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 287
--> 10 Fertility and Education: What Do We Now Know? Parfait M. Eloundou-Enyegue Introduction The preceding chapters in this volume have examined various associations between education and fertility. This chapter addresses the broad question of what these associations mean in light of current scholarship. The title of this chapter poses essentially the same question as that of a classic review by Cochrane (1979): Fertility and Education: What Do We Really Know? Presumably, such a similarity in the question posed reflects a similarity of preoccupation. In both cases, the objective is apparently to summarize the existing evidence on education-fertility relationships. Both works also reflect a critical stance and imply an invitation to transcend superficial interpretations and seek deeper understanding. Perhaps as important is a common rhetorical flavor. While apparently calling for literal answers, both works could usefully be addressed by focusing on their raison-d'être: Why, in spite of countless empirical studies on the subject, may one still entertain doubts about the meaning of education-fertility connections? And does such soul searching indicate a lack of scientific progress? Such questions are unavoidable today in view of the knowledge gains that can be assumed to have occurred in the decades since Cochrane's review. Accordingly, they constitute the focus of the present discussion. Specifically, this chapter addresses the persistent difficulty of deriving general conclusions from existing studies on education and fertility. Because most of these difficulties—including those of definition; functional form; and the interactive, contextually variable, multifaceted, and cross-generational nature of relationships—are well known, the emphasis is not on their enumeration, but on
OCR for page 288
--> their integration within a systematic analysis of historical changes in the education-fertility discourse. The main argument developed throughout is that the persistence of interpretive difficulties—and the resulting ambiguity in the education-fertility discourse—arises largely from an uneven growth in the four pillars that sustain this discourse: policy agenda, theory, methodology, and empirical evidence. On the one hand, the volume of empirical evidence has greatly expanded in the last two decades as the result of steady progress in research technology and the increased availability of large data sets. On the other hand, this empirical expansion has uncovered new complexities (interactions, nonlinearities, and contextual variability), some of which remain unexplained because of a lag in theoretical development. Furthermore, this accumulation of evidence has been accompanied by an even more rapid raising of methodological standards (notably, a greater demand for data and statistical tools that address issues of temporal sequence, unobserved heterogeneity, and endogeneity). The result has been dissatisfaction with much of the earlier evidence that fails to conform to these high standards. Finally, new policy demands have encouraged the study of additional linkages, including reverse, indirect, and intergenerational influences between education and fertility. In short, it has become more difficult to assess what education-fertility associations mean because the body of evidence has grown at once too complex in comparison with existing theories and too crude in light of current methodological standards and heightened policy demands. Together, these trends have resulted in a paradoxical situation in which the education-fertility discourse has become more diffident as the facts accumulated and analytical methods have improved. To state the point differently, the current uncertainty does not reflect a lack of scientific progress, but a lag in the progress achieved on certain fronts, including theoretical arguments and actual methodological practices. The rest of the chapter develops this argument by reviewing and comparing the progress achieved on the policy, theory, methodology, and empirical fronts. This evolution is summarized in Table 10-1, which distinguishes three time periods, ending in the mid-1970s, mid-1980s, and mid-1990s, respectively.1 Until the mid-1970s, the key policy objective of education-fertility research was to assess how one's education affects an individual's fertility. In fact, studies focused even more narrowly on the effects of the formal schooling of women on their fertility. The theoretical expectation, derived from modernization and microeconomic theories, was a negative effect. Consistent with this policy emphasis and theoretical expectation, analysis methods consisted of recursive mod- 1 Clearly, this is an arbitrary breakdown. Because history does not proceed in orderly decennial jumps, this historical discontinuity must be viewed only as a didactic necessity. Similar didactic considerations require overstating the consensus on the definition of key concepts, notably theory and methodology.
OCR for page 289
--> TABLE 10-1 Education and fertility research in developing countries: Changes in underlying policy issues, theory, methodology, and empirical evidence.
OCR for page 290
--> els featuring fertility as the dependent and education as the independent variable. These analyses relied on cross-sectional data on either individuals or countries. The empirical evidence at the time was rather limited in both volume and geographic coverage, most data being drawn from the more industrialized nations. By and large, this evidence indicated negative associations between education and fertility. Thus, there was a good match among (1) the scope and precision of the policy question addressed (whether education reduces fertility), (2) confidence in the use of methodology based on regression and cross-sectional analysis to answer this question (despite some qualifications), (3) the scope and precision of prevailing theories (education should reduce fertility), and (4) the nature of the available evidence (findings showed negative relationships most of the time). Because of this match, the education-fertility discourse was relatively unequivocal. During the next decade (1975-1985), several developments occurred. First, policy demands required better specification of the education benefits that affect fertility, as well as processes through which these benefits operate. Second, more data became available (mainly through the World Fertility Surveys [WFS]), some of which indicated atypical cases in which education had no or a positive effect on fertility (see also Diamond et al., this volume). These atypical patterns suggested the possibility that education may have differing effects on the various proximate determinants of fertility and on the demand for children. It therefore became necessary to distinguish the effects of education on, say, nuptiality, contraceptive use, and fertility demand. Accordingly, while statistical analysis still ran in one direction, there was greater emphasis on the paths through which education effects operate. Overall, the discourse became more qualified and acknowledged differences depending on paths and context. Since 1985, aided by the Demographic and Health Surveys (DHS), evidence has continued to accumulate. At the same time, the policy agenda has progressively expanded to include the reciprocal effects of fertility on education, as well as intergenerational links between education and fertility. That is, high fertility is no longer the unique policy issue sustaining education-fertility research. Instead, this research is increasingly motivated by policy concerns related to women's educational attainment, the welfare of children, labor force quality, economic inequality, and social stratification. Along with these changes in substantive focus, research methods have also improved. For the most part, these advances address previously recognized but unresolved methodological issues of temporal order, endogeneity, contextual variation, and heterogeneity. As a side-effect however, these advances confirm the need to correct for potential biases before inferring causal connections. Therefore, they cast doubt on the reliability of earlier results that are based on less refined methods. Together, these substantive and methodological developments further complicate the set of links that require explanation, even as they suggest caution against hasty and broad generalizations.
OCR for page 291
--> The remainder of this chapter is organized as follows. The next section reviews alternative interpretations of education-fertility associations. The following four sections examine in turn what were referred to earlier as the four pillars of the discourse on education and fertility: policy agenda, theory, methodology, and empirical evidence. The final section presents a summary, conclusions, and recommendations for future research. Alternative Interpretations Of Education-Fertility Associations Correlation is a necessary but not sufficient criterion for causation. In the absence of other evidence, researchers finding a negative statistical association between education and fertility generally suspect one of several explanations: causation, heterogeneity, reverse causation, or endogenous association. As Table 10-2 suggests, these alternative interpretations differ along two key dimensions: (1) whether education is viewed as a choice and (2) whether education is viewed as having a causal influence on fertility. The first interpretation, referred to here as causal, emphasizes what Carter (this volume) terms "the productive aspect of education" and neglects individuals' active choices in determining the level and content of their educational outcomes. In this view, schooling is largely an exogenous influence that transforms individuals and confers some attribute that modifies fertility preferences and/or one's capacity to actualize preexisting preferences. A few researchers have extended the analysis to examine the specific benefits that are responsible for the effect of schooling on fertility. For instance, working in the context of South Africa, Thomas (this volume) concludes that "the impact of comprehension skills ... may be important in affecting family decision making." In a similar vein, Oliver (1997, cited by Glewwe, this volume) indicates that literacy but not numeracy contributes to reduce fertility. Conversely, one may recognize that individuals do make deliberate choices about their schooling, sometimes overcoming major obstacles.2 However, such schooling choices need not involve a conscious anticipation of fertility implications, nor do the distinctive characteristics of the educationally driven necessarily shape fertility outcomes. In essence, individuals decide how much, when, and what type of schooling to obtain, but later experience unexpected fertility consequences of these schooling choices. Although exposure to schooling is a choice, this interpretation is still causal in the sense that the schooling experience affects fertility independently. 2 Such obstacles exist wherever schooling opportunities are limited by a lack of schooling infrastructure, limited family resources, blacks during the apartheid regime, when "state policies actively or active discriminatory policies, as was the case for South African sought to limit the education opportunities of blacks" (Thomas, this volume, citing Samuel, 1990).
OCR for page 292
--> TABLE 10-2 Alternative interpretations of education-fertility associations. NOTES: E and F indicate education and fertility outcomes, respectively. Bracketed letters indicate that the outcome reflects largely individual choice.
OCR for page 293
--> A third possibility is that schooling in itself does not cause fertility outcomes. Rather, the educated differ from the noneducated in meaningful ways that have a bearing on their later fertility. Educated individuals may come from distinct social or family backgrounds, and this explains subsequent differences in fertility behavior (see Thomas, this volume). Educated individuals may also tend to have educated spouses with distinctive fertility preferences. Hence this assortative mating, rather individuals' education per se, may explain differences in fertility outcomes (see Basu, this volume). Should one correlate individual education with fertility without considering these distinctive characteristics, a spurious association will be observed. This association should disappear with adequate controls. In practice, however, it is difficult to control for all distinctive factors, whether because corresponding data were not collected or because some factors are inherently difficult to measure. Fourth, without data on the timing of fertility and schooling events, an equally tenable interpretation is that educational attainment results from, rather than causes, fertility. The initiation of childbearing might disrupt or interrupt schooling. Ultimately, therefore, late childbearers who successfully avoid the economic and social burdens of childbearing are more likely to achieve higher levels of schooling. This interpretation is referred to as reverse causation. A final possibility is endogeneity. In the fertility-schooling literature, parents' decisions about fertility and their investments in the human capital of individual children are simultaneous and involve a tradeoff—a choice of high fertility ipso facto implying a lesser amount of resources per child (Becker and Lewis, 1973; Hanushek, 1992). This dual decision is viewed as a rational choice influenced by the prevailing socioeconomic environment. In particular, Lloyd (1994) argues that four contextual factors are likely to be important in that respect: stage of economic development, the role of the state, the phase of the demographic transition, and the nature of the family system. In this context, an association between education and fertility does not mean that one affects the other in a causal sense, but simply that the two decisions are simultaneous and interdependent. Table 10-2 thus illustrates the possible ambiguity in the meaning of education-fertility associations. Depending on assumptions and evidence on the timing of fertility and schooling, randomness in the distribution of schooling outcomes, and the relatedness of schooling and fertility choices, one can infer causation, unobserved heterogeneity, reverse causation, or endogeneity. While a different typology may be offered, the key point here is that, a priori, any given association between education and fertility can be interpreted as having several equally plausible meanings. Without additional evidence, a fair arbitrage among these meanings is difficult. Furthermore, these various interpretations are not necessarily mutually exclusive. For example, Thomas' analysis (this volume) concludes that "a small part of the correlation [between maternal education and fertility] can be attributed to the role of the family and community resources, a larger part to the
OCR for page 294
--> role of husband's schooling, and part to the acquisition of cognitive skills. An indeterminate fraction of the correlation is associated with unobserved heterogeneity...." As this research indicates, the ultimate question therefore is not whether an association reflects a causal influence, heterogeneity, reverse causation, or endogenous choices. Rather, it is how much each of these processes contributes to the association. Such fine distinctions, it must be added, are not meant solely to satisfy researchers' penchant for complication, but in fact bear on the identification of effective policies in the areas of education and population. Thus far, this section has focused narrowly on the issue of interpreting the results of a single study. However, a full analysis must also summarize the diversity of findings in the education-fertility literature. Throughout the chapter, the first issue is referred to as a question of interpretation, while the second is termed a question of generalization. As indicated earlier, the objective of the discussion is not so much to reveal what education-fertility associations mean as it is to analyze why difficulties of interpretation and generalization have persisted over the years, despite considerable research attention. To answer this question, we examine historical changes in the four foundations of the scientific discourse on education and fertility noted earlier: policy agenda, theory, methodology, and empirical evidence. An Expanding Policy Agenda Research questions circumscribe findings. Therefore, a possible starting point in evaluating education-fertility findings is to explore the underlying policy agenda and its evolution over time. If one defines a research agenda as the set of policy questions that, at a given time, are deemed worthy of research attention, the trend has clearly been toward expansion. As shown in Table 10-1, the policy focus has shifted progressively from an exclusive concern for whether one's education affects one's fertility to a larger set of questions, including whether one's fertility affects one's educational attainment, whether parents' fertility affects their children's schooling outcomes, and whether the schooling experience of children affects their fertility. As noted earlier, this evolution in research themes reflects a growing concern of population policy with issues of women's status and educational attainment, child welfare, labor force quality, economic inequality, and social stratification. Prima-facie evidence for this agenda expansion can be found in a comparison of Cochrane's (1979) review and the agenda for the workshop that generated this volume. While posing the same question as the workshop (What is known about the connections between education and fertility?), Cochrane's assessment focused entirely on the effects of one's education on one's fertility, while the workshop explored a broader set of linkages.3 To be sure, this classic relationship still held center stage during the workshop, but a few presentations ad-
OCR for page 295
--> dressed the reverse relationship (see Fuller and Liang, this volume), as well as links across generations (see Montgomery and Lloyd, this volume). Additional evidence for this substantive expansion may be found in a comparison of publication dates for key studies or reviews. For instance, Cochrane's (1979) review, which focuses on the effects of one's education on one's fertility, precedes by 15 years a similar review of the effects of parents' fertility on child schooling (Lloyd, 1994). To date, no comparable review has been done for the effects of child schooling on parents' fertility, even though such analyses are warranted by rapid changes in schooling costs throughout the developing world. This remark aside, the fact of an expansion in research focus is hardly disputable. Granted that the research agenda has indeed broadened to incorporate reverse and intergenerational influences, what drove this expansion, and what are its ultimate implications for our understanding of the relationships between education and fertility? Shifts in research agenda result from intricate political processes that are difficult to reduce to a few isolated causes. Nonetheless, the following societal changes in developing countries and sociological changes within the field of demography are likely to have played important roles. One key societal change concerns women's rising levels of educational attainment. It makes little scientific sense to search for fertility effects on women's schooling in low-education societies where schooling precedes childbearing with little or no overlap in the life course of most women. Following recent gains in the duration of females' school enrollment in most developing countries, the overlap between schooling and childbearing years has increased, making it possible that the initiation of childbearing may compromise further schooling. This development should promote research on the socioeconomic consequences of adolescent fertility. In addition to trends in female schooling, ideological changes are also worth mentioning. To be considered worthy of research attention, a dependent variable must reflect a socially valued outcome. Therefore, to the extent that societies increasingly value women's education and their participation in the labor market, this should promote research on the determinants of female schooling.4 Along with these societal transformations, the contours of demography have expanded. While formal demography restricts analysis to the three cardinal processes of fertility, migration, and mortality, a wider range of outcomes (including schooling) may be studied under the umbrella of population studies 3 Because of this expansion, there is a growing distinction between ''education-fertility" research, which takes fertility as the dependent variable, and "fertility-schooling," in which education outcomes are the dependent variable. 4 A recent article by Presser (1997) also highlights such shifts in policy agenda, although the author argues that this new policy emphasis on issues of gender equity (evident, for instance, at the 1994 International Conference on Population and Development) has yet to translate into a strong feminist research agenda in demography.
OCR for page 296
--> (McNicoll, 1992; Preston, 1993). Therefore, as the distinction between formal demography and population studies has blurred, research on schooling increasingly becomes a legitimate area of inquiry within demography. This rapprochement of formal demography and other social sciences is likely to continue and remain beneficial for both camps, especially in education-fertility research. In that symbiosis, social sciences improve demographic analyses of education's effects on fertility by clarifying the meaning and measurement of education, drawing attention to important control variables, or contributing complementary theoretical perspectives or special methods of data collection or analysis (Bogue, 1993; Crimmins, 1993; Carter, this volume). While the social sciences complement demographic analysis, formal demography can reciprocate by contributing powerful tools to the analysis of schooling outcomes. In particular, life-table methods may be usefully applied to research on educational attainment. Such applications are likely to grow in popularity when longitudinal data on schooling become available. In summary, three main factors contributed to an expansion of the policy agenda sustaining education-fertility research, in a direction that gives greater emphasis to the effects of fertility on schooling both within and across generations. These three factors are the rising levels of female enrollment, changing ideological climates in developing countries, and the blurring of the distinction between formal demography and population studies. A Theoretical Lag If one overlooks nuances among empirical generalizations, frameworks, principles, and formal theories, a theory can be defined as a "systematic explanation for observed facts" (Babbie, 1989:46). Generally, theories are expected to play both predictive and explanatory/organizing roles in the course of normal science. In their predictive role, theories focus empirical investigations by generating relevant and testable hypotheses. In their explanatory role, they help organize findings from different studies into a cumulative body of knowledge. Fulfilling this second role, however, requires that the scope and precision of theory exceed the scope and precision of findings (Kuhn, 1970). Unfortunately, this requirement is not currently met in education-fertility research, in which empirical findings have clearly outgrown the scope and precision of existing theories. With regard to scope, empirical studies have increasingly covered a wider variety of settings and highlighted the contextual dependency of the education-fertility relationship, whether this relation is examined within or across generations (see Diamond et al., this volume; Jejeebhoy, 1995; Lloyd, 1994). However, existing theories do not typically consider, let alone explain, these contextual variations. Implicitly the (questionable) assumption is that the content and meaning of education vary little across countries (see Glewwe and Carter, this volume). With regard to precision, empirical studies yield quantitative estimates of
OCR for page 297
--> education's effects, specifying when those effects are nonlinear or interactive, and in fact showing that their size and functional form vary substantially with the socioeconomic context and the aggregate educational level of a society (see Diamond et al., this volume; Singh and Casterline, 1985). In contrast, the theories that are supposed to explain these precise, quantitative, interactive, and context-dependent findings afford only hypotheses that are vague, verbal, monotonic, and context-invariant (Burch, 1996; Eloundou-Enyegue and Stokes, 1997). Such limitations in the scope and precision of theories restrict the extent to which one can integrate findings from different studies and build a cumulative body of knowledge, unless results happen to be uniform or follow an obvious pattern. When such uniformity is lacking, better theory is needed to reconcile seemingly disparate findings, or alternatively to refute ostensible similarities. By separating analysis of the effects of education into effects on demand, supply, and fertility regulation, various frameworks may explain why the effects of education on fertility outcomes are nonlinear. However, the effects of education on these proximate factors are themselves likely to be nonlinear and context dependent, and such variations also require explanation. For instance, discussions at the workshop outlined the need to investigate how education-fertility relations may depend on the stage of development, educational transition, or fertility transition. Context dependency is also important in studying the effects of fertility on schooling, both within and across generations. For instance, the quantity-quality tradeoff provides a common argument for expecting an inverse relationship between sibship size and child schooling (Becker, 1981; Kaplan, 1994). However, the nature of this tradeoff seems to vary across contexts, and the patterns of these variations need to be understood. A meta-analysis of existing studies suggests that the tradeoff depends on the level of development, state policies, the culture of the family, and the phase of the demographic transition (Lloyd, 1994). Although these empirical generalizations represent a major advance, the ultimate step is to develop contextual theories that can generate quantitative hypotheses on the effects of specific contextual variables on the quantity-quality tradeoff and optimal fertility and parental investment choices (Eloundou-Enyegue and Stokes, 1997). In sum, regardless of the particular linkage investigated, theoretical arguments exist for expecting a negative relationship between education and fertility. The problem, however, is that these arguments are not as precise in their predictions or as broad in their scope as the evidence they are required to organize. While this limitation is perhaps innocuous in interpreting the findings of a single study, it constrains the ability to reconcile findings from different settings. The main challenge, therefore, is to develop quantitative and contextual theories that will enable quantitative predictions regarding the mutual effects of education and fertility while recognizing the influence of socioeconomic context.
OCR for page 298
--> Methodological Advances Classic definitions of methodology emphasize its technical dimension, that is, the tools, rules, and procedures used to gather and analyze evidence. The dominant methodologies in education-fertility research include survey and ethnographic methods, each of which comprises distinct procedures for selecting and interrogating informants, as well as compiling and presenting the resulting information. However, methodology may also be viewed in a broader sense that includes the social organization of research. The present review adopts the latter perspective, and covers both technical and sociological changes in research practice. Technical Changes During the last three decades, the technical aspects of demographic research have markedly improved. Major advances concern both research tools and techniques, that is, both the hardware and software of demographic research. A major hardware change concerns the advent of computers and their extensive use in the analysis of demographic data. That computers have greatly facilitated data storage, retrieval, merging, computing, multivariate analysis, and reanalysis needs no elaboration. Demographers have also noted steady improvements in the statistical arsenal available for demographic analysis. Looking just at survey data, the commonly used tools have evolved progressively from simple correlations, to multiple regression, to path analysis, to contextual and dynamic modeling (Teachman et al., 1993). These statistical advances have facilitated the empirical distinction between correlation and causation in education-fertility research. Multivariate analysis enables extensive statistical controls, reducing the confounding effects of education correlates. Controls are further extended to unmeasured variables through the specification of fixed community or family effects, while other methods are used to address endogeneity (Bollen et al., 1995). Finally, longitudinal analyses consider the timing of fertility and schooling events, making it less likely that a spurious association will receive a causal interpretation. While pervading all social sciences, these changes, especially the development of event-history methods, have profound implications for the study of education-fertility relationships, which involve two intrinsically dynamic, cumulative, and endogenous processes. In particular, the practice of using simple associations to draw causal inferences has become less acceptable. The change, it must be stressed, has been technical rather than epistemological: like current researchers, social scientists in the 1970s knew the distinction between correlation and causation. Lacking then, however, were the technical tools for choosing between competing interpretations of an observed correlation. Without these tools, researchers could more easily drift from (legitimately) considering causal-
OCR for page 299
--> ity as a possible interpretation to (abusively) retaining it as the final explanation. This drift may have been reinforced by an ideological belief in the efficacy of education (Carter, this volume) and by processes of "interactive myth-making"5 through which researchers validate the thesis of a causal influence on the basis not of critical reviews of evidence, but the pronouncements of colleagues, themselves influenced by other colleagues' opinions. At the same time, it is important to note that despite the availability of improved statistical tools, actual practices still lag behind cutting-edge methodology. For instance, the lack of schooling histories in most major data sets still limits the application of event-history techniques in studies of educational attainment. Likewise, little research on the effects of education on fertility in developing countries has adequately addressed endogeneity issues. Existing research contains "serious methodological problems that could invalidate the results" (Ainsworth et al., 1996). Overall, as noted earlier, the unfortunate consequence of the higher methodological standards has been to breed dissatisfaction with much of the previous evidence that was derived through less refined methods. The Social Organization of Demographic Research Since the 1970s, the makeup of the demographic research community has evolved substantially, notably with increasing numbers of women and researchers from developing countries. In themselves, such changes may have little influence on methodological practices and findings (Watkins, 1993). Perhaps more influential is the increasing bureaucratization of demographic research. Manifestations of this bureaucratization include the concentration of research production within a few institutions and a division of labor that has increasingly separated data collection from analysis. The first of these patterns is noted in the production of major research articles (Teachman et al., 1993),6 as well as in the production of major data sets: DHS and World Bank data sets have thus far supported most large-scale studies on education and fertility in developing countries. And while 1970s researchers frequently collected and analyzed their own data, today's researchers are often spared the effort of designing questionnaires, constructing samples, training interviewers, and interacting with respondents. 5 The term is used by Caldwell et al. (1987) in a paper addressing the mutual reinforcement of speculation and research in generating "knowledge." They suggest, for instance, that the disciplines of anthropology and demography often feed on each other, each accepting ideas from other disciplines more readily and less critically than it would accept ideas emerging from its own members. Something similar could be said about researchers' inclination to accept conclusions from colleagues even when aware of the limitations of the data. Often, the assumption may be that others must know something that one does not. 6 As of 1992, for instance, more than 50 percent of all papers published in Demography were contributed by fewer than 10 percent of all organizations (Teachman et al., 1993).
OCR for page 300
--> Instead, they can concentrate on the ultimate phase of data analysis. Although this division of scientific labor presumably improves the efficiency of each activity in the research chain, it presents new challenges to researchers attempting to interpret data they did not collect or data originating from unfamiliar settings. Mounting Evidence Because of the methodological improvements noted earlier, raw empirical evidence has accumulated, mostly in the form of correlations, regression coefficients, and odds ratios depicting micro-level associations between education and fertility. Beyond the sheer number of case studies, geographical coverage has improved markedly thanks to the implementation of WFS and DHS surveys throughout the developing world.7 However, although empirical evidence is necessary to advance scientific knowledge, neither clarity nor consensus necessarily results from the mere accumulation of findings. This is especially the case when findings from individual studies do not warrant unambiguous interpretations or when findings appear to be inconsistent from study to study. To give meaning to raw findings, researchers must deal with problems of interpretation and generalization. As noted earlier, the term interpretation as used here refers to the meaning ascribed to the findings of a single study, while generalization denotes the overall picture framed by compiling all available evidence. The challenge of the workshop was precisely to reconcile the variety and richness of existing findings into a coherent, concise, and accurate summary. As was argued earlier, methodological advances have improved researchers' capacity to arbitrate between alternative interpretations of observed associations. On the other hand, the reconciliation of findings from different settings has become increasingly difficult because of the geographic expansion of education-fertility research, the multiplication of substantive links of interest, and the lack of contextual theories. The expansion of geographic coverage has restricted broad generalizations by providing hard evidence of atypical patterns. On the other hand, the increasing policy attention to intergenerational links, to reverse effects, and to paths of influence has made it more difficult to summarize the complexity of the education-fertility relationship with a concise statement. Increasingly, qualifications are needed as to which path/link is being examined. Finally, the lack of contextual theories limits understanding of observed contextual variations. In the absence of such theories, researchers have resorted to meta-analysis to derive empirical generalizations (see Diamond et al., this volume; Jejeebhoy, 1995; Lloyd, 1994). However, such analyses are problematic 7 As of 1997, about 42 WFS and 50 DHS surveys had been implemented in developing countries (Presser, 1997).
OCR for page 301
--> when they mix studies of varying methodological soundness (see Diamond et al. and Glewwe, this volume). Overall, as a result of these trends, summary statements on education-fertility relationships have tended to become more qualified and less definitive even as the volume of empirical evidence has expanded. In contrast with these difficulties in generalization, methodological advances have progressively reduced the difficulties involved in interpreting the results of a single study. Yet even in this area, problems remain with concept clarification and data. First, greater attention needs to be paid to the local meaning and content of education. As the chapters by Carter, Diamond et al., and Glewwe in this volume suggest, little is known about the skills, knowledge, and attitudes pupils acquire in school and how they vary across different school institutions. Also, while dynamic modeling is widely advocated in education-fertility research, the major data sets that support these analyses do not contain schooling histories and thus do not allow the application of cutting-edge methodology (Lloyd and Blanc, 1996; Knodel and Jones, 1996). Overall, however, these difficulties of definition and data represent minor impediments, if only because their remedies are known and manageable. Certainly more daunting problems reside in explaining contextual variations. Summary, Conclusions, And Research Recommendations With a few exceptions, empirical studies around the world report negative statistical associations between education and fertility, both within and across generations. However, the meaning of these associations remains elusive and may in fact have become more so over the years. At issue are the twin difficulties of interpretation and generalization. Interpretation difficulties concern individual studies and the basis on which researchers determine whether an association between education and fertility reflects a causal influence, heterogeneity, reverse causation, or the endogeneity of schooling and fertility choices. Generalization difficulties concern how to reconcile findings from different studies. To date, both difficulties continue to plague education-fertility research despite policy interest in reliable interpretations, and despite the progress made since the publication of Cochrane's (1979) seminal work. The purpose of this chapter has been twofold. First, it has reviewed some of the difficulties involved in ascribing meaning to education-fertility associations. More important, it has attempted to explain why these difficulties have persisted in spite of considerable research effort over the last three decades. The chapter's main argument has been that the enduring uncertainty about the meaning of education-fertility associations does not indicate a lack of scientific progress. Indeed, much progress has been achieved on all four pillars that sustain the education-fertility discourse: policy agenda, methodology, empirical evidence, and theories.
OCR for page 302
--> In particular, the policy issues underlying education-fertility research have expanded from an almost exclusive focus on high fertility to concern for women's educational attainment, children's welfare, labor force quality, and economic inequality. This expanding agenda has drawn attention to the multiple facets of the education-fertility relationship, including reverse and cross-generational effects, the paths through which education affects fertility, as well as to the particular features of the schooling experience that are most relevant to these effects. Over the same period, substantial methodological improvements have occurred, whether they concern the technical tools of research—the greater use of computers in demographic research, the development of statistical techniques to address problems of statistical control, unobserved heterogeneity, endogeneity, and the study of dynamic processes—or the social organization of demographic research, with a growing concentration of data and research production within a few institutions and an increasing separation of data collection and analysis activities. Finally, the implementation of large-scale surveys across the developing world has generated a voluminous database that permits extensive analyses and cross-country comparisons. Yet, for all these methodological improvements, the scientific discourse on education-fertility relationships has remained surprisingly tentative. My interpretation is that this does not reflect an overall stagnation, but the fact that progress has been uneven. While advances have been achieved on the four pillars that sustain the education-fertility discourse, this progress has followed an odd sequence in which theory trails rather than leads empirical investigations, and in which methodological standards have surpassed the actual practices permitted by most existing data sets: while methods have drastically improved, the associated rise in methodological standards has bred dissatisfaction with previous findings that were based on less refined methods. While the specialization of data collection and analysis activities presumably improved the output efficiency of the demographic research industry, researchers now face the challenge of interpreting data that they did not collect or data that originated from unfamiliar settings. While the number, the geographical coverage, and the precision of empirical studies improved, researchers' ability to reconcile findings from different studies is restricted by the lack of quantitative, contextual theories that generate precise predictions about the expected contextual variations in education-fertility relationships. The ultimate outcome of this asynchronous evolution is that even though data have become more available and research tools have sharpened, the conclusions derived from existing evidence have become increasingly qualified and tentative. A number of recommendations for future research flow from this diagnostic. One concerns data collection. Longitudinal data on the fertility and education experiences of both parents and children are needed. While fertility surveys contain fertility histories, similar historical data are not available for schooling,
OCR for page 303
--> and cross-country differences in patterns of grade progression make it impossible to reconstruct schooling histories from available information. Where longitudinal surveys cannot be carried out, a retrospective reconstruction of schooling histories can provide a reasonable substitute. Such reconstruction can be facilitated by the design and use of life-history calendars and the combined use of ethnographic methods and survey techniques. A second recommendation concerns the distribution of research effort across different facets of the education-fertility relationship. While the effort has become somewhat diversified, most of the emphasis remains on the effects of one's education on one's fertility. Even within this agenda, more studies should examine the specific schooling benefits that matter most to fertility outcomes. There is also a need for studies on the reverse linkage—the consequences of fertility for the educational attainment and other socioeconomic outcomes of young women. This research could build on the methodological experience of similar research in developed countries. Likewise, research must reexamine the effects of high fertility on schooling at the family level within a dynamic framework that acknowledges time variations in sibship size and family context, as well as unobserved heterogeneity in family background. Finally, very little research has examined the consequences of child schooling for parents' fertility.8 This focus is particularly warranted by the current changes in schooling costs and returns in developing countries under structural adjustment. A third broad class of recommendations concerns theory. Given the current state of methodological development, and assuming that longitudinal data become more available, theory is likely to become the key limiting factor. Without the guidance provided by good theory, empirical investigations are likely to proceed haphazardly, and generalizations will remain unduly constrained by the observation of contextual differences. The greatest theoretical challenge is to anticipate contextual variations in education-fertility relationships. This includes, for instance, variations across fertility transition stages and across educational development stages. Meta-analyses such as those attempted by Jejeebhoy (1995) and Diamond et al. (this volume) for contextual variations in education effects and by Lloyd (1994) for fertility effects on child schooling are steps in the appropriate direction. Ideally, however, such steps would be complemented by truly quantitative and contextual theories (see, e.g., Burch, 1996). Without significant theoretical developments, current methodological advances may not be put to their best use, making it likely that a decade from now, Cochrane's nagging question will resurface. 8 See Axinn (1993) for an exception.
OCR for page 304
--> Acknowledgments I wish to thank the editors of this volume, workshop participants, and the anonymous reviewers, as well as Julie DaVanzo and Shannon Stokes, for their insights or comments. Remaining errors are my own. References Ainsworth, M., K. Beegle, and A. Nyamete 1996 The impact of women's human capital on fertility and contraceptive use: A study of fourteen sub-Saharan countries. World Bank Research Observer 10(1 ):85-122. Axinn, W.G. 1993 The effects of children's schooling on fertility limitation. Population Studies 47(3):481493. Babbie. E. 1989 The Practice of Social Research. Belmont, Calif.: Wadsworth Publishing Company. Becker, G., and H. Lewis 1973 On the interaction between the quantity and quality of children. Journal of Political Economy 81(2):s279-s288. Becker, G.S 1981 A Treatise on the Family. Cambridge, Mass.: Harvard University Press. Bogue, Donald J. 1993 How demography was born. Demography 30(4):519-532. Bollen, K.A., D.K. Guilkey, and T.A. Mroz 1995 Binary outcomes and endogenous explanatory variables: tests and solutions with an application to the demand for contraceptive use in Tunisia. Demography 2(1): 111-131. Burch, T.K. 1996 Icons, strawmen and precision: Reflection on demographic theories of fertility decline. The Sociological Quarterly 37( ):59-81. Caldwell, J., P. Caldwell, and B. Caldwell 1987 Anthropology and demography: The mutual reinforcement of speculation and research. Current Anthropology 28( ):25-43. Cochrane, S.H. 1979 Fertility and Education. What Do We Really Know? Baltimore, Md.: The Johns Hopkins University Press. Crimmins, E.M. 1993 Demography: The past 30 years, the present and the future. Demography 30(4):579-591. Eloundou-Enyegue, P., and C.S. Stokes 1997 Davis & Blake and Becker in Discrete-Time: From Verbal To Quantitative Theories Of Fertility. Paper presented at the 1997 Annual Meeting of the Population Association of America. Washington D.C., March 27-29. Hanushek, E. 1992 The trade-off between child quantity and quality. Journal of Political Economy 100(1 ):84117. Jejeebhoy, S.J. 1995 Women's Education, Autonomy and Reproductive Behavior: Experience from Developing Countries. Oxford: Clarendon Press. Kaplan, H. 1994 Evolutionary and wealth flows theories of fertility. Population and Development Review 20(4)753-791.
OCR for page 305
--> Knodel, J., and G.W. Jones 1996 Does promoting girls' schooling miss the mark? Population and Development Review 22(4):683-702. Kuhn, T.S. 1970 The Structure of Scientific Revolutions. Chicago, Ill.: University of Chicago Press. Lloyd, C.B. 1994 Investing in the Next Generation: The Implications of High Fertility at the Level of the Family. Working Paper No. 63. New York: The Population Council. Lloyd, C.B., and A.K. Blanc 1996 Children's schooling in sub-Saharan Africa. Population and Development Review 22(2):265-298. McNicoll, G. 1992 The agenda of population studies: A commentary and complaint. Population and Development Review 18(3):399-420. 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. Presser, H.B. 1997 Demography, feminism and the science-policy nexus. Population and Development Review 23(2):295-331. Preston, S.H. 1993 The contours of demography: Estimates and projections. Demography 30(4)593-606. 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. Singh, S., and J. Casterline 1985 The socioeconomic determinants of fertility. Pp. 199-222 in J. Cleland and J. Hobcraft, eds., Reproductive Change in Developing Countries: Insights from the World Fertility Survey. London: Oxford University Press. Teachman, J.D., K. Paasch, and K.P. Carver 1993 Thirty years of demography. Demography 30(4):523-532. Watkins, S.C. 1993 If all we knew about women was what we read in Demography, what would we know? Demography 30(4):551-578.
OCR for page 306
This page in the original is blank.
Representative terms from entire chapter: