5
School Quality, Student Achievement, and Fertility in Developing Countries

Paul Glewwe

Introduction

Investments in education are increasingly viewed as essential for economic growth in developing countries (World Bank, 1990; United Nations Development Programme, 1990; Becker, 1995). Education also has nonmarket effects, such as better health and lower fertility, that may not be captured in income numbers at the national or individual level (Haveman and Wolfe, 1984). Parents' decisions regarding their children's education depend in part on the characteristics of local schools, most of which are public schools. Unfortunately, in many countries the low quality of schools severely limits households' opportunities for educating their children. Examples of countries with school-quality problems are Brazil (Harbison and Hanushek, 1992), Ghana (Glewwe, 1998), and Pakistan (World Bank, 1995). Many more developing countries also have poor school quality; these three examples simply illustrate that the problem exists in each region of the developing world.

Low school quality can take various forms; recent studies have shown that schools in developing countries suffer from many deficiencies that lead to reduced learning among students (Lockheed and Verspoor, 1991; Hanushek, 1995). While remedying these deficiencies should raise school quality and lead to substantial increases in student learning, much remains to be discovered about which policy options are most effective in achieving this goal. Moreover, evidence is incomplete concerning the likely impacts of improved school quality on fertility and other socioeconomic outcomes.



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 105
--> 5 School Quality, Student Achievement, and Fertility in Developing Countries Paul Glewwe Introduction Investments in education are increasingly viewed as essential for economic growth in developing countries (World Bank, 1990; United Nations Development Programme, 1990; Becker, 1995). Education also has nonmarket effects, such as better health and lower fertility, that may not be captured in income numbers at the national or individual level (Haveman and Wolfe, 1984). Parents' decisions regarding their children's education depend in part on the characteristics of local schools, most of which are public schools. Unfortunately, in many countries the low quality of schools severely limits households' opportunities for educating their children. Examples of countries with school-quality problems are Brazil (Harbison and Hanushek, 1992), Ghana (Glewwe, 1998), and Pakistan (World Bank, 1995). Many more developing countries also have poor school quality; these three examples simply illustrate that the problem exists in each region of the developing world. Low school quality can take various forms; recent studies have shown that schools in developing countries suffer from many deficiencies that lead to reduced learning among students (Lockheed and Verspoor, 1991; Hanushek, 1995). While remedying these deficiencies should raise school quality and lead to substantial increases in student learning, much remains to be discovered about which policy options are most effective in achieving this goal. Moreover, evidence is incomplete concerning the likely impacts of improved school quality on fertility and other socioeconomic outcomes.

OCR for page 105
--> This chapter examines the relationship between school quality and fertility in developing countries. A large amount of research has shown convincingly that the quantity of schooling in those countries is associated with significant reductions in fertility (Schultz, 1993), but very little research has examined the relationship between school quality and fertility. Given the low levels of school quality in many developing countries and increasing recognition of the need for improvement, two questions arise: First, do we really know how best to improve school quality? Second, what impact will increased school quality have on fertility? To answer these questions, the chapter first provides a critical assessment of the literature on the determinants of school quality in developing countries. It then examines the likely impact of improved school quality on fertility. Particular attention is given to the role played by cognitive skills, such as literacy and numeracy, and on the need to address a host of statistical estimation problems and data inadequacies that, regrettably, often receive insufficient attention in the literature. The chapter is organized as follows. The next section begins by presenting a simple economic model of the determinants of educational outcomes, focusing on the role of school quality, and uses this model to draw implications for empirical work. The third section provides a critical review of the current state of knowledge on the impact of school quality on learning. The following two sections address the relationship between school quality and fertility, reviewing, respectively, what economic theory says about the impact of school quality on fertility and the scant empirical evidence that is available. The final section presents concluding remarks. A Simple Economic Model Of Educational Outcomes Overview of the Issues Schooling provides children with many benefits. Most obvious are the skills taught explicitly as a part of the curriculum, such as literacy, numeracy, scientific knowledge, and advanced thinking skills. Schooling also provides social skills and values that can help children succeed in the adult world. Finally, a certain prestige may be associated with completing particular levels of education, so that one may be able to obtain better employment or a ''better" spouse (see Basu, this volume). The present discussion focuses on the basic cognitive skills school curricula are designed to impart, but occasional reference is made to other benefits of schooling as well. The cognitive skills acquired by a child per year of schooling depend on the characteristics of the child, of his or her household, and of the school attended. The variation in acquired cognitive skills that is due to school characteristics can serve as an indicator of school quality. In particular, school quality can be defined as those school (and teacher) characteristics that increase the cognitive

OCR for page 105
--> skills children acquire per year of schooling. This concept of school quality is fairly intuitive. As discussed below, however, much remains to be learned about which school characteristics lead to the acquisition of cognitive skills. The cognitive skills children acquire in school can play a very important role in determining their standard of living as adults. The best example is the role played by those skills in determining income, with better-educated individuals generally having higher incomes. At the same time, however, the exact relationship between cognitive skills and income is poorly understood and probably quite complex because different skills are likely to have different effects on income, and there has been very little research in either developed or developing countries on which skills are most important in this regard. Cognitive skills also affect individuals' living standards by helping to determine many other socioeconomic outcomes, such as health status, marriage prospects, and fertility. Yet almost nothing is known about which skills have the most important effect on these outcomes. Knowledge of the impact of different skills on income and on other socioeconomic outcomes could have implications for school curriculum. For example, if literacy skills were identified as much more important than, say, scientific knowledge in determining future income, it might be desirable to reduce some of the classroom time devoted to science while increasing the time devoted to language skills. Figure 5-1 provides a visual framework for conceptualizing the relationships among school quality, school attainment, cognitive skills, and socioeconomic outcomes. At the top of the diagram, school, child, and family characteristics influence both schooling and other socioeconomic outcomes. One can think of these characteristics as exogenous (beyond the control of the child and the family), at least initially.1 Given these exogenous characteristics, parents (perhaps considering their child's wishes) decide how long to send their child to school. School quality may influence this decision (arrow a in Figure 5-1) because higher-quality schools should provide more benefits per year of schooling, making additional years of schooling more attractive. In contrast, school costs (arrow b) will tend to reduce years of schooling. Many child characteristics can affect years of schooling (arrow c); on the positive side, more talented children are likely to go to school longer, while on the negative side, children with greater potential to contribute to household income may receive less schooling. Finally, several family characteristics can affect years of schooling (arrow d); two examples are household income and parental tastes for their children's schooling, both of which may increase educational attainment. Once a child enters school, the acquisition of cognitive skills begins. Time spent in school should increase skill acquisition (arrow f in Figure 5-1 ), as should 1   In particular, school quality may not be fixed when parents can choose from more than one school for their children. This issue is discussed in detail in the next section.

OCR for page 105
--> FIGURE 5-1 Relationship among school quality, school attainment, and student achievement.

OCR for page 105
--> school quality (arrow e). In addition, child and family characteristics can play important roles (arrows g and h, respectively); for example, child ability should increase learning, holding other factors constant, and parents' education can do the same if they help their children with schoolwork. Note that school costs have no impact on learning once years of schooling are accounted for; there is no reason that school fees paid should affect skills acquired beyond their role in determining years of schooling.2 The bottom of Figure 5-1 shows the impact of schooling and other factors on socioeconomic outcomes, of which fertility is one example. The cognitive skills a child acquires in school should have substantial effects on many socioeconomic outcomes (arrow i). Family and child characteristics may have separate effects (arrows j and k, respectively). For example, children with more motivation or a better work ethic may be more successful even after controlling for cognitive skills, and parents may use social connections to help their children obtain better jobs. School quality and years of schooling may also affect socioeconomic outcomes beyond the direct effect of cognitive skills (arrows I and m). For example, years of schooling may signal employers about traits they seek in workers (e.g., innate ability, motivation, work ethic) that are difficult to measure.3 Similarly, enrollment in a higher-quality school may improve a child's socioeconomic prospects because of the prestige attached to that particular school. Although the direction of causality in Figure 5-1 generally flows from top to bottom, some reverse causality is possible. These effects are indicated by dashed lines. For example, cognitive skills may determine years of schooling (arrow n) if schools prevent students from advancing to the next grade until they pass a standardized test; a child who fails such tests two or three times may be forced to drop out of school. As another example, one socioeconomic outcome that could affect years of schooling (arrow o) is pregnancy among female students; pregnant students may be forced to leave school because schooling is thought to be incompatible with their childcare responsibilities. Obviously, the simple scheme presented in Figure 5-1 could be made more realistic through the addition of more detail. However, it is intended only to provide a basic framework, not a complete picture. The next subsection presents a formal economic model of the determinants of school quality, school attainment, and the acquisition of cognitive skills that will prove useful in the literature assessment presented in the next section. 2   School costs can have an effect if parents spend money on extra classes or better material inputs. However, such additional inputs can be included as part of the influence of income, a family characteristic measured by arrow h. 3   In fact, a child's motivation and work ethic may be influenced by his or her schooling. These effects can be thought of as being among the social skills and values acquired from schooling.

OCR for page 105
--> The Model This chapter, while somewhat eclectic, approaches education issues from a (neoclassical) economic perspective. In particular, it takes the position that a model of rational behavior is needed to ensure that proper statistical procedures are used in attempting to estimate the impact of school characteristics and school policies on educational outcomes and of schooling and cognitive skills on socioeconomic outcomes. The argument is quite simple. Explicit models of human behavior provide significant insight into whether assumptions underlying specific statistical methods are satisfied. If a plausible theory suggests that some assumptions are not satisfied, the empirical results based on those methods may be invalid. The model may also suggest how to test to determine whether the statistical assumptions hold and what statistical procedure should be used if they fail to hold. Note that this is not intended to be the definitive model on schooling, only a plausible model that can illuminate several statistical issues. To formulate the model, it is assumed that parents make decisions for their children and that their objective is to maximize a utility function that has two arguments: household consumption (of goods and services) and child cognitive skills. More specifically, it is assumed that there are only two time periods. In time period 1, a child may attend school, work, or both (if both, the child first goes to school and works only after schooling is completed). In time period 2, the child becomes an adult and works.4 Whenever the child works in time period 1 or 2, part or all of the child's earnings may be given to his or her parents. The following simple utility function takes consumption in both time periods (C1 and C2) and child cognitive skills (A) as its arguments: (5.1) Note that δ is effectively a discount factor for future consumption, and σ is a parameter indicating tastes for educated children (higher values indicate a greater desire for educated children). Thus parents may value educated children for two distinct reasons: (1) educating children may increase the parents' levels of consumption, and (2) educating children directly affects utility (through σ). The next step is to explain how cognitive skills, A, are produced. The following simple production function keeps the mathematics at a relatively elementary level: (5.2) where a is the "learning efficiency" of the child, Q is school quality, and S is 4   To keep the exposition simple, it is assumed that there is only one child per family. The possibility that the number of children is another choice variable is considered in a later section.

OCR for page 105
--> years of schooling. Increases in either Q or S should increase the cognitive skills produced, so f( ) and g( ) are increasing functions of Q and S, respectively. The learning efficiency of the child, α, can represent several different factors, such as innate (i.e., genetically inherited) ability, child motivation and tastes for schooling, and parents' motivation (and capacity) to help their children with their schoolwork. For simplicity, these different factors are combined into a single parameter, α. The model is completed by specifying the relationships between consumption and child schooling and between child schooling and child income. Consumption levels in each time period are given by: (5.3) (5.4) where p is the price of schooling (e.g., annual tuition); Y1 and Y2 are (exogenous) parental income in periods 1 and 2, respectively; Yc is the income of the child when he or she works in either time period; and k is the fraction of that income given to the parents.5The last term in equation (5.3), (1 - S)kYc, requires some explanation. For convenience, S has been rescaled to represent the fraction of time spent in school by the child in time period 1. The remaining time during that period, 1 - S, is spent working. This rescaling is purely for notational convenience and has no effect on the results; however, to keep the vocabulary simple, S is still called "years of schooling." Finally, equation (5.5) relates the child's cognitive skills to employment income in either time period: (5.5) where π can be thought of as the productivity of cognitive skills in the labor market. By substituting equation (5.2) into equation (5.5), then equation (5.5) into equations (5.3) and (5.4), and finally equations (5.2), (5.3), and (5.4) into equation (5.1), parents' utility is expressed as a function of years of schooling (S): 5   Note that it is implicitly assumed that there is no borrowing or saving; the only way to transfer income between periods 1 and 2 is to invest in children's education. This assumption is made for simplicity. Allowing for other possibilities would complicate the mathematics and is not pursued here. In general, allowing for borrowing or saving would reduce the need for parents to invest in their children's education. However, it would not completely eliminate the investment motive for educating one's children because almost all investments are risky, and thus it is prudent to diversify one's investments among several different alternatives, including children's education.

OCR for page 105
--> (5.6) One can show (see the appendix) how the optimal (utility-maximizing) value of years of schooling is affected by changes in the model's various parameters. The findings are all intuitively plausible. Years of schooling increases when the following increase: (1) children's learning efficiency (α), (2) school quality (Q), (3) the weight (δ) parents give to future relative to current consumption, and (4) parents' tastes for schooling (σ). Years of schooling decreases when the price of schooling (p) rises. Finally, years of schooling is likely, though not certain, to rise when parents expect to receive a larger proportion (k) of their children's income from working and when the value of cognitive skills in the labor market (π) is higher. Sometimes parents can choose not only years of school, but also school quality. The model can easily be extended to the case in which both years in school (S) and school quality (Q) are choice variables. To make the model more realistic, assume that the price of a school depends on its quality6: (5.7) where po is the "base" price of schooling. Thus high-quality schools have higher costs (p rises as Q increases). Equation (5.7) may at first seem to have an arbitrary functional form: If one school has a level of quality twice as high as that of another, why should the price be exactly twice as high, as opposed to less or more than twice as high? In fact, equation (5.7) is simply a normalization; Q should be interpreted as an index of expenditures on quality. Whether or not, say, doubling expenditures doubles the impact of school quality on learning (i.e., doubles f(Q)) depends on the functional form of f(Q). After replacing p with poQ in equation (5.6), one obtains the following expression, which is to be maximized with respect to both S and Q: (5.8) Without further assumptions about the functional forms of f( ) and g( ), one cannot determine how changes in the various parameters, such as po, α, δ, and σ, affect S and Q. For ease of exposition, assume that f(Q) = Qß, where ß > 0, and g(S) = Sg where γ > 0. Differing values of ß and γ allow for a wide range of the shapes of both f( ) and g( ). Both ß and γ must be greater than zero to guarantee that f( ) and g( ) increase as Q and S, respectively, increase. Using these func- 6   Making this assumption is not only more realistic, but also necessary for the model to make sense. If higher quality could be obtained at no additional cost, all households would choose the highest quality possible.

OCR for page 105
--> tional forms, one can show (see the appendix) that parents' utility as given in equation (5.8) is maximized by setting S and Q as follows: (5.9) (5.10) where the asterisk indicates the utility-maximizing (optimal) level for each variable. The optimal level of cognitive skills (A) is obtained by inserting equations (5.9) and (5.10) into equation (5.2). Note that the optimal levels of S and Q in equations (5.9) and (5.10) are intuitively reasonable. Both years of schooling (S) and school quality (Q) are higher when parents give higher weight (δ) to future relative to current consumption, and parents have higher tastes for schooling (σ). School quality (Q) increases as learning efficiency (α) increases, but decreases as the base price of schooling (po) increases. A less plausible result is that years of schooling depends on neither the base price of schooling nor the innate ability of the child. This result admittedly is due to the functional forms used for f( ) and g( ), but it is not as unreasonable as it may first appear to be. What happens is that parents, in response to a lower base price or higher child learning efficiency, shift to higher quality, which raises their children's cognitive skills without changing years of schooling. By opting for higher quality instead of more years of schooling, parents avoid a cost associated with the latter—a reduction in the length of time a child works during time period 1 (see equation (5.3)). If school quality were not a choice variable, greater learning efficiency (α) or a lower price would lead to increased years of schooling, as explained above. A final result that also appears counterintuitive is that an increase in the propensity of children to give money to their parents (k) or in the market return to cognitive skills (π) decreases years of schooling. Here again intuition suggests that the best response to such changes is to increase school quality and reduce time spent in school, which will increase the time the child spends working in time period 1.7 Implications of the Model for Statistical Analysis The model presented above is useful for discussing several statistical issues involved in measuring the impact of school quality on learning. Most empirical studies of the impact of school characteristics on the acquisition of cognitive 7   School quality is likely to increase when k or π increases, but it may decrease. In the event of a decrease, total cognitive skills attained must decline, but this loss in income to the parents is outweighed by the increase in income due to the child's working longer in the first time period.

OCR for page 105
--> skills are based on linear statistical models, an approach that simplifies estimation. Taking the logarithm of both sides of equation (5.2) and assuming exponential functional forms for f( ) and g( ) yields an equation that is linear in the logarithms of the variables.8 For ease of exposition, assume this latter equation is linear in the original variables; this assumption can be made because the following line of reasoning also applies to the linear-in-logarithms case. Thus the equation of interest is: (5.2') where the µ coefficients are unknown parameters to be estimated. The residual term e is added to account for the fact that in empirical work, no data fit the model perfectly. Equation (5.2') can be estimated using simple statistical methods, but its specification of school quality is oversimplified. It is more realistic, and more useful for policy analysis, to decompose school quality into a function or index of the different school characteristics that promote learning. Doing so yields: (5.2") In equation (5.2"), Q has been replaced by an index of n different school characteristics (Q1, Q2 etc.) that affect the acquisition of cognitive skills. What policy makers want to know, and analysts need to estimate, is the magnitude of the various τ's. These estimates can be combined with data on the costs of the different school characteristics to assess the cost-effectiveness of each characteristic in promoting learning. Equation (2") implicitly assumes that learning efficiency, α, can be observed. In fact, there are often few data on the factors determining learning efficiency, so equation (5.2") must be rewritten as: (5.11) where α (and its coefficient) is combined with e to produce u, a residual term that represents both random "noise" from imperfectly fitting data and the impact of unobserved aspects of learning efficiency (α) on cognitive skill acquisition (A). Examples of learning efficiency variables that are difficult to observe are the child's innate ability and motivation and the parents' motivation (and ability) to help children with schoolwork.9 8   In particular, assuming that f(Q) = ΘQß and g(S) = ΦSg for some parameters Q, ß, F, and γ. 9   While some of these factors may be measured (e.g., using an IQ test to measure innate ability and using parental schooling to indicate parents' ability to assist their children), it is highly unlikely that

OCR for page 105
--> Equation (5.11) is commonly estimated using ordinary least squares (OLS) regression. OLS provides unbiased estimates of all parameters (the µ's and τ's) in equation (5.11) only if the residual term, u, is uncorrelated with S and the various Q's. However, the simple model presented earlier shows that such correlation is very likely; in equation (5.10), higher learning efficiency (α) increases school quality (Q), implying that u, which represents (in part) variation in α, is positively correlated with the various Q's. The result in general will be overestimation of the associated parameters (τ's). Few empirical studies do anything to avoid this statistical problem. If school quality is not a choice variable, the estimation problem discussed above can be avoided. In rural areas of many developing countries, school quality may well be exogenous because each village has only one school, and villages are too far apart for children to attend school in a neighboring village. In such cases, the Q variables in equation (5.1 1) may not be correlated with the error term, u. Yet even under this scenario, parents may still be able to influence school quality. First, they may directly affect quality at the sole local school by participating in the parent-teacher association (PTA) or using political connections to obtain better educational services. Second, they may send their children to live with relatives or at a boarding school, thus allowing them to attend a nonlocal school.10 Third, families with higher tastes for educated children may migrate to areas with better schools, a common occurrence in the United States. Since parents may affect school quality even in rural areas, then, overestimation is possible. Yet it is also possible that endogenous school quality leads to underestimation. Even if parents cannot affect school quality, it could be correlated with the error term because governments may provide better schools to areas with unobserved education problems (Pitt et al., 1993). These unmeasured problems would again be relegated to u in equation (5.11), producing negative correlation between the error term and the school quality variables (Q's) and thus underestimating the impact of school quality. Even when school quality is completely uncorrelated with the error term in equation (5.1 1), years of schooling (S) may be correlated. Note that in equation (5.9), parents with higher tastes for schooling (σ) send their children to school longer. These tastes are rarely measured, and any effect they have on learning efficiency (e.g., such parents help children more with schoolwork) would be reflected in the error term u, leading to positive correlation with S. This in turn     one can measure all of them. Indeed, it is not even clear that innate ability can be measured; any test that purports to do so (in the sense of genetic endowment) is likely to reflect environmental factors (American Psychological Association, 1995). One possible way to get around this problem is to use data on twins (see Behrman et al., 1994). However, such data are very rare in developing countries. 10   About 19 percent of secondary students in rural Peru live away from their families (Gertler and Glewwe, 1990), and the same applies to 27 percent of middle school students in Ghana (Glewwe and Jacoby, 1994).

OCR for page 105
--> In summary, there is a small amount of evidence on the impact of school quality on the fertility decisions of today's parents, which can be inferred by looking at the impact of school availability on fertility. One study finds an insignificant effect of primary schools, while another finds a small but significant effect of primary schools, but no significant effect of middle or high schools. While these studies constitute a sample size of two and can be criticized on methodological grounds, they are relatively careful as compared with the literature reviewed earlier. Thus the little evidence available suggests no sizeable effects of school quality on the fertility decisions of today's parents. Of course, school quality may affect fertility in the future. This possibility is examined in the next subsection. Impact on Future Decisions of Today's Children We turn now to the impact of improvements in school quality on the future fertility decisions of today's children. Figure 5-3 shows that the relationship between education and fertility, while complex, must work through one of the following four pathways: (1) changes in values brought about by schooling, (2) changes in fertility knowledge, (3) cognitive skills learned in school, and (4) a noncausal relationship whereby an individual's tastes determine both schooling outcomes and fertility. Knowledge of the relative importance of these pathways and of the distinct contributions (if any) of specific types of cognitive skills would provide a much clearer picture of the role played by school quality in affecting future fertility, and might even demonstrate how changes in school curriculum could lead to reductions in fertility. There appear to be only three studies that attempt to disentangle the various pathways through which schooling leads to reduced fertility. The first, Lam and Duryea (1999), uses Brazilian data to look at the impact of years of schooling on both fertility and labor force participation. At low levels of schooling, the authors find strong effects of years of schooling on fertility, but only weak effects on labor supply. They also find strong effects of low levels of schooling on child health. While many studies have examined the impact of schooling on both fertility and labor force participation, the Lam and Duryea study is distinctive in that it uses the results to say something about the pathways by which schooling affects fertility. In particular, the authors interpret the evidence, especially the effects of low levels of schooling, as indicating that the effects of schooling on fertility work, at least in part, through pathways other than those based on the value of a mother's time in the labor market. Although the analysis of Lam and Duryea is intriguing, some of the findings can be criticized. First, their regressions on the determinants of fertility assume that a mother's schooling and that of her spouse are exogenous, which can be questioned. In particular, the positive relationship between higher schooling and lower fertility may reflect, at least in part, the impact of tastes as depicted in

OCR for page 105
--> Figure 5-2. Second, part of the impact of mother's schooling may work through spouse's schooling, in that a better-educated woman will choose a better-educated spouse (see also Basu, this volume). Third, many other variables that may determine fertility, such as local family planning or health services, are not included in the regression. Fourth, the authors ignore the possibility that part of the causality underlying the relationship between schooling and fertility may run from the latter to the former as a result of pregnancy forcing young women to drop out of school. A second study addressing the pathways by which schooling leads to reduced fertility is by Oliver (1997). It uses data on women's cognitive skills to understand the relationship between education and fertility. In particular, it uses the 1988-1989 Ghana Living Standards Survey to show that a woman's level of literacy, but not her level of mathematical ability, leads to reduced fertility.25 In addition, it finds that years of schooling has a strong negative effect over and above the effect due to literacy. This finding suggests that either values, fertility knowledge, or perhaps cognitive skills other than literacy and numeracy can play a role in reducing fertility beyond the role played by the acquisition of basic cognitive skills. Oliver's findings can be interpreted in terms of Figure 5-3. She shows that improvements in school quality reduce fertility by raising cognitive skills (arrow d in Figure 5-3). In addition, she shows that more years of schooling reduces fertility even after controlling for the effect that works through cognitive skills. This could occur because schooling changes students' values, their value in the marriage market, or their fertility knowledge (arrows a, b and c, respectively). Unfortunately, the data used by Oliver cannot distinguish among these alternative pathways. Another intriguing aspect of Oliver's study is that it relates the impact of literacy and years of education on fertility to specific components of school quality. In particular, the study presents estimates of the impact of six middle school quality improvements on the number of children ever born, and for three of those a cost-effectiveness ratio is given. These results are shown in Table 5-2. The total impact of the different school improvements varies from a fertility reduction (measured in terms of children ever born) of 0.04 due to providing textbooks to a reduction of 0.64 due to providing blackboards. Note that much of this effect comes about because these schooling improvements also raise years of schooling (see column 2 of Table 5-2). Of the three schooling improvements for which cost-effectiveness figures are calculated, provision of blackboards is by far the least expensive avenue to reduce fertility. Repair of leaking classroom roofs is more expensive, and the costliest of the three is provision of textbooks. It would be interesting to compare these cost figures with the cost of reducing 25   Both literacy and numeracy are measured in terms of scores on cognitive skills tests.

OCR for page 105
--> TABLE 5-2 Impact of Middle School Improvements on Children Ever Born in Ghana   Impact on Children Ever Born     Middle School Improvement Reading Scores Years of School Total Cost-Effectivenessa Reducing travel time from 2 to zero hours 0.121 0.171 0.292 — Raising average teacher experience from 2 to 10 years 0.112 0.158 0.270 — Providing a school library 0.066 0.094 0.160 — Repairing classrooms that cannot be used when it rains 0.288 0.153 0.441 1,273-2,545 cedis Providing blackboards in schools where none exist 0.440 0.200 0.640 100-200 cedis Providing 50 textbooks per room in schools that now have 25 per room 0.044 — 0.044 36.364-60.605 cedis a These figures indicate the cost of reducing total predicted (future) children ever born to students in the improved classroom by 1. Note that the data presented here were collected in 1988-1989, at which time the exchange-rate value of the Ghanaian cedis was about 200 cedis per U.S. dollar. SOURCE: Oliver (1997). fertility through typical family planning programs; unfortunately, Oliver does not do this in her study. Oliver's study does suffer from some methodological shortcomings. The schooling of the mother is assumed to be exogenous, which ignores the possible role of tastes in determining the relationship between schooling and fertility. It is also likely that omitted-variable bias is present because several variables that may determine fertility are not included, such as the availability of family planning and health services. Third, the possibility that fertility could reduce schooling because pregnant girls quit school is not considered. Finally, although the attempt to relate fertility reduction to specific changes in school quality is particularly valuable, the study does not address aspects of school quality that are likely to change a child's values. This is an admittedly difficult task, and it is questionable whether any existing data could be used to investigate this aspect of schooling and fertility.

OCR for page 105
--> A final study that sheds light on the impact of school quality on fertility is that of Thomas (see Chapter 6), who uses recent household survey data from South Africa to examine how schooling affects fertility. In particular, Thomas examines the effect of years of schooling and of scores on mathematics and reading comprehension tests on children ever born. His findings are similar to those of Oliver. Reading comprehension has an important effect, but mathematics skills have no significant effect, and even when both of these variables are included in the regression, years of schooling still has a strong independent effect. Unfortunately, Thomas' South African data cannot be used to examine how specific aspects of school quality determine the acquisition of cognitive skills. As with the other two studies, Thomas' work suffers from several shortcomings, most of which cannot be corrected with the available data. The role of tastes in determining both schooling and fertility may seriously bias the estimates, or at least reduce what can be inferred from them. The possibility of reverse causality, that is, of women dropping out of school because of pregnancy, is not considered. Finally, the general problems of omitted-variable bias and measurement error are not addressed. In summary, these three studies support the common finding that schooling is associated with reduced fertility. The studies of Oliver and Thomas show that reading comprehension skills, but not mathematics skills, directly affect fertility, and Oliver's paper shows the association between specific changes in school quality and reduced fertility. Finally, the results of all three studies imply that the impact of schooling on fertility is not just a matter of attaining mathematics and reading skills. However, much more remains to be learned. What can be done about this need is briefly discussed in the following subsection. Recommendations for Future Research The most obvious recommendation for future research on the impact of school quality on fertility is that much more of such research should be done. First, more research is needed on the impact of school quality on the current fertility decisions of today's parents. This work should examine the impact of school quality directly, as opposed to relying on the price of schooling or the availability of a local school. Second, much more can be done to examine the impact of current changes in school quality on the future fertility decisions of today's children. As in the study by Oliver (1997), this work can be divided into two parts: the impact of a variety of cognitive skills, knowledge, and even attitudes on fertility, and the impact of improvements in school quality on all of those determining factors. While some initial work has been done on the most basic cognitive skills, almost nothing has been done on other cognitive skills or on knowledge and attitudes. As with the literature on the determinants of school quality, another obvious

OCR for page 105
--> recommendation is that more useful data should be collected. Data are needed on school quality, fertility, and cognitive skills, plus attitudes and fertility knowledge (and other knowledge as well, such as health knowledge), from the same households. If data that follow young women over several years can be collected, a link can be made between specific schooling outcomes and subsequent fertility outcomes. Alternatively, if only cross-sectional data can be collected, the data should include the skills of adult women (determined by administering achievement tests) to see how those skills are related to fertility. Both types of data go well beyond what is usually collected in developing countries, which explains the paucity of research on the pathways by which school quality affects fertility. Finally, the call for more randomized evaluations of education policies made earlier also applies here. Although this section has not gone into detail on the statistical problems that complicate attempts to estimate the impact of school quality on fertility (see Strauss and Thomas, 1995, for a discussion of this issue), many such problems do arise that are difficult to solve. Randomized evaluations can, in principle, get around almost all of these problems.26 There have apparently been no randomized trials relating schooling to fertility in developing countries, perhaps because the time lag between a schooling intervention and the future fertility outcomes of today's children may be many years. Again, bilateral and multilateral aid agencies need to take the lead on this matter by building randomized evaluations into their development projects. Conclusion The main message of this chapter is that very little is known about what determines school quality in developing countries, and even less is known on how school quality affects fertility. This state of affairs exists even though a large number of studies have been done on the first topic—the relationship between school quality and learning. There are two main reasons that little is known about this relationship: (1) statistical tools have not been used very carefully, and (2) the data available are usually inadequate for the task. Both of these problems need to be addressed in future work. In addition, randomized evaluations of schooling interventions hold promise for addressing some of the more intractable statistical problems. Given the apparent inefficiencies in the way schools operate in developing countries, improving the general state of knowledge on which interventions are most cost-effective has the potential to bring about sizeable increases in learning, and eventually in the standard of living, in those countries. While the relationship between the quantity of schooling and fertility has 26   However, recall that poor design and/or implementation can compromise results based on randomized evaluations.

OCR for page 105
--> been documented numerous times, the effect of school quality on fertility has rarely been examined. Better knowledge of the various pathways involved in the latter could lead to policy recommendations that would ultimately help bring about further reductions in fertility. For example, the little evidence produced thus far suggests that spending more class time on reading skills and less on mathematics may reduce fertility. However, much more must be learned before such broad policy recommendations can be made, and one must bear in mind that fertility reduction is only one of many different schooling outcomes to be considered when contemplating such changes. Finally, since this chapter was written from an economist's perspective, there is one more difficult question to raise: Why should the government concern itself with trying to alter households' fertility decisions? There is a tendency among many demographers to assume at the outset that fertility levels in developing countries are too high. This may be so, but it should be demonstrated instead of merely assumed. If demographers want to win the support of economists on this issue, more theoretical and empirical work may be needed. Acknowledgments The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They do not necessarily represent the views of the World Bank, its executive directors, or the countries they represent. I would like to thank Andrew Foster, Bruce Fuller, Hanan Jacoby, Emmanuel Jimenez, the editors of this volume, and two anonymous referees for very useful comments on and/or discussion of previous drafts. References Ainsworth, M., K. Beegle, and A. Nyamete 1996 The impact of women's schooling on fertility and contraceptive use: A study of fourteen sub-Saharan African countries. World Bank Economic Review 10(1):85-122. Alderman, H., J. Behrman, S. Khan, D. Rose, and R. Scott 1995a Public schooling expenditures in rural Pakistan: Efficiently targeting girls and a lagging region. In D. van de Walle and K. Nead, eds., Public Spending and the Poor: Theory and Evidence. Baltimore, Md.: The Johns Hopkins University Press. Alderman, H., P.A. Chiappori, L. Haddad, J. Hoddinott, and R. Kanbuv 1995b Unitary versus collective models of the household: Is it time to shift the burden of proof? World Bank Research Observer 10(1): 1-19. American Psychological Association 1995 Intelligence: Knowns and Unknowns. Report of a task force established by the Board of Scientific Affairs of the American Psychological Association. Washington, D.C.: American Psychological Association. Becker, G. 1981 A Treatise on the Family. Cambridge, Mass.: Harvard University Press.

OCR for page 105
--> 1995 Human Capital and Poverty Alleviation. HRO Working Paper No, 52. Human Resources Development and Operations Policy Vice Presidency. Washington, D.C.: The World Bank. Behrman, J. 1997 Intrahousehold Distribution and the Family. In M. Rosenzweig and O. Stark, eds., Handbook of Population and Family Economics. Amsterdam: North Holland. Behrman, J., M. Rosenzweig, and P. Taubman 1994 Endowments and the allocation of schooling in the family and in the marriage market: The twins experiment. Journal of Political Economy 102(6)1131-1174. Cox, D., and E. Jimenez 1991 The relative effectiveness of private and public schools: Evidence from two developing countries. Journal of Development Economics 34:99-121. Fuller, B., and P. Clarke 1994 Raising school effects while ignoring culture?: Local conditions and the influence of classroom tools, rules and pedagogy. Review of Educational Research 64(1)119-157. Gertler, P., and P.Glewwe 1990 The willingness to pay for education in developing countries: Evidence from rural Peru. Journal of Public Economics 42:251-275. Glewwe, P. 1998 The Economics of School Quality Investments in Developing Countries: An Empirical Study of Ghana. London: Macmillan. Glewwe, P., M. Grosh, H. Jacoby, and M. Lockheed 1995 An eclectic approach to estimating the determinants of achievement in Jamaican primary school. World Bank Economic Review 9(2):231-258. Glewwe, P., and H. Jacoby 1994 Student achievement and schooling choice in low income countries. Journal of Human Resources 29(3):843-864. Glewwe, P., M. Kremer, and S. Moulin 1998 Textbooks and Test Scores: Evidence from a Prospective Evaluation in Kenya. Draft paper. Washington, D.C.: The World Bank. Godfrey, L.G. 1988 Misspecification Tests in Econometrics. Cambridge: Cambridge University Press. Hanushek, E. 1986 The economics of schooling: Production and efficiency in public schools. Journal of Economic Literature 25:1141-1177. 1995 Interpreting recent research on schooling in developing countries. World Bank Research Observer 10(August):227-246. Hanushek, E., C. Benson, R. Freeman, D. Jamison, H. Levin, R. Maynard, R. Murnane, S. Rivkin, R. Sabot, L. Solomon, A. Summers, F. Welch, and B. Wolfe 1994 Making Schools Work: Improving Performance and Controlling Costs. Washington, D.C.: Brookings Institution. Harbison, R., and E. Hanushek 1992 Educational performance of the poor: Lessons from rural northeast Brazil . New York: Oxford University Press. Haveman, R., and B. Wolfe 1984 Education and economic well-being: The role of non- market effects. Journal of Human Resources 19:377-407. Heckman, J., and J.A. Smith 1995 Assessing the case for social experiments. Journal of Economic Perspectives 9(2):85110.

OCR for page 105
--> Hedges, L., R. Laine, and R. Greenwald 1994 Does money matter? A meta-analysis of studies of the effects of differential school inputs on student outcomes. Educational Researcher 23:5-14. Heyneman, S.P., D.T. Jamison, and X. Montenegro 1984 Textbooks in the Philippines: Evaluation of the pedagogical impact of nationwide investment. Educational Evaluation and Policy Analysis 6(2): 139-150. Jamison, D., B. Searle, K. Galda, and S. Heyneman 1981 Improving elementary mathematics education in Nicaragua: An experimental study of the impact of textbooks and radio on achievement. Journal of Educational Psychology 73(4):556-567. Jimenez, E., M. Lockheed, E. Luna, and V. Paqueo 1991 School effects and costs for private and public schools in the Dominican Republic. International Journal of Education Research 15:393-410. Jimenez, E., M. Lockheed, and N. Wattanawaha 1988 The relative efficiency of public and private schools: The case of Thailand. World Bank Economic Review 2:139-164. Kagitcibasi, C., D. Sunar, and S. Bekman 1993 Long-Term Effects of Early Intervention. Department of Education, Bogadzdi University, Istanbul, Turkey. Kremer, M. 1995 Research on schooling: What we know and what we don't, a comment on Hanushek. World Bank Researcher Observer 10:247-254. Kremer, M., S. Moulin, D. Myatt, and R. Namunyu 1997 Textbooks, Class Size and Test Scores: Evidence from a Prospective Evaluation in Kenya. Cambridge, Mass.: Department of Economics, Massachusetts Institute of Technology. Lam, D., and S. Duryea 1999 Effects of Schooling on Fertility, Labor Supply, and Investments in Children, with Evidence from Brazil. Journal of Human Resources. Lockheed, M. 1995 Educational assessment in developing countries: The role of the World Bank. In T. Oakland, ed., International Perspectives on Academic Assessment. Norwell, Mass.: Kluwer Academic Publishers. Lockheed, M., and A. Verspoor 1991 Improving Primary Education in Developing Countries. New York: Oxford University Press. Oliver, R. 1997 Fertility and women's schooling in Ghana. In P. Glewwe, ed., The Economics of School Quality Investments, in Developing Countries: An Empirical Study of Ghana. London: Macmillan. Pitt, M., M. Rosenzweig, and D. Gibbons 1993 The determinants and consequences of the placement of government programs in Indonesia. World Bank Economic Review 7(3):319-348. Rosenzweig, M. 1982 Educational subsidy, agricultural development and fertility change. Quarterly Journal of Economics 97(1):67-88. Schultz, T.P. 1993 Returns to women's education. In E. King and A. Hill, eds., Women's Education in Developing Countries: Barriers, Benefits, and Policies. Baltimore, Md.: The Johns Hopkins University Press.

OCR for page 105
--> Strauss, J., and D. Thomas 1995 Human resources: Empirical modeling of household and family decisions. In J. Behrman and T.N. Srinivasan, eds. Handbook of Development Economics, Volume IIIA. Amsterdam: North Holland. United Nations 1995 Women's Education and Fertility Behavior: Recent Evidence from the Demographic and Health Surveys. New York: Department for Economic and Social Information and Policy Analysis, United Nations. United Nations Development Programme 1990 Human Development Report. New York: Oxford University Press. Velez, E., E. Schiefelbein, and J. Valenzuela 1993 Factors Affecting Achievement in Primary Education. HRO Working Paper No. 2. Washington, D.C.: The World Bank. World Bank 1990 World Development Report. New York: Oxford University Press. 1995 Improving Basic Education in Pakistan: Community Participation, System Accountability and Efficiency. Washington, D.C.: The World Bank, South Asia Region.

OCR for page 105
--> Appendix A Simple Two-Period Model Of School Choice To determine how the optimal (utility-maximizing) value of years of schooling is affected by changes in the parameters of the model given in the main text, assume first that school quality is given exogenously. As given in equation (5.6) in the text, the expression to be maximized with respect to S is: (A-1) The first and second derivatives of U with respect to S are: (A-2) (A-3) Totally differentiating the first-order condition (the condition that equation (A-2) = 0) yields: (A-4) It is immediately clear that the coefficient associated with dp (i.e., -1) is negative and that the terms associated with dσ and dδ are positive. The fact that the second-order condition, (A-3), must be negative implies that the term associated with dS is positive. The fact that the first-order condition, (A-2), equals zero implies that the terms associated with dα and dQ are positive. Since k and π always appear together, their product can be treated as a single variable, denoted by kπ. The term associated with dkπ cannot be signed unambiguously, but it will be positive if σ is relatively small. Now turn to the case where school quality, Q, is a choice, and higher school quality implies a higher tuition fee. As given in equation (5.8) in the text, under the assumption that f(Q) = Qß and g(S) = Sg, the expression to be maximized with respect to S and Q is: (A-5) The first and second derivatives of (A-5) with respect to S are: (A-6)

OCR for page 105
--> (A-7) The first and second derivatives of (A-5) with respect to Q are: (A-8) (A-9) Note that the requirement that both second derivatives equal zero implies, by (A-9), that ß < 1, so that Qß is a concave function of Q. Setting (A-6) and (A-8) equal to zero, dividing (A-6) by Q and (A-8) by S, and substituting (A-6) into (A-8) (i.e., substituting out Po) yields: (A-10) Dropping Qß-1 from both sides of (A-10) and multiplying both sides by S1-γ yields the following optimizing solution for S: (A-11) Clearly, this solution is only plausible if y> ß. Substituting (A-11) into (A-7) and noting that y > ß implies that (A-7) is negative, which means that the second-order condition for maximization is satisfied. Setting (A-8) equal to zero, solving for Q, and replacing S with (A-11) yields the optimizing solution for Q: (A-12) The impact of most of the parameters in (A- 12) is clear. The exceptions are k and π , which again can be treated as a single variable kπ. Differentiating Q with respect to kπ yields: (A-13) The expression in (A- 13) will be positive if (1 + δ) + (1 - γ)σ/kπ > 0, and negative otherwise. Thus the impact of k and π on Q will depend on the values of δ, γ, σ, k, and π. Note that if σ is sufficiently small, this term will be positive even if αγ > 1, which is a similar condition for the case when Q is exogenous.