PolIy, PInnln, enW Rsewch WORKING PAP%RS Educeton and Employment Population and Human r;esources Departmert The World Bank AugJst 1 988 WPS 29 The Relative Effectiveness of Single-Sex and Coeducational Schools in Thailand Emmanuel Jimenez and Marlaine E. Lockheed Single-sex schooling is more effective for girls, but coeduca- tional schooling is more effective for boys in improving student performance in mathematics. The differences are due to peer group effects, rather than to school or classroom characteristics. hePolicy.Plannig,and RearchComplexdistibuteaPP WoiingPapermtodiasntetheflndingaofwo inprogress ndto enoge the exhange of ideas ong Bank staff and aU others interested in development issues. Thcc papers carry the names of the authors. refleamny their views, and should be used and cited accordingly. Thb findings, irztrpstatics. and cerclusions rc the authores own. They should not be attributed to the World Bank, its Board of Di,ctors its manageaent, orany of its manber omnties. Poloy,Pbkh, and Rueh |h Educotlon nd Empboyment I Several studies of the relative effectiveness of previous achievement. It does this by measuring single-sex and coeducational schools have performance at the beginning and again at the shown that single-sex secondary education end of the year to focus on the educational promotes both academic achievement and "value added" during that year. orientation, particularly for girls. "Single-sex" education also includes coeducational schools Girls in single-sex Thai schools scored where students are separaZ ed into single-sex higher in mathematics achievement at the end of classes for instruction. the eighth grade, but the reverse was true for boys, who exhibited higher scores in coeduca- Most studies of single-sex education have tional schools. Why was this so? made litde or no attempt to control for factors such as student background, school type (public The largest factor affecting student perform- or private), and school selection by parent or ance was the student's peer group. The data did students. These factors can affect achievement not permit an analysis of how peer groups affect and skew analyses that seek to compare only the achievement, but studies in developed countries effect of single-sex education versus coeduca- suggest that class participation and leadership tion. opportunities are suppressed for girls in coedu- cational settings and for boys in single-sex Mathematics test scores of Thai eighth- settings. graders, obtained during the 1981-82 academic year, are compared for students in coeducational This paper is a product of the Education and and single-sex schools. The study overcomes Employment Division, Population and Human the methodological problems by holding con- Resources Department. Copies are available stant student background, school type, and free from the World Bank, 1818 H Street NW, school selection. Moreover, the study mini- Washington, DC 20433. Please contact Teresa mizes the effects of non-measured variables Hawkins, room S6-224, extension 33678. such as a student's ability, motivation, or e PPR WorCng Payrg Se0es dissemes the fiindgs of work ender way ir the Bank's Policy, PlamngB and Resekrch Complex. An objective of the series is to get these fmdings out quickly, even if presentations are' ,ess than fully polished. rne fuindings, interpretations, and conclusions irt thee papers do not necessarily represent official policy of the Bank. Copyright 0 1988 by the Intemiational Bank for Reconstructi.on and 11-velopnienVlIte World Bank The Relative Effectiveness of Single-Sex and Coeducational Schools in Thailand by Emmanuel Jimenez and Marlaine E. Lockheed Table of Contents Page Introduction .......................................................3 Background .........................................7 Education in Thailand ...................................................... .7 Single-Sex Education ...................................................... 9 Academic Achievement ...................................................... 9 Academic Orientation .................................... 13 Why are Single-Sex Schools Effective? ...................................................... 14 The Basic Model and Data ...................................................... 16 Model ...................................................... 16 Sample ...................................................... 17 Mathematics Achievement ....................................................... 18 Student Background Characteristics ...................................................... 19 Peer Group, Class, Teacher and School Characteristics . . 20 Single-Sex Learning Envirorment ...................................................... 21 The Effect of Background on Achievement in Single-Sex and Coeducational Schools 22 How We Control for Background ........ .............................. 25 What determines the Choice of School Type? . ........................................... 27 How Does Socio-Economic Background Affect School Achievement?.. 29 Male Achievement ....................................... 32 Female Achievement ..................................... 32 The Nature of the Single-Sex Coeducational Differential ...................................... 36 Differences in School Attributes . . ....................... 36 Males .......................................... . 37 Females ........................................... 38 School Practices and Achievement Gain .......................... 39 Male Achievement ....................................... 39 Female Achievement ..................................... 42 Peer Group Effects . . .4........ ..4 Conclusions ..... . .47 Appendix The Basic Value Added Model ...................................................... 49 . ..... . ....... 53 Footnotes ...................................................... . 57 INTRODUCTION The relative effectiveness of single-sex versus coeducational schools on student attitudinal and cognitive outcomes has become a question of considerable interest for educators in both developed and developing countries. Whereas coeduca4.lon is seen as improving educational efficiency (Woody, 1920) and promoting positive social development for students (Dale, 1969, 1971, 1974; Schneider & Coutts, 1982), recent studies suggest that single-sex education has strong positive effects on student achievement and self-perceptions, particularly for girls (Carpenter, 1985; Finn, 1980; Hamilton, 1985; Hennesy, 1985; Jones & Shallerass, 1972; Le & EBryk, 1986; Price & Rosemier, 1972; Riordan, 1985). In developing countries where eoeducation may be culturally unacceptable, single-sex education holds the additional promise of enabling or increasing female school participation (Lycette, 1986). Contemporary literature on the effectiveness of single-sex versus coeducational schools, however, suffers from several problems of methodology, interpretation and analysis. The most important methodological issue is the difficulty in attributing differences betweeeu the attitudes and cognitive abil!ties of students in single-sex versus coeducational schools to school characteristics alone, since a variety of non-school factors also affect achievement. These include soe'io-economic background, innate ability and individual motivation. Moreover, these non-school factors also affect school choices made by families, such that students with one type of background and ability attend single-sex schools, while students with different backgrounds and Single-sex schools in Thailand ........ ... page 3 abilities attend coeducational schools. Unless non-school factors are controlled appropriately, estimates of school effects will be contaminated by what has become known as "selectivity bias." In general, research on the effects of single-sex versus coeducational schools has failed to control for differences in students selecting the two types o' schools. We are aware of only two recent studies, both of which utilize data on Catholic schools in the United States (Lee & Bryk, 1986; Riordan, 1985), that make an attempt to control for entry level characteristics of students attending single-sex versus coeducational schools; neither of them directly address the issue of selectivity bias. 1 A second shortcoming of the available lito %ture is that the causal mechanisms underlying the differences in at ig outcomes for single-sex and coeducational schools have not been ;i...;yzed. What do single-sex and coeducational schools actually do that is different? Do they employ different input-mixes or use the same input-mix differently? Is the single-sex school or classroom responsible for the apparent effect? How does the presence or absence of opposite-sex peers change the nature of classroom interchange or the amount of student effort directed towards academics? Third, the vast majority of previous analyses of single-sex/ coeducational school differentials have utilized cross-sectional data only, which does not allow for a direct measure of value added. If the achievement relationship (i.e., the educational production function) holds at both points in time, it is possible to concentrate on exactly what happens educationally between those points when outcomes are measured. Differences in achievement can be related to specific material inputs and teaching processes over a shorter time perici. Single-sex schools in Thailand....................... page 4 Moreover, the effect of omitted unmeasured factors, such as student ability or motivation, are lessened in the 'value added' as opposed to the 'level" formulation. Any "level, effects of these unmeasured variables have already been incorporated into prior (e.g., at the beginning of the school year) achievervent. This does not mean that unmeasured variables no longer affect the estimating equation. However, their effect is mitigated because only "growth" effects of omitted variables such as innate abilities would influence "value added". Technically, the error term of a "value added" equation does not contain unmeasured personal characteristics that affect achievement similarly in both periods. This ?aper contributes to the literature on single-sex/ coeducational schooling in several important dimensions. First, it extends the empirical evidence for developing countries by analyzing data from the Second International Mathematics Study (SIMS) conducted by the Inte':national Association for the Evaluation of Educational Achievemeut (IEA) in Thailand during the 1981-82 academic year. We are not aware of any other rigorous comparisons of single-sex/coeducational schools in determining achievement in developing countries, although one descriptive study was conducted in Jamaica (Hamilton, 1985). Second, to our knowledge, this is one of the first comparison studies, in developing or developed countries, of single-sex/ coeducational achievement that uses longitudinal data. (See Lee & Bryk, 1986, for a just published analysis using U.S. data.) In SIMS, students were tested at the beginning and end of the eighth grade school year. We are thus able to obtain better controls for unmeasured variables because the data base contains a direct measure of "value added' of a year of schooling. Single-3ex schools in Thailand......... . .... page 5 Third, the paper also confronts the difficult methodological questions that have arisen in other studies. An individual's status as single-sex or coeducational school student is a choice made by student and parent. If this choice is systematically correlated with personal characteristies, there may be sample selection bias. We model the effect of student and school characteristics on "value added", (i.e. gains in achievement over time) as well as control statistically for the possibility of selection bias. Fourth, we make further inquiries into the nature and single-sex/coeducational differentials in school achievement. In particular, we document differences in the availability and use of school-inputs between single-sex and coeducational schools, and examine the effects of peers on achievement prior to drawing conclusions about effectiveness. Single-sex schools in Thailand ..............0 ........ page 6 3AKG10ND Educetion In Thailand With a population of approximately 50 million and a size of 514,000 square kilometers, Thailand is one of the larger countries in Southeast Asia, antb is ne-r he size of France in both population and land area (World Bank, 1986) The present education system includes six years of primary school, followed by three years of lower secondary education and three years of upper secondary education.2 Approximately 13% of primary students are enrolled in private schools. The main types of private schools are Chinese, Muslim and Thai; a few international and religious schools also operate. Government guidelines for curricula and syllabi are followed in all schools, both government and private, although foreign language instruction is more widely avsilable in pri;ate schools than in public ones. In Grade 8, four periods of mathematics and four periods of science per week are compulsory; of the 35 total hours per week available, other compulsory subjects are Thai language (4 hours), social studies (5 hours), physical education (3 hours), art education (2 hours), various personality development activities (2-5 hours) and work and occupation (4 hours). Secondary education is selective, with admission dependent upon examinations taken at the end of Grade 6. Although western-style education in Thailand dates from King Chulalongkorn's modernization efforts in the latter part of the 19th century, formal schooling dates from the thirteenth century. For 600 years, education was based on monaatic schools situated in Buddhist temples; it concentrated on teaching boys reading, writing and religion, and girls were not educated (Buripakdi and Mahakhan, 1980). Single-sex schools in Thailand.......*.........* page 7 The introduction of Western education in Thailand changed such of the character of the former education system. One important aspect of cisnge was the inclusion of girls; as early as 1921 the Thai government introduced a Compulsory Education Act, which by 1932 covered in theory 802 of the country. Girls appear to have been included significantly from the outset. For example, in 1965, less than 50 years after the Compulsiry Education Act, 82Z of primary aged boys and 74% of primary aged girls were enrolled in school. By 1984, 7.3 million students were enrolled in primary schools and 2.3 million students were enrolled in secondary schools; primary education was virtually universal, although only about 301 of the relevant age group attended secondary school (Un-qco, 1987). At both primary and secondary levels, females comprise more than 401 of the students (481 in primary in 1980 and 421 in secondary in 1970, the last year for which published data are available). A study of determinants of school participation in rural Thailand confirms this apparent gender equity in schoooling; this study of 400 households in 22 villages found that sex was not a constraint to education participation (Cochrane and Jamison, 1982). While female education is widespread, and coeducation tolerated where alternatives are not available, single-sex education is apparently preferred for girls, particularly at the secondary level. Thus, "old" Thai, Chinese and other middle-class families typically choose to send their daughters to single-sex Catholic schools operated by nuns. One explanation given for this preference is the opportunity for students to study English more intensively than in the public secondary schools. Single-sex schools in Thailand .......page 8 'lnglo-dex education Th. evidence regarding the effectiveness of single-sex education in general is limited, and research xemininng its effectiveness vis-a-vis mathematics achievement, Dgr se, is rare* Moreover, since single-sex education is in most cases inextricably confounded with private education generally and Catholic education in particular, most research on its effectiveness has been nonfounded by selectivity with regard to sector. Nevertheless, there is growing evidence that single-sex education at the secondary level promotes both academic achievement and orientation, particularly for girls, when compared with mixed-sex education. In this paper, we use the term 'single-sex education" to include both single-sex schools and single-sex classes within coeducational schools. We do this because academic achievement is the outcome of interest, and there is evidence that single-sex classes can function effectively to improve female achievement within coeducational environments (Fox, 1976; Harvey,1985). If gender-related socialization were the outcome of interest, we would treat single-sex schools separately from single-sex classes. Academic achievement. There is considerable evidence that single-sex tducation is positive for girls; its effects on boys is mixed. Where all-male and all-female schools are grouped into "single-sex" schools, no effects are frequently observed. For example, a school-level study from Thailand (Coomber and Keeves, 1973) found no single-sex/coeducational school effect for science achievement of 14-year-old students When all-female schools are distinguished from all-male schools, however, positive effects for girls are typically found. Single-sex schools in Thailand ................. .... page 9 For example, a second study of achievement in southeast Asia surveye* 89 secondary schools and 7,674 students in West Malaysia (Beebout, 1972). In this study, a "value addded" approach was used, with performance indicated by student-level differences between secondary school entrance and completion examination scores; no other student-level characteristics were examined, and tnalyses were conducted with schools as the uni._ of analysis. The results showed that students in all-female schools outperformed those in coeducational schools in English and outperformed those in both all-male and coeducational schools in Malay. Both of the two previous studies analyzed data at the school level, and made little made attempt to control student background or public/private sector, which--in light of recent research on the relative effectiveness of private schools in comparison with public schools--is a serious shortcoming. We are aware of only one large-scale study that compares the effects of single-sex and mixed-sex education, while controlling for student background and sector. This is a recent reanalysis of High School and Beyond data froz the U.S. by Lee and Bryk (1986), who controlled for sector by selecting from the data archive 1887 students enrolled in 75 Catholic schools only (21 all-male, 24 all-female and 30 coeducational). Unlike most studies of single-sex education, Lee and Bryk also made extensive adjustments for background characteristics of students. Even controlling for sector and adjusting for student background, the effects of single-sex education on achievement were particularly strong for girls. Specifically, girls in single-sex Catholic schools gained more in reading achievement and science achievement than girls in coeducational Catholic schools; no differences in achievement were observed for boys. Single-sex schools in Thailand .................. page 10 Similar results for girls weze found by Riordan (1985) in a reanalysis of National Longitudinal Study data from the U. S. on white students in 1212 public and 37 Catholic school who were enrolled in college, business and general education tracks. This study, however, did not control for sector. Instead, comparisons were made between the performance of girls in Catholic single-sex schools and girls public coedu^ational schools. In this case, sector differences are undoubtedly partially responsible for the significant single-sex effect that was observed. Failure to take into account sector and student background may also account for results of studies in which female performance in single-sex schools is not superior to that of students in coeducational schools. For example, Carpenter (1985) studied a stratified random cample of 1286 Grade 12 students in 26 schools in Queensland, Australia. Comparisons were made between girls in non-Government single-sex schools (N - 75) and in Government coeducational schools (N = 428). No differences in overall academic achievement were found, but sector effects were not controlled. In a related study of girls in single-sex and coeducational schools in Queensland (N - 503) and Victoria (N - 632) Australia, Carpenter and Hayden (1987) found that with paternal occupation and education and maternal education held constant, the sex composition of the school had a positive effect on girls' average external public examination sccre in Victoria but not on girls' teachers' grades in Queensland. This study, however, included students in public, private and Catholic schools, with no allowance made for sector. Other studies make no mention of sector and little attempt to control for student background; most report superiority of sir.gle-sex Single-sex schools in Thailand ................. ...... page 11 education. One such recent study comes from Jamaica (Hamilton, 1985). In this stratified random sample of 1146 Grade 11 students (529 boys and 617 girls) attending 15 of the 41 high schools in Jamaica (3 all-boy, 5 all-girls and 7 coeducational), students in single-sex schools significantly outperformed their same-sex counterparts in coeducational schools on the General Certificate of Education "O" level examinations. In mathematics, in particular, the mean score of girls in single-sex schools was 5.2, compared with a mean of 3.9 for girls in coeducational schools; since the standard deviation of these scores was quite large (s.d - 6.6 for single-sex schools and 3.4 for coeducational schools) these differences were not statistically significant). A secondar analysis of IEA data from three countries came to the same conclusion regarding the superiority of single-sex schools in promoting achievement. Finn (1980) studied 14-year old students in the United States (4 all-male schools, 4 all female schools, 118 coed achools), England (19 all-male schools, 28 all-female schonle, 88 coed schools) and Sweden (95 coed schools). In the U.S. and England, both girls and boys in single-sex schools outperformed their same-sex counterparts students in coeducational sc4ools on a variety of tests, including tests of reading comprshension, word knowledge, biology, chemistry and physics. Effects were stronger for Grade 9 students than for Grade 8 students, and not all effects were statistically significant. Finally, a study comparing the effects of single-sex classes with those of single-sex schools controlled for background effects by matching studente on verbal reasoning scores (Harvey, 1985). In this study, acience achievement of students in 17 secondary schools in southwest England was e minedj results showed that: (a) girls in Single-sex schools in Thailand ....................... page 12 single-sex science classes in mixed-sex schools outperformed girls in mixed-sex classes in mixed-schools in physics, but not in biology or chemistry, and (b) girls taught in mixed-sex schools (mixed or single sex classes) outperformed girls taught in single-sex schools in physics and chemistry, but not in biology. Overall, however, the evidence suggests that girls in single-sex schools do outperform girls in coeducational schools, but with few exceptions, the previous literature does not clarify whether this is due to differences in sector and selectivity of single-sex and coeducational schools. The same comment applies to most studies of non-cognitive outcomes. Academic orientation. Recent studies have demonstrated the positive effects of single-sex schools on student attitudes and orientations toward academic activities. For example, in the study by Lee and Bryk described above, girls in single-sex Catholic schools took more mathematics courses, spent more time on homework than their peers in coeducational schools, and reported greater interest in mathematics. Positive effects were found for boys with respect to homework and mathematics course-taking only. Similarly, Jones, Shallcrass and Dennis (1972) studied 1,223 New Zealand secondary school students in two single-sex schools and one coeducational school. In comparison with students in the coeducational school, both boys and girls in single-sex schools reported spending more time on homework outside school and preferring to spend an extra hour of school studying rather than doing something else; girls in single-sex schools were more likely to want to be remembered as a brilliant student in comparison with girls in coeducational schools. Single-sex schools in Thailand ....................... page 13 Similar results come from Canada. Schneider and Coutts (1982) studied 2029 Grade 10 and Grade 12 students from five coeducational, four all-female and four all-male high schools. They found that in terms of contributing to status among their same-sex peers, students ranked "getting high grades, honor roll" more highly in single-sex schools than in mixed-sex schools. Finally, in the study by Carpenter and Hayden (1987) mentioned above, girls in single-sex schools in Victoria were more likely to be taking science courses as seniors than were girls in coeducationAl schools, suggesting that motivation toward science might be higher. Why are single-sex schools effective? What explanation does previous research provide for the greater effectiveness of single-sex schools, particularly for girls? Three explanations have been offered: differences in resources available within the schools, differences in governance and organizational characteristics, and differences in classroom "climate" for girls (see Arnot, 1983; Spender and Sarah, 1983; Lee and Bryk, 1986 for recent reviews). In addition to these, we add sector and selectivity; that is, students enrolled in single-sex schools may come from more advantaged backgrounds than those from coeducational schools, since a higher proportion of single-sex schools are in the private sector. The only study that has examined either differences in resources or differences in organizational characteristics was the Catholic school study reported above (Lee and Bryk, 1986). They found little support for the claim that differences in either accounted for the single-sex school effect. That single-sex learning environments--either single-sex classes or single-sex schools--could Single-sex schools in Thailand ............ .. ......*$.. page 14 benefit girls in particular is strongly suggested by research on classroom interaction that shows diffences in male and female participation rates in class. The majority of this research has been conducted in the United States and Canada, but the findings appear quite stable. It suggests that girls in mixed-sex classes have less opportunity to learn, both from their teachers and from classmates, than do boys in mixed-sex classes. First, in coeducational mathematics classes, teachers direct more of their attention to male students (Leinhardt, Seewald & Engel, 1979); in part, this may be due to greater disruptive behavior on the part of boys (Lockheed, 1984). Second, while girls in mixed-sex classes are equally likely to provide information to male or female classmates, boys are less likely to help their female classmates than their male classmates (Webb, 1982). This combination of less teacher attention and less peer help for girls in coeducational classes could account for the differences in les ning observed. In this paper, we will be able to examine the differential effects of resources and some organizational qualities, but will not be able to examine classroom interactional effects. Single-sex schools in Thailand ........ 0...0....too. page 15 TBE BASIC HODEL AND DATA Model In this paper we use the following final estimating formt (1) Ai8 - go + gl Ajj + 82Xi8H + g3Zi + el8 This is a basic value added model where Ait--t - 7,8--represents the i.hievement score of the ith child at the end of year t; Xi8H represents a vector of variables measuring the ith child's learning environment during 8th grade; Zi represents a vector of variables affecting achievement but which are invariant over time; aiS is a random error term and the g's are parameters to be estimated. This basic value added model is derived from more complicated models of the level of achievement, as outlined in the Appendix The interpretation of the estimated parameters depends crucially upon the assumption of the "level" models. In our version, g1 is interoreted as the rate at which the impact of seventh grade characteristics affects eighth grade achievement; g2 is the current period effect of a contemporaneous variable (e.g., parental encouragement during eighth grade) on achievement; and g3 is the cumulative effect of a fixed variable (e.g., parental background) on eighth grade achievement. There are several remaining methodological issues which require further discussion. First, note that it is important to distinguish between variables that change during the eighth grade (and thus belong to the X vector) and t},ose that are invariant over the child's schooling career (and thus belong to the Z vector). The coefficient of a Z-type variable (such as sex) cannot be interpreted as the marginal effect on Single-sex schools in Thailand ................. page 16 eighth grade achievement; rather, the coefficient is the marginal offect on eighth grad- achievement less its effect on achievement in the seventh grade. Second, the estimate of f might be biased if (1) is estimated by OLS because one of the explanatory variables, Ai7 has a random component. We do not believe that this is a major problem that would greatly affect the results regarding differential achievement of single-sex and coeducational schools, since both types of schools would be equally affected. Third, the us' of 'value added" does not necessarily make the problem of omitting unobserved variables go away -- although we would expect the problem to be mitigated. The problem is important if variables such as ability and motivation are correlated with the X's and 's (e.g., more able children are given more attention at school and at home). The coefficient of the measured variable would be biased upward or downward, depending upon its cor.elation with the unmeasured characteristics. Since we are focusing on one particular environmental effect --the single-sex/coeducational dimension -- the problem can be couched in terms of selection bias. If students are systematically selected (or self-selected) into one type or school or another on the basis of some unobserved criteria (such as ability), estimates of achievement within each school type would be contaminated by this selection effect. This problem is corrected using (now) standard statistical techniques. Sample The IEA SIMS sample comprised 99 mathematics teachers and their 4030 eighth-grade students and was derived from a two-stage, stratified Single-sex schools in Thailand ....................... page 17 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 resulting sample represented a 1Z sample of eighth grade mathematics classrooms within each region. This paper reports data from the 3265 students for whom complete data were available. A; both the beginning and end of the school year, students were administered a mathematics test covering five curriculum content areas (arithmetic, algebra, geometry, statistics and measurement). Students also completed a short background questionnaire at the pretest and a longer one at the posttest administration. Teachers completed several instruments at the posttest, including questionnaires on their backgrounds, teaching practices and characteristics of their randomly selected "target" class. Data about the school were provided by a school administrator and were supplemented by additional information provided by the Ministry of Education. Mathematics achievement The IEA developed five mathematics tests for use in SIMS. One of the tests was a forty-Stem instrument called the core test. The remaining four tests were thirty-five item instruments called "rotated forms" and designated A through D. The five test instruments contained roughly equal proportions of items from each of the five curriculum content areas, except that the core test contained no statistics items Single-sex schools in Thailand .................. *,. page 18 (Wattanawaha, 1986). For purposes of this analysis we regard the instruments as parallel forms with respect to mathematics content. The IEA longitudinal design called for students to be administered both the core form and one rotated form chosen at random at both pretest and posttest. In Thailand, students were pretested using the core test and one 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. Approximately equal numbers of students took each of the rotated forms in both administrations. One goal of our analysis was to predict posttest achievement as a function of pretest performance plus other dete-minants. Since students took the core form twice, the core form posttest score reflects, to some degree, familiarity with the core test items. Instead of using the core test, therefore, we analyzed scores obtained from the rotated forms after they were equated to adjust for differences in test length and difficulty. In this analysis, we used equated rotated form formula scores for both pretest and posttest measures of student mathematics achievement. A complete description of the equating procedure is provided in Lockheed, Vail and Fuller (1987). Student background characteristics To conform with the value added model outlined above, student characteristics are divided into two categories: time-invariant or fixed (ZH); and variable or Grade 8-specific (XH). Fixed background information about each student included his or her sex, age, number of older siblings, maternal education, paternal occupational status, educational aspirations and correspondence between home language and Single-sex schools in Thailand .......... .............. page 19 language of instruction.3 Definitions and summary statistics for each of those variables are provided in Table 1. Student characteristics thought to vary over the course of the school year include amount of out-of school tutoring in mathematics, perceived parental encouragement, and home use of a four-function calculator (a proxy for family educational resources). Parental encouragement was measured by an index which comprised four items of the type "My mother thinks that learning mathematics is very important for me," with a five-point Likert-type response alternative ranging from 1 - 'Exactly like" to 5 - "Not at all like." On this index, a low score represents high parental encouragement. Peer group, class, teacher and school characteristic. We use three measures of a student's peer group characteristics: average pre-test score, proportion of classmates having mothers with greater than primary education, and proportion of classmates having fathers with professional occupations. No data on actual peer interaction were available for analysis. Class characteristics include class size (number of students), and whether or not the curriculum was described as being "enriched". Teacher background characteristics include his or her sex and participation in in-service training. Teacher classroom teaching practices include using workbooks, maintaining discipline, and administering tests and quizzes. School characteristics include information on regional per-capita income, school size, public/private status, proportion of teachers qualified to teach mathematics, and proportion of teachers who are female. Definitions and categories for these variables are provided in Table 5. Single-sex schools in Thailand ............. ..... page 20 SLngle-sex learning envirokent. In this paper we use the term "single-sex school" to refer to single-sex learning environments: schools, grade levels, or mathematics classes. The achievement effect of single-sex schools is presumably due to the absence of other-sex peers within academic classes. Since many Thai secondary schools are "coeducational' -- that is, they enroll both girls and boys -- but segregate students by sex for instruction, we utilize the segregated learning environment as our indicator of single-sex school. Single-sex schools in Thailand ...................... page 21 THE EFFECT OF BACKGROUN 'IN ACHIEVEMENT IN SINGLE-SIK AND COEDUCATIONAL SCHOOLS A critical policy question is: Would a student, randomly chosen from the general population, do better in a coeducational or in a single-sex school, and are the effects similar for both male and female students? According to Table 1, the school composition effects for male and female students are quite different. Coeducational schools appear to enhance male achisvement, whereas single-sex schools appear to enhance female achievement. In terms of level, we note that the average scores of boys in coeducational schools are higher than those of boys in single-sex schools, whereas the average scores of girls in single-sex schools are higher than those of girls in coeducational schools, for both pretest and post-test. Moreover, simple gains in achievement (average post-test score minus average pretest score) also appear to favor coeducational schools for boys (3.82 points vs. 2.32 points) and single-sex schools for girls (3.61 points vs. 3.18 points). However, these differences are not as sharp when gain is expressed in units of pretest standard deviation, a common method of comparison. Then, while average male gain in coeducational schools remains greater than that in single-sex schools (half a standard deviation, .51 and one-third of a standard deviation, .32, respectively), female gain is approximately the same in both single-sex and coeducational schools, four-tenths of a standard deviation, with a slight advantage for coeducational schools (.40 and .42, respectively). Thus, it appears that the gross measure of "value added" during eighth grade is clearly higher for boys in coeducational schools, but that the effects for girls are less Single-sex schools in Thailand ....................... page 22 Table 13 Background and Achievemont in Sngle-sez and Coeducational Schools in Thailand, 1981-82 Male Female Variable Descriptlon& SIVlie Coed Single Coed Achievement score after 8th grade (A18) 10.50 12.65 14.94 12.19 (8.46) (9.05) (10.62) (8.82) Achievement score after 7th grade (A17) 8.18 8.83 11.33 9.01 (7.29) (7.57) (8.95) (7.62) Time-Invariant baCkgrourd (ZW). (duov.1 If): Father's oceipation: uskillIed 0.16 0.16 0.12 0.15 skilled 0.39 0.51 0.29 0.41 clerical 0.27 0.22 0.34 0.26 professional 0.18 0.11 0.25 0.13 Mother's educatIon: none 0.27 0.31 0.17 0.24 prImary 0.53 0.58 0.55 0.63 sec ndary 0.13 0.07 0.13 0.08 university 0.07 0.04 0.15 0.05 Expectations for further educatlon: < 5 years 0.43 0.53 0.27 0.35 5-8 ore years 0.35 0.39 0.41 0.45 > 8 more years 0.22 0.18 0.32 0.20 Eldest child 0.24 0.20 0.26 0.22 Language of Instruction used at home 0.39 0.50 0.46 0.51 Age In months 172.52 172.12 168.17 170.05 (9.65) (8.57) (8.92) (8.21) Private school 0.25 0.10 0.30 0.11 Barkground dirlng 8th Grada (XW): Hours of extra tutoring 1.54 1.62 1.92 1.64 (3.20) (2.63) (3.66) (2.54) Index of parental encouragement (11hlgh; 5.1mw) 2.25 2.23 1.90 1.97 (0.99) (0.99) (0.82) (0.87) Home use of 4-functlon calculator (uAmy.1) 0.26 0.28 0.37 0.30 Lambda 1.04 -0.53 1.02 -0.47 NUiber of observations: 567 1,120 502 1,076 aStardard deviations In parentheses for continuou variables Single-sex schools in Thailand ................ . ..* page 23 consistent. However, because students in coeducational and single-sex schools are also different, these gross figures should not be used to conclude that one school type is more or less effective than the other for either sex. Students in single-sex schools come from somewhat more advantaged backgrounds than their coeducational school counterparts. This is not surprising given that a higher proportion of students in single-sex schools (27Z in comparison with 11% of students in coeducational schools) attend fee-charging private schools. Approximately 21% of single-sex school students (18% of males and 25% of females) had fathers with professional occupations, compared to 122 of coeducational school students (112 of males and 132 of females). Also, a greater proportion of single-sex school students had mothers with secondary school education or above (24Z versus 122); 15Z of girls in single sex schools had mothers with some university education. These trends are reinforced by the slightly higher expectations that students in single-sex schools had regarding further education: 22% of the boys and 322 of the girls expected to complete university, as compared with 182 of the boys and 21% of the girls in coeducational schools. Girls in single-sex schools also benefitted more from parental inputs than their coeducational school counterparts: more home use of calculators, out-of-school tutoring and perceived parental encouragement. A lower proportion of students in single-sex schools spoke the language of instruction at home (39% of male students and 46% of female students as compared with 50% of male students and 51% of female students in coeducational schools). Most of these differences are modest however, and it is not clear from a simple comparison of Single-sex schools in Thailand .....................*.. page 24 means how they would affect diffirences in learning sains (as opposed to learning prior to eighth giade), although they have to be taken into account when comparing achievement scores. How we control for background To control for student or household characteristias, such as socioeconomic status (SES) and ability, while comparing achievement test scores, we use the value added achievement model developed earlier. We estimate equation (1) for students in coeducational and single-sex schools separately, under the assumption that they come from different populations4: (la) Ai8 - g8O + g"1*AE$7 + 82'*Xi8H + g 3'*Z"iH + e*i8 (lb) AAi$ - g8o + gAl*AAi7 + gA2 *X'i8H + gA3 *ZjiH + *Ai8, where the superscripts - and ^ represent single-sex and coeducational sectors, respectively. Note that only household subvectors (subscripted by H) of X and Z are of concern and to simplify notation. OLS regressions on (la) and (lb) for coeducational and single-sex school students might lead to misleading results because of the selection bias. Suppose that students and parents are free to choose whichever type of school they prefer. One type of selection results if students sort themselves into those institutions where they think they can perform the best. There would be positive selection in both single-sex and coeducational school samples. Another alternative is that students are hierarchically sorted. For example, if there is excess demand for places into the coeducational schools and the best Single-sex schools in Thailand....... page 25 students are selected, there would be positive selection into coeducational schools but negative selection into single-sex ones. A third alternative is that the selection process operates differently for male and female students. In all cases, the analyst cannot observe the characteristics of single-sex school students among the coeducational school sample or vice versa. Because the subsamples are not a random draw from the student population, the assumptions of the basic linear model and could lead to biased estimates of the achievement 5 effect To correct for sample selection, we use Heckman's two-step methodology (Heckman 1979). The first step in this methodology is to estimate what determines the choice of type of school (see Cox and Jimenez 1987 for a model of school choice): (2) Ji - ko + klYi +wi, where Ji- 1 if the ith child learns in a single-sex learning environment, and J = 0 otheIrwise; Y indicates the explanatory variables and w is a random error term. The second step is to use the results of the first step to correct for the selection bias in (la) and (lb). If we assume that (wi, ei) are jointly distributed, then Heckman (1979) has shown that: (3a) E(e-ilIi>0) - g"4*X)i, and (3b) E(eAilIi 8 we years 0.05 0.61 0.24 2.53 Age -0.09 -1.19 -0.07 -0.92 Age squared 0.00 1.29 0.002 0.84 Eldest child 0.09 1.07 0.06 0.69 Language at home -0.45 -6.61 -0.42 -5.64 Private school 0.75 8.29 0.68 7.32 Nkber of observatlons: 1,687 1,578 Log-lIkelIhood: -1002.1 -900.3 language as the language of instruction. Sons of skilled blue-collar fathers are less likely to choose single-sex schools, and sons of mothers with secondary or university level education are more likely to choose single-sex schools. For girls, the most important background variables in determining single-sex school choice are maternal education, educational expectations, home language and private school choice. Daughters of mothers with secondary or university level education are more likely to choose single-sex schools, and those with expectations to complete college are also more likely to do so. Again, as for boys, girls tend Single-sex chools in Thailand ............... .... page 28 to choose private single-sex schools that do not use their home language as the language of instruction. The parameters of the probit equations in Table 2 can now be used to estimate the terms that will be used to correct for the selection bias. The average X in equations (3a) and (3b) are shown in the penultimate row of Table 1. How does socio-economic background affect school achievement? The variables that are used to explain achievement scores in Thailand (i.e., the vectors Xi8H and ZiH) include many of the same variables that are used in Table 2. However, the variables that represent XiBH should affect achievement scores only, since the decision to select a single-sex or a coeducational school was taken well before the student started eighth grade. This set includes variables that measure parental encouragement for mathematics, out-of-school tutoring during eighth grade, as well as the availability of school aids, such as calculators. Finally, the achievement equation includes a term that holds constant for the selection bias --i.e., for the probability that a given student will be in single-sex schools. This term is derived from parameters in the choice equation, as described earlier. The estimated achievement equations (la) and (lb) are presented in Tables 3 and 4, for male and female single-sex and coeducational school students, respectively. These equations can be used to estimate whether or not a school achievement advantage exists for students in coetducational or the single-sex schools after holding constant for student background and selection. Single-sex schools in Thailand.......... ......... page 29 Table 3: Male .Achievasent Functions for Single-sex and Coeducatioual Schools in Thailand, 1981-82 Cnmfficients Varlable SILg.s -.Coed Value t-stats Value t-stats Constant 16.41 0.41 59.80 1.20 Past achievement 0.68 17.79 0.79 28.94 Time-invariant bCkarOrund (ZH) Father's occuatIon skIled 1.75 2.16 -1.02 -1.69 clerical 2.07 2.38 -0.82 -1.24 professlonal 2.16 2.19 -0.27 -0.32 Mother's education priNary 0.99 1.58 0.28 0.64 secondary -0.22 -0.22 0.05 0.06 university 1.53 1.23 -0.97 -0.87 Educatlonal Expectations 5-8 more years 0.91 1.46 0.48 1.08 > 8 more years 1.46 2.02 1.80 3.15 Age -0.07 -0.15 -0.56 -1.03 Age squared -0.001 -0.05 0.001 0.93 Private school -2.00 -1.98 0.47 0.50 Backgroutd ducin; gth Grade (XH) Tutorlng -0.04 -0.51 -0.17 -2.30 Parental encouragement -0.13 -0.47 0.01 0.06 Home calculator 0.51 0.83 -0.24 -0.52 Labda: -0.31 -0.17 1.55 0.89 R-scuared 0.459 0.481 F-stats 29.184 63.823 Single-sex schools in Thailand ...................... page 30 Table 4t Fm_le Achievmeent functions for Single-sex and Coeducational Schools in Thailsad, 1981-82 Coaffielents Varilable Sin la Coad Value t-stats Value t-stats Constant -24.68 -0.36 -28.31 -0.75 Past aehlevement 0.82 20.60 0.78 30.12 Time-invariant bahkground (ZHj Father's oc4uatlon skiIled 1.15 1.09 0.05 0.10 clerieal -0.60 -0.55 -0.12 -0.19 professional 0.15 0.12 0.04 0.05 Mother's ed*aticn prliary 1.g1 2.18 u.3i O.68 secondary 1.56 1.23 1.41 1.72 university 0.78 0.54 0.65 0.58 Edat I onal ExIpectatIons 5-8 more years 0.88 1.11 1.t67 3.82 ; 8 more years 0.95 1.00 2.16 3.70 Age - 0.44 0.55 0.45 1.03 Age seuared -0.001 -0.60 -0.001 -1.21 Private school -2.57 -2.09 1.31 1.51 la*kround dirin 8th Grada (Xgj) Tutoring -0.04 -0.46 -0.04 -0.54 Parental encouragement -0.15 -0.38 0.46 2.10 Home caclulator 1.81 2.68 -0.13 -0.32 Lambda -4.65 -1.99 2.34 1.36 R-squ ured 0.584 0.513 F-stats 42.587 69.722 Single -sex schools in Thailand .................. ...... page 31 As explained earlier, the interpretation of the coefficients of the stock variables (ZiH) differs from that of the flow variables (XiH)* The former are non-marginal effects -- they represent the cumulative effects on past achievement as well. The estimate of the decay rate of the effect on current year achievement of a previous year's characteristic is .68 for single-sex schools; and .79 for coeducational schools. The achievement results will be discussed separately by sex. Male achievement. For the single-sex school students, paternal occupation (reference category: unskilled) and educational expectations (reference category: less than five more years of school) are statistically significant predictors of achievement. However, maternal educational attainment (reference category: no education), age, tutoring, parental encouragement and presence of a home calculator are all insignificant. Surprisingly, enrollment in a private school is negatively related to achievement gain. For the coeducational sample, educational expectation is the only variable positively and significantly related to achievement gain, and the effect of out-of-school tutoring is negative. The selection term (coefficient of Lambda times its mean) is negative for students in both single-sex and coeducational schools, but the effect is not statistically significant. Thus, for males, the impact of selection on the achievement equation is unimportant. Female Achiev emnt. For the single-sex school sample, the only background variable that has any statistically significant effect on achievement gain is maternal education (reference category: no education). Presence of a home calculator is significant, but enrollment in a private school is negatively related to achievement gain. For the coeducational school sample the only statistically Single-sex schools in Thailand ......................... page 32 significant variables are educational expectations and parental encouragement; private schools have no effect on achievement for this group. The selection term (coefficient of Lambda times its mean) is negative for both single-sex schools and coeducational schools, and the effect is significant for single-sex schools. Background constant, is there a single-sex school effect? The estimated differential in coeducational and single-sex school students' achievement scores can be computed from the parameters presented in Table 3 to hold constant for the effect of background. We do this separately for male and female students. Because single-sex and coeducational school achievement equations differ in terms of intercept and slope, the comparison would be affected by the values of the other explanatory variables, as well as the coefficients in these equations. To clarify this, we compute the following unconditional single-sex school effect: From the entire sample of single-sex and coeducational students, consider a randomly chosen pupil with the average characteristics of coeducational school students (i.e., standardize according to coeducational school means). The unconditional effect measures the increment (or decrement) in test score had that student been sent to a single-sex school.6 The same calculations can be performed standardizing at the single-sex school means. The question would be: how would the average single-sex school student have done had he/she been in coeducational school? There is no theoretical reason to prefer one method of standardization over another. However, as in the index number problem, there is no guarantee that the results will be consistent with one another. The results are summarized in Table 5. Single-sex schools in Thailand..................... page 33 Table 5: Single-sex School Effects After Holding Constant for Background Characteristics, Thailand, 1981-82- Predicted scores of average single-sex school student if that student were in: Single-sex-Coed Single Coed Difference Unconditional effects for: Male students 10.81 13.10 -2.29 Female students 19.66 15.84 3.82 Predicted scores of average coeducational school student if' that student were ifi: Single=sex-Coed Single Coed Difference Unconditional effects for: Male students 11.55 13.44 -1.89 Female students 18.25 13.44 4.81 aCalculated from Tables 1, 3 and 4. The results from the top panel of Table 5 indicate that, after holding constant for past achievement and socioeconomic background, girls in eighth grade single-sex schools have an unconditional advantage in mathematics test performance of about four points and boys have an unconditional disadvantage of about two points. This implies that a Thai eighth grade girl with the background of an average single-sex school student, chosen randomly from the population, would improve her achievement by about 40% by attending a single-sex school, while a boy would reduce his score by 20%. To check the robustness of this result, we calculated the single-sex school effect for a randomly chosen student with the average Single-sex schools in Thailand .. ........... ........... page 34 coeducational school characteristics. The results, shown on the lower panel of Table 5, are not substantially different. Single-sex schools in Thailand ....................... page 35 THE NATURE OF THE SINGLE-SEX COEDUCATIONAL DIFFERENTIAL The previous section has shown that, in Thailand, girls in single-sex schools score higher in mathematics achievement at the end of eighth grade than do girls in coeducational schools, but that the reverse is true for boys, after controlling for previous achievement, socioeconomic background and systematic selection by school type. For policy makers, the remaining question is, what accounts for these achievement differentials? Is it possible to identify school characteristics that contribute most to these achool effects? 'Wthat do administrators and teachers do that is different? What is the influence of a student's peers on relative achievement? This section attempts to answer some of these questions for Thailand. The method is to redo the estimates of the previous section. However, this time, the full achievement equation (3) is estimated; that is both Xig - 1Xi8H Xi8S1 and Zi - (ZiH ZiSl are included in the equation. We will, then discuss how the school-specific components in these vectors of explanatory variables affect achievement in coeducational and single-sex schools. Differences in School Attributes: Table 6 presents, by type of school, the mean values of the school, classroom and teacher attributes that will be used in the estimation procedure. They indicate some basic differences between single-sex and coeducational schools. Again, since the single-sex schools were more effective for female students, while coeducational schools were more effective for male students, this section discusses each type of school separately by sex. Single-sex schools in Thailand...................... page 36 Table 63 Characteristics of Male and F_ ale Single-sex and Coeducational Schools In Thailand, 1981-82 Means and (SO) Male Fouale Varlable Descrlption Single Coed Single Coed Schnol-lovel darantmristios Average district Per capita Income In baht 13770.0 12408.0 14951.0 13052.0 (3311.1) (4908.3) (46B1.8) (4944.1) School enrollment 2024.5 1213.1 1620.8 1290.9 (1178.6) (973.7) (1003.3) (935.3) Prnonrtiln of teachers oalifled 0.55 0.53 0.45 0.54 to teach ath In student's school TIMd ard clasm Mlaractorlst[ls: Teacd 's age In years 28.38 29.63 32.35 29.80 (4.69) (7.98) (8.73) (8.13) Proportion ale 0.50 0.38 0.28 0.28 Proportion hiving In-servlce training 0.05 0.12 0.17 0.14 Proportion teaching enriched math class 0.13 0.20 0.24 0.27 Proportion uaing workbook often 0.17 0.21 0.42 0.23 Proportion spending > 15 sins/week maintaining order 0.72 0.43 0.58 0.43 Minutes/week spent on qulizzes and tests 47.96 29.37 35.91 28.79 (83.13) (22.76) (25.51) (21.57) Nuiber of students In target class 44.70 41.13 44.00 41.36 (4.04) (12.09) (4.70) (11.74) Per gmm characteristics Average of average pre-test scores 8.18 8.40 11.33 9.47 (3.72) (4.27) (8.95) (4.70) Average proportion mothers > prlwy .dKcatIon 0.20 0.12 0.28 0.13 Average proportion fathers prof owocatlon 0.18 0.11 0.25 0.13 Males. In comparison with single-sex schools, coeducational schools appear advantaged in one respect and disadvantaged in two others. On the one hand, they are smaller --suggesting greater Individualization of program. On the other hand, they are located in poorer regions -- suggesting less access to resources--and have somewhat fever teachers qualified to teach mathematics. Single-sex schools in Thailand .................... .... page 37 Characteristics of teachers and their teaching practices, however, gen.rally tend to favor coeducational over single-sex schools. Teachers in coeducational schools are older than those in single-sex schools--suggesting greater experience--and a higher proportiou of students in the coeducational sample have teachers who have undergone some form of in-service training. Less time is spent on maintaining order in coeducational classes. A higher proportion of students in coeducational schools are enrolled in enriched mathematics classes and use commercially-produced workbooks; class sizes in coeducational schools are smaller. On the other hand, students in single-sex classes spend more time taking quizzes and tests. Females. Despite the apparent advanta&e given to female students by single-sex schools, the schools themselves do not appear particularly advantaged in comparison with coedsucational ones. Single-sex schools are more likely to be located in more wealthy regions of the country, to have older teachers and to ha,e students using commercially-produced workbooks. On all other characteristics, they appear disadvantaged relative to coeducational schools. Single-sex schools are larger, have fewer teachers qualified to teach mathematics, have teachers who spend more time maintaining order, and have larger classes. The rough picture provided by this comparison of means is that schoolcharacteristics might account for male advantage in coeducational schools, but that there are few -- if any --characteristics that explain the positive effects of single-sex schools on female achievement. Single-sex schools in Thailand .................... page 38 School practices and achievement gain The re-eatimation of the student achievement functions includes the additional variables listed in Table 6. As before, single-sex and coeducational school functions are estimated separately. Aside from statistical reasons for not assuming homogeneity of slope and intercept coefficients, separate estimations reflect the fact that unmeasured management practices and "school culture" could differ between coeducational and single-sex schools. Teachers and administrators probably face an entirely different set of constraints, depending upon which type of school they work in. Thus, we expect the coefficients of each of the school or teacher related variables to differ for coeducational and single-sex school students. The coefficients are presented in Table 7; they demonstrate considerable differences in effects for male and female students and for those enrolled in coeducational and single-sex schools. We therefore again discuss the results separately by sex. Hale Achievement. School level characteristics have greater effects on male achievement in single-sex schools than in coeducational schools. Of the eleven variables examined, seven have statistically significant effects for single-sex school students, while only three have such effects for coeducational school students. In single-sex schools, larger schools and older teachers have positive effects on student achievement. By comparison, student achievement is lower in schools with more qualified teachers, more male teachers, smaller classes, in classes in which the teacher spends more time maintaining order, and in classes considered to have an enriched curriculum. In coeducational schools, teacher qualifications and enriched classes are Single-sex schools in Thailand ..................... page 39 positively associated with achievementj time spent maintaining order is still negatively associated with achievement. Single-sex schools in Thailand ....................... page 40 Table 7t Achi-v_imnt Functions for Single-sex and Coeducational Schools In Thailand, 1981-82 Male Feale SIngle-sex Coed Single-sex Coed VarIable Value T-stats Value T-stats Value T-stats Value T-stats Constant 1 10.15 0.28 152.54 1.14 142.83 0.68 1-11.09 -0.30 Past achievement 1 0.59 13.29 t 0.68 20.98 1 0.58 12.73 0.68 21.00 Tlme-lnvariant badcround (Z) Father's occupatlon IIt skIlled 1 2.17 2.69 t-0.77 -1.30 1 0.&6 0.69 1t-0.05 -0.09 clerical 1 1.93 2.27 1-0.79 -2.22 t1-06.2 -0.61 t1-0.38 -0.62 professIona I t 2.33 2.40 1-0.58 -0.68 1 0.33 J.27 14-0.03 -0.04 Mother's education O t I : prlmary I 0.93 1.51 1 0.30 0.69 1 1.42 1.71 1 0.28 0.57 secondary 1 -0.371 -0.66 41-0.50 40.5S I 0.94 0.79 11.15 1.44 university 1 1.01 0.80 1-1.83 -1.62 1 0.33 0.24 1 0.10 0.09 Educatlonal expectations 1 1t 5-8 more years 1 0.90 1.47 t 0.37 0.85 I 0.59 0.78 11.48 3.47 D 8 more years 1 0.95 1.32 1 1.48 2.60 1 0.91 1.01 11.75 3.04 fne 1 -0.03 -0.08 10-0.51 -0.96 14-0.40 -0.52 1 0.22 0.51 Age sqiared 1 -0.002 -0.13 1 0.001 0.85 1 0.001 0.41 1-0.001 -0.68 Private school 1 -0.96 -0.89 1-0.65 -0.70 1-1.36 -1.10 1 0.01 0.00 Background durlng 8th Grade (XH) Tutoring 1 -0.04 -0.54 1-0.12 -1.67 1-0.03 -0.33 1 0.02 0.27 Parental encourage. t -0.04 -0.14 1 0.04 0.20 14-0.13 -0.36 1 0.42 1.98 Home calculator 1 0.40 0.66 1-0.52 -1.12 1 0.88 1.37 1-0.50 -1.21 Pear gcm clrlng *th Grade ( Average pre-test scot 0.44 4.23 1 0.41 6.90 t 0.73 8.24 t 0.39 7.25 Prop mothers > prim 1 1.48 0.47 1-1.32 -0.46 1-3.04 -1.02 1.39 0.50 Prop fathers prof t -3.32 -1.12 1 1.52 0.62 1t4.29 -1.09 1-0.21 -0.10 I I I I Lambda 1 -0.50 -0.27 1-0.33 -0.19 1-1.15 -0.52 1 1.90 1.01 Single-sex schools in Thaland ....... .... page 41 Female Achievement. Characteristics play an important role in explaining female achievement gain in single-sex schools, but the signs of three of the five statistically significant coefficients are reversed. Thus, positive achievement effects are found for teacher qualifications and larger classes -- both of which were negatively related to achievement in all-male schools-- and negative achievement effects are found for school size -- which was positively related to male achievement in single-sex schools. Inservice training and time spent on maintaining order in the classroom are negatively relate.d to achievement for girls. In coeducational schools, students in enriched classes and those in schools with more qualified teachers score higher on tests of mathematics, while those with teachers who spend time maintaining order score lower. Is there still a single-sex school effect even after differences in teaching practices and school characteristics are held constant? In order to answer this question, we compute the unconditional single-sex school effect for a randomly chosen student with the average background and the average peer group characteristics of single-sex students. We do this separately for male and female students. According to the first and second panels of Table 8, the achievement advantage of single-sex schools for females is lessened with the addition of teacher and school variables, but there is still a residual effect of two to three points on average. Similarly, the advantage of coeducational schools for boys is maintained. Such differences that remain indicate unmeasured differences between male and female single-sex and coeducational schools that influence achievement and that may have to do with differential peer effects. Single-sex schools in Thailand ............... page 42 Table 8: Achievement Functions for Single-sex and Coeducational Schools Thailand, 1981-82 Male Female Single-sex Coed Single-sex Coed Variables Value T-stats Value T-stats Value T-stats Value T-stats Constant 130.21 0.77 :63.57 1.36 177.58 1.18 1-19.39 -0.51 Past achievement 1 0.63 16.15 0.74 26 25 , 0.64 14.85 1 0.75 27.43 I I I. I Time-invariant background (Z7H Father's occtvation I skilled 1 1.42 1.62 1-0.44 -0.71 1 1.06 0.96 1 0.08 0.14 clerical 1 1.63 1.94 1-0.65 -1.00 1-0.89 -0.87 1-0.11 -0.18 professlonal 1 2.24 2.35 1-0.05 -0.08 10.08 0.07 1 0.06 0.08 Mother's educatlon II primary 1 0.59 0.97 I 0.30 0.70 1 1.35 1.61 1 0.43 0.84 secondary -0.62 -0.61 1 0.05 0.06 0.98 0.79 1.43 1.77 university 1 1.00 0.80 1-1.23 -1.10 1 0.18 0.13 10.95 0.87 EducatlonaI expectatIons Is 5-8 more years 1 0.84 1.36 1 0.25 0.57 1.32 1.73 1 1.60 3.71 . 8 aore years 1 1.39 1.94 11.62 2.88 1.78 1.82 2.03 3.50 Age I -0.10 -0.22 14-0.62 -1.16 t-1.01 -,.31 0.35 0.80 Age squared I 0.00 0.07 1 0.001 1.06 1 0.003 1.21 1-0.001 -0.98 Pivate school 1 -3.13 -1.64 1 0.72 0.63 1 7.08 2.01 1 1.99 2.00 Background durlng 8th Grade (Xe) Tutorlng I -0.10 -1.25 1-0.15 -1.97 1-0.01 -0.16 1-0.04 -0.49 Parental ercworage. I 0.03 0.12 1-0.01 -0.03 1-0.09 -0.25 1 0.46 2.12 Hume calculator 1 0.37 0.63 1-0.42 -0.90 1 1.25 1.95 1-0.25 -0.61 I I I I Grada 8 school characteristics District-level per 1 0.00 0.53 1-0.00 1.86 1-0.000 -1.48 1 0.000 0.72 capita Income I I I I School enrollment 1 0.002 3.98 1-0.00 -0.20 1-0.003 -4.84 I 0.000 0.91 Teacher-s math qual.l -4.58 -2.23 1 1.84 2.55 1 4.65 2.55 1 1.51 2.04 Class size I -0.54 -3.15 1 0.01 0.32 1 0.93 5.96 1-0.02 -1.27 Teacher's age 1 0.52 5.66 1 0.01 0.35 1-0.10 -1.80 1-0.02 -0.50 Male teacher I -4.79 -4.50 -0.11 -0.24 1 0.49 0.33 1-0.24 -0.50 Teach. In-serv. trg.1 -2.97 -1.65 1 0.59 0.91 -10.45 -4.23 1 0.58 0.96 Enriched math class I -5.57 -4.01 1 1.86 3.40 1-1.48 0.90 1 1.38 2.74 Used workbook often 1 -0.11 -0.08 1-0.44 -0.86 1 0.76 0.40 1 0.13 0.25 Maintain order 1 -4.31 -4.99 1-1.24 -3.00 1-4.84 -4.67 1-1.45 -3.35 Mine. test/taking 1 -0.00 -0.03 1-0.01 -1.49 1-0.06 -1.58 1-0.01 -0.71 Lambda 1 0.77 0.37 l 1.24 0.69 1-3.30 -1.18 I 3.78 1.93 Single-sex schools in Thailand ............ ........O.. page 43 Peer group effects Because students interact with each other while in school, the ab4iity and *oc4o-economic status of fellow students could affect individual achievement. Thus, peer group differences could account for the observed single-sex/coeducational school effect differences. However, on average, the peer groups of both boys and girls in single sex schools are more advantaged than those in coeducational schools; mothers are more educated and a higher proportion of fathers have professional employment. Nevertheless, we reran the achievement equations with the addition of three classroom-level variables: average pretest score, proportion of mothers with greater than primary education and proportion of fathers with professional occupations in the school where the student is studying. The results of these regressions are shown in Table 8, separately by sex. They show that average schoolmate pretest score is highly correlated with the individual's posttest score, for both girls and boys, and for both single-sex and coeducational schools. In order to determine the extent to which peer groups affect the single-sex-coeducational differential, we compute the unconditional single-sex school effect for a randomly chosen student with the average background and the average peer group characteristics of single-sex students. A comparison of the third and fourth panels of Table 9 with that of Table 5 reveals that, for males, the single-sex disadvantage declines to less than one point after peer group characteristics are taken into account, and the single-sex advantage for girls is even further eroded. A similar analysis, standardizing at the coeducational Single-sex schools in Thailand....................... page 44 school mean characteristics produces similar results. We conclude that peer groups account for the bulk of the difference in achievement effects between coeducational and single-sex school achievement. Single-sex schools in Thailand .......... o.......... page 45 Table 9: Singl*-aex School Effects After Holding Constant for School Characteristics and Peer Group, Thailand, 1981_82a Predicted scores at average background of single-sex school student and single-sex school characteristics if that student were in: Single-sex-Coed Single Coed Difference Unconditional effects for: Male students 9.73 12.28 -2.55 Female students 18.29 16.15 2.14 Predicted scores of average background of coeducational school student and coeducational school characteristics if that student were in: Single-sex-Coed Single Coed Difference Unconditional effects for: Male students 12.66 13.28 -0.62 Female students 16.88 13.99 2.89 Predicted scores at average background and peer group characteristics of single-sex school students if that student were in: Single-sex-Coed Single Coed Difference Unconditional effects for: Male students 11.01 11.84 -0.83 Female students 16.11 15.73 0.38 Predicted scores at average background and peer group characteristics of coeducational school characteristics if that student were in: Single-sex-Coed Sinale Coed Difference Unconditional effects for: Male students 11.83 12.48 -0.65 Female students 14.46 13.09 1.37 aCalculated from Tables 1, 7 and 8. Single-sex schools in Thailand ....................... page 46 CONCLUSIONS This paper provides evidence regarding the relative effects of single-sex and coeducational school in enhancing eighth grade mathematics achievement in Thailand. It uses pre-eighth grade and post-eighth grade test scores to estimate value added equations for single-sex and coeducational schools, after controlling for the effect of selection into those schools, as well as the direct effect of background on achievement. its preliminary conclusions are the following: First, girls in single-sex schools do significantly better than their coeducational school counterparts, after holding constant for selection and background factors, while boys in coeducational schools do better. Thus, there is not a unique single-sex/coeducational school effect on enhancing achievement, but this effect interacts strongly according to the sex of the student. Second, although we expect that solection biases are mitigated in "value added" formulations of achievement functions (as opposed to "level" formulations), they are still significant in Thailand. Moreover, failure to correct for selection biases results in a singificant underestimate of the positive effect of single-sex schools in enhancing female achievement. Third, even after measured inputs and school practices are held constant, a single-sex school advantage for females and a coeducational school advantage for males persist. Fourth, peer "quality" effects in single-sex and coeducational schools appear to account for most of the difference between the two Single-sex schools in Thailand ...................... page 47 types of schools and their relative effectiveness for male and female students. With the present data, we are unable to explore precicesly how the peer group affects achievement; studies in developed countries suggest, however, that class-level participation rates and leadership opportunities are suppressed for girls in coeducational setting and for boys in single-sex settings, particularly for stereotypically male activities (for reviews, see Lockheed, 1983; Lockheed, 1985). Finally, even after controlling for peer effects, a small residual remains. This suggests that there are important managerial incentives and practices in female single-sex settings and male coeducational settings that result in enhanced achievement. Single-sex schools in Thailand..................... page 48 APPENDIX: THE BASIC VALUE ADDED MODEL Boardman and Murnane (1979) have shown that, in value-added models, the interpretation of the estimated coefficients and the properties of these coefficients depend crucially on the nature and validity of the assumptions required to derive the estimating equation. Accordingly, we devote the following section to a discussion of the implicit assumptions in our model. Factors which affect achievement can be quantified through the use of statistical inference. The "ith" pupil's score in the eighth grade mathematics achievement test is characterized by the following equation: (Al) Ai8 - aO + a8 *Xi8 + a7'*XL7 + + al'*X + C8'*Zi + da*Ii + Ui8, and the ith pupil's score in the seventh grade test can be similarly expressed as: (A2) Ai7 - bo + b7'*Xi7 + b6'*XL6 + -.. + al"Xil + c71*Zi + d7*Ii + ui7- The symbols represent: Ait Achievement score of the ith child a the end of year t - 7,8; Single-sex schools in Thailand ............ page 49 Xit A vector of variables deseribing the ith child's learning environment during year t, such as non-school tutoring, parental encouragement, availability of study mat'^!4Jls at home, characteristics of teache other school-related characteristics specific to year t, peer characteristics; the vector describing learning environment can be partitioned into two vectors, one which is school-related (S) and another which is child or household-related (H), or: Xit - [XitS XitH]; Zi A vector of variables affecting achievement but which 15 invariant over time, such as the quality of the home environment or parental inputs (such as parents' education), educational aspirations, and student characteristics (sex, age); Ii A vector of variables describing unobserved influences, such as innate ability or pre-school care; uit A random disturbance term with a zero mean and variance v2; at' A vector of coefficients describing the effect on achievement at the end of the eighth grade of a one unit change in the child's environment at year t (e.g., a7 is the marginal effect on eighth grade achievement of environmental characteristics during the seventh grade); bt' A similar interpretation as the a's for achievement at the end of the seventh grade; ct' A vector of coefficients of the marginal effect on eighth grade achievement of a unit change in one of the Z variables; dt The marginal achievement effect of the unobserved component; Indicates a transposition of a column vector into a row vector. It is not feasible to estimate equations (Al) and (A2) because researchers rarely have much information on past characteristics, such as class size or parental tutoring four or five years before. An ilternative is to estimate (Al) and (A2) as "value added" equations by Single-sex schools in Thailand ....................... page 50 subtracting (A!) from (Al) (see Hanushek 1986 for a more thorough review of the arguments). The resulting equation could groatly simplify the specification if some of the Xit's cancel each other out. Howev-r, for this to happen, additional assumptions are necessary. Boardman and Murnane (1979) have demonstrated the importance of deriving the empirical form of the "value added" equations carefully since each specification imposes behavioral restrictions. The specifications in (Al) and (A2) imply that school and student characteristics in previous years also affert current achievement. For example, the size of a student's class in the seventh grade (on down to the 1st grade) affects his/her eighth grade achievement. However, we do not expect characteristics in previous years to have the same effect on current period achievement as current period characteristics. A more reasonable expectation is that the effect of past characteristics on current achievement diminishes over time. If this effect diminishes geometrically, then, a simple "value added" equation can be derived from equations (Al) and (A2). Let the unsubscripted variables "a", "c", and "d" represent the "true" current period effect of a component of X, Z and I, respectively, on school achievement. For example, "a" is the marginal effect of an increase in one unit of class size during the eighth grade on eighth grade achievement. Thus, a8 - a. Let f be the amount by which the effect of seventh grade characteristics on eighth grade achievement diminishes relative to a, or a7 - f*a.1 The critical assumption is that the effect of previous years diminishes geometrically thereafter. !We would expect f < 1. However, this is not a restriction since f is a parameter to be estimated. In the unlikely event that the estimated f > 1, we conclude that past characteristics have greater importance than present characteristics in explaining present achievement. Single-sex schools in Thailand ....................... page 51 Or, *6 * f2*aB ..., a1 f7*a, for f < 1. By similar reasoning, the impact of previous years' characteristics on 7th grade achievement ist b7u a, b6 * f*a, ..., bl _ f6*a. These definitions can then be substituted into (Al) and (A2).2 Then, if (.%2) is multiplied by f before it is subtracted from (Al), and terms are cancelled out, the following simple specification is obtained: (A3) Ai8 - g0 + gl*Ai7 + 92'*Xi8H + g3'*ZiH + ei8 where go - (ao-f*bO), g1 - f, g2'- a', ga'- c'*(l-f) and *ei - do*(l-f)*Ii + (ui8-f*ui7). This estimating equation is intuitively appealing because the terms describing previous environments Xit, t-1,...7, are deleted and the (unrestricted) coefficient of Ai7 can be easily interpreted as f. 2The properties of this lag structure are well known in the applied econometrics literature. Single-sex schools in Thailand ..................... page 52 uEgREINCgs Boardman, Anthony E. and Richard J. Murnane, "Using panel data to improve estimates of the determinants of educational achievement," Sociology of Education, 52: 113-121, 1979. Beebout, H. The Droduction surface for academic achievement: An economic study of the Malaysian secondary education. Unpublished doctoral dissertation, University of Wisconsin, 1972. Bossiere, Maurice, John Knight and Richard Sabot, 'Earnings Schooling Ability and Cognitive Skills," American Economic Review, 75(2): 1016-1030, 1985. Buripakdi, Chalio and Pratern Mahakhan, "Thailand" In T.N. Postlethwaite & R.M. Thomas, (Edo.), Schooling in the ASEAN Region. Oxford: Pergamon Press, 1980. Carpenter, Peter, "Single-sex schooling and girls' academic achievements", Australian and New Zealand Journal of Socioloay, (November, 1985), 21(3): 456-472. Cochrane* Susan and Dean T. Jamison, "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, 1982. Comber, L.C. and J. Keeves. Science Education in Nineteen Countries. Stockholm: Almquist & Wiksell, 1973. Coven, R. and M. McLean (Eds.). International Handbook of Educational Systems. Vol.III. New York: John Wiley F Sons, 1982. Cox, Donald and Emmanuel Jimenez, "Private-public differences in secondary school performance: the role of selection effects in Colombia and Tanzania," World Bank: Education and Training Department (processed), 1987. Dale, R. R. Mixed or single-sex school? Volume I: A research study about pupil-teacher relationshis. London: Routledge and Kegan Paul, 1969. Dale, R. R. Mixed or single-sex school? Volume II: Some social aspects. Londons Routledge and Kegan Paul, 1971. Singlo-sex schools in Thailand......... page 53 Dale, R. R. Mixed or sinxle-sex school? Volume III: Attainment, attitudes and overview. London: Routledge and Kegan Paul, 1974. Finn, Jeremy D. 'Sex differences in educational outcomes: A cross-national study", Sex Roles (1980), 6(1)s 9-26. Fox, Lynn H. The effects of sex role socialization on mathematics participation and achievement, Final Report. Washington, DC: National Institute of Education, 1976. Hamilton, Marlene A. "Performance levels in science and other subjects for Jamaican adolescents attending single-sex and co-educational high schools", Science Education (1985), 69(4): 535-547. Hanushek, Eric. "The economics of schooling: Production and efficiency in the public schools," Journal of Economic Literature (September 1986): 1141-77. Harvey, T.J. "Science in single-sex and mixed teaching groups." Educational Research (1985), 27(3), 179-182. Reckman, James J. "Samplei selection bias as specification error," Econometrica (January 1979): 153-161. Hennessey, Martha Scott. Female adolescence as a function of educational context: Coeducational and single-sex schooling. Unpublished doctoral dissertation, University of Pennsylvania, 1985. Jones, J. Charles, and Jack Shallcrass. "Coeducation and adolescent values," Journal of Educational Psycholoay (1972), 63 (4): 334-341. Lee, Valerie E, and Anthony S. Bryk. "Effects of single- sex secondary schools on student achievement and attitudes", Journal of Educational Psychology, (1996), 78 (5), 381-395. Lockheed, Marlaine E. Legislation against sex discrimination: Implications for research. Paper presented at the 60th Annual Meeting of the American Educational Research Association, San Francisco, 1976. Single-sex schools in Thailand ................ ....... page 54 Lockheed, Marlain. E. "Sex equity in classroom organization and climate." In S.S. Klein (Ed.), Handbook for Achieving Sex Equity Through Education, Baltimore: Johns Hopkins University Press, 1985. Lockheed, Marlaine E. "Sex and social influence: A meta-analysis guided by theory." In J. Berger and M. Zelditch (Eds.), Status Attributions and Rewards. San Francisco: Jossey-Bass, 1985. Lockheed, Marlaine E., et al. Sex and ethnic differences in middle-school mathematics, science and computer science: What do we know? Princeton, N.J: Educational Testing Service, 1985. Lockheed, Marlaine E., Stephen C. Vail and Bruce Fuller, "How textbooks affect achievement in developing countries: Evidence from Thailand", Educational Evaluation and Policy Analysis, (Winter, 1986). 8: 379-392. Lyeette, Margaret A. Improving basic educational opportunities for women in developing countries. Washington, D.C.: International Center for Research on Women, 1986. Mellow, Wesley, "Unionism and Wages: A Longitudinal Analysis," Review of Economics and Statistics, (February 1981): 43-52. Murnane, Richard J. "Comparisons of public and private schools: Lessons from the uproar," Journal of Human Resources (Spring 1985): 263-267. Murnane, Richard J., StuartNewst3ad and Randal Olsen. "Comparing Public and Private Schools: The Puzzling Role of Selectivity Bias," Journal of Business and Economic Statistics 3(1): 23-35, 1985. Price, Eleanol, and Robert Rosemier. "Some cognitive and affective outcomes of same-sex versus coeducational grouping in first grade," The Journal of Experimental Education (Summer, 1972) 40(4), 70-77. Psacharopoulos, George and William Loxley, Diversified Secondary Education and Development: Evidence From Colombia and Tanzania. Baltimore: Johns Hopkins University Press, 1985. Single-sex schools in Thailand ........... so ....... p)age 55 Riordan, Cornelius. *Public and Catholic schaolingt The effects of gender context policy," American Journal of Education (August, 1985), 518-540. Schneider, Frank W. and Larry M. Coutts. 'The high school environmentt A comparison of coeducational and single-sex schools," Journal of Educational Psycholoav (1982), 74(6): 898-906. Unesco. Statistical Yearbook. Parist Unesco, 1987. Wattanawaha, Nongnuch. A study of equitv in mathematics teachins and learning in lower secondary school in Thailand. Unpublished doctoral dissertation, University of Illinois, Urbana-Chanpaign, 19P.6. Willis, Robert J. and Sherwin Rosen. "Education and self-selection," Journal of Political Economy (October 1979 supplement): S7-S36. Woody, Thomas. A History of Women's Education in the United States. New York: Science Press, 1929. World Bank. World Development Report. Washington, DC: World Bank, 1986. Single-sex schools in Thailand .......... ............ .. page 56 Footnot-s 1. Modern statistical techniques help in controlling for this bias, altlough recent research has revealed that it is also important to keep track of one's assumptions in modelling (see Murnane, Newstead and Olsen, 1985, for a careful assessment of the results of Coleman et al. and their critics). Moreover, it is not possible to measure all relevant characteristics; apparent differences by school type, therefore, could be due to some unmeasured aspect of the student's background, ability and/or motivation. Several studies have attempted to use direct measures of ability through the use of tests specifically designed to measure innate ability (e.g., an I.Q. test) rather than *cognitive achievement (Psacharopoulos and Loxley, 1985; Boissiere, Knight and Sabot, 1984 among others). Many analysts have questioned the validity of these tests in distinguishing between ability and achievement. In any case, no one has ever suggested that such tests fully control for both ability and motivation. 2. This section draws heavily from "Thailand" in the International Handbook of Educational Systems (Cowen & McLean, 1982), pp.515-555. 3. Since the correlations between paternal and maternal occupational status (r - .39) and paternal and maternal educational attainment (r - .58) were high, we analyzed the effects of paternal occupational status and maternal educational attainment only. There were also fewer missing cases for these variables. 4. Alternatively, equation (1) can be estimated as one equation, with a dummy variable for single-sex and coeducational types of schools. However, statistical (F-) tests lead us to reject the Single-sex schools in Thailand ......... page 57 hypothesis that the coefficients of all the other variables are equivalent in both types of schools. Results are available from the authors. 5. The presence of I as a component of e is the critical factor in this problem. If there were no unmeasurable influences on achievement, or if unmeasurable effects were uncorrelated with school type, as well as other components of Z and X, there would be no selection bias. Present samples would be random draws from the population. Of zourse, costly strict experimental designs would also obviate this problem -- i.e., if students were assigned randomly to coeducational and single-sex schools. 6. This unconditional effect nets out the selection term from both the coeducational and single-sex equations in calculating school effects. In contrast, the conditional single-sex school effect would measure the following: from the sample of students who have already selected coeducational school, the increment (or decrement) in test scores had that student gone instead to a coeducational school. This conditional effect leaves in the selection term in calculating school effects. Single-sex schools in Thailand ............ page 58 PPR Working Paper Seriles Title Author Date Contact WPS9 Can We Return to Rapid Growth? Andrea i3oltho June 1988 J. Israel 31285 WPSIO Optimal Export Taxes for Exporters of Perennial Crops Mudassar lmran June 1988 A. Kitson-Walters Ron Duncan 33712 WPSII The Selection and Use of Pesticides in Bank Financed Public Health Projects Norman Gratz June 1988 C. Knorr Bernhard Liese 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 Ian Timaeus 33612 WPS14 The Optimal Currency Composition of External Debt Stijn 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 33792 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 PPR Working Paper Series Title Author Date Contact 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 Patrick Guillaumont 33706 Sylviane Guillaumont WPS23 Cost-Effective Integration of Immunization and Basic Health Services In Developing Countries: The Problem of Joint Costs A. Mead Over, Jr. July 1988 N. Jose 33688 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 July 1988 W. Ketema 33651 WPS26 Changing Patterns in Vocational Education 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. Campbel 33769 WPS29 The Relative Effectiveness of Single-Sex and Coeducational Schools In Thailand Emmanuel Jimenez July 1988 T. Hawkins Marlaine E. Lockheed 33678 WPS30 The Adding Up Problem Bela Balassa July 1988 N. Campbell 33769 WPS3; Public Finance and Economic Development Bela Balassa July 1988 N. Campbell 33769