Click for next page ( 52


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 51
51 2 . 2 THE PROBLEM m e primary goal of this chapter is to present the results of exploratory macro analyses in which the estimated intercepts and micro parameters for the socioeconomic variables in the onset (AFB), early fertility (EF), and later fertility (LF) structural equations are treated as functions of macro variables. This focus reflects two basic decisions: to select certain structural equations for examination; and, within these equations, to examine macro variability in the intercepts and in the effects of the socioeconomic variables. m e reason we restrict our attention to a subset of the entire model is that our empirical research has just begun. The results reported in this chapter are thus in the nature of a progress report. We concentrate on the AFB, EF, and LF equations because the three endogenous variables in these equations operationalize the decomposition of fertility which is at the heart of the theory presented in Chapter 1. Within these equations, we focus on the effects of the socioeconomic variables because the theory is concerned with the socioeconomic deter- minants of fertility. Hypotheses presented in Chapter 1 about variability in the effects of these variables are novel, and their empirical testing is a priority research concern. Analysis of the remaining variables, although essential to a complete evaluation of the model, must await another occasion. Finally, our focus on the AFB, EF, and LF structural equations provides insight into the main or additive effects of the macro variables in the determination of AFB, EF, and LF, as discussed in further detail below. 2 . 3 MACRO HYPOTHESE S Chapter 1 examined possible changes in the directions of association between socioeconomic and fertility variables as a function of location on the traditional/transitional continuum. In that discussion of the implications of the theory for commonly used endogenous variables such as children ever born (CEB) , we treated sign changes from traditional to transitional settings as indicators of changes in magnitude as well as in sign. Thus, the coefficient of a particular socioeconomic variable which is positive in an extreme traditional setting but negative in an extreme transitional setting may be negative and smaller in absolute value, or even zero, in a less transitional setting. In developing the macro hypotheses to be operat~onal~zed and tested, we must first be more precise about the sources of macro variability in the micro parameters. Next, we develop a macro specification, and translate our notions of coefficient variability into hypotheses about the effects of macro variables on micro parameters. Subsequently, we operationalize the micro and macro formulations. The traditional/transitional continuum is based on the notion that at the traditional extreme there is no explicit idea or practice of fertility control, while in transitional settings, some explicit idea and practice exist. In general, to operatzonalize this continuum would seem to involve measuring knowledge, attitudes, and practices concerning fertility control at both the micro and macro levels. For a given

OCR for page 51
52 country, a time series of this kind of information would provide a profile of a society's degree of transitionality over time.4 However, such an operationalization is currently beyond our grasp. It may nevertheless be possible to measure a society's degree of transitionality indirectly. For one thing, indicators of socioeconomic development should be related to the continuum: the greater the develop- ment, the further along the transitional end of the continuum. This assertion is based on the idea that socioeconomic development, broadly construed, increases the relevance of family-size targets and birth limitation, and thus of fertility control. It does so by changing the social and familial environment in which the advantages and disadvantages of children are determined, although not necessarily equally or evenly throughout a society (Leibenstein, 1974; Caldwell, 1976; Bulatao, 1979; Notestein, 1953; Coale, 1973; Freedman, 1975~. Although this relationship need not be monotonic for all phases of socioeconomic development, the least socioeconomically developed societies should be relatively tradi- tional and the most socioeconomically developed societies relatively transitional. Because macro data on socioeconomic development are relatively plentiful, it should be possible to study variability in the micro parameters by relating them to level of development. A second macro dimension relevant to the degree of transitzonality encompasses the presence, scope, and vigor of official family planning programs. Indeed, information on these programs should bear directly on knowledge and practice of fertility control in Third World countries. m e effectiveness of this influence is of course at issue (Tsui and Bog ue, 1978; Demeny, 1979~. For precisely this reason, it is of particular importance to examine the macro-level impact of these programs on vari- ability in the effects of the micro socioeconomic determinants of fer- tility, as well as in the intercepts of the micro structural equations. There are a number of mechanisms by which family planning programs can have an impact on fertility. For a variety of reasons, these mechanisms are not readily distinguished from each other empirically. Perhaps the chief difficulty is that such an analysis would require a large sample of micro data from countries located across virtually the entire range of socioeconomic development. That is, the effects of a program may well depend on the socioeconomic environment within which it is implemented. Unfortunately, the WFS data available reflect a relatively narrow range of development. Therefore, we will restrict our attention to a single mechanism underlying potential family planning program effects on the micro parameters of our model, as descr ibed below. A fundamental hypothesis about family planning programs is that over time they reduce socioeconomic differentials in fertility. m is hypothe- sis implies that in an ideal-typic traditional society, family planning programs will primarily benefit those of higher socioeconomic status, who, because of health and other advantages, are likely to have the highest fertility. (Programs in such settings are, of course, extremely unlikely.) In a transitional society, on the other hand, those of lower socioeconomic status will have higher fertility and so be prime program beneficiaries. Of the three fertility components outlined in Chapter 1, family planning programs are most likely to relate to the last--later fertility

OCR for page 51
53 (LF). There is no reason to suppose that these programs are presently focused directly on timing and spacing, factors relevant to age at first birth (AFB) and early fertility (EF). muS, the effect of family planning programs in suppressing socioeconomic differentials in fertility should be visible for the LF structural equation, but not for the other two equations. This point is supported by Table 2.1. This table extends Table 1.1 to show the sign hypotheses for socioeconomic effects in the AFB, EF, and LF structural equations ~ conditional on the extremes of socioeconomic development and family planning program strength. The left-hand column of the table, corresponding to settings with no family planning programs, displays the same pattern of signs found in Table 1.1. The right-hand column contains the extension of Table 1.1. In this column, the damping of socioeconomic differentials in conjunction with strong family planning programs occurs only for LF. Although other macro forces may be hypothesized to drive the micro socioeconomic parameters, we consider the two discussed above--socio- economic development and family planning program strength--to be on a Priori grounds worthy of particular consideration. The empirical research reported in this chapter will therefore depend on measurement of these two dimensions. If these dimensions affect the socioeconomic micro parameters of the AFB, EF, and LF structural equations, they do so differentially depending on the particular structural equation being considered. That is the force of the argument thus far advanced and summarized in Table 2.1. To trans- late these hypotheses into specific expectations capable of empirical TABLE 2.1 Expected Signs of Micro Relationships Between Socioeconomic Variables and Fertility Components in Particular Macro Settings Fertility Component Lowest Extreme of Socioeconomic Development No Program Very Strong Program AFE - EF ~ LF + Toward Zero Considerable Socioeconomic Development AFB EF LF + + Toward Zero

OCR for page 51
54 testing would require the specification of functions linking the micro parameters to the macro dimensions. The simplest functions compatible with our substantive micro and macro reasoning are as follows. Let Djkm be the effect in the jth (j = 1,...,J) country of the kth {k = 1,...,K) micro predictor on the mth (m = 1,...,M) endogenous variable. Since our analysis is focused only on the micro intercepts and the effects of the socioeconomic determinants on fertility, assume that k subscripts only the intercepts and socioeconomic variables, and not othe r variables potentially in a structural equation, such as adjustment variables. Furthermore, let m = 1 denote AFB, m = 2 denote EF, and m = 3 denote LF. Labeling socioeconomic development by S and family planning program strength by FP, the specification in its most general form becomes (2.1) ~jkm = nOkm + ~lkmSj + n2km(FP Id ~ n3km~s.Fp~i + Ajkm for k = 1~...,K; m = 1,2,3, where the e-coefficients are macro regres- sion coefficients, and the \-term denotes error in the determination of the micro coefficients.5 Thus, when ~ km = Djl:, the dependent variable is the intercept of the AFB micro equation, and observations are over countries. When Djkm = Djk3 for k > 1' the dependent variable is the coefficient of a socioeconomic variable in the LF micro equation, and observations are again over countries. Expression {2.1) specifies 3K macro equations. Our hypotheses about the effects of S and FP (i.e. the n ' s ~ in these equations can now be stated. Macro Determinants of Micro Slopes The reasoning in Chapter 1 about the signs of the effects of the socio- economic variables at the micro level leads to identical hypotheses about the signs of the effects of each of those variables in a given structural equation. For example, there should be identical sign variation across the traditional/trans~tional continuum for wife's and husband's education in the determination of LF. For this reason, in stating our expectations about the signs, but not the magnitudes, of the effects of S and FP in expression (2.1), it is unnecessary to distinguish among the socioeconomic variables (K > 1~. Where we expect the effects of S and FP to be nonzero, there are no hypotheses about magnitude. Hence, if we restrict our attention to the directions of the S and FP effects, it suffices to reduce expression (2.1) to just two equations for each micro structural equation--one for modeling the micro intercept, and another for modeling micro socioeconomic parameters. Table 2.2 shows our sign hypotheses about the effects of S and FP. This table contains three kinds of entries. Where we have a directional hypothesis, it lists the appropriate sign. Where we hypothesize a null effect, it lists either a zero or a ~*.. me n*n entries are for coeffi- cients which, if not zero, can be either positive or negative. For these entries there are no directional hypotheses, and no a priori interpreta- tion of a nonzero coefficient. These entries always pertain to FP and S.FP, variables which are relevant only for modeling the LF micro parameters. The empirical macro analysis reported in Section 2.6 below

OCR for page 51
55 TABLE 2.2 Expected Relationships Between Macro Variables and the Intercepts and Slope Coefficients of the Micro Equations Macro Coef f icient S Main Effect AFB Equation FF Main Effect S FP Intercept ~ Micro SES Effects + * * EF Equation Intercept + Micro SES Effects O * * LF Equation Intercept Micro SES Effects - Note: The asterisks indicate that there is no reason to expect a relationship. does not test the null hypothesis that the n-coeff icient associated with a particular n*. entry in Table 2.2 i ~ 7.-r~ baa i nod The ~ ~ ~ ~ ~ _ _ ~ _ ~ . . ~ _ . - + -em "~ernaClve. again' a nonzero result would require post hoc speculation, and inclusion of FP and S.FP in the macro equation amounts to specification error resulting from inclusion of irrelevant variables. Finally, where Table 2.2 lists a zero, we do inspect the data for a relationship. The reason for this, which will become clearer below, is that one of the sign hypotheses depends intimately on the validity of the hypothesized null effect. The sign hypotheses displayed in Table 2.2 are listed separately for the intercepts and socioeconomic ef feats of the AFB , EF , and LF struc- tural equations. As for the hypothesized signs of the macro variables on the micro socioeconomic effects, these can be deduced immediately from Table 2.1. Consider the AFB equation. The hypothesis that the socioeco- nomic determinants of AFB will be negative in traditional settings and positive in transitional settings regardless of family planning program strength translates into ~ positive effect of S on 8.kl for k > 1. The hypothesis of no impact of fa'.~:Ly planning program strength on the directionality and magnitude of the effects of socioeconomic status on

OCR for page 51
56 AFB translates into a null effect of both FP and S.FP. For the reasons discussed above, we do not test hypotheses about these null effects, and Table 2.2 reflects this with arm entries. The hypothesis that socioeco- nomic differentials in EF are independent of any variability along the traditional/transitional continuum leads to the hypothesis that the effects of S. FP, and S.FP are all zero. In this instance, the FP and S.EP entries contain Ares for the reasons discussed above, but the S entry contains a zero. The explanation is that the macro hypothesis about the dependence of the EF intercept on S rests in part on the assumption that the effect of S on the socioeconomic coefficients in the EF equation is nil. This assumption is examined below in Section 2.6. It is only in the macro equations determining the levels of the socioeconomic differentials in LF that we expect to see effects for S. FP, and S FP in expression (2.11. The implications of the argument for the joint effect of S and FP can be determined by taking partial deriva- ti~res of expression (2.1~. With no loss of generality, let k = 2. Then (2 .2a) and (2.2b) a523 aS = "123 + Dl23FP a623 aFp = n223 +n323S. We can determine the signs of nl23 and q323 from expression (2.2a) as f allows. Taking FP = 0 to indicate total lack of family planning program effort, the derivative must be negative (as suggested by the left-hand column of Table 2.1~. Taking FP at a positive value indicating a high degree of family planning program effort, the derivative must not be positive, and must be closer or equal to zero. Thus, nl23 0, and the appropriate signs for these coefficients are accordingly entered into the bottom panel of Table 2.2. This leaves the sign of the "additive" component of FP to be deter- mined. In expression (2.2b), we know that n323 > 0 according to our hypotheses. Taking S = 0 to indicate the lowest extreme of socioeconomic development, it is clear that the derivative must be negative at that point, based on The reasoning that at the lowest levels of socioeconomic development, the introduction of family planning programs will lead to a damping of positive socioeconomic effects on fertility. At the other extreme, taking S at a high positive value indicating considerable socio- economic development, expression (2.2b) should be positive. This implies that n223 must be negative and Table 2.2 thus indicates a negative ~additive" effect of FP in the bottom panel. Macro Determinants of Micro Intercepts In considering the directionality of the effects of S and FP on the intercepts of the AFB, EF, and LF equations' we completely characterize the hypothesized macro variability of all of the micro parameters in the AFB structural equation. Since we have no hypothesis about setting

OCR for page 51
57 variability in the effect of AFB on EF, we also characterize all of the hypothesized macro variability in the micro parameters of the EF struc- tural equation. This is not true, however, for the LF structural equation. For this equation, there are predetermined variables whose effects we are not studying, some of which interact with the micro socioeconomic factors. To complicate matters further, a number of the micro effects not included in the present analyses are hypothesized to vary systematically over the traditional/transitional continuum. To the extent that our hypotheses about these effects are incorrect, our expecta- tions about macro variability in the LF intercept will be subject to distortion. For purposes of the present analyses, then, our expectations about the impact of S and FP on the LF intercept are less certain than those about the AFB and EF intercepts. In our later research, we will be examining all of the coefficients in the LF equation. Outside of the analysis of covariance, the intercept is perhaps the most neglected quantity in analyses of generalized linear models (e.g., regression and logistic response models). Nevertheless, an understanding of macro variability in the micro intercepts is of critical importance for understanding the main effects of the macro variables on the micro endogenous variables. This point is disguised as long as we think of micro equations on one hand, and macro equations on the other. However, it can be revealed by the following simple example. With no loss of generality, suppose the micro model is bivariate and can be written as a single regression equation: Yii Boj + BljXij + ij, (i = 1,...,nj; j = 1,...,J), where i subscripts individuals in countries and j subscripts countries. Then the macro model consists of two equations. For them, we suppose, again with no loss of generality, that there is only one macro explana- tory variable, and that it is the same for both equations: ~Oj = nOO ~ nQlGj + loj; clj = ~10 + DllGj + Fiji Then substituting for 60j and Dlj in the micro equation yields (2.3) Yij = DOG + n01Gij + DloXi; + n~lKijGii + (error)ij, where Gij = G for i = 1,...,nj and "error. denotes a complex error structure that plays no role in the derivations below. This combined micro~macro nKxlel shows that the effects of the micro and macro variables are inseparable. The effect of G on Y is given by (2.4a) aG = 001 + nllX' while the effect of X on Y is given by (2.4b) aX n1O + nllG.

OCR for page 51
58 As can be seen, it is the macro equation for the intercept that contri- butes the main effect of the macro variable e As noted above, this result generalizes to any number of micro and macro predictors. In the linear model, used to estimate the AFB, EF, and LF structural equations, the intercept is a linear combination of the mean of the response variable and the effects of the predictors multiplied by the means of the predictors (the DX terms). Therefore, hypotheses about the dependence of the intercepts of the micro structural equations on S and FP must depend on the following: hypotheses about how the means of the response variables vary as a function of S and FP; hypotheses about the dependence of the micro coefficients on S and FP; hypotheses about the dependence of the X terms on S and EP; and hypotheses about the dominance of variability in the means of the response variables over variability in the OX terms, or vice versa. AS summarized in Table 2.2 for the AFB intercept, we hypothesize that S will have a positive effect, FP will have no effect, and there will be no interaction between S and FP. For the Third World countries sampled by the WFS, there should be a positive relationship between AFB and socioeconomic development. Since we have already hypothesized a positive effect of S on the socioeconomic determinants of AFB, we expect that the greater the AF8 mean, the greater the quantity resulting from micro socioeconomic effects that will be removed (the means of the micro socioeconomic variables should increase as socioeconomic development increases). If this reasoning is correct, the relationship between the AFB intercepts and S should be attenuated relative to the relationship between the AFB means and S. However, we do not expect it to become negative, since if X is socioeconomic status and G is socioeconomic development in expression (2.4a), this partial derivative should be nonnegative at all levels of micro socioeconomic status. The hypothesis about the relationship between the AFB intercept and socioeconomic development is strengthened by recalling that this relationship also describes the "mains effect of socioeconomic development in the combined micr~macro model (2.3 ~ . There is no reason to expect that FP will have an ef feet on the AFB intercept, both because none is hypothesized for macro variability in the socioeconomic determinants of AFB and because there is no reason to suppose that FP should affect the AFB means. Family planning programs have not, in general, been aimed at this component of fertility in Third World countries, particularly for the years of concern to us, as will become clear when we define the cohorts to be studied. For the EF intercept, we expect S to have a positive impact at the macro level. Since we hypothesize no effects of S. FP, and S.FP on the micro socioeconomic effects in the EF structural equation, it follows that hypotheses about the nature of macro variability in the EF intercept must depend on the following: expectations about between-country vari- ability in the EF means {(EF)j); expectations about the dependence of mean AFB {AFB) on S and of mean micro socioeconomic status (SES) on S.; and expectations about the dominance of the AFB effect relative to the SES effects on EF at the micro level. In general, we_have no compelling hypothesis about the dependence of EF on S. SES and AFB should be positively correlated with S. even if the micro effects of SES and AFB do

OCR for page 51
Be not depend on s.6 However, the micro effects of SES and AFB are offsetting--one hypothesized to be positive, the other negative. Further- more, the effect of exposure (AFB) should dominate the SES effect, regardless of level of socioeconomic development. Putting all of these expectations together, it follows that, since the effect of AFB on EF should be negative and dominate that of SES, and since AFB should be positively correlated with S. the effect of S on the EF intercept should be positive. There is no strong reason to expect a relationship between the EF means and FP because family planning programs in the Third World have not been targeted at, or markets successful in slowing, early fertility. Although this is less likely to be the case for recent cohorts, it seems reasonable for the older cohorts-women aged 40 to 44 in 1974-on which we focus. m us, we expect a systematic' positive effect of S on the EF intercept, and no relationship for FP or S.FP. The middle panel of Table 2.2 summarizes this conclusion. Turning now to hypotheses about the effects of socioeconomic development and family planning strength on the LF intercepts, as noted earlier, these hypotheses are the most uncertain because the present discussion focuses only on the socioeconomic determinants of LF and not the full LF equation. To derive hypotheses about the macro determinants of the LF intercept, it suffices to define the micro and macro models as (2.5a) (LF) id = DOj + ~lj (SES) ij + ~ ij ; fib) Do; Too ~ nOlSj + D02(EP); + no3(s.Fp~i + Aoj; and (2.5c) ~lj = D10 + Ells; + nl2(EP); + nl3(s-Fp~i + ll Then, substituting for the as gives (2.5d) (LF)i; nOO + nOlSi; + ~02 (FP)i; + no3 (S.FP)i + n 10 (SES) i j + n 11 (S.SES) i j + n 12 (FP.SES) i + nl3(FP.S.SES)~i ~ (error) ij' where Sij = Sj, i = 1,...,nj; that is, the 5 values are identical within countries, as they are for the other terms involving macro variables. There are essentially two ways to attempt to derive hypotheses about n01, no2, and nO3. One is to reason directly about equation {2055~ ; the other is to recognize that these coefficients are the Main effects" of the macro variables `n the combined micro-macro equation (2.5d), and then to derive the signs of these coefficients from inspection of the partial derivatives. We will take the latter course here since it explicitly uses information discussed earlier about the effects of the macro variables on the micro socioeconomic effects. Regardless of which alternative is selected, the inclusion of S.FP in equation (2.5b) must be justified. Actually, this justif ication is neither more nor less than that required in support of including S and FP in the micro-macro model, equation

OCR for page 51
60 (2.5d). The reason is that, since the micro socioeconomic effects on LF are held to vary as a function of S. FP, and S.FP, and since this vari- ability is necessarily incorporated into the micro intercepts as well, we must allow for the dependence of the intercepts on S. FP, and S.FP. Operating on equation (2.5d), and taking first partial derivates with respect to S and FP, gives (2.6a) a(LF) = nO1 + nO3FP + nlltsEs) + nl3(FP.SES) and (garb) a {LF) = ~02 ~ DOSS + nl:{SES) + nl3`S.SES,, where already derived signs have been placed above the appropriate coefficients. Examining expression (2.6a) first, and setting FP = 0 to indicate total absence of family planning program effort, we see that the partial derivative of LF with respect to S depends critically on Sol. It is at the heart of the theory presented in Chapter 1 that this coeffi- cient be negative. If socioeconomic development leads to a reduction in fertility, it should certainly be seen in later fertility. A similar point holds for expression (2.6b). If family planning programs have an effect on fertility in developing nations, that effect should certainly be on LF. Thus, setting S = 0 to indicate the lowest level of socioeco- nomic development, it is clear that if the partial derivative of LF with respect to FP is always to be negative, it must be the case that nO2 < 0. However, n02 = 0 is too weak as a hypothesis. The effect of FP on LF should not have to depend on the magnitude of the interaction effect between micro socioeconomic status and family planning program strength (n :21. Above and beyond this relationship, there should be a negative effect of FP on LF. Thus, it is hypothesized that ng2 ~ O The sign of nO3 can be derived either from (2.6a) or (2.6b), given that we have the signs for nO' and n02. Focusing on (2.6a) yields the general hypothesis that the nigher the level of socioeconomic development, the lower LF should be regardless of the level of family planning program effort. Fixing SES at any level and setting nO3 > 0 leads to the result that as FP increases, the effect of S on LF decreases in absolute value and may even become positive if FP is large enough. On the other hand, fixing SES at any level and setting n03 < 0 leads to offsetting compo- nents : n O,P + n 1/P.SES, where the first term is negative and the second positive. Of course, everything depends on the relative magnitudes of nO3 and n13; however, hypothesizing nO3 ~ ~ is the only alternative if (2.6a) is to have a chance of remaining constant or nonpositive as FP increases. We have now reviewed all of the directional hypotheses about the effects of socioeconomic development and family planning program strength on macro variability in the intercepts and micro socioeconomic parameters of the AFB, EF, and LF structural equations. Essentially, this has amounted to an expansion and translation of the reasoning embodied in Table 1.1 of Chapter 1. In addition, the discussion of macro variability

OCR for page 51
61 in the intercept of the EF structural equation was based on another aspect of the theory developed in Chapter 1: expectations about the hypotheses that socioeconomic effects on EF are weak, relative to both the effect of "exposure" on EF and the effects of the socioeconomic variables on other fertility components. Of the three components of the fertility process, it is posited that LF should be most susceptible to socioeconomic posi- tion, onset less so, and early fertility least. This hypothesis is, of course, specific to the range of countries most likely to be encountered in the WFS, and is not meant to apply to the most industrialized or socioeconomically developed societies. Finally, this hypothesis points to another aspect of the theory developed In Chapter 1 which can be tested by means of replications over countries, as illustrated later in this chapter. 2.4 DATA AND VARIABLES EMPLOYED As noted earlier, the empirical analysis presented here is based on the 15 countries for which WTS data are currently available. These countries include Colombia, Costa Rica, Fiji, Guyana, Indonesia, Jamaica, Jordan, Kenya, Korea, Lesotho, Malaysia, Panama, Peru, Sri Lanka, and Thailand. The measure used for family planning program scope and vigor is that of Mauldin et al. (1978), which refers to the state of the program circa 1972. In the macro analyses that follow, this variable is used in its actual scored form, and also as a dichotomy with countries coded as none to weak versus moderate to strong. Socioeconomic development is measured variously. In our macro analyses, we experiment with gross domestic product per capita circa 1965 (Hagen and Hawrylyshyn, 1969~. In addition, we attempt two transformations of the social setting index of Mauldin et al. (1978), which refers to conditions circa 1970. In the first of these, we rank-order the available 15 WFS countries on this index. In the other, we dichotomize the index into low to upper middle versus high. Of course, for any macro analysis only one measure of family planning program effort is considered at a time, and only one measure of socioeconomic development. The micro model is cohort-specific. The results reported here are based on a single birth cohort of currently married respondents aged 40-44 in 1974. We use this year because the survey dates range over a four-year period (1974-771. m is particular cohort was chosen, first, to permit study of between-country variability in the socioeconomic deter- minants of all three components of the fertility process. Selection of an older cohort allows sufficient time for the implications of socioeco- nomic characteristics to develop. In addition, older cohorts experienced first-hand the effects of the family planning programs introduced in many countries during the 1965-74 decade. Respondents aged 40-44 in 1974 were aged 30-34 in 1964 ; according to the logic underlying the decomposition of the fertility process, these women were entering the potentially discretionary stage of their reproductive life as major policy shifts were taking place at the macro level. m is cohort may thus have played a critical role in determining the success or failure of family planning programs in many Third World countries.