shows that first-generation landholding and first-generation literacy were both significant determinants of second-generation RPM scores. It is plausi- 290 Economic Development and Cultural Change TABLE 5 DETERMINANTS OF ADULT ABILITY" (RPM) AND SCHOOLING (SYRS) DEPENDENT VARIABLES INDEPENDENT VARIABLES RPM SYRS SYRS FLAND . 100*+* .092*** .082*** (13.959) (29.753) (27.138) FLIT 2.796*** 2.780*** 1.963*** (16.355) (40.724) (23.316) C1 3.754** 4.989*** 4.477*** (6.618) (29.430) (28.136) DIST .984* .476 .283 (5.485) (3.236) (1,366) RPM ... ... .2048** (46.108) AGE ... ... .055*+* (30.877) C 12.089 .302 .091 R2 .140 .291 .419 Adjusted R2 .130 .283 .409 N 369 369 369 NOTE.-F-statistic in parentheses. * P < .05. ** p < .01. *** P < .001. ble, of course, that, as we have modeled the situation, an individual's RPM score is both an outcome of schooling as well as a determinant of schooling. We thus properly have a simultaneous system, which we have identified in this case by assuming that the coefficient of schooling on RPM is zero. Though this is at best an approximation, it does have justification in that the intention of the RPM is to measure "innate" ability, and, in our sample, this intention seems to be partially fulfilled in that relative differences between males' and females' RPM scores are small compared with, say, numeracy scores.23 Schooling Attainment The second and third columns of table 5 report the results of two multiple regression analyses to estimate the determinants of second- generation school attainment. These regressions indicate that school attainment was significanitly affected by first-generation landholding and literacy as well as by caste status. The effect of "innate" ability as measured by RPM score was also statistically significant, considered both independently (regression coefficient = .317; F = 96.765; r2 = .207) and in combination with the exogenous variables. Overall, more schooling was attained by younger persons with relatively more literate and landed fathers from the Brahmin, Rajput, and Kayastha castes and with higher "innate" ability. Higher-status caste members on the aver- TABLE 6 DETERMINANTS OF ADULT LITERACY (HHLIT), READING COMPREHENSION (HHNCOMP), NUMERACY (HHNUM), AND ATTITUDINAL MODERNITY (MOD) DEPENDENT VARIABLE INDEPENDENT VARIABLE HHLIT HHLIT HHCOMP HHCOMP HHNUM HHNUM MOD MOD FLAND ... .002 ... -.006 ... -.Q01 .. 025* (1.464) (.696) (.693) (5.379) FLIT ... .030 ... .196 ... .047 ... .865** (.447) (1.011) (1.763) (9.459) CL ... .053 ... .468 ... .032 ... 1.594** (.337) (1.319) (.190) (7.359) DIST ... .020 ... .089 ... .013 ... - .683** (.602) (.625) (.415) (17.612) AGE ... .001 ... -.005 ... -.003** ... -.021** (.369) (1.281) (12.997) (9.184) SYRS .116** .113* .383** .373** .036** .025** .222** .096* (758.337) (417.409) (422.185) (232.472) (105.244) (31.966) (60.057) (7.440) RPM ... -.001 ... -.008 ... .010* ... .040 (.169) (.303) (12.725) (3.573) C .069 .040 .191 .488 .622 .630 13.312 13.922 R2 .674 .677 .535 .542 .223 .281 .141 .258 Adjusted R2 .673 .671 .534 .533 .221 .267 .138 .244 N 369 369 369 369 369 369 369 369 NOTE.-F-statistic in parentheses. < < .05. **p < .001. 292 Economic Development and Cultural Change TABLE 7 DETERMINANTS OF SECOND-GENERATION DESIRE FOR THIRD-GENERATION SCHOOLING ALTERNATIVE SPECIFICATIONS I.NDEPENDENT - VARIABLES (1) (2) (3) DIST .000 .093 .46 (.000) (.527) (.094) FLAND .006 .016 .016 (.382) (1.601) (1.476) FLIT .501 ** .403 .409 (6.673) (3.132) (3.179) C I .089 -.223 - .262 (.049) (.226) (.307) SYRS -.045 .018 .017 (2.115) (.235) (.214) HHNUM .224 .221 .199 (.573) (.456) (.363) HHCOM -.021 -.044 -.0441 (.168) (.612) (.505) RAVENS .002 - .009 .008 (.015) (.300) (.209) MOD .072* .062 .070 (3.852) (1.996) (2.483) LAND .003 .082** .081** (.018) (7.473) (7.184) AGE .004 .002 .002 (.598) (.089) (. 100) SAP .459 (1.412) SAS ... ... .210 (.802) DISP ... ... 385 (.7 13) DISS ... ... -.087 (.251) C .876 1.065 .261 R2 .109 .136 .143 Adjusted R2 .081 .102 .095 N 369 285 285 NOTE.-F-statistics in parentheses. * p < .05. ** p < .001. age attained nearly 4.5 more years of school than lower-status caste members; individuals with literate fathers attained nearly 2 more years of schooling than individuals with illiterate fathers; and, finally, older individuals attained fewer years of schooling than younger individuals. Adult Literacy Second-generation adult literacy was assumed to be a function of back- ground characteristics, "innate" ability, and schooling. Although first- generation landholding, literacy, and caste had significant effects on Dean T. Jamison and Marlaine E. Lockheed 293 second-generation literacy when assessed independently, these effects operated largely through second-generation school attainment (cols. I and 2 of table 6). Likewise, the effects of RPM score on literacy were mediated by schooling. Overall, schooling alone accounted for 67% of the variance in second-generation adult literacy; the inclusion of vari- ous exogenous and first-generation factors in the regressions did little to improve the overall estimation. Adult Reading Comprehension Second-generation adult reading comprehension was also assumed to be a function of background characteristics, "innate" ability, and schooling. Columns 3 and 4 of table 6 report two of several alternative models of the determinants of second-generation reading comprehen- sion. From these regressions we see that the most significant predictor of reading comprehension is school attainment. The effects of both first-generation literacy and second-generation RPM score, statistically significant when considered independently, disappeared when consid- ered simultaneously with school attainment. Numeracy Again, the significant effects of first-generation literacy on second- generation adult numeracy was through the school attainment of the second generation. Second-generation school attainment accounted for 22% of the variance in second-generation numeracy, but, unlike the models for literacy and reading comprehension, RPM score and age were also significant predictors of numeracy (cols. 5 and 6 of table 6). Attitudinal Modernity Columns 7 and 8 of table 6 show that the significant effects on second- generation attitudinal modernity of first-generation landholding, first- generation literacy, and caste remained significant when second-gen- eration schooling and RPM score were taken into account. Both second- generation school attainment and RPM score were significant determi- nants of attitudinal modernity, and, although the independent effect of each was reduced when both were considered simultaneously, each remained a significant predictor of attitudinal modernity. Attitudes toward Schooling Second-generation attitudes regarding third-generation schooiing were assumed to be determined by both first- and second-generation charac- teristics and by school availability. These effects were assessed twice, once on the full sample of household heads and once on the sample of household heads from villages for which school availability data were obtained. The results of these analyses are presented in table 7. From the first column of table 7 we see that, for all household 294 Economic Development and Cultuiral Change heads, first-generation literacy and second-generation attitudinal mo- dernity were the only factors related to second-generation attitude to- ward schooling. The second column of table 7 reports the results of this analysis when it was repeated for the sample household heads residing in villages from which school availability data were obtained. For this second sample, only landholding was a significant determinant of de- sire for schooling. This difference suggests that the two samples may not be similar, Column 3 of table 7 presents the coefficients for four indicators of school availability-the availability of primary schools in the village, the availability of secondary schools in the village, the distance to primary school, and the distance to secondary school. It is evident that school availability was not a factor in determining desire for schooling for this sample. V. Determinants of School Participation In this section we analyze the determinants of third-generation (child) school attendance. The analysis is conducted first with the data ag- gregated at the household level and second with the data aggregated at the individual child level. The dependent measures for these analyses differ. At the housc hold level, school participation was operationalized as the proportion of children ages 6-16 who completed at least 1 year of school (PCE 1); the mean value of this indicator for all households having children ages 6-16 was 19.1%. At the individual level, school participation was operationalized two ways: first as the child's devia- tion from the age-specific mean school attendance for all children in the sample (PCE 2)-these means are shown in table 8 and indicate a low level of school attendance-and, second, as a dichotomous indicator TABLE 8 AVERAGE YEARS OF SCHOOL ATTENDED BY AGE COHORT Average Years of Age Cohort School Attended 6 years old .179 7 years old .273 8 years old .455 9 years old .471 10 years old .616 11 years old 1.621 12 years old 1.086 13 years old 1.133 14 years old 2.154 15 years old 2.292 16 years old 1.125 Dean T. Jamison and Marlaine E. Lockheed 295 TABLE 9 MEANS, STANDARD DEV!ATIONS, AND SAMPLE SIZES OF VARIABLES ANALYZED IN SECTION IV CHILDREN IN VILLAGES ALL CHILDREN WITH SCHOOLS (N - 443) (N = 335) VARIABLES Mean SD Mean SD District (DIST) .533 .500 .648 .478 Grandfather literate (FLIT) .102 .302 .090 .286 Grandfather land (FLAND) 4.504 8.777 3.609 4.073 Caste (Cl) .029 .169 .030 .170 Father's schooling (SYRS) 1.519 3.036 1.591 3.206 Father's numeracy (HHNUM) .680 .239 .681 .243 Father's reading comprehension (HHCOMP) .847 1.662 .925 1 735 Father's Raven's score (RPM) 13.402 4.237 13.481 4.304 Father's modernity (MOD) 13.529 1.697 13.548 1.679 Father's land (LAND) 2.213 3.488 1.936 1.985 Father's age (AGE) 43.713 0.779 43.713 10.752 Father's attitude toward school (SD) 2.139 1.112 2.128 1.110 Deviation from age mean school (PCE 2) -.075 1.699 .088 1.707 Child's sex .601 .490 .585 .493 Primary school (SAP) ... ... .633 .483 Secondary school (SAS) ... ... .203 .403 Distance to primary school (DISP) ... ... .300 .421 Distance to secondary school (DISS) ... ... .778 .544 of school participation (CS). The means and standard deviations of the variables used in the following analyses are presented in table 9. DATA AGGREGATED AT THE HOUSEHOLD LEVEL Method Once again we have ordered our variables recursively. Referring back to figure 1, the variables in boxes A, E, and B-school availability- are entirely exogenous and are assumed to affect the variables in boxes F-second-generation cognitive competencies and attitudes-and G- second-generation attitude toward school-both separately and in combination. These variables, Laken together with the number of chil- dren in the family and the child's sex, are assumed to affect third- generation school participation. These data were also analyzed using the SPSS ordinary least squares regression. Results Because of the very low rates of enrollment in our sample, household school participation by children was operationalized as the pecentage 296 Economic Development and Cuiltural Change TABLE 10 SCHOOL ATTENDANCE IN BARA AND RAUTAHAT DISTRICTS SCHOOL LEVEL Lower Primary Secondary Secondary Age of students (years) 6-8 9-12 13-15 Ministry of Education: Total population: In Bara and Rautahat 52,627 26,679 24,186 Estimated female 26,314 13,339 12,093 Estimated male 26,314 13,339 12,093 Total enrolled: 19,152 4,826 1,872 Female en-rolled 2,948 (15%) 480 (10%) 148 (8%) Male enrolled 16,204 (85%) 4,346 (90%) 1,724 (92%) Total enrollment rate (%): 36.4 18,1 7.7 Female enrollment rate 11.2 3.6 1.2 Male enrollment rate 61.5 32.6 14.3 New ERA/World Bank: Sample: In Bara and Rautahat* 355 417 225 Female sample 157 170 109 Male sample 198 247 116 Sample completed some school: 42 74 44 Female 7 (16.7%) 9 (12%) 8 (18.2%) Male 35 (83.3%) 66 (88.0%) 36 (81.8%) Female 1 year completion (%) 2.2 2.2 3.6 Male I year completion (%) 9.9 15.8 16.0 NOTE.-Comparing Nepal Ministry of Education Data with New ERA/World Bank Survey Data. * All children in household, by ages 6-8, 9-12, 13-16. of children in the household, ages 6-16, who completed any year of school. This measure provided overall statistics not dissimilar to na- tional statistics for these districts (table 10). Household determinants. As the first column of table 11 indi- cates, household children's school participation was determined by the household wealth (as indicated by household landholdings), the school- ing attained by household head, the numeracy of the household head, the years of schooling of the household head, the RIM score of the household head, the attitudinal modernity of the household head, the caste of the household, and the percentage of female children. With the exception of the percentage of female children, these effects were positive. To examine the question of whether rates of female participation in school were affected by the presence of young children in the house- hold, we constructed the interaction term PGC, or the percentage of girls in the household multiplied by the number of children ages 0-6. Introducing this term (which had a statistically significant coefficient Dean T. Jamison and Marlaine E. Lockheed 297 TABLE II HOUSEHOLD CHARACTERISTICS, SECOND-GENERATION COGNITIVE COMPETENCIES, ATTITUDES, AND SCHOOL AVAILABILITY AS DETERMINANTS OF THIRD-GENERATION SCHOOL PARTICIPATION ALTERNATIVE SPECIFICATIONS INDEPENDENT VARIABLES (I) (2) (3) (4) DIST 4.178 3.492 4.749 3.231 (1.524) (1.134) (1.461) (.521) LAND 1.679** 1.683*** 2.773** 2.263* (8.602) (9.157) (8.440) (5.489) AGE .237 .173 .209 .192 (2.272) (1.281) (1.383) (1.145) SYRS 1.748* 1.343 1.348 1.352 (5.653) (3.511) (2.610) (2.557) HHNUM 21.777* 21.958** 22.256* 23.160* (6.534) (7.105) (5.351) (5.876) RPM 1.054* 1.086* 1.112* 1.131* (5.533) (6.276) (4.886) (5.122) MOD 3.358*** 2.558* 3.192* 2.529* (10.143) (6.029) (6.540) (3.992) C1 33.607*** 33.753*** 33.661** 36.017** (10.454) (11.122) (9.298) (10.685) PG616 - .224*** - .128* - .141 - .154* (28.951) (3.927) (3.618) (4.323) C06 1.732 6.073* 3.583 4,166 (.927) (5.838) (1.536) (2.100) PGC ... - .097* - .067 - .070 (5.390) (1.828) (2.005) SD ... 6.069*** ... 4.303* (16.55) (6.226) SAP ... ... ... 1.437 (.114) SAS ... ... ... - 2.145 (.142) C - 67.982 - 71.986 - 71.174 - 70.783 R2 .4!1 .454 .435 .454 Adjusted R2 .390 .430 .404 .415 N 282 282 212 212 NOTE.-Dependent variable is percentage of school-aged children in household hav- ing completed at least I year of school (PCE 1). F-statistics in parentheses. * P < .05. ** p c .01L *** P < .001. equal to -.088, F = 4.136) into the regression dramatically reduced the effect of the percentage of girls in the household on childr6n's school participation, while increasing to statistical significance the ef- fects of children ages 0-6 (coefficient = 5.518, F = 4.570). From this we concluded that girls were being utilized in the household to care for smaller children rather than being encouraged to attend school. The effects of the cognitive abilities and modern attitudes of the household heads' spouses were estimated in separate equations. These 298 Economic Development and Cultural Change effects were limited to a positive effect of spouses' RPM score on children's school participation; spouses' modernity was not related to children's school participation. Attituide toward school effect. The second column of table II reports the effects on school participation of entering the respondent's attitude toward schooling into the regression, For heads of households (but not for spouses), attitude toward schooling was worth approxi- mately 5% increased enrollment for each additional level of desired schooling beyond the primary level. Holding attitude toward schooling constant did little to change the relative impact of other household characteristics on household children's school participation. School availability effect. Two indicators of school availability were used in this analysis: presence or absence of a primary or lower secondary school in the village and the distance to a primary or lower secondary school that served the village, should none be available in the village. As the third column of table II indicates, holding house- hold characteristics constant, neither primary nor secondary school availability was found to have a statistically significant effect on chil- dren's school participation, though the direction of the observed effect was generally what would be expected. We note, however, that school availability for this sample may be atypical for Nepal as a whole, insofar as 68% of the households were located in villages reported to have primary schools and 20% were located in villages with lower secondary schools. The average distance to primary school was, fur- thermore, only .26 kosh, or about one-quarter mile. Household characteristics, attitude toward school, and school availability. As the last column of table 11 indicates, when household characteristics, attitude toward school, and school availability were considered simultaneously, the head of household', cognitive abilities and attitudes continued to affect children's school participation, as did the attitude toward schooling. School availability had no effect inde- pendent of attitude toward school. For comparison purposes, it is interesting to note that being in a high caste group-a totally exogenous variable-was worth 36% more children participating in school, while being numerate-a potentially determined variable-was worth 23% for a perfect test, or 1.6% for each correct answer. Being modern was worth 2.5% more children enrolled for each modern response. A positive attitude toward school- ing was worth 4.3% more children enrolled for each level of schooling desired beyond the primary level. Assuming that each household with children had four school-aged children, two of which were females, each girl was worth 3.8% fewer children enrolled; if all the school-aged children were female, 15% fewer children would participate in schooling. If primary school availability were a statistically significant factor- Dean T, Jamison and Marlaine E. Lockheed 299 which it was not-a school in the village would be worth only 1.4% more children enrolled. DATA AGGREGATED AT THE CHILD LEVEL Method Determinants of third-generation school participation were identified as a combination of fixed background characteristics, first-generation factors, second-generation factors including a stated desire for third- generation schooling, and school availability. The determinants of school participation at the child level of analysis were estimated in two ways: first, using an ordinary least squares linear multiple regression approach with school participation expressed as a continuous variable, and, second, using a logistic regression with school participation ex- pressed as a dichotomous indicator. For the first analysis, the depen- dent variable was the child's deviation from the mean years of school- ing completed by his or her age cohort. Results of the Linear Multiple Regression Analyses Sixty-two different equations testing alternative specifications for the determinants of the child's school participation were estimated. In general, first- and second-generation factors explained about 16% of the variance in third-generation school participation, while third- generation sex explained an additional 4% of the variance. Table 12 reports several of the more interesting specifications. First-generation effects. The first column in table 12 presents the coefficients for the effects on third-generation school participation of first-generation literacy and landholding, holding constant caste, dis- trict of residence, and sex of the child. Although caste is fixed and may be considered exogenous, it is important to note that members of higher castes are substantially more likely to send their children to school; caste alone accounted for 5. 1% of the variance in school atten- dance. First-generation literacy accounted for 5.8% of the variance in third-generation general school participation, while first-generation landholding accounted for another 3.5% of the variance. Second-generation effects. The second column in table 12 pre- sents the coefficients for the effects on third-generation school partici- pation of first-generation landholdings and literacy and of second- generation RPM score, school attainment, cognitive competencies, landholding, attitudinal modernity, and fixed background characteris- tics. In column three, the child's sex is also included. The most significant determinant of third-generation school participation was the sex of the child, boys receiving on the average three-quarters of a year more schooling than girls. Although both second-generation RPM score and attitude toward school were significant determinants of third- TABLE 12 DETERMINANTS OF THIRD-GENERATION SCHOOL PARTICIPATION ALTERNATIVE SPECIFICATIONS INDEPENDENT VARIABLES (1) (2) (3) (4) DIST .411** .249 .222 .022 (7.863) (2.887) (2.293) (.012) FLAND .032*** - .003 - .002 .043 (13.801) (.030) (.013) (3.277) FLIT .837*** .465 .426 .206 (10.647) (3.259) (2.734) (.383) Cl 2.277*** 1.366** 1.315** 1.420* (26.027) (8.412) (7.839) (5.442) SYRS ... .057 .061 .046 (2.061) (2.383) (.880) HHNUM ... ,799** .665 .474 (5.051) (3.425) (1.240) HHCOMP ... .011 -.002 .042 (.032) (.001) (.343) RAVENS ... 049** .051 ** 049* (5.916) (6.360) (4.145) MOD ... .066 .053 .083 (1.790) (1.145) (1.719) LAND ... .064 .065 .061 (2.986) (3.057) (1.484) AGE ... .011 .010 .010 (2.672) (2.164) (1.222) SD ... .736*** .142* .051 (25.384) (4.477) (.386) SEX .687*** ... 773*** .757*** (21.098) (27.864) (19.128) SAP ... ... ... -.484 (.809) SAS ... ... ... .045 (.020) DISP ... ... ... -.765 (1.492) DISS ... ... ... .280 (1.308) C - 1.003 - 3.559 - 3.570 - 3.291 R .197 .275 .282 .291 Adjusted RI .188 .255 .261 .253 N 443 433 433 335 NOTE.-Dependent variable is child's deviation from age cohort mean school atten- dance (PCE 2). F-statistics in parentheses. * P < .05. ** p < .01. *** p < .001. Dean T. Jamison and Marlaine E. Lockheed 301 TABLE 13 MEANS AND STANDARD DEVIATIONS OF SELECTED VARIABLES BY SCHOOL ATTENDANCE STATUTS OF CHILD SCHOOL SCHOOL ATTENDEES NONATTENDEES (N = 69) (N 361) VARIABLES Mean SD Mean SD Exogenous: District (DIST) .652 .480 .507 .501 Caste (CI) .159 .369 .006 .074 First-generation factors: Literacy (FLIT) .275 .450 .064 .245 Landholding (FLAND) 9.969 17.800 3.273 4.531 Second-generation factors: Age (AGE) 43.768 11.661 43.859 10.672 Years of schooling (SYRS) 4,449 4.00 .920 2.259 Literacy (HHLIT) .638 .484 .194 .396 Comprehension (HHCOMP) 1.957 1.818 .612 1.522 Numeracy (HHNUM) .853 .186 .647 .232 Raven's Progressive MAtrices (RPM) 17.087 5.575 12.665 3.536 Landholding (LAND) 4.597 6.947 1.752 2.104 Modernity (MOD) 14.764 2.167 13.278 1.461 Attitude toward schooling (SD) 2.667 1.421 2.002 1.005 Child factors: Sex (SEX, .899 .304 .529 .500 generation school participation, the effects of attitude toward school were markedly reduced when the sex of the child was included in the specification, 3uggesting that this attitude did not generalize to daugh- ter's schooling and was restricted to attitudes toward son's schooling, Effects of school availability. Column 4 of table 12 presents the results for the effects of school availability on child school participa- tion for the subsample of children for whom school availability data were obtained. Holding other variables constant, neither the presence or absence of either primary or secondary schools in the village nor the distance to these schools had any effect on child school partLicipation. Logistic Regression Analysis For this analysis, the dependent variable (CS) was a dichotomous indi- cator of whether or not the child had obtained any schooling. The means and standard deviations of selected variables by child schooling status are presented in table 13. Several dozen logistic regressions were run; table 14 reports the results of some of the more interesting ones. First-generation effects. As was noted in the linear multiple re- gressions, the first column of table 14 indicates that caste was a statisti- cally significant predictor of child school participation, with members of higher castes being twice as likely to attend school than members of 302 Economic Dev;elopment and Cultural Change lower castes. Neither first-generation literacy nor first-generation landholdings were significant predictors of third-generation school participation. Second-generation effects. With first generation and exogenous variables held constant, the second column of table 14 indicates that parental schooling and modern attitudes were strongly related to child school participation; parental attitudes toward school were not related to child school participation in this analysis. Other exogenous effects. Columns 3-4 of table 14 report the effects of two other exogenous factors on child school participation: child's sex and school availability. Two variables related to child's sex were created. The first, SEXC, was an indicator that the child was female in a household having other children ages 0-6 years; the sec- ond, SEXO, was an indicator that the child was a female in a household with no other small children present. In columns 3 and 4 of table 14, we see that the effect of being female was consistently negative, with girls 3.5-4.5 times less likely to attend school than boys. Small children in the household decreased the girls' likelihood of attending school, but all girls were less likely to attend school than were boys. Two variables related to the availability of schools were created; these were similar to those used in the linear multiple regression. The first was whether or not a primary school was available in the village, and the second was, if no primary school was available, the distance to the nearest school. In this analysis, availability of a primary school in the village was not related to school participation. VI. Summary and Conclusions In this article we have examined the determinants of adult cognitive competencies and of child school participation in the Terai region of Nepal. Data on three generations from 795 rural farm households were analyzed. Four analyses were conducted: a linear multiple regression analysis of the determinants of adult cognitive competencies and at- titudes, a linear multiple regression analysis of the determinants of household child school participation, a linear multiple regression of the determinants of individual child school participation, and a logistic regression of the determinants c zhild school participation. The results of these various analyses are highly consistent and may be summarized as follows: (1) first-generation (grandparent) land- holdings, literacy, and caste status were significant determinants of second-generation "innate ability" as measured by the Ravens Pro- gressive Matrices (RPM) test; (2) RPM was a significant determinant of second-generation schooling- (3) second-generation schooling was a significant determinant of second-generation literacy, numeracy. read- ing comprehension, and attitudinal modernity; (4) attitudinal moder- TABLE 14 LOGISTIC REGRESSION ESTIMATES OF THE DETERMINANTS OF THIRD-GENERATION SCHOOL PARTICIPATION ALTERNATIVE SPECIFICATIONS INDEPENDENT VARIABLES (1) (2) (3) (4) DIST - 1.345*** .478 .655 .664 (-6.653) (1,155) (1.334) (1.357) FLIT .488 .678 - .057 - .106 (1.034) (1.241) (-.078) (-.145) FLAND -.030 .111** .079 .097 (-.974) (2.810) (1.618) (1.932) Ci 2.405** 1.333 2.917* 2.647 (2.757) (1.067) (2.073) (1.889) AGE ... -.026 -.016 -.015 (-1.722) (-.850) (-.853) SYRS ... .231*** .281** - .275** (2.962) (2.778) (2.758) HHNUM ... .311 1.163 1.091 (.308) (1.033) (.979) RAVENS ... ,036 .030 .026 (.787) (.558) (.486) LAND ... .039 .i89 .174 (.451) (1.698) (1.573) MOD ... -.211** -.311** -.303** (-2.611) (-3.018) (-2.952) HHCOMP ... - .065 .140 .136 (-.558) (.929) (.912) SD ... .124 .485* .434* (.823) (2.453) (2.221) SEXC ., ... -4.881*** - 4.776*** (-4.872) (-4.773) SEXO ... ... -3.415** -3.451** (-2.623) (-2,651) SAP ... -.254 (-.576) DISP ... ... - .363 (-.654) N 326 326 326 326 Log likelihood at convergence -181.9 -118.9 -89.57 86.51 Likelihood ratio 88.04 214.1 278.8 278.9 df 322 314 311 311 NOTE.-Dependent variable is dichotomous indicator of child schooling (CS). t- statistic in parentheses. * P < .05. ** P < .01. P < .001. 304 Economic Development and Cultural Change nity was a significant determinant of attitude toward school for all households but not for the subset of households in villages for which school availability data had been obtained; (5) in these latter villages, there was no relationship between school availability and a positive attitude toward school-only household landholding was a determi- nant of attitujde toward school; (6) third-generation child school partici- pation was determined jointly by second-generation landholdings, caste, schooling and numeracy, attitudinal modernity, and the child's sex; (7) girls were significantly less likely to have completed at least 1 year of school than were boys; (8) the presence of small children in the household only slightly decreased girls' chances to participate in school; (9) school availability had no effect on child school par- ticipation. Data from the same households shed light on two other important determinants and outcomes of school. Moock and Leslie found that child nutritional status was positively and significantly related to male and female school enrollment and to male grade attainment.24 Jamison an" Moock found that grade attainment and numeracy improved the efficiency of rural farmers by economically meaningful amounts.25 A compl:tely parallel study conducted at the same time in Thailand found generally similar results, with only one or two exceptions, most nota- bly that schooling and ability affected parental aspirations for chil- dren's schooling.26 Notes * This paper was prepared under the auspices of World Bank research project RPO 671-49, "Education and Rural Development in Nepal and Thai- land," which was undertaken by World Bank staff and consultants in collab- oration with staff of New ERA, a Nepalese research and consulting firm. A number of our colleagues-Bal Gopal Baidya, Susan Cochrane, Joanne Leslie, Peter Moock, and Rajendra Shrestha-have provided valuable advice and as- sistance on various aspects of our work, and we wish to acknowledge our indebtedness to them. Erwin Chou and Kalpana Mehra provided expert com- puter assistance on all aspects of the work. Views and conclusions expressed in the paper are those of the authors and do not necessarily reflect those of the World Bank group. 1. Urban labor markets: George Psacharopoulos, "Returns to Education: An Updated International Comparison," Comparative Education 17 (1981): 321-41. Agricultural productivity: Marlaine E. Lockheed, Dean T. Jamison, and Lawrence J. Lau, "Farmer Education and Farm Efficiency: A Survey," Economic Development and Cultural Change 29 (1980): 37-76. Human fertil- ity: Susan Cochrane, Education and Fertility: What Do We Really Know? (Baltimore: Johns Hopkins University Press, 1978), Health and nutritional status: Susan H. Cochrane, Joanne Leslie, and Donald J. O'Hara, "Parental Education and Child Health: Intracountry Evidence," Health Policy and Edu- cation 2 (1982): 213-50, 2. J. R. Sheffield, "Retention of Literacy and Basic Skills: A Review of the Literature" (paper prepared for the education department of the World Dean T. Jamison and Marlaine E. Lockheed 305 Bank, 1975); Alex Inkeles and David H. Smith, Becoming Modern (Cam- bridge, Mass.: Harvard University Press, 1974). 3. Kjell Harnqvist, "Enduring Effects of Schooling-a Neglected Area in Educational Research," Educational Researcher 6, no. 1 (1977): 5-1 1; Sheffield. 4. Literacy: Jennifer Noesjirwan, "Permanency of Literacy in In- donesia," American Educational Research Journal 1 (1974): 93-99; Psacharopoulos; Everett M. Rogers and L. Svenning, Modernization among Peasants (New York: Holt, Rinehart & Winston, 1969); Prodipto Roy and J. M. Kapoor, The Retention of Literacy (Delhi: Macmillan Co. of India, 1975); H. Schuman, Alex Inkeles, and David H. Smith, "Some Psychological Effects and Noneffects of Literacy in a New Nation," Economic Development and Cultural Change 16 (1967): 1-14. Reading and writing: M. Ahmed, "Retention of Literacy Skills in Adults," Literacy Work 11 (1973): 43-53; G. Flores, "A Study on Functional Literacy for Citizenship in the Philippines," Quarterly Bulletin of Fundamental Education 2 (1950): 24-28; D. Kosak, "The Problem of Reading in Elementary Education," Adult Education, nos. 5-6 (1963), pp. 35-38; J. Landy-Tolwinska, "Farmer Participants in Literacy Courses- Poland," Literacy Discussion 1 (1970): 107-10; H. F. Lionberger and H. C. Chang, Communication and Use of Scientific Farm Information by Farmers in Two Taiwan Agriculture Villages (Columbia: University of Missouri, College of Agriculture, 1968). Numeracy: Russell Beirn, David C. Kinsey, and Noel F. McGinn, Antecedents and Consequences of Early School Leaving: An Ana- lytic Survey of Research Activities, Occasional Papers in Education and Devel- opment no. 8 (Cambridge, Mass.: Harvard University, Graduate School of Education, 1972); Flores; Noesjirwan. Modernity: Alex Inkeles, "The School as a Context for Modernization," International Journal of Comparative Sociology 14 (1973): 163-79; R. Sultzman and 1. Amnon, "The Effects of Schooling on Psychological Modernity: A Longitudinal Study" (unpublished manuscript, 1978). 5. Nancy Birdsall, "Child Schooling and the Measurement of Living Stan- dards," World Bank Living Standards Measurement Study Working Paper no. 14 (Washington, D.C.: World Bank, February 1982); Nancy Birdsall and Susan Hill Cochrane, "Education and Parental Decision-making: A Two Generation Approach," in Education and Development, ed. Lascelles Anderson and Douglas M. Windham (Lexington, Mass: D. C. Heath & Co., 1982), pp. 179- 210. 6. Mary Jean Bowman and C. Arnold Anderson, "The Participation of Women in Education in the Third World" (unpublished manuscript, University of Chicago, Comparative Education Center, October 1978). 7. Peter Moock and Joanne Leslie, "Childhood Malnutrition and School- ing in the Terai Region of Nepal," Jouirnal of Deve7opment Economics 20 (1986): 33-52. 8. Peter Sellar, David Sprague, and Virgil Miedema, U.S. Aid to Educa- tion in Nepal: A 20-Year Beginning (Kathmandu: U.S. Agency for Interna- tional Development, 1981). 9. H. M. G. Nepal, Ministry of Education and Culture, unpublished data (Kathmandu, 1983). 10. Rajendra B. Shrestha, Accessibility to Educational Opportunities in the Remote Areas (Kathmandu: Tribhuvan University, Center for Economic Development and Administration, 1976). 11. Beirn, Kinsey, and McGinn. 12. CERID/WE, Determinants of Educational Participation in Rural 306 Economic Development and Cultural Change Nepal (Kathmandu: Tribhuvan University, Center for Education Research, Innovation and Development, 1983). 13. Barbara Butterworth, Dibya Karmacharya, and Richard Martin, "Radio Education Teacher Training Program, Final Evaluation Report" (un- published manuscript, Kathmandu, April 1983). 14. Shrestha. 15. Birdsall; Birdsall and Cochrane. 16. Meena Acharya and Lynn Bennett, "The Rural Women of Nepal: An Aggregate Analysis and Summary of 8 Village Studies," in The Status of Women in Nepal, vol. 2 (Kathmandu, 1981), part 9. i7. Equity and Discrimination among Children: Household Decision- making Strategies for the Provision of Schooling in Rural Nepal (unpublished manuscript, 1983). 18. Prem Kasaju, "'Learning to Read and Write: A Major Task for Devel- opment," in On Education in Nepal, ed. M. Mohsin and P. Kasaju (Kath- mandu: National Education Committee, 1974). 19. Shrestha. 20. Gwendolyn L. Lewis, A Model of Turkish Adults' Educational Aspi- rations for Their Children (Ankara: Hacettepe University, Institute of Popula- tion Studies, 1978). 21. Prem Kasaju, Parents' Attitude toward and Expectations from Educa- tion (Kathmandu: Tribhuvan University, Center for Education Research. In- novation and Development, 1982). 22. Norman H. Nie, C. Hadlai Hull, Jean G. Jenkins, Karin Steinbrenner, and Dale H. Bent, Statistical Package for the Social Sciences, 2d ed. (New York: McGraw-Hill Book Co., 1975). 23. Dean T. Jamison, Joanne Leslie, and Bal Gopal Baidya, "Determi- nants of the Literacy and Numeracy of Adults in the Terai Region of Nepal" (paper presented at the annual meeting of the Eastern Economic Association, Boston, May 1979). 24. Moock and Leslie (n. 7 above). 25. Dean T. Jamison and Peter Moock, "Farmer Education and Farm Efficiency in Nepal: The Role of Schooling, Extension Services and Cognitive Skills," World Development 12 (1984): 67-86. 26. Susan Hill Cochrane and Dean T. Jamison, '"Educational Attainment and Achievement in Rural Thailand," in New Directions for Testing and Mea- surement: Productivity Assessment in Education, ed. A. Summers (San Fran- cisco: Jossey-Bass, Inc., 1982), pp. 43-59.