THE WORLD BANK Discussion Paper EDUCATION AND TRAINING SERIES Report No. EDT82 Social Selection in Education and Economic Development C. Arnold Anderson May 1987 Education and Training Department Operations Policy Staff The views presented here are those of the author(s), and they should not be interpreted as reflecting those of the World Bank. Discussion Paper Education and Training Series Report No. EDT82 SOCIAL SELECTION IN EDUCATION AND ECONOMIC DEVELOPMENT C. Arnold Ander3on Education Research Division Education and Training Department May 1987 The World Bank dbes not accept responsibility for the views expressed herein, which are those of the author(s) and should not be attributed to the World Bank or its affiliated organizations. The findings, interpretations, and conclusions are the results of research or analysis supported by the Bank; they do not necessarily represent official policy of the Bank. Copyright t 1987 The International Bank for Reconstruction and Development/ The World Bank ABSTRACT This is a quantitative, cross-sectional analysis of the level of equity among students in secondary and higher education in countries at all levels of development. Equity is assessed in terms of access to schooling by parental education, occupation and family income. The survey finds that there is no deterministic relationship between the occupational structure or economic level of developing countries and social selectivity for schools. The author concludes that increasing enrollments overall would be the most effective way to achieve equity. TARLE OF CONTENTS CHAPTER 1. INTRODUCTION A. Education, equity, and development B. Previous cross-national surveys on social selectivity of schooling C. Data sources and quality D. Methodology E. The multiple facets of selection processes F. The structure of this report CHAPTER 2. FATHER's EDUCATION AND SELECTIVITY OF.ENROLLMENTS in secondary schools and universities A. Profiles of education of fathers and of base populations B. An overall view of degrees of selectivity C. Who is over- and who is under-represented? D. Enrollment rates and educational selectivity E. The economy and student selectivity by parental schooling CHAPTER 3. FATHER'S OCCUPATION AND SELECTIVITY OF ENROLLMENTS IN SECONDARY SCHOOLS AND UNIVERSITIES A. The data and occupational classifications B. Profiles of occupations of fathers and of base populations C. An overview of degree of selectivity D. Who is under- and who is over-represented? E. Selectivity of farmers' children and the place of agriculture in the economy F. Selectivity and per capita income CHAPTER 4. COMPARING SELECTIVITY ON FATHER'S EDUCATION AND FATHER'S OCCUPATION CHAPTER 5. DIFFERENTIAL ENROLLMENT RATES A. Relationship between selectivity indexes and background-specific enrollment rates B. Ratio and arithmetic measures of the magnitude of contrasts among background-specific enrollment rates C. Measuring differences in a developmental perspective D. Per capita income and selectivity in enrollment rates TABLE OF CONTENTS (Cont'd) CHAPTER 6. LONGITUDINAL PERSPECTIVES ON SELECTIVITY A. Cbanges over time in degrees of selectivity B. Background-specific enrollment rates and changes in educational opportunities C. Actual changes in background-specific enrollment rates relative to changes required to reach specific targets CHAPTER T. SELECTIVITY BY TYPE OF SCHOOL AND CURRICULUM A. Social heterogeneity and homogeneity in the classroom B. Selectivity by type of secondary curriculum C. Public, private, and religious schools D. Differentiation within higher education CHAPTER 8. INCOME SELECTIVITY OF EDUCATIONAL PARTICIPATION CHAPTER 9. CONCLUSIONS Appendix A. Concerning the sample for paternal schooling and occupation Appendix B. Specifying the base population Appendix C. Constraints on selectivty indexes Appendix D. Uses and limitations of indexes of dissimilarity Appendix E. Probit and logit functions Appendix F. Selectivity into elementary schooling Appendix G. Tables on mother's education Appendix H. Reserves of talent, school marks, and educational selection in third-world countries Appendix J. Supplemental tables for Chapter 2 Appendix K. supplemental tables for Chapter 3 Appendix L. Supplemental tables for Chapter 5 Sources of data by countries References LIST OF TABLES 1.1 Indexes of status of university students in four countries 1.2 Percentage distributions of schooling of mothers related to schooling of fath'rs; secondary pupils in Senegal, 1967 1.3 Selectivity indexes of Ghanaian Secondary boys by paternal occupation and schooling, with associated percentage dis- tributions of selectivity indexes among tribes 1.4 Students in higher education by size of family and occupation of father: Hungary, 1930 1.5 Percentage distribution of paternal occupation by middle- school grade from which pupils transferred to secondary school: Ghana 2.1 Frequency distributions of percentages of fathers of designated amounts of schooling by world regions; secondary and university students 2.2 Frequency distributions of indexes of dissimilarity between education of fathers and of male base populations by world regions; secoandary and university students 2.3 Frequency distributions of selectivity indexes by fathers' schooling; ssecondary and university students 2.4 Analysis of selectivity indexes and constraints on them for fathers with higher education; secondary students 2.5 Analysis of selectivity indexes and constraints on them for fathers with at least some secondary education; secondary students 2.6 Analysis of selectivity indexes and constraints on them for fathers with higher education; university students 2.7 Analysis of representation of the relatively uneducated among fathers of secondary students 2.8 Effects of per capita income and enrollment rates on indexes of dissimilarity by father's education 2.9 Swimmary of effects of per capita income on proportions of fathers with at least some secondary education 2.10 Effects of per capita income, enrollment rates, and base populations on proportions of fathers with selected educational attainments 2.11 Effects of per capita income and enrollment rate on selec- tivity by father's education 3.1 Frequency distributions of occupations of fathers; secondary students 3.2 Frequency distributions of occupations of fathers; university students 3.3 Frequency distributions of indexes of dissimilarity between occupations of fathers and of base populations 3.4 Frequency distributions of selectivity indexes by father's occupation -iv- LIST OF TABLES (Cont'd) 3.5 Selectivity indexes and constraints on them for fathers white- collar (excluding proprietors and traders); secondary students 3.6 Selectivity indexes and constraints on them for fathers white- collar and proprietors-and-traders; secondary students 3.7 Selectivity indexes and base-population constraints; white- collar fathers of university students 3.8 Effects of per capita income and enrollment rates on indexes of dissimilarity by father's occupation 3.9 Effects of per capita income, enrollment rates, and base popu- lations on proportions of fathers white-collar, manual, and farm 3.10 Effects of per capita income and enrollment rate on selectivity by father's occupation 4.1 Backgrounds of pupils aged 17-18 and age 13; IEA countries, 1963 4.2 Effects of per capita income, enrollment rates, and income dis- tribution on indexes of dissimilarity by paternal education and occupation; secondary and university students 5.1 Comparisons of background-specific enrollment rates by father's education; secondary students 5.2 Comparisons of background-specific enrollment rates by father's occupation; secondary students 5.3 Comparisons of background-specific enrollment rates by father's occupation; universty students 6.1 Occupational distributions of fathers, selectivity indexes, and indexes of dissimilarity at earlier and later dates 6.2 Comparison among alternative measures of changes in inequality of occupational representation; secondary and university students 6.3 Distributions and selectivity by father's schooling at three student levels; four male cohorts in the United States 6.4 Assessment of progress toward parity of unviersity enrollment rates by social status; Sweden, 1930-70 6.5 Assessment of progress toward parity of rates of college entry by paternal occupation in Japan 1953-1968 and for four United -States male cohorts; progress toward parity of college graduation for four United States male cohorts 6.6 Assessment of progress toward parity of secondary enrollment rates by father's occupation; Hungary 1931-1963 and United States male birth cohorts I (1898-1907) to IV (1928-1937) 7.1 Percentage distribution of paternal status by repute of secondary school; Ghana 7.2 Household characteristics of boys and of girls in elite and in all secondary schools in Kenya, 1968 LIST OF TABLES (Cont'd) 7.3 Distributions of secondary schools by proportion of fathers and of mothers who had completed secondary school; Buenos Aires 7.4 Distributions of secondary schools by percentage of students' fathers in designated occupations; Sao Paulo 7.5 Distributions of education of parents by type of school or course; secondary students 7.6 Distributions of occupations of fathers by type of school or course; secondary students 7.7 Distributions of education of parents of secondary students in public, private, and religious schools 7.8 Distributions of occupations of fathers of secondary students in public, private, and religious schools 7.9 Occupations of fathers of students in various types of schools; Serbia, 1885 7.10 Distributions of education of parents by type of institution; university students 7.11 Distributions of father's occupation by type of institution; university students 8.1 Income selectivity to higher education in Japan, 1968; class- ification by absolute income levels 8.2 Income selectivity into Japanese universities in 1961 and 1976; classification by income quintiles 8.3 Income and educational selectivity in Colombia 8.4 Income and educational selectivity in Malaysia A.1 Distributions of percentages of adult males with des,ignated schooling by world regions: all countries and sampled countries B.1 Estimates of selectirity index using alternative estimates of base population; Canada, 1975 B.2 Selectivity to higher education by education of father: 18-21 year-olds n Santiago, Chile B.3 Income selectivity to higher education in Santiago B.4 Effects of type of base-population comparison on socio- economic selectivity i-ndicatore: French university students, 1974 C.1 Possible range in percentage of students enrolled in category j who are from background i C.2 Possible range in selectivity ratios (S for students enrolled in j by base population category i D.1 Illustration of indexes of dissiLnilarity and selectivity: Turkish secoridary students -vi- LIST OF TABLES (Cont'd) F.1 Selection into elementary school by parent's education F.2 Selection into elementary school by father's occupation G.1 Distributions of education of mothers; secondary students G.2 Distributions of education of female base populations; secondary students G.3 Selectivity indexes and indexes of dissimilarity; mothers of secondary students G.4 Distributions of education of mothers and indexes of dissimilarity; university students J.1 Indexes of dissimilarity between distributions of education of fathers of secondary and university students and of base populations J.2 Percentage distributions of education of male base population; secondary students J.3 Percentage diRtributions of education of male base popula- tion; university students J.4 Pertentage distributions of education of fathers of secondary students J.5 Percentage distributions of education of fathers of university students J.6 Selectivity indexes by father's education; secondary students J.7 Selectivity indexes by father's education; university students K.1 Indexes of dissimilarity between distributions of father's occupation and of occupations in base populations; secondary students K.2 Indexes of dissimilarity between distributions of father's occupation and of occupations in base populations; university students K.3 Percentage distributions of occupations of base populations; secondary students K.4 Percentage distributions of occupations of base populations; university students K.5 Percentage distributions of occupations of fathers of secondary students K.6 Percentage d.istributions of occupations of fathers of university students K.7 Selectivity indexes by father's occupation; secondary students K.8 Selectivity indexes by '.ather's occupation; university students -vii- LIST OF TABLES (Cont'd,c L.1 Estimated background-specific rates of enrollment in secondary school per hundred fathers by lather's education L.2 Estimated background-specific rates of enrollment in secondary school per hundred fathers by father's occupation L.3 Ratios of background-specific rates of enrollment in university by father's education L.4 Estimated background-specific rates of enrollment in university per hundred fathers by father's occupation -viii- CHARTS Chart 2.1 Selected comparisons of fathers' education with education of base populations; secondary students 2.2 Selected comparisons of fathers' education with education of base populations; university students 2.3 Selectivity indexes by base populatioDs with secoDdary education or more; secondary students 2.4 Selectivity indexes by base populations with secondary education or more; university students 2.5 Selectivity indexes by percentage of male base population with higher education: secondary students 3.1 Occupational distributions of fathers of secondary students (triangle grid) 3.2 Occupational distributions of athers of university students (triangle grid) 3.3 Percentages of fathers by percentages of base population; secondary students A. White collar B. Proprietors and traders 3.4 Percentages of fathers by percentages of base population; uDiversity students A. White collar B. Proprietors and traders 3.5 Percentages of fathers by percentages of base population; secondary students C. Manual D. Farming 3.6 Percentages of fathers by percentages of base population; university students C. Manual D. Farming 3.7 Selectivity indexes by base population white collar (excluding proprietors and traders); secondary students 3.8 Selectivity indexes by base population white collar (excluding proprietors and traders); university students 3.9 Selectivity indexes by base population farmer; university students 3.10 Index of dissimilarity between occupations of fathers and of base populations by per capita income; A. Secondary students B. University students 4.1 Ratios of (S -1) to (MaxS -1); fathers with at least some secon ary education ty white-collar fathers (excluding proprietors and traders 4.2 Inexes of dissimilarity; university by secondary studerits A. Comparisons by father's education B. Comparisons by father's occupation 4.3 Education and Occupation indexes of dissimilarity com- pared; secondary students 5.1 Comparisons of enrollment rates in secondary school of youth whose fathers had at least some secondary schooling and of those whose fathers lacked such schooling; ratios of rates by arithmetic differences 5.2 Logit differences in secondary enrollment rates of children of men with at least some secondary education and children of men with no secondary education, by per capita income 5.3 Logit differences in secondary enrollment rates of children of white-collar men (including proprietors and traders) and of farmers, by per capita income 5.4 Logit differences in university enrollment rates of children of white-collar men (including proprietors and traders).and of farmers, by per capita income 6.1 Ratios of white-collar to manual and to farm enrollment rates: later by earlier dates A. White-collar/manual ratios of rates B. White-collar/farm ratios of rates D.1 Illustration of a Lorenz approach to the index of dissimilarity FOREWORD One common problem of higher education in the developing countries is that cn unequal percentage of places is filled by privileged sections of the population. Some observers regard this problem of biased participation to be so serious in the developing countries that in particular cases they question whether public investments should be diverted to purposes other than higher education, whether within the educational system or elsewhere. The World Bank commenced educational lending in 1963. As with Health, Rural Development, Water Supply, and other sectors, the operational staff in education are in constant dialogue with their professional colleagues in the developing countries. The purpose has been not only to ensure that investments, in higher education for example, are productive to ,he national economy, but that the participants (the direct beneficiaries) broadly represent the make-up of the national population. This paper was supported by the Education Department in hopes of improving that dialogue. A substantial amount of research had been conducted on the exi-tence of unequal participation in secondary and in higher education; this paper investigates whether there nre systematic patterns to this problem. Does the degree of unequal participation change with the level of national economic development? Does it differ substantially across regions? Is there evidence of changes in the degree of unequal participation over time, and if so in which countries or situations? To what extent does unequal participation in higher education extend to (or reflect) unequal participation in secondary education? Some of the findings with regard to these questions might be summarized as follows: First, the measured equality of participation in secondary and in higher education is heavily influenced by how equality is specified. Results can be quite different when representation is expressed in relation to father's or mother's education or to father's or mother's occupation. Results can be quite different too when the appropriate age group specification has been made in the adult population used for comparison with parents of secondary and university students. Second, previous interpretations of statistical measures have often been incorrect. Since sharp contrasts exist both in the proportion covered in primary and secondary education across countries and in the geographical coverage of primary and secondary education within countgies it is difficult to make comparisons using standard measures of inequality in higher education. Third, no country has realized full equality of socio-economic status participation in higher education or even in upper secondary education. Some countries have come closer than have others at comparable levels of per capita income. However, after allowing for effects stemming from variations in primary and secondary school enrollments, the differences observed in the equality of participation in higher education between the less and the more developed countries are minor. -xi- Fourth, while policies to provide greater equality in educational opportunity deserve strong support, such policies can be costly. Moreover, no policies have so far been found to have resulted in a perfect equality in higher education participation. In sum, policies to promote greater equality of participation in higher education should not anticipate the attainment by a developing country of what has been successfully attained by no country. These findings reflect the author's investigations, not World Bank policy. Nevertheless, we think them to be of sufficient importance to warrant publication and open discussion, which is the purpose of this Staff Working Paper. Aklilu Habte Director Education Department -xii- PREFACE This report was commissioned by the Education Department of the World Bank, at the initiative of Dr. Stephen Heyneman. Many aspects of social selection might have been considered in addition to those dealt with: parental schooling, paternal occupation, family income. One could examine also ethnicity, religion, region or province of residence, and others. But the three traits of secondary and university students that were considered proved to be sufficiently revealing of the diversity of patterns of selectivity around the world. The patterns of relationship between selection for school and coun- tries' levels of development are -- as anticipated - loose-textured. The influence of local traditions, institutional structures, and educational policy would seem to have large scope. Equal in importance to the substantive results are some methodolog- ical considerations that have been little discussed hitherto in the litera- ture on selectivity. These considerations point to the need for careful use of conventional indexes of selectivity in opportunity for schooling; one must specify the mathematical constraints on indexes. Appendices have been used to discuss the derivation of statistical techniques w4'U hout interrupting the flow of the exposition. In addition, a few topics that are important supplements to the main focus of the report are discussed in appendices, and more detailed data in appendix tables supplement the tables included in this text. Regression analysis has been used sparingly, primarily LO illuminate the interactive effects of enrollment rates and of per capita incomes upon selectivity. -xiii- Assembling data and analyzing them systematically has proven to be a much more laborious task than had been anticipated. Just because the relation- ships are not clear cut it was important to display the particularities of signations of different countries using several indicators and with tables and charts that reveal the diversities. Data by counries may challenge some readers to draw on their special knowledge to go behond the strictly statis- tical findings in more summary form. Meanwhile, it is believed that this report advances the study of selectivity for education to a new stage. The opportunity to carry out this investigation has been greatly appreciated. Professor Mary Jean Bowman has contributed her mathematical capabi- lities at many points. June ,, 1982 C. Arnold Anderson -xv- EXECUTIVE SUMMARY This is a quantitative, cross-national survey of the level of equity in utilization of secondary and higher education. Equity is assessed in terms of three aspects of family background: parental schooling, paternal occupation, and family income. This is the first comprehensive examination of this topic, and it is the first survey that looks at this problem with respect to countries at all levels of development. The strategy of analysis embodied in this report enables one to quantify the effects of different policies with more appro- priate indicators than usually have been employed. Those techniques canndt resolve any policy issues but they can give a definiteness to the, empirical evidence which will clarify options for policy. Some widely held beliefs about selectivity in access to education have been shown to be unfounded. Several generalizations are supported. Many tables and analytical charts have been used to emphasize the particularities of relationships in different countries. Selectivity to schooling varies greatly in degree among countries within each world region and at each level of economic development; The degree of selectivity is only loosely associated with either national per capital income or the proportion of the labor force engaged in agricultlre. It was demonstrated how quantitative indicators used to delineate the diversity in social selection can be adapted to the methodological task, and innovations have been made in statistical techniques. It is essential to interpret indicators of selecti'vity in the light of the mathematical constraLnts on their possible values that arise from the distribution of occupacions and from the distribution of schooling among adults who are potential fathers of students (the "base population"). It is of central importance to ask, further- more, to what extent educational selectivity has been reduced by rises in enrollment rates or, instead, is not dependent on those rates. Nevertheless, after allowing for these structural constraints upon indicators of social selectivity countries still differ greatly in how strongly family background affects the use of schools. The upper possible limits of selectivity indexes are smaller where schooling is widespread among the generation of parents and where the occupa- tional structure has become more differentiated. Actual selectivity in nearly every society is less than the maximum selectivity possible given the occupa- tion structure, the distribution of schooling among the generation of parents, 'and the present rates of overall enrollment. There are large differences among societies, however, in the gaps between maximum possible and observed selectivity. Even taking enrollment rates into account in specifying the maxima, those variations reflect differ- ences in degree of stratification or egalitarianism beyond what is picked up in current overall enrollment rates. -xvi- Wherever democracization of schooling is a salient aim for a nation, there is scope for some complementarity betwen criteria of efficiency and equity. For many countries, however, a close approximation to proportional representation of students from the various status categories would not be a feasible immediate goal. In most countries, rapid approach to that goal would require a substantial expansion of expensive secondary and higher (and often even of elementary) schooling. The following pages contain a terse summary of this investigation, together with a suggestion for a next step in policy-oriented research. Coverage Usable information could be obtained for a fourth of the world's nations; the main shortfall was in Asia. In only a few cases were data available for both secondary and higher institutions in the same countries. Among the less-developed countries data are as likely to relate to parental schooling as to paternal occupation; in high-income countries data more often have been reported only for paternal occupation. Only scanty data are ia hand for elementary schools or for characteristics of mothers. Typically data are given for national aggregates, though disaggregation for sex of students is common. Selectivity in Relation to Economic Development There are no acceptable indexes of general political or social development of societies. For economic development use was made of per capita income and of the proportion of the labor force in agriculture. The indicators of social selectivity of schools were only loosely correlated with the economic measures. Selectivity typically was greater for higher than for secondary education, as one would anticipate; but selectivity indicators for secondary schools can be larger ia some countries than indicators for higher institutions elsewhere. Moreover, there was little association on overall indicators between occupational selectivity into secondary and into higher education. At each level of school selectivity varied widely, and on each school level countries at similar levels of economic development had disparate degrees or selectivity, Over-representation fom white-collar families was diverse even among the comparatively homogeneous "western" countries, and association of "democratization" with per capita income is quite uneven among that set of countries. Children of farmers tend to be better represented than do children of manual workers in secondary and higher schools of the high-income countries. In the third world, by contrast, Lt is the off-spring of manual workers who have the better chances of attending post-elementary schools--reflecting the early stages of urbanization and of modern-sector employment. It cannot be assumed that cross-section data are reliable evidence as to how educational selectivity will evolve with developmuenit, hut reliable evidence spanning dates far enough apart to identify trends could be found foc -xvii- only a handful of societies. Generally, in those cases, there was a decline over time in the excess representation for children of white-collar men relative to children of farmers or manual-workers--but in markedly varying degree from one country to another. There is some evidence that countries manifest a stable pattern for edu'cational selectivity: relative to maximum possible values, selectivity by father's schooling or occupation was not closely linked to changes in occupational structure. In countries on which a Gini index of inequality in income was available, the Gini index was correlated with educational selectivity overall as measured by the index of dissimilarity between distributions of occupation and of schooling of students' fathers and of the comparable male base popula- tions. Countries with greater inequality of income displayed more educational selectivity for either secondary or university students. A predominantly negative association between Gini indexes and per capita incomes (like the positive association of enrollment rates with incomes) accounts statistically for most of the negative association between per capita income and selectivity into secondary school but not into higher education. Logic of the Analysis Two main approaches were used in assessing selectivity of educational opportunity. The first approach compared the status distribution of students' fathers with that of the corresponding cohort in the adult population. The principal questions addressed were (1) what is the makeup of the pool of young people from which the country will draw in filling the major posts of responsi- bility in the future and (2) how closely do their family backgrounds matcti those of youth in general? On these points use was made of three indicators of selectivity: (1) the index of dissimilarity, which is an overall indicator, (2) conventional selectivity indexes by family background (and modifications of these), and (3) differences in the logits of father and base-population percentages for a particular category of status. The last of these, while less easily understood intuitively, is superior statistically and theoretically to the selectivity index and is used as the dependent variable in various regressions. The second approach focuses directly on relative chances of receiving comparatively advanced schooling for youth who were born into one or another category of family status. The best evidence on this question comes from estimates of background-specific enrollment rates. Since ratios of enrollment rates specific to any pair of backgrounds equal the ratios of selectivity indexes on those backgrounds, the latter ratio could be used as an indicator of relative chances. The same relationship makes possible the estimation of background-specific enrollment rates whenever data both permit computation of selectivity indexes and provide overall enrollment rates. Given background- specific-enrollment rates, it is possible to go on to the superior indicator of disparities in enroJlment chances: differences in the logits of the enrollment rates. -xviii- Selectivity Measurement and Assessment of Representational Equity Sociologists have been using the "selectivity index" to measure opportunities for schooling, for mobility, etc. If the schools being compared are different segments of the overall system in one society or belong to similar societies, comparisons based on that index may not be misleading. But when one compares the degree of selectivity over long periods of time in a society or across diverse societies, that index has shortcomings., Only wtieri the constraints on possible values of the indexes are taken into account is it proper to draw inferences about equity. When one surveys schools across the world's gamut of societies, two features are central in the analysis. First, representativeness of students by parental occupation alters because the society's occupational structure changes. Second, schooling spreads among successive cohorts oE youth at different rates and this gradient varies among societies. These societal changes are embodied in shifting rates of enrollment in the aggregate and for youth from different sorts of families. This incer-generational dynamic is at the heart of the changing relationship between development and schooling. We will obtain only an unclear notion of what is happening if we use inappropriate indexes and if we misinterpret them. The two main constraints in this study were the base-population proportions of fathers having the designated statuLs (which is the denominator of the selectivity'index) and the aggregate race or enrollment in a given type or level of school for a particular age cohort. The over-representatCon in schools for children of advantaged groups catches our attention beca'use the selectivity indexes are dramatically large. But it is apparent that the proportion of the base population in any given occupation or educational-attainment category determines the maximum possible values of the corresponding selectivity index, and the lower that proportion the higher the maximum index. The greater the gap between the observed index and the maximum possible index given the base-population structures, the greater must have been the influence of democratizing forces upon the schooling of youth. One can identify societal features that work to bring selectivity for children of privileged families down from the maximum possible values uander Siven base-population constraints. Rising enrollment rates, while only one factor, can be crucial. Constraints arising from the overall or aggregate enrollment rates may set upper limits on selectivity that are lower than the limits set by base-population distributions, especially for secondary schuols. One can then identify the upper limits of the selectivity indicatocs wh=Li taking into account both enrollment rates and base-population distributions. There are analogous constraints on how low indexes might be for the status categories that normally are under-represented in school, but these conistraints normally are weak. Indeed, those low indexes warrant special attention. One innovative feature of this study has been to compare indicators of the degree of selectivity on paternal schooling with that on paternal occupation. For the countries on which data for both aspects of paternal status were available (mainly low- and middle-income countries) tdaera was a -xix- clear positive association between seleccivity to secondary school by father's occupation and by father's educacion. Universal enrollment perforce would mean enrollment in proportion to status background. When (as is usual), overall enrollments are low or moderate, students from particular backgrounds may or may not approach pro- portionate representation. When rates of enrollment are high, inevitably representation will approach proportionately. A gap between the maximum possible selectivity-taking into account both base-population and enrollment constraints-reflects one or anocher equalizing irnfluence: namely, the institutionalized "democratic" or "elitist" traditions of education, pervasive political structures, the ways that schooling is perceived within various segments of the society, and an assorcment of particular educational practices. If academic selectivity is strict (in respect to examinations, promotion, etc.), one will not expect to observe a strong tide toward widening opportuni- ties to remain in school. Meritocratic selection can winnow out inept children from favored homes as well as identify "high flyers" from humble homes. Modification of pedagogic practices can have fewer equalizing effects than does expanding enrollments, but the latter policy unaccompanied by efforts to treat fairly those children from non-elite homes who have actually entered the schools can be fiscally overwhelming. The Index of Dissimilarity As a Summary Indicator This is perhaps the only summary indicator of representativeness of youth from different status backgrounds that is appropriate for cross- national comparisons. (It- can, moreover, legitimately be compared with the Gini coefficient of inequality of income.) This index is robust when one compares the distribution of schooling among students' fathers with that of maen in the corresponding base population. The categories of schooling are monotonically ordered over the frequency distribution. The breaking points given in published data usually are put where under-representation is separated from over-representation. This index of dissimilarity is less suitable for comparison of socio-economic categories because occupational rubrics are internaily more heterogenous and they conceal the breakinv points respecting representation. Accordingly, use of this index for paternal occupations faces the same problems in comparisons between less- and more-developed countries that affect comparisons using selectivity indexes by father's occupation. The index of dissimilarity cannot exceed 100 minus the aggregate enrollment rate in schools. Some readers might view this as not a serious defect since enrollment rates are an instrumental variable for policy. But rates of enrollment usually cannot be raised much in the short term, and ic is politically difficult to lower enrollments from hitherto educogenic families. A positive correlation between the Gini coefficient of income inequality and the index of dissimilarity for paternal occupation is almost certainly in part a reflection of differences among the high-income societies in degrees of social stratification in other respects. -xx- Background-specific Enrollment Rates in a Development Perspective The sequences of changing utilization of schools may or may not move synchronously as among subpopulations, and the relative pace of diffusion of utilization across and within subpopulations (or status categories) can liffer considerably among societies. - Several measures of inequality in background-specific enrolmuent rates were compared: arithmetic differences in rates, ratios of rates, dald differences in their logits. The arithmetic diff'.rences are a poor indictctor, highly sensitive to overall levels of enrollment,, Ratios of rates have analogous limitations but in the opposite direction; they have the advancage that they can be computed with no information beyond ratios of selectivicy indexes, and logarithms of those ratios reduce their biases and bring them closer to measurement of inequalities by differences in logits. As a linear transform of a cumulative S curve, the expression of enrollment rates in logit form is inherently appropriate to the study of inequalities in the spread of educational participation. Differences in logits of enrollment rates for youth from white-collar and farm families, for example, are insensitive to the proportions of these families in the base population; they are insensitive also to aggregate rates of enrollment. These advantages apply to comparisons among societies and also over intervals of time in a gLven country. As has been mentioned, there are many policies and pedagogic practices that operate to widen opportunities for schooling among the families of a society. Each change of practices (even if not an explicit ouccome of policy) has pedagogical and fiscal implications. Enrollment changes are the principal manipulable factor designed to affect equity. But substantial changes of enrollment can have awesome fiscal implications. It is prudent, therefore, to consider the balance between the equity and the efficiency implications of alternative policies. It follows, indeed, that consideration of the fiscal consequences flowing from changes in enrollment may well persuade policy makers that removal of inequity may have to proceed slowly. Subpopulation Differences in Selectivity The overall degree of social selectivity may be quite high in the schools of a society, yet differentials in selectivity between the sexes can be moderate; Latin America exemplifies that pattern. Sex differentials are quite large in many African and most Muslim societies. Every society displays a multiplicity of differentials (of varying size) in the utilization of schools. Ethnic differentials may be small while regional constrasts remain large; differentials for social status are likely to be substantial generally. Virtually everywhere, girls come from higher social-status backgrounds than do boys except when attendance approaches universality for a given level of school. Whatever groups are under- represented, shortfall for girls will be large in those groups.' Daughters or men at the upper status levels are likely to be correspondingLy over represented and the location of this "tipping point" varies by society. -xxi- Constrasts in Selectivity within Levels of School As enrollment rises toward universality, selectivity within secondary education may persist in the form of streaming or by favoring children from upper-status families for entry to prestigious schools. These differentials-perhaps more for ethnic group or religion than for social status-can occur early or late in the establishment of a society's system of schools. Enrollees in vocational courses in secondary schools, or in pedagogy and medicine in university, come in different proportions from various strata of families. Different universities may contribute to distinct aspects of national life; notice the dissimilar backgrounds of students at Cairo and at Al Azhar universities in Egypt. Variations in the social composition of students is a major feature of any educational system, and those assortments of students' backgrounds color the social life of individual classrooms and institutions. In the present study only cursory attention was given to selectivity for different types of school and even less to contrasting entrants to various courses of study. There simply was not sufficient time to explore many fascinating byways. No doubt those other differentials often biased the data that could be used. Longitudinal Perspectives on Selectivity In the countries for which data were available for secondary schools at two dates, convergence over time toward parity of access was marked, con- firming the picture obtained by comparing countries at lower and higher levels of income. In each case a substantial rise in overall enrollment at the secondary level had played a large part in democratization. Meanwhile, struc- tures of occupations and of adult distributions of schooling were changing. Among university students also (in the handful of available cases), disparity of access diminished, although societies were quite diverse in the extent of these transformations. Even when selectivity indexes were adjusted for base-population constraints, over-representation of children from white- collar families diminished, although in the adjusted indexes changes were seldom large. Some changes were dramatic however: notably Puerto Rico and Greece. One is not surprised that these trends toward equalization are less impressive for higher than for secondary schools. One can hardly avoid the inference that each country (or associated set of countries) manifests a distinctive ethos respecting. democratization of schooling. The general (though often small) declines in over-representation from white-collar homes among university students was facilitated by rising enrollments, but the constraint of aggregate enrollments on the upper limits of the selectivity indicators was not of major importance for higher educatioa. Such statistical effects of higher enrollmeni:s on selectivity as were found for university were unquestionably in part a reflection of deliberate efforts to extend access, which led directly to increased enrollment from previously disadvantaged categories of youth. -xxii- From a political standpoint, to reiterate, usually it is easier to alter the aggregate enrollmnent rate than to diminish "excess" enrollment from advantaged fdmilies. In due course overall expansion erodes the over- representation from certain sectors of the population. One gets a more complete picture of how different factors affect particular differentials in utilization of education when one follows the procedure of analysis exemplified in this report: use of appropriate indicators of selectivity, use of background- specific enrollment rates if possible, and transformation of rates into logits. That strategy of analysis yields clear perspectives on the progress toward parity that has occurred and on the effort that must lie ahead to attain stipulated targets of equity. It also lays a basis for consi4ering- whether resources should be put into arousing aspirations in dormaat groups as against capitalizing on known differences in readiness to uttlize places in schools as soon as they are opened. Data necessary for this analysis are nLot available in most of the third-world countries but since the mode of analysis is particularly fruitful for those countries, this report can be used to stimulate improvement of data in many countries. Implications for Next Moves in Policy-oriented Research There is no deterministic relationship between the occupational structure or economic level of developing countries and the pattern or degree of social selectivity for schools. This conclulsion stands out unambigu- ously from these explorations. Subject to considerations of cust and of political acceptability, virtually every country could alter its present pattern of educational opportunity in a more democratic direction. Without substantial increases in rates of enrollment overall, however, a rapid move toward parity of chances among all youth would require restrictions on enroll- ment by children of the families that have been enjoying the better opportun- ities. The equity implications of such a policy would not be all in one direction, and the sacrifice of efficiency could be large. - Nevertheless, there are prospects of impressive gains in both equity and efficiency by focusing on the reduction of regional differentials in availability of schooling, including reduction of urban-rural differentials. In most countries, and especially in less-developed countries in which communi- cation gradients are steep, regional differentials in.enrollment rates are associated with differentials between ethnic and religious groups. In turn, those differentials are linked to imparities among social-status categories. Lower-status individuals in the more developed areas of a society often enjoy better opportunities for schooling than do upper-status inidividuals in back- ward areas. Except where boarding schools are used, this handicap can be large even for children residing only ten kilometers or less from a town. Diffusion of schooling is a continual building-up and satisfying of aspirations for schooling. Though aspirations may not be tightly-linked to social status, children in any population who first take advantage of new opportunities typically will come disproporttonately from families who already enjoy other advantages. Whatever the effects of equalization of opportunities for schooling on economic development in the short-run, -xxiii- positive long-run effects of increasLai- enrollments and of wider partici- pation in the ongoing economic life of a developing country are clear enough. Among the longer-run effects of major inequalities of represeation acrob!; regwons and tribes can be rigidificatton of emerging status structures and intensified internal political stresses. Indeed, democratization of opportunity is itself an aspect of "social," political, and even economic development. Chapter 1 INTRODUCTION A. Education, Equity, and Development This investigation focuses on a crucial relationship between education and sociietal development: as societies become "more developed," do opportunities for schooling come to be distributed more equitably? This is no broad review of how education is related to development, neither to social nor to economic development; it deals with only one important aspect of this rela- tionship. The "and" in the title of the report is deliberately unspecified. There is a minor theme in much of the literature about development-- perhaps a residue of old ideas about primitive communism-and espousal of this theme does not depend on the academic discipline of the writer. The notion is that low incomes and undifferentiated patterns of living are asso- aiated with comparatively uniform opportunities-prior to the emergence of "capitalism." (I specifically leave out any speculative connection to "the class system.") The deep appeal of the notion that "free" university educa- tion is equalitarian-when usually it constitutes a subsidy by the poor to the better off-manifests this outlook. In an inquiry twenty-five years ago, dealing mainly with western countries, I concluded that economic level of a country had little to do with equality of educational opportunity for university. Casual examination of data for new nations had led me to believe that the same inchoateness might prevail also in the third world. And even if a definite association between level of economic development and selectivity of participation in secondary -2- and higher education were to emerge, the question of causality would remain. An invitation from the World Bank Education Department to undertake this synthetic report led to an exhilirating intellectual adventure. Within the context of contemporary research about development I have had, perforce, to look at formal western-type schools. Both the scarcity of data and the vagueness of nonformal education categories precluded atten- tion to the latter. There is considerable evidence, however, that in large part these two aspects of learning and training are more complements than substitutes. No taxonomy of societies has been elaborated, although charac- teristics of the various societies included are reflected in the characteris- tics of the base populations with which the fathers of students are compared. Associations between selectivity measures and per capita income, and between those measures and proportions of the total labor force engaged in agricul- ture, were explored, with essentially negative results. The important exception is the indirect effect of per capita income oa enrollment rates and through them on the mathematically constrained limits of possible selectivity of children of highly educated parents and of rrents from high-level occupa- tions. Fortuunately the third-world countries for which we have data are sufficiently diverse to warrant confidence in the findings. Data for the advanced countries also are more numerous and more diverse than in any previous investigation, and effects of institutional or historical factors upon patterns of educational selectivity emerge clearly. These variations strengthen my earlier assessment--that there are no simple associations between level of economic development or societal type and the distribution of educational opportunity. In examining development it is a temptation to interpret cross-country contrasts as indicators of changes over time; that is an unwarranted assumption. -3- Equally dubious is the assumption that contrasts between earlier generations and today in the now-developed societies can be interpreted as heralding what direction events will take in the newly-developing societies. I have sought to resist those temptations except where one could link educational selec- tivity to changing levels of base-population status and to rising rates of enrollment. B. Previous Cross-national Surveys on Social Selectivity of Schooling As quantitative data accumulate about an important topic, suimmaries and assessments of that information on a world scale begin to appear. Concern as to whether development programs by international 2,gencies have led to a widening of income differentials within the less-developed societies has led to world surveys of income distribution. There are similar surveys about changing patterns of labor force distribution, especially among women. Other writers have compared dev-lopment policies in civil and military regimes, and reviews of the incidence of civil violence in the new politics are numerous. There have been previous, quite limited, overview reports on the present topic. In the 1950 Year Book of Education social selectivity in higher education was compared over a number of countries. Unfortunately that compilation was made by different authors and there was little alignment of occupational rubrics among the countries. This 1950 report was the starting point of my own survey in 1956 for university students; I used a larger set of countries but had little data from the third world. The problem of noncomparable rubrics was solved by a compromise; in one set of countries farm laborers were included with urban manual labor and in others with farm operators. That report emphasized the diverse balance of privilege between children of professional men and those from entrepreneurs' families, and it -4- underlined evidence for "a new hereditary aristocracy" of civil servants. The level of national income was unrelated to the degree of social selectivity for higher education. Though women students are on average from better family backgrounds than are men, the greatly varying proportion of women among students was not reflected in the overall selectivity of attendance for different countries. To a modest degree, where rates of attendance were higher social selectivity was lower--an anticipation of the strong influence of enrollment rates found among the more heterogeneous countries reported on here. Several countries were surveyed by OECD (in a study directed by Charles Nam) a decade later, and this study recently was updated. The authors of those reports strove to cope with the incommensurable rubrics of occupations, but this problem was not solved. OECD found that the member countries were extremely diverse in degree of social selectivity. A few years later the ministers of higher education in Europe (sponsored by Unesco) widened the scope of the OECD inquiry. Individual authors have made careful reviews of the situation among Eastern European "socialist" countries and the USSR. (The reports on Hochschulstatistik for prewar Germany were models of scope and of detail.) There were several historical surveys of varying breadth for the states of nineteenth century Germany; these and more recent data (also for France) are elaborately analyzed by Fritz Ringer. John Craig has returned to the older German data; by disaggregating information by German state and by detailed paternal socio-economic category, he for the first time shows trends related to the populations at risk. One can obtain data for each institution in the United Kingdom, but those data cannot be fitted into a cross-national compilation; elaborate institutional information is needed to codify and interpret them. It has -5- become routine in the United States to report miscellaneous pieces of information about family status for students in hundreds of colleges, along with much supplementary information. While these data demonstrate the bewildering diversity of our system of higher education, they cannot be fitted into the present sort of overview. Just recently David Angus (1980) has been collating information about status of students in high schools of the United States before 1940, including studies well back into the nneteenth century; no effort has been made to rework his findings for inclusion here. None of the foregoing surveys looked at the world of the "developing countries." There is, not surprisingly, no set of data on a common basis for a representative set of countries. However, several factors have led to the multiplication of compilations about students' parental status during the last quarter century. The inauguration of UNESCO and the expansion of its statistical activities have been a stimulus, especially for new nations-- though only to a slight extent with reference to the social characteristics of pupils. The vogue for educational planning and the spread of official affirmations about equality of educational opportunity have stimulated investigations of pupil characteristics in secondary schools and in institu- tions of higher education. These and parallel influences have been augmented by the strong development of the field of sociology of education, a field that has been brought to bear on educational problems in advanced and in less-developed countries alike. Older sets of data, produced by innovative officals or statisti- cians, sometimes set a standard for quality that seldom is met today in the third world. A remarkable example is the survey of pupils in all levels of schools for the Kingdom of Serbia in 1885. Historical series for Germany have been more utilized than those in the Swedish archives. Alumni directories -6- for Oxford and Cambridge continue to yield new perspectives (Schnaper and Anderson). This report does not include a systematic historical search, but references to these investigations will be included in Chapter 6, on changes through time in educational selectivity. The present report includes more countries than any previous survey; at a high cost in time for search, many third-world countries have been incorporated. Doubtless many sets of data were missed; within ministerial archives here and there additional material remains to be disinterred. 1/ Tentative generalizations for the enlarged spectrum of societies covered in this report have been made despite the difficulties of ensuring comparability and despite the non-representative nature of the sample of countries. Those generalizations are made cautiously, however. C. Data Sources and Quality 1. Locating materials. Because I had long been interested in social selectivity for schools and had published a number of articles on the topic, I began with a moderately good bibliography for a mixed set of colin- tries. Published bibliographies are of limited use, for the rubrics on this topic are too unspecific. Most of the literature is non-empirical and not quantitative much of it is tendentious and only superficially analytical. By combing the shelves at IIEP in Paris much material about third-world countries was obtained. The Economic Program in American University and the Education Department of the World Bank furnished some special ECIEL tabulations for Latin America. 1/ Appendix A discusses the sample of countries for which data on selec- tivity by parents' education are presented; see also the appended list of source-s-of data. -7- Many reports provide no information about base populations with which information on student backgrounds can'be compared; even when such base-population data are shown often they are inappropriate, giving distorted results. Because of this lack of 'base data it had been expected that much of the report could give only profiles of parental status or schooling. Recently UNESCO has prepared compilations of school attainments for adult populations. The unpublished tabulations give data for each sex separately (which the Yearbooks do not). The new compilations give for many countries schooling in the population by age and sex categories, thus providing a more suitable specification of the relevant base population. This is extremely important where there have been rapid changes in schooling over recent decades. Receiving these data only recently, laborious recomputations of selectivity measures were required, but the utility of the analysis was greatly enhanced. No such systematic compilation is at hand for occupational distributions, and the occupational rubrics used in the various studies of student backgrounds are diverse. Undoubtedly it would be possible to improve the analysis of selectivity by father's occupation by making use of some of the data brought together by the ILO or in national censuses--not easily accessible. That was too time consuming an undertaking for this report. 2. Data quality. Compared with research on the economics of development or on demographic correlates of development, this project must deal with unsatisfactory data. Useful demographic tabulations go back a couple of cen- turies; both the extent and the quality of information in this area improves steadily. To be sure, demographers always face a choice between broader geographic coverage and deeper coverage for each area; they seek continually to refine their data by multiplying the variables used. Statistics about -8- pupils normally cover only age, sex, and rural-urban residence. Though much information is collected about other characteristics of students, publication occurs seldom. Most of the data used in this report come from special studies (often available only in processed reports) made since the second world war. Continuity of data is rare, and tabulations for successive cohorts of youth usually are not available. Few reports utilize systematically representative (or random) samples of pupils or students or of schools. Withal, the results in this report are plausible. Fortunately, political or ideological considera- tions have not often distorted the statistics available; the topic is important to countries regardless of their political orientation. (That problem intrudes in some cases in the data by parental occupation, but not in those by parental schooling.) Thus three major deficiencies exist in such data as do exist: (1) sample biases, (2) misspecification of the "base population" with which to compare students' parents, and (3) non-comparable rubrics among countries for categories of schooling or of occupation. Wherever the coverage of students at a given level of school is incomplete, one has to be alert for biases in the sample; moreover, "secondary school students" is a heterogeneous categorization in terms of level as well as of types of schools. Studies vary in whether they include all secondary pupils, entrants, or those in the last year of school--which itself may be 10th or 12th year. If a country is small and post-compulsory schooling is not widespread, one may have full coverage of pupils but inadequate data for a base population. Moreove:, selectivity into higher education may be understated because many pursue their studies abroad. One may find data only for "academic secondary" pupils, who everywhere are more selected than others. Data on those in teacher training are sometimes available, but often with inadequate specifi- cation of the level of education involved. While it is possible to identify many of these variations, seldom has it been possible in this report to assess the quality of a sample for any given level or type of schooling. 3. Data requirements. Two sorts of information are crucial for inclusion of a society in this report: (1) the distribution of schooling (or of occupation or income) for fathers of pupils in secondary schools or of students in higher institutions; and (2) comparable information for the "base population," the appropriate cohort of fathers (or mothers) in the adult population. As mentioned earlier, information on parents may be informative even where base data are not available, but in such cases no assessment of degree of social selectivity is possible. It is highly desirable to know what proportion of youth are enrolled in the designated level of education. The best estimates of enrollment rates for most countries were derived from UNESCO compilations of schooling attained by age and sex categories of the population. In many.cases one can estimate a coumtry's enrollment rat-s for secondary (or higher) education by choosing the appropriate five-year category, taking into account the date of the study of students and the date of the census. This approach avoids most of the problems associated with over-age students in data on current enrollment rates by either age or level of schooling. Correct specification of the base population is crucial for any selectivity analysis; yet often we must accept approximate information. Here again the painstaking work by UNESCO wa.S extremely valuable. An alternative approach, to construct distributions of adult schooling by accumulating - 10 - enrollment data over a generation, is too laborious for other than demonstra- tion countries.* Deficiencies in data are more serious for analyses of select- ivity by fathers' occupation than for parental schooling; often one has to use data for all adult males. A special examination of the problem of correct specification of a base population, with some illustrations of degree of sensitivity to these specifications, is given in Appendix B. Any attempt to make comparisons across countries multiplies the difficulties of appropriate specification and interpretation. Here again the problem is less serious for parental schooling than for other traits such as father's occupation, for the UNESCO compilations were set up to maximize comparability. In addition, where finer details are missing, collapsed and alternative categories of e7hooling will be quite satisfactory. This is because the schooling categories fall in a meaningfully ranked sequence, and the available breaking points between the categories on which data are available tend to coincide with the points at which there are shifts from under-to-over representation among secondary or university students. The fact that for the least developed countries details are fuller at the bottom of the educational pyramid, whereas they are fuller at higher levels for the advanced countries, reflects their respective underlying distributions and tells us where the meaningful breaking points in representation usually will be located. This actually facilitates cross-country comparisons if one thinks in terms of relative positions of parents in a total population. Data limitations may preclude comparisons (for example) of degrees of selectivity among children of illiterate fathers in the advanced nations with children of illiterate fathers in the less-developed nations, but these are very different * J. van L. Maas of the World Bank has provided such estimates for Tanzania. sorts of subpopulations. Furthermore, in any such comparison one runs into sharp differences in the mathematical constraints on possible values of selectivity indicators, a problem discussed in subsequent chapters and in Appendixes C and D in particular, along with ways of interpreting the indicators in the light of those constraints. Occupational classifications pose problems that are more critical for both within-country and across-country comparisons than are those we encounter in working with edixcational classifications. In fact, even within a country and even where census base data are detailed, information for students may be reported in broad and poorly defined categories or in categories not commensurate with those in the census. Even where internal matching of census and study rubrics is satisfactory, reported categories of occupations often cannot be arranged in rank-order. Furthermore, not only do occupational distributions differ substantially across societies; activities included under the same rubric may be very different. A farmer in Germany is quite unlike a peasant in Tanzania. Extremely crude consolidated occupational classifications will often be the best that we can make even within a country, and hence of course between countries. A demonstration of how occupational categories may be consolidated for comparative purposes is presented in table 1.1 for Snain, Germany, Sweden, and the United States. It will quickly be seen that some questions cut to the foundations of modern societal transfonma- tions. Civil servants often are merged with clerical workers: in view of the rise in hereditability of civil servants during the last century and the associated predominance of public employment for well-educated persons over- much of the third world, casual handling of this categorization will yield misleading-conclusions. Most elusive of all, virtually everywhere; are - 12 - rankings for business enterprisers and for farmers, both of which are exceed- ingly important for developing countries. Indeed, popular as discussions may be about selectivity by parental occupations (often treated as indicators of "socio-economic status"), parental education may turn out to be a better -.indicator of occupational status in many of the LDC's than the occupational rubrics themselves. This will be most clearly the case where certification isa pervasive criterion for recruitment to jobs. Education-occuption associations are much weaker in most of the economically advanced and in the non-socialist countries than in those with small cadres of well-qualified manpower or with bureaucratic controls over manpower allocations. Every social scientist spends much time trying to match data with concepts. "Social status" is such a concept, however ambiguous its treatment in various theories and in empirical specifications. In this report parental schooling and paternal occupation are the two empirically specified variants of this concept. Perforce one must use data from special investigations on sets of schools, for information about even these characteristics rarely is routinely collected. It will be many years before Unesco will begin to collate surveys of parental background in its reports on education statistics. It follows that the tabulations in this report are not much like those in demographic reports or in surveys of the labor force. Nor is this jigsaw- puzzle assembling of data from particular studies like a summary of medical research on a particular topic: a "disease" usually can be specified with a sufficient stability that small samples about therapy can be more informative than the partial samples reporting on characteristics of pupils. As stated in the opening paragraph of this report, the task is to ascertain whether -degree of selectivity in school enrollment can be related to countries' - 13 - xI rying "level of development." By not letting oneself be so meticulous about quality of data that he abandons the venture, the rough picture that comes out is credible and it adds new chapters to our knowledge about compara- tive education and about development. D. Methodology 1. Indicators of selectivity. Distribution of fathers or mothers of students may be compared with the base population of the parental age cohort in several ways. Most commotily used are "selecti,'ity indexes" and "indexes of dissimilarity." The selectivity index (SI) is the ratio of percentages of fathers of pupils to that of males in the adult base population for a given characteristic: e.g., secondary schooling or being a farmer. if the designated category makes up only a small fraction of the base population, the selectivity index can be very large. But a category that encompasses a large proportion of the base population (e.g., peasants in a predominantly agricultural society) could not be over-represented by a large muiltiple even under the unrealistic assumption that all pupils were children of peasants. On the other hand, large absolute differences of percentages can occur even where the selectivity index (as a ratio) varies narrowly--e.g., between .75 and 1.33. (With propor- tionate representation among students all selectivity indexes would be unity and all differences in percentages would be zero.) In addition to the very different aspects of disparities that are highlighted by ratios of percentages and differences of percentages (as between parents of students and base population) there is the important feature that one can sum up the positive differences (or alternatively the negative differences) to get a summary "index of dissimilarity" (ID). That index tells us how many individuals would need to be shifted (what percentage - 14 - of students would have to come from a different background) for the distribu- tion of fathers of pupils to match the distribution of the relevant base population. Thus the index of dissimilarity automatically displays the overall degree of contrast between students' fathers and the base population, irrespc.ctive of where in the educational or the occupational distribution the major contrasts appear between youth who might be and those who actually are enrolled. No such ovc all summarization is possible with selectivitv indexes. Most of the data that have been found with diligent search take the form of distributions of fathers of students and a distribution of an approxi- mately matched base population on the same trait. Studies that provide direct information on enrollment rates of children of men in each occupation or with each educational attainment are rare, and where they do exist the same problems of matching the students against the appropriate base population commonly arise. It is sometimes possible, however, to make rough estimates of specific enrollment rates (e.g., the proportions of children of farmers who enter institutions of higher education, the proportions of children of professional men who do so, and so forth). This can be done if there are sufficiently comparable data for the distribution of student fathers, the distribution of the base population, and the overall enrollment rates in the level of schooling of the students being studied. Specifically let E be the enrollment rate for youth whose fathers have a given trait i, let Si be the selectivity ratio for that category of fathers, and let Et be the overall enrollment rate. We then have the simple equation Ei = Si. Et. Limited use will be made of this relationship in this report, especially for the few cases that permit compari- sons over time within a country; at that point the proof of this equation will be given (in chapters 5 and 6). -.15 - One may compare these specific enrollment rates in various ways. The simplest is just the absolute difference between any two such rates (in the above example, the enrollment rate of children of farmers compared with that of professional men). A better measure in such cases, however, is a comparison of the logits of the enrollment rates; among the advantages of using logits is the fact that they are more appropriate if we want to view enrollment rates at any given time as observations along an underlying potential growth curve. Illustrations of such applications and the insights we can gain from them are set out in Chapter 5. The transformation of percentages into a logit form gives a cumula- tive distribution that is one of several variants of the familiar S form of- most growth curves. A closely related transform is a probit formulation; in this report we will make use of the "latter only as a charting device, using probability grids. This has substantial advantages over any other chart for visual presentation of percentages of students' fathers against percentages of a base population with a given trait. In addition, it provides a quick picture of variations in degrees of selectivity that can be interpreted as standard-deviation units of departure from a fully representative distribution of student fathers. Although logit (and probit) transforms are less easily understood intuitively than the more familiar selectivity indexes, they have several special advantages for comparative analyses of selectivity. These transforms take into account the fact that possible ratios and differences are limited by the percentile range in a distribution against which comparisons are made. Use of logits on grouped data is a simple way of avoiding this problsm, both technically and conceptually. Accordingly, in regression analyses of Lffects - 16 - of per capita income and enrollment rates on selectivity of educational participation (in Chapters 2 and 4) the selectivity measure used is the difference between the logit of proportions of students' fathers with a given trait (scho'oling or occupation) and the logit of the proportion in the corre- sponding base population of adults. (In addition to discussions in Chapters 5 and 6, a formal statement about probits and logits is given in Appendix E.) If asked about equality of opportunity (without utilizing the discipline of numbers) perhaps few people would object if large proportions of the children of university-schooled men finish secondary school or university. Yet when one perceives the high selec;vity indexes for such children, the reader may be startled by the seeming disparities in opportunity. It is all too easy to overlook the fact that high ratios would be mathematically impos- sible for children from families that make up the broad population. The selectivity indexes of greatest importance, however, are just those relating to the lower end of the distribution of parental schooling or occupation. Concern about the distribution of opportunities should be'focused upon the less-advantaged parts of the population if we are to learn what is happening to those who may be "left b2hind". Such also must be a focus of attention if there is concern about whether spread of education in a less- developed country is building a new class system-or is maintaining something like comparable chances for further education of children born in poorer circumstances--however low the overall enrollment rates may be. It is essential in seeking to interpret any of these indexes that they be viewed in relation to the constraints on their possible values that are inherent not only in the distributions of the base populations but also in enrollment for the given level of schooling. Implications of these -17- constraints are set out formally in Appendixes C, D, and E for selectivity indexes, indexes of dissimilarity, and logit comparisons, respectively. Those constraints and relationships of observed to maximum possible (or minimum possible) values of the indexes are illustrated in the interpretations of the data in ensuing chapters. Expansion of enrollments is of course one instrument for the enlargment of opportunities, and it is a comparatively easy way to diminish selectivity although it is not the only way to do so. For any given overall enrollment rate, selectivity would be maximized if there were strict queuing by socio-economic background for places in the post- compulsory levels of education. An important part of this report will be to trace how far variations in selectivity indicators may reflect effects of expanded schooling, how far other factors (or policies not explicitly identified) contribute to a modification of social selectivity of educational participation. 2. Avoiding simplistic assumptions about national patterns of selectivity. The present investigation was begun with the expectation that one would find no close correlation between the degree of selectivity for schooling and the level or type of society. To be sure, this viewpoint reflects earlier work using mainly "advanced" countries, but it reflects also several years of work on the broad topic of "education and development." I see societies as configu- rations of loosely connected institutional structures. The following para- graphs link an earlier study on selectivity of university attendance to the topic of the present report in order to give readers a certain skepticism about generalizations. For those who must be concerned with policy about the place of education in development, a conclusion that societal type and educational selectivity are tenuously connected could be more congenial than a deterministic conclusion. - 18 - Data for four countries are chosen to illustrate contrasting situa- tions (Anderson, 1956). Setting aside the United States for a moment, the other countries were similar ini overall enrollment rates and in the proportion of students who were women (Table 1.1). Students from manual families (agriculture plus labor) were about equally infrequent in two quite dissimilar countries: Spain and West Germany; by contrast such families furnished nearly a fourth of the Swedish and over a third of the U.S. students. (In all four countries farm laborers had to be in-lu1ed with farmers rather than with other laborers, a difficulty often met with in this sort of investigation.) Students from farm families were about equally common in the two least- agricultural countries and not greatly less frequent in either Spain or Germany. Students from non-agrarian backgrounds, however, made up varying proportions in each country. Germany was like Spain rather than like either Sweden or the U.S. in representation from broadly manual families; that Germany is highly industrialized was not reflected in the share of students coming from laborers' homes. In Germany and Spain agricultural families supplied a larger propor- tion of all students than did urban laborers, whereas the latter groulp in Sweden supplied a half more and in the U.S. three times as many students as did farmers. "White-collar" families furnished nine-tenths of the students in Spain as well as in Germany, three-fourths in Sweden, but only three-fifths in the U.S. It would seem that the status profile of a society gives no firm basis for anticipating how accessible higher education will be for youth from different social backgrounds. Sweden was similar to Spain and Germany in overall enrollment rate yet the profile of parental status in Sweden resembled that in the U.S.; see the indexes of dissimilarity. The comparatively high proportion of women among U.S. students should have produced an upward tilt of status among - 19 - TABLE 1.1 Indexes of i:a:us of University Students ir. .our CountrLes 1/ LUniteLi Stater-Sweden Spain Germany 1 j7 1930 1945 1928 Students per 100 aged 2U-'24 (bothi sexes) '0.9 1.8 2.5 1.7 Femnales as percentage of all students 33.0 13.0 14.0 15.0 1t. prere raye di. strilhiuwon '! paternal status A,griculture + labor 37.3 23.1 9.4 7.5 aericulure (9.8) (9.5) (7.0) (5.8) non-turm labor (27.5) (13.6) (2.4) (1.7) All others '2.7 76.9 90.6 92. Total 100.0 100.0 100.0 100.0 B. Indexes ot-dissi,ipt!ari rC vs. U.S. (wh:res) -14.2 27''9 -.8 vs. Sweden 14.2I - 13.7 15 . vs. Spaini 27.9 13.7 - 1.9 vs. Germany 29.8 15.6 1.9 - C. Ratio of acrual ro expecrel percenitages bY parentral background Avriculture + labor 0._7 0.29 0.11 0.1 1 Agriculture 0.62 0.29 0.13 0.3 2 Non-farm labor 0. 5 5 0.30 0.08 0.04 All otlier 1.8 2.8 5.2 2.8 Professions 3.4 25.6 13.2 1.' Entrepreneurs 3.0 2.1 - 23 Whites only. Inclusion of Negroes (Blacks) would not greatly alter the figures; nearly 5% of the cohort were enrolled, of whom half were female. Almost half were from non-manual homes, 19% from farms and 36% from non-farm labor homes. The respective ratios to expectancy (in the order given for whites) was 0.64 (0.48, 0.79), 5.9, 9.3, N.A. - 20 - students (since women normally come from higher-status families than do men). The status profiles for students in Germany and Spain are similar--surely not what one would have anticiapted--each being about equally unlike Sweden and even more unlike the profile for the U.S. One relates the profile of status among students to that for the labor force by computing the selectivity index (given at the bottom of Table 1.1). Manual families are indeed poorly represented in the higher schools of Spain and Germany, but urban workers were even more under-represented (and especially in Germany). The two categories of manual labor were about equally represented (except in Germany) but at a comparatively low level in Spain and at a high level in the U.S. Throughout the following report attention will be called to the importance of measuring opportunity for the "populace" (non-white collar families) and comments such as those just made will recur. The other (broadly white-collar) families are far less numerous in a population but usually vastly over-represented among students. Nonetheless, differences among countries in the pattern of this over-representation are impressive. In the present illustration, the "elitist" pattern for Spain is marked, at least for the aggregate of non-manual families. By this criterion Sweden and Germany are on an intermediate level while the U.S. is distinctly non-exclusive. It is perhaps surprising that entrepreneurial families are only moderately in excess of quota. And surely the very low selectivity index for professional families in the U.S. (barely greater than for enterprisers' children) stands out. This brief comparison among four countries illustrates the complexity of the notion of "access" to higher education. It would seem that historical or institutional influences play a large part in the degree and pattern of social selectivity for higher schools. - 21 - 3. Abstemious use of theoretical and regression models. This report may be described as "analytical-descriptive." It does not expound on any of the social-political theories that so often arouse acrimonious debates; no grand structure, ideological or otherwise, is presented. Furthermore, it must be emphasizad that all that can in fact be directly measured even with the best of data is selectivity of educational participation. This is not the same thing as inequality of opportunity or of access; opportunities for schooling are not always utilized, for a variety of reasons. Furthermore, accessibility of a family to education for its children is not the same thing as accessibility to the child; parental controls and attitudes intervene here. It can be assumed that there is some correlation between participation rates and accessibility, but the fact that this is not a close correlation is demonstrated by the sex contrasts in participation rates in many countries. It is shown even more unambiguously by the efforts and frustrations accompanying attempts to induce certain sub-populations to make use of available places even in primary schools. Analytical explanatory models very different from those of the "grand" theories could be and are being applied in studies of the spread of schooling in some of the LDC's. These studies may refer to geographic units or they may use primarily survey data on households. We have studies of the effects of household traits on rates of attendance (or enrollment) in secondary schools or universities for several less-developed countries, including Malaysia, Colombia, and a sample of villages in India. These studies employ econometric or sociometric models of varying sophistication. Those based on household surveys are rooted in micro-economic theories of family decisions with respect to numbers of children and investments in them. Those using area - 22 - data for India, Mexico, and Brazil combine theories of the diffusion of innovation developed by human geographers with micro-economic decision theory. Such studies are valuable for the interpretation of the selectivity measures in this report, but they require intensive study for particular countries, which is impossible in a broad comparative study. Regression analysis is limited in view of the priority that must be given to examination of the data and of the uses and constraints on conven- tional indicators of selectivity. Thus, emphasis is placed first on seeking comparability of data and on the use of scattergrams to examine relationships and to identify underlying situations that affect the nature and meanings of observations in widely differing social and economic environments. An important part of this probing is the specification of variations in mathe- matical constraints on possible values of conventional selectivity indexes and of indexes of dissimilarity. Only against such a background do I deem it legitimate to use regression analyses, and even then only with caution, recognizing the defects and many error components in the data. Where used, the regression analysis follows upon questions raised in the prior examination of two-way relationships, and primarily to identify the interrelationships of per capita incomes and enrollment rates in their association with selectivity. The dependent variables in the regression are indexes of dissimilarity, logits of proportions of students' fathers with given characteristics, and the logit- difference measure of selectivity mentioned earlier. E. The Multiple Facets of Selection Processes Families in every society are assorted or classified into many categories, using diverse criteria or bundles of criteria. Because the time and place of contact with the intrusive West can be so important in the way - 23 - development occurs in a third-world society, the main classificatory criterion may well be place of residence. Closely associated with locality is ethnic group or tribe, and (because of the uneven incidence of contact with mission- aries) religious affiliation may be important. If formal schools are accepted, whether parents or older siblings or other kinfolk have used the schools can become a major basis for selectivity. Once development begins to "take", occupation or income or even just participation in the monetary sector becomes a major line of differentiation in the society. The scope of this report is limited to the effects of parental schooling and occupation upon children's education, supplemented for a few countries by examination of the effects of parental income. Though it is appreciated that data aggregated for a whole society must be misleading, the nature of sources usually dictates use of data in that form. Pupils in a school actually reside in a particular locale (even where nationally supported boarding schools are used). Their parents have a parti- cular ethnic or religious affiliation and can be classified by the manner of their connection to the established or emerging modern economy. In some degree the positive advantages for utilization of the schools cumulate or pyramid; certain tribes received schools earlier, more often engage in rnon-traditional ways of earning a living, and so on. But in other respects the various classificatory characteristics are offsetting or crosscutting; members of disadvantaged ethnic groups or residents of remote hinterlands or those who have a humble occupation but who live in towns may receive better opportunities than do the rural members of more advantaged ethnic categories. These advantages (or disadvantages) cumulate within individual families. Eliou supplies a tabulation of paternal by maternal schooling for - 24 - boys and girls (Table 1.2). The structures of cumulative advantages are not symmetrical. Because fewer mothers have good schooling, if a mother is schooled the father will usually be; however, a large propotrtion of the well-schooled fathers have unschooled wives. Bearing in mind that influences on participation in schooling to successive levels may be cumulative but also can be partially offsetting, it may be helpful to break this down into a few more manageable components, beginning with geographic location. 1. Geographic effects. As schools are inaugurated successively in the more isolated districts of a country, the selective acceptance of them that occurred in initial regions is repeated--with attenuated selectivity as reports about the benefits from schooling become diffused into the hinterland through kinship ties and the reports of migrants (Berry and Jackson, 1981). This process of leveling up lagging districts is accompanied by recurrent differentials between early and late acceptors among families. In the widening labor markets and in boarding schools with nationwide catchment areas, this diffusion of schooling continuously shifts the balance of social selectivity. As families in the hinterland accept schooling, the more eager arid possibly more properous families there may win opportunities (such as places at boarding schools) in competition with lower-status aspirants from the areas that had schools earlier. There will also be exercise of privilege in the use of personal influence to secure places for sons (and daughters) even where their performance is relatively poor; such influence operates in socialist and communist societies as in capitalist ones--in part through similar, in part through different processes. Net effects on measured selectivity of educational participation can be diverse, but an underlying - 25 - TABLE 1.2 PERCENTAGE DISTRIBUTIONS OF SCHOOLING OF MOTHERS RELATED TO SHCOOLING OF FATHERS; SECONDARY PUPILS IN SENEGAL, 1967 Schooling of Schooling of Mothers Fathers None Primary Secondary Higher Total B oys None 98.6 1.3 0.1 - 100.0 Primary 85.8 12.3 1.6 .3 100.0 Secondary 66.3 21.5 11.7 .5 100.0 Higher 49.5 19.2 - 17.6 13.5 99.8 All None 68.5 12.2 3.9 - All Primary 21.1 40.6 14.2 8.8 Secondary 8.2 35.6 52.8 8.8 Higher 2.2 11.6 29.1 82.4 Total 100.0 100.0 100.0 100.0 Girls None 97.3 2.6 0.1 - 100.0 Primary 78.2 20.2 1.6 - 100.0 Secondary 57.4 26.1 15.4 1.1 100.0 Higher 38.7 23.2 23.8 14.3 100.0 All 77.5 15.1 5.9 1.5 100.0 None 46.4 6.3 1.0 - All Primary 34.4 45.4 9.5 - Secondary 14.5 3309 51.4 14.3 Higher 4.7 14.4 38.1 85.7 Total 100.0 100.0 100.0 100.0 Source: Eliou International Review of Education Vol. 19 No. 1, 1973 pX40 Simhilar data for Kenya and Tanzania are available in Sabot 1981. - 26 - thread in educational development is always the effects of increasing rates of enrollment, remarked above. 2. Ethnic identity. The importance of ethnic membership as a line of cleavage in third-world societies varies considerably from one country to another; so do policies directed to this problem (which may either alleviate or aggravate it). While this report is not addressed to the question of selectivity between ethnic groups, it does include a few sets of data relating to intra- ethnic selectivity. This can be a complex phenomenon. Thus in Ghana Fost.r r found that boys from some tribes were present in secondary school at well below their random proportion, while boys from other tribes were over- represented by a wide margin, and some seemed to hold places at about random frequency. Using adapted data that show parental schooling and occupation against tribal over-representation (Table 1.3), one sees that the supply by tribe of the prospective leaders of Ghana is strongly affected by the selective entry of boys into secondary school according to the distributions of paternal occupation and education in the various ethnic groups. I assume that for the most part father's schooling and occupation have greater influence than tribal membership. However, this raises questions as to why some tribal groups substantially exceed others in both education and economic modernization; historically tribal membership may be a prior attribute that encompasses both characteristics of the traditional culture and-early contacts with western influences. 3. Kinship ties and family size. Like ethnic membership, kinship extends beyond social status and may affect educational opportunity. Among the best known examples are the importance of clan membership for educational and other opportunities in precommunist China (and indirectly in China even - 27 - TABLE 1.3 Selectiveity Indexes of Ghanaian Secondary *Boys by Paternal Occupation and Schooling with Associated Percentage Distributions of Indexes Among Tribes. - Percentage distribution among tribes by tribal selectivity index Father's Selectivity (Each row totals 100) Occupation Index Over 1.5 1.0-1.4 Under 1.0 Professional 4.9 61 29 10 Business 2.9 38 39 23 Skilled 1.1 48 41 11 Semi-/unskilled 0.1 36 36 28 Farmer, etc 0.6 15 54 31 Father's Selectivity Schooling Index University 13.0 49 43 8 Teacher training 10.1 67 26 7 Secondary 1-6 6.9 73 25 2 Middle 1-4 2.5 47 38 15 Primary 1-6 1.5 26 56 13 None 0.1 15 47 38 28 - today). A micro study into the distribution of educational o'portunity would have to take account of diverse influences through kinship quite apart from associations between family size and level of income. In many societies financial or other assistance may be available from kinfolk outside the nuclear family (Sinclair). Different systems of polygyny have varying arrangements for priority among siblings in privileges (including now going to school) or for economic assistance in asscorted ventures. Quite apart from economic ramifications of kinship, educational aspirations for a child are mediated by kinship structures. Here one confronts the familiar dilemma: expand our ethnographic information about kinship relationships or focus first upon the more quantitative features of selectivity into schools. Taking account of kinship ties may make little improvement in how well we can predict the intensity of selectivity for schooling in a society even though we may gain further understanding of how selection operates. The kinship and nonkinship factors interact in various ways in different societies. One need only recall the outstanding record in western societies of clergymen's and teachers' children in making use of schools and universities despite modest economic resources and social position. Examination of some of the factors in selectivity of girls relative to boys in opportunities to attend schools to various levels is included in another report. A few remarks about relationships between family size and school attendance of girls may be worth noting here, however, along with observations concerning the sibling correlates of educational participation. In her study of Papua New Guinea, Palmer found that secondary girl pupils were more likely than boys to have siblings who also had schooling. Muckenhirn - 29 - reported that in Western Nigeria girls in secondary school were more likely than boys to have siblings (age 13-20) attending, and for each sex this likelihood was greater among those in grammar than in secondary-modern schools. These findings are what we should expect given higher rates of enrollment among boys than among girls. Less predictable, however, is Palmer's finding that for each sex in Papua New Guinea there were more oldest than youngest children attending secondary schools, whereas a random dis- tribution would have yielded equal numbers. Ihdeed, in many societies either the oldest or youngest daughter is expected to care for parents, and older daughters are likely to have greater responsibilities for helping with younger siblings, while middle daughters are less bound by either sort of responsibilitity; sons may escape such claims almost entirely (but not claims on their time to help in agriculture, for example). Ben Salem's report on Tunisian university students contains some surprises. Overall the median size of student's families was six: a tenth came from families with at least ten siblings while only a sixth were from families have less than four children. In total, 44 percent of students were from families having seven or more children, and this proportion was under forty only for upper-status and white-collar families. Youth from humble backgrounds were not rare. Moreover, the youth from lower-status homes who did attend higher schools were not from distinctively smaller families. Of course, all aspects of a family are intertwined with economic features; we need to know the combined distribution of family size and family status both for students and for the population at largfa--data rarely found. (This can be of some importance even for appropriate specification of the base population in comparatively simple estimates of educational selectivity, as is - 30 - shown in Appendix B.) Such a tabulation was published for Hungary for the year 1930 (for mothers aged 40-59 who had been married at least twenty years). In those data (Table 1.4 ) status outweighs family size and is of roughly similar effect for each size of family. That pattern need not prevail at other periods or in other societies, and the relationship would be more complicated for polygynous families. It is only to be expected that family size would have least influence on attendance among the "educated classes" Among single-child families at this level one child in five enrolled in contrast to Qne in two hundred from farm families; in the largest families the comparable proportions were one in thirty versus one in a thousand. "Educogenic families" has become a vogue word. In some families a strong aspiration for schooling of their children is taken for granted because older members of the family have been well schooled, while in other families this taste for schooling has barely begun to be assimilated. In most societies only a minority of secondary pupils and a tiny fraction of students in higher schools come from families in which a parent had attended the given level (or even the just-lower level) of school. The importance of this differential incidence of parental schooling for interpretation of indices of selectivity underlies the strategy of analysis for this report. Data about schooling of siblings, cousins, or older kin such as uncles or aunts is virtually non- existent. In much of the third world the present pupils (often even in elementary school) are the first of their family to have the now-favored variety of formal schooling. Where we have data showing parents of pupils classified by schooling, we can estimate the proportion of first-generation pupils or students. (A careful study of this topic would try to compare trends in schooling possessed by the best-schooled fourth or third of adults - 31 - TABLE 1. 4 STUDENTS IN HIGHER EDUCATION PER 100 CHILDREN BY SIZE OF FAMfILY AND OCCUPATION OF FATHER: IUNGARY 1930 Number of Children 1 2-3 4-5 6 all Agriculture .51 .31 .18 .11 .20 Mining and Industry 3.55 1.70 .65 .37 .99 Commerce and Credit 6.26 3.42 1.57 .72 2.31 Transport 8.17 3.79 1.43 .53 2.21 "Educated Class" 19.50 12.44 6.97 3.67 9.19 All 3.25 1.72 0;69 .31 0.96 - 32 - since "college graduation" (e.g.) has come to have less indicative value over time and one needs to have relative schooLing.) If a society has had a strongl.,y r elective secondary (or higher) system ov,;r many decades, the proportiouJ oFl students with well-schooled parents could be higher while the propchl.:i.on of students who are first-genera- tion could be low. However, where attendanuce at a given level of school is expanding rapidly in ratio to population, the proportion of first-generation pupils must rise and the proportion of first-generation students will vary with social status of their families. Thus, for Yugoslavia Martic reports that the percentage of university students who had a relative also possessing higher education was for workers 12 percent, for farmers 15, for artisans 19, for clerical and civil servants 28, and for professionals 39. A quarter-century ago, Jahoda found for the Gold Coast (Ghana) that schooling among parental and grandparental generations of the students was as follows: Grandfather Grandmother Father Mother University 5 - 6 - Secondary 27 3 65 16 Elementary 113 48 105 66 None 331 425 62 156 Total 476 476 238 238 In neither generation was the correlation between schooling of fathers and of mothers more than negligible. Moreover, for both husband and wife, having more formal schooling tended to be associated with polygamy (in opposite ways), with social status, and also with age. But these networks of factors remain to be explored (on other than special samples) for any country. - 33 - 4. "Normal age" students, overage students, and the pace of change. Associations between the ages at which children entter and complete various levels of education -,d their parental backgrounds are quite common in third world (and other) countries. This is part and parcel of the spread of schooling to various levels through a population and of the "catching-up" process among those who did not at first have opportunities or did not respond to them. For individuals, the process starts with entry to primary school at a standard (even though not yet typical) age instead of more belatedly, and it is associated with regularlity of attendance as well, contributing to rapid completion of an educational cycle and to low drop-out rates. Those from comparatively educated families and with fathers in white collar occupa- tions tend to have the advantage. The educational situation in Ghana has been changing rapidly, and some aspects of the recent conditions are revealed by a special inquiry into the working of examinations in that country (Addae-Mensha)--a study occasioned by the rapid opening of special fee-paying preparatory schools. Pupils who attend these special schools transfer into secondary school two or three years younger than do those who have their pre-secondary work in a state school. Although the new schools have neither lower pupil/teacher ratios nor better- trained teachers, they teach in English with little attention to vernaculars. Shinman's part of the study (p. 116) found that pupils transferring into secondary school earliest had parents of higher occupational status; neverthe- less, among the youngest transferees children of farmers and fisherman were as numerous as those from white-collar families (Table 1.5). -34 - TABLE 1.5 PERCENTAGE DISTRIBUTION OF PATERNAL OCCUPATION BY MIDDLE-SCHOOL GRADE FROM WHICH PUPIL TRANSFERRED TO SECONDARY SCHOOL Grade 4 Grade 3 Grade 2 or earlier Paternal occupation White Collar 22 28 -37 Farmer, fisherman 50 48 37 Laborers 14 12 10 Other 15 13 16 Total 101 101 100 Contrasts of age of entry and steadiness of progress through the schools do not depend on the availability of cram schools or of elite streams in secondary schools, however. Many examples can be cited from the earliest years in primary schools, and from many parts of the third world. Indeed, this phenomenon commonly plagues those attempting to identify enrollment rates for the supposedly relevant age groups in a population, as has been demon- strated in a number of studies including work by Phyllis Goldblatt on Mexico, by Devindra Sharma on India, and by Schiefelbein and Farrell for Chile. The problem is one to which much attention has been given by those working in Francophone Africa, though not in most cases with explicit attention to which youth proceed smoothly or with delays or do not complete a cycle. Investiga- tion of these aspects of educational expansion is important in its own right, with or without specification of social selectivity into "normal" or deferred or interrupted progression through the schools. That topic is not addressed here, but it has been taken into account (as it must be) in the specification of enrollment rates and their constraints on selectivity indicators, mentioned earlier. - 35 - F. The Structure of This Report The broad structure of this report is indicated in the table of contents. Chapter 2 lays out in considerable detail the data and findings on educational selectivity with respect to the education of parents. That chapter sets the basic pattern for Chapters 3 on selectivity by occupation. in that most of the problems in specification and interpretation of indicators of selectivity are encountered and discussed. In Chapters 2 and 3 alike, summary information is provided on distribution of parental backgrounds of students even when base-population distributions were not available, and charts display some of the relationships between those distributions. Indexes of dissimi- larity and selectivity indexes are also presented. The selectivity indexes are examined in relation to their mathematically possible maximum values (for over-represented population groups), atid where relevant they are examined also in relation to minimum values that exceed zero in the representation oI under-represented segments of the population. Problems of comparability with base populations and across countries are more severe in assessments of selectivity by parental occupation than by father's schooling. One consequence is that in Chapter 3 more explicit attention must be given to such problems and to discussion of the relativity c,f occupational rubrics and the divergen- cies in their meanings for a cross-country analysis. Both chapters 2 and 3 end with applications of regression models to analysis of relationships between per capita incomes and selectivity by parental background, with special attention to over-all enrollment rates as an intervening variable. Chapter 4 compares results on selectivity indicators for parental schooling and for parental occupation in countries for which both types of information were available. - 36 - Chapter 5 shifts focus slightly to turn the indicators around in an examination of background-specific enrollment rates, asking how the chances of attaining secondary or higher education are affected by family characteristics. This is where logit analysis becomes most interesting, since it is a superior approach to an analysis of the diffusion-of a trait. This same theme is then taken up, along with the types of selectivity measures used in previous chapters, in Chapter -6. Here the emphasis is on changes over time in each country for which this was possible. .In chapter 7 attention is turned briefly to variations in selectivity by types of institutions and curricula. That theme is continued in Chapter 8, where a few illustrative studies of educational selectivit,'y by family income are examined. Although-brief and limited in country coverage, Chapter 8 includes examples from both third world and advanced countries, and from countries that differ substantially in other respects, levels of development aside. Finally, Chapter 9 attempts to pull together the main findings and to point up promising and important avenues of further research to illuminate issues in public policy. - 37 - Chapter 2 FATHER'S EDUCATION AND SELECTIVITY OF ENROLLMENTS IN SECONDARY SCHOOLS AND UNIVERSITIES This chapter has five sections. First, the profiles of distri- butions of schooling of parents and those of the base population are pre- sented; relationships between these profiles are shown for selected levels of paternal schooling (using double probability paper). Second, overall degrees of selectivity are examined using indexes of dissimilarity between the fore- going pairs of distributions country by country. Third, selectivity indexes are then presented for under-and over-representation in secondary and higher schools for children of men with specified amounts of schooling--the status hierarchy of educational opportunity in a sense. Fourth, constraints on possible values of selectivity indexes are specified. Thereby one makes explicit the extent to which the relative size of a category of schooling among adult males and the rates of enrollment among young people will auto- matically constrain the measured degree of selectivity. This analysis also shows the range of variations in observed selectivity within those constraints. Finally, the association between selectivity of enrollments and the economy--as measured by per capita incomes and proportions of the labor force in agri- culture--is explored. That analysis includes regressions with estimates of direct partial effects of income and indirect effects via enrollment rates. A. Profiles of Education of Fathers and of Base Populations Distributions of education of parents of students are informative irrespective of variations in degree of selectivity. The profiles tell us something about the mixtures of family backgrounds of youth who will shortly play leading roles in the national policy and economy. At the same time, they provide clues concerning the communication networks that may link future - 38 - leaders to the larger population of the less schooled, as backgrounds of the newly educated overlap with those of the population at large. And they provide clues concerning communication networks among people of various backgrounds who will be thrown together in their adult lives. There are some close analogies in some respects with the growing literature on measurement and interpretation of ethnic segregation in American cities, and on relationships between physical and social distance. Here, however, the chances of interaction across social groups are delineated through selectivity of participation in post-compulsory schooling by parental schooling (or occupation) instead of by ethnic origin. As in the studies of residential segregation, measurements of relative likelihoods of meeting or communicating with people of quite different backgrounds overstate the actual relative frequency of such communications to the extent to which communication networks are selective within the designated residential areas (in studies of residential segregation and ethnic "isolation") or within levels of schooling (in the present study). Variations among schools and classrooms in the social composition of student bodies are more likely to be important at secondary than at higher levels, and where an elitist educational system with formal tracking has been inherited from the past than where it has developed over recent decades. (Some of the limited data available on selectivity by schools and curricula are discussed in Chapter 7.) There can be no doubt about the existence of associations between social and physical distance during the school years. Youth who continue into secondary and into higher education are brought together and separated spatially from those who do not continue. Moreover, their communicatioa networks are reinforced by common goals and subsequently by clustering of their activities as adults. The chance of meeting with others in the same age - 39 - cohort who come from quite different backgrounds are clearly dependent in large degree on the socio-economic mix in student bodies. Thus sons and daughters of well-educated men, for example, will be much more likely to know and communicate with youth whose fathers were unschooled when the latter make up a substantial minority or even a majority of the student body. One of the things this study cannot tell us, however, is how far the newcomers from relatively unschooled families who enter the labor markets from secondary schools and universities will maintain active communication with people who share their origins, how far they will divorce themselves from those origins. It is not without reason that concern about the social distance that emerges between secondary youth and people whose origins they share is often expressed in some of the less developed countries. A summary of distributions of father's education for secondary and for university students is provided in Table 2.1 by region and by sex. (Details by country on both education of fathers and of men in the age cohorts of students' fathers are included in Appendix J. The scanty data on schooling of mothers are presented in Appendix G.) Where information by sex of pupil is not available, the undifferentiated information is included in Table 2.1 in the column for males, with footnotes specifying the number of such cases. Usually, especially for university students, the preponderance of males makes these combined figures fair approximations to the males' paternal schooling, but with varying degress of inderstatement of selectivity since socio-economic selectivity of girls is almost universally greater than of boys. As Table 2.1 shows, all countries in which 30 percent or more of fathers of students had no schooling were in sub-Saharan Africa with two exceptions for secondary and one exception for university students. The exceptions were Tunisian males in secondary schools and both males and females - 40 - TABLE 2.1 FREQUENCY DISTRIBUTIONS OF PERCENTAGES OF FATHERS WITH DESIGNATED AMOUNTS OF SCHOOLING; SECONDARY AND UNIVERSITY STUDENTS BY WORLD REGIONS Africa South of Europe, Japan, Percentages All Sahara Mediterranean Asia Latin America North America H F M F M F H F M F M F Secondary Students Fathers with no schooling Under 10 4 5 - 2 - 2 2 a 1 2b - 10 - 29 4 3 b - 1 - - - - - - - 30 - 49 6 3 5 b 2 - - 1 1 - - - - 50+ 3 - 2 a - - - - - Total 18 8 10 4 2 2 3 2 2 - - - Atleast some secondary Under 10 4 1 4 a 1 - - - - - - - - 10 - 29 11 4 4 b 2 1 - 4a 2 2 - - - 30 - 49 6 6 2 b 1 - - - 1 2 b 2 2 2 504 9 5 - 1 1 2 - - 7 c 1 1 1 Total 30 16 10 5 2 2 4 3 11 3 3 3 Beyond secondary Under 10 12 5 6 f 2 1 - 4a 3 1 a - - - 10 - 29 13 5 - 1 1 1 - - 5 d - 7 3 30 - 49 i 2 1 - - - 1 - - 2 b - - - 50+ - - - - - - - - - - - Total 27 11 6 3 2 2 4 3 8 - 7 3 University Students Fathers with no schooling Under 10 5 2 - - 3 d 1 - - 2 a 1 - 10 - 29 6 1 4 d 1 1 - 1 - - - - 30 - 49 1 1 1 - - - - 1 - - - 50+ 2 - 1 a - - - 1 - - - - Total 14 4 6 1 4 1 2 1 2 1 - Atleast some secondary Under 10 - - - - - - - - - - - 10 - 29 7 2 5 e 1 1 - 1 1 - - - 30 - 49 3 - I a - - - 1a - 1 a - - 50+ 8 2 - - 2b 1 - - 2 a 1 4 f - Total 18 4 6 1 3 1 2 1 3 1 4 - Beyond secondary Under 10 7 1 4 b 1 1 - 1a - 1 a - - - 10 - 29 13 2 - - 3d 1 - - 3 b 1 7 b - 30 - 49 11 - - - - - - - - - a - 50+ - - - - - - - - - - - - Total 21 3 3 1 4 1 1 - 4 1 8 - a/ One case undifferentizted by sex b/ Two cases undifferentizted by sex c/ Six cases undifferentiated by sex d/ All cases undifferentizted by sex e/ Four cases undifferentiated by sex f/ Three cases undifferentizted by sex - 41 - in higher institutions in Papua New Guinea. Sex differentiation in enrollment rates and in selectivity are extreme in Tunisia, and there has been a dramatic educational leap over the past generation in Papua New Guinea. The tables in Appendix J display even greater diversity in proportions of students from unschooled homes than can be read directly from Table 2.1. For university students the range is from 3% in Portugal (and 2% of females in Tunisia) to 65% of male students in Papua New Guinea. The proportions of secondary students whose fathers lacked any schooling range from 2% for girls in Turkey to 74% for boys in the Ivory Coast. Everywhere fewer fathers of girls than of boys were unschooled. Dispersions in the proportions of fathers who had "at least some secondary" schooling is equally large. It should come as no surprise that most countries in which 30% or more of the parents of secondary students had attended secondary schools were in Latin America or in advanced industrialized nations. With the exception of sub-Saharan Africa, fathers of a majority of the university students of both sexes had gone beyond elementary school. Fathers with post-secondary schooling were a minority in all cases, however, even among university students, although they exceeded 10 percent in all the advance courntries while falling well below that figure in sub-Saharan Africa, in Malaysia and Papua New Guinea, and for males in Tunisia. Indeed, in a few countries (notably Papua New Guinea) it would have been literally impossible to fill even one secondary-school classroom each year with sons and daughters of college men. Some correlation between the distribution of schooling among fathers of students and in the base population of the father's age cohort is to be expected and is indeed observed. Nevertheless, there is also considerable diversity in the degrees of closeness or disparity Qf those distributions even among countries at similar levels of economic and of educational development. - 42 - Societal implications of any given distribution of the family backgrounds of students depend in important ways on the distributions in the population at large -- whether our interest is in the spread of communication networks and linkages between the highly schooled and the rest of the population or in questions of equity in access to and participation in post-compulsory levels of education. Detailed data on education in the base populations and among students' fathers are shown in Appendix J, where they can be compared country by country. Selected data for base-population proportions plotted against proportions plotted against proportions of students' fathers are shown in Charts'2.1 and 2.2 for secondary and for university students respectively. These charts are drawn on double probability paper for two reasons. First and simplest, the probit scale gives the best visual comparisons because of the way in which entries spread out on that scale. More important and more subtle is the fact that a comparison on doilble probability paper gives us auto- matically a superior way of looking at selectivity in the context of educa- tional development. This point is pursued in Chapter 5, which makes further use both of probit charting and of logits; see also Appendix E. For the moment it suffices to make three points about the reading of these charts. First (as with any scattergram), one can read the frequency distributions of the variables directly from the horizontal and vertical scales. Here the scaling is in what amounts to standard deviation units around a normal hypothetical distribution centering on 50 percent. Second, the deviations of percentages of students' fathers in the designated educational category from proportional representation relative to the base population is shown by the vertical distances from the diagonal. Third, each chart shows two different comparisons: the one for a relatively over-represented category (fathers with . -43 - Chart 2.1 Selected Comparison of Fathers' Education with Male Base Population; Secondary Students Percentage of Fathers 99.99 99.9 99.8 A Male/ * Female r '.360/ 99 - All IP - .077)/ e/ 290 10 70, 600 0.1. 9e8 eta. ofMl/aePplto so ./o0 40 - -A 30 t dS 1 - C/ 05 ---- - Sk 0.051 0.01 /|II11111IIIIII:. 0.01 0.05 0.1 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99 99.8 99.9 99.99 Percentage of Male Bas Population - 44 - Chart 2.2 Comparison of Fathers' Education with Maie Bae Population; University Students Percentage of Fathers 99.9., rA Male / . Female r-.8/ 9,9.9 - All (p - .02) 99.8 98t V/X 615 80 ~36 5 600~44 0 A,-. t 20 I 021 1367 60 t- 2 C5 50 65 54 0 032 0_102a A1026 .S 47 82 8113 205 / 192 089 (< 8) 65Q 1 /02 797 10 1952 20 (< 8) 0.01 0.0}50.1 0.2 0.5 1 2 5 10 20) 30 40 50 60 70 80 90 95 98 99 99.8 99.9 99.99 Percentaqe of Male Base Population - 45 - post.secondary education) gives a scattergram lying entirely above the diagonal; the other, for a: relatively under-represented category (elementary or less) gives a scattergram that lies entirely below the diagonal. On each chart the points for males (triangles) and for females (circles) in a country are connected by vertical lines. We can see clearly that there is a greater over-representation of females than of males from relatively advantaged homes and much greater under-representation of daughters of men with little or no schooling. Equally evident, especially on the charts for secondary pupils (about whom there is more information by sex), is the great diversity from country to country in disparity between male and female students in their selectivity. If degrees of selectivity were highly correlated in probit form with the percentages of the male base population having a designated schooling, the scatters of points on a chart would approximate a straight line running above the diagonal for over- and below it for under-representation. That line would portray constant degrees of selectivity from one country to another when exprensed in probit form. The lack of any such constancy is apparent, even though plotting on double-probability paper gives a "better' correlation than does plotting on arithmetic or logarithmic scales. B. An Overall View of Degrees of Selectivity The index of dissimilarity provides an overview of selectivity into schools that takes into account the entire spectrum of parental schooling; this index is shown by countries and by sex for secondary and for university students in Appendix J.1 and summarized in Table 2.2. (The technical features of the index of dissimilarity and its robustness are discussed in Appendix D.) - 46 - For virtually all countries covered in this report, the conditions necessary for robustness in the index:are met; where there is suspicion that they are not met (leadiThg to a downward bias in the index) a plus sign accompanies the entry in the table. It should be remembered that the index of dissimilarity is based on arithmetic differences between percentages; it tells us what proportion of pupils would have to come from other backgrounds of parental schooling to bring the dis.tribution of students' fathers into conformity with that in the base population. The Table 2.2 summary of the varying size of these indexes from country, to country by world region follows the same format as was used for distributions of paternal schooling in Table 2.1. Scanning Table 2.2, it appears that both the lowest and the highest indexes occur mainly among secondary pupils, while for university students the indexes are of intermediate size. It might be supposed that the index of dissimilarity would be uniformly higher for university students; we are accustomed to believe that social selectivity intensifies as one moves up the schooling ladder. Indeed, for most countries (Appendix Table J.1) the index is lower for secondary students, but there are notable exceptions. In part what we are seeing in Table 2.2 reflects the different sets of countries in the secondary and in the university samples; the sample for university students included more of the educationally advanced countries but fewer from Latin America. This is not the whole story, however. Countries differ in the relative importance of academic performance or social background as selective factors, and in the extent to which each of these is stressed at one or another stage in progress through the educational system. In some countries social selectivity may take place primarily at entry to the secondary schools, TABLE 2.2 FREQUENCY DISTRIBUTIONS OF INDEXES OF DISSIMILARITY BETWEEN EDUCATION OF FATHERS AND OF MALE BASE POPULATION BY WORLD. REGION; SECONDARY AND UNIVERSITY STUDENTS Africa South of Europe, Japan, Index of Dissimilarity All Sahara Mediterranean Asia Latin America North America M F M F M F M F M F M F Secondary Students Under 20 11 4 - - - - 2 a 1 2 - 7 3 20 - 39 9 3 3 - 1 - 2 1 3 b 2 - - 40 - 59 7 3 2 a 2 1 - - 1 4 b - - - 60 and over 4 5 1 a 2 - 2 - - 3 c 1 - - Total 31 15 6 4 2 2 4 3 12 3 7 3 University Students Under 20 1 - - - - - - - - lb - 20 -39 9 e - 2 a - 1 a 2a- - - 4e - 40 -59 12 3 4 b - 2 a - - 1 4 d 1 2b - 60 and over - 1 - - - 1 - - - - - - Total 22 4 6 1 3 1 2 1. 4 1 7 - a/ One case is undifferentiated by sex b/ None are differentizted by sex c/ Two cases are undifferentiated by sex d/ Three cases are undifferentized by sex e/ Four cases are U.S.A. male birth cohorts (1.898-1937) - 48 - with selection by examination then becoming the predominant factor in the determination of which secondary-school youth will enter the universities. Where this is the pattern, selectivity indexes into higher education when measured against parental schooling may be close to the corresponding indexes for selection into secondary school. The variations in systems are many and well worth study; such an analysis could not be attempted for the present report. Access to secondary education can vary greatly in its social dis- tribution from one society to anot' .r. It is interesting that in most of the Latin American countries secondary students are highly selected from better- schooled homes. These educational systems are not recently inaugurated, and primary schooling is widespread (though often through only a few grades), but the educational pyramids often narrow sharply by entry into secondary schools. Yet sex disparity in access is slight in most of this region. Low indexes of dissimilarity occur mainly in the more fully developed educational systems. Looking at particular countries (Appendix J), one is struck again by the contrast between the sexes in Tunisia at the secondary level and the extraordinarily high index, of 81, for young women in university. Malaysia, on the other hand, has a remarkably unselective enrollment overall, with an index of only 23. Among secondary students in Finland and in Denmark, the usual situation between the sexes is reversed, with female indexes of 17 and 10 and male indexes of 19 and 17. The Japanese index of 44 for higher education is high in view of the large fraction of young people who go on to higher educa- tion? In part this is because the proportions of youth of both sexes going beyond secondary school were lower in 1968 (to which the figures refer) than they are today, but there is another reason. Better-educated men and women - 49 - in Japan invest heavily, including the time of mothers, to make sure that their children will excell, and the pervasive drive to attain educational goals leaves little room for any lapses among children of better educated parents. Meanwhile, it is the increase in secondary enrollment rates rather than a loose association between family background and educational attainment that has brought the indexes of dissimilarity down to only 10 for girls and 8 for boys among Japanese secondary students. Indexes of dissimi- larity between the sexes in fathers' schooling (not shown) are consistently higher in the less developed countries of Africa and in Tunisia and Turkey than in countries of other regions. It was suspected that the index of dissimilarity might be affected especially by a concentration of fathers in the ."no schooling" category, but plotting the latter percentage against the index of dissimilarity revealed no association foreither secondary or university students. Essentially the same lack of association occurred also when "more than elementary" ("at least some secondary") schooling of the base population was related to the index of dissimilarity, although there was a slight tendency toward lower indexes when larger percentages of the base population had gone beyond elementary school. This association, though weak, undoubtedly reflects the fact that the maximum possible index of dissimilarity is constrained by the enrollment rate -- a constraint of considerable importance for selectivity of students in secondary school eveni though every index of dissimilarity is in greater or lesser degree below the maximum possible value it would reach in a strictly ranked queuing for entry to secondary school, C. Who is Over- and Who is Under-Represented? Selectivity indexes relate the percentage of pupils' fathers to that for the corresponding category of the adult population (percentages of - 50 - fathers divided by percentages in the base population). Frequency distri- butions of selectivity indexes by father's education are presented in Table 2.3. A value of 1.00 in such an index indicates proportionate representation; entries below that value show greater negative selectivity the smaller they are, while indexes above unity reflect more selectivity the larger they are. The entries in the upper part of Table 2.3, which refer to the less educated populations, are fully as important as those in the lower half-- perhaps they are more important. Yet ratios below unity rarely have received the attention that has been given to the higher ratios of fathers' percentages to base-population percentages for the better-schooled men. In part, no doubt, this bias in emphasis has reflected the fact that indexes for the better-schooled parents, because they relate to smaller proportions of the base population, show dramatic fluctuations in size and thereby catch the attention. Indexes for the least educated may be extremey low, but do not have the same attention-getting quality-nor do they focus attention on who might be regarded as the "villain" of the piece. I suggest that both the relative disregard of the selectivity indexes for the representation of the youth from less favored backgrounds and the "villain" attitude sometimes characterizing the emphasis on ratios at the top distort our understanding. This will become increasingly evident- in following sections. Looking at Table 2.3, broad patterns can be generalized and will evoke no surprise. Among secondary-pupils the selectivity indexes for "less than full elementary" are consistently higher (closer to unity) than are those for "no schooling," and the indexes for "no secondary" are above those for "less than full elementary." This is what should be expected since we are moving up the educational scale to include in the base progressively better-schooled men. This pattern is less distinct among university students - 51 - TABLE 2.3 FREQUENCY DISTRIBUTIONS OF SELECTIVITY INDEXES BY FATHER'S SCHOOLING; SECONDARY AND UNIVERSITY STUDENTS Father's Schooling < Full No 4 Full No None Elementary Secondarv None Elementary Secondary Selectivity Secondary Students University Students Index M F M F M F M F M F M F Under .10 2 a/ 2 3 a/ - - - lb/ 1 - - - - .10 - .19 3 a/ 3 - 1 - - - .20 - .29 4 c/ 1 I b/ 1 - 1 2 b/ - - - 1 b/ - .30 - .39 3 c/ I 1 b/ - 3 c/ 2 3a/ 1 - - 2 a/ - .40 - .49 3 2 2 a/ - 5 e/ 2 ib 1 1 - 2 a/ 1 .50 - .59 4 c/ - 4 b/ - 2 b/ - I b/ - 3 1 1 - .60 - .69 '1 - 2 b/ 1 5 e/ 1 2 b/ - 3 a/ - 2 c/ - .70 - .79 - 2 - 4 f/ 3 1 - 3b/ 2 a/ 2 .80 - .89 - - 3 b/ - 8 b/ 5 - - - - 5c/ - .90 - .99 - - - - 8c! 1 - - - - - - Total 20 9 18 3 35 15 11 3 10 1 15 3 Father's Schooling At least At least some Beyond some Beyond Secondary Secondary Secondary Secondary Secondary Students University Students M F M F M F 1< F 1.00 - 1.99 6 c/ 4 6 b/ 1 2 c/ - 3 a/ - 2.00 - 2.99 6 - 2 2 4 - 3 - 3.00 - 3.99 4 b/ 3 3 c/ 1 i - 4 c/ - 4.00 - 4.99 3 a/ 1 4 f/ 1 1 - ib! - 5.00 - 9.99 9 f/ 4 5 c/ 2 4 f/ - 4 f/ - 10.00 - 19.9- 3a/ 3 5c! 3 3c 2 3a/ 20.00 - 29.99 1 b/ - 2 a/ - - 1 - 1 30.00 -39.99 - - - - 1b! - b - 40.00 - 49.99 - - 2 c/ 1 - 50 and over - - - - -1 1 Total 32 15 29 11 16 3 19 2 a/ All cases undifferentiated by sex. b/ One case undifferentiated by sex. c/ Two cases undifferentiated by sex. d/ Five cases undifferentiated by sex. e/ Four cases undifferentiated by sex. f/ Three cases undifferentiated by sex. - 52 - taking all countries together, but it prevails within any given country. The ambiguity here arises from the biases in availability of information on the base "no schooling" and for the base "less than full elementary school- ing"- biases that distort the sample when used for a summarization such as that in Table 2.3. Starting from the other direction, with selectivity indexes for per- centages of fathers with higher education relative to base population, the same generalization operates but in reverse. Since the category "at least some secondary" includes not only higher education but also fathers who had not attained that level, the selectivity indexes decline in moving from the "higher education" to the "at least some secondary" base. Whether we start from the bottom or fom the top thse cumulative specified categories may conceal the breaking points at which the shift from under- to over-representation occurs. More detailed analysis shows that in many countries the break is between parents with no schooling and those with even a little primary schooling; in Africa south of the Sahara few of the generation cf fathers of students had had any schooling. Later the break is between those who have and have not gone beyond elementary school. Children of men who had completed more than secondary school are over-represented ia secondary or unversity enrollment to a greater degree in the less developed societies (as is evident with even a hasty scanning of Appendix Tables J6 and J7). In Japan there are and have been virtually no dropouts from secondary school and the breaking points for both secondary and university stuents are in fact between fathers who have not and those who have gone beyond secondary school; this reflects the same factors that explainedthe rather high index of dissimilarity for University students in Japan, as discussed above. - 53 In Table 2.3 the general tendency for indexes to be larger for females than for males is concealed by the facts that entries under males include countries in which only data for the two sexes combined were available, and that those figures are included with males (as specified in the footnotes). This limitation is especially serious in the tabulation for university students. Selectivity indexes are highly sensitive to the distribution of per- centages for schooling categories among the base population. If 0.1% of the base population had any schooling beyond the secondary level, and if all students in university were offspring of fathers with at least some higher schooling, the selectivity index would be 1,000. However, if i0% of the base population had some post-secondary schooling the selectivity index could not exceed 10, other constraints aside. Such factors explain in part (not wholly) the contrast, for example, between a selectivity index of 1.47 for Japanese secondary school sons of fathers with post-secondary schooling and the index of 42.7 for Ghanian secondary school girls with such fathers (Appendix Table J.6). Notice; however that even though fathers of Ghanaian girls are drawm -from the same base population of adult males as the fathers of Ghanian boys, the selectivity is much lower for the boys than for the girls. Among secondary students the range in selectivity indexes on fathers with higher education is striking - from scarcely more than 1.00 in Puerto Rico to an estimated 63.00 for girls in Tanzania. The Tanzanian figures surely are incorrect, given sensitivity to a difference, for example, between an estimate of 0.1% and 0.3% of the base population as possessing higher education. There can be no doubt, however, that the true selectivity index for university students with university fathers must be extremely high in that country; despite efforts to spread opportunities more evenly, the establishment for - 54 - higher education is very small relative to population. Whether it is too small or sufficiently large in terms of needs for high-level manpower is quite another question. Charts 2.3 and 2.4 relate selectivity indexes for fathers of secondary and university students to percentages of the base population with at least some secondary schooling. The diagonal line on these charts defiDes in each case the maximum possible selectivity index, given the base-population percentage. (Horizontal lines mark off the constraint oD possible selectivity indexes associated with enrollment rates in secondary or in university schooliDg, respectively, depending upon which group of students is being looked at. I return to this in the next section.). There is a negative correlation between selectivity indexes with respect to parents with at least some secondary education and the magnitude of the percentage with secondary education among the base population. This association is somewhat closer for university than for secondary students. It is equally evident, however, that the selectivity indexes fall well below the diagonal defining the constraints of base-population percentages on those indexes, especially among male students in both secondary and higher insti- tutions; selectivity of female students comes closer to the maximum possible value given the distribution of schooling among man in the age cohort of their fathers. Essentially similar patterns (not shown) are revealed when the analysis refers to selectivity of children of college men, but with fewer cases close to the diagonals--a reflection of affects of the comparatively low rates of enrollment in higher education. Diversities at any given base-popula- tion percentage are most striking on the base "at least some secondary school- ing." There are many paths in development of educatioD and in its selectivity dimensions. - 55 - Chart 2.3 Selectivity Indexes by Base Populations with Secondary Education or More; Secondary Students Selectivity Index . . AMale * Female aAll 50 Maximum Possible Index at E - 5% 20 10 OMaximum at E - 10% A \ 0 5 nMaximum at E 20% 0 o 0 ° \ at C-= 50Y% 1 2 5 1 0 20 50 100 Percentage of Bass Population with Secondary Education or More - 56 - Chart 2.4 Selectivity Indexes by Base Populations with Secondary Education or More; University Students Selectivity Index 100 AFemale All 50 Maximum Posuible Index at E - 5% 20 01 o . Maximum at E 10% 10 A A Maximum at E -20% A Maximum 2 \ at E - 50% 1 2 5 10 20 50 100 Percentage of Base Population with Secondary Education or More - 57 - Since the percentage of a population with "no secondary schooling" is simply 100 minus the percentage with "at least some secondary schooling," it is to be expected that a relatively high selectivity index on one of these bases would be associated with a relatively low selectivity index on the other. So, in a very loose sense, it is, but because of variations in the relative sizes of the two base percentages these indexes are not mirror images of each other. Even when relative representation of youth from advantaged homes is comparatively high (a high selectivity index), the representation of children from less-advantaged homes may nevertheless be comparatively close to the norm of 1.00. With Chart 2.5 we turn back to examine relationships for secondary students between selectivity indexes for children of men with no schooling in relation to the proportion of such men in the base population. The most impressive feature o'i this chart is the scatter along the vertical scale whatever the base population percentages may be. If one divides the chart into quadrants at .50 on the horizontal scale and at .35 for the selectivity index, a rather substantial positive association appears. The poorest and educationally least advanced countries tend to have the higher relative representation of the children of illiterate men in their secondary schools. Considering that the three points that are deep in the southeast quadrant are all female, this association is the more impressive. Information on represen- tation of youth whose fathers are unschooled is too scant in the data on university students to justify a comparative charting. Furthermore, in advanced nations the tiny fractiouis of adults with no schooling are a negatively selected segment of the population whereas in the least developed countries they may be a majority and be much like everyone else. Chart 2.5 Selectivity. 'ndexes by Percentage of Male Base Population with Higher Education; Seconidary Students Selectivity Index 70 .e-.- 20 ._______________________ - - Maximum Possible Index at E = 5% -_____-____ 10 Pt Maximum Possible Index at E = 5 2 - * 10 IM - >laximum Possible Index at E =1096 7A _Maximum Possible Index 5 at E =20% A Me4s I | 3 0| Females A a All A 2 - ._ . - _-_--_._\E =_50_ X. L . . . . .I .....L.. .. I . . . ... ..I i . . .I I . x1 .15 .2 .3 .4 .5 .1 1 2 3 4 5 7 1t0 1E5 20 30 40 50 70 100 Percentage of Male Base Population with Higher Education - 59 - D., Enrollment Rates and Educational Selectivity We come now to a set of questions that may have direct and important policy implications. How important are rises in overall enrollmreat rates as a road to reduction in selectivity indexes at the top and to raising those for the children of the less-educated parents? How easily can these changes be made and under which circumstances? Even if a rise in enrollment rates seems to be the surest path to the convergence of selectivity indexes for different strata, the cost may be prohibitive and other ways to implement development may be judged to be more urgent; wisdom may call for toleration of wide dis- parittes among selectivity indexes for different segments of the population. How large is the variation of selectivity indexes even within the bounds of the constraints on maximum possible indexes arising from a combination of base-population percentages and enrollment rates? Which societies attain a relatively egalitarian opportunity situation over and above what would be reached simply by raising rates of enrollment among successive cohorts of children? A first step toward answering these questions is to identify the maximum possible selectivity indexes under the combined constraints of the relevant base-population percentages and the enrollment rate in secondary school (or university). Those maxima are easily defined algebraically, as is shown in Appendix C, which also includes a table illustrating the importance of these combined constraints. Define the following: max the maximum possible selectivity ratio for students S in education category j with fathers with schooling ij attaiament i. E the enrollment rate in education category j (all youth, j regardless of background) in percentages. - 60 - B1 the percentage of the base population with education i. It then follows (as is shown algebraically in Appendix C) that max S. 100/B when Bi > E i j/ i i j 100/E. when Bi < E The hori.zontal lines (in Charts 2.3 and 2.4) defining maximum possible selec- tivity indexes-call them Max S, dropping subscripts--were based on these simple equations. Enrollment rates can also define constraints on the lowest possible selectivity indexes; I come to that shortly. Enrollment rates are estimated with greater reliability for secondary than for higher education. Furthermore, enrollment rates in higher education are likely to exceed base population percentages only for individuals whose fathers have higher education. This means that with rare exceptions the E. constraint will be operative for university students on the selectivity index against fathers with higher edulcation only. Accordingly, for university students only the analysis against the higher-education base population is included (Table 2.6). Most of the discussion will relate to secondary pupils and to the analysis laid out in Tables 2.4 and 2.5. In each of those tables the first column repeats the selectivity index shown in Appendix Tables J-7 and J-8: the observed index S b . The second column shows the maximum possible selectivity index given the percent- age of the base population who possess the designated schooling: Max SBi. Column three specifies the limitation on S due to enrollment rates for the cohort of the study of secondary students (or in Table 2.6 the corresponding - 61 - TABLE 2.4 ANALYSIS OF SELECTIVITY INDEXES AND CONSTRAINTS ON THEM FOR FATERHS WITH HIGHER EDUCATION: SECONDARY STUDENTS Sobs-1 Max 5bs - Max S)- Sobs Max Sbc Max Sec Max S Max S-1 (Max Sbc - S obs) (1) (2) (3) (4) (5) (6) AFRICA Ghana 1961 All 18.67 333.33 34.48 34.48 .53 .95 Male 13.67 20.83 20.83 .64 .98 Female 42.67 90.91 90.91 .49 .83 C-hana Form 6 All 45.00 50.00 50.00 .88 .99 Kenya 1961 All 6.00 1000.00 27.78 27.78 .19 .98 Male 5.00 16.13 16.13 .26 .98 Female 11.00 55.56 55.56 .18 .95 Kenya 1968 Male 10.00 16.95 16.95 .63 .99 Tanzania 14.50 250.00 20.83 20.83 .68 .97 MEDITERRANEAN Tunisia 1963 Male 1.33 66.67 9.71 9.71 .04 .87 Female 12.67 " 47.62 47.62 .25 .20 Turkey 1962 All 13.05 52.63 14.29 14.29 .89 .97 Male 10.31 11.36 11.36 .90 .97 Female 18.00 21.74 21.74 .82 .89 SOUTHEAST ASIA W. Malaysia 1966 All 1.41 34.48 11.11 11.11 .04 .71 1972 Malays Male 6.86 a 71.43 10.00 10.00 .65 .95 Female 8.86 a 20.00 20.00 .41 .82 1972 Chinese male 2.75 a 22.73 8.33 8.33 .24 .72 Female 3.43 a 11.11 11.11 .24 .60 SOUTH AND CENTRAL AMERICA Argentina All 3.10 17.24 8.33 8.33 .29 .63 Bolivia 1970 All 6.32 16.95 6.67 6.67 .94 .97 1975 All 4.24 5.00 5.00 .81 .94 Brazil All 11.18 38.46 12.82 12.82 .86 .94 Male 6.92 9.52 9.52 .69 .92 Female 9.62 17.54 17.54 .52 .73 Chile All 1.78 20.41 5.88 5.88 .16 .78 Colombia All 3.14 19.61 7.14 7.14 .35 .76 Mexico All 21.36 45.45 23.81 23.81 .89 .90 Paraguay All 8.67 33.33 11.76 11.76 .71 .87 Peru All 4.55 18.18 5.88 5.88 .73 .90 Trinidad 1967 All 4.23 38.46 7.69 7.69 .48 .90 EUROPE, JAPAN, AND UNITED STATES Denmark 1964 Male 4.44 20.00 * * * * Female 4.36 * * * * Finland 1965 male 3.88 16.95 8.33 8.33 .39 .66 Female 2.73 5.88 5.88 .35 .78 Japan 1967 Male 1.47 10.63 1.54 1.54 .87 .99 Female 1.54 1.59 1.59 .92 .99 United States Male Chohort of 1898-1907 2.36 15.63 3.30 3.30 .59 .93 1908-17- 1.74 13.16 2.26 2.26 .59 .95 1918-27 1.65, 11.63 1.80 1.80 .8'. .98 1928-37 1.41 9.17 1.55 1.55 .75 .98 * Not available a) Completed secondary or more - 62 - TABLE 2.5 ANAYSIS OF SELECTIVITY ILIDEXES AND CONSTRAINTS OF THEM FOR FATHERS WITH ATLEAST SOME SECONDARY EDUCATION; SECONDARY STUDENTS Sobs-1 Max Sbs - Max S) Sobs Max Sbc Max Sec Max S Max S-1 (Max Sbc - S obs) (1) (2) (3) (4) (5) (6) AFRICA Ghana 1961 All 10.71 35S.71 34.48 34.48 .29 .05 Male 8.68 ". 20.83 20.83 .39 .55 Female 19.46 90.91 35.71 .53 .00 Ghana Form 6 All 14.04 50.00 35.71 .38 .00 Kenya 1961 All 4.78 111.11 27.78 27.78 .14 .78 Male 3.67 " 16.13 16.13 .18 .88 Senegal 1962 All 6.00 333.33 * 1967 Male 5.00 * 1967 Female 9.67 * Tanzania 1980 All 6.48 32.26 20.83 20.83 .28 .44 Uganda 1969 All 4.21 52.63 12.50 12.50 .28 .17 Zaire 1972 All 18.79 71.43 25.00 25.00 .74 .88 MEDITERRANEAN Tuniria 1963 Male 3.62 17.24 9.71 9.71 .30 .45 Female 10.17 47.62 17.24 .56 .00 Turkey 1962 All 6.09 9.80 14.29 9.80 .58 - .00 Male 5.41 11.36 9.80 .55 .00 Female .7.63 21.74 9.80 .75 .00 SOUTHEAST ASIA W. Malaysia 1966 All 1.62 6.06 11.11 6.06 .12 .00 1972 Malays Male 2.01 12.20 10.00 10.00 .11 .22 Female 3.28 " 20.00 12.00 .20 .00 1972 Chinese Male 1.21 4.37 8.33 4.37 .06 .00 Female 1.44 11.11 4.37 .13 .00 Papua New Guinea 1975 All 4.34 34.48 5.13 5.13 .81 .97 Male 3.98 4.22 4.22 .77 .96 Female 4.97 6.99 6.99 .66 .93 SOUTH AND CENTRAL AMERICA At&entina All 2.31 4.37 8.33 4.37 .39 .00 Bolivia 1970 All 5.18 7.62 6.67 6.67 .74 .39 1975 All 4.66 7.63 5.00 5.00 .92 .89 Brazil All 6.17 9.71 12.82 9.71 .59 .00 Male 5.82 9.52 9.52 .57 .05 Female 6.70 17.54 9.71 .65 .00 Cayenne Male 7L52 10.87 * Female 7.73 . * Chile All 1.45 3.15 5.88 3.15 .21 .00 Colombia All 2.38 4.48 7.14 4.48 .40 00 Guyana 1960 Male 2.28 8.13 * Female 3.67 * Mexico All 10.61 15.15 23.81 15.15 .68 .00 Netherlands Antilles 1960 Male 2.65 10.64 10.20 10.20 .18 .05 Female 3.26 22.22 10.64 .23 .00 Paraguay All 4.37 6.62 11.76 6.62 .60 .00 Peru All 3.11 4.78 5.88 4.78 .56 .00 Trinidad 1967 All 9.94 6.25 7.69 6.25 .37 .00 EUROPE, JAPAN, AND UNITED STATES Denmark 1964 Mlale 1.47 2.44 * * * * Female 1.42 * * * * Finland 1965 Male 1.66 4.29 8.33 4.29 .20 .00 Female 1.42 5.88 4.29 .13 .00 Japan 1966 Male 1.25 3.04 1.54 1.54 .46 .84 Female 1.30 1.59 1.59 .50 .83 * Note available - 63 - TABLE 2.6 ANALYSIS Of SELECTIVITY INDEXES AND CONSTRAINTS ON THEM FOR FATHERS WITH HIGHER EDUCATION; UNIVERSITY STUDIENTS Sobs-1 Max Sbs - Max S) Sobs Max Sbc Max Sec Max S Max S-1 (Max Sbc - S obs) (1) (2) (3) (4) (5) (6) AFRICA Kenya 1970 All 13.33 333.33 25.00 25.00 .51 .96 Sierra Leon Provincials 16.00 250.00 142.86 142.96 .11 .46 r Tanzaniat 1975 All 34.00 1000.00 250 250 .13 .78 male 29.00 1000.00 167 167 .17 .86 Female 63.00 1000.00 500 500 .12 .53 MEDITERRANEAN Portugal 1967 All 15.50 55.56 50.00 50.00 .30 .14 Tunisia 1963 Male 5.44 111.11 9.71 9.71 .51 .96 Female 21.00 " 47.62 47.62 .43 .70 Yugoslavia 1965 All 3.61 15.62 15.70 15.62 .18 0 SOUTHEAST ASIA W. Malaysia 1974 All 5.40 100.00 33.33 33.33 .14 .71 SOUTH AND CENTRAL AMERICA Chile All 5.32 24.39 14.29 14.29 .33 .53 Colombia All 6.60 20.00 14.29 14.28 .42 .43 Puerto Rico 1944 All 1.35 29.41 * * * * 1960 All 1.06 12.20 7.69 7.69 .01 .40 JAPAN AND NORTH AMERICA Japan 1967 Exams. Male 2.89 14.71 3.70 3.70 .70 .93 Japan 1968 All 4.10 4.41 4.41 .91 .97 United States 1920 All 3.67 13.07 * * * * United States Birth Cohorts 1898-1907 Male 3.14 15.-2 6.94 6.94 .79 .70 1908-17 Male 3.14 13.16 5.21 5.21 .51 .79 1981-27 Male 2.78 11.63 4.03 4.03 .54 .67 1928-37 Male 2.52 9.17 3.30 3.30 .ti6 .88 Canada 1975 All (under age 26) 1.35 3.61 2.20 2.20 .29 .64 - 64 - constraint for university students). Column four shows the maximum possible value of S given both of these constraints. Ma; S is the lower of the two entries in columns two and three. With these specifications set out, we can then make two interesting comparisons. Subtracting 1 from the observed selectivity index and from the maximum possible index as specified in column (4) gives us measures of the degree of over-representation; call these the "degree of selectivity." We can now compare the observed degree of selectivity with its maximum possible value under the Bi and E. constraints. The larger that ratio (column 5) the more nearly do the two constraints explain the observed selectivity index; the lower that ratio the more important is the modification of selectivity by other forces. These other forces may be informal or they may be built formally into the educational structure and into the rules for admission of students. A ratio of 1.00 would imply that there was a systematic ranking such that no child of a man with less than higher education could be found among the students until all children of such men had enrolled; no child of a man with less than secondary schooling would enroll until all children of men with secondary schooling had enrolled, and so on down the line. The higher the enrollment rates in the cohorts of young people, the further down this rank ordering it would be necessary to go, and so, taking enrollment rates as given, the lower must be Max S. In actuality, there is a wide variation in the ratios given in column (5) of Table 2.4, which analyzes selectivity indexes for fathers with higher education among secondary pupils. Those ratios are close to 1.00 for Turkey, for Bolivia, for one of the studies in Brazil, for Mexico, aad for Japan; in these cases the combined Bi and E. constraints dominate the observed selectivity. Distinctly low, by contrast, are the ratios for Tunisia (even - 65 - for females), for West Malaysia in 1966, for Kenya in 1961, for Chile in the early 1970's; in these cases there is substantially less selectivity than would result simply from the constraints of base populations and of enrollment rates on the indexes. One could speculate on what lies back of some of these variations. It may be more prudent, however, to refrain from entering upon so lengthly a task; specialists on individual countries may venture explanations. Notice that in Table 2.5 the enrollment rate E. is not always-the J determining constraint, as it is in Table 2.4; this is due to the fact that the relevant enrollment rates were for the level of secondary school whose pupils were studied, whereas the base population may include a lower cycle among those classified as "at least some secondary." Most of the ratios in column (5) of Table 2.5 are lorwer than in column (5) of Table 2.4 for this reason. The last column in each of these tables specifies the relative importance of the enrollment-rate constraint in explaining the deviation of observed from maximum possible selectivity indexes under the constraint of base-population percentages alone. The numerator of the ratios in this column is the difference between the constrained maximum based on Bi only (column 2) and the maximum taking enrollment rates into account as well (column 4). The larger that difference the bigger the effect of enrollment rate in reducing the upper bound on the selectivity index. The denominator is the difference between the maximum under the base-population constraint only and the observed selectivity index. Thus the ratio tells us what fraction of the difference between Max SB and the observed selectivity ratio may be explained by E (that is, by tde excess of E. over Bi). - 66 - Looking again at Table 2.4, the expansion of enrollment appears to explain most of the dampening of selectivity indexes with respect to children of men with higher education. The most striking entry in columi (6) is the low ratio of .20 for women (though not for men) in Tunisia; this deviates so much from others that it cannot be attributed simply to error. No doubt this is matter-of-fact to specialists on North Africa; it appears in work Bowman and I have been doing on education of girls and women. The likelihood that a Tunisian girl will be enrolled in secondary school is not neatly explained by parental schooling or social status. Other than for women in Tunisia, the lowest ratios are .60 or higher. That ratios for women in column (6) tend to be lower than those for men reflects the lag of enrollments for women behind those for men, sometimes dramatically so. It may be not too difficult to cut down the highest selectivity. indexes (for children of schooled parents) simply by expanding enrollments. Suppose that 1% of the base population of adults has at least some higher education and that enrollment rates in secondary school are 2% of the relevant cohort. Already, that 2% implies a maximum possible selectivity index of 50 instead of 100; if enrollment rises to 4% the maximum possible index drops to 25. This is still a high index, but quadrupling the percentage of the cohort attending secondary school has reduced the maximum possible index by 75 points compared with the maximum set by the base-population distribution of schooling. Each doubling of the enrollment rate cuts thes maximum index in half. But each doubling of the enrollment rate entails a bigger absolute rise in proportions who enroll in secondary school (or higher education, as the case may be). One cannot continue to cut selectivity indexes in half in this way although one can continue to reduce their maximum possible values 67 - absolutely. Assuming the expansion of enrollments is fiscally bearable, the reduction in selectivity ratios need not entail cutting back in the proportion of children from any status group who enter secondary (or higher) education. To reduce the degree of selectivity without expanding enrollment requires a more difficult adjustment whereby enrollment rates of children from better- schooled parents would actually decline. We can make a parallel analysis of representation from the less- advantaged homes. What sets the lower bounds on selectivity indexes for children of men with little or no schooling? Could the minimum possible representation from a particular set of fathers be above zero? Again one must look at enrollment rates, for there is nothing about the base-population percentages as such that would preclude zero representation among pupils. An index above zero will be inevitable if the sum of the percentage for the base population (ejg., for men with no secondary schooling) plus the enrollment rate exceeds 100%. (For a formal analysis see Appendix C.) This situation does not often occur for fathers who have no school- ing because in the extreme situation of large illiterate majorities among adults the secondary-school enrollment rates of the young are likely to be low (Table 2.7). The only case for which the minimum possible selectivity index of those with no schooling exceeded zero was Papau New Guinea, where the spread of schooling in recent years has been phenomenal. Bolivia (1975) and Tunisia (1963) for males give approximately the same minimum values, but through a different combination. In Tunisia 94.2% of the base population had no secondary schooling but 10.3% of young men were enrolled in secondary school; these percentages sum to 104.5%. With an enrollment rate of 10.3% at least 4.5 % of those enrolled in secondary school must have parents who had not gone - 68- Page 1 of 2 TABLE 2.7 ANALYSIS OF REPRESENTATION OF THE RELATIVELY UNEDUCATED AMONG FATHERS OF SECONDARY STUDENTS Base Population Min S Min S Actual S.I. Percentage No No No No No No Enrollment Schooling Secondary Schooling Secondary Schooling Secondary Rate AFRICA Ghana 1961 All 0 .03 .37 .72 '78.8 97.2 2.9 Male 0 .40 .43 .78 78.8 97.2 4.8 Female 0 0 .11 .47 78.8 97.2 1.1 Ghana Fam 6 All 0 0 .30 .36 79.0 97.5 2 Kenya 1961 All 0 .76 .41 .97 72.7 99.1 3.6 Male 0 .86 .47 .96 72.7 99.1 6.2 Female 0 .45 .08 .91 72.7 99.1 1.8 Kenya 1968 Male 0 .82 .49 .93 70.7 98.9 5.9 Tanzania 1980 0 .02 .26 .51 51.6 96.9 4.8 Uganda 1969 All 0 .78 .56 .93 55.1 98.1 8 MEDITERRANEAN Tunisia 1963 Male 0 .46 .68 .84 80.4 94.2 10.0 Female 0 0 .17 .44 80.4 94.2 2.1 Turkey 1962 All 0 0 .24 .42 38.8 89.8 3.6 Male 0 0 .32 .50 38.8 89.8 5.7 Female 0 0 .05 .25 38.8 89.8 1.4 SOUTHEAST ASIA W. Malaysia 0 0 .33 .88 24.7 83.5 8 1972 Malays Male 0 .20 .27 .91 33.3 91.8 10 Female 0 0 .21 .80 33.3 91.8 5 1972 Chinese Male 0 0 .53 .94 18.4 77.1 12 Female 0 0 .34 .87 18.4 77.1 9 Papua 1975 All .17 .88 .51 .90 83.8 97.1 19.5 Male .33 .90 .59 .93 83.8 97.1 13.7 Female 0 .82 .46 .88 83.8 97.1 14.3 SOUTH AND CENTRAL AMERICA Argentina 0 0 * .61 4.7 77.1 12 Bolivia 1970 All 0 .15 .05 .37 34.9 86.9 10.5 1975 All 0 .40 .11 .45 34.9 86.9 16.7 Cayenne 1959 Male 0 * * .34 * 90.8 * Female 0 * * .32 * 90.8 * Chile All 0 0 .36 .79 10.7 68.3 17 Colombia All 0 0 * .60 18.5 77.7 14 Guyana Male 0 * * .82 8.8 . 87.7 1.7 Female 0 * * .62 8.8 87.7 7 Mexico All 0 0 * .32 40.7 93.9 4.2 Page 2 of 2 - 69 - Base Population Min S Min S Actual S.I. Percentage No No No No No No Enrollment Schooling Secondary Schooling' Secondar Schoo.ing Secondary Rate SOUTH AND CENTRAL AMERICA (cont'd.) Neth. Antilles Male 0 .07 * .83 13.0 90.6 Female 0 0 * .77 13.0 90.6 5.8 Paraguay All 0 0 * .40 13.3 84.9 Peru All 0 0 * .44 18.7 79.1 17 Trinidad All 0 0 * .63 4.9 84.0 13 EUROPE, JAPAN AND NIITED STATES Finland 1965 Male 0 0 * .80 * 76.7 .'2 Female 0 0 * .37 * 76.7 17 Japan 1966 Male 0 .74 * .88 * 67.1 65 Female 0 .71 * .85 * 67.1 63 U.S. Birth Cohorts 1898-1907 Male 0 0 * .59 a/ * 28.7 a/ 30.3 1908-17 Male 0 0 * .70 a/ * 32.4 a/ 74.2 1918-27 Male 0 0 * .73 a! * 37.2 a/ 55.6 1928-37 Male 0 0 * .68 a! * 22.8 a! 64.7 a/ Less than Elementary 8. * Min S is the mathematically lowest possible selectivity index for a given Bj (base percent) and Ej (enrollment rate). - 70 - beyond elementary school--even if all sons of men who had done so (5.8% of the total population) were indeed in secondary school. This means that at least 4.5/10.3 or 43.7% of the youth enrolled in secondary school must have parents with no secondary schooling. The ratio 43.7/94.2 is the selectivity index; the minimum possible index is .46. A similar computation gives a minimum for Bolivia of .40; in this case there is a somewhat lower proportion of the base population with less than secondary schooling but twice as large a pro-portion of the yout.h cohort enrolled in seconclary school. A quite different situation? has prevailed in the South and Central American countries included. There change has been slower and the spread of secondary schooling among youth has been less extensive than the spread of at least a few years of primary schooling in the generation of parents. Con- sequently, most of the entries in column (2) are zero. The combination of enrollment rates in secondary schools and the proportions of the base popula- tion without any secondary schooling does not set mathematical bounds on the minimum possible value of the selectivity index. Japan of 1966 represents the opposite situation, with high secondary enrollments. Even at that date almost two-thirds of the youth (of both sexes) began secondary school and remained until the senior year, although two-thirds of the generation of fathers had not gone beyond the compulsory levels. For males this gives 67.1 + 65,0 - 132.1. Even if all sons of men who had at least some secondary schooling were now enrolled in secondary school, they could account for only 32.9/65.0 or 50.6 percent of those students. The other 49.4 percent (32.1l/,r'.0) of male secondary pupils must then have fathers who had no secondary schooling. But that 49.4 percent divided by the base population of 67.1 percent implies a minimum possible selectivity index of 49.4/67.1, or .74. - 71 - Some of the ratios in columa (2) are surprisingly high. They indicate that where a low-level base population is large. the selectivity index for that secondary-school population may be pushed up very substantially by expansion of enrollments. This is not the case if the low-level base population is comparatively small; see the figures on the four birth cohorts for the United States at the bottom of Table 2.7. In this latter case the selectivity indexes refer to fathers with less than eight years of elementary schooling, and selectivity indexes were reasonably high (around .70) but the minimum possible indexes were nevertheless still zero. Looking again at LDCs, even if the mathematically determined minimum index remairns zero, there may be a tendency for relative representation of the less-schooled to be greater the larger their percentage of the base population and the higher the overall enrollment rate. There are two reasons in addition to any mathematical constraints. First, the larger is the proportion of unschooled adults, the less will they be differentiated (in ability, culture, etc.) from the rest of the population. The pool of youth with unschooled fathers out of which able and ambitious individuals may emerge is relative large. Second, the larger the enrollment rate (or the "places" made available) in secondary schools, the greater the chance for energetic young people from humble homes to secure one of those places. In addition, pressure from the populace can itself bring about more secondary places. This populist movement has been a common experience in the third wc ld, even where the pressures of so-called "social" (really individual or private) demand are regarded as a major and political "problem". It is not accidental that Chart 2.5 showed higher selectivity indexes (closer approxi- mations to proportional representation) for children of unschooled men - 72 - when those unschooled made up more than half of the base population than where their proportion was smaller. This does not mean that secondary-school enrollment rates for sons or daughters of men who are unschooled will be high; their enrollment rates will remain low so long as the overall enrollment rate of the cohort is low. E. The Economy and Student Selectivity by Parental Education One of the questions that gave rise to this research was whether levels of economic development had systematic effects on the democratization of secondary and higher education. One might ask why level of income in a country should affect the social selectivity of its schools. For economically advanced countries it might be more appropriate to ask whether selectivity of schools affects level of income, not to mention affecting political behavior. National traditions and policIes about attendance at schools and about rules and practices for admission or continuation in schools are strongly rooted and might be expected to persist despite many changes in the economy. Much of what has been said in this chapter is relevant to these broad questions. It has become clear that any association between economic development and selectivity of schooling will depend upon the spread of education through a society as indicated by the distribution of schooling among the parental generation and by the present-day enrollmen.t levels. Here I discuss: 1) the relationship between degree of under-representation of unschooled fathers and the proportion of the labor force in agriculture; 2) effects of per capita incomes and enrollment rates on indexes of dis- similarity; 3) components of the correlation between per capita income and the proportions of students' fathers having given amounts of schooling; and 4) effects of per capita income and enrollment rates on selectivity - 73 - measured by differences of logits between paternal and base-population proportions. - 1. Proportions in agriculture and degree of under-representation of unschooled fathers. The proportion of t labor force in agriculture is a simple index of economic structure. It has a moderate negative association with per capita incomes and also it is an index of rurality; rurality is associated with inaccessibility of schools and with weak exposure to urban enhancement of demands for schooling. For university students there was virtually a zero relationship between countries' proportions of farmers and the indexes of dissimilarity. Among secondary students there was no systematic relationship, but in the three countries with less than 15% in agriculture, the indexes of dissimi- larity for both sexes were under 20%--half the average index. There was a distinct positive association between selectivity indexes for representation of unschooled fathers and proportions working in agriculture. Since these indexes all are under unity, this means less under- representation of unschooled men's children in the rural than in the urban societies. 2. Effects of per capita income and enrollment rates on indexes of dissimilarity by father's schooling. Is the level of national income and extent of social selectivity among students greatly mediated by enrollment rates? Do the direct effects of income and those via enrollment rates operate in the same or opposite direction? Zero-order correlations between income and indexes of dissimilarity approximate the "total" effects; between log of per capita income (LNY) and index of dissimilarity (IDX) we find for secondary students the negligible negative correlation--127, but for university students -.420. - 74 - A path model is used to analyze income and enrollment effects on indexes of dissimilarity. (While I would not wish to emphasize a causal interpretation, a recurrent model is a useful way of sorting out the relation- ships of most interest here.*) Table 2.8 displays the path data (controlling for sex) and separates direct effects of income from indirect effects via associations between income and enrollment rates. The path from income (LNY) to enrollment rate (ET) is strongly positive; the standardized beta coefficients are .698 for secondary and .896 for university students. Among secondary students (the countries being heavily weighted with LDC and especially African countries) ET has a strong direct negative effect on the index of dissimilarity. The result is a stronger negative effect of income among university than among secondary students. Summing up, higher levels of per capita income are associated with lesser selectivity (for paternal schooling), but for secondary students this is attributable entirely to association of level of income with enrollment rates. For secondary and university students alike, there is great diversity in the level of social selectivity at any given level of income. Enrollment rates, income, and sex combined leave 70% of the variance in the indexes of dissimiliarity unexplained. * Present enrollment rates of the students cannot "explain" present levels of income. Serial correlations of present enrollments with the schooling of the whole earlier generation could introduce a spurious "causal" component, but for our set of countries these relationships are loose. TABLE 2. 8 EFFECTS OF PER CAPITA INCOME AND ENROLLMIENT RATES ON INDEXES OF OISSIILARMY BY FATHER' S EDUCATION Secondary Students Universicy Studencs .698 ET (Enrollment rate) .896 ET LNY \ . 641 LNY . 153 (lIn. of4 per capita .215 IfX - .286 DX income) (Index of di3simi1ariry) N 33 37 Mean TIfX 37.2% 41.7% 2.279 .3096 P (.0163) (.0940) Effects of LTY Direct .215 -.286 indirect via ET -.447 .137 Total -.232 -.423 All equations control for sex. - 76 - 3. Components of correlates between per capita income and proportions of students' fathers having given amounts of schooling. In turning to select- ivity from particular educational backgrounds, one must first ask what determines the dis'tribution of,schcoiinE among students' fathers. The propor- tion of the base population of adult males with a particular schooling must be of definite effect. The appropriate scaling for these proportions requires use of logits (or alternatmively probits) of the base-population proportions (LB) and logits of proportions among students' fathers (LF). A linear regres- sion using proportlGtI.. would often give absurd results, predicting over 100% or less than zero. (The argument in developmental terms for use of logits is set out in Chapters 5 and 6.*) Today's per capita income hardly causes the base population to possess particular amounts of schooling; causation would run the opposite way, from educationlal attainments of adults to high (or low) per capital income. Interest here, however, is in the interpretation of associations between per capital national income and the distribution of schooling among faithers of secondary and university students. A recurrent model is useful for this purpose regardless of effects of education on growth. Again there are controls for sex, but regressions are not run separately for males and females. Three models are tested for secondary and for university students. These are: 1) LNY and LB only; 2) LNY and LET (logit of enrollment rate) only; and 3) LNY, LB, and LET. Table 2.9 di'splays these models with respect to proportion of fathers or of base population having "at least some secondary schooling" (LPS and LBS). Since logits are symmetrical around zero at 50 percent, the * To modify sensitivity of logits to errors and sample biases at the extremes, the formula used in all such regressions is ln P+ .01 instead of simply ln - 77 - results shown in Table 2.9 can be converted into an analysis for fathers with no secondary education simply by appropriate changes of sign in the arrows reaching to LFS from LNY and LET and to LBS from LNY. Further details for the data underlying Table 2.9 and for fathers having higher education and those having no schooling are shown in Table 2.10. Model I makes it clear that simply introducing LBS (logit of the proportion of the base population with at least some secondary schooling) absorbs most of any effects of per capita income on the proportions of students' fathers with at least some secondary schooling-among both secondary and university students. It is predominantly through its association with LBS that per capita income affects LFS. When we examine the proportion of students' fathers with higher education, however, the direct effects of per capita income match or exceed the indirect effects via base-population associations. Direct effects for secondary students (at .324) account for 60% of the .525 "total effect" of income on LFH. Among university students the direct income effect on "father with higher education" is larger (.610), accounting for three-fourths of the .841 total income effect on LFH. The pattern with respect to fathers lacking any schooiing is shown in the third and sixth columns of Table 2.10. Among secondary students it is the base-population of unschooled that accounts for the modest negative association between per capita income and proportions of unschooled students' fathers. Among university students, on the other hand, base population proportion (LBO) exercises little influence, with direct income effects the main component of the strongly negative (-.622) total income effect. - 78 - *TABLE 2.9 SJIMARY OF EFFECTS OF PER CAPITA INCOMES ON PROPORTIONS OF FATV> RS WITH AT LEA,.T SMA qFrnA1y F!'ITCATION Se.er4awiW ;wt St ue At st s2 p LET LET rLS T - StT,- LN_ LW' ...JL - 3 Effects of LNY (1) Direct -.045 .122 Indirect via LBS .621 .662 Total .576 .784 (2) Direct .861 .227 Indirect via LET - .285 557 Total .576 .784 (3) Direct .138 - .263 Indirect via LBS .936 .615 Indirect via LET - .499 .430 Total .575 .783 TABLE 2.10 EFFECTS OF PER CAP.TA INCOME, ENROLLMENT RATES, AND BASE POPULATIONS ON PROPORTIONS OF FATIIERS WITH SELECTED EDUCATIONAL ATTAINMENTS Secondary Students Univeirsty Students Fathers with Fathers with Fathers Fathers with Faihers with Fathers IIigher at least some with no Higher at least some with no Education Secondary Schooling Education Secondary Schooling (LFN) (LFS) (LFO) (LFH) (LFS) (LFO) Hodel I N 26 34 22 17 19 13 LB (Logit of base Mean of LF -1.769 -.491 -1.332 -1.869 -.271 -1.648 population proportion) It2 .3319 .4624 .4601 .7569 .7528 .6836 Prob F (.0201) (.0001) (.0063) (.0001) (.0001) (.0097) Path coefficients Involving LB LN LF LNY to LB .6q .859 -.462 .925 .862 -.761 (Per capita (Logit of father LB to LF J3. .723 .728 .250 .768 .206 Income) proportion) Effects of LNY Direct .324 -.045 .065 .610 .122 -.465' Control for sex via association with LB .201 .621 -.336 .231 .662 -.157 Total .525 .576 -.271 .841 .784 -.622 Model 2 N 26 34 22 17 19 13 LET (Logit of enrollment Mean of LF -1.769 -.491 -1.532 -1.869 -.271 -1.648 rate) g2 .3405 .4097 .0839 .7551 .6839 .6781 Prob F (.0173) (.0004) (.7020) (.0002) (.0002) (.0105) Path coefficients involving LET LN LF LNY to LET .677 .687 .465 .917 .914 .782 (per capita (Logit of father LET to LF -.348 -.415 -.002 .243 .609 -.179 income) proportion) Effecta of LNY Direct .760 .861 -.270 .619 .227 -.482 Control for sex Indirect via LET -.236 -.285 -.001 .223 .557 -.140 Total .524 .576 -.271 .842 .784 -.622 Model 3 N 26 34 22 17 19 13 LET Mean of LF -1.76'9 -.491 -1.532 -1.869 -.271 '-1.648 .R2 5358 .4619 .5025 .7591 .7816 .6878 Prob F (.0013) (.0001) (.0104) (.0008) (.0001) (.0318) Path coefficients involving LB and LET LNY _ F LNY to LA .663 .859 -.462 .925 .862 -.761 LNY to LET .677 .687 .465 .917 .914 .782 LB to LF .730 1.090 .811 .187 .714 .164 LET to LF -.792 -.726 .353 .150 .471 -.112 LB Effects of LNY Direct .576 .138 -.061 .531 -.263 -.410 Indirect via LET -.536 -.499 .164 .138 .430 -.088 Control for sex Via association with LB .484 .936 -.375 .173 .615 -.125 Total .524 .575 -.272 Q42 .783 -.623 - 80 - Model II drops base-population proportion but introduces the logit of enrollment rates (LET). By definition overall enrollment rates are not specific to any particular family background. The variation in the path coefficients LNY to LET for secondary students (across the first three columns of Table 2.10) reflects the mixture of countries on which information was available, and similarly for variations in coefficients on LNY or LET across the university columns. For secondary students, Model II displays significant but small direct negative effects of enrollment rates (controlling for income) on proportions of students' fathers with higher and with at least some secondary schooling. The positive direct effects from income are substantially larger, however, and overall these direct effects predominate in the "total effect" of per capita income. Enrollment rates have no direct effect on proportions of students' fathers with no schooling, and the negative direct effects of income on LFO are significant but small. For university students there is both a strong effect of LNY on LET and a strong positive effect of LET on LFS. The net result is a large positive indirect effect of income via LET on the proportion of students' fathers having at least some secondary education. LET in Model II (like LBS in Model I) accounts for most of the "total effect" of per capita income on proportion of fathers with at least some secondary schooling. The direct effect of LNY predominates (as in Model I) in explaining the proportion of university students whose fathers had higher education. Predominance of the direct income effect also shows up for university students with unschooled fathers, much as it did in Model I. Model III brings together a fuller interactive system. For secondary students LET carries on a strong direct negative coefficient on LFS, as is - 81 - easily seen in Table 2.9. (Turning this around, to its mirror image, the negative coefficient of LET on LFS says that given controls for income and base-population distributions of schooling, increasing enrollment raises substantially the proportions of secondary students whose fathers lacked any secondary schooling.) The negative sign on LET gives a large substantial negative indirect effect of LNY via LET on LFS, which partially neutralizes the small positive direct income effect and the strong positive indirect effect via association of LNY with LBS. For fathers having at least some secondary schooling, the pattern among university students seems very different-as it did also in Model II. LET continues to have a positive effect on proportions of university students' fathers with at least some secondary schooling, even controlling for LBS as well as for per capita income. (Turning the interpretation around, increasing university enrollment ratets reduces the proportion of university youth whose fathers had not gone beyond elementary school--a perverse effect.) Us:ing Model III with both LET and LBS in the equation, the direct effect of LNY becomes slightly negative, but the combination of strong indirect positive effects of income via both LET and associations with LBS dominates the results giving a strong positive total effect of income on LFS. Given the multicol- linearity embodied in equation III, it would be a mistake to put too much stress on the coefficients displayed, but the equations analyzing selectivity, below, support the foregoing evidence. In any case, it is clear enough that' base-populati6n traits have systematic effects on proportions of students' fathers possessing at least some secorndary schooling. It is equally clear that after controlling for base-population proportions, variations in propor- tions of students' fathers with given schooling remain substantial. - 82 - The patterns for procortions of fathers having secondary schooling occur in modified form for fathers with higher education. For secondary students the effects on LFH via LET remain strongly negative, whereas among university students these effects are weakly positive. Effects of per capita income via its association with the base LBH are moderate for secondary students and negligible for university students. For proport.Zons of fathers having no schooling, patterns for secondary and for university students are again different. For secondary students the moderate negative total effects of income on proportions of fathers with no schooling (LFO) operate mainly through the negative associa- tion of income with LBO (proportion of base population unschooled). By contrast, for university students that base-population proportion (LBO) is of little importance though the total nelgative effect of income on FLO is sub- stantial; of the -.623 total effect, two-thirds is the direct negative effect of income and a fifth is indirect effects via LET. 4. Logit analysis of selectivity related to income and to enrollment rates. Much attention has been given in preceding paragraphs to disentangling the determinants of the distribution of schooling among parents of students. It is now time to trace out the associations among per capita income, enroll- ment rates, and disparities in the representation among students of the various base-populations. The indicator of selectivity used is the difference between the logit of a father'es proportion and the logit of the associated base-population: designat:ed "LF-B" referring to LFH-LBH, LFS-LBS, and LFO-LBO for the three categories of base and parental schooling. Simple equations are used to ask: to what extent does per capita income explain selectivity, and what part of that influence operates through effects of income on enrollment rates? (See Table 2.11.) -83- TA R 1. 2 .211 EECT'S OF PERt CAPITA INJCUME tNl) AN i F (.0032) (.0001) (.0317) Path coefficients invoiving LET: LNY to LET .b73 .bd7 .465 LET to LF-B -.d99 -.o23 .514 Effects of LNY Direct .631 .084 .012 Indirett (via LET) -.609 -.565 .239 Total .022 -.481 .251 UNIVERSITY STUDENTS Number of observactions 17 19 13 MSan LF-B 1.405 2.024 -2.165 R- .3133 .5261 .3631 Prob > F (.1576) (.0060) (.2311) Path coefficients involving LET LN4Y t o LET . 9 L 7 .9 l4 .782 LET' t( LF-B .027 .732 .121 ' fects of L'Y ir:ect -i160 -1.157 .137 Tnd irect .025 .669 .095 Total -.135 -.488 .232 T'he effects on LF-B shown in the second column are cmirrer i.mages of effects on selectivity (underrepresentation) of chLldren of men with no secondary schooling; tits this column may be read to refer to the latter simply bv revecsing, tiie signs - 84 - The signs on the coefficients must be read wicn care. Wherever the proportion of fathers typically exceeds that of the corresponding base popula- tion, any factor that raises the difference LF-B increases relative over- representation-as in the first two columns of Table 2.11. In the third column the proportion of fathers falls short of the proportion in the corre- sponding base-population; anything that raises that difference algebraically brings it closer to zero and thus closer to proportional representation. The patterns for secondary students are unambiguous. For fathers with higher education the direct effects of national per capita income are strongly to increase over-representation, but the indirect negative effect of income via LET is almost as strong, giving a total effect of income that is virtually zero. For fathers with at least some secondary schooling direct income effects are positive but negligible whereas indirect effects via LET are strongly negative and hence equalizing; this equalizing effect dominates in the negative total effect of -.481. For fathers with no schooling there is again little direct effect of income on selectivity (in this case, degree of under-representation) but enrollment effects clearly are positive, and they account for most of the total income effect, raising representation of youth from homes with unschooled fathers in the countries in which proportions of unschooled adults were large enough to be recorded; this cuts out countries with long-established mass schooling. The patterns for university students (shown at the bottom of Table 2.11) are quite different. For each set of fathers, income is closely and positively related to LET, but enrollment rates have negligible di,.dct effects on selectivity except for fathers in the broad category "at least some secondary schooling". Putting this the other way, the' income effect via LET - 85 - on representation of fathers lacking any secondary schooling is negative; the higher are enrollment rates, the lower is. the relative share (for a given per capita income) among university students of youth from homes in which fathers had no more than primary schooling. Among university students there is, however, an exceptionally strong negative coefficient of -1.157 for the direct effects of per capita income on proportions of fathers with at least some secondary education. This more than neutralizes the positive indirect income effect via LET to give a total income effect that is unambiguously negative and hence equalizing. Neither per capita income nor enrollment rates, nor interactions between them, significantly affects selectivity of university students from homes with the most-schooled or the least-schooled fathers. At best, results are not perverse in equity terms. In the most general terms, there are wide variations in degree of inequality of representation in both secondary and higher education within levels of per capita income, even where per capita income seems to be most predictive. Processes of development appear to operate somewhat differently at the secondary and at the higher-education levels in step with the stages of educational expansion represented by the available samples of countries. - 86 - CHAPTER 3 FATHER'S OCUPATION AND SELECTIVITY OF ENROLLMENTS IN SECONDARY SCHOOLS AND UNIVERSITIES Paternal schooling and paternal occupation are the most widely used indexes of family background for delineating access to educational opportu- nities. By scanning the list of sources of data one can see that on secondary pupils information about paternal schooling was available for 27 countries and about paternal occupation for 52 countries; the corresponding numbers for students in higher education were 19 and 86. The distinctly large number of V'advari,,d" countries about which there are data on occupations. of fathers of university students affected the percentages of fathers from different occupa- tions. There are regional differences also as to countries from which data are available; see the totals in Tables 2.1 and 3.1 and 3.2. The present chapter is organized similarly to chapter 2. In this empirical investigation, there is little point in reviewing the conceptual lilterature about the merits of various indexes of social status or about the justification for using composite indexes of status. One must use the available data. Occupation is associated not only with income and schooling but also with a range of opportunities for learning outside the schools and for accupational inheritance or mobility through informal learning. Farming is a prime example. (On the other hand, handicaps commonly associated with farm residence for a child are compounded by the pronounced spatial isolation of farm families. Over much of the world use of boarding schools or pensions mitigates this isolation.) - 87 - Traits associated with participation in a particular occupation vary considerably despite a broad similarity in occupational structures among societies at any given "stage" of economic development. Where secondary schools are "exclusive" and almost everywhere in the world of higher edu- cation, the opportunities for schooling associated with paternal occupation (or strata) can be very particular. (See the abundant historical material on this topic in Ringer.) Such configurations partly explain the great diversity in the degree of social selection for higher education to be observed among even the comparatively similar "western" countries. Though exploration of sex differences in selectivity for schooling is treated mainly in another report, readers will easily observe a parallelism with observations made in the chapter on parental schooling: e.g., the pre- vailing greater degree of social selectivity among girls. No analysis is made of the scanty data on maternal occupation for elementary schools (sum- marized in Appendix F). A. The Data and Occupational Classifications However difficult it is to establish a satisfactory classification of occupations for any particular society, the difficulty is multiplied when one tries to devise a classification that will be usable across many societies. It is easier to have comparability over time in a given country than between countries. Studies of educational selectivity have made little use of International Standard Classification of Occupations. Plagued by small samples and led astray by notions about an "underlying status structure," the compilers of data have served us poorly in many cases. The seemingly simple categorization of "white collar, manual. farming" does not tell us where "services" were placed nor whether farm labor was put with farmers - 88 - small samples and led astray by notions about an "underlying status structure," the compilers of data have served us poorly in many cases. The seemingly simple categorization of "white collar, manual, farming" does not tell us where "services" were placed nor whether farm labor was put with farmers or with manual labor; these decisions vary from one country to another and often the data are reported in such broad rubrics that one cannot reallocate individuals. The classifications of students' fathers often are made with only casual regard to the country's census categories for occupations. This problem of suitable categories of paternal occupations arises also because it is desirable to find a "base population" that can be compared with fathers of students. In most instances four categories of occupation could be identified: (1) farmers (including peasants), (2) manual workers (with the usual ambiguity about artisan proprietors), (3) traders and small proprietors, and (4) white collar (also a heterogeneous category). Inevitably the work of classification for this report involved a certain circularity; for example, where "service" workers displayed patterns of educational selectivity similar to those among manual workers the former were called "manual," or conversely white collar. Where this criterion was unhelpful, service workers were put with "miscellaneous." The distributions of paternal occupations and of males in the base population (where available), along with selectivity measures, are given in Appendix J country by country. Specifying the base population was an important but difficult task. Care in identifying the occupational distribution among the generation of potential parents has been rare in studies of educational selectivity. Usually an approximate distribution was given by an author; however, -uomen sometimes were included and usually the total labor force of males was used - 89 - rather than that part of it in the appropriate ages. (There was no composite compilation as useful as the UN1'SCO distribution of schooling for adults by age.) The age of the base population has differing importance for separate categories of occupations. Managers and proprietors are more likely to be in the age range of fathers of secondary or university students than are clerical workers; "proprietors and traders" are a poor proxy for the former group. In countries where the structure of the economy has not changed greatly during recent decades there will be little mobility out of manual into white-collar jobs, and "white-collar" may be an acceptable rubric. The situation is similar for farmers but not for the fluctuating and youthful set of farm laborers. In most countries the occupational structure has altered greatly during the last generation. When one could compare occupations for the whole male labor force with those of men in the age cohort of potential fathers of pupils or students, the distortion for farmers is seen to be large; France and especially Japan are examples. When farming hat, become a less populous occupation, selectivity indexes become biased upward if no age adjustment is made in the base popula- tion. For other sectors the effect is a downward bias in the indexes. For white-collar men in the advanced countries the bias is slight. In much of the third world there has-been rapid expansion of white-collar and of manual- wage workers; in those countries the upward bias in the selectivity indexes for farmers will be negligible, but there may be a pronounced downward bias in the selectivity indexes for white collar men. Problems in interpreting the indexes of dissimilarity arise more from crude classifications of occupations than from undifferentiated age data for the base population. - 90 - B. Profiles of Occupations of Fathers and of Base Populations Frequency distributions of the occupations of fathers are tallied by world regions in Tables 3.1 and 3.2. (It was possible to include more entries in those tables than the details of Appendix Tables K.1 and K.-2 might suggest since assignment to a Percentage category was often possible even where one lacks preis&e percentages of fathers.) For university students we have a fair representation of Eastern Europe (including some data for years before the revolutions). As for parental education but: to a greater degree, data for university students cover more economically advanced countries than do data for secondary students. Later in this chapter an estimate will be made (as was done earlier for parental schooling) of how much apparent selec- tivity of students has been affected by the underlying structure of occupations. Looking at Table 3.1 for secondary students, one infers that economic structures contribute to the comparatively large proportions of African students who come from farms, whereas farmers' offspring are comparatively fewer in Latin America or "Europe." Manual workers, by contrast, are fewer among fathers of African pupils as they are in Africa generally; this group is numerous among students in much of Latin America (as among the base populations in that region). The proportion of manual workers' children in the secondary schools is diverse even among western societies. "White-collar" fathers are relatively numerous among fathers of students everywhere; in this instance representation exemplifies "selectivity" and runs counter to the modest size of this group in most of the base populations. The percentages for farmers and for manual workers tend to be smaller among university than among secondary students; conversely, offspring - 91 - TABLE 3.1 FREQUENCT DISTRIBUTIONS OF OCCUPATIONS OF FATHERS: SECONDARY STUDENTS South and Japan, Ocu inadce'ntral Eastern W. Europe. Percentage and Category Africa Mediterranean Asia America Europe S erica hi Total Far1ers 1 1 2 1 2 - 11 2 3 10-19 3 1 3 1 1 6 1 1 3a/ 5 1 1 21 : 20 - 29 2 - 1 - 1 - - 3 2 2 30-39 - 91 1 - ----- -- - - -- 1 3 2 40 - 49 1 - 1 - - - - --1----- -- - - -- 1 - 2 50 and over 2 3 - - - - L 1 ------ -- - - -- 1-4 Total 8 5 5 3 - - 5 4 4 a 3 3 3 - - 13 2 1 4'1 1- 13 Manual Workers 10 4 1 1 1 1 1 1 1 1. 1 /- - 3 - - :0 3 3 10- 19 5 3 3 2 0 0 0- 0 O 0 0 O 00 0 3 1 1 19 4 4 20 - 29 - - 1 1 - - 3 3 3 1 - - - - - - - 9 3 4 30 - 39 1 1 -- - -- - - - - 3 2 2 - - - 2 1 - 6 4 2 40-49----------- 4 3 2 1 - - 3 1 1 8 4 3 50 - 59 - Total 10 5 5 3 - - 4 4 4 9 6 5 3 - - 7 3 2 53 18 6 Proprietors and Tr0ders 5 5 5 - - - 1 1 1 3 3 3 1 - - 7 - - 1'7 9 9 10-19 4 5 5 1 - - - - 4 2 - - 7 - - 16 7 7 20 - 29 - - - 2 - 1- - - 1 1 4 t 30 - 39-- ------ -2 1 1-----------2 l 40 - 49 Total 9 10 10 3 - - 4 2 2 8 7 7 1 - - 12 - - 40 19 19 White Collar (excluding proprietore and trade) 10 -19 4 2 1-----------------4 2 1 20-29 - 1 1 1 5 1 1 - 6 3 1 30 -39 1 2 2 2 2 5 2 2 1 - 10 6 4 40 -49 1 - 2 1 4 2 2 4 1? 2 3 50 -59 - - 2 1-------- - 1 1 - - 3 60 -69 2 - -------- 2 - - 4 - 1. 70 -79 -----2 -- - - Tta8l 8 5 5 3 - - 2 3 3 14 5 5 2 - - 15 - - 44 13 13 20T-o29 2 - - - - - - - - - - 22 2 30 -39 3 1 - - - - 1- - 2 2 - - 6 4 40 - 49 3 1 1 2 2 1 - 4 1 10 4 3 50-59 1 - 3 - - - - 1 1 3 - - - - - 2 - - 6 1 4 60 -69 1 - 1 2 - - 3 1 2 2 1 2 -- - - 4 - - 12 2 5 2 - - - - - - - - 1 - - 4 - 1 70 -79 2 - 2------------------------------------- 3 - 80 - 89 - - - Total 12 6 6 3 - - 4 3 3 7 3 3 3 - - 15 1 1 44 13 13 aJ kIngery, 1931 included. i/ 'North America' includes only the U.S. 1920 (All). Education Table/3.1 -92- TABLE 3.2 FREQUENCY DISTRIBUTIONS OF OCCU'PATIONS OF FATHERS: UNIVERSITY STUDENTS South and Japan, Occupation and Asia Central Eastern W. Europe, Percentage and Category Africa Mediterransan (icl. Japan) America Europe N.. America b/ Total Farmers 10 I - 1 - 2 2 2 2 2 4 6 b/- 7 13 9 16 10-1 - 19- 1 4 3 1 2 2 6 3 6 4 1 20 9 -6 20 -29 - - - 2 - - - - 2 2 1 I 1 2 5 1 5 30 -39 3 - I 2 1 - I - - - - - - 1 6 2 L 40 -49 2 - - - - - 1 1 ----- ------- - - - 3 - 50 and over 2 1 - - - - - 1 ----- ------- - - - 2 2 - Total 9 1 1 6 2 2 7 5 5 4 4 2 10 1 8 13 Il 10 49 24 28 H!anual Workers 10 6 3 5 2 2 2 3 3 3 l a/ - - 5 3 4 17 1 1 14 10 -19 1 1 1 5 3 3 I - - - 8 3 2 13 7 6 20 -29 . - 2 1 3 I - - 3 1 1 9 2 1 30 -39 2--------------------------------------------------- 1 - - - 9 1 1 40-49------------------------------------------------------------- - - - -I I - - 50 and over 1 1 - - - I I I- - -I2 1 1 Total 9 3 5 5 3 3 9 7 7 3 1 11 I 1 16 7 7 53 22 23 Proprietors and Traders 10 5 2 2 1 1 1 1 I 2 2 2 1 :0 5 6 10 -19 2 2 - 4 2 2 2 2 2 1--------------------7 5 4 16 11 8 20 -29 - - - 1 22 - 2 2 2 7 2 L 30 -39 1 1 1 I I 2 2 1 40 -49 - - --- - - - - --- --- - - ----- - - - 50Oand over----------------------------I---------------------------------------------- - - Total 7 4 2 6 5 3 7 3 4 3 1 - 2 2 2 1 1 5 5 36 20 16 White Collar (excluding proprietors and trade) to0-19 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 20 -29 1 - - - 1 - 1--------------------------------------------------2 1 30 -39 3 - - 2 - - 2 2 2 1 I 11 2 40 -49 - I - - I - I - - 1 - 1 - - 4 - 50 -59 1 - -- - 1 1 1 2 - -. - - 2 1 - 6 2 2 60 -69 - I I 11 --------------3 2 - 5 4 - 70 and over - - I I - 2------------------------------5 1 4 7 1 7 Total 5 - - 6 2 2 7 4 4 4 - - 2 - - ~ 12 4 4 36 10 1 0 All Non-manual, Non-farm 10 -19 I I ----------------------- - - - ------------- 1 1 20 -29 30 -39 1 -- - - 1 - -I 1 - - - 2 2 - 40 -49 3 - 3 1 1 - - 2 - L - - - 9 1 1 50 -59 4---------------------- -- - -----------7 1 - - 12 3 2 60 -69 - - - - - 2 3 1- 3- - - - - 4 - 1 9 3 2 70 -79 - -- - L 1 - 1 ----- - ------7 2 1 7 4 2 80Oand over 1 - - 4 3 4 5 2 6 - I - 1 'i/- - 7 3 3 18 9 13 Total 9 - - 7 5 5 7 a 8 3 1 - 11 I 1 19 5 5 58 20 19 a/ Hungary 1931 included. b/ 'North America' includes only the U.S. 1920 (All) and c. 1950 sales. Mexico is Srouped vith South and Central America. Education: Table/3.2 - 93. - of non-farm/non-manual fathers are more pervasive among university students. The pattern for proprietors and traders is not greatly different for the two levels of school. Scanning these two tables identifies many anomalous coun- tries. While the foregoing contrasts exist in most countries, oDe can dis- cern no overall or general contrast in the profiles of students' fathers as between secondary and higher education. These profiles are set out in a summary form on Charts 3.1 and 3.2, using triangular grids. One observes a very high representation of "white collar" fathers in the higher institutions, but there is also a broad scattering down to about a third of total enrollment. 'The distribution for secondary students lacks such a concentration of white-collar youth, and it displays a more distinct contrast between representatives from farm and manual-worker homes. Where less than half the students in secondary school come from white-collar families, there is a stringing out of young people (mainly males) with less than 20 percent from manual homes but with 30 to over 70 percent from farm backgrounds. The cases with fewer than 20 per- cent from farm homes and less than half from white-collar backgrounds are more numerous, and they are not predominantly male pupils. (Only four cases-- oDe male and three undifferentiated by sex--fall between those two groups.) Chart 3.2, for university students, has no such demarkation. IDstead we see a cluster of cases that range from 30 to 65 percent white-collar, 25 to 40 percent manual, and 5 to 30 percent farming. There are wide differ- ences among countries in patterns and degree of selectivity to schools, though the contrasts between these two charts also reflect differences in the respective sets of countries. 100 0 Chart 3.1 Occupational Distributions of 90 10 Fathers of Secondary Students /3 Not Distinguishing Sex & Males 80 20 Females 6 60 400 40 3 a a0arm 30 70 20 10 - - 0 100 100 90 80 70 60 50 40 30 20 10 0 Percentage Farm 100 0 Chart 3.2 Occupational Distributions of 90 10, Fathers of University Students 13 Not Distinguishing Sex AMales 180 Q 20Females A 40 40 30. 500 0 100 100 90 80 70 60 50 40 30 20 10 0 Percentage Farm - 96 - As was demonstrated in the previous analysis of selectivity by parental schooling, interpretation of selectivity indexes must take account of the underlying distribution of occupations in the base population; see Charts 3.3 through 3.6, using probability scales. Equal proportions of occupations among fathers of pupils. and within the base population would be represented by entries on the diagonals. In Chapter 2 high-education and low-edl,;7ation scattergrams could be put on the same chart because in no case were pupils with well-schooled fathers under-represented, and vice versa. The categories of parental schooling were systematically monotonic, but in this respect the situation for occupation is less neat. White-collar families generally are over- represented in both secondary and university populations, but there are exceptions as there are also for traders. (On Charts 3.1 and 3.2 traders and proprietors had been merged with white-collar in order to obtain three categories totaling 100 percent; on Charts 3.3 and 3.4 white-collar excludes traders.) The relationship is very tenuous between proportions of students' fathers who are in white-collar jobs and the proportions of adult males in such work, especially among secondary students. The vertical distance to the diagonal on the charts indicates the degree of selectivity, taking account of the base population constraints. (These patterns closely resemble the differences in logits discussed in Chapters 5 and 6 and in Appendix E.) Social selectivity obviously differs among countries for each level of school. Observe that the distribution of the base population for white-collar men includes cases with smaller percentages and with higher selectivity (greater vertical distances from the diagonall on the chart for secondary schools 97 - than on that for university: This reflects the fact that the sample for secondary students includes more of the least developed and fewer of the industrialized countries. On each chart, proportions of fathers who were traders ran generally parallel to the line of equal representation and closer to it than for white- collar men. Indeed, selectivity for offspring of traders and proprietors is more uniform among countries than is selectivity for children from white- collar families. Children of traders appear to be under-represented in secondary schools of the Cameroons, though the data may be unreliable; at university level it appears that offspring of traders are grossly over- represented in Mexico but there is a shortfall in Lebanon, or so it appears. Parts C and D of Charts 3.5 and 3.6 deal with children of manual laborers and of farmers--for whom selectivity indexes normally will be below unity. Among fathers of secondary students under-representation of manual workers (shown by vertical distance below the diagonals of Part C) was most marked in a few countries having 40 to 60 percent of the base population in such work; by the 1960s Puerto Rico and Hungary had moved toward more equal representation for children of manual workers. In the three African countries that display over-representation for manual families we perhaps are seeing the effects of urbanization and of participation in the modern sector among a previously rural and.unskilled population. One's inference from this situation is quite different from what it would be with corre- sponding over-representation from manual homes in an industrialized nation. Children from manual families are more often under-represented in university (Chart 4.6) than are such children in secondary school; however, - 98 - Chart 3.3 Percentages of Fathers by Percentages of Base Populations; Secondary Students A. White Collar A. WHITE COLLAR B. Proprietors and Traders Percentage of Fathers 70 500 ou . o_o____ 40 a a 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 B. PROPRIETORS AND TRADERS 40 0 10~ 00 Feal 10~~~ Allg UK 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 PrOenPtaE of BA PopulAtion -99- Chart 3.4 Percentages of Fathers by Percentages of Base Populations; University Students A. White Collar B. Proprietors and Trader A. WHITE COLLAR 90 80 A Male j0 70 -Female 60 - 0 All 50 - 40 Cj 30 0.2 0.5 1 ;! 5 10 20 30 40 50 60 70 80 90 B. PROPRIETORS AND TRADERS 60 50 40 200 30- Mcxico ^ Lebwon Ito~ 10 0 =0 0.2 0.5 1 2 5 10 20 30 40 50 60 70 80 90 Percentage of BSe Population 100 - Chart 3.5 Percentages of Fathers by Percentages of Base Populations; Secondary Students C. Manual D. Farming G MANUAL 60 / | A Male 60 50 All 40|(P-000) | Bo 40 0 Senegpl / 30- 20Oo X O 9 13 Puerto Rico 20ol 19440 1 0 -- - Coast , 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 n). FARMING 00 40. F 70- r .755 600 (p.000) 000+ a 50 30 0 10-0A 300 200 1 0 , XX - 50 I . I I I I I I -I I I l I, 0.5 1 2 5 1 0 20 30 40 50 60 70 80 90 95 98 Percentage of Base Population - 101 - Chart 3.6 Percentages of Fathers by Percentages of Base Populations; University Students C. Manual C. MANUAL 70 * Papua A Male 60 * Female A iPapua 0 All 50 30 20 1 0 A Belgium / A. I Austria France 2 - . 0 Netherlands 1 a Kenya t I II II I I l- I I II I 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 D. FARMING .50 40 r- 400 30 - m ° ^0 A 200 1 00Portugal °° ° 0 a Mexico 5 / a 2 * Malaysian Indian 0.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 Percentage of Base Population - 102 - the countries here are more heterogeneous. Under-representation was most marked in Western Europe (and in Kenya). Papua New Guinea displays in Chart 3.6 the situation remarked for secondary pupils in West Africa (Chart 3.5); the few newly-modernized manual workers seem to be eager participants in the remarkable educational progress of that new nation. Representation from farm families will be examined in detail later in this chapter; here children of farmers are compared with those from manual- worker homes (lower sections of Charts 3.5 and 3.6). Among university students the extreme under-representation for manual families (seen in parts of Western Europe and in Kenya) does not occur for farm families-presumably because agriculture has become a highly capitalized sector. (Mexican peasants and Indian peasants in Malaysia are distinctly disadvantaged.) For children of farmers attending secondary school we observe an astonishingly wide range among countries in degree of under-representation. This diversity is marked even in countries with a high proportion of base-population farm families. The most extreme cases almost surely reflect low rates of secondary enrollment, but it is for just those countries that we have poorest information about enrollment rates. It is a commonplace in the literature on educational selectivity to bracket children of farmers and of manual laborers as "non-white-collar." But as the foregoing comments indicate, the counterpoint between farm and manual groups has some perhaps surprising features, as perusal of selectivity indexes shown in Appendix K reveals. Comparison of these two groups is restricted by the comparative absence of manual fathers in the countries for which we have data about secondary schools in contrast to data for university students. Among less-developed countries the selectivity indexes for -103- children of manual laborers tend to be greater than for farmers; in these countries manual laborers are an emergent and relatively privileged small population. Among the set of countries for which there are data on university students, the picture is more complicated. In the developing countries (speaking broadily and without using charts) the selectivity index tends to be larger for manual workers than for farmers. But in the more developed countries-and where farming typically is a highly capitalized sector with approximation to parity of income for farmers-the selectivity index is definitely larger for children of farmers. The improvement in the position of farm children in association with "development" is not surprising. C. An Overview of Degree of Selectivity The index of dissimilarity (between distributions of occupations in the base population and among parents of students) is a convenient summary indicator of degree of selectivity into schools. It was a robust indicator to use in analysis of selectivity for parental schooling because of the monotonic selectivity pattern for parental education; in most instances there was information on the relevant breaking points between under- and over- representation. Such a tidy pattern can rarely be anticipated when the criterion of status is paternal occupation. The index of dissimilaxity will be an underestimate if some occupations in the white-collar category are scantily represented while some sorts of manual workers are over-represented. It is in these situations that degree of aggregation can limit comparative use of this index. Merely aggregating into three, presumably hierarchically structured, categories of occupation does not resolve the difficulty. Fortunately, the ranking of occupations need not be the same in all countries; the index of dissimilarity allows re-ordering to fit the particular case. It - 104 - does not matter that in some countries some manual workers are over- represented, that traders occasionally are under-reprsented, or even that once in a while we find a white-collar group deficient (Chart 3.3A). The available indexes of dissimilarity are summarized in Table 3.3 (Western Europe, North America, and Japan are combined). The indexes for individual countries are given in Appendix tables K.1 and K.2. To recall, this index specifies what percentage of pupils' fathers would have to change occupation in order to bring their distribution into alignment with that of the base population. Most of the indexes shown as below 30 percent were really between 20 and 30 percent; those over 50 percent were spread into the 60s and even into the 70s. D. Who is Under- and Who is Over-represented? Distributions of selectivity indexes by occupation of father are summarized in Table 3.4. Though selectivity for girls tended to be greater, sex contrasts were not impressive in the few cases for which separate data by sex were available. This summary therefore combines the sexes. (Details by country and sex are shown in the Sobs column of Tables 3.5 and 3.6.) The upper part of Table 3.4 refers to occupations that usually are over-represented among students; most decidedly this is the case for "white collar excluding proprietors and trade." Except for these traders the selectivity indexes are highly variable for each category of parental occupation. (The highest ratios here do not match those found in some less- developed countries for the best-schooled fathers; only two cases exceed 20.00 and both were for secondary pupils.). - 105 - TABLE 3.3 FREQUENCY DISTRIBTUION OF INDEXES OF DISSIMILARITY BETWEEN OCCUPATIONS OF FATERS AND OF BASE POPULATION Region and Sex Under 20 20-29 30-39 40-49 50 and over Total Secondary Students Both sexes W. Europe, Japan, U.S. 6 1 6 2 - 15 E. Europe 1 1 - - I a 3 Other 4 2 6 5 18 Total 36 Males W. Europe, Japan, U.S. 1 - 1 - 1 3 Other (Africa & Asia) 1 3 3 - - 7 Total 10 Females W. Europe 1 - 1 - 1 3 Other (Africa & Asia) - 2 2 - 3 7 Total 10 University Students Both Sexes W. Europe, Japan, U.S. - 3 4 8 3 18 E. Europe 1 1 3 1 1 a 7 Other - 2 3 3 4 12 Total 37 Males W. Europe, Japan, U.S. - 1 6 1 2 5 Other (Asia and Mediterranean) 1 - 2 2 2 7 To tal 12 Females W. Europe, Japan, U.S. - 1 - 2 3 6 Other (Asia and Mediterranean) - - - 2 5 7 Total 13 a/ Hungary, 1931 - 106 - TABLE 3.4 FREQUENCY DISTRIBUTIONS OF SELECTIVITY INDEXES BY FATHER's OCCUPATION White White All Collar All Collar Selectivity Non-farm (excl. Proprietors Non-farm (excl. Proprietors' Index Non-manual proprietors) and Traders Non-manual proprietors and Traders Under 1.00 1 1 3 - 2 1.00 - 1.99 11 6 18 9 5 15 2.00 - 2.99 9 8 3 17 14 4 3.00 - 3.99 4 1 2 5 6 3 4.00- 5.99 4 4 2 3 1 2 6.00 - 7.99 2 1 - I I - 8.00 - 9.99 1 4 - - 10 00 and over 1 4 - 1 - - Total 33 29 28 36 27 26 Manual Farm Manual Farm Under .10 - 1 .10- .19 2 1 9 - .20- .29 5 4 7 3 .30- .39 2 4 6 4 .40- .49 3 4 2. 7 .50- .59 4 5 3 8 .60- .69 1 2 5 8 .70- .79 2 4 4 2 .80 -.89 5 1 1 2 .90 -.99 4 5 2 2 1.00 - 1.99 4 3 1 - 2.00 - 2.99 - - 1 3.00 - 3.99 - 1 - 4.00 and over 3 - - Total 35 34 42 35 - 107 - There are two closely related reasons why the distributions of selectivity indexes for paterual occupation differ from those for paternal scho'oling. First, none of the occupational categories in the source material referred to base populations constituting as small a proportion of the total population as occurred in several countries for some of the educational categories. Second, fewer less-developed countries were included among those for which data were available on occupations; this altered distributions of base populations in such a way as to constrain selectivity indexes to more moderate levels. Table 3.4, like charts 3.3 and 3.4, showed that the selectivity in- dexes are generally lower for proprietors and traders than for white-collar men; this indeed might have been predicted. What the data (and charts) demonstrate that perhaps was less predictable is that selectivity indexes for traders and proprietors exceed 1.0 w.ith few exceptions. The lower part of Table 3.4 presents selectivity indexes for occupa- tions that normally are under-represented. (This distribution is divided more finely than is that for over-represented occupations.) ,Both the manual and the farmer indexes in these distributions are spread out remarkable evenly and over a wide range for both secondary and university students, but for offspring of farmers attending university there are fewer extreme values. For each occupation of fathers the degree of selectivity is highly diverse among countries. In part, though only in part-as was shown in the analysis of selectivity by parental schooling-this diversity reflects the fact that the selectivity indexes on a given category of fathers are constrained by the proportions of such men in the base population. Charts 3.7 and 3.8 show the relationship between that constraint and the observed - 108 - selectivity indexes for white-collar origins (excluding traders). For both secondary and university students the observed indexes parallel the diagonal that defines the maximum constraint due to base population propor- tions. Nevertheless, the observed index departs in widely differing degrees from that maximum. Some observations are close to the maximum, whereas others fall substantially below it, in some cases coming closer to the egalitarian ratio of 1.0. The horizontal lines on these charts (as on the parallel charts in Chapter 2) specify the upper limits of possible selectivity indexes at selected enrollment rates. Those constraints are less often the operative ones, however, when selectivity is by parental occupation than when it is by parental education, given the occupational categories on which data are available. Tables 3.5 and 3.6 present an analysis for secondary pupils of observed and maximum possible selectivity indexes. This analysis, which parallels analyses in Chapter 2, refers to selectivity from the ranks of white-collar families excluding traders (Table 3.5) and then alternatively including them (Table 3.6). Since lack of data on enrollment was more of a problem in the sample of countries for which occupational selectivity data were available than in the samples discussed in Chapter 2, this analysis is incomplete. In most instances for which we have enrollment rates, those rates were not high enough to be the binding constraint. In some cases it is clear even without precise information that enrollment rates would not be high enough to be the constraining factor on the upper limit of,selectivity indexes. Where enrollment rates are lacking, the fifth column of the table Page 1 of 2 -109- TABLE 3.5 ITMEXES AND CONSTRAINTS ON THEM FOR FATHERS WHITE COLLAR (EXCLUDING PROPRIETORS AND TRADERS; SECONDARY STUDENTS Sobs-l Max Sbs - Max S) sobs Max Sbc Max Sec Max S Max S-1 (Max Sbc - S obs) (1) (2) (3) (4) (5) (6) AFRICA E. Cameroun 1971 All 2.41 3.97 a/ 3.97 .47 0 W. Cameroun 1971 All 2.92 4.69 a/ 4.69 .52 0 Ghana Form 5, 1961 All 5.99 14.49 34.48 14.49 .37 0 Male 5.10 14.49 20.83 14.49 .30 0 Female 9.81 14.49 90.91 14.49 .65 0 Form 6, 1964 - All 8.49 14.49 a/ 14.49 .56 0 Ivory Coast 1963 All 9.10 50.00 (.17) Male 7.15 50.00 (.13) Female 4.35 50.00 (.47) Mali 1965 All. 38.36 90.90 (.42) Niger 1966-67 All 41.00 142.86 (.30) Male 31.71 142.86 (.22) Female 73.43 142.86 (.51) W. Nigeria 1967 All 4.91 33.33 (.12) Senegal 1962 All 8.50 50.00 (.15) MEDITERRANEAN AhND SOUTHEAST ASIA Turkey Lyce 1962 All 7.92 16.39 14.29 14.29 .52 .25 Malaysia Form 5, 1967 All 1.95 7.19 11.11 7.19 .15 0 Male 1.73 7.19 9.09 7.19 .12 0 Female 2.33 7.19 14.29 7.19 .21 0 Malaysia 1972, Form 5: Malays Male 2.33 9.90 10.00 9.90 .15 0 Female 3.55 9.90 20.00 9.90 .29 0 Chinese Male 1.54 6.62 8.50 6.62 .10 0 Female 1.96 6.62 11.11 6.62 .17 0 Spain 1970 All 2.26 5.08 a/ a/ .31 Thailand 1978(1st yr) All 4.33 11.11 6.67 6.67 .59 .65 SOUTH AND CENTRAL AMERICA Chile, Liceo 1970 All 3.30 8.90 5.88 5.88 .47 .54 Puerto Rico 1944 All 4.40 9.71 4.76 4.76 .90 .93 1960 All 1.27 4.24 3.12 3.12 .13 EUROPE AND JAPAN E. Germany 1960 Full-time .64 2.58 1.33 1.33 Part-time 1.17 2.58 Not possible to combine full-time 1967 Full-time .78 2.58 and part-time. 1.30 1.30 Part-time 1.64 2.58 110- Page 2 of 2 Sobs-1 Max Sbs - Max S) Sob. Max Sbc Max See Max S Max S-1 (Max Sbc S oba) (1) (2) (3) (4) (5) (6) EUROPE AND JAPAN (cont'd.) France 1959 All 2.60 4.83 (.42) W. Germany 1972 All 93 2.72 (Full-time) Male .94 2.72 under 1.00 Female .92 2.72 Sobs Norway 1969 All 2.43 5.08 4.00 4.00 .48 .41 Sweden 1960 All 2.27 4.08 3.33 .55 .44 ADDENDUM ON I .E.A. COUNTRIES 1963 Belgium 1A12 2.86 2.56 .08 .17 England 2.90 4.55 8.33 .54 0 Finland 1.96 4.00 4.17 .32 0 W. Germany 2.51 3.03 * .74 * Japan 1.31 3.28 1.79 .39 .90 Sweden 2.17 3.70 2.94 .60 .50 U.S. 1.10 2.50 1.35 .29 .38 a/ Max Sej not known but definitely exceeds Max Sa. b/ Entries in parentheses are all (S,bs-l)/(Max Sbc+l) when It is not known whether Max Sej exceeds or is less than Max Sbc. Education:Table/3.5 (1-3) -111- Page 1 of 2 TABLE 3.6 SELECTIVITY INDEXES AND CONSTRAINTS ON THEM; FOR FATRERS WHITE COLLAR AND PROPRIETORS AND TRADERS; SECONDAIty STUDENTS Sobs-I Max Sbs - Max S) Sobs Max Sbc Max Sec Max S Max S-1 (Max Sbc - S obs) (1) (2) (3) (4) (5) (6) AFRICA E. Cameroun 1971 All 1.56 2.12 a/ 2.12 .50 0 W. Cameroun 1971 All 1.58 2.26 a! 2.26 .46 0 Ghana Form 5, 1961 All 4.86 9.35 34.48 9.35 .46 0 Male 4.34 9.35 20.83 9.35 .40 0 Female 7.05 9.35 90.91 9.35 .72 0 Form 6, 1964 All 6.28 9.35 a/ 9.35 .63 0 Ivory Coast 1963 All 4.49 18.87 .20 0 Male 3.64 18.87 .18 .30 Female 10.91 18.87 76.42 18.87 .55 0 Kenya 1961 Male 3.48 11.11 16.13 11.11 .25 0 Female 6.92 11.11 55.56 11.11 .59 0 Mali 1965 All 5.19 14.29 100.00 14.29 .54 0 Niger (see year) All 14.60 40.00 66.67 40.00 .33 0 1966-67 Male 12.04 40.00 43.48 40.00 0 Femle 23.48 40.00 142.86 40.00 .58 0 U. Nigeria 1967 All 2.64 8.85 a! 8.85 .21 0 Senegal 1970 All 5.73 12.66 33.33 12.66 .41 0 MEDITERRAiN AND SOUTHEAST ASIA Turkey (Lyce) 1962 All 6.49 9.52 14.29 9.52 .64 0 Malayala 1966-67 All 1.72 3.16 11.11 3.16 .33 0 Form 5 Mcle 1.62 3.16 9.09 3.16 .29 0 Female 1.89 3.16 14.29 3.16 .41 0 Malaysia 1972. Form 5: Malays Male 2.05 6.94 10.00 6.94 .18 0 Female 3.11 6.94 20.00 6.94 .36 0 Chinese Male 1.57 2.61 8.50 2.61 .35 0 Female 1.66 2.61 11.11 2.61 .41 0 Spain 1970 All 2.05 3.39 a/ 3.39 .44 0 ThaIland 1978 All 3.83 5.56 667 5.56 .62 0 SOUTH AND CEPNTRAL AMERICA Bolivia 1975 Grade 12 All 3.91 8.06 5.00 5.00 .73 .74 Chile 1970 Lysl 4 All 2.69 5.15 5.88 5.15 .41 0 Puerto Rico 1944 All 3.38 6.54 4.76 4.76 .61 .56 1960 All 1.53 3.60 3.12 3.12 .25 .23 -112- Page 2 of 2 Sobs-1 Max S - Max S) Sobs Max Sbc Max 'SeC Max S Maz S-1 (Max Sbc - S cbs) (1) (2) (3) (4) (5) (6) El-STERNt EUROPE E. Germany 1960 Full-time .74 3.4S .3 1.33 * * Part-time 1.11 1967 Full-time .82 2.35 1.30 1.30. * * Part-time 1.50 Hungary Lyc6 1931 All 3.19 3.73 9.09 3.73 .80 0 1963 All 2.21 5.65 3.70 3.70 .45 .57 WESTERN EUROPE, JAPAN, UNIT'PD STATES Denmark All 1.83 2.42 7.35 2.42 .58 0 Male 1.84 2.42 11.11 2.42 .59 0 Female 1.82 2.42 3. 2.42 .S8 0 Finland 1965 2.08 3.33 * * (.46) 0 France 1959 All 2.22 3.50 .76 .44 Norway 1969 All (1.19) 3.98 (1.25) a! (1.25) b/ (.76) .98 All 2.28 3.98 4.00 3.98 .43 0 Sweden 1960 All 2.11 3.10 3.353 3.10 .48 0 U.S. Birth Cohort of 1928-37 (Grade 12) Male 1.18 2.48 1.55 1.55 .33 .72 ADDENDUM ON I.E.A. COUNTRIES 1963 Belgium 1.11 2.13 2.56 2.13 .10 0 England 2.53 3.23 8.33 3.23 .69 0 Finland 1.64 3.33 4.17 3.33 .27 0 W. Germany 2,14 2.38 * 2.38 .83 0 Japan 1.31 2.13 1.79 1.79 .39 .41 Sweden 2.08 2.94 2.94 2.94 .56 0 U.S. 1.12 2.17 1.35 1.35 .10 .78 8/ NOt known but greater than Max bc b/ Not known but around 1.25. Education:Table/3.6 (1-4) - .113 - has been filled in with entries in parentheses entirely on the basis of base-population constraints; those entries give minimum estimates of the ratios of observed to maximum possible selectivity indexes. 1/ The estimates in Table 3.6 are more complete than in Table 3.5 primarily because Table 3.6 (using the broader definition of white collar) entails a larger base-population proportion and in most cases it is possible to specify definitely that enrollment rates in secondary school will not be high .enough to become the determining constraint. The higher the ratio in the fifth column of Tables 3.5, and 3.6, the more nearly does the selectivity index press toward or against the upper bound. The lower the ratio in the fifth column, the stronger must be other forces bringing an educational system toward a more nearly proportional representation by occupational background in the secondary schools. Typically indexes for males are lower than those for females, whose selectivity into secondary education is greater (outside of a few countries), but most of the sex differences are surprisingly small. Sex contrast is larger in Africa where even for boys the indexes are high. Data are available for sex comparison for only one Muslim country, Malaysia; there the selectivity index for girls is only 40 percent of the maximum possible. In the few cases for which enrollment rates set the upper limit or white-collar selectivity for secondary pupils, they account for as much as 1/ Where higher education is expanding more rapidly than white-collar employment in the cohorts of students' parents, enrollment constraints will become increasingly important in setting the upper limit of selec- tivity indexes; many of the low indices in parentheses in column 5 might be raised if they were computed against enrollment-constrained maxima. - 114 - 98 percent and as little as 23 percent of the difference between the maximum index (constrained only by the base population MwaSBi) and the observed index. Among countries there are wide variations in the extent to which reduced over- representation from white-collar homes is associated with expanded secondary enrollments as a constraint on the selectivity index. Educational policies and other societal factors operating within the constraints from the occupa- tional structure and even from enrollment rates can alter substantially the degree of selectivity by parental occupation into secondary schools. Readers who are familiar with particular countries can interpret the situation in those countries. It mtLy be of interest to comment on Japan and the two Germanies. Looking only at East Germany one might suppose that the contrast between the full-time and part-time secondary pupils (with higher white-collar selectivity into the latter) reflected postwar political develop- ments and deliberate favoring of workers' children. But apparently even in West Germany white-collar families are under-represented among full-time pupils, and so one infers that there may be a structural feature of West German schools to be identified. For Japan, the higlier 1966 figure in the last column of Tables 3.5 and 3.6 is easily understood; in that education-centered society it is only the humblest or the least successful whose children were not being pressed by parents into secondary school--even when they can gain access only to schools or curricula that are second or third choices. Since 1966 the proportion of boys enrolled in the last year of ,uiFr-secondary school has risen from two-thirds to over 90 percent of the age cohort; the maximum possible selectivity index is now only 1.10 and there remains little scope - 115 - for further equalization of representation. Selectivity in Japan remains, as it has always been, largely by type and prestige of school rather than by level of school; policies for equalization must deal with the former issue. In higher education the situation with respect to enrollment rates is quite different. Rarely do these rates exceed the proportions of the base population who are white-collar workers (even narrowly defined).. The base- population proportion defines the maximum possible selectivity index, and it is so treated in Table 3.7. The ratio of observed to maximumn possible indexes is higher for the broader definition of white collar than for the stricter definition (columns 3 and 6). Given the larger base population for column 3, this contrast is to be expected so long as normally children of traders and proprietors have "precedence" for entry into secondary school. Putting this the other way round, the ratios in column 3 will be high so long as few children of farmers and manual workers enter higher institutions. If none of those children enrolled, the observed selectivity index would equal the maximum as constrained by the base population and the ratio in column 3 would become 1.00. This will be the case whatever the policies or processes by which the scarcity of lower-status children is brought about. Those policies need not be legislative but may reflect customs, traditions, and incentives. The noteworthy sex contrast in Tunisia, for example, re- flects those sorts of influence, which in turn affect governmental policies. The most extreme opposite example is Papua New Guinea where the base- population constraint is inoperative. Overall there is a moderately close association between occupa- tional selectivity indexes and base-population proportions even though few -116 Page 1 of 2 TABLE 3.7 SELECTIVITY INDEXES AND THE BASE-POPULATION CONSTRAINGS; WHITE COLLAR FATHERS OF UNIVERSITY STUDENTS White Collar including White Collar excluding Proprietors and Traders Proprietors and Traders Sobs-l Sobs-l Sobs Max Sbc Max Sbc-1 Sobs Max Sbc Max Sbc-1 AFRICA Kenya 1970 All 17.33 33.33 .41 2 8 1 China 1970 All 4.99 9.17 .49 5.61 14.29 .35 Ivory Coast All 2.76 4.78 .47 2.81 5.29 .42 MEDITERRANEAN Greece 1962 All 2.71 5.88 .35 3.72 11.36 .26 1980 All 1.89 4.17 .28 1.91 6.41 .17 Leba.on 1961 Male 1,57 1.75 .68 4.22 6.45 .59 Female 1.69 1.75 .92 5.22 6.45 .77 Portugal 1964 All 4.82 5.85 .79 6.10 10.10 .56 Spain 1962 All 4.19 4.78 .84 * Male 4.11 4.78 .82 ** Female 4.41 4.78 .90 * * * Tunisia 1965 Male 5.82 13.51 .39 7.10 25.64 .25 Female 10.09 13.51 .81 15.23 25.64 .58 Yugoslavia 1965-6 All 2.92 6.29 .36 3.38 8.00 .33 SOUTH AND SOUTHEAST ASIA Malasia. 1968 Malays Male 2.29 7.63 .19 2.91 12.72 .16 Female 3.89 7.63 .44 5.19 12.72 .35 Chinese Male 1.92 3.04 .45 2.23 7.69 .18 Female 2.31 3.04 .64 3.31 7.69 .33 Indians Rale 2.60 4.07 .52 4.0' 7.69 .46 Female 3.21 4.07 .72 4.7i 7.69 .56 Papua New Guinea All 7.50 166.67 .04 * * * 1975 Male 6.50 166.67 .03 * * * Female 9.83 166.67 .05 * * * South Korea 1970 All 2.17 3.37 .49 * * * SOUTH AND CENTRAL AMERICA Mexico 1950 Male 8.81 10.75 .80 7.51 14.49 .48 Puerto Rico 1944 All 3.96 6.54 .53 3.63 9.71 .30 1960 All 2.38 3.68 .52 -2.12 4.35 .33 RUSSIA AND EASTERN EUROPE U.S.S.R 1938 All 2.38 5.65 .30 * * * 1964 All 2.35 4.76 .41 * * * 1970 All 2.12 4.00 .37 * * * East Germany 1967 Full time 1.24 1.17 2.35 2.58 Part time 2.03 2.19 -117- Page 2 of 2 White Collar including White Collar excluding Ptoprietors and Traders ProprietJrs and Traders Sobs-1 Sobs-1 Sobs Max Sbc Max Sbc-1 Sobs Max Sbc Max Sbc- -RUSSIA AND EASTERN EUROPE (cont'd.) Hungary 1931 All 3.11 3.73 .77 * * * 1963 All 3.18 5.65 .47 * * * Poland 1960-61 All 2.31 4.17 .41 * * * 1974-75 All 2.18 3.76 .43 * * * WESTERN EUROPE AND SCATDINAVIA Austria 1965-66 All 3.49 3.85 .87 3.82 5.18 .67 Male 3.48 3.85 .87 4.07 5.18 .73 Female 3.54 3.85 .89 4.17 5.18 .76 Belgium 1966-67 All 1.98 2.82 .54 2.53 4.72 .41 Male 2.40 2.82 .77 3.23 4.72 .60 Female 2.62 2.82 .89 3.81 4.72 .76 Denmark 1964-65 All 1.66 2.38 .48 1.95 4.00 .32 Male 1.65 2.38 .47 2.02 4.00 .34 Female 1.66 2.38 .48 1.92 4.00 .31 France c. 1955 All 2.79 3.08 .86 3.46 4.63 .68 1961-62 All 2.13 2.60 .71 2.53 3.98 .-1 Male 2.11 2.60 .69 2.48 3.98 .50 Female 2.14 2.60 .71 2.59 3.98 .53 1967-68 All 1.96 2.44 .67 2.20 3.42 .50 W. Germany 1972 All 2.06 2.44 .74 2.03 3.36 .44 1964-65 All 1.78 2.99 .64 1.76 2.74 .44 Male 1.74 2.22 .61 - 1.71 2.74 .44 Female 1.90 2.22 .74 1.93 2.74 .53 Ireland 1963 All 2.35 3.25 .60 * * * Netherlands 1964-65 All 2.01 2.31 .77 2.42 3.36 .60 Male 1.94 2.31 .72 2.28 3.36 .54 Female 2.15 2.31 .88 2.81 3.36 .77 Norway 1964-65 All 2.85 5.18 .44 3.37 7.19 .38 Sweden 1960 All 2.37 3.13 .64 2.58 4.08 .51 Switzerland 1959/60 All 3.05 3.88 .71 * * * U.K. 1961 All 2.60 3.51 .64 * * * Penale 2.67 3151 .67 * * * 1978-79 All 2.24 2.88 .66 * * * (Household survey) JAPAN AND UNITED STATES Japan 1968 All 1.68 2.12 .61 1.98 2.83 .i54 U.S. 1920 All 2.43 3.86 .50 2.21 7.14 .20 U.S. Birth Cohort of 1928-37 (college c. 1946-60) College Entrants Male 1.53 2.48 .36 * * * College Graduation Male 1.64 2.48 .44 * * * Education:Table/3.7 (1-4) - 118 - countries press closely aginst their maximum possible values. The main exceptions for higher education are in column 3 of Table 3.7; for Spain and Austria and for France in the 1960's; for Mexican males; and for females in Lebanon, Belgium, and the Netherlands. Charts 3.7 and 3.8 reveal a marked association between selectivity indexes and base-population proportions in white-collar employment. They display also wide variations among countries in the extent to which the indexes deviate from their maximum possible values under the base-population constrat.nt alone--especially for secondary, but also for university students. Again, there are real differences among countries in the degree of social selectivity for schooling that are not accounted for by base-population constraints. E. Selectivity of Farmers' Children and the Place of Agriculture in the Econiomy There is no more heterogeneous occupational designation than "farmer." This fact is often acknowledged by distinguishing between "com- mercial" and "subsistence" farmers or between "big" farmers and peasants. The wide diversity in representation of farmers' children among students in secondary or higher schools is due partly to these ambiguities. Everywhere one would expect larger representation among the more sophisticated farmers than among small farmers or peasants, and where "farmers" are a small fraction of the labor force it is the former sort who would predominate. A scattergram not shown here related the percentage of fathers of secondary students who were farmers to the proportion of the labor force in agriculture. There was a marked dichotomy in the sample countries between those in which 20 percent or less of the labor force was in agriculture (all in western Europe, North America, or Japan) and those with 50 percent or ore engaged in farming. All but one of the former had farmer representation among students' fathers - 119 - Chart 3.7 Selectivity Indexes by &se Populations White-Collar (Excluding Proprietors and rraders); Secondary Students Selectivity Index 70 a Sex Not Distinguished 50 A Male * Female 30 - 20 7x at E, \ 0 A 7C 5 -Ma__x. at EMit 20 00 3a 5 Mx atx. a2 * 0 Maa 05 06 07 1 2 3 4 5 6 7 8 910 20 30 ° 40 50 60 70 80 100 Percentage of Bas Population White-Collar (excluding proprietors and traders) - 120 - Chart 3.8 Selectivity Indexes by Base Populations White-Collar (Excluding Proprietors and Traders); University Students Selectivity Index 100\ 70 A Male 50 e Female 30 20 15 10 =0 a Max at Eax at 9 8 7 A A 6 50 1 at E, 20 . - ..'''-'. 0 3 A A0 00 1 1.5 2 2.5 3 4 5 6 7 8 9 10 15 20 30 40 50 60 70 80 100 Prcentage of Be" Population White-Collar (excluding propretors and traders) - 121 - Chart 3.9 Percentage Farm Fathers by Percentage of Labor Force in Agriculture; Secondary Students Percentage of Fathers Who were Farmers 80 Sex Not Distinguished A Males . Females 70../O 60 300 50 / Cl 400 30 -30 200 300 1A 3 100 0 0 I 0 10 20 30 40 50 60 70 80 90 Percentge of Labor Fome in Agriculture -122- close to or in excess of proportions of the population in the~--abor force. A relatively high ratio for Japan reflected the fact that although there had been a rapid decline in farming among younger men, farmers are still numerous among men of an age to be fathers of students. (Women now make up more than half the agricultural labor force whereas men often are in off-farm work.) Japanese farmers' sons were well represented among students but not over- represented if one uses an appropriate base population; the Japan data provided an extreme example of the distortions that may arise in the absence of age controls on the base population. What caught the eye among countries with over 50 percent in agricul- ture is the way the points fan out as one moves rightward toward larger proportions of men working in farming, a pattern that is replicated for universityr students. Even among countries that are preponderantly agricultural, representation of farmers' chi:Ldren in secondary schools varies substantially though in no case did they match the proportion of the labor force in agriculture. Among most of the industrialized countries, by contrast, the proportion of farmers' offspring among secondary students exceeded proportions of the labor force in agriculture and this was common even for university students. It must be remembered, however, that given the age structure of the farming population and the sizes of farm families a random student selection would give just such results. Shifting back from proportions of the labor force in agriculture to the proportion of men in the parental age cohort who were farmers, the most striking result is the extraordinary spread in selectivity indexes in countries in which a third or less of the base population were farmers. In Belgium, Finland, Japan, and as of 1967 (not 1960) in East Germany, the farmer selectivity - 123 - indexes in countries in which a third or less of the base population were farmers. In Belgium, Finland, Japan, and as of 1967 (not 1960) in East Germany the farmer selectivity indexes for secondary students were crowding proportionality at ratios between .90 and 1.00 and in the United States as of the 1950's they were already exceeding 1.00. Denmark is particularly in- teresting in the strong representation of f-amers' daughters among university students, at a 1971 ratio of 1.10 (and .90 for farmers' sons); farmers' daughters exceeded representation of farmers' sons in the Danish gymnasia as well, at 1965 ratios of e89 for. girls and .74 for boys. One may speculate that the sex difference in Denmark reflects different. life prospects for farmers' daughters as those prospects may relate t'6 their education. The spread of selectivity indexes against base populations in farming is shown for university students in Chart 3.9, where entries for Eastern and Western Europe and Japan are distinguished from those referring to other parts of the world. (There is no entry for the United States on this chart.) The Camerouns (1971), with low percentages classified as "farmers" in the base population is the extreme example of farmer over-representation in Chart 3.9 as it was for students in secondary schools as well; one can suspect that the designation farmer is highly selective within the larger agricultural population in the Camerouns data. Figures for the Ivory Coast, however, lend some support to those for the Camerouns; the farm base population for the Ivory Coast was over 86 percent of the total base population. Nevertheless, in the Ivory Coast (1963) there was a remarkably high representation of farm families among male secondary and university students; because few girls attended secondary school this meant a relatively high proportion of all students from farm backgrounds. In general the sex contrasts in representation of children of farmers are especially striking. - i24 - Chart 3.10 Selectivity Indexes by Base Population Farmer; University Students Selectivity Index A Male E Eastem Europe * Denmark 0 Femaie W -Western Europe 1.36 2.26 0 All J Japan 1.00 El 0 Denmark .90 A Denmark W 0 .80 A O 0 WO .70 A WW 0 w0 E w .60 O A 0 A A .50 E wO E 00 oE .40 oW w 0 0 E0 . w .30 0 oW .20 .100A 00 0 10 20 30 40 S0 60 70 80 90 Percentage of Base Population Farmers - 125 - F. Selectivity and Per Capita Income With analysis by occupations of fathers we are looking at a different set of countries, especially for university students, than were used in analyzing selectivity by education. If one interprets cross-national observa- tions as indicators of differentstages of economic development, in the analysis of the next few pages we are focused largely through not exclusively on variations among countries at intermediate and later stages of development. The data show, however, that there is nothing deterministic about associations between national levels of economy and occupational selectivity for education except--and the exception is important--as extremely unequal occupational selectivity for school is incompatible with the levels of enrollment needed to support an economically advanced society. 1. Per capita income, enrollment rates, and indexes of dissimilarity. Chart 3.10 gives plottings of indexes of dissimilarity against per capita incomes (estimated to match survey dates) and Table .3.8 summarizes the effects of per capita income and of enrollment rates on indexes of dissimi- larity between the distributions of occupations of students' fathers and in the base populations. (As in Chapter 2, these regressions include a control for sex.) The results for secondary students are clear and unambiguous: per capita income is closely associated with enrollment rates and where enrollment rates are higher- the index of dissimilarity is lower. With enrollment rates in the regression, per capita income has only a small direct effect on IDX, but the effect is nonetheless negative. The total effect is - .499 in an equation that is highly significant, even though it leaves three-fourths of the variance in IDX unexplained. - 126 - Chart 3,. Index of Dissimilarity betvi -.1-i, Occupations of Fathers and of Base Populationl by Per Capita Income A. SECONDARY STUDENTS |A^ Mal Female O Al Index of Dissimilarity 80- - - 70F 60O soo 00 O 00 20- 10 0o 0-0 B. UNIVERSITYt STUDENTS 80 70 rO 40 O 30 O o. UNI.ERS,T.STUDENT 80 10 60 100 200 300 0oo 1,000 2,000 5,00 10,000 20,000 Per Capita Incofnw - 127 - TABLE 3.8 EFFECTS OF PER CAPITA INCOME AttO EVROLLˇENT RATES ON INDEXZS OF DISSIMILARITY BY FATHER'S OCCUPATION Secondary Students University Students .765 ET (Enaollmeat rate) .643 ET -.232 LlY- \- . 563 LtY. (In. of per capita -.068 ID-DX income) (Index of diss imilarity) N 37 47 Mean IDX 35.1% 44.5Z R' .3319 .3160 P (.0003) (.0003) Ef fects of LNY Direct -.068 - .264 Indirect via ET -.431 -.149 Total -.499 - .413 All equations control for sex. - 128 - For university students, income is again positively associated with level of enrollment, though not quite as strongly as for secondary students. The direct negative effect of enrollment rates on the index of dissimilarity is much weaker; this might be anticipated in view of the comparative lack of university-enrollment constraints on the maximum possible size of the index of dissimilarity. Negative direct effects of per capita income are slightly greater than the direct effects of ET and are of course greater than the indirect effects of income via enrollment rate. The total effect of income is significantly negative but less strongly so than for the secondary students and more of the variance in IDX remains unexplained, two-thirds rather than three-fifths. These outcomes are despite the fact that effects of income embody associated variations in occupational distributions in the base populations. 2. Components of the correlation between per capita income and occupations of students' fathers. This analyais was carried out on three consolidated categories of occupation (due to limitations in original reports): white collar inclusive of proprietors and traders, manual workers, and farmers. Logits of proportions are used in all cases. For students' fathers the codes are respectively: LFWC, LFM, and LFAG; and for the corresponding categories of the base population LBWC, LEM, LBAG. Charts 3.3 and 3.4 (especially the former) showed that associations between base-population and fathers' proportions are closer for the traders & proprietors than for other white-collar men, and over--representation of the former is considerably less than for the more strictly-defined white-collar category. On the other hand, the occupational, economic, and social-status composition of what is here called "white collar" - 129 in the broad sense differs with national levels of economic development. Low-income countries possess only a thin layer of white-collar workers, few manual wage-earners, and usually large proportions of poor peasants. Distri- butions in the base-popultaion reflect these structural variations and their associations with per capita income. An effort is made by use of regression analysis to sort out some of these complexities and their implications for changing selectivity into schools. Table 3.9 parallels the display of Table 2.10, but fGr occupation rather than for schooling of fathers. Looking first at Model I, controlling for sex and income it turns out that base-population proportions are strongly predictive of proportions of white-collar fathers (especially for university students) and of farm fathers (especially for secondary students). On the other hand, in the same matrix there is no association between base proportions and father proportions who are manual workers among university students. Even among secondary students that association is lower than for either white-collar or farm fathers. Effects of enrollment rates are different (Model II). In these samples the associations between income and rates of enrollment are uniformly strong, but the enrollment rate alone has little effect on proportions of students' fathers from different occupations. The partial exception of farm fathers of secondary students entails a modest positive indirect effect of income on proportions of farm fathers via LET, but this is not enough to neutralize the negative direct and other indirect effects of income. In Model II per capita income may be regarded as essentially a proxy for proportions of farmers in the base population, given that income had negligible direct effects in Model I except for its negative effect on the proportions of white- collar fathers (LFWC) among university students. TAnLE 3.9 EFFECTS OF PER CAPITA INCOME, ENROLLMENT RATES, AND BASE POPULATIONS ON PROPORTIONS OF FATHERS - WHITE COLLAR, MANUAL AND FARM Secondary Students University Students White White Collar Manual Farm Collar Manual Farm (LFWC). (LFM) (LFAG) (LFWC) (LFM) (LFAG) Model 1 N 34 31 34 . '7 47 44 Mean of ILF 37.186 -1.542 -1.271 44.879 -1.755 -2.002 LB (logit of base population f2 .5117 .5276 .5499 .6063 .0679 .5932 proportion) Prob F (.0001) (.0001) (.0001) (.0001) (.3955) (.0001) Path coefficients involving LB LNY to LB .699 .819 -.890 .757 . .539 -.749 LNY LF LB to LF .815 .577 .929 1.018 -.050 .715 (per capita (logit of father Effects of LNY income) proportion) Direct -.158 .092 .214 -.457 .179 .024 Via association vith LB .570 .473 -.827 .771 -.027 -.536 Control for sex Total .412 .565 -.613 .314 .152 -.512 Model 2 N 34 31 34 47 47 44 Mean of LF 186 -1.542 -1.271 .879 -1.755 -2.002 LET (logit of enrollment K2 .1872 .4239 .4036 .1645 .0662 .3720 rate) Prob F (.0814) (.0008) (.0001) (.0364) (.4099) (.0001) -Path coefficients involving LET LNY to LET .806 .831 .821 .720 .720 .717 0 LNY LF LET to LF .077 .106 .301 .058 .008 -.001 (per capita (logit of father Effects of LiY income) proportion) Direct .349 .476 -.860 .272 .146 -.511 Indirect via LET .062 .088 .247 .042 .006 -.001 Control for sex Total .411 .564 -.613 .314 .152 -.512 Model 3 N 34 31 34 47 47 44 Mean of LF .186 -1.542 -1.271 .879 -1.755 -2.002 LET R2 .5387 .5302 .5845 .6201 .0679 .5933 Prob F (.0001) (.0002) (.0001) (.0001) (.5799) (.0001) Path coefficients involving LB & LET LNY to LB .699 .S19 -.890 .757 .539 -.749 LNY LF LNY to LET .806 .831 .821 .720 .720 .717 LB to LF .899 .606 .964 1.058 -.050 .715 LET to LF -.349 -.116 .370 -.218 .007 .017 Effects of LNY LB Direct .064 .164 -.059 -.331 .174 .013 Indirect via LET -.281 -.096 .304 -.157 .005 .012 Via association with LB , .628 .496 -.858 .801 -.027 -.536 Control for sex Total .411 .564 -.613 .313 .152 -.511 Education: Table/3 .9 - 131 Model III brings together these interactive elements. Here we see much the same partial (derect) effects of base populations on fathers' occupa- tional distribution as in Model I. Among secondary students the indirect effects of income via LET are increased slightly for farm fathers and become mildly (but significantly) negative for white-collar fathers of secondary students. The indirect effects of income via LET partially neutralize the still dominant influence of base-population proportions on the "total effects" of per capita income. For university students results with Model III differ little from those with Model I. 3. Logit analysis of occupational selectivity. While the findings set out in Table 3.9 have implications with respect to associations between income levels and educational selectivity for occupation, they do not provide direct evidence on these relationships. Table 3.10 provides that evidence. Among secondary students per capita income has a consistent and highly significant negative (total) association with the differences between fathers' anrd base-population proportions not only for white-collar but also for manual occupations. For the white-collar families this implies an equalizing effect of income since the mean value of the difference is positive. A reduction in that difference implies a reduction in relative over-representation. For manual occupations the picture is more complex. Among the sets of secondary students the range in per capita incomes and economic structures is large, and the mean differential between logits of fathers' and of base- population proportions is close to zero. The negative total effect of income on selectivity from manual homes must be seen in the light of the fact that in less-developed societies men counted in a census as manual workers are commonly the urban manual wage earners in the "modern sector", and have - 132 TABLE 3.10 EFFECTS OF PER CAPITA INCOZ¶E AIND ENItOLLMENT KATE ON SELECTIVITY BY FATIIER'S OCCUPATION - LET (Logit of enrollment rati:) MUDEL LNY (In. per capita > LF-B (Logit of proportion of fathers Lncome) minus logit of prop,.LtLon of base popuilation) Equations inciude control variable for scx. White Collar Manual Agriculture Occupations Occupations and Related SECONDARY STUDENTS Number of observations 34 31 34 Mean LF-B 1.549 -.125 -1.391 R2 .5613 .4312 .5566 Prob > F (.0001) (.0006) (.0001) Path coefficients hivolving LET LNY to LET .806 .831 .821 LET to LF-B -.562 -.413 .396 Effects of LNY Uirect -.149 -.297 .388 Indirect - .453 .343 .325 Total -.602 -.640 .713 UNIVERSITY STUDENTS Number of observations 47 47 44 Msan LF-B 2.003 -1.526 -.931 R- .2901 .0754 .3366 Prob> F (.0007) (.3380) (.0003) i?ath coefficients involving LET LNY to LET .720 .720 . 717 LET to LF-B -.211 .017 .034 Effects of LNY Direct -.268 -.194 .503 Indirect -.152 .012 .024 Total -.430 -.182 .527 - 133 - a comparatively good status. For offspring of farmers, total effects of income are positive, presumably because of the rising status or economic complexity of that occupation with general economic advance. Higher enrollment rates have a generally equalizing effect for all occupations among the secondary students. Indeed, the association of rising enrollments with rising incomes accounts for three-fourths of the negative effect of income on selectivity from white-collar homes and for almost half of the positive total effect of income on representation of youth from farm familes. The seeming anomaly in the negative indirect effect through LET on representation froia manual families points once again to the non-comparability of the "manual" category as between many less-developed and the developed countries. A negative coefficient on LET and hence on the indirect effect of LNY via LET implies in part a reduction of the over-representation of "manual" workers in the lowest-income countries. But there is more to the story. Data for secondary students in the advanced countries usually refer to youth who have completed an advanced-secondary cycle; selectivity in favor of white-collar families unquestionably is strong in many European countries even though secondary enrollment rates are comparatively high. The fact that the proportion of workers who are in manual occupations declines at later stages of development is another piece of this complex picture. The pattern for university students is much simpler. Total effects of income are negative for white-collar and positive for farm families, imply- ing an equalizing trend in both cases. Enrollment rates in university have no direct effect (and hence contribute no indirect income effect) for either manual or farm families. There is a negative effect LET for representation - 134 - families; LET thus contributes somewhat over a third of the modest negative "total effect" of income in diminishing over-representation of white-collar families in universities. The negative effects of per capita income on selectivity from homes with manual fathers are negligible, but direct and total income effects on selectivity from farm hones are positive and strong enough to have a significant equalizing effect for farmers' sons and daughters. The most striking and unpredictable features of Table 3.10 are imbedded in the results for manual occupations (irrespective of the size of the coefficients) when compared with the other two categories of occupation. Only careful attention to distributions such as have been provided in scatter- grams along, with attention to country particularities that lie behind the statistical analysis can provide a basis for interpretation of that analysis or offer clues for probing into the diversity of potential paths in the democratization or the inegalitarian evolution of educational partIC.t-;dtion. - 135 - CHAPTER 4 COM4PARING SELECTIVITY ON FATHER'S EDUCATION AND FATHER'S OCCUPATION The importance of taking into account base-population structures in any interpretation of the commonly used indexes of social selectivity is reiterated throughout this work. Selectivity indexes are highly sensitive to the distribution of education and of occupations in the adult population. This being the case., there is only a limited possibility of making direct comparisons, even within the same country, between selectivity indexes for different aspects of status. Three sorts of comparison are helpful, however: these are used in this chapter. Ratios of observed selectivity indexes (S bs-1) to their upper limits as constrained by base populations (MaxS Bi-1) are examined in Chart 4.1 for white-collar fathers compared with fathers having at least some secondary education. (This comparison is made for secondary students only; data for such a comparison of university students were available for even fewer countries; even on secondary pupils the possible comparisons are few and limited entirely to less developed countries.) Fathers with white-collar jobs and with at least some secondary schooling were chosen because data are available on these and also.because some correlation betwe'-n them is reasonable a priori. The particular adjusted indexes control in an appropriate way for differences in the underlying structures of education and of occupation. The correlation exemplified in Chart 4.1 is consistent at least with the proposition that among the less developed countries selectivity on the - 136 - Chart 4.1 Ratios of (Sobs-1) to (Maxsbl-1); Fathers With At Least Some Secondary Education by White-Collar Fathers (Excluding Proprietors and Traders) White Collar Ratios 70 * Ghana/ GhsnJn/ Form 6/ O Ghana Ghana 30 Malay ed Chile Male * Female All 20 Chinese Malaysian Senegal * / O A 0 All Malaysian Malay 10 Chines Malaysian 0 10 20 30 40 50 60 7) Education Ratios - 137 - education base examined is associated with that on the white-collar base (exclusive of proprietors). This must be the case if there is a close cor- relation between father's occupation and schooling for the designated cate- gories. Turkey is exceptional in displaying higher selectivity by education than by occupation relative to the base-population constraints given the breaking point, taken here on education of fathers (at schooling beyond the elementary years). Indexes of dissimilarity allow for wider application in that they do not entail direct comparisons between particular categories. Limitations in the maximum values those indexes can take are largely insensitive to variations in the distributions of base population. 1/ Like selectivity indexes, they may. be mathematically constrained by enrollment rates: this constraint differs from MaxSBi, however, in that enrollment rates can be affected by the current educational policy whereas the schooling or occupation of the base population is essentially a given. The index of dissimilarity is an indicator that summarizes selectivity in a single figure, the upper bound of which depends only on the over-all enrollment rate. Since expansion of enrollment rates is in itself a component of the democratization of educational opportunity, use of the index of dissimilarity for comparisons provides a useful measure of situations in different countries (for given sorts of students) relating to various aspects of paternal status. Unfortunately, there are few countries for which data are available on both educational and occupational selectivity and for the same level of school. 1/ As was emphasized in Chapter 3 and in the discussion in Appendix D, the indexes of dissimilarity on occupational backgrounds are more sensitive to the classifications used in their estimation than are those on educational backgrounds because of the monotonic associations between enrollment rates and paternal schooling. - 138 - Secondary and- university students are compared for education and for occupation of father. On Chart 4.2, Part A, for paternal schooling, displays little correlation and the cases are few; it is Tunisian females and Malaysian youth that give the visual impression of positive correlation. The position of particular countries is interesting: Papua New Guinea is exceptional while Chile and Colombia are not out of line for their region, with a high index for university students and a comparatively low index for secondary school. Tunisian males resemble Latin American males, while Tunisian girls attending secondary schools have extraordinarily well-educated fathers. Part B of Chart 4.2, for occupations, looks quite unlike Part A, but the assortment of coun- tries contains many more from Europe. (The points for "full-time" students for East and West Germany seem implausible; notice the contrast with East German part-time students.) Overall one would expect the index of dissimilar- ity to be larger for university than for secondary students, as is generally the case for the occupation-based indexes. Direct comparisons between educational and occupational indexes of dissimilarity are presented in Chart 4.3 for secondary students. We observe a clear, quite unambiguous positive association. Countries that are low for educational selectivity of students are low also for occupational'selectivity, and vice versa. Japan (1966) and the United States (males born 1928&37) manifest exceptionally low selectivity. For university students there was a tidy enough monotonic ordering of the two sorts of selectivity in the four countries for which data permitted such a comparison, but this does not justify even.tentative generalization. Chart 4.2 Indexes of Dissimilarity; University by Secondary Students ;Malis *Famd i AiI A. COMPARISONS ON FATHER'S EDUCATION B. COMPARISONS ON FATHER'S OCCUPATION Secondary Students Socondary Students so so 70 70 6C -60 Ghani P a p a g_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __Pa p u a_ _ _'. 4C L ~~Papuo / 4] l //swe:nJ Puerto Rico 1944 Donmarklch OFvnce 0.1960 Norway 0 East Germany Part-Time -30 < -30 4 Chiron Malaysian l Chib1 ATunia MoyA * ChIlnsaIMalaysisn /IvoryCoast3 0 Hungarv 1963 a OMslaysi # 10 East Germany Japan Full-Tirne West Gerrany Full-Tifme 0 * 0 10 20 30 40 50 60 70 80 0 10 20 30 40 60 60 70 sO University Students Univrasity Students - 140 - Chart 4.3 Education and Occupation Indwxes of Dissimilarity Compared Secondary Students Occupational Indexes 80 70 A Male a Fele * Ghana 60 All Seneag / a Ghuan 40 /" Bei Chile a / Malay A Denmark A A Gh.w 30 A Malaysia Chin"se|- 0 Malaysian A Malay 20 / 10 - Japan A U.S. o I I 'I 0 10 20 30 40 50 60 70 80 Educational Indexes - 141 - The 1963 IEA study of mathematics achievement, undertaken in twelve economically advanced countries, provides some slightly different perspectives on educational and occupational selectivity into secondary schools. The samples, which were carefully drawn to be representative of the full range of secondary education in each country, were taken at two levels. The first was the level in which normal enrollment age was 13. Since educattion to that age and over is virtually universal in all the countries studied, the fathers of 13-year-olds constitute a good approximation to the base populations for youth still in school at the age of 17 to 18 - the age norm for selection of the second sample. The relevant data are set out in summary form in Table 4.1, where countries are entered in the order of the mean years of fathers education for the age-13 samples. There is amazingly little association between that ordering (column 1) and percentages of young people still in school at the agel of 17 in 1963 (column 5). The countries with the highest enrollment rates at age 17, United States and Japan, have low indexes of occupational dissimilarity (column 6) compared to the other countries in the study, but so does Belgium, with a 39 percent enrollment rate. England and Scotland are at the other extreme, reflecting the early ages at which children begin school and the small proportions of their youth who continue in school after age 15; there and in West Germany continuing one's education to the age of 17 or 18 signals university expectations to a degree not matched elsewbere even when large proportions of secondary-school graduates go on into higher institutions. Comparisons of column (2) with column (1) and of column (4) with column (3) provide another slant In looking at the selectivity of education in these countries. The mean schooling of fathers of the youth finishing secorndary school differed much less among countries than did the schooling of TABLE 4.1 BACKGROUNDS OF PUPILS AGE 17-18 AND AGE 13 COMPARED; IEA COUNTRIES, 1963 Index of Percentage of Fathers Percentage Dissimilarity Column (4)-(3) Mean Years of in Top Three Occupa- in School in Father's Divided by Father's Education tion Categories at Age 17 Occupation Column (2)-(1) Age 13 Age 17-18 Age 13 Age 17-18 Age 18 vs. 13 (1) (2) (3) (4) (5) (6) (7) Finland 11.2 11.7 24 34 23.5 35 40 Sweden 7.6 10.7 22 52 34.5 35 10 France 8.1 11.7 16 45 39.0 42 8 Net;herlands 8.7 11.1 23 53 33.7 42 13 Japan 8.9 10.9 19 40 56.5 24 10 W. Germany 8.9 12.5 17 68 53 14 Australia 9.3 9.8 20 39 15.8 40 Israel 9.7 11.0 14 25 8 Scotland 9.8 11.3 17 49 13.0 45 25 England 10.1 11.4 13 48 12.0 44 27 Belgium 10.4 10.9 20 33 38.8 21 27 United States 11.2 11.7 24 34 74.0 22 20 Source: T. Husen et al., Study of Achievement in Mathematics, 1967, vol. 2, p. 205. a/ Years in school for which ages 13 and 17-18 were the norm. In all the-se countries, enrollment at age 13 was universal. Education: Table/4 .1 - 143 - the fai:hers of the 13-year-olds, Finland and West Germany excepted. Whether one measures the differences in absolute years of schooling or in ratios of those years, the countries with the highest mean years of schooling in their base populations (fathers of the 13-year-olds) would appear generally to have L½er the least selective on father's education, but not necessarily in terms of fathers in relatively high-status occupations. Contrasts among countries are more easily seen in columns (3) and (4), which refer to the percentages of fathers in the top three of a longer scale of occupational categories. It is quite easy merely by glancing down this table to spot the countries with the most highly selective upper-secondary systems. West Germany is extreme in the magnitude of the differentials on both proportions of high-status fathers and mean years of father's schooling. England, France, and Scotland all display big contrasts in proportions of high status fathers, whether measured in absolute percentage differences for the two age groups (at around 30%) or in the ratios of the two proportions (high status fathers were three times as large a proportion among the older as among the younger pupils). Belgium and the United States turn up with by far the least occupational selectivity however one looks at the figures. The last column of Table 4.1 shows the differentials in percentages of fathers with high occupational status as a ratio to the differentials in mean years of fathers' education. This column says nothing about overall degrees of selectivity; instead it pinpoints differences in the relative importance of selectivity by father's education and occupation. Thus Germany, which has by far the most highly selective system measured by age contrasts in either mean schooling of fathers or proportions of fathers with high occupa- tional status, comes out in a median position in column (7); the two aspects of - 144 - selectivity would appear to operate together without any marked contrast in their relative importance. In Australia the contrasts in mean years of schooling are exceptionally small; as a result the moderate age 17-18 versus age 13 Australian difference in the proportion of high-status fathers consti- tutes by far the greatest occupational effect relative to effects of fathers' mean schooling. France shows up with the low ratio of 8 in column (7) primarily because of the relatively large difference in that country between mean years of schooling of fathers of the older compared with the younger pupils, not because of an exceptionally low differential on proportions of high status fathers. Isreal matches the French figure in column 7, but with a very different combination and generally lower selectivity; indeed, for reasons distinctive to that country the proportion of fathers with high occupational status is exceptionally small relative to mean educational attainments for both the older and the younger students. Finally, regression analyses were used to examine and compare effects of per capita income, enrollment rates, and the Gini coefficient for income distribution on the indexes of dissimilarity by education and by occupation for secondary and for university students. In this case there is no pairing of the data for the same country; the regressions are run on whatever data were available (Table 4.2). There can be no doubt of the negative effects of enrollment rates on both the education and the occupation indexes of dissimi- larity for secondary students, or of the fact that it is through effects of per capita incomes on enrollment rates that national per capita incomes have an overall negative effect on degree of selectivity. For university students, it is the direct rather than the indirect effect of per capita income that is important. TABLE 4.2 EFFECTS OF PER CAPITA INCOME, ENROLLMENT RATES, AND INCOME DISTRIBUTION ON INDEXES OF DISSIMILARITY BY FATHER'S EDUCATION AND BY FATHER'S OCCUPATION; SECONDARY AND UNIVERSITY STUDENTS IDX on Education IDX on Occupations Secondary University Secondary University Studenats Students Students Students Model 1 N 33 20 37 47 Mean IDX 37.2 41.7 35.1 44.5 ET (Enrollment rate) R2 e .2729 .3096 .3879 .3160 Prob F (.0163) (.0940) (.0003) (.0003) Path coefficients Involving ET LNY IDX LNY to ET .698 .896 .765 .643 (ln of (Index of ET to IDX -.641 -.153 -.563 -.232 per capita dissimilarity) Effects of LNY Income) Direct .215 -.286 -.068 -.264 Control for sex Indirect via ET -.447 -.137 -.431 -.149 Total -.232 -.423 -.499 -.413 Model 2 N 17 12 15 34 Mean IDX .345 .385 .335 .458 R2 .2317 .3655 .2246 .3304 GINI (income distribution) Prob F (.3183) (.2801) (.3765) (.0035) Path coefficients involving GINI LNY to GINI -.686 -.741 -.783 -.442 LNY IDX GINI to IDX .643 -.094 .671 .280 (ln of (Index of Effects of LNY per capita dissimilarity) Direct .339 -.491 .401 -.175 income) Indirect via GINI -.441 .070 -.525 -.124 Control for sex Total -.102 -.421 -.124 -.299 Education:Table/4.2 - 146 - The samples on which Gini coefficients were available were much smaller, and there is the further problem of the dating of those coefficients; it was necessary to assume stability for each country in the size of the Gini coefficient over the time period between the dates of the student surveys and the date for which a Gini coefficient on income distribution was at hand. There was a strong negative association between per capita income and the Gini coefficient in all samples, although this association was least strong for the sample of university students that included information on parental occupations. In the samples of secondary students, the negative association between per capita income and the Gini coefficient was matched by a positive association between per capita the Gini coefficient and the index of dissimilarity, controlling for income. This combination of relationships yielded an indirect negative effect of income via association with the Gini coefficient of the same order of magnitude as the negative indirect effects 'via enrollment rates. Overall, broad tendencies emerge clearly enough, but the associations between income and educational selectivity are mediated by enrollment rates and by the relatively egalitarian or stratified structure of the society in other respects, which include income distributions. Although paternal school- ing is not interchangeable with paternal occupation as an index of social status, in many societies there seems to be a pervasive ethos of high, inter- mediate, or low social selection for schooling. - 147 CHAPTER 5 DIFFERENTIAL ENROLLMENT RATES The discussion to this point has related most directly to two topics: 1) the origins of individuals who are receiving secondary or higher schooling, and 2) how representative these better-schooled are of their age cohort. The first question is answered by the profiles for schooling and for occupations of the fathers of students. The second question is addressed overall by the indexes of dissimilarity between distributions of characteristics of students' fathers and of comparable "base populations;" it is addressed for given categQries of paternal schooling or occupation by use of selectivity indexes and of differences in logits of proportions among fathers and the base popula- tion. Enrollment rates have been considered mainly with respect to whether they constituted constraints on the selectivity indicators; thus we have asked what part increases in overall enrollment rates might have played in reducing social selectivity by each of these measures. Differentials in enrollment rates by socio-eco.nomic background are of importance in themselves, however. These differentials put the equity question in a direct way: how do chances of obtaining a particular amount or level of schooling differ for individuals who were born into one or another category of parental status? This discussion of enrollment rates has four parts. To start with, the relationship between selectivity indexes and background--specific rates of enrollment is set out in general terms. Second, enrollment rates for specific socio-economic backgrounds are compared in terms of their absolute differences and the ratios of one rate to another; these two perspectives give very difFerent results. In the third section the use of differences in logits for such comparisons is explored. Finally, relationships between selectivity - 148 in enrollment rates and per capita national income are reviewed in terms of the question: are the the relative chances of differentially situated young people systematically correlated with per capita incomes of their society? A. Relationship between Selectivity Indexes and Background-specifc Enrollment Rates Given information on selectivity indexes and on overall enrollment rates in a particular level or type of education, it is a simple matter to derive background-specific enrollment rates. Let us define: P. . . . the number of pupils whose father are i pt . total pupils B. . . . number of base population who are i BL .. total base population E . . enrollment rate of children of B. E . . . overall enrollment rate Si . . selectivity ratio for category i We then have, by definition (1) E. = P./B. 1. 1 l. (2) Et = P /Bt (3)i / P. B t t Substituting (1) and (2) in (3) gives: (4) . Ei/Et, and hence (5) E = S.E In other words, all our Si are easily translated into Ei provided we put enough faith in the estimates of overall enrollment rates to use them as multipliers for this purpose. 1/ Clearly we get better Ei estimates the more appropriate are the base population figures used in the computation of selectivity indexes as well. 1/ Provided also that there are no substantial family-size biases relating to the various B.. On this and related problems concerning base popula- tions see AppenAix B and the discussion for Chile in Chapter 8. - 149 - Notice also that the ratios of selectivity indexes and of correspond- ing background-specific enrollment rates necessarily will be the same since Ei is obtained in each case by multiplying Si by the same overall enrollment rate E . This enables one to estimate relative enrollment rates from selectivity t indexes even when one lacks information about actual enrollment rates, either general or background-specific. Thus we can treat ratios of selectivity indexes as equity indicators of relative participation rates in secondary or in higher education. B. Ratio and Arithmetic Measures of Differences among Background-specific Enrollment Rates It is difficult to match enrollment rates of youth with the data available on their social backgrounds, especially for higher institutions. Unless level of school can be specified for a particular age cohort, reported enrollment rates can seriously distort the picture in some situations, and especially in some of the less developed countries. That limitazion was less serious in the analysis of constraints on measured selectivity (in Chapters 2 and 4) than it can be here. In the earlier analyses it was often sufficient merely to know that the overall enrollment rate was lower than the percentage of the base population in a given socio-economic category. In the nomenclature used earlier, Max SBi might clearly be lower than Xiax SEJ even when one did not know the value of E (here called Et). 1/ Analysis here is more sensitive, furthermore, to distortions in the student surveys, which often provide biased samples, especially of secondary students; here, as earlier, it has been necessary to adjust the overall enrollment estimates in some cases to exclude 1/ In the earlier analysis the subscrIpt j referred to level of school, whereas here E refers to the overall enrollment rate at any particular level of education, without specifying which level. Henceforth E will be used since there is no need to introduce a subscript to identify the level of education under consideration. - 150 - certain streams where the secondary system is highly differentiated; this problem arises primarily in some of the countries of Europe. Distortions in the data on this account appear to have had surprisingly little effect, how- ever, however, on selectivity rankings of the countries. Estimates of background-specific enrollment rates for secondary students at the levels of secondary school covered by the surveys are given in Appendix Tables L.1 and L.2 for father's schooling and father's occupation. One ca7 quickly discern that in some countries the arithmetic differences in rate by parental background are large, in oth-rs quite small. These differ- ences are one way of measuring the avantage in educational chances of coming from a well-educated rather than a poorly educated family, or from the home of a farmer rather than a manual worker or a while-collar man, as the case may be. A second measure, which is the equivalent of taking ratios of selectivity indexes, i3 to ask how many times greater are the chances of attending secondary school if one comes from background A as compare.d with background B. A third measure is the differences in logits of enrollment rates between youth from background A versus B. All three sets of measures are shown in table 5.1 for secondary enrollment rates of children of men with at least some secondary schooling compared with children of men who had no secondary education and with children of men who were totally unschooled. The disparities in rankings with the first two measures are dramatically illustrated in Chart 5.1. 1/ Relationships be- tween the two measures are not, however, simply random. The southeast quadrant 1/ A logarithmic scale is used for the ratios to provide better spacing on the chart. This gives a visual comparison of arithmetic differece with ratios, but does not alter the rankings of ratios relative to arithmetic differences. - 151 - TABLE 5.1 COMPARISONS OF BACKGROUND - SPECIFIC ENROLLMZNT RATES BY FATHER's EDUCATIONS SECONDARY STUDENTS Atleast Soso Secondary Atleast Same Secondary and No Schooling and No Secondary Arithmetic Logit Arithmetic Logit Difference Ratio Difference Difference Ratio Difference AFRICA Ghana 1961 All 30.0 28.3 3.542 29.0 14.8 2.961 Male 39.6 19.9 3.419 38.0 11.3 2.375 Female 21.3 214.0 4.516 20.9 42.8 3.665 Ivory Coast 1963 All Male Female Kenya 1961 All 15.7 11.5 2.500 13.7 4.9 1.701 Male 19.9 7.9 2.232 16.8 3.1 1.507 Female 20.3 204.0 4.456 18.8 12.8 2.649 Kenya 1968 Male 40.5 15.0 3.181 37.9 7.9 2.545 Malt 1964 All Sonegal 1967 All 16.6 9.7 2.371 16.0 7.4 2.114 Male 19.4 -7.3 2.151 18.6 5.8 1.925 PFeale 13.9 24.2 3.061 13.4 13.2 2.541 Tanzania 1980 All 29.9 25.9 3.467 27.2 8.0 2.364 Uganda 1969 All 29.2 7.5 2.340 26.2 4.5 1.814 Zaire 1972 All 74.4 107.3 5.805 72.1 25.0 4.513 MEDITERRANEAN Tunisia 1963 All 25.4 8.5 2.362 24.3 6.4 2.084 Male 29.7 5.6 2.074 28.1 4.5 1.843 Femle 21.1 71.3 4.003 20.5 23.8 3.209 Turkey 1962 All 21.3 36.5 3.558 20.4 14.6 2.796 Male 29.6 25.7 3.453 27.9 10.6 2.640 Femle 10.6 107.0 3.704 9.9 13z4 2.493 SOUTEAST ASIA W. Malaysia 1967 All 10.2 4.9 1.645 5.8 1.8 0.655 Papua New Guinea 1975 All 42.0 7.9 2.628 30.5 2.7 1.459 Male 42.8 5.9 2.390 29.5 2.3 1.319 Femle 40.5 10.6 2.871 32.1 3.5 1.711 SOUTH AND CENTRAL AMERICA Argentina All * * * 20.5 3.8 1.567 Bolivia 1970 Alt 64.1 92.6 5.314 60.1 13.8 3.578 Chile 1970 All 18.6 4.0 1.606 11.2 1.8 0.739 Colombia 1974 All * * * 24.9 4.0 1.676 Guyana 1959 All * * * 23.1 3.5 1.533 Male * * * 24.9 2.8 1.358 Female * * * 21.3 5.8 1.979 Mexico 1974 All * 43.3 34.3 3.973 Netherlands Antilles 1959 All * 15.3 3.5 1.410 Femle 40.5 10.6 2.871 32.1 3.5 1.711 SOUTH AND CENTRAL AMERICA Argentina All * C * 20.5 3.8 1.567 Bolivia 1970 All 64.1 92.6 5.314 60.1 13.8 3.578 Chile 1970 All 18.6 4.0 1.606 11.2 1.8 0.739 Colo-bla 1974 All * C * 24.9 4.0 1.676 Guyana 1959 All * * * 23.1 3.5 1.533 Male * * 24.9 2.8 1.358 Femle C 21.3 5.8 1.979 Mexico 1974 All * 43.3 34.3 3.973 Netherlands Antilles 1959 All * 15.3 3.5 1.410 Male 18.2 3.2 1.366 Female 12.4 4.2 1.530 Paraguay 1974 All 39.7 10.9 2.879 Peru 1974 All 5 53.4 7.1 2.814 Trinidad 1974 All 30.0 4.7 1.915 EUROPE, UNITED STATES, JAPAN Denmark 1964 All * 16.1 2.1 0.950 al e 18.1 2.2 1.014 Feale 13.7 2.0 0.869 Finland 1965 All * 9.9 1.8 0.707 Male 10.4 2.1 0.845 Female * 9.3 1.6 0.597 United States 1953 Male (21.2) (1.5) (0.865) Japan 1966 All * 26.2 1.3 1.265 Male 24.0 1.4 1.166 Femle * 28.3 1.5 1.357 Chart 5.1 Comparisons of Secondary Enrollment Rates of Children Whose Fathers Had at Least %'ome Secondary Schooling and of Those Whose Fathers Lacked Such Schooling; Ratios of Rates by Arithmetic Differences Ratios of Rates 50.0 *A 40.0 1 Mexico- 30.0 0 Zaire 20.0 - A ° O A ai Boliva A A A ~ A 10.0 AAu D A AA O Peru A c AO 5.0 j GA A CA 4.0 OI DA AAal 3.0 - A Maio A * Female A CI All AW AW A 2.0 5 oC a W 33 W *W ~U.S. A OJ 1 .0 1 I I I I p 0 10 20 30 40 50 60 70 80 Arithmetic Differences in Percentage Rates - 153 - is empty; one does not find large arithmetic differences between enrollment rates for children of the relatively educated and the less educated in com- bination with low ratios of those rates. In fact there are only four cases with arithmetic differences in rates exceeding 40 percentage points: Mexico, Peru, Bolivia, and Zaire, all in the Northeast quadrant. Such a situation indicates a marked bifurcation in the society, with attendance in secondary school generally expected for youth whose parents had (in this case) at least some secondary schooling, but with continuation into secondary schools extremely rare among the rest of the population. The northwest quadrant of the chart is inhabited mainly by girls, though there are girls in the southwest quadrant as well. The relatively low arithmetic differences for girls are inevitable when attendance of girls in secondary school is not yet widely accepted even by better schooled parents. If virtually no girls from less educated homes attend secondary schools the ratios of enrollment rates from the more favored to those from the less favored backgrounds can be high even with low arithmetic differences. This is the situation in a number of African countries, both north and south of the Sahara, but there are also African and other less developed countries with only moderate ratios. The lowest part of the southwest quadrant is a very different story, however; both ratios and arithmetic differences here may be low because secondary-school attendance has been spreading throughout the population, although there can be special exceptions. 3/ 3/ All the entries in Chart 5.1 with ratios below 2.2 are European, the United States (males in 1953), or Japan, with two exceptions: Chile and Malaysia. -154- . Malaysia is an intresting case. The drive to equalize opportuni- ties within and between ethnic groups in Malaysia is clearly refl:-ted in the position of that country near the extreme southwest corner of Chart 5.1. Only data for all ethnic groups combined in Malaysia were plotted, but details by ethnic groups reinforce this picture. The Chinese Malaysians were the least differentiated by parental schooling in secondary enrollments for males, and virtually so for females also. Even among the Malay males the desparities in enrollment rates in 1972 were not high (on Chart 5.1 they would have been entered with a ratio of 2.1 and an arithmetic difference' of 11.0); class differences are more important among Malay girls, but are moderate compared with Muslim girls in many other countries. The Japanese situation as of 1966, while also egalitarian, is very unlike that in Malaysia. The Japanese (along with males in the United States in the early 1950's) had the lowest ratios of enrollment rates though not the lowest arithmetic differences in enrollment rates among the countries repre- sented in Chart 5.1. By 1966 almost two thirds of the youth in this education- centered society were completing upper-secondary school, and secondary-school attendance was spreading in epidemic proportions from children of the better schooled on through the ranks of those with less schooled parents, Today completion of upper-secondary school is almost universal, but in 1966 a substantial minority of the children of men lacking education beyond the compulsory 8-9 years still were not entering senior secondary schools. It is at that late stage in the thrust toward universal secondary education that Chart 5.1 records the selectivit:y of secondary enrollments for Japan. That history explains also the approximation to maximum enrollment constraints on educational selectivity at secondary level observed for Japan in Chapter 2. - 155 - Variations in occupation-specific secondary enrollment rates are more complex than those by father's education in that educational categories have an intrinsic monotonic ordering, which is not true for occupations. Moreover, as has been stated before, the countries covered for occupational backgrounds are quite different from those covered on paternal education, which contributes also to giSling a different impression; in particular, Latin America was better represented in the data on selectivity by paternal education whereas "Europe" is better represented in data on selectivity by occupation. The effect of including proprietors and traders with "white collar" or of excluding them is marked for most of the African countries and especially for West Africa, but ccmparatively minor for the other countries. Proprietors and traders, mainly traders in West Africa, include large numbers pursuing traditionai activities tt'1at generally lie outside of the "modern sector" and account for a larger fraction of t:he base population than in the other coun- tries for which data were available. The relatively high status of manual wage wbrkers in urban Africa compared with the average peasant was mentioned in Chapter 3; this shows up more clearly in the high white-collar to farm secondary enrollment ratios compared with the white-collar to manual ratios in most of Africa (Table 5.2 part A). Indeed, for two African countries, Mali and Senegal, the enrollment rates for children of white collar men inclusive of proprietors and traders were lower than those for children of manual workers (Table 5.2 part B). There is a clear negative relationship for African countries between ratios of an arithmetic diffeences in enrollment rates of children of men in manual and in white-collar employment, whereas the two measures tend to be positively - 156 - TABLE 5.2 COMPARISONS OF BACKGROUND - SPECIFIC ENROLLMENT RATES BY FATHER's OCCUPATION; SECONDARY STUDENTS A. White Collar Excluding Proprietors and Traders tWhte Collar and Manual White Collar and Farm Ari5hiuet±c Logit Arithmetic Logit Difference Ratio Difference Difference Ratio Difference AFRICA Ghana 1961 All 15.8 10.9 2.454 15.9 11.6 2.513 Male 21.7 8.8 2.360 21.6 8.4 2.326 Female 10.3 21.6 2.864 10.6 54.0 3.427 Ivory Coast 1963 All 41.3 10.8 2.903 41.4 11.1 2.927 Male 44.3 8.6 2.760 44.3 8.6 2.760 Female 29.7 19.6 3.220 30.9 78.5 4.331 Mali 1965 All 27.3 3.5 1.594 38.1 128.0 4.826 Niger 1967 All 47.6 4.4 2.278 60.7 76.9 5.065 Hale 53.4 3.7 2.395 71.4 48.6 5.047 Female 43.4 6.4 2.476 51.2 257.0 5.578 Senegal 1970 All 39.8 3.0 1.775 59.3 119.6 5.355 OTHER LDC's Turkey 1962 Lyce All 47.6 7.1 2.663 53.7 32.6 4.164 Malaysia 1967 All 10.5 2.6 1.062 13.0 4.2 1.555 Male 11.4 2.3 0.970 14.2 3.5 1.389 Female 9.1 3.1 1.191 11.1 5.6 1.782 Thailand 1978 All 78.2 10.4 4.226 79.1 11.7 4/347 Chile 1970 All 40.5 3.6 1.922 52.7 16.5 3.536 EUROPE, JAPAN, AND U.S.A Belgium 1963 All 9.0 1.3 0.378 4.7 1.1 0.194 Finland 1963 All 32.8 3.3 1.667 24.2 2.1 1.094 France 1959 All 64.3 3.7 3.158 77.9 8.4 4.142 Spain 1970 All 63.9 3.6 3.124 68.8 4.6 3.411 Sweden 1963 All 57.1 4.4 2.629 61.6 6.0 2.991 Japan 1963 All 44,.3 2.5 1.897 19.1 1.4 0.838 U.S.A. 1963 All 19.2 1.3 0.972 -12.6 0.9 -1.253 AFRICA Ghana 1961 All 12.i 8.7 1.889 12.6 9.1 1.916 Male 18.0 7.5 1.956 17.9 7.1 1.914 Female 7.3 16.3 1.855 7,5 35.1 2.048 Ivory Coast 1963 All 14.4 4.4 1.485 14.6 4.5 1.511 Male 19.6 4.4 1.584 19.7 4.4 1.590 Female 12.6 8.9 1.901 13.7 31.9 2.499 Kenya 1961 All 10.8 3.5 1.343 12.6 6.0 1.873 male 13.6 2.7 1.085 17.6 5.4 1.706 Female 7.9 11.9 1.801 7.6 3.9 1.667 Mali 1965 All -2.9 0.7 -0.305 7.9 27.8 2.041 Niger 1967 All 8.0 1.6 0.528 21.1 26.0 2.756 male 8.1 1.4 0.439 26.2 19.1 2.767 Female 8.4 2.0 0.750 16.2 72.1 2.829 Senegal 1970 All -2.8 0.9 -0.176 16.7 32.8 2.662 OTHER LDC's Turkey 1962 All 37.6 5.8 2.184 45.2 26.7 4.280 Malaysia 1967 All 8.7 2.3 0.867 11.1 3.8 1.287 Male 10.1 2.2 0.839 12.8 3.2 1.215 Female 7.3 2.7 0.938 9.3 4.9 1.422 Th#iland 1978 All 68.3 9.2 3.456 69.2 10.3 3.568 Bolivia 19'5 All 61.6 4.7 2.813 71.5 11.6 3.744 Chile 1970 All 6.9 1.4 0.401 19.S 4.9 1.669 Puerto Hlcq 1960 All 54.7 3.8 2.410 65.6 8.3 3.275 EUFOPE, JAPAN, AND U.S.A Belgium 1963 All 8.5 1.2 0.349 3.5 1.1 0.139 Denmark 1964 All 36.9 7.3 2.347 24.4 2.3 1.167 (gyan.) Male 20.2 6.8 2.154 21.5 2.6 1.1'78 Female 50.0 7.9 2.713 30.2 2.1 1.253 Finland 1963 All 25.3 2.8 1.327 16.8 1.7 0.778 France 1959 All 52.1 3.2 2.254 65.7 7.3 3.202 Hungary 1963 All 38.6 2.8 1.669 42.6 3.5 1.914 Spain 1970 All 61.9 3.6 2.878 66.7 4.4 3.152 Sweden 163 All 55.7 4.7 2.543 56.3 4.9 2.585 U-S.A. 1953 Male 13.6 1,2 0.630 28.7 1.6 1.232 1963 All 21.1 1.3 1.071 -11.0 0.9 -1.069 Japan 1963 All 44.2 2.5 1.859 20.5 1.4 0.877 - 157 - associated for the other countries. As of 1963 Belgium and the United States were distinctively egalitarian among this sample of countries in the repre- sentation of children of manual workers in secondary education, whereas despite its educocentric character, Japan displayed substantial arithmetic differences. (In 1963 manual workers in Japan still had a low representation in entry to the secondary schools.) The extreme case in arithmetic contrasts between enrollment rates for children of white collar and of manual fathers was Thailand (1978), which had been experiencing rapid expansion in higher education, but where secondary enrollment rates were still at 20 percent or less of the age cohort. In brief, each country in a comparison such as this is at a single point in a changing scene, with some countries shifting more rapidly than others and each following its own development path despite commonalities in expansion processes. Table 5.2 brings out the sharp contrast between some extremely high African ratios of secondary enrollmernt rates of children of white collar men relative to farmers and the generally more moderate ratios in the economically advanced counries. The salient result for the less developed countries, however, may be their diversity. There seems to be a fairly systematic positive correlation between the ratio and afithire-i measures for the West and Japan, which is perhaps to be expected with higher enroll- ment rates; by contrast, no such relationship appears for the less-developed countries, where a small white-collar population can send large proportions of its children to secondary school before secondary education spreads to - 158 - the rural population. The entry for the United States is extreme in displaying higher representation for farm than for white-collar families 3/. The estimation of background-specific enrollment rates in higher education is hampered by the unreliability of the over-all estimates of enrollment rates at the tertiary level. This problem is most severe for the sample of countries that give information on education of parents of univer- sity students. In the African countries in particular, enrollment rates in higher institutions are extremely low. There can be no-question about the general orders of magnitude of these rates, but converting them into estimated background-specific rates is another matter, given the high sensitivity of such estimates to small differences in overall enrollment percentages. This being the situation, I have limited estimation in that sample to ratios of background-specific enrollment rates, which can be derived directly from ratios of selectivity indexes without regard to absolute enrollment rates. The various ratios range widely (Appendix Table L.3). For the comparison of origins "at least some secondary" and no schooling" they range from lows around 2.0 for Papua New Guinea to over 100 in Portugal and for Tunisian females. Occupational background-specific rates of enrollment in higher education can be estimated with somewhat greater reliability, given the country coverage in the student surveys. Table 5.3 shows these estimates in comparisons using arithmetic differences, ratios, and differences in logits of enrollment rates. (More detail is provided, again, in Appendix Table L.4). 3/ For some countries, including the United States, there is probably a bias in 'Pqvor of farm representation in th'Ja figures because farm families tend to be large compared with urban families and the base population refers to potential fathers unweighted by the numbers of their children. The general pattern would not be altered, however, by adjustments on this account. - 159 - TABLE 5.3 COMPARISONS OF BACKGROUND - SPECIFIC ENROLLMENT RATES BY FATHER's OCCUPATION; UNIVERSITY STUDENTS (White Collar Includes Proprietors) White Collar and Manual White Collar and Farm Arithmetic Logit Arithmetic Logit Difference Ratio Difference Difference Ratio Difference AFRICA Kenya '70 All 11.0 200.0 2.581 10.6 27.5 2.260 Ghana '70 All 5.2 12.2 1.573 4.9 7.0 1.359 Ivory Coast '75 All 4.1 3.9 1.032 4.5 5.5 1.231 Tanzani,e '75 A1- 3.9 19.9 1.478 3.8 12.6 1.383 MEDITERRANEAN Egypt '62 All 26.9 27.7 2.981 27.4 51.; 3.250 Greece '61 All 9.4 4.3 1.335 9.3 4.1 1.308 '80 Al 21.0 2.6 1.198 20.7 2.6 1.167 Portugal '63 All 33.1 48.1 3.418 18.0 2.1 0.964 Spain '62 All 20.1 25.8 2.718 20.2 27.7 2.751 Male 25.3 28.2 2.939 24.9 19.8 2.749 Female 15.5 43.7 2.683 15.5 44.5 2.688 Tunisia '65 All 5.8. 9.3 2.068 6.0 13.0 2.321 Male 8.6 7.9 1.666 9.0 10.8 1.834 Female 2.9 26.6 1.315 3.0 48.6 1.363 Yugoslavia '65 All 12.0 3.0 1.128 15.3 6.5 1.786 OTHER LDC S. Lorea '70 All 7.5 3.2 1.029 8.8 5.2 1.658 Papua New Guinea '75 All 14.8 2.9 1.163 21.0 15.1 2.483 Male 21.2 2.7 1.216 30.9 11.8 2.579 Female 5.5 3.3 1.015 7.6 27.7 2.011 Mexico '50 Male 16.7 19.3 2.457 17.4 89.6 2.934 Perto Rico '60 All 25.9 6.0 1.964 10.7 1.5 0.552. EASTERN EUROPE U.S.S.R. '64 All 10.0 2.T 1.005 9.5 2.5 0.928 '70 All 17.6 3.2 1.300 19.1 4.0 1.498 E. Germany '67 All 5.4 1.4 0.376 9.7 2.0 0.760 Hungary '62 All 16.1 5.4 1.676 17.1 7.6 1.946 Poland '74 All 20.4 2.0 1.300 25.3 5.9 1.938 WESTERN EUROPE Austria '65 All 17.0 35.6 2.701 16.1 12.4 2.211 Male 25.4 35.4 3.039 24.0 12.3 2.439 Female 8.6 39.8 2.176 8.3 15.2 1.915 Belgium '67 All 41.4 18.3 3.379 34.9 4.9 2.058 Male 45.0 14.8 3.065 34.2 3.4 1.683 Female 37.9 28.5 3.309 35.6 10.6 2.604 Denmark '64 All 11.6 4.6 1.444 6.0 1.7 0.545 Male 15.9 4.2 1.487 9.3 1.8 0.668 Female 7.3 5.2 1.375 3.0 1.5 0.395 Finland '68 All 27.3 4.4 1440 9.5 1.4 0.438 France '50 Al' 17.7 37.5 2.755 16.2 8.8 2.015 '65 All 23.2 9.0 2.208 20.1 4.3 1.589 Male 27.0 8.5 2.247 23.5 4.3 1.648 Female 19.4 9.6 2.156 16.7 4.4 1.526 W. Germany '65 All 8.4 5.4 1.449 5.6 2.2 0.744 Male 13.0 4.9 1.530 7.5 1.8 0.647 PFmle 3.7 8.2 1.268 2.5 2.5 0.688 1970 All 19.9 8.9 2.126 16.1 3.5 l1346 Ireland '63 All 15.9 10.0 2.085 12.7 3.6 0.438 Natherlands '64 All 12.5 10.6 1.996 8.5 2.6 0.945 Male 18.8 7.7 1.993 12.1 2.3 0#905 Femle 5.5 41.-6 1.824 4.7 6.0 1.281 Norway '64 All 32.9 5.7 2.062 31.4 4.7 1.869 Sweden '60 All 20.9 8.4 2.106 17.3 3.7 1.412 Switzerland '60 All 33.5 11.6 2.623 32.5 8.8 2.398 U.K. '61 All 20.1 7.1 1.971 a a a Male 28.8 6.8 2.118 a a a Female 11.3 8.3 1.811 a a a '78 All 37.9 6.5 2.280 a a a Japan and U.S.A Japan '53 All 20.0 16.4 2.888 16.6 4.S 1.549 Male . 29.1 14.3 2.664 23.0 3.8 1.529 Female 11.2 29.3 2.315 10.4 9.7 1.857 '68 All 29.2 4.2 1.743 29.6 4.4 1.788 U.S.A. '53 Male 24.6 2.1 1.107 30.7 3.0 1.490 Logits stabilized at extreem by taking in P+.002 1-P+s.002 -160 - As was demonstrated in Chanter 3, the representation in higher in- stitutions of children of farmers relative to children of manual workers varies greatly among countries. Represnetation of children of farmers in the universities is higher both absolutely and compared with offspring of manual workers the more economically advanced a country -- a reflection mainly of changes in the economic nature of farming. There are also marked differ- ences among countries in the characteristics of "proprietors and traders," as was the case for secondary students as well. This hetrogeneity is reflected in differences between the last two columns of Appendix Tables L.2 and L4, which also show incidentally one of the limitations of our data for dis- tinguising "proprietors" from "white collar" fathers (and base populations). What cannot be seen directly in this table are the difficulties encountered in separating out the proprietors and traders even in many of those cases in which this was done. In view of these problems, Table 5.3 uses the combined white-collar and proprietors category throughout for coG parisons with manual and with farm origins. (The reader can easily identify from Appendix Tables L.2 and L.4 the countries for which choice of white collar inclusive versus exclusive of traders makes the greatest difference.) Overall, there would seem to be virtually a random relationship between the arithmetic and ratio indicators shown in Table 5.3, but some systematic country and sex contrasts lie back of these loose patterns. IThite-collar/manual ratios tend to be lower in Eastern Europe and the Soviet Union than in Western Europe, with East Germany at the lower extreme. There are overlaps with other regions in the white-collar/manual ratios, however, with a range outside of Eastern Europe&from, 2.1 in the United States back even in 1953 to 48 in Portugal (1963). Arithmetic - 161 - differences between white-collar and manual enrollment rates range from 3 and 4 percent (Tanzania 1975 and Tunisian females 1965) to 45 for Belgian males in 1967. (Notice the sharp contrast with the low selectivity observed in Chapter 3 with resperat t-o secondary education in Belgium when all secondary students were taken together, without regard to possible selectivity among secondary streams.) Combinations of high ratios and low arithmetic differences in enrollment rates are predominantly for women, f'rom whatever part of the world, reaflecting their generally lower rates of continuation into higher education and the extreme selectivity in some cases of the few who do enter such institu- tions; however, ratios for women are not invariably high. The overall distributions of white-collar/farm ratios and arithmetic differences differ sharply from white-collar/manual comparisons in the large proportion of countries in which ratios of enrollment rates are around 2.00 or less and the corresponding (but less obvious) smaller fraction of countries in which the ratios approach 10.00 or more. Thiis is primarily a manifestation of agricultural progress and efficiency in the West, and the associated participa- tion of farmers' children in higher education -- both for those who may return to farming and those who enter other spheres of activity. It is not surpris- ing that in the comparisons between white-collar and farm enrollment rates East European countries lose the position they took in the white-collar/manual comparisons. The lowest white-collar/farm ratios and the lowest arithmetic differences show up in Western Europe. The highest ratios appear in Mexico, Egypt, and Spain (and for females again in Tunisia). The diversity in these findings has some systematic underlying elements, but this does not alter the fact that there are important varia- tions independent of levels of economic development. - 162 - C. Measuring Differences in a Developmental Perspective That ratios and absolute differences may gi-e different rankings for degree of selectivity is to be expected. But is there a better way to make these comparisons? Can we look at background-specific rates of enrollment more explicitly in a context of development? Many people are familiar with the use of cumulative S curves for depicting growth. S curves can differ in steepness, in their calendar scale, in their asymptotic limits. S curves may be used for whole populations or for subpopulations: e.g. children from different status categories of families. Different subpopulations would be expected to stand at different positions in development and to occupy separate positions on an S surve. It can be useful to identify equal distances along the S c-urve in terms of a linear transform. Among the much-used linear transforms of S curves are. the probit function (linear in standard deviation units of a cumulative normal probability distribution) and logits; these are specified mathematically in Appendix E. Logits are the logarithms of the "better's odds." If we designate the enrollment rate of children of white-collar men as E , the better's odds of being enrolled are then E w/(I-Ew), and the logit is ln(E w/l-Ew ). The possible values range from minus to plus infinity with a value of zero when the odds are even, implying a group enrollment probability of 50 percent. Consider an analogy in epidemiology. At first few acquire the disease, but with increasing exposure the likelihood of infection rises and the proportion infected rises at an icnreasing pace; the pace slows down when the most resistant individuals are reached. Similar patterns occur for diffusion of a cultural or social practice, such as attending secondary school. Accelerating and retarding factors are of different magnitudes for different subpopulations and the factors may be differently timed, with the result that we observe leading and lagging individuals and sets of individuals. - 163 - The logit transform takes account of the greater difficulty of a given absolute percentage change at either the initial or the final stage as compared with the central range of a cumulative distribution. The difference between the logits for two enrollment rates indicates the developmental gap between two subpopulations: e.g. children of white-collar workers and of farmers. Two further steps were taken to compare results on different measures of selectivity in enrollment rates. First countries were ranked by absolute differences in percentage rates, by ratios of those rates, and by differences between the logits of the same two rates. In all comparisons by father's education and in most of those by father's occupation, the rank on differences in logits falls between the rankings on the other two measures or coincides with one of them. The second step was the charting of differences in logits against ratios of enrollment rates, the latter again on a logarithmic scale. Visually, in other words, this is a comparison between the logarithms of ratios and differences in logits. For comparisons by father's education, those two measures were closely related with only minor deviations from the main linear pattern. Comparisons of university enrollmedt rates of children of white- collar men and of farmers also displayed a systematic positive relationship of the two indicators throughout, though not as tight an association as that observed in comparisons of seconsiary en:ollment rates by father's education. An analysis analogous chart comparing children of white-collar men and of manual workers spread out in the higher ranges of ratios of enrollment rates, primarily though not solely because at ratios above 10.0 the logit differences for females were much smaller than those for males or those that did not distinguish by sex. A mathematical relationship between the logarithms of the ratios and the differences in logits lies behind these patterns. This can be stated - 164 - simply, i.n terms of the source of deviations between the two indicators. Let us designate the enrollment rate for children of white-collar men as E and w and that for children of farmers as Ea and use the same subscripts on the selectivity indexes. We then arrive (as shown in Appendix E) at the equation: Logit E - logit E = [ln S - lnS ] - [ln(1-S E ) - ln(1-S E Xi (6) w w a w t a t But the first term on the right side of equation (6) is the logarithm of the ratio df S to S , which is equal to the logarithm of the ratio of Ew to E since these ratios are indentical, as was shown earlier. Furthermore, S Et = E and Sw Et = E . Substituting in (6) we then have Logit E - logit Ea = ln(E /El + [ln(l-Ea) - ln(l-Ewl (7) - 165 D. Per Capita Income and Selectivity in Enrollment Rates Pursuing a developmental pespective, differences in logits of enrollment rates may be compared with per capita incomes for the appropriate dates in each country. There is a moderate negative association between selectivity so measured and GNP for secondary enrollment rates of children of men with at least some secondary educatiorn and of men lacking any such schooling. That relationship is shown in Chart 5.2. Unlike previous charts, this one (and Chart 5.3) omits entries for the two sexes combined when data are available separately by sex. (In reading this chart it is possible in most cases to identify male and female for the same country since the two entries will be in the same position with respect to per capita GNP.) The main outlying case is for Mexico, with exceptionally high selectivity relative to income. Generally there are wide variations in degrees of selectivity in the lower-income countries whereas for this limited sample the variations narrow to a tidy negat4ve slope above a per capita income of $1,000 (Mexico excepted). A comparison of secondary enrollment rates of children of men having at least some secondary education with children of men with no school- ing at all also shows that the disperity has a negative association with per capita income (chart 5.2) but data on enrollment rates for children of totally unschooled men are not available at higher per capita incomes. Selectivity of enrollment rates for children of white-collar men compared with children of manual workers (not shown) displayed no association - 166 - with per capita income when the measure of selectivity used waa difference in logits. At every per capita income level the variations in logit differences was high and for university enrollments very close to the variation for the entire sample. It is evident enough that selectivity between white-collar and manual families for enrollment into secondary or higher education depends primarily on factors other than per capita incomes. Differences between the logits of enrollment rates of children of white-collar men (excluding proprietors and traders) and of farmers tell a different story. For participation in secondary schools (shown in Chart 5.3) there is a negative association between the differences in logits and per capita income. There is also a marked gap however, between a few economically advanced countries with large differences in logits relative to their per capita incomes and other countries (including Japan and the United States) in comparable income brackets. The scatter in Chart 5.3 takes on quite a different appearance when white-collar is defined more broadly to include all who are not farm or manual. The main effect is a clustering of the countries of Sub-Saharan Africa in the middle range on differences in logits, refUecting relatively low representation of children of African traders in the secondary schools along with the importance of small-scale proprietors and traders in the base populations of the African countries included. Although the scattergram using the broader definition includes all the countries shown in Chart 5.3 along with several other African (and advanced) countries, none in Africa displayed a logit difference above 3.0. By contrast, as is readily seen, most of the A's in Chart 5.3 are above 3.0 and they run up to values of around 5.0, an extremely high figure for a difference in logits. - 167 - Chart 5.2 Logit Differences in Secondary Enrollment Rates of Children of Men with at Least Some Secondary Education and Children of Men with no Secondary Education, by Per Capita Income Differences in Logits 5 0 4 L a * A Malel L All l 3 a0 2-A & O -1 - - ' I 'I O A 100 200 300 500 1,000 2,000 3,000 5,000 10,000 Per Capita GNP (1977 U.S. Dollars) - 168 - Chart 5.3 Logit Differences in Secondary Enrollment Rates of Children of White-Collar Men and of Farmers, by Per Capita Income Differences in Logits A G1 A 5. MA A* 4 3 W A A 2 1 A Male * Fernale 0All ow 0 oJ -1 . I 1 .II,; 100 200 300 500 1,000 2,000 3,000 5,000 10,000 Per Capita GNP (1977 U.S. Dollias) U.S. 1963 - 169 - Chart 5.4 Logit Differences in University Enrollment Rates of Children of White-Collar Men (Including Proprietors) and Children of Farmers, by Per Capita Income Differences in Logits of Enrollment Rates 5 A Male Female A l 2 - * a O O .A a a 0O O 0O a O O 100 200 300 500 1,000 2,000 3,000 5,QOO tO,OOO Per CGapita Income (1977 U.S. Dollars) - 17.0 - Chart 5.4, which refers to university enrollments, uses the broader definition of white-collar and looks somewhat more like an analogous scattergram for secondary students using that definition (though the logit differences in the middle and higher income ranges for secondary students displayed a wider spread than is indicated in Chart 5.4). In any case, no direct comparisons between the scatters for secondary and for university students are justified given the small overlap in countries represented. Most of the African countries drop out and both Eastern and Western Europe are relatively well represented in Chart 5.4. In all variants bn definitions of white collar and for secondary or university students some negatiye relationships between the logit differences and per capita incomes is displayed nevertheless. And in all cases the principal factors at work are rooted in the nature of farming. The modest negative associations reflect the increasing sophistication of farming and the higher incomes in farming relative to other economic activites that are associated with higher per capita incomes. Background-specific enrollment rates are readily understood indica- tors of relative participation. Arithmetic differences among those rates, their ratios, and differences in their logits tell us different things. The contrasting situations of children of farmers and of manual workers, each compared to the children of white-collar men, highlight contrasting status relationships and shifts in those relationships across countries and over time. -171 - CHAPTER 6 LONGITUDINAL PERSPECTIVES ON SELECTIVITY Analysis of social selectivity in schools almost intrinsically has a dynamic aspect. While actual differences in degree of selectivity among societies is of great interest, whether opportunities are widening or narrowing in particular societies may be even more important. In this chapter the few cases on which data are available for differing dates are analyzed; here we have actual changes over time, not proxies for such changes in the contrasts among countries variously situated at one time. This analysis exemplifies the more full-bodied exploration of changing opportunities for schooling that becomes possible when we can look at differences in date and not just at countries which happen at a particular moment to stand at different levels of selectivity. Discussion is organized in three main sections. The first section focuses on changes in the degree of representativeness of students in secondary and in higher education: are students becoming progressively more or less representative of the general population? Fundamental to this analysis, as to that in Chapters 2 and 3, is consideration of how distributions of status in base populations may constrain the values of selectivity indexes, but now with the added question of how those base-population distributions may be altering the situation through time. The second section builds on the analysis introduced in Chapter 5 in that the focus is on background-specific enrollment rates, but for most countries the best approximation available is estimates of ratios of those rates for any two status origins; as shown in Chapter 5, ratios of enrollment - 172 - rates equal ratios of selectivity indexes. The relation of the logarithms of those ratios to differences in the logits of enrollment rates and the logic of using the latter are discussed further here, with particular reference to questions of convergence over time in enrollment ratas of youth from different backgrounds. Finally, in the third section logit measurement is used to examine how actual changes in enrollment rates compare with the changes that would be required to reach specified goals in progress toward greater equality of educational participation. That analysis includes assessments of the magnitude of the task that would lie ahead in further efforts to realize stated goals. Such an analysis is necessarily confined to the few countries for which background-specific enrollment rates, not just their ratios, can be reliably estimated for widely separated dates. A. Changes Over Time in Degrees of Selectivity How the degree of selectivity or representativeness of student populations changes over time is affected by several factors: (1) Broad changes in occupational distributions for a society are embodied in the distributions of occupations in the adult base population; those changes shift the mathematically constrained bounds of the selectivity indexes. But in addition there are changes in occupational structures within the broader categories that could affect observed selectivity even when there is underlying stability in differentiated selectivity from more narrowly defined sub-categories of occupations. - 173 - (2) Aspirations for schooling spread within and among subpopulations in response to anticipated relative private advantages of investing time and other resources in schooling. These changes can be dramatic within a single lifetime among residents of what we call "developing nations." Recorded changes in overall rates of enrollment are a blurred reflection of the results of this complex process. (3) Beyond a broad policy to expand schooling, more specific educational policies affect the provision of facilities. active recruitment of students, and promotion of students through the system. Some educational practiceb <,WouraTe, some inhibit changes in social selectivity. (4) Changes in the linkage of educational with labor-market poiicics operate not only through c1langes in incentive structures and in associated aspirations, but also through direct controls. (5) The leveling up of proportions of students who are female can retard the measured decrease in social selectivity. Almost universally, social selectivity is greater for girls, sometimes by a wide margin. This effect happens to be minimal in the countries represented in Table 6.1, given what is known about female representation in those countries, although none of the longtitudinal data provide separate information for each sex. (Those for the United States are for males only; the others do not distinguish by sex.) 1. Overall Changes in Occupational Selectivity; Indexes of Dissimilarity. Most of the data in hand with'a time dimension refer to paternal occupations of students. The available data on distributions of fathers and on selectivity by occupation for the same countries in earlier and later years are set forth in columns (1) through (9) of Tables 6.1. It is well to take an over-all view to start with, however, so I turn first to TABLE 6.1 OCCUPATIONAL DISTRIBUTIONS OF FATHERS, SELECTIVILITY INDEXES, AND INDEXES OF DISSIMILARITY AT EARLIER-AND LATER DATES Percentage of Fathers Selectivity Indexes Manual White Collar White White Index of (including I Including II Excluding Proprietors Collar Collar Disaimi- Farm farm labor) Proprietora Proprietors Fare Kanual end Tradera I II larity (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) SECONDARY STUDENTS Ghana 1974 33.9 15.8 47.8 42.3 ***C* 1961 33.4 14.0 43.6 43.2 ~ .53 .53 .10 3.96 6.26 36.3 .37 1970 15.2 33.3 51.5 37.1 .17 .15 2.44 6.44 17.95 71.8 .47 1962 32.0 a 17.0 .44 - (1.96) - *8.50 a Chile, Santiago 1969 - (10.0) - 90.0a 80.0 - (.16) - 2.29 2.44a 2.46 54.0 .81 1929 - (4.1) - 95.9a 57.8 - (.05) - 3.02 4.8la 17.52 80.0 .72 Puerto Rico 1960 7.7 27.5 64.8 47.3 1.54 .41 b 2.33 b 39.7 .51 1944 12.6 24.3 63.0 41.1 .82 .35 b 4.06 b 47.5 .56 Hungary 1963 16.6 44.2 39.2 a .63 .79 2.21 C21.5 .26 1931 13.5 3.8 82.7 a .26 .18 *3.19 *55.9 .80 Denmark 1961 9.6 23.6 66.68 .54 .63 *1.62 C25.4 .44 1912 5.6 11.0 83.4 *.20 .20 C3.34 *58.4 .78 Norway 1960-66 - (52.8) - 47.2 a C a a A 1938-50 - (36.1) - 63.9 a C U.S.A. Cohortac IV 15.9 36.7 47.4 C .73 .97 C1.18 C7.2 .12 (Hales) liI 19.3 39.6 41.1 C.67 1._ C 1.31 C9.8 .14 II 21.2 34.1 44.7 C.62 .97 a 1;46 C14.0 .20 I 23.1 25.5 51.3 C.57 .96 C1.54 C13.0 .27 UNIVERSITY STUDENTS Uganda 1968 50.6 4.0 45.4 34.3 CCCCCC C 1958 48.0 11.7 40.3 32.5 C Graece 1980 30.0 24.9 45.1 29.8 .73 .71 1.82 1.89 1.91 23.1 .28 1962 33.7 19.9 46.4 33.2 -.68 .59 1.63 2.71 3.07 29.3 .35 Puerto Rico 1960 6.8 27.8 65.4 23.0 1.36 .41 ob 2.38 b 38.2 .51 1944 12.1 27.3 60.6 10.0 .79 .39 b 3.96 b 45.3 .67 - Hungary 1963 10.8 33.0 56.2 C.42 *54f 3.18 C38.5 .47 1931 12.8 3.9 83.3 C.25 .18 a 3.11 C56.5 .77 Poland 1975 11.2 30.9 57.9 a .37 .72 2.18 C31.3 .41 1961 18.3 26.2 55.5 C .50 .67 C2.31 C31.5 .39 U.S.S.R. 1970 10.7d 36.2 53.1 C.54 .66 C2.12 C28.1 .40 1964 11.6d 34.9 53.5 C.47 .65 C2.55 a 32.S .44 1938 21.7 36.1 42.2 C.45 C'l 2.38 lb 31.1 .54 France 1965 6.7 11.7 81.6 64.3 .44 .25 1.26 2.13 2.60 43.2 .37 1950's 5.7 4.5 89.8 74.8 .27 .10 1.38 2.79 3.46 57.3 .52 West Germany 1975 3.7 12.0 84.3 60.0 .58 .23 2.12 2.06 2.03 43.3 .74 1965 3.4 6.9 89.7 64.2. .89 1.85 1.85 1.78 1.76 44.7 .64 Ireland 1979 19.3 10.7 69.0 Ck Ca a Ca 1963 17.5 10.0 72.5 C .65 .24 2.35 *6 41.7 '.60 United Kingdoq 1979 Urban 22.3 77.6 C Urban .34 C2.24 a 43.0 .66 1961 Surveys 26.0 74.0 C Survey .36 a 2.60 a 45.5 .64 Japan 1968 12.4 8.1 79.5 C.43 .29 (.80)0 1.82 1.9Bf 35.7 .64 1953 18.4 4.71 77.0 C.53 .14 (1.16)e 2.37 3.04f 44.5 .66 U.S. Freshmen Cohortc IV 11.2 27.2 61.5 C.52 .72 C1.53 C 21.3 .36 III 12,3 30.0 56.8 C.42 .78 I-18ei 25.5 .37 II 15.11 27.2 577C.44 .77 C1.88 C27.1 .39 T. 19.7 22.3 58.0 C.49 .84 C1.74 a 24.6 .37 U.S. Graduates Cohort . IV 8.5 25.4 66.1 C.39 .67 C1.64 a 25.8 .43 III 10.8 25.1 64.1 P.37 .63 C2.04 a 32.7 .47 II 11.6 23.7 64.7 C.34 .67 a 2.11 C34.0 .49 I 17.3 20.3 62.4 C.43 .76 C 1.87 C39.0 .44 a/ Large landed proprietors Included; virtually all have city residences. They were 2 percent of fathers In 1969 hB ass figures were not age-controlled, which seriously distorte them for proprietors in Pueto Rico. The distortion is largmly rounved. however, for White Collar I. cI Birth cohorts are: 1 1898-1907, II 1908-17. III 1918-27, IV 1928-37 d/Includes workers on collective faresa 'ISaleas ElExcluding 'Sales' - 175 - column (10) of Table 6.1 and the first column of Table 6.1, the range in indexes of dissimilarity is extremely wide, especially for secondary students, but its is substantial for university students as well. In the few cases for which data are available on secondary students, the declines in this index are marked. In most cases it declined for university students also, but less dramatically. For the U.S.S.R. the index first rose, froTm 1938 to 1964, but then declined between 1964 and 1970. Reductions in the index of dissimilarity were comparatively large in Hungary and in France. In the United States shifts in the index were generally small, especially for secondary students, but so were the indexes of dissimilarity for those students even in the older cohorts. 2. Changes in representation of farmers' children. As can be seen from Table 6.1, in a majority of cases the proportionE of students coming from farm families decreased; decline does not necessarily mean a reduction in selectivity indexes for such stuidents) however. In all cases in which the proportions of students from farm homes fell there was also a fall in the proportion of the base population who were farmers (the unreliable figures for Senegal aside) and in most cases selectivity indexes on farm origins rose. At the secondary level there were notable increases in both Denmark and Hungary in the proportion of students from farm homes despite declines in farmer proportions of the base populations over the long time spans covered for those countries; both forward surges in secondary-school attendance and advances in the sophistication and prosperity of the agricultural sector contributed to this result. Among university students there were no comparable cases, but estimated farm selectivity indexes rose in all cases except Poland, West Germany, and Japan. - 176 3. Changes in Representation o. Children of Manual Workers. The predominant pattern for manual workers is the reverse of that for farmers in that their proportions rose both among fathers of students and in the base population. (The samples for the United Kingdom were for urban households only, and are not comparable with the others.) Also, in the U.S.S.R. between 1938 and 1964 there was a slight decline in the proportions of fathers of university students who were manual workers, along with a marked increase over that period in the base-population proportions in manual jobs. The associated decline in the selectivity index (from 1.11 in 1938 to .65 in 1964) was a substantial reversal, and that index hardly changed between 1964 and 1970. What is indicated here is a second-generation readjustment following the effects of extreme quota policies of the first two decades of the Soviet state. By contrast, the 1931 to 1963 increase in proportions of Hungarian university students from manual workers' homes was dramatic; traditionally this country had been decidedly elitist in its educational system and occupational structure. Despite some increase in their manual selectivity indexes, low representation from such homes persisted over the years shown among university students in France and in Japan. In West Germany the selectivity indexes for both farm and manual workers' children declined between 1965 and 1975. The most nearly proportional representation of students from homes of manual workers (the highest selectivity indexes) were in the United States, Poland, and (despite the decline since 1938) the U.S.S.R. 4. White-Collar Selectivity Indexes and Changes in Them. These indexes exceed 3.00 in all earlier-year entries for secondary students with the marked exception of United States males. Among university students, indexes over 3.0 occurred in Puerto Rico in 1944 and in Hungary in both 1931 and 1963. In most cases the white-collar selectivity indexes declined over time. Most striking were the declines among university students in Puerto Rico, followed at some distance by Greece, France, and Japan. The - 177 - white-collar indexes rose in Hungary between 1931 and 1963 -- a somewhat surprising phenomenon until one takes into account the sharp decline in the reported white-collar proportion in the base population, a decline that may have been fictitions and was not in any case quite matct1ed by the decline in their share of university places. In the U.S.S.R. the white-collar index rose between 1938 and 1964, but then declined as the growth in white-collar workers caught up with the prior increase in white-collar representation inr the universities. White-collar indexes below 2.0 characterized Danish secondary students in 1961 in contrast to the high index 50 years earlier, .and they characterized all cohorts of U.S. males in both secondary and higher education, university students in West Germany in 1965 (but not in 1975), Japanese atudents in 1968, and Greek students in 1980. Part of the reductions in white-collar selectivity indexes must be traced I;.o increases in the proportions of the base population who were in white-collar employment. TRhis matter is addressed in column (11) of Table 6.1 and column (2) of Table 6.2, which present the estimates of (Sobs-l)/(MaxSbi-i). It will be remembered from Chapters 2 and 3 that this ratio specifies the excess in selectivity from the more favored backgrounds (Sobs-!) as a proportion of the maximum possible oven-representation given base population constraints (MaxSbi-1). A low ratio in column (11) indicates' that the observed selectivity ratio is low relative to its base-population- constrained maximum. For both secondary students and college entrants in the United States the ratio are distinctly low; they are low also for Puerto Rico in 1960, for Greece in both years, and for the U.S.S.R. in 1970. The ratios -are over .70 (indexes 70 percent or more of their maximum possible values) among secondary students in Santiago in both years, in Hungary in 1931, and in Denmark in 1970. Among university students they exceeded .70 in Hungary in 1931 and even at the later dates in France and in West Germany -- but nowhere else. - 178 - TABLE 6.9 COMPARISON AMONG ALTERNATIVE MEASURES OF CHANGES IN INEQUALITY OF OCCUPATIONAL REPRESENTATION: SECONDARY AND UNIVERSITY STUDENTS DIFFERENCES IN S.I: RATIOS OF S.I. Index of (Sobs-I)' WRITE COLLAR I vs WHITE COLLAR I vs Dissimilarity (Max Sbi-1) Farm Manual Farm Manual (1) (2)b (3) (4) (5) (6) SECONDARY STUDENTS Chile, Santiago 1969-29 -26.0 -.09 * * * * Hungary 1963-31 -34.4 -.54 -1.35 -1.59 -8.76 -14.92 Denmark 1961-12 -33.0 -.35 -1.99 -2.15 -1.12 -14.13 Puerto Rico, 1969-44 -7.8 -.05 -2.45 -1.79 -1.20 -5.92 U.S. Males Grada 12 Cohortsa IV-IIL -2.6 -.02 -.19 -.10 -.34 -.09 III-II -4.2 -.06 -15 -.20 -52 -.18 -37 -.39 -1.08 -.20 -38 II-I -4.0 -.07 -.13 -.09 -.35 -.09 UNIVERSITY STUDENTS Greece 1980-62 -8.2 -.07 -.87 -.94 -1.40 -1.93 Puerto Rico 1960-44 -7.1 -.16 -2.15 -1.60 -.27 -4.34 Hungary 1963-31 -18.0 -.30 -0.10 -.34 -4.87 -11.89 Poland 1975-61 -0.2 +.02 Zero -.18 +1.27 -.37 U.S.S.R. 1970-64 -4.4 -.04 -.50 .44 -1.50 -.71 1964-38 +1.4 -.10 +.15 +.63 - +.14 +1.78 France 1965-50's -14.1 -.15 -.83 -.81 -5.49 -19.38 W. Germany 1972-65 -1.4 +.10 +.59 +.38 +1.44 +3.13 U.K. 1979-61 -2.5 +.02 * -.34 * -.63 Japan 1968-53 -8.8 -.02 -4.5 -.70 -.24 -10.65 U.S. Males: Freshmen (including junior colleges) Cohortsa IV_III -4.2 -.01 -.38 +.20 -1.37 -.19 III-II -1.6 -.02 -.05 -.24 -.50 -.09 +.04 -.71 -.12 +.06 II-I +2.5 +.02 +.19 +=21 +1.38 +.37 U.S. Males: Graduates (16 + years) Cohorts IV-III -6.9 -.04 -.42 -.44 -1.30 -.79 III-II -1.3 -.02 -.10 -.19 -.03 -.14 -.70 -.14 +.09 -.01 II-I +5.0 +.05 +.33 +.33 +1.86 +.69 a/ Births cohorts are: I: 1898-1907, It 1908-1917, III 1918-1927, IV 1928-1937 bI All figures refer to white collar I. For the U.S. cohorts the base population constraints dominated 9setting lower maximum) than the enrollment constraint for all cases except Grade 12 for cohorts II, III and IV, for which MaxS was less than MaxSbi. The corresponding entries in column (2) comparing later with earlier (Sobs-W)~(MaxS-1) would be IV-III - -.07, III II - +.02 and II-I - +.10, reflecting the effects of rises in enrollment rates on ration of selectivity indexes to their maximum possible values given those rates. A similar estimate for puerto Rico drops the 1960-44 entry for secondary students from -.05 to -.24 equalization was substantially more than accounted for by rising enrollments. For Hungary, on the other hand, the figure -.54 would be replaced by -.35 - 179 - The extent to which selectivity from white-collar homes is cut back below the maximum possible value with given base-population distributions may have two components: constraints associated with expansion in overall enrollment rates (which could reduce the possible maximum even with a rigidly ranked status queuing for entrance to the secondary and higher institutions), and features of social and educational structures and policies that reduce effects of status on selectivity beyond the constraining effects of expansion in enrollments. At the university level enrollment rates are usually much less than the proportions of base populations in white-collar jobs. This means that for university students the entrants in column (11) can be interpreted as ratios to absolute maximal possible over-representation even taking overall enrollment rates into account. For stcondary students in the later cohorts, however, enrollment rates constrain the possible maxima in several cases; to incorporate them would raise the figures in column (11) of Table 6.1 slightly for the United States in the three younger cohorts and substantially for Hungary in 1963 and for Puerto Rico in 1960. While changes over time in the ratio of observed to maximal over-representation of white-collar youth can be seen in Table 6.1, for con- venience I have shown differences in these ratios in the second column of Tables 6.2. It is easy to see here that with few exceptions there has been a decline, but the shifts in most cases are small. The sizeable declines in Denmark and Hungary among secondary students exceed any differences for uni- versity students, although the decline for Hungarian university students almost matches the Danish decline for secondary students. At the university level the changes were generally small and in most cases negligible, despite unevenness of changea in xiniversity enrollments from country to country and despite the fact that there are large differences among countries on each indicator of selectivity. h.e stability over time in ratios of actual to maximum white-collar over-representation is impressive evidence of the conti- nuity through time in underlying selectivity processess in most countries. - 180 - Up to this point discussion has focussed on either the overall indexes of dissimilarity or the movement of selectivity indexes on particular occupationsl categories toward (or away from) proportional representation. The last four columns of Table 6.2 are addressed to comparisons of selectivity indexes on white-collar with those on farm and those on manual status. Columns (3) ad (4) show changes over time in the arithmetic differences of the pairs of selectivity indexes, whereas columns (5) and (6) show changes over time in the ratios of the same pairs of selectivity indexes. It is evident enough from inspection of these 'columns that the two measures are only loosely related. Those in columns (5) and (6) are the more interesting, since the ratios of selectivity indexes are the same as ratios of enrollment rates. I shall return to this matter shortly, after a few remarks about changes over time in selectivity by father's schooling. 5. Selectivity by Father's Schooling Over Four White Male Cohorts in the United States. Distributions of paternal schooling along with associated selectivity indexes and indexes of dissimilarity are shown for the four cohorts in the upper half of Table 6.3. These indexes of dissimilarity are quite close to those for paternal occupations despite the fact that the schooling distributions had more rubrics and separated out smaller segments of the base population at the upper levels than were available for occupations. (This reflects both the robustness of indexes of dissimilarity where selectivity bears an unambiguously monotonic relationship to socio-economic categories, as it does for schooling, and the associated insensitivity of the index to selectivity at the extreaes--whatever the country under consideration.) - 181 TABLE 6.3 DISTRIBUTION AND SELECTIVITY BY FATHERS' SCHOOLING AT THREE STUDENT LEVELS: FOUR MALE COHORTS IN THE UNITED STATES Percentage Distribution of Selectivity Indexes (Sobs) Years of Father's Sechooling by Year of Father's Schooling Education Level and Under 8 8-11 12 Over 12 Under 8 8-11 12 Over 12) Cohort (Higher (Higher) (1) (2) (3) (4) (5) (6) (7) (8) SECONDARY, Grade 12 IV 22.8 42.4 19.4 15.4 .68 1.04 1.28 1.41 III 30.8 38.0 17.0 14.2 .73 1.02 1.39 1.65 II 32.4 36.0 18.4 13.2 .70 1.02 1.69 1.74 I 28.7 36.9 19.2 15.1 .59 1.06 1.86 2.36 TERTIARY, Freshmen IV 15.9 3.54 21.2 27.5 .48 .87 1.40 2.52 III 24.0 30.9 21.2 30.9 .57 .83 1.74 2.78 II 25.5 32.6 18.0 23.9 .55 .92 1.65 3.14 I 25.0 33.6 21.3 20.1 .51 .97 2.07 3.14 TERTIARY, Graduates IV 12.5 36.0 20.4 31.1 .38 .88 1.35 2.85 III 19.8 26.6 22.3 31.3 .47 .72 1.83 3.64 II 26.2 29.9 15.4 28.4 - .57 .84 1.41 3.74 I 28.2 32.3 19.5 20.0 .58 .93 1.89 3.12 ,adex of Max SBL Max sa (Sobs-l)/(Max S-Il)a (Max Sbi-Max S)t Di-similarity (Max Sbi-Scbc) Higher Grade 12 Higher Grade 12 Higher Grade 12 Higher Grade 12 (9) (10) (ll) (12) (13) (14) (15) (16) (17) SECONDARY, Grade 12 IV 7.2 9.17 6.62 1.56 1.56 .90 .82 .98 .95 III 9.8 11.63 8.20 1.80 1.80 .92 .77 .98 .94 II 14.0 13.16 9.17 2.26 2.26 .77 .75 .95 .92 I 18.0 15.63 9.71 3.30 3v30 .72 .56 .93 .82 TERTIARY, Freshmen IV 21.3 9..17 6.62 3.30 3.30 .76 .42 >88 .64 III 25.5 11.63 8.20 4.03 4.03 .69 .43 .86 .65 II 27.1 13.16 9.17 5.21 .60 .32 .79 .53 I 24.6 15.63 9.71 6.94 6.94 .45 .30 .70 .42 TERTIARY, GraduAtes IV 25.8 9.17 6.62 5.65 5.65 .50 .24 .56 .15 III 32.7 11.63 8.20 7.04 7.04 .52 .26 .57 .18 II 34.0 13.16 9.17 10.42 9.17 .36 .27 .29 Zero I 29.0 15.63 9.71 11.76 9.71 .27 .27 .25 Zero I - 182 - Among secondary students in the Unitcd States, as we move from the oldest cohort (I) to the youngest cohort (IV) there is a steady decline in indexes of dissimilarity. For the university students this occurs from the second to the youngest cohort, but not between the oldest, born in 1989-1907, and the next oldest, born between 1908 and 1917. Moreover, the sharp drop in the index of dissimilarity occured between cohorts III and IV -- the men who would have been entering college in the late thirties and the war years (or who entered college later under the special educational subsidy bill for veterans) and the cohort of men who would have been entering college in the immediate post-war years directly from secondary school. Looking at selectivity indexes across cohorts, there is remarkably little change in representation of youth from the less educated backgrounds among students in the secondary schools, but indexes referring to selectivity .¼to college graduation among children of men with less than eight years of schooling declined substantially between cohort II and cohort IY; that decline reflects underlying changes in the characteristics of the shrinking proportions of the cohorts of fathers with less than eight years of schooling. At the same time, there were steady drops in the over-representation of sons of men with 12 or more years of schooling among students who were enrolled in the senior years of secondary school. Among university students the youngest cohort evinced consistently the lowest selectivity on fathers with 12 and on fathers with more than 12 years of schooling. Columns (10) through (17) give a fuller analysis of the indexes of selectivity relative to their maxima than was possible for most of Lhe countries included in Tables 6.1 and 6.2. Thus columns (12) and (13) specify - 183 - maximum possible indexes as constrained by the combination of base-population proportions and overall enrollment rates. The bidding constraint was the enrollment rate in all cases except the two older cohorts of college graduates against fathers who completed highschool. Columns (14) and (15) show the ratios of over-representation in the observed selectivity indexes to the maximum possible over-representation (from columns (12) and (13) minus 1.00). The high ratios for secondary students in column (14), referring to children of college men, reflect the near-universal highschool completion in this group for the two younger cohorts. Even among college graduates, the, younger cohorts display rates of over-representation of sons of college men as much as half of their maximum possible values. The ratios for sons of men with just 12 years of schooling are much lower, at approximately a fourth of maximal possible values in all cohorts. For sons of college men the principal upward shifts occurred between cohorts II and III at all school levels. The last two columns of Table 6.3 show the fraction of the gap between observed selectivity indexes and the maximum under constraints inherent in base-population percentages that could be accounted for simply by the constraints due to overall enrollment rates. This proportion is extremely high for grade 12 enrollments of youth whose fathers had 12 years of education or more; again, this reflects an almost universal rate of secondary-school completion in all cohorts, with only the children of men having less than eight years of schooling substantially underrepresented. Among college graduates, however, the maxima due to overall enrollment rates (in this case rates of college-completion) accounted for more modest proportions of the gaps between observed and base-population-constrained over-representation of sons of college men, and for little or none of that gap for sons of men with grade 12 schooling. Taking base-population constraints into account, the picture is one of low status selectivity even - 184 - in the oldest cohorts, with that selectivity declining through the 1930's and the early post-war years. The openness of the system found expression in quite different ways at the secondary and higher levels, setting the stage for the drives toward even higher and more widely spread participation and "compensatory programs" in higher education that followed in the 1960's and 1970's. 1/ B. Background-Specific Enrollment Rates and Changes in Educational Opportunities Behind the deviations of representat,ion in student bodies from distributions of occupations (or of schooling) in the populations of men in the age cohorts of students' parents are the processes that affect opportunities for educational participation and the choices made with respect to the investment of time and money in secondary and higher education. Fundamentally, the educational development process goes back to those opportunities and decisions. The heart of what is going on is the expansion of enrollments, in general and within particular socio-economic segments of a population. The conceptual advantages of using logit comparisons for a developmental analysis are most obvious when attention is focused on enrollment rates, which is the reason I began to discuss that logic in Chapter 5. A recent pertinent article by Mare (1981) starts from the sociological studies of status attainment, beginning with analysis of the determinants of educational attainments. In those studies educational attainment appears as a first-stage dependent variable. (Mare refers .1/ While a major component in that development stemmed also from the civil rights movement and the drive for racial equalization, it was not limited to racial considerations and sex differentials were not involved. - 185 - primarily to analyses of the data on U.S. male cohorts used in the imme- diately preceding discussions, but he begins with what I designated as Cohort II and he includes men who were 20-24 years of age at the time of the sur- vey.) Regression analyses of the 1907 to 1941 cohorts manifested what he regarded as remarkable stability in the effects of father's schooling and occupation on educational attainments of their sons. Mare asks the question, closely related to the focus of this investigation: "what are the reasons for stability in the educational process, and to what extent does the evi- dence for that stability turn on the specific measures and statistical models commonly used?" He argues for use of the "logistic response" model as a way of getting at the process of educational development. Pursuing this purpose he has broken the process down into a sequence of transitions between grades and the logit analysis is then applied at each stage. Thus the dependent variable at each transition is expressed as the logit of the proportions con- tinuing into the next grade. By using the logit function, he frees the ana- lysis from dependence on average continuation rates (in contrast to simple linear probability models). This is analogous to my use of logits of enroll- ment rates at any given level, though I do not build-up from a cumulative series of continuation rates. Both analyses permit inter-cohort comparisons of the effects of social background, in his worda (p. 74) "because differ- ences in effect result from genuine differences in the associations between measured variables over populations" -- in this case, over time. He points out (p.83) that if an analysis takes as the indicator simple differences among groups in proportions continuing in school, change over cohorts will reflect mainly the effects of changes in .the average continuation rates rather than response to changes in the principles by which schooling is allo- cated. By contrast, "statistical models that measure the association between school continuation and social background, net of the marginal distribution - 186 - of schooling, are sensitive to changes in the principles by which schooling is allocated ..." (Further comments on this aspect of the use of logits will be found in Appendix E.) It is not possible in most cases to get the estimates of background-specific enrollment rates that are needed to apply an analysis of differences in logits cf those rates to studies of changes in a country over time. With the few exceptions examined later, the best option open to us is to make use of the fact that ratios of enrollment rates for youth from any two background categories equal the ratios of their selectivity ratios. This relationships was set out in Chapter 5 and applies equally here, Ratios of selectivity indexes (and hence of enrollment rates) are not subject to the errors that inevitably slip into direct estimates of enrollment rates, especially in higher education and for dates at some time in the past. Fortunately, the logarithms of ratios of selectivity inde:xes are closely related to differences in logits over a wide range of the two indicators. The mathematical relationship was specified in Chapter 5. To repeat: (logit Ew-logit Ea) - ln Sw + [-ln(l-SaEt)-ln(1-SwEt)] where E refers to enrollment rate, S refers to selectivity index, the subscripts w and a refer to particular backgrounds (e.g. white-collar fathers and fathers in agriculture), and Et is the overall enrollment rate in education level t. But the first term on the right side could be written ln(Ew/Ea) since the ratio Sw/Sa equals Ew/Ea. Thus it is the bracketed term on the right that accounts for imperfect correlations between differences in logits and in the natural logarithms of ratios of enrollment rates. Carrying - 187 - this a step beyond the discussion in Chapter 5, it is particularly important here to take notice of the mathematical constraints on the relationship between the ratio Sw/Sa and the bracketed term in the equation. If Sw/Sa exceeds 1.00 it follows that (l-SaEt) will exceed (l-SwEt) and the bracketed term must be positive; conversely, if Sw/Sa is less than 1.00 the bracketed term must be negative. More generally, designating that term as G we have the followig implications: G > O when Sw/Sa > i G-= 0 wheni Sw/Sa = 1 G < O when Sw/Sa < 1 and dG > I d(SwISa) As Sw/Sa approaches 1 from either direction the bracketed term G approaches zero, the logarithm of Sw/Sa approaches zero, and so also does the difference in logits. Convergence toward equality between any pair of enrollment rates measured by comparisons of ratios of rates at one date and another, later, date, will imply such covergence when measured by comparisons of differences in the logits at the early date and those differences at the later date. Columns (5) and (6) of Table 6.2 gave the ratios of selectivity indexes and hence of enrollment rates for youth from white-collar/farm and white-collar/manual backgrounds. Those figures constitute the basis of the entries on Chart 6.1, where they are plotted on a double logarithmic grid. 2/ 2/ The figures for Senegal are non-comparable and extremely unreliable. They are not included on Chart 6.1 - 188 - Chart 6.1 Ratios of White Collar to Manu;a and to Farm Enrollment Rates; Later by Earlier Dates [Seondary,Students A. WHITE COLLAR/MANUAL RATIOS OF RATES o Cllege and Univeiiiy Students Later Ratios 9.0 West Germany 8.0 - 0 France 7.0 6.0 O U.K. O Japan 6.0 - /13Puarto Rico * aa 5.0 Puarto Rico 0 Hungary 4.0 , USSR 38-64 Sneagl ° /A/l USSR 38-70 a Poland 3.0 OGrec Demr Zs 0 D nmark U5/ 0 Hungary 2.0 US 0 IIY US. ll-IV 1.5 0 U.S. 1-11I * U.S. Il-IV 1.0 I I I I . I I I I I I . B. WHITE COLLAR/FARM RATIOS OF RATES 1.5 10.0 - 9.0 7.0 - O Hungpry 6.0 6.0 - Polwnd; 5.0 zPUSSR 38B6 U0 - 1 -111 (C1,M an France 4.0-OUS387 3.0 - / =Denmark 25 - /3 OU.S. I I-IV (C1) Puerto Ricob U.S. I-Ill * U.S. Il-IV 1.5 - /* Puerto Rico 1.0 1.5 2.0 2.5 3.0 4.0 5 6.0 7 8 9.10.0 15.0 2 2.5 3 4 5 Earlier Ratios - 189 - Contrasts among countries are clear enough to give a good visual impression of the diversity in the ratios and in the extent to which there has been movement toward their equalization. On this chart countries lying above the diagonal experienced actual increases in relative disparities of enrollment rates among youth whose fathers fell in different occupational categories, whereas points below the diagonal signal movements in an equalizing direction. The vertical distance from the diagonal measures the amount of progress (or retrogression) in terms of egalitarian goals. (These vertical distances are differences in the logarithms of ratios of enrollment rates, whereas t1he entries in columns (5) and (6) of Table 6.2 were the simple time differences in ratios of rates, not the differences in their logarithms.) In reading Chart 6.1, notice that secondary students are designated by black%dots, students in higher education by an x. For the United States, as for most other countries, the data plotted here refer to entrants to higher institutions, not to their graduates. Children of manual workers manifest substantial movement toward parity of enrollment rates in all cases in which the initial ratios of white-collar to manual rates exceeded 10.0 (which, to repeat, is exactly the same as the ratios of two selectivity indexes exceeding 10.0). For secondary schools in Denmark and Hungary, the changes are striking, and only in part due to the long span of years covered. These two societies have of course had very different socio-politf-tal histories. At the university level -Hungary and Japan are quite close to each other in both earlier and later ratios of white-collar to manual enrollments. Tne French shift in an egalitarian direction is large for the short time span, even though France remains at a higher white-collar to manual ratio of enrollment rates in university than any other country except West Germany. Apart from earlier extreme ratios (exceeding 10.0) there is definite stability in the logarithms -.190 of the ratios, even as enrollment rates have been generally rising. This stability is shown by the clustering of entries near the diagonal, even though most are below it, manifesting some progress toward parity. The pattern for white-collar to farm enrollment rates is less stable through time, although there are fewer really extreme early cases. The most notable progress for farmers' children was in Denmark and in Puerto Rico at the secondary level. C. Actual Changes in Background-Specific Enrollment Rates Relative to Changes Required to Reach Specified Targets The increase in overall enrollment rates that would be required to reach equality among background-specific rates depends on the enrollment rate of the most over-represented background category, the initial overall enrollment rate, and the proportion of the population of youth who are in the most over-represented group. A simpler way of expressing this is to specify these relationships in a slightly different way. Define No= Nh+Nio=ThEh+TiEio where No = initial total enrollments Nh = enrollments of the most advantaged youth Nio = initial enrollments of all other youth Th = number of all h youth in the population Ti = number of all other youth in the population En = enrollment ratio of h youth Eio= initial enrollment ratio of all other youth - 191,- The increase in enrollments required to make all enrollment rates the same without reducing Eh (that is to bring them all up to Eh) is entirely an increase from Nio; call this Ni. We then have, by the stipulation of equality of rates: Ni = Ti (Eh-Eio) But with that increase we must have the overall rate equal to Eh. This gives the equation: Nh+Nio+ Ni = Eh where T = Th+Ti T and the increase in overall enrollment rate is then Ni 5 Ti (Eh-Eio). T T For this sort of estimation one needs to know enrollment rates, but no matter what those rates may be, the top background-specific rate sets the level for equality in enrollment rates if there is no cutting back at the top. What, then, are feasible.goals? One common approach is to emphasize meritocratic selection, even though in practice this will normally still favor the socio-economically advantaged in at least some degree. Other approaches (which may be combined with meritocratic selection) include quotas and various combinations of means-tested (ability-to-pay) subsides to help in the financing of further education; at the university level such subsidies may include student maintenance while in attendance. Free tuition for all is no answer; indeed, it generally favors those who already have more, since they are the youth most likely to be in a position to go on into higher institutions; they are also in the best position to forego earnings while studying. Whatever the policy pursued, there is no way of avoiding the questioti: "how far toward equality of opportunity -- or even, a much stronger criterion, equality of participation -- should a country in a particular situation strive to go? - 192 - In this section I discuss measurements of degrees of progress toward assumed targets, but without attempting to give policy prescriptions with respect to what those targets or goals should be, or how they might best be implemented. Such prescriptions cannot have general validity, nor can they be imposed by outsiders on any particular society. What can be done is to offer realistic assessments in a developmental perspective of the degree of "effort" -- both public and by individuals and families -- manifest in observed progre-; toward diminishing inequality of educational participation. Eight years ago, while teaching in Sweden, I concluded that local disappointment with efforts to widen access to higher education was unjustifiably pessimistic. Despite conspicuous improvements, local commentators seemed to me to be expecting unrealistically-rapid advance, especially in view of the generally "sticky" character of the Swedish status structure. So I explored ways to measure in a more realistic perspective the progress' toward greater parity among social-status groups in attendance of their youth at university. The available data were reported in a conventional three-status scale. The contrasts between the three different ways of measuring progress are striking (Table 6.4). Group I, at the top, had the largest gain in absolute percentage terms, but the lowest in ratio of 1970 to 1950 enrollment rates. Group III had the least absolute gain, but the highest ratio of 1970 to 1950 enrollment rates. Viewing the groups as moving along development curves (increases in the logits of enrollment rates), the lowest-status group made the most impressive gains from the starting point, although Group I made an almost equal logit gain. These measures may be variously interpreted for policy purposes, but the Swedish performance over this twenty-year period surely was not lacking in democratization, despite the dramatic absolute increase of enrollment for the very small Group I. -1.93- TABLE 6.4 ASSESSMENT OF PROGRESS TOWARD PARITY OF UNIVERSITY ENROLLMENT RATES BY SOCIAL STATUS; SWEDEN, 1950-'970 Social Status Group I II III High Intermediate Low A. Enrollment Rates of 20-year- olds in Higher Education 1950 28.30 4.76 0.95 1970 79.49 20.10 8.92 3. Logits of Proportions Entering College 1950 -0.9304 -2.9958 -4.6471 1970 4-1.3548 -1.3899 -2.3234 C Measures of Changes in Rates of College Entry, 1950-1970 1970 minus 1950 Rates 51.19 15.34 7.87 Ratio of 1970 to 1950 2.809 4.223 9.389 Logit 1970 minus + 2.2852 + 1.6059 t 2.3237 logit 1950 ), Hypothetical Logit Changes from 1950 to ileach Designated 1970 Targets a. Overall 1970 rate - (21.0%) -9.3939* + 1.6715 + 3.3228 b. Group I 1950 rate. (28.30) ... 4-2.0654 + 3.7167 c. Group I 1970 rate (79.49) + 2.2852 4 4.3506 + 6.0019 . Hypothetical Logit Changes to Reach Designated Targets Minus Observed Changes a. 21.00% -2,6791* + 0.0656 4- 0.9991 b. 28.30% -2.2852 i 0.4595 + 1.3930 c. 79.497% ... + 2.7447 4 3.6782 , Observed Changes as Proportion of Changes Required to Reach Designated Targets a. 21.00% 0.961 0.699 b. 28.30% 0.778 0.625 c. 79.49% 0.369 0.387 Source: C. Arnold Anderson: Expanding Educational Opportunities: Conceptualization and Measurement, Higher Education 4 (1975), 393-408 (Tables V and VII) - 194 - One may view the distance travelled in logits as a,rough indication of progress in units of "veffort." The entries in section D of Table 6.4 are then the "effort distances" from 1950 required to reach specified 1970 goals. In section E these effort distances are compared with the actual logit gains: The figures specify how many more logit units would have been required to meet each target. In section F the actual logit gains are expressed as proportions of the gains that would have been required to meet those targets. The youth from the lowest st!tus category had the furthest'to travel initially (from 1950) in both logit and absolute percentage terms, and they still had the furthest to go in 1970. But equally impressive are the facts that the lowest-status group had made the greatest gain in logit units and that for any except the extreme target (c) -- which could not be sustained as an overall rate of college enrollment -- the youth of Group III had covered well over half of the logit units of gain needed to meet targets. This approach puts events in a perspective quite unlike assessments based on observations of arithmetic differences in enrollment rates. A similar analysis of data on occupational selectivity in enrollment rates is presented in Table 6.5 for Japanese and for American youth in higher institutions and in Table 6.6 for Hungarian and, for American youth in secondary schools. Because the data for the United States refer to white males only, they might suggest less selectivity.into higher educaton than would be the case if they referred to the total population and to the two sexes combined. Sex differences in this respect are! small in the United States, however, and for secondary students in grade 12 they have even been reversed. TABLE 6.5 ASSESSMENT OI PROGRESS TOWARD PARITY OF RATES OF COLLEGE ENTRY BY PATERNAL OCCUPATIONAL IN JAPAN 1953-68 AND FOR FOUR UNITED STATES MALE COHORTS; PROGESS TOWARD PARITY OF RATES OF COLLEGE GRADUATION FOR FOUR UNITED STATES MALE COHORTS. ENTRY TO HIGHER INSTITUTIONS COLLEGE GRADUATION JAPAN 1953-68 U.S. Males, Cohort I-IV U.S.-Males, Cohort I-IV Farm Manual White White Farm Manual White Collar Farm Manual Collar Collar A. Enrollment Rates Later date 8.7 5.6 36.8 16.1 22.3 47.6 7.1 12.1 29.7 Earlier date 5.0 1.3 22.4 7.3 12.4 25.7 3.8 6.7 16.4 B. Logits of Enrollment rates Later date -2.3508 -2.8428 -0.5408 -1.6508 -1.2483 -0.096 -2.5714 -1.9830 -0.8616 Earlier date -2.9484 -4.3297 -1.2425 -2.5415 -1.9551 -1.0616 -3.2314 -2.6337 -1.6215 C. Measures of change Later minus Earlier +3.7 +4.3 +14.4 8.8 9.9 21.9 3.3 5.4 13.3 H Later ratio to Earlier 1.74 4.31 1.64 2.21 1.80 1.85 1.87 1.81 1.81 Change in Logits +.5936 +1.4869 +.7017 +.8907+7068 +.9655 +.6600 +.6507 +.7599 D. Hypothetical Logit Change to Reach Designated Target a. Earlier white collar rate +1.7019 +3.0872 ... +1.4799 +.8935 ... +1.6099 +1.0122 b. Later average rate 1/ +1.5706 +2.9559 (-0.1313) +1.7461 +1.1579 +.2662 +1.7218 +1.124 +.1119 E. Hypothetical Required for Designated Targets minus Observed Logit Change a. To earlier white collar rate +1.1083 +1.6003 ... +.5892 +.1867 ... +.9499 +.3615 b. To later average rate +1.9770 +1.4690 (-0.8330) +.8554 +.4511 (-.6993) +1.0618 +.4734 (-.6480) F. Observed Change as Propprtion of Hypothetical Required a. To earlier white-collar rate .349 .482 ... .602 .791 ... .410 .643 b. To later average rate .378 .503 N/A .510 .610 (3.627) .383 .579 (6.791) 1/ Average entry rates for Japan were 9.5% in 1953 and 20.2% in 1968 (Logits -2.2541 and -1.3738) Average entry rates for U.S. Males were 14.8% in Cohort I and 31.1% in Cohort IV (Logits -1.7504 and -.7954 Average proportions who graduated from college among U.S.males were 8.9% in Cohort I and 18.1 percent in Cohort IV. (Logits -2.3259 and -1.5096) - 196 - TABLE 6.6 ASSESSMENT OF PROGRESS TOWARD PARITY OF SECONDARY ENROLLMENT RATES BY FATHER'S OCCUPATION: HUNGARY 1931-63 AND UNITED STATES MALE BIRTH COHORTS I (1898-1907) Hungary United States, Males Farm Manual Wi-hite White Collar Farm Manual Collar A. Enrollment Rates Later date 17.0 21.3 59.7 48.0 63.5 77.4 Earlier date 8.5 2.0 35.1 17.9 29.7 47.6 B. Logits of Enrollment Rates Later date -1.5856 -1.3069 +0.3930 -0.0800 +0.5537 +1.2368 Earlier date -2.3763 -3.8918 -0.6146 -1.5931 -0.8616 -000961 C. Measures of change Later minus Earlier 8.5 19.3 24.6 30.1 33.8 29.8 Later ratio to Earlier 2.00 10.65 1.68 2.68 2.14 1.63 Change in Logits +.7907 +2.5849 +1.0076 +1.4431 +1.4153 +1.3329 D. Hypothetical Logit Change to Reach Designated Target a. Earlier white collar rate +1.7617 +3.2772 . +1.4270 +.7655 ... b. Later average rate 1! +1.3787 +2.8972 (-.3800) +2.731 -+1.5116 +.7461 E. Hypothetical Required for Designated Targets minus Observed Logit Change a. To earlier white collar rate +.7910 +.6923 ... (-.0161 (-.6498 oo b. To later average rate +.5880 +.3123 (-1.3876) +.7300 +.0963 (-.5868) F. Observed Change as Proportion of Hypothetical Required a. To earlier white-collar rate .449 .789 ... (1.011) (1.849) b. To later average rate .574 .892 Inapplic. .664 .936 (1.786) 1/ Average enrollment rates for Hungary were 11% in 1931 and 27% in 1963. Average rates for U.S. males were 31.0% in Cohort I and 65.7% in Cohort IV (at grade 12) Corresponding logits for Hungary are -2.0907 and -0.9946; for U.S. males they are 0.8001 and +.6500 - 197 - The important part of Table 6.5 is in the analysis of college entrants in Japan (1953 and 1968) and for the oldest and youngest cohorts of U.S. males who were of college-entry age around 1920 and 1950 respectively. The last three columns of the table, which refer to proportions of youth completing college, are essentially similar in patterns of change to those on college entry, although of course at somewhat lower rates throughout; comments are therefore concentrated on the first six columns of the table. The years 1953 to 1968 in Japan span a period of consolidation in adjustments to post-war changes in the structure of the educational system and of higher education in particular; around 1968 the expansiion of both secondary and higher education was accelerating, and to a degree that was causing considerable concern by the mid-1970s. The average logit gains in college entry rates between 1953 and 1968 were very nearly the same magnitude as the gains over the 1920 to 1950 generation of U.S. males: + .8803 and + .9550 respectively. The principal differences between the two countries were in their starting poirnts; overall the college entry rates of the oldest cohort of U.S. males exceeded the 1953 rates for Japan, and children of manual workers were in a stronger initial relative position in the United States. In Japan it is clearly children of manual workers who wade the greatest gains (the white-collar advantage measured in absolute percentage increases aside). Nevertheless, children of manual workers in Japan still had gone only half way in logit units to the 1968 average college entry rate and they were still far behind the 1953 rate for children of white-collar men. (The 1968 college entry rate of children of manual workers was still only 15 percent of the 1968 entry rate of children of white-collar men.) Among the U.S. males, the logit gains across a generation were actually least for sons of manual workers, though differences in gains among occupational - 198 - categories were relatively small. Sons of manual workers advanced by three-fifths of the logit units required to reach the overall college completion rate of the youngest cohort, but they did not have as far to go to that goal as the sons of manual workers in Japan. The time span between the earlier and later years for the secondary students (in Table 6.6) is three decades for both Hungary and the United States -- from 1931 to 1963 in Hungary and again from around 1920 to 1950 for the U.S. males. The thrust for greater representation of manual workers in Hungarian secondary schools is reflected unambiguously in their logit gains compared with both sons of farmers and of white-collar men. Indeed, between 1931 and 1963 sons of manual workers had made an impressive + 2.5849 logit advance, gaining 90 percent of the total logit units that would have been required to attain the 1963 overall secondary enrollment rate. It must be remembered, however, that a substantially larger part of the 1963 than of the 1931 base population were classified as "manual workers" and sons of such workers therefore had a greater weight in the average for all youth. Also, the overall secondary enrollment rate in 1963 still was less than the white-collar rate in 1931, the absolute percentage gains in enrollment rates were higher for sons of men in white collar than in manual employment, and as of 1963 sons of manual workers still had attained enrollment rates only slightly over a third of the rates for children of white-collar men. Sons of farmers meanTwhile dropped behind sons of manual workers. Cctrasts between selectivity of the Hungarian and the United States secondary students are obvious enough ina absolute percentage terms; even in the oldest U.S. birth cohort, reaching grade 12 around 1920, 199. enrollment rates of sons of manual workers exceeded their 1963 rates in Hungary. Though both the ratio of later to earlier rates and differences in their logits were greatest in the United States for sons of farmers and least for sons of white-collar men, advances in enrollment rates were remarkably similar across parental occupations. By the later date, sons of manual workers had surpassed by a substantial margin the rates for sons of white-collar fathers in the previous generation, which was not the case in Hungarian counterparts in the proportion of the "logit distance" to parity covered over the span of a generation. There is more than one path to equalization of educational prticipation, but the "effort" entailed depends very much on the starting points both in enrollment rates and in socio-economic heritage. - 200 - CHAPTER 7 SELECTIVITY BY TYPE OF SCHOOL AND CURRICULUM There can be and often is as much social selectivity among schools at a given level as into that level as a whole. It has long been known, for example, that students of pedagogy and of theology come disproportionately from homes of farmers and of urban artisans. "Grammar school" pupils are more often offspring of high-status parents than are pupils in "secondary modern" or vocational schools. Pupils in "modern" lines are less selected than those in "classic" lines. Most of the data on this topic come from Europe, but status differentiation among schools and curricula is not confined to Europe. It is probably safe to say that historically among developed countries exclusive- ness in entry into upper-secondary institutions (and hence in due course into a higher school) tends to be accompanied by status hierarchies of secondary schools that maintain selectivity even as secondary education in general is opened to and used by the broad populace. But the!re are exceptions. Even so democratic and mass oriented an educational system as that of Japan does not ensure against social selectivity among secondary schools, despite the support of an elaborate meritocratic examination system. Indeed, as educational systems expand the interplay between mass attendance and meritocratic selection becomes increasingly complex and subtle, whether or not expressed in an elaboration of formalized procedures. Similar strains appear early in the development of modern educational systems even in the least developed countries, as can be seen in both Kenya and Tanzania, for example, despite the contrasting approaches that have characterized those two countries. - 201 - Diversity in selectivity among types of institutions and curricula has to be treated cursorily here, for it is quite impossible to cover system- atically such aspects of socio-economic selectivity in a cross-national exploration at a highly aggregated level. The most that can be done is to bring together information that was unearthed incidentally in pursuit of the main topic of this report. There are many studies of background characteristics (and even more of ability) of students registered in one or another field in higher education for example. These studies do not include the aggregative information needed for the rest of this report, and there could be no search of that literature. One has always to remember that deliberate selectivity for academic capability need not correspond closely with incidental selectivity for social status. In many ex-colonial countries with large proportions of expatriate teachers and often extensive use of boarding schools, the separation between those two sorts of selectivity may be marked. The pupils from peasant homes or from remote districts sometimes display academic superiority over many of their classmates because it is mainly the brighter and more ambitious of those children who enroll first (Heyneman). If under those circumstances selection tests are used for promotion, there may be only weak social selection into school and only a low correlation between background and performance in school. At the same time, ambitious and better-placed parents may manage to get their children into preferred primary schools in the first place or to circumvent barriers at later stages when their children do not excell in the competition for preferred places. An understanding of selectivity into secondary and higher institutions of one sort or another may have to be pushed back to an examination of entry into elementary schools: quantitative - 202 - (in an educationally backward country, who gets elementary instruction at all), and qualitative (the advantage of having been in a superior elementary school). Social selectivity into elementary education as a whole is discussed briefly in Appendix F, but there has been no attempt to delineate variations in prestige or social selectivity into different primary schools. Four topics are dealt with in this chapter: (1) social heterogeneity or homogenity in the secondary classroom, (2) selectivity by types of secondary schools, (3) differentiation by agencies controlling the secondary schools, and (4) differentiation within higher education. The reader may find it useful to look at other data as particular countries are mentioned, particularl.y in chapters 2 and 4. *A. Social Heterogeneity or Homogeneity in the Classroom If there is marked social segregation by type of school or of curri- culum, one expects to find a homogenous clustering of pupils in each school or classroom; the more heterogenous the composition of-classrooms the more open is the system and the broader is the range of experience through associations with classmates. Often there is more of a classroom mixture than might be supposed in the light of simp3lified indexes of family background. In his study in Ghana, Foster found that pupils in those secondary schools with a top national reputation more often came from families with higher status, yet in all but a few "elite" schools farmers' sons made up a plurality (Table 7.1). The important questions in such a case are just what role the "elite" schools play in the system and in the placement of their graduates and whether in the future composition of classrooms will become more or less differentiated. Across the continent in Kenya, Olson contrasted - 203- TABLE 7.1 PECENTAGE DISTRIBUTION OF PATERNAL STATUS BY REPUTE OF SECONDARY SCHOOL; GHANA Repute of School Father's occupation High Low Elite Professions 35 35 53 Trade 11 12 12 Skilled 12 17 15 Other manual 2 1 1 Farming 39 37 19 Uniformed services 1 1 - Total 100 100 100 TABLE 7.2 HOUSEHOLD CHARACTERISTICS OF BOYS AND OF GIRLS IN ELITE AND IN ALL SECONDARY SCHOOLS IN KENYA, 1968 Boys Girls Household or Paternal Traits All In Elite All In Elite School School Percentages Father resides in town 9 28 18 37 Father's schooling: Secondary or more 5 41 11 28 None 42 4 21 16 Father's occupation: Upper white collar 8 41 13 35 Manual 36 6 23 9 TV in home 1 18 1 26 Radio in home 42 85 54 70 - 204 - African pupils of two elite schools with those in the other secondary schools in 1968 (Table 7.2). As usually is true, secondary girls came from more advantaged homes than did boys; whether a reversal of this pattern in the elite schools would be found in other societies is not known. In a study of secondary schools in Buenos Aires, Petty reported the proportions of parents in each school who had completed secondary school or more (Table 7.3). Fathers typically were better educated, but the schools were quite varied for both fathers and mothers, and schools with better-schooled fathers tended also to have better-schooled mothers. Using a full distribution of parental schooling in each school, in half the cases in index of dissimilarity between fathers and miothers was between 20 and 30; it was over 30 in a quarter and under 20 in a quarter of schools. Hutchinson (p. 81) provides a similar tabulation for ginasios of Sao Paulo in the late 1950s. Children of businessmen made up a sizeable minority or a majority in all seventeen schools, whatever the classroom composition otherwise. Children of manual workers were few and never a large minority although in one school they comprised a quarter of the students. One or two distinctively elite schools are suggested by the distribu- tions of children of professional men and businessmen. In Argentina, as in some studies from Africa, the general pattern is one of some heterogeneity in most schools and of an elite clientele for a few schools. The Sao Paulo schools are indicative of considerable selectivity overall, given the scanty representation from homes of manual workers. To understand what is happening in an educational system, both within the schools and for selectivity into them, it is necessary to know much more about classroom composition than usually is recognized, despite - 205 - TABLE 7.3 DISTRIBUTIONS OF SECONDARY SCHOOLS BY PROPORTIONS OF FATHERS AND OF MOTHERS WHO HAD COMPLETED .5ECONDARY SCHOOLING OR MORE; BUENOS AIRES Percentage completing Number of Schools secondary school or more Fathers Mothers 0- 19 8 20 20 - 39 17 7 40 - 59 6 9 60 - 79 6 1 80 and over 2 2 Total 39 39 TABLE 7.4 DISTRIBUTIONS OF SECONDARY SCHOOLS BY PERCENTAGES OF STUDENTS' FATHERS IN DESIGN.AtED OCCUPATIONS: SAO PAULO Percentage Father's Occupation Other Professional Businessman White Collar Manual 0- 9 2 2 10 10 - 19 6 3 6 20 - 29 5 3 6 1 30 - 39 1 6 5 40 - 49 5 1 50 - 59 1 2 60 - 69 1 70 - 79 1 1 Total 17 17 17 17 Percentage of total 27 41 24 8 - 206 - widespread assumptions about the effects of peer groups on student attitudes and performance. One cannot assume, furthermore, that classroom composition as measured by parental schooling or occupation will operate in the same ways in different settings or in the same country over time. B. Selectivity by Type of Secondary Curriculum Socially dissimilar schools are most readily categorized by type of curriculum. Such distinctions underlay the long debate in Britain about 'parity of esteem" between types of schools. German Gymnasia enjoy special esteem even though they certainly do not recruit a socially uniform student body. Even in populist Finland, despite proverbially small sex differences in opportunity, status differentiation by type of secondary school is far from minimal. This sort of differentiation differs greatly from one country and one time to another; see Tables 7.5 and 7.6 with reference to parental schooling and occupation, respectively. Looking first at the African countries, the most striking feature is the weak association between type of curriculum and father's schooling. Indexes of dissimilarity by type of course ranged from 2 (between the long and the short modern curricula in Senegal) to 16 (between grammar and, modern schools in Western Nigeria). Selectivity into secondary schools by father's education in these countries is substantial, nevertheless. In Zaire the overall index of dissimilarity was a high 64, although the indexes of dissimilarity between general and vocational schools and between those in rural and in urban areas were quite low. The low indexes on these between-' type comparisons cannot be dismissed merely as a reflection of the early development of secondary education in view of the much higher selectivity on the same indicators into secondary schooling as a whole., 207 - TABLE 7.5 DISTRIBUTIONS OF EDUCATION OF PARENTS BY TYPE OF SCHOOL OR COURSE: SECONDARY STUDENTS Parents Education Some of Indexes- of Some Full Full Dissimilarity Overall Country and Type Elemen- Elemen- Elemen- Higher Among School Index of Secondary Education None tary tary tary Education Total Categories Dissimilarit- SENEGAL 1962 Fathers Long-classic course 47 23 18 12 100 14 Long modern 59 25 12 4 100 14 Short 59 27 12 4 100 2 WEST NIGERIA Fathers Grammar 53 25 5 12 5 100 16 Modern 69 26 2 3 0 100 Mothers Grammar 74 18 6 1 100 14 Modern 88 11 1 0 100 ZAIRE Fathers (All 64) Rural Schools General 32 28 40 100 7 Vocational 38 29 33 100 Urban Schools General 9 36 20 35 100 11 Vocational 10 27 18 45 100 All Rural 35 28 37 100 10 All Urban 40 18 42 100 WEST GERMANY 1979 Fathers Boys 15+ Gymnasium 48 24a . 28a 100 30 Realschule 78 i8a 4a 100 31 Hauptschule 79 128 ga 100 Girls 15+ Gymnasium 42 26a 32a 100 35 Realschule 77 188 5a 100 36 Haupteschule 78 12a lo 100 6 All 15+ Gymnasium 46 24a 30a 100 31 Realschule 77 18a 5a 100 33 Hauptschule 79 12a ga 100 6 FINLAND 1965 Fathers (Male 17) Boys: To gymnasium 61 16b 23b 100 16 Other Jr. Sec. Leavens 74 lgb 7b 100 (Female 10) Girls: To Gymnasium 67 17b 16b 100 2 Others Jr. Sec. Leavens 87 gb 4b 100 ° a/ Fathers who were beyond realschule are entered as "higher". b/ Fathers who nassed examinations at completion of gymnasiym are entered as "higher" Continued (Japan) - 208 - TABLE 7.5 Parents Education Some of Some Full Full Elemen- Elemen- Second- Junior None tary tary ary College University Total JAPAN 1966 Fathers Boys: Academic General 43 32 13 12 100 Other General 71 23 4 2 100 Commerce 71 23 4 2 100 Agriculture 74 22 3 1 1OO Technical 67 25 5 3 100 Girls: Academic General 34 37 16 13 100 Other General 65 24 7 4 100 Commerce 65 27 6 2 100 Agriculture 77 21 2 0 100 Domestic Science 67 23 6 4 100 Mothers Girls: Academic General 31 63 6 100 Other General 66 32 2 100 Commerce 67 32 1 100 Agriculture 63 36 1 100 Domestic Science 59 41 6 100 JAPAN: INDEXES OF DISSIMILARITY AMONG COURSES Overall Index of Dissimilarity Academic Other Commerce Agriculture General General All Fathers of Boys 8 Non-academic general 28 Commerce 28 0 - Agriculture 31 3 3 - Technical 24 4 4 7 Fathers of Girls 10 Non-academic General 31L - Commerce 31 3 - Agriculture 43 12 12 - Domestic Science 33 2 4 10 Mothers of Girls Non--academic General 35 - Commerce 36 1 - Agriculture 32 4 4 - Domestic Science 28 9 9 5 - 209 - Although selectivity by father's occupation between types of secondary schools in Zaire is slightly greater than selectivity by father's education, it does not alter the main picture (Table 7.6). Western Nigeria is another story, however; even disregarding sex, the indexes of dissimilarity by type of school exceed the estimated selectivity into secondary education as a whole. Moreover, when the data distinguish by sex the differentiation between grammar and modern turns out to be much sharper, at 47 percent for males and 41 percent for females. This shift in the indicators reflects differences in the sex composition of the modern and grammar schools and the small proportions of girls compared with boys in each type of school who were offspring of farmers and of unskilled workers. In contrast to Nigeria, in the Ivory Coast (as of 1963) -the between-curriculum index of dissimilarity is lower on father's occupation than is the index for all secondary schools combined. The relatively egalitarian character of education in Malaysia was stressed in earlier chapters. As of 1972 the overall secondary-school index of dissimilarity by father's occupation was 25. Here again, as in Nigeria, differentiation within secondary education appears more important when a distinction is made by sex, but that contrast is almost entirely between the urban and rural academic schools, not between academic, technical, and voca- tional schools-a situation very different from the modern-grammar contrast in Nigeria. As is to be expected, the rural-urban contrast in pupils' socio- economic status is greater for females than for males. Having access to voluminous 1966 data about parents of Japanese secondary pupils, several comparisons are given in Tables 7.5 and 7.6. Academic secondary schools are distinctive for parental background; the sharp contrasts between the academic-general courses and all others compared with - 210 - TABLE 7.6 DISTRIBUTION OF OCCUPATIONS OF ':ATHERS BY TYPE OF SCHOOL OR COURSE; SECONDARY STUENTS White Collar Proprietor, Including Excluding Indexes of Farm Manual Traders Prorprietors Proprietors (Misc.) Dissimilarity Among School Over Categories All AFRICA Ivory Coast 1963 Modern 72.9 5.3 5.1 19.6 14.5 2.1 Classical 64.5 9.2 1.3 25.0 23.7 1.3 W. Nigeria 1967 21 Modern All 8.4a 4.3a 8.0 14.3 6.3 26.9 Graumzk All 54.5a 7.8a 18.7 37.3 19.0 Modern, Male 77.9a 6.4a 11.4 15.2 3.8 0.5 47.2 Grammar Male 30.7a 21.0a 12.3 44.6 3.7 Modern Female 42.7 19.7 21.2 36.6 15.4 1.0 Grammar Female 8.3a 25.1 14.9 63.2 48.3 3.4 40.7 Zaire 1972 62 Urban 34.8 15.6 12.4 46.3 33.9 3.3 16.8 Rural 23.9 26.8 20.1 48.1 28.0 1.2 MEDITERRANEAN AND SOUTHEAST ASIA Turkey, Lyce 1962 60 Public 24.5 19.5 14.4 56.0 41.6 23.1 Private 9.2 11.7 24.9 79.1 54.2 2 Malaysia 1972 . 25 Male: Urban academic 7.5 22.3 30.9 57.4 26.5 12.8 36.2 Rural Academic 41.8 15.2 17.3 28.3 11.0 14.7 Technical 22.8 20.5 16.9 44.9 28.0 11.8 Vocational 23.9 29.6 21.4 37.1 15.7 9.4 Female: Urban Academic 5.6 22.8 29.8 64.7 34.9 6.9 49.6 Rural Academic 48.0 10.7 15.0 27.2 12.2 14%1 EUROPE AND AUSTRALIA East Germany 1967 Full-time 11.9 53.1 5.0 35.0 30.0 5.0 33.4 Part-time 5.1 31.2 9.3 63.7 63.4 0.3 1960 Full-time 9.4 59.0 6.9 31.6 24.7 6.9 1960 Full-Time 9.4 59.0 6.9 31.6 24.7 6.9 20 5 Part-time 8.8 44.0 2.0 47.2 45.2 2.0 2 West Germany 1972 50 All: Hauptachul 7.4 63.5 7.8 29.1 21.3 Realschul 5.8 45.3 14.7 48.9 34.2 48.2 Gymnasium 3.7 19.0 20.7 77.3 56.6 8 Male: Haupt. 7.3 63.0 8.0 29.7 21.7 Real. 4.1 48.4 14.8 47.5 32.7 47.2 Femi G Hf -aupt 7 3' 6449 2): 8 0 6 8 Real 7.4 42.6 14.6 50.0 35.4 49.9 Gymn. 4.3 17.8 22.8 77.9 55.1 Sweden 1960 37 Gymnasium 1at year 8.3 24.4 11.6 67.3 55.7 General 7.5 22.5 10.4 70.0 59.7 21.9 Technical 7.5 22.5 17.7 55.5 37.8 Australia, 1961, Melbourne All private 5.8 11.5 82.7 All public 1.9 26.1 72.0 23.0 Protestant 9.2 3.1 87.7 Catholic 3.1 18.3 78.6 a/ Unskilled lapor included wih farmers - 211 - the low overall selectivity into upper-secondary schools highlight the fact that social selectivity in education is much more than selectivity into successive years in school. Understandably, there has been tremendous pressure in Japan over the past fifteen years to get one's children into the general- academic curricula, even .as upper-secondary enrollment rates have approached universality.- C. Public, Private, and Religious Schools Private secondary schools exist in many countries and in some countries there is a variety of religiously controlled schools. Some data on selectivity by agencies of control are brought together in Tables 7.7 and 7.8. Mission schools have of course played a central part in the history of education in Sub-Saharan Africa, and their direct descendants in the school system are still important in some African countries. Moreover, private demand for secondary schooling has generated private schools not only in countries that have generally favored private enterprise but even in Tanzania, despite efforts to restrain that development. The only African country for which information on social selectivity by type of control was found was the Cameroons, and there sponsorship of schools did not appear to have much effect on backgrounds of pupils. The greatest contrast is between the private religious schools (having the most elite students) and the public schools, but even these differences are only half of the selectivity for secondary education as a whole. Much greater contrasts between public and private schools are manifest in Turkey and in Bolivia; one knows the differences are considerable in many other countries but supporting data are lacking. - 212 - TABLE 7.7 DISTRIBUTIONS OF EDUCATION OF PARENTS OF SECONDARY STUDENTS Y PUBLIC PRIVATE, AND RELIGIOUS SCHOOLS Parents Education Some of Some Full Full Indexes of Country and Type Elemen- Elemen- Second- Dissimi- of Secondary Education None tary tary ary Higher Total larity E. CAMEROON Fathers Public 18 25 25 32 100 18 Private Religious 8 17 28 .47 100 10 12 Private Secular 9 26 22 43 100 10 W. CAMEROON IFathers Public 19 19 30 32 100 17 Private 12 9 30 49 100 TURKEY Fathers All 43 Public 38 21 33 8 100 29 Private 18 12 38 32 100 Mothers Public 60 21 18 1 100 37 Private 27 17 46 10 100 BOLIVIA Fathers All 55 Public La Paz 2 44 35 19 100 Public Provincial 10 71 15 4 100 3 Catholic 1 17 28 54 100 63 Other Private 1 19 33 47 100 CHILE Fathers All 26 Public 4 23 °5 32 6 100 28 Private 2 11 21 44 22 100 Liceo 1 7 23 59 10 100 Top over primary 7 30 39 22 2 100 TRINIDAD Fathers Government - 22 32 33 13 100 Assisted - 24 24 52 14 100 81 Private - 31 32 32 5 .100 15 TABLE 7.8 DISTRIBUTION OF OCCUPATIONScOF FATHERS OF SECONDARY STUDENTS IN PUBLIC, PRIVATE, AND RELIGIOUS SCHOOLS White Collar Unclas- Proprietor, Including Excluding sifted Indexes of Farm Manual Traders Prorprietors Proprietors (Misc.) Dissimilarity Among School Over Categories All AFRICA Ea3t Cameroon 1971. 44 Public 20.2 13.2 17.8 66.6 48.8 Private Religious 4.8 12.7 13.1 82.5 69.5 20.7 Private Secular 1.5 17.3 6.6 71.2 64.5 West Cameroon 1971 47 Public 23.1 10.7 9.,2 66.2 57.0 9.6 Private 20.6 6.4 6.4 73.0 66.6 SOUTH AMERICA Boliv4-a 1975, Grade 12 a a 39 Urban public 17.4 45.0 37.6 45.8 Urban privatel 3.1 13.5 83.4 Rural 17.9 53.5 28.6 Bolivia 1971 59 La Paz city, Public 7 36 12 57 45 30 L.a Paz Department: All 7.8 22.6 8.8 69.6 60.8 Catholic 4 13 8 83 75 54 Others private 4 18 8 78 70 Provincial Public 39 32 5 29 24 Chile 1969, Public Liceo 39 Santiago 2 10 8 88 80 Provincial Cities: Large 3 16 7' 81 74 19 Medium 5 14 12 81 69 Small 7 15 17 78 61 - 214'- Considerable income selectivity into private secondary schools is almost inevitable unless the private schools are privatey subsidized--only occasionally are public subsidies substantial. Private schools are not everywhere the "best" schools; indeed, the opposite situation may prevail where private schools mainly instruct children who have not been particularly successful academically. Furthermore, in some countries private religious schools serve populations that diverge culturally from those whose children attend the public schools, The latter parents often are seen as under-educated or as engaged in humble occupations. But private schools (lay or religious) need not compromise egalitarian aims. There are large differences across societies in this respect and in the conditions that foster private schools. By the accidents of foreign contact or of the stuggle of disparaged minorities to thrive, religious schools often launch a minority onto strong patronage of formal schools. The geographic accident of where missions were located can foster schooling among groups that enjoy only weak traditional positions in their society. Myers and authors of investigations of secondary schools in the capitals of three Latin American countriess-Fischer, Cleary, Petty--delineated the staLus patterns of public and private secondary schools. For each school they computed an average SES score for pupils and also the standard deviation among these scores. Scattergrams of their data show no neat pattern of association between mean and variance, though each capital was different. Academic were distinguished from vocational schools. Generally, in each country the private-academic schools were distinctively selective compared with the public-academic, and vocational schools were still different, tending to draw pupils from lower-status homes. The selectivity index for white-collar - 215 fathers was for each city as a wl-ole: La Paz 4.0, Santiago 2.9, and Buenos Aires 1.3 while for private-academic schools the indexes were 5.1, 4.8, and 1.7. Each country had its own institutional arrangement for utilizing different sorts of secondary schools by families of various status levels. Of all schools, in La Paz 78%, in Santiago 70%, and in Buenos Aires only 42% had standard deviations for paternal status of 1.4 units or more. The dispersion was about the same for public as for private schools and about the same overall for the academic as for the vocational schools. Public- secondary were less exclusive than private-secondary schools in La Paz and in Santiago but the reverse was distinctly the case in Buenos Aires. In none of these cities were academic schools of uniformly high status in clientele. The authors concluded that eliminating private schools may not reduce social selectivity of secondary education, nor will opening more vocational schools necessarily make secondary education more representative. D. Differentiation within Higher Education Nearly a century ago statisticians in the Kingdom of Serbia published information on the social backgrounds of students (and much other information) about the various segments of the educational system. I have computed select- ivity indexes relative to the distribution of father's occupation of boys in primary school (Table 7.9). In the first row one observes a clear selectivity of girls for beginning primary school; girls from any but privileged homes represented small proportions of the children who moved up through the system. At all post-primary levels most pupils were males. Throughout, children of farmers were decidedly under-represented, except in teachers' colleges where the ratio reached .57; for farm children other indexes, regardless of level of 216- TABLE 7.9 OCCUPATIONS OF FATHERS OF STUDENTS IN VARIOUS TYPES OF SCHOOLS; SERBIA, 1985 Clergy Officials Teachers Businessmen Artisans Farmers Percentage Distribution of Boys in Primary Schools (Base population) 1.1 2.0 0.6 6.4 10.9 79.0 Selectivity Indexes against Primary Boys Primary girls 1.9 5.2 3.2 4.2 3.6 .24 Gimnasium-lower 6.4 7.5 5.4 4.5 2.1 .29 -upper 5.8 11.7 11.1 3.4 2.2 .22 Ecoles reales 3.8 9.7 6.4 3.4 2.8 .26 Teachers Colleges 5.2 2.3 16.2 1.2 2.9 .57 University 9.1 15.5 15.6 3.4 1.3 .18 - 217 - school, were under .20. The position of the students in teachers colleges a century ago in- Serbia iS particularly interesting in view of the prevalence of just such a situation over most of the world--though perhaps with somewhat less inheritance in teaching than is suggested by the data for Serbia. It is of interest also that relative representation of children of officials and of businessmen in the teachers colleges was lower than in any other category (even than girls in primary school), though for all other schools the highest selectivity ratios were for children of officials. Along with the occupational inheritance suggested by the high selectivity of children of teachers into teachers' colleges is a selectivity index on teachers that matches that on officials for enrollment in universities. These data relate to a phenomenon that has been widely observed: upward mobility first into teaching and then in the ensuing generation through higher education to other occupations with high status. The only comparable data (in Tables 7.10 and 7.11 which display the evidence gleaned indirectly in the present study) are for Kenya and for the United Kingdom. In Kenya the contrast in social selectivity between entrants into the university and into training for secondary teaching is clear, with indexes of dissimilarity of 27 against both father's occupation and father's schooling. The discrepancy on father's schooling is concentrated entirely in the marked difference between universities and normal schools in the proportions of students whose fathers were totally unschooled. In the urban household survey for the United Kingdom in 1961, 43% of the students in teacher training against 26 percent of those in universities were from homes of manual workers. - 218 - TABLE 7.10 DISTRIBUTIONS OF EDUCATION OF PARENTS BY TYPE OF INSTITUTION; UNIVERSITY STUDENTS Parents Education Some of Some Full Full Inexes of Country and Type Elemen- Elemen- Elemen- Dissimi- of Higher Institutions None tary tary tary Higher Total larity KENYA Fathers University Students 21 57 18 4a 100 Secondary-Teacher Training 48 47 5 0 100 27 GHANA Fathers Entry via Form 6 24 42 34 100 28 Entry by other routes 36 58 6 100 NIGERIA Fathers Bello 37 48 12 3 100 14 IBadan Mothers 653 31 11 4 100 Bello6336 04 09 Ibadan 72 22 6 0 100 EGYPT Fathers Cairo University 7 8 19 42 24 100 48 Alazhar 28 .28 33 21 18 0 100 JAPAN (1968 entrants) Fathers 44 Universities National 34 36 30 100 Public 33 40 27 100 3 Private 30 41 29 100 3 Junior College Public 38 40 22 100 10 Private 28 45 27 100 Night University or College Mothers Universities National 37 56 7 100 Public other 34 59 7 100 3 3 0 Private 34 59 7 100 Junior College Public 41 55 4 100 10 Private 31 63 6 100 Index of dissimilarity for night university or college versus: National University 16 Private Junior College 22 - 219 - TABLE 7.10 continued Parents' Education Indexes of Dissimilarity Compulsory to Grade Upper Among 8 or 9 or less Secondary. Higher Total Institutions Overall UNITED STATES 1920 Fsthers: Bys Non-state Co-ed 43 35 22 100 16 State univ. 27 54 19 100 19 Men's College 24 37 39 100 20 Girls Non-state Co-ed 53 30 17 100 28 State Univ. 25 49 26 100 14 40 Womens' College 13 47 40 100 Mother: Boys Non-state Co-ed 53 34 13 100 State Univ. 39 50 11 100 14 33 Mens' College 20 57 23 100 19 Girls Non-state Co-ed 28 57 15 100 State Univ. 21 62 17 100 7 17 Women's College 11 65 24 100 10 Index of Elemen- Elemen- Dissini- CANADA, 1975 tary only Secondary Higher Total tary only Secondary Higher larity All under Age 26; Father'z Education 23.9 39.8 37.4 100.0 .63 1.12 1.35 13.8 All student ages; both sexes Fathers of University Undergraduates 22.9 48.5 38.6 100.0 .61 1.41 1.03 14.8 Comunity College Students 29.4 51.9 18.7 100.0 .78 1.50 .68 17.4 Fathers: of Graduate Students Full-time: Male 22.4 36.4 41.2 100.0 .59 1.06 1.49 15.3 Female 14.4 31.0 54.6 100.0 .38 .90 1.97 26.9 Part-time Male 31.9 33.7 34.4 100.0 .85 .98 1.24 6.7 Female 26.0 33.7 40.3 100.0 .69 .98 1.45 12.6 Fathers of students in Professional Schools Male 14.3 36.7 49.0 100.0 .38 1.06 1.77 23.4 Female 14.4 30.9 54.7 100;0 .38 .90 1.97 27.0 Mothers: of University undergratuates 21.3 54.8 23.9 100.0 .61 1.34 1.00 1.38 Community College students 28.5 57.3 14.2 100.0 .81 1.40 .59 16.3 Graduate Students Full-time Male 25.3 45.6 29.1 100.0 .72 1.11 1.22 9.8 Female 12.7 42.2 34.1 100.0 .36 1.04 1.89 22.4 Part-Time Male 31.9 44.9 23.2 100.0 .91 1.10 .97 3.9 Female 21.9 38.6 39.5 100.0 .62 .94 1.65 15.6 Students in Professional Schools Male 16.1 47.5 36.4 100.0 .46 1.16 1.52 19.0 Female 16.3 45.4 38.3 100.0 .46 1.11 1.60 18.8 - 220 - TABLE 7.11 DISTRIBUTIONS OF FATHER'S OCCUPATION BY TYPE OF INSTITUTION UNIVERSITY STUDENTS Indexes of Uhite Collar Dissimilarity Indexes of Proprietors, Including Excluding Among Dissimilarity Farm Manual Traders Proprietors Proprietors Misc. Institutions Overall AFRICA AND MEDITERRANEAN Kenya 1970 Univ. of Nairobi 44.0 1 55 27 Teacher Training 60.0 12 28 Ghana 1965 University 40.6 6.5 3.8 52.9 49.1 University College 50.8 9 8 5 2 39 4 34 2 13.5 Legon 1963 41.9 11.4' 14.4 46.7 32.3 Cape Coast 50.7 11.8 13.0 37.5 24.5 Nigeria Univ. of Ibadan 1960-65 31.6 8.3 16.9 60.1 43.2 Ibadan 1972 55.2 4.0 16.1 40.8 24.7 A.B.U. Univ. 1972 60.8 3.9 16.6 35.3 18.7 Nsukka 1972 38.2 4.2 24.5 57.6 33.1 Akmodo Bello 1972 All (mostly male) 57.3 23.7 10.l 19.0 8.2 Female: All 16.8 31.9 17.8 51.3 .33.5 Christian 20.2 31.4 21.8 48.4 26t6 Huslim 10.3 29.3 10.3 60.4 50.1 23.5 Egypt 1960 Cairo Univ. 6.0 5.8 88.2 Allezar Univ. 49.4 7.8 42.8 45.4 EUROPE U.S.S.R. 1964 Full-time 19.6 39.4 41.0 Part-tims 2.3 50.6 47.1 26.3 Correspondence 7.0 25.7 67.3 East Germany 1960 Full-time 4.3 51.8 8.2 43.9 35.7 Part time 0.8 7.3 2.0 91.9 89.9 54.2 1967 Full-time 840 39.5 7.3 52.5 45.2 Part-time 1.8 11.9 1.4 86.3 84.9 39.7 Latvia 1970 Day University 11.5 39.9 48.6 Night University 5.6 48.9 45.5 11.6 Correspondence 15.7 37.3 47.0 France 1967 43 Grandas Ecoles 2.4 a/ 2.5 10.1 86.3 76.2 8.8 Universities 9.5 a/ 14.5 14.6 70.7 56.1 5.3 37.8 IU.T.s 12.8 */ 22.8 12.6 51.0 38.4 13.4 Switzerland 1960 53 University 5.0 13.2 80.5 6.3 Technical Institute 5.7 18.0 75.4 0.9 5.5 United Klngdom 1961. University 26.0 74.0 Teacher Training 43.0 57.0 17.0 'Further Education" 39.6 60.4 (Advanced) JAPAN AND UNITED STATES Japan 1968 32 National University 14.3 6.1 R8.6 79.1 70.5 Public University 10.8 12.0 9.9 77.2 67.3 7.2 Private Univereitj 10.8 9.7 9.3 79.5 70.2 2-year Public 18.2 7.3 8.4 74.1 65.7 7.4 2-yemr Private 13.3 5.3 9.8 81.4 71.6 7.3 United States 1920's 38 Private College & Univ. 21.1 12.5 33.5 66.4 32.9 State University 27.4 12.6 31.9 60.0 28.1 6.8 Private Junior College 27.6 8.6 37.3 63.8 26.5 14.5 Public Junior College 14.9 19.3 29.9 65.8 35.9 20.5 a/ Includes fars labor. - 221 - Studies of socio-economic, backgrounds of college and university students enrolled in different courses have been made in a variety of countries, along with the more frequent studies with one or another measure of "ability" and of performance on examinations. There also are studies of the backgrounds (usually the occupations) of fathers of men in particular professions. Few of these sets of data are compatible with the analyses in this report. 1/ What has been gathered together displays the distinctiveness of each system of higher education rather than common featvtres of such systems. Ghanaian students who in 1971 entered university from Form 6 came more often from relatively educated homes than did those who entered by less standardized routes; this is common enough in most countries. Peil reports (p. 22) that the minority of entrants to a Nigerian university who came in at a normal age were more likely to acquire a superior degree, especially with high honors; those individuals were more often born in cities and had attended an elite secondary school. Among Ibadan students (according to van den Berghe's report, p. 155) offspring of farmers made up only 18% of entrants below age 20 but 55% of entrants after age 30; the latter were likely to have spent an interval in teaching. Among the Ibadan students in medicine 44% had professional or semi-professional fathers in contrast to 18% who were sons of farmers; in pedagogy the respective percentages were 13 and 59. The contrast between a "western" university in Egypt and the very traditional Al Azhar is dramatic; students in the latter were likely 1/ Intensive research undoubtedly would unearth data about curricula that could be turned to that use. This has been done for France, but time did not allow pursuit of this topic (Millot). Much more is known about selection for ability than for socio-economic status in higher education, and there is a risk that generalizations will confound these two aspects of selectivity. - 222 - to have little-schooled fathers due largely to the heavy representation from peasant homes in the study of theology (Shafshak 1964).. In most (but not all) countries there are numerical limits on admissions to certain curricula in universities while expansion has occurred mainly in the humanities and social sciences. The restricted or "Eprivileged" faculties are heavily subsidized and they offer the best prospects of a high lifetime income. In my study of social selection into Swedish universities I drew on data showing that admissions to the restricted courses showed a distinctive proportion of students from high-status homes. The ratio of restricted entries to those in other faculties were (in 1956) as follows by "social class" of father: I, 1.31; II, .86; III, .75, and during the following dozen years Group I widened its advantage. Distinctions between part-time and full-time students, between those in day and in night courses, and between those studying in residence and those taking correspondence courses display patterns that can be interpreted only in the light of accessibility when at work and in the light of government controls. (There are data for the USSR, East Germany, and Latvia, information about the very different situation of night students in Japan, and part-time students in Canada.) The distributions of parental status for college sttidents in the United States sixty years ago (Table 7.11) portray the traditional hierarchy among types of colleges and individual colleges loosely associated with varying degrees of social selectivity. Not only was the system of higher education at that time already comparatively large; it had long been distinc- tively populist. One-sex colleges (especially those for girls) had the least representative rosters, although few of these schools were socially elite and - 223 - several were academically outstanding,, For the one-sex schools the indexes of dissimilarity with the base population were distinctively high. Japan today broadly resembles the United States of a generation ago in its system of higher educatioh; there are also continuing differences: intensity of examina- tion competition, distinctive junior colleges for women, and limited proportions of women in the four-year institutions. Contrasts between the short-course junior colleges and four-year colleges in Japan are not always as sharp as might be expected. There is a major contrast in social status selectivity Into the elite French Grandes Ecoles on the one hand the the IUTs (technical institutions attached to universities), to be sure. But in Canada the universities are only slightly more selective for father's education than the community colleges. In Japan public junior colleges take in a larger proportion of students whose parents had no more than compulsory schooling than do the four-year institutions. But the private junior colleges (with their large female contingent) are more selective for social status than any of the other categories of private institutions. (The famous national universities are extremely selective on examinations but not on social background.) In the United States back in 1920 the public junior colleges had smaller proportions of farmers' children among their students than did any other category of higher institution (bottom of Table 7.11). The private junior collegtes had the largest proportions of farmers' children and the smallest representation of manual workers' children. An interesting twist to these figures shows up in the column on proportions of students from white-collar homes (exclusive of proprietors and traders who were numerous); the largest representation of the white-collar families was in the public junior colleges followed by private four-year colleges - 224 - and universities. These distributions reflect the structure of the economy at that time, the regional distribution of higher institutions, the contrasts over the nation in the varieties of education offered in and expected from the private and public junior colleges, and the earlier history of educational developments in each region of the country. Few readers would be surprised to learn that social selectivity usually is greater for tertiary than for secondary schools, even where the secondary enrollments are low. Yet this is not always the case; a shift toward a meritocratic selection at a given stage in the schooling can substantially alter the selection pattern. In extreme cases (as with Japanese national universities) it can even reverse some of the relationships. Over much of the third world tendencies toward rising selectivity in successive stages of the schools are muted by the use of examinations and by public stipends. The proliferation of a variety of post-secondary institutions further complicates the flows of youth through the educational system and on into their post-school lives. - 225 - Chapter 8 INCOME SELECTIVITY OF EDUCATIONAL PARTICIPATION Parental schooling, income, and occupation are moderately inter- related, but they are distinct attributes. The degree of selectivity may well differ for each, and the association among such characteristics can difft-r sub- stantially among countries. Sources of information about income selectivity overlap little with sources of information on selectivity by schooling or occupation. Most of the information about income distributions for families of students and Lor corresponding base populations has to be taken from studies conducted for other purposes. There are several recent studies estimating the distributive incidence of public expenditures, often including information about the incidence of taxes. Relevant information has been obtained from reports on Colombia and Malaysia and from ongoing studies for France and for Chile--all of which have been or are being sponsored by the World Bank. (There may be other untapped materials even within the Bank.) The material presented here is illustrative only; it relates to Japan, Colombia, Malaysia, and West Germany. (Some material on income selectivity for Santiago, Chile is pre- sented in Appendix B as part of the discussion of the importance of using an appropriate base population for selectivity measurement whether one is deal- ing with parental schooling, income, or "SES".) I have not taken time to attempt reorganization of such data in recent studies for individual states in the United States. Tables 8.1 and 8.2 analyze data for Japan, distinguishing income selectivity by types of post-secondary institutions. (For neither table were there data by sex of student.) The classification of types of institutions is more detailed for Table 8.1 and in this table the distributions of parental - 226 - Table S:l INCOME SELECTIVITY TO HIGHER EDUCATION IN JAPAN, 1968; CLASSIFICATION BY ABSOLUTE INCOME LEVELS Income Class (million Yen) Under 0.5-0.9 1.0-1.4 1.5-1.9 2.0-2.9 Over Total 0.5 3.0 Percentage Distributions: Base Population 12.2 54.5 23.4 6.4 2.5 1.0 100.0 All Student Families 3.6 33.4 33.3 15.0 9.1 5.6 100.0 Families of Students in: National Universities 9.1 42.5 31.1 10.5 4.9 1.9 100.0 Other Public Universities 5.4 38.1 36.2 11.6 6.0 2.7 100.0 Private Universities 2.4 30.9 33.7 16.1 10.2 6.7 100.0 Junior Colleges: Public 5.6 45.3 33.3 10.8 3.4 1.6 100.0 Private 2.2 31.0 34.2 16.1 10.2 6.3 100.0 index of Selectivity Indexes: Dissimilari All Students .30 .61 1.42 2.34 3.64 5.60 32.7 Students in: National Universities .75 .78 1.33 1.64 1.95 1.90 15e1 Other Public Universities .44 .70 1.55 1.81 2.40 2.70 2302 Private Universities .20 .57 1.44 2.52 4.08 6.70 33,4 Junior Colleges: Public .46 .61 1.42 1.69 1.36 1.60 15e8 Private .18 .56 1.47 2.52 4.08 6,30 33.5 Source: Teichler, 1972 - 227 - Table 8.2: INCOME SELECTIVITY INTO JAPANESE UNIVERSITIES IN 1961 AND 1976; CLASSIFICATION BY INCOME QUINTILES Index Income Quintiles of - Low 1 2 3 4 5- Dissimilarity, Percentage_Distribut.ons All Base Populations 20 20 20 20 20 1961: All Universities 11.0 13.1 13.5 19.1 43.2 23.2 National 19.7 20.2 15.4 18.5 26.2 6.4 Private 6.4 9.2 12.3 19.2 52.9 32.9 1976: All Universities 7.2 8.7 11.6 25.2 47.3 32.5 National 12.7 12.3 15.1 24.5 35.4 19.9 Private 5.6 7.7 10.6 25.4 50.7 36.1 Selectivity indexes 1961: All Universities .55 .65 .68 .96 2.16 National .99 1.01 .77 .93 1.31 Private .32 .46 .62 .96 2.65 1976: All Universities .30 .44 .58 1.26 2.36 National .64 .62 .76 1.22 1.77 P-rivate .28 .39 .53 1.27 2.54 Changes in Percentages 1961 to 1976 (1976 minus 1961): All Universities -3.8 -4.4 -1.9 +6.1 +4.1 +9.3 National -7.0 -7.9 -0.3 +6.0 +9.2 +13.5 Private -0.8 -1.5 -1.7 +6.2 -2.2 +3.2 Source: Cummings, 1980, p. 226. Table 8.3: INCOME ANMI EDUCATIONAL SELECTIVITY IN COLOMBIA Percentage Distributions Selectivity Indexes and Indexes of Dissimilarity Household Children All Institutions Public Institutions Private Institutions income (youth) in All Public Private S.I, or, S.T. on S.I. on S.I. on S.I. on S.I. on Quintile Households Age Cohort Institutions Institutions Institutions Household Child Household Child Household Child a/ Quintiles Population Quintiles Population Quintiles Population (1) (2) (3) (4) (5) a (6) (7) (8) (9) (10) A. ELEMENTARY SCHOOL 1 low 20 33.56 28.4 32.1 11.9 1 1.43 .85 1.61 .96 .60 .35 2 20 24.55 25.5 21.9 13:8 1 1.28 1.04 1.40 1.14 .69 .86 3 20 18.97 9.9 20.0 19.3 I 1.00 1.05 1.00 1.05 ."7 1.02 4 20 13.26 15.2 14,2 19 8 .76 1.15 .71 1. 0 .rig 1.1!' 5 high 20 9.66 10.8 15.8 19:8 1 .54 1.18 .29 .60 1.76 3.64 Total 100 100.00 100.0 100.0 100.0 Index of Dissimilarity 14.1% 4.96% 20.0% 5.32% 15.2Z 2.4% Far-ing Farming poor poor B. SECONDARY SCHOOL 1 low 20 25.28 13.4 16.1 9.8 .67 .53 .84 .66 .49 .39 ~ 2 20 23.10 11.7 23.0 9.8 .84 .72 1.15 1.00 .49 .42 s 3 20 19.86 17.9 21.1 14.5 .90 .90 1.06 1.06 .73 .73 4 20 17.32 23.8 24.9 22,5 1.19 1.37 1.25 1.25 1.13 1.30 20 14.44 28.2 14.3 13.4 1.41 1.95 .71 .71 2.17 3.01 'Total 100 100.00 100.0 100. 0 100.0 Tn,d of Pi6siilarity 12.0% 20.24% 9.0% 8.82% 25.9% 34.14% C. HIGHER INSTITUTIONS 1 low 20 (25.28) 0.5 0.9 zero .03 02 .05 .04 zero zera 2 20 (23.10) 3.8 4.9 2.3 .19 .16 .25 .21 .11 .10 3 20 (19.86) 8.1 10.9 4.7 .41 .41 .54 .55 .24 .24 4 20 (17.32) 20.4 23.8 16.5 1.18 1.18 1.19 i.37 .83 .95 5 20 (14.44) 67.2 59.5 76:5 4.65 4.65 2.98 4.12 3.13 5.30 Total 100 I Index of Dissimilarity 17.6% 55.84% 43.3% 51.54% 56.5% 62.06% a. Age 6-12 for elementary; age 13-19 for secondary and for higher. The age 13-19 base is of course far too young for higher education but was all that was available. Source: Computer form data in Marcelo Selowsky's Who Benefits From Public Expenditure; A Case Study of Colombia, Oxford University Press, 1979 (for the World Bank). - 229 - Table 8.4: INCOME AND EDUCATIONAL SELECTIVITY IN MALAYSIA Percentage Distributions Household Quintile of Per Capita Household's Pupils D1ifference Income Per Capita or in Selectivity Quintile Income Students Percentages Index (1) (2) (3) (4) (5) A. ELEMENTARY SCHOOL: 1 low 20 18.76 -1.24 .94 2 20 18.98 -1.02 .95 3 20 20.53 + .53 1.03 4 20 21.85 +1.85 1.10 5 high 20 19.87 - .13 1.00 Total 100 Index of Dissimilarity 2.39 B. SECONDARY SCHOOL: 1 low 20 16.67 -3.33 .84 2 20 16.67 -3.33 .84 3 20 20.20 + .20 1.01 4 20 22.24 +2.22 1.11 5 high 20 24.24 +4.24 1.21 Total 100 Index of Dissimiliarity 6.66 C. HIGHER EDUCATION: 1 low 20 5.69 -14.31 .28 2 20 9.76 -10.24 .49 3 20 18.70 - 1.30 .94 4 20 21.14 + 1.14 1.05 5 high 20 44.72 +24.72 2.24 Total 100 Index of Dissimilarity 25.86 Source: Computed from data in Meerman, 1980. - 230 - and base incomes are by absolute intervals of income. 1/ For the total of students and for those in most segments of higher education the selectivity indexes rise monotonically with fa,mily income, as might be expected. However, selectivity at the top income level is less than at the next-to-top category in national universities; also there is no systematic pattern for public junior colleges in selectivity from the highest three income categories. National universities and public junior colleges also display the lowest indexes of dissimilarity from the base population--less than half the indexes for private universities or private junior colleges (the latter being attended mainly by girls). To one who is familiar with the Japanese system these results are no surprise but they nonetheless warrant comment. Over half of the students in national universities and in public junior colleges come from families with incomes under a million yen, whereas the analogous proportion among students in the other institutions ranged from a third to two-fifths. The reasons for relatively strong representation of lower-income families in the natioaal universities are not the same as for the public junior colleges. Students in the national universities are the most throughly winnowed by examinations though they pay the lowest tuition--at that date by a wide margin. Despite 1. Here the base is parents of all secondary graduates of 1968; excluding parents of children who were not in secondary school understates the actual income selectivity into higher education. On the other hand, we are dealing with direct entrants into higher institutions from upper secondary school; this excludes individuals who entered subsequently after being for a year or more a "ronin" (in crash programs to retake examinations). The direction of bias from this exclusion is uncertain. For individuals going directly from secondary to higher schools, the estima-tes of Table 3.1 seem to permit comparisons among types of institutions--whatever the common bias because from the base population were excluded parents of youth who did not attend secondary school. (Dropouts from secondary schools in Japan are miniscule.) - 231 - effects of family income upon access to the special programs preparing for entrance examinations, selectivity into the national universities by academic performance modifies the inhibitions of low income on entry to university among the lowest two-thirds of the population (incomes under one million yen). Academic selection is least rigorous for the public junior colleges; these tend to be the least costly higher institlutions and they serve youth with weaker academic aspirations. Students in the private junior colleges are more selected for income, but partially this is a sex contrast since most of these students were girls, Comparing the low indexes of dissimilarity relating parents of public junior college students to the base population (and the low selectivity indexes among these students for individuals from homes with over a million yen) with the indexes for other higher institutions provides a good approximate indication of the pure effects of relative cost on income selec- tivity within higher education (see Bowman, 1981; Chapters 6 and 8). Table 8.2, also referring to Japan, deals only with university students, but now the incomes are specified in quintiles (for 1961 and 1976). Selectivity indexes at the top are lower in Table 8.2 than in Table 8.1 because the top income category in Table 8.2 includes 20 percent of the base population whereas the top category in Table 8.1 embraces only Qne percent. It is evident again that income selectivity is less for the national than for the private universities. Nonetheless, in terms of income quintiles selectivity as measured by indexes of dissimilarity increased substantially between 1961 and 1976 for both national and private institutions. The overall increase in in- come selectivity also reflects in part the much greater proportion of univer- sity students who were entering private universitites in 1976 as compared with - 232 - 1961. It is interesting to notice the shift over time in the income-quintile breakingpoint between over- and under-representation in higher schools; in 1961 this break occurred between the fourth and fifth quintiles but in 1976 between the third and fourth. Thls change reflects the spreading of higher education among the population, with a rising attraction of relatively less affluent (though absolutely well-situated) youth into the higher schools. The absolute income level has been as important as the relative income level in this selectivity. Even though analysis by income quintiles facilitates comparing countries for educational selectivity we need to take into considera- tion the level of income along with its relative size. Absolute income levels bring us closer to the sort of analysis used in selectivity for parental. schooling. Shifting to Colombia (Table 8.3) we compare income selectivity to different levels of education. As in Table 8.2 we use household i-neome quintiles but for Colombia one could take account of family size, allocating children of the relevant age group percentagewise among income quintiles. This gives us two sets of selectivity indexes for each level of schooling and for each type of school. Effects of specifying ehe distribution of children among income quintiles are most striking for children in elementary school. (For a further discussion of this aspect of specification of an appropriate base population, see Appendix B.) Indeed, the elementary school selectivity indexes in Column (6) reverse the order of those in column (5) precisely because of the decline of children per household as family incomes rise. Correction for numbers of children aged 13-19 raises the selectivity gradients by household income for secondary-school pupils but without the dramatic - 233 - reversal displayed for the pupils in elementary school. Using the distribution of child population as the base, indexes of dissimilarity rise from elementary to secondary and in turn to higher education in all types of institutions. Throughout, selectivity at the top of the income scale is substantially greater for private than for public schools; at the secondary level the highest income quintile is actually underrepresented in the public institutions, though the contrast is even more striking between public and private elementary schools. These facts tell us something important about Colombian society. For Malaysia, Table 8.4 (like Table 8.3) uses income quintiles, though here for per capita incomes. To thtis extent family size is taken into account; the intent is to reveal the real economic situation of families. A possible additional step would be to distribute children among households by per capita income. The indexes of dissimilarity and the selectivity indexes for the higher levels of income are modest for university students. This reflects policies to bring Malays and the less affluent of all groups into a more favorable position in the economy. The contrast of overall pattern with Colombia is impressive. This becomes increasingly important as enrollment rates for a given level of schooling move toward universality. Measured income selectivity into successive levels of education and into one or another type of institution or curriculum, like selectivity by father's education or occupation, is inevitably in part a reflection of family background. Income selectivity, however, comes as close as is possible to identification of more strictly "economic" factors: "ability to pay" and relative costs to the individual or his (her) family. Those costs are of three kinds: direct outlay on books, tuition, transportation and so forth; foregone earnings; and the interest rates that must be paid directly or - 234 - in the ranking of types of higher institutions by the incomes of their students as shown in Table 3.1 for Japan. They are also involved, though less well identified, in the other data presented in this chapter. This brings me to a point that must be mentioned at least in passing: the effects of public subsidy to higher education, and of "free tuition" on selectivity into higher education. In some countries the public expanditure per student in higher education is as much as 100 times (or even more) the per pupil outlay in village primary school. Far from being egalitarian, as is often assumed, this "free" schooling constitutes a subsidy to the privileged from the large majority of the population who do not gain access even to secondary education. The subsidies may modify the income selectivity indexes for higher education, and there may be severe limitations on alternatives where most people are poor but where there is a societal "need" for more well qualified people--although there are a number of ways of modifying the perverse over-all distributive effects. Nevertheless, this situation gives a special pathos to the fact that educational places often tend to be overfilled for higher and even for secondary schools while there is chronic failure to meet targets for the expansion of elementary schooling. 1/ Interest rates are not counte'i as a cost in benefit-cost analysis; instead the interest rate paid directly or implicitly by the individual is compared with the "internal rate of return" in judging the desirability of the educational investment on strictly "economic" grounds. However, differ- ences among individuals and families in income and wealth affect the interest cost to the family of financing further education (whether by borrowing or simply foregoing interest on alternative investments). For discussions of this see Becker (1965) and Bowman (1981, Chapter 8). - 235 - CHAPTER 9 Conclusions This has been a comprehensive and comparative report on a central topic in the sociology of education. Does the degree of selectivity for schooling vary substantially from one country to another? Does selectivity change as we move from less to more developed societies? For this first thorough survey on the subject the sample of countries is varied though not representative; about a third of all countries were included. At the end of the task, despite all its frustrations, it has been possible to challenge common assumptions about selectivity and to map out main strands in the network of underlying forces. The contributions of this investigation are almost equally in substance and in methodology. Data have been displayed in a prodigality of tables and charts, for attention to the particularity of situations is essential for understanding the broader picture. Regressions were used to explore the interrelated effects of occupational structure, per capita income, and enrollment rates upon educational selectivity. This has not been a study on "education and development." Nevertheless, it is widely believed that with "development," opportunities for education become more selective and that status lines crystalize. Indeed, many writers contend that the period during which "new nations" enjoy a widening of education opportunities will be shorter than that enjoyed by western societies. This view has been shown to be simplistic, probably valid only in the earliest stages of economic and educational development. There is only a tenuous association between per capita income and the degree of social selectivity into schools. At each level of per capita income -236 - selectivity varies greatly among societies. It follows that variations in selectivity could make only a moderate difference for economic development. The main effects will be upon opportunities for schooling in itself. Democratization of those opportunities remains a worthy goal. That goal can be increasingly realized with economic development even as the degree of democratization of educational participation both reflects and shapes political and social aspects of development. Ambiguities about sequences in development cannot, however, be resolved in this study, which is focuseed on the delineation of patterns of selection rather than the longrun effects of inequalities in opportunity. Discussions about educational opportunity have not been bulwarked heretofore by systematic and comprehensive data for a large assortment of countries. At each forward step in studies of this topic, there are new laments about the poor quality of information. The empiricist meanwhile asks his critics to examine the quantitative evidence before concluding that it is either too crude to use or is factitious. The present work has been guided by a sociologist's familiarity with analysis of social stratification and it reflects the author's awareness of the glacial changes in systems of schools. At the same time, indicators of selectivity are interpreted and modified in the light of the spread of schooling among successive cohorts and of changes in occupational structures. The innumerable influences upon access to schooling are submerged in aggregated analyses. Clignet and Foster characterized French and British colonial policies for education in Africa as "egalitarian centralism" and - 237 - "voluntarism". Under the British there was a proliferation of vernacular schools with minimum expenditure and few controls on quality -- allegedly the more flexible system. French administrators insisted on use of French in classroom, which entailed employing more teachers from France; they used examinations to regulate the flow of pupils from the first grade onwards, and they did not encourage reliance on private funds or fees. By 1965 in Ghana two-thirds of the cohort attended primary school in contrast to only a third in the Ivory Coast. In the Ivory Coast a larger proportion of pupils were from rural areas and had illiterate parents, yet selectivity for paternal occupation was greater in the Ivory Coast. The gradient of selectivity (measured against base population) was steeper in Ghana for ethnic affiliation than for urban origin. While such comparisons are suggestive, they provide only limited perspective on the larger questions of commonalities and diversities among societies and their relationships, if any, to development. This work seeks insight through examination of a broad spectrum of societies, concentrating primarily on selectivity by father's education and occupation, but with illustrative cases of selectivity by family income. Many countries, from the newest nations to the old educational centers of Europe, are compared with the use of several kinds of indicators. In a few cases reliable longitudinal data were available, and those estimates were consistent with observations using cross-country comparisons. The simplest summary indicator of the degree of selectivity of a student population is the index of dissimilarity between the distribution of fathers of students and of a base population of men in the age cohort of - 238 - students' fathers. That index tells us the proportion of students who would have to be replaced by other youth for full parity of representation. In pursuit of the main question that gave rise to this study, the scattering of observation displayed in charts and tables was summarized by regression analyses in which indexes of dissimilarity were the dependent variable. Using a simple path model, enrollment rate was treated as an intervening varible between per capita income and the index of dissimilarity. These regressions were run on each of the four samples, with the following results (from Tables 2.8 and 3.8): Dissimilarities by Dissimilarities by Father's Education Fathsr's Occupation Secondary University Secondary University Mean Index of Dissimilarity 37.2% 41.7% 35.1% 44.5% RZ .237 .310 .382 .316 Effects of per capita income: Direct .215 -.286 -d068 -.264 Indirect -.447 -.137 -.431 -.149 Total -.232 -.423 .499 -.413 In considering these results it is important to remember that the sample for secondary students by father's education was biased strongly toward the less developed countries. This may account for the positive direct effect of income in that sample and the smaller total income effect than in the other samples. The negative indirect effects through enrollment rates are much greater for secondary than for university students, as is to be expected since everywhere rates of enrollment in secondary schools exceed those in universities. Income and enrollment rates explained just under a third of the variance in indexes of dissimilarity between origins of university students and base population, traits, whether measured by paternal education or 239 occupation. This leaves two thirds of the obs2rved differences in selectivity of university attendance to be explained in other ways: income level and enrollment rates are only part of the study. The coefficients of determination are lowest in the sample that is mos.t heavily weighted by LDCs. While the errors in the data bias the correlation estimates downward, such errors can hardly account for the positive direct effects of per capita income on selectivity in this case. Here we are observing effects of early-stage changes in occupational structures and hence in the distribution of education in the base population. The sample of secondary students for which data were available on paternal occupations has a more balanced country spread; here even the direct income effect ceases to be positive and more of the variance is explained by income, with a strongly negati-ve total effect on the index of dissimilarity. Anyone who has followed the analysis in this report will appreciate how important it is that one adjust and interpret indicators of selectivity to take into account the distribution of educational and occupational status in the cohorts of which the parents of students were members. A selectivity index cannot possibly exceed the ratio 1/Bi where Bi is the proportion of the base population with trait i; it will equal 1/Bi only if all the students' fathers are in category i. In addition, rates of enrollment constrain the possible value of a selectivity index to a maximum of l/Ej where Ei is the overall enrollment rate in educational level j. mne maximum for a given Bi and Ej will be determined by which of these denomenators is larger, making its reciprocal the smaller. The importance of base population constraints, though often ignored, is recognized by anyone who has given careful attention to the -240- ,estimation and interpretation of selectivity-indexes. Enrollment constraints are less often recognized. They are also different in that enrollment levels are more susceptible to controls by educational policy. Suppose that 2 percent of the base population is in the advantaged base population categoty A, say for father's schooling, and that 2 percent of a youth cohort enrolls in secondary school. If all those students came from background A, the selectivity index would be 100/2 or 50.0, while for all others (call them Z) the index would be zero. Raising enrollment rates to 4 percent would reduce the maximum possible selectivity index for Group A from 50.0 to 25.0 while raising the index for all others from zero to 50/98 or .51. With universal enrollment in a given level of schooling there would be of course no selectivity, and no selectivity index could then exceed (or fall short of) 1.00. One measure of the extent of over-representation of an advantaged group is simply the observed selectivity index minus 1.00, which I have compared with the maximum possible over-representation for a given base population and for given overall enrollment rates. Virtually no country displays selectivity indexes for the most privileged that are as high as would be poss-ible takAin into account proportions of men in the bas'e population who were relatively well educated or in white-collar occupations. Few approrimate the maximum consistent even with high over-all enrollment rates, but there are substantial differences among countries in the gap between observed selectivity and the maximum selectivity possible for any given enrollment rate. One way of avoiding the troublesome problems in use of selectivity indexes is to measure selectivity from particular origins in terms of the difference between logits of a base population proportion and of the - 241 - proportion of students' fathers in a given category. Taking those di.,Cerences as the dependent variables a path analysis was again used from per capita income to degree of selectivity (logit differences), with the logit of the overall enrollment rate as the interventing variable. First three categories of father's schooling were compared with base-population proportions in this way: 1) fathers having higher eduation, 2) fathers with at least some secondary schooling, and 3) fathers with no schooling. The R2 and the direct and indirect effects of per capita income were as follows (from Table 2.11): Dependent variables (Logit differences between proportions of fathers and of base populations) with higher with at least some with no education secondary education schooling R2 .428 .586 .360 Effects of per capita income: Direct .631 .084 .012 Indirect via logit of secondary enrollment rate -.609 -.565 .239 Total .022 .481 .251 The importance of enrollment rates as an intervening factor between per capita income and the logit-difference measures of selectivity is clear enough in all cases. The negative indiredt income effect via enrollment neutralizes positive direct effects of income for selectivity of children of men with higher education. They completely dominate the results for children of the larger category of men with at least some secondary education, - 242 making for a total income effect that sharply reduces over-representation from that category. In the third column a positive sign implies movement toward parity in representai:ion of children of men with no schooling: here again per capita income operates in an equalizing direction primarily through increased overall rates of enrollment in secondary schools. The proportion of variance explained in the second column, with an R2 of .586, is substantial; even here, however, two fifths of the variance in degree of selectivity remains unexplained. Among university students only one of the three path analyses gave significant results. This was the analysis for children of men with at least some secondary schooling (]12 .526, p = .0006). Tne results are the reverse of those for secondary students in that the effect of enrollment rates turned out to be perverse, but the direct effect for per capita income had a large negative value. The net total effect of per capita income was negative, with approximately the same equalizing impact as among secondary students. Selectivity by parental occupation raises more complex questions because there is not the almost automatic ranking of the major occupational categories that is normal for paternal schooling. Using the same procedure as for selectivity by father's education, the results on white-collar fathers were consistently a reduction of over-representation with rising per capita incomes, mediated primarily through enrollment rates for secondary students but with little effect of those rates for university students (Table 3.10). There were also consistent gains (less under-representation) with rising incomes in the relative representation of farmers' children, but even among secondary students the mediating effect of rising enrollment rates was much less in this case. -243 The most unexpected result was that for children of manual workers among both secondary and university students, but especially among the former. With rising per capita income and rising over-all enrollment rates the relative under-representation of children of manual workers among secondary students was actually aggravated, and substantially so. Given the use of logit differences, this is not a factitious mathematical result of changing proportions of manual workers in the base populations. It is unquestionably a reflection of these changes, nevertheless. In the least-developed countries the categorization "manual worker" commonly refers to wage workers in urban settings, and these are privileged men relative to the peasants and the uncounted family workers and hawkers. That relative status is not maintained in later stages of community development, even though "manual workers" become increasingly skilled. Quite another process lies back of the gains in representation of farmers' children with rising per capita national income. The varying situations for children of farmers and of manual workers are particularly interesting. Representation of these two non-elite groups in post-elementary schools differs markedly from one to another country and at different levels of per capita income. In the higher-income countries the selectivity index for children of farmers exceeds that for children of manual workers; this appears to be especially the case where agriculture employs only a small fraction of the working population but yields them average or higher incomes. By contrast, in the third world in which one speaks of "peasants" even unskilled manual workers may be part of what is called the "modern" sector, and their offspring may enjoy comparatively good opportunities for attending school beyond the compulsory level. - 244 Given selectivity indexes and figures for overall enrollment rates in a given level of schooling, it is possible to obtain estimates of background-specific enrollment rates and to compare them. The enrollment data are much more reliable for secondary than for higher education, and this analysis could therefore be carried somewhat further for selection into secondary schools. It is in comparison of background-specific enrollment rates that logit analysis becomes most important. Expressing rates in log-it terms and comparing the logits has two critical advantages. In regressions it avoids the prediction of impossible values below zero or above 1.00 (100%); this is crucial when the data cover above 80 or under 20 percent, where the curvatures on cumulative growth or diffusion paths change sharply and rapidly. More fundamentally, the logit transforms give greater weight to a given absolute percentage increase at the start, before the spread of a trait (in this case schooling) accelerated and at the other extreme as more resistant (or oppressed) groups are reached and enrollment approaches universality. The use of logits of background-specific enrollment rates avoids the main problems encountered in comparisons of selectivity indexes. Several sorts of analysis were made comparing differences in logits of background-specific enrollment rates. Those differences were analyzed across countries and in a few cases for particular countries over time. The latter included examination of progress toward democratization of an educational system, viewing each category of the population of further education as in itself a development process. Cross-section and longitudinal comparisons of background-specific enrollment rates carry us a step further in the study of particular cases, where the degree of progress toward less - 245 - inequality of representation may be assessed in terms of changing life chances in the particular societal context and in response to educational policies. Expansion of school enrollments is but one of the ways in which selectivity favoring children of advantaged families may be reduced and shares of less advantaged groups are raised toward parity. Almost any pedagogical practice can become a manipulable factor leading to greater or lesser democratization of a school system. Rates of transition between grades or between levels of school are-not in the nature of things but vary among systems. Admission and promotion policies can be set up on a meritocratic basis or quotas may be used. A particular style of examinations together with norms for passing are characteristics of each educational system; these traits affect the rate of repetition and in turn the amount of "wastage." Each practice affects the distribution of pupils among regions, ethnic groups, religions, or social-status sets of families. Sometimes just making sure that every pupil has textbooks can raise the proportion of children from less advantaged backgrounds who continue in school and who will be acknowledged by teachers to be reasonably competent. Possibilities for greater pedagogical efficiency ma,y sometimes be a major means to realize greater equity. Each new international team that is invited to recommend changes in a country's schools pinpoints aspects of inequity in opportunity. The IL0 team that reported on Colombia in 1980 pointed out that dropouts within any cycle of schools were very large while pupils completing primary school had a good chance (9 in 10) of continuing into secondary school and almost as good - 246 - a chance (3 in 4) of then going on from completion of secondary school to higher education. Whether attrition is mainly within or between levels of school differs among countries and no doubt among subpopulations in any country. Over much of the world this situation has been compounded by the fact that often village schools offer only two or three grades whLle urban schools offer six or seven grades. In the end, zealous officials of any country must turn to diraggeegated data to identify what kinds of detailed information will be of diagnostic use in modifying educational policies. Whatever the policy, it may have different eventual outcomes than initially was anticipated, and these outcomes can vary greatly among societies that we often judge to be similar. Direct reversal of the association between schooling and social status in a revolutionary context may establish a new bureaucratic privileged class in succeeding cohorts. Different indicators of selectivity can have quite dissimilar implications for a judgment about equity. Different conclusions can be drawn, for example, if one focuses on absolute differences among background-specific enrollment rates, on ratios of those rates, or on differences in their logits. Overall enrollment rates are more important as a constraint on the degree of selectivity in secondary than in higher schools; few countries envisage truly mass higher education. Selectivity indexes for-e small elite can be large, yet the index for offspring of modest-status families may nevertheless be close to parity. These shifting patterns could be expressed in another way: what is the proportion of today's students (in higher or even in secondary schools) who are the first representatives of their families to attend? That - 247 - proportion is astonishingly diverse even among western societies, and also among different types of secondary school or courses of study in university. The speed of upgrading of the general population in either schooling or occupation is also speed of change between cohorts of youth and the parental cohorts. Those individuals who are the first of their family to obtain post-compulsory schooling will have a different fund of embodied "human capital" than those who come from a long line of university graduates. The latter will embody a greater stock of human capital and much of it will have been acquired informally.- There is a definite ambivalence associated with our eagerness to give greater opportunity for schooling. From one point of view the "democrat" will laud the rising selectivity indices for lower-status children--or the declining indexes of dissimilarity between pupils' parents and the base population. From another point of view it is just the pride in schooling, rising aspirations for schooling, and canny appreciation of the vocational and cultural utility of schooling that one expects to be nourished in growing proportions of families by the democratization of schooling. But these much-desired qualities of family life are exactly the privileges -- when possessed by "other" families -- that we measure in high selectivity index for children of well-schooled parents. It takes some appreciation of the relationships between shifting trajectories -- past and prospective profiles of schooling -- to grasp the dynamics of democratization. National leaders in many societies must fear that citizens' thirst for schocling is unquenchable and frighteningly responsive to opportunities. Leaders are torn between demands for equity and awareness of the financial - 248 - costs of.extending educational access. At the same time, it may be important to capitalize on readiness for schooling without having to instill. motivations for schooling in backward areas. Such dilemmas are most pressing in poor countries, but everywhere selectivity indexes for the lower-status section of the population can be raised substantially only as enrollment expands -- at perhaps prohibitive costs -- or by curtailing attendance of individuals from top-status categories. If one seeks policies to achieve wider representation in secondary and higher education, the focus must be mainly on what happens to opportunity among populations normally having enrollment below average rates: children of the relatively uneducated, of peasants, of manual workers. It would be easy to reduce selectivity indexes on the small, most over-represented segments of a population by making rather small absolute decreases in their numbers enrolled -- easy, that is, if they were not articulate people. But efforts to move in that direction almost everywhere meet with strong opposition from the most alert, articulate, and educogenic families. Few people compare distributions of students' fathers and base populations; they ask, rather, how particular policies may affect the chances for schooling for themselves and their children. These are related but quite different perspectives. Issues concerning equity and meritocratic selection intersect in complex ways. A meritocratic emphasis can be justified in its owni right, even as it is also used to relationalize a policy that will in fact ensure continuance of relative advantage in eductional access to children from .privileged homes. Only with a quota system such as that introduced between - 249 - Malays and Chinese in Malaysia can egalitarian and meritocratic emphases be reconciled, and inevitably this entails, for the short run at least, a downgrading of the meritocratic criteria. Even where no such quota system is introduced, a meritocratic system will extend some of the most cherished educational opportunities to bright and motivated young people from disadvantaged homes, however, just as it may also appear as a threat to a few of the socially privileged. These and related questions are important ones, and differences among countries in the way in which they are answered lie back of some of the variation among advanced countries observed in this report. For example, touched upon only lightly (in Chapter 7 and 8) is selection among institutions or courses of