6
Fertility, Education, and Resources in South Africa

Duncan Thomas

Introduction

South Africa has emerged from three-quarters of a century of apartheid, which placed comprehensive restrictions on the geographic and economic mobility of its population. As those restrictions are relaxed and the society moves toward integration, one of the greatest challenges facing the people of South Africa will be the absorption of a young and rapidly growing labor force into an economy undergoing dramatic restructuring. While South Africa appears to be ahead of other African countries in its demographic transition, there is tremendous diversity within the country. It seems safe to conjecture, therefore, that population policy will play a key role in influencing the success of social and economic development in the country (Chimere-Dan, 1993; African National Congress, 1994, 1995).

Using recently collected household survey data, this chapter examines an important consideration in the design of that policy: the relationship between fertility and resources, with a focus on the role played by maternal education. A vast number of studies have demonstrated that education and fertility tend to be negatively correlated in a wide array of contexts. In South Africa, education is a key correlate of fertility among all black women, but the association is less clear among women of other races, except, perhaps, those who are better educated. Over and above simply documenting these findings, this chapter examines three potential mechanisms in an attempt to provide insight into what underlies the observed association between education and fertility among South African blacks.

First, education is not randomly assigned within a population; rather, people



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 138
--> 6 Fertility, Education, and Resources in South Africa Duncan Thomas Introduction South Africa has emerged from three-quarters of a century of apartheid, which placed comprehensive restrictions on the geographic and economic mobility of its population. As those restrictions are relaxed and the society moves toward integration, one of the greatest challenges facing the people of South Africa will be the absorption of a young and rapidly growing labor force into an economy undergoing dramatic restructuring. While South Africa appears to be ahead of other African countries in its demographic transition, there is tremendous diversity within the country. It seems safe to conjecture, therefore, that population policy will play a key role in influencing the success of social and economic development in the country (Chimere-Dan, 1993; African National Congress, 1994, 1995). Using recently collected household survey data, this chapter examines an important consideration in the design of that policy: the relationship between fertility and resources, with a focus on the role played by maternal education. A vast number of studies have demonstrated that education and fertility tend to be negatively correlated in a wide array of contexts. In South Africa, education is a key correlate of fertility among all black women, but the association is less clear among women of other races, except, perhaps, those who are better educated. Over and above simply documenting these findings, this chapter examines three potential mechanisms in an attempt to provide insight into what underlies the observed association between education and fertility among South African blacks. First, education is not randomly assigned within a population; rather, people

OCR for page 138
--> choose how much they invest in schooling given the constraints and opportunities they face. In the extreme, it might be argued that schooling provides no value-added, but is simply a sorting mechanism: those who are more able spend more time at school in order to signal their ability to their future employers (or spouses). Placing the spotlight on educational attainment around natural exit points in the South African education system, the evidence suggests that self-selection in education does play some role in explaining the effect of education on fertility in a regression context. Thus, a naive interpretation of the effect as being entirely causal would be misleading. Second, education and household resources tend to be correlated, so a woman's education may simply be a proxy for her, or her family's, income. To the extent that it can be tested with the available data, this interpretation does not appear to tell the entire story in South Africa. While household resources do affect fertility outcomes, even after controlling for spousal characteristics, household income, labor market choices, and community characteristics, female education continues to have a powerful negative association with fertility. Thus one can conclude that a substantial part of the effect of female education on fertility operates independently of resources. Spousal education is also negatively correlated with fertility (see Basu, this volume), and the impacts of male and female education are, in most cases, close in magnitude. In South Africa, men appear to have an important influence on demographic outcomes. Third, one can attempt to isolate the relationship between skills likely to be learned in school and demographic outcomes. Drawing on a sub-sample of women who completed a short comprehension and quantitative test, one finds that, after controlling for income and education, performance on these tests has an independent impact on fertility. The impact of comprehension skills is particularly large in magnitude, suggesting that the acquisition and assimilation of information may be important in affecting family decision making. The next section sets the stage for the main analysis of the chapter by placing fertility and educational attainment in South Africa in historical context. This is followed by three sections that use recently collected household survey data to examine in turn the three mechanisms discussed above. The final section presents conclusions. Fertility And Education In Historical Perspective Given the central role played by population policy in the apartheid system of South Africa, it is remarkable how little solid evidence exists on the demography of the country. Even today, few of the data sources predating 1993 that contain information on fertility have been placed in the public domain. As a result, one must rely on reports that have not been subjected to the sort of scientific scrutiny they warrant (see Caldwell and Caldwell, 1993, for an excellent discussion). Fortunately, as the government has embraced the concept of openness and access

OCR for page 138
--> to information, this situation has been remedied. Data collected by the Central Statistical Service and, in many cases, by units that are independent of the government are now routinely placed in the public domain; the Project for Statistics on Living Standards and Development (PSLSD) and the annual October Household Surveys are good examples. The 1994 October Household Survey estimates South Africa's population at about 40 million, of whom over three-quarters are black; half the rest are white, nearly 9 percent are mixed-race coloreds, and fewer than 3 percent are Asians, mostly of Indian descent. That survey estimates the total fertility rate (TFR) to be 4.1, although the Central Statistical Service views this as a substantial underestimate, particularly among blacks. Much of the historical data on fertility and family planning in South Africa is described by van Zyl (1994), who also documents the methodologies used for the main surveys conducted by the Human Sciences Research Council that form the basis of these estimates. There are good reasons to be skeptical about the quality of some of those surveys, not least of which is the fact that population policy was an important political issue in South Africa, and it is far from clear that, at least at that time, the Human Sciences Research Council played a role that was entirely divorced from the political system. It is also prudent to treat comparisons across time with considerable caution since the surveys are not always comparable, and several focused on specific subpopulations. For example, in many surveys, the so-called homelands (where many of the poorest South Africans lived) were excluded from the samples, and in some surveys, fertility questions were asked only of married women. Setting aside the fact that the definition of marriage is complex in this society, many women who would not declare themselves as married have borne children, and teenage pregnancy rates are very high. According to the 1994 October Household Survey, about 33 of every 100 women have given birth out of wedlock. (See, for example, Preston-Whyte, 1990, for an insightful discussion.) While remaining mindful of these caveats about data quality, it is useful to attempt to place South Africa's fertility rate in historical context. TFR estimates for 1950 through 1990 for each of the four main racial groups in South Africa are presented in the upper panel of Figure 6-1.1 According to these estimates, in 1950, the TFRs of blacks, coloreds, and Indians were all high (above 6) and reasonably close to one another, whereas that of whites was much lower (slightly above 3). By 1990, the picture was dramatically different: TFRs among coloreds and Indians had been reduced by more than half and were very close to that of whites. The TFR of blacks, however, remained much higher, around 5.0. While fertility among blacks had declined sharply during the 1970s and early 1980s, it had remained virtually constant during the earlier two decades. It appears that 1   The data are drawn from Lucas (1992) and van Zyl (1994).

OCR for page 138
--> FIGURE 6-1 Total fertility rates and education of females, South Africa, 1950-1990. among blacks, the demographic transition began to accelerate only around a quarter-century ago. In sharp contrast, there were rapid declines in fertility among Indians dating back to the 1950s, and their fertility stabilized around 1980. Coloreds fall between these two groups: their fertility actually rose in the 1950s, so that in the early 1960s, the TFRs of blacks and coloreds were almost identical (around 7), but the TFR of coloreds declined during the following quarter-century to a level that is now very close to that of Indians. The fertility of whites changed very little during the period, having risen slightly during the 1950s and declined slowly since then to below replacement levels.

OCR for page 138
--> If these estimates reflect reality, several questions arise. First, what underlies the differences in the timing of fertility decline among Indians, coloreds, and blacks'? Second, why has the fertility of coloreds and Indians fallen so dramatically, while that of blacks has remained at levels comparable with those of neighboring Botswana and Zimbabwe, whose populations are substantially poorer and less well educated? A plausible hypothesis centers around differences in community resources, specifically access to family planning services. There is, however, little direct evidence on the relationship between the fertility of individual women in South Africa and their access to services, and the evidence in both Botswana and Zimbabwe suggests that access to services alone cannot explain fertility decline there. Nevertheless, casual inspection of the evidence in South Africa does suggest that service availability may have played a role. Prior to the mid-1960s, government had little involvement in family planning, and it seems likely that only whites, and possibly Indians, would have had good access to services. Public investments in family planning began in the mid-1960s and were targeted largely toward urban dwellers: this coincides with the decline in fertility among colored women, the majority of whom live in urban areas. In the mid-1970s, the National Family Planning Programme was established, and access to services was massively expanded, with special attempts being made to serve rural women. In 1972, for example, there were fewer than 250 family planning clinics in the country, but by the early 1980s, that number had risen to more than 36,000 (Lucas, 1992). It was also during this period that the decline in black fertility took hold. Yet even today, access to family planning services is probably not universal in South Africa. Evidence for this conclusion is provided by data from the PSLSD indicating that there is a family planning clinic in the community for over three-quarters of Indian women aged 15 to 49, about half of colored women, but only one-third of black women.2 2   These discrepancies are likely to be underestimates of differences in service availability for at least two reasons. First, many women in South Africa obtain family planning services from places other than family planning clinics, and blacks are the least likely to use alternative sources. Second. the estimates are based on data collected at the community level, and it is well known that those data are difficult to collect in a survey setting and are therefore prone to serious measurement error. (See Frankenberg and Sudharto, 1995, for a general discussion, and Thomas and Maluccio, 1996, for a specific description of the nature and extent of problems that are evident in a community survey conducted in Zimbabwe in 1990.) Two issues are of special importance. First, it is not entirely clear how to define ''community," particularly in urban and metropolitan areas, and the definition may not mean the same thing to all respondents. Second, the so-called "community informants" who are "prominent members of the community such as school principals, priests and chiefs" (Project for Statistics on Living Standards and Development, 1994:v) may themselves not be very well informed. For example, the PSLSD reports a family planning clinic in two-thirds of urban communities, one-third of rural communities, but only one-quarter of metropolitan communities. Yet it is the metropolitan communities that are likely to be best served of all.

OCR for page 138
--> Racial differences in fertility rates may also be explained by differences in household resources. The benefits of growth and development in South Africa have not been distributed evenly across races, and the society is marked by very high levels of inequality. This inequality is reflected, for example, in differences in education as depicted in the lower panel of Figure 6-1, which displays mean educational attainment for women, by birth cohort, as reported in the 1993 PSLSD.3 Among women born in 1935, whites are much better educated than coloreds, Indians, and blacks: the average white woman had completed 9 years of schooling, whereas the average black had completed less than 2. The educational attainment of all groups except whites—especially Indians and, to a lesser extent, blacks—has increased substantially since then. Thus among women born in 1975, Indians are at least as well educated as whites, and blacks are nearly as well educated as coloreds. While the convergence of average educational attainment across races is striking, it is well known that there are substantial differences in the quality of schooling between racial groups, so it is unlikely that a year of education means the same thing to a black and a white. A comparison of the upper and lower panels of Figure 6-1 does suggest that educational attainment and fertility are negatively correlated. However, it remains difficult to explain the timing of declines in TFRs for each race using an argument based on the average education of women. Digging a little more deeply and examining changes in other parts of the education distribution, a more complex picture emerges. Thomas (1996) exploits the large sample sizes of the 1991 South African Census to explore rates of growth in educational attainment at each quartile of the education distribution. He reports that among those born between 1920 and 1970, an Indian woman would have spent 1.5 more years in school if she had been born a decade later, and the growth was most rapid among the least educated; for example, growth rates were 1.8 years at the bottom quartile and 1.2 years at the top quartile of the distribution. A similar pattern emerges for coloreds, although their average growth rate was only 0.7 years per decade. In sharp contrast, however, the least-educated blacks were largely excluded from the increase in education, which was, instead, concentrated among the better educated 3   Completed years of education is defined as I if the woman completed sub-A, sub-B, or Standard 1, with another year being added for each grade completed until Standard 10, when most students take the matriculation (matric) examination. In the survey, post-secondary schooling is not reported in years, but in terms of qualifications. Diplomas, technical degrees, teacher training, nursing training, and completion of some university education are assigned 12 years of schooling. Completion of a university degree is assigned 14 years of schooling. Since only 5 percent of women report attending school beyond Standard 10, the impact of making different reasonable assumptions about years of schooling for these women is likely to be small. Moreover, the specifications for the regressions discussed in detail below are sufficiently flexible to ensure that coefficient estimates are not biased by this assumption.

OCR for page 138
--> of that group. There is a positive note in this dismal picture for the poorest black South Africans since, among those born in the latter part of the period, increases in education have occurred across all black women. This more recent extension of the fruits of development to the poorest may well be related to the tardiness of the fertility decline among blacks. While they are suggestive, these aggregate data can give only a very incomplete picture of the relationship between fertility and education at the micro level. Therefore, in the following sections, survey data from the PSLSD are used to examine not only the correlation between fertility and education, but also some of the mechanisms that underlie that correlation. Before doing that, however, it is useful to take a slight detour and examine the determinants of educational attainment in South Africa. Ideally, one would like to be able to say something about how much of the increase in educational attainment in South Africa over the last half-century can be attributed to public policies such as investments in school infrastructure, teachers, and teacher training. It is not possible to answer that question directly with the available data, but it is possible to turn the question around and assess the extent to which resources in the home and in the community have affected the educational outcomes of adults and children. We begin with completed schooling of adults. Respondents in the South African Social Stratification Survey (1991-1994)4 report their own education; a migration history, including information on the type of house and community they lived in at birth; and characteristics, including education and occupation, of the people they viewed as their father- and mother-figure at age 14. Regressions in the upper panel of Table 6-1 report the impact of maternal and paternal years of schooling on the completed years of education of adult children aged 20 to 70 at the time of the survey. Parental education is a powerful predictor of own education: fully 26 percent of the variation in educational attainment among blacks is explained by parental education alone. Intergenerational mobility has risen over time, as indicated by the larger coefficients on education for older than younger blacks; this change is particularly striking for the impact of maternal education. While mobility remains lower among blacks relative to whites, even for whites an additional year of parental schooling is associated with one-fifth of a year more schooling for the respondent. How much of the impact of parental education can be attributed to the role of household and community resources? It is difficult to imagine any household survey measuring resources in the home 50, 40, or even 20 years prior to the interview with any degree of accuracy. Fortunately, the South African Social 4   The survey was conducted by Don Treiman in collaboration with the Human Sciences Research Council. One (randomly chosen) adult respondent was interviewed from each of 9,000 households in South Africa. See Treiman (1996).

OCR for page 138
--> Stratification Survey does provide proxies for resources in the home and community when the respondent was a child. Drawing on this information, the regressions were reestimated to include controls for the occupation of the parents when the respondent was age 14, location of birth, type of residence at birth, and language spoken at home. In addition, to capture variation in access to schools and possibly quality of schools, as well as other infrastructure at the community level, the reestimated regressions include fixed effects for town of birth. Results are reported in Panel B of Table 6-1. Among blacks, over half the variation in educational attainment is explained by the controls, and the effects of parental education, particularly that of fathers, are significantly reduced. This suggests that household and community resources do matter and that paternal education is, in part, proxying for those resources. Further support for this interpretation is provided by the fact that among the covariates, paternal occupation is key. If a respondent's father had a job working in the wage sector or for the government or if the father was self-employed, the respondent is likely to have completed half a year of additional schooling, but a third of a year less if the father was a farm laborer; the excluded category includes informal-sector workers, communal farmers, and those not working at all.5 The additional controls do little to diminish the impact of maternal education, especially among older blacks, suggesting that a woman's education has an independent effect on the human capital of her children over and above the role of resources. Among whites, the additional controls do reduce the impact of parental education, but the differences are relatively small and not significant. This may be because variation in household and community resources is less important among whites or because the measures of these resources are too crude to be informative. To provide some evidence on that question, one can examine the effects of parental education and resources on the educational attainment of children still resident at home using data from the PSLSD.6 An advantage of these data is that they contain information on household per capita expenditures, a longer-run measure of resources available to the family; the disadvantage is that one is forced to restrict attention to only those children who are still resident in the 5   The regressions include controls for whether the respondent was born in an urban, peri-urban, or squatter area, along with town-of-birth fixed effects. Paternal occupation is not, therefore, simply proxying for rural-urban differences. The negative effect of occupation if the father was a farm laborer is greater (and significant) among the younger cohort, suggesting that it may partly capture the effect of the Verwoerdian policy of "putting a school on every farm." The quality of those schools has been roundly criticized (see, for example, Department of Education, 1996). 6   The PSLSD is a nationally representative survey of some 9,000 households conducted in South Africa during October through December 1993, a few months before the first democratic elections in April 1994.

OCR for page 138
--> TABLE 6-1 Intergenerational Transmission of Education ADULTS BLACKS (by age in years) WHITES (by age in years) ALL 20-39 41-70 ALL 20-39 41-70 A. Education only Maternal education 0.296* 0.258* 0.419* 0.201* 0.224* 0.185* [0.02] [0.02] [0.04] [0.02] [0.03] [0.04] Paternal education 0.264* 0.235* 0.337* 0.178* 0.164* 0.194* [0.02] [0.02] [0.05] [0.02] [0.03] [0.03] B. Full set of controls Maternal education 0.258* 0.211* 0.397* 0.176* 0.156* 0.194* [0.02] [0.03] [0.05] [0.03] [0.04] [0.05] Paternal education 0.178* 0.155* 0.215* 0.177* 0.160* 0.187* [0.03] [0.03] [0.08] [0.03] [0.04] [0.04] R2 0.52 0.46 0.57 0.42 0.43 0.50 Sample size 4,533 2,964 1,569 2,161 1,133 1,028

OCR for page 138
--> C. CHILDREN BLACKS (by age in years) WHITES   ALL 10-11 12-13 14-15 16-17 ALL Maternal education 0.073* 0.038* 0.076* 0.102* 0.115* 0.003   [0.01] [0.01] [0.01] [0.02] [0.02] [0.01] Paternal education 0.056* 0.021+ 0.046* 0.063* 0.073* 0.015   [0.01] [0.01] [0.02] [0.02] [0.02] [0.01] In(per capita expenditure) 0.373* 0.200* 0.289* 0.415* 0.529* 0.066   [0.05] [0.05] [0.06] [0.09] [0.10] [0.12] R2 0.56 0.26 0.32 0.35 0.34 0.85 Sample size 6,609 1,706 1,729 1,658 1,516 577 NOTE: Dependent variable is years of completed education. Adult regressions based on respondents aged 20-70 in South African Social Stratification Survey, 1991/94. Adult "education only" regression includes controls for gender and (linear spline in) age of respondent, whether education of parent is missing, and whether father-figure and mother-figure are biological parents. "Full set of controls" adds controls for occupation of father and mother (at respondent's age 14), language spoken in home at respondent's age 14, type of place lived in at birth, type of house lived in at birth, and fixed effect for town of birth. Child regressions based on children aged 10-17 in PSLSD, 1994. Regressions include dummies for gender of child, each year of age of child, and whether urban or rural dweller.

OCR for page 138
--> home. The focus here is on children aged 10 through 17, but recognizing that a substantial fraction of older children are likely to have left the nest, the analyses are stratified into 2-year age bands. Panel C of Table 6-1 presents the regression results. Educational attainment of black children is significantly correlated with parental education and household per capita expenditures, and all the effects increase with the age of the child. Among whites, there is no evidence that parental education or household resources are correlated with child educational attainment. One can conclude, therefore, that while there has been substantial intergenerational transmission in educational attainment among whites, resource constraints and limited access to schools are unlikely to have had a large impact on this group's educational attainment. Among blacks, however, resources do seem to matter, indicating that there is substantial scope for public interventions designed to raise schooling levels. Corroborating evidence is reported by Case and Deaton (1995), who show that even crude measures of teacher-pupil ratios are correlated with educational attainment among blacks, but not whites. The interpretation has considerable intuitive appeal in view of the history of South Africa, where racially segregated education was a central pillar of the apartheid system, state policies sought to restrict blacks' access to schooling, and a very well-funded system for whites coexisted with a substantially poorer system for blacks. In 1975, for example, the year before the Soweto school riots, public expenditures on education of the average white school-age child were more than 15 times greater than expenditures on the average black child. Relationship Between Fertility And Education Having shown that parental education and, for blacks, family and community resources play an important role in determining the educational attainment of the next generation, the discussion now turns to the correlation between education and fertility. The following analyses are based on the PSLSD, in which female respondents aged 15 to 49 reported the number of children ever born (CEB), how many survived to age I and how many to age 5, and how many were alive at the time of the survey. Unfortunately, fertility in the last year (or last 5 years) was not recorded. We must therefore focus on CEB as our sole measure of fertility.7 Summary statistics for the 10,500 female respondents are reported in Table 6-2. Three-quarters of the women in the survey were black and, as indicated in Figure 6-1, they had borne more children than coloreds, Indians, or whites. But 7   It is also unfortunate that the limited set of information reduces the scope for undertaking internal consistency checks in these data—an issue of considerable import given the concerns raised about data quality above.

OCR for page 138
--> Role Of Other Measures Of Human Capital As noted above, there were vast differences in public investments in education across racial groups under apartheid in South Africa, and according to data in the PSLSD, these gaps were reinforced by private expenditures on schooling. There are therefore dramatic differences in school quality across races, regions, and socioeconomic groups. Eight years of schooling probably does not mean the same thing to every South African woman. Even the standard examinations, such as matric, had their own race- and region-specific boards, and standards were not uniform across those boards. The Indian board, for example, has a reputation of setting the highest standard. In an effort to capture some of these differences in the meaning of schooling, a Literacy Assessment Module (LAM) was fielded in conjunction with the PSLSD. The LAM sought to measure three sets of basic skills: reading comprehension (6 questions), listening comprehension and practical mathematics (2 questions), and computational skill (6 questions). The specific questions appear in Appendix A. Fuller et al. (1995) provide an extensive description of the data. The LAM was intended to be given to a random subsample of 25 percent of all households. Within each household, the enumerator was instructed to choose randomly two household members for the test; one was to be aged 13 to 18 and the other aged 18 to 50. Unfortunately, the sampling scheme that was implemented in the field was not random. The LAM was completed in only 15 percent of households, and because the sample sizes are small, the present analysis focuses on black women. The characteristics of the LAM subsample are reported in the second column of Table 6-2. A regression of the probability that a woman is in the LAM subsample provides a simple way of summarizing the differences between the subsample and the women in the full sample. That regression indicates that enumerators tended to select younger, never-married, urban women. In large part, this presumably reflects the fact that these women were more likely to be at home and that younger (and therefore better-educated) respondents would complete the module more quickly. However, conditional on age, there is no evidence of differences between women in the LAM subsample and the full sample in terms of their education and household income. The effect of the selection is reflected in the first column of Table 6-6, which reports the results of regressing fertility on education, controlling for age and location, based on data from the 778 black women who completed the LAM. The comparable regression, based on the entire sample of black women, is reported in the first column of Table 6-5A. The correlation between education and fertility is stronger among women in the LAM subsample and, as is apparent from column 2 of Table 6-6, so are income effects. Within this subsample, the inclusion of raw test scores has a substantial

OCR for page 138
--> TABLE 6-6 Children Ever Born and Cognitive Test Scores, Black Women Aged 15-49   Baseline HH Income Test: Total Test: Separate Income & Tests Access Information All Years of education Woman -0.147* -0.102* -0.089* -0.087* -0.082* -0.096* -0.079* [0.016] [0.016] [0.018] [0.018] [0.018] [0.017] [0.018] Spouse   -0.057* -0.066* -0.066* -0.056* -0.057* -0.056*   [0.025] [0.025] [0.025] [0.025] [0.025] [0.025] In(HH income) — -0.147* —   -0.133* -0.142* -0.130*   [0.045]     [0.045] [0.045] [0.045] Test scores Total — — -0.052* —   — —     [0.017]         Comprehension — — — -0.076* -0.068* — -0.067*       [0.033] [0.033] — [0.033] Quantitative — —   -0.033 -0.030 — -0.028       [0.028] [0.028]   [0.028] (1) if read paper in last 2 weeks — — — — — -0.099 -0.061           [0.096] [0.097] F(all human capital) — — 28.27 19.09 15.87 20.12 11.99     [0.00] [0.00] [0.00] [0.00] [0.00] F(test scores)       5.04 4.06   3.72       [0.01] [0.02]   [0.02] F(all covs.) 279.14 179.93 179.49 164.54 154.09 165.04 143.00 R2 0.69 0.72 0.72 0.72 0.72 0.72 0.72 NOTES: In each regression, 778 women were included. Sample characteristics are reported in column 2 of Table 6-2. Standard errors in parentheses below coefficients; p-values below F test statistics. Regressions include controls for age (spline): spouse present. absent, or dead; and location of residence.

OCR for page 138
--> impact on the effect of education: the effect is reduced by more than one-third.18 Moreover, the tests themselves are negatively correlated with fertility. Specifically, it is the reading comprehension test that is of particular relevance. A woman with better-developed comprehension skills (who answered all six questions correctly) would, on average, have nearly half a child less than a woman who failed to answer any of the questions correctly. Since women who score well on such tests are likely to earn more and live in higher-income households, the impact of the test scores may simply reflect the role of household resources. Controlling for income, however, has very little impact on the estimates. More directly, in a regression of (log) wages on age, education, and test scores among black women, it is computational skills that appear to be rewarded in the labor market: an additional correct answer to the computational questions is associated with a 15 percent higher wage (t statistic is 2.5), whereas a correct comprehension answer is associated with only a 7 percent higher wage (t statistic is 1.3). Thus if test scores are simply a proxy for income, computational skills should have a larger impact on fertility than do comprehension skills. In fact, the reverse is true. Why do comprehension skills affect demographic outcomes, but not wages? Women with better comprehension skills may be better able to access and assimilate information in the community. They may thus be likely to be better informed than their peers and therefore better able to use community services effectively. In one attempt to examine this hypothesis, test scores were replaced with a control for whether the woman reported that she had read a newspaper in the previous 2 weeks. Slightly fewer than half of the women in the sample had done so. Clearly, reading a newspaper is a choice, so the interpretation of the regression is not ambiguous; however, the covariate is not significantly correlated with fertility, and its inclusion in a regression with the test scores does not affect the estimated impact of comprehension skills. An alternative strategy for examining the information hypothesis is to include community fixed effects that capture all community-level information and services available to women. If women who score well on the tests live in communities with services that affect fertility, these effects will be absorbed by the fixed effects. While the impact of the total test score is robust to their inclusion (its impact is -0.047 with a standard error of 0.022), the impact of the comprehension score is halved, and it is not significant. (In contrast, the impact of the quantitative score rises.) This result suggests that access to information may play an important role in explaining the link between fertility and test scores, and therefore education. 18   Put another way, test scores and education are highly correlated in these data. For example, among black women, controlling for age and location, an additional year of schooling is associated with a 0.21 higher score on the quantitative tests and a 0.22 higher score on the comprehension test. Standard errors are 0.02 in both cases.

OCR for page 138
--> Conclusions Female education is a powerful predictor of fertility outcomes among women in South Africa, particularly those who are black or colored. Drawing primarily on survey data from the PSLSD, this chapter has investigated some of the mechanisms that underlie this relationship. The discussion has focused on black women, among whom an additional year of schooling is associated with about 0.12 fewer children ever born. Part of this effect reflects assortative mating in the marriage market (see Basu, this volume). The education of husbands and wives is positively correlated, and controlling for the education of the spouse, the effect of maternal education on children ever born is reduced by about a third. This reduction is greatest among the least-educated women. The results indicate that with regard to fertility choices, men matter in South Africa, particularly among the poorest groups, suggesting that family planning outreach should be directed toward both men and women. It is difficult with the available data to determine the precise magnitude of the effect of household resources on fertility with much confidence. There can be no doubt, however, that income significantly depresses childbearing. Higher rates of female labor force participation and higher wages are also associated with lower fertility. Thus, policies designed to increase growth in income, particularly female income, are likely to lead to slower population growth. There is suggestive evidence from aggregate data on the timing of fertility declines in South Africa that greater access to family planning services may have been associated with lower fertility. Evidence from the PSLSD demonstrates that community characteristics do indeed affect fertility outcomes, although the impact apparently varies within communities. It is, however, not possible, using the PSLSD, to identify the specific factors that are important or who benefits most. This is unfortunate since investments in community infrastructure and services can play a central role in public policy interventions, particularly in the social sector, and in policies that are targeted toward particular subpopulations, such as the poorest. While household and community resources are important determinants of fertility outcomes, the key finding for this analysis is that they can explain at most half of the correlation between maternal education and fertility, and most of the difference can be attributed to the role of husband's education. Parental education, particularly schooling of the mother, is also a powerful predictor of the educational outcomes of children, after controlling for family and community resources. Thus as maternal education increases, there is both a decline in the quantity of children desired and an increase in their quality. One can conclude, therefore, that maternal education has an important independent association with fertility and investments in the human capital of the next generation. Part of the influence of maternal education is related to skills that are likely

OCR for page 138
--> to be learned in school; in particular, comprehension skills appear to play a special role in affecting demographic outcomes (see also Diamond et al., this volume). Thus raising cognitive functioning by improving the quality of schooling of today's students and also, perhaps, by providing adult education programs may yield returns beyond the direct economic benefits through the additional impact on demographic and social outcomes. However, part of the association between education and fertility appears to have nothing to do with the education production function per se, but is a reflection of the fact that education is not randomly distributed across the population. Specifically, women and their families choose their educational attainment given the opportunities and constraints they face, so the better educated are a self-selected sample of the underlying population. This is an important insight as it suggests that a naive interpretation of the correlation between education and fertility as entirely causal is likely to be misleading and to lead to an overstatement of the potential impact of increased levels of female schooling on population growth. In sum, this chapter has identified several mechanisms through which the correlation between fertility and maternal education operates. A small part of the correlation can be attributed to the role of family and community resources, a larger part to the role of husband's schooling, and part to the acquisition of cognitive skills. An indeterminate fraction of the correlation is associated with unobserved heterogeneity, reflecting the fact that educational attainment is ultimately a choice made by individuals. Acknowledgments I am grateful to SALDRU and The World Bank for providing access to the survey data used here. The comments of Caroline Bledsoe, John Bongaarts, John Casterline, Barney Cohen, Elizabeth Frankenberg, Simon Mpele, Ingrid Woollard, and three anonymous referees have been very helpful. References African National Congress 1994 The Reconstruction and Development Programme. Johannesburg, South Africa: Umanyano Publications. 1995 Population Policy for South Africa? Green Paper. Pretoria, South Africa: Ministry of Welfare and Population Development. Ashenfelter, O., and A. Krueger 1994 Estimates of the economic return to schooling from a new sample of twins. American Economic Review 84(5): 1157-1173. Caldwell, J. 1977 The economic rationality of high fertility: An investigation illustrated with Nigerian survey data. Population Studies 31(1):5-27.

OCR for page 138
--> 1979 Education as a factor in mortality decline: An examination of Nigerian data. Population Studies 33(3):395-413. Caldwell, J., and P. Caldwell 1993 The South African fertility decline. Population and Development Review 19(2):225-261. Case, A., and A. Deaton 1995 School quality and educational outcomes in South Africa. Princeton, N.J.: Princeton University. (mimeographed) Chimere-Dan, O. 1993 Population policy in South Africa. Studies in Family Planning 24(1):31-39. Cochrane, S. 1983 Effects of education and urbanization on fertility. In R. Bulatao and R.D. Lee, eds., Determinants of Fertility in Developing Countries. New York: Academic Press. Department of Education 1996 The Organisation, Governance and Funding of Schools. Education White Paper 2. Pretoria: Department of Education. Frankenberg, E., and P. Sudharto 1995 Community-Facility Data Collection in the Indonesian Family Life Survey. Santa Monica, Calif.: RAND. (mimeographed) Fuller, B., P. Pillay, and N. Sirur 1995 Literacy Trends in South Africa: Expanding Education While Reinforcing Unequal Achievement? Cambridge, Mass.: Harvard University. (mimeographed) Hyslop, D. 1996 Estimation of a Dynamic Model of Female Labor Force Participation. Industrial Relations Working Paper. Princeton, N.J.: Princeton University. Lam, D., and S. Duryea 1995 Effects of Schooling on Fertility, Labor Supply and Investments in Children, with Evidence from Brazil. Ann Arbor, Mich.: University of Michigan Population Studies Center. (mimeographed) Lucas, D. 1992 Fertility and family planning in Southern and Central Africa. Studies in Family Planning 23(3): 145-158. Preston-Whyte, E. 1990 Qualitative perspectives on fertility trends among African teenagers. In W.P. Mostert and J.M. Lotter, eds., South Africa's Demographic Future: Pretoria: Human Sciences Research Council. Project for Statistics on Living Standards and Development 1994 South Africans Rich and Poor: Baseline Household Statistics. Cape Town: Southern African Labor and Development Research Unit. Samuel, J. 1990 The state of education in South Africa. In B. Nasson and J. Samuel, eds., Education: From Poverty to Liberty. Cape Town: David Philip. Thomas, D. 1996 Education across generations in South Africa. American Economic Review 86(2):330334. Thomas, D., and J. Maluccio 1996 Fertility, contraceptive choice and public policy in Zimbabwe. World Bank Economic Review 10(1):189-222. Treiman, D. 1996 On the backs of the blacks: Apartheid and Afrikaner upward mobility. Unpublished manuscript. Los Angeles, Calif.: Department of Sociology, University of California, Los Angeles.

OCR for page 138
--> van Zyl. J. 1994 History, scope and methodology of fertility and family planning surveys in South Africa. Pretoria: Human Sciences Research Council. (mimeographed) Weiss, A. 1995 Human capital vs. signalling explanations of wages. Journal of Economic Perspectives 9(4): 133-155.

OCR for page 138
--> Appendix ASaldru, University Of Cape Town Project for Living Standards and Development LITERACY ASSESSMENT MODULE [LAM] Section A: Listening, Comprehension, and Practical Math (about 5 minutes) Question 1: ''Imagine that you are taking a trip on the bus to a township that is located 280 kilometres away. The bus driver says that he will drive at a speed of 80 kilometres per hour. How many hours will the trip take to complete?" Question 2 : [in Mother Tongue] "Meshack would like to set up a fruit stand in the town market. The manager of the market says that Meshack must pay him R 100 to set up his stand. He must also pay R 25 per month to the market manager. Over the first year, how much in total must Meshack pay to the manager?" Section B: Reading Comprehension (15 minutes) Question 3: "When Mbaya was a child, he got very excited when his mother, Corfu, asked if he would like to go to the meat market with her. As they walked into the centre of town, the wonderful odours of meat-both fresh and spoiled—could be smelled up to one kilometre away. The hundreds of market stalls formed a row of almost 1 and 1/2 kilometres long. It took almost one hour to walk slowly from one end of the meat market to the other. "Sometimes Corfu would let Mbaya choose what meat they would buy that morning. The smell of fresh beef was Mbaya's favorite. But sometimes Mbaya would accidentally choose the beef that was not fresh. Corfu would go up close to the big piece of meat hanging from the rack and smell it. Once she was close to it, Corfu could tell immediately that the beef was not fresh. Then she would laugh at Mbaya and tease him for picking spoiled meat. The meat seller would be angry, as Corfu let on to other shoppers that his beef was not fresh. Mbaya would then start looking around for beef that seemed more fresh, no longer trusting that his nose is the best instrument for finding fresh meat."

OCR for page 138
--> Q3a. How long was the row of meat stalls, from one end to the other end'? a. 1 and 1/2 kilometres long. b. 1 kilometre long. c. It was very close from one end to the other end. d. Hundreds of stalls were lined up. Q3b. What did Mbaya most like to do inside the meat market? a. Try to find spoiled meat. b. Walk from one end to the other end of the market. c. Have his mother tease him when he found spoiled meat. d. Find fresh beef for his mother to buy. Q3c. What did not happen when Mbaya would choose a spoiled piece of meat'? a. The meat seller would get angry. b. Mbaya and his mother would leave the market. c. Corfu would tease Mbaya. d. They would keep shopping for fresh meat. Question 4: "Zenariah was riding to work in her usual combi. The driver and the woman sitting next to him, named Roseline, were arguing over whether it was any use for the woman's son to stay in school. The son, named Philemon, was 16. His secondary school had been closed for many days over the past 6 months. Teachers often did not show up for work. But the woman felt that if he could pass matric, Philemon could eventually find a good job, perhaps as a clerk or office worker. The driver, however, claimed that even university graduates were having difficulty finding jobs as clerks. Zenariah had graduated from the University of the Western Cape, and it had taken her 3 months to find her job as an assistant accountant. She was sympathetic to the woman's position, but also had to agree that until the economy improved, education would not guarantee a good job." Q4a. When Zenariah goes to work in the morning ... a. She usually takes the combi. b. She always sits next to Roseline. c. She tries to find different drivers and combis. d. She usually talks about her son, Philemon, in the combi.

OCR for page 138
--> Q4b. What kind of job did Philemon's mother hope he would find? a. Assistant accountant. b. Combi driver. c. Office worker. d. Teacher. Q4c. What was the combi driver's position? a. Schools are of high quality. b. Philemon should go to the University of the Western Cape. c. Completing school will lead to a good job. d. Schooling does not guarantee a good job. Section C: Computational Problems (15 minutes) Question 5:103 kg - 37 kg = _____ kg Question 6: R 35.50 x 7 = R _____ Question 7:25% of R 228 = R _____ Question 8: R 22.25 - R 7.88 = R _____ Question 9: "According to the doctor, the mother must buy 0.30 litres of cough mixture for her two sick children. She can either buy three bottles, each containing 0.10 litres, for R 9.50 per bottle or she can buy four bottles, each containing 0.08 litres, for R 7.00 per bottle. What is the least amount of Rand she needs to spend to get the 0.30 litres required by the doctor? Question 10: "Namane was trying to figure out her transport costs from the township to the city to get to her job. The combi cost R 2.00 each day. If she took a combi, then a taxi for part of the trip, she would have to spend R 3.50 each day. How much more would the taxi plus the combi cost for the week, than just taking a taxi, if she went to work five days during the week?"

OCR for page 138
--> Section E: Spoken Languages and General Reading Question 11: What language do you speak? a. At home or when talking with friends? b. At work (if currently working, otherwise go to Q12). Question 12: During the past two weeks, did you read any parts of the following items? a. A newspaper? If yes, in what language? b. A magazine'? If yes, in what language? c. A book of any kind'? If yes, in what language?