Click for next page ( 222


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 221
Appendix A Adapting the Easterlin-Crimmins Synthesis Mode} to Sub-Saharan Conditions This appendix explores whether the Easterlin and Cnmrnins (1985) synthesis model of reproductive decisions provides an appropriate framework for the study of contraceptive use in Africa. The synthesis perspective centers attention on an individual couple of reproductive age. What is at issue is the number of children the couple expects to bear if no contraception is em- ployed, compared with the number of births that are wanted or demanded. If the potential supply of births exceeds the quantity demanded, there exists a motivation for contraceptive use. Thus, the model draws together three essential elements: reproductive capacity, the demand for reproductive out- comes, and the principal means contraception by which couples bring actual reproductive outcomes into agreement with their reproductive desires. Socioeconomic determinants enter the model via the concept of de- mand, which encompasses factors such as the risks of child mortality and the access to and costs of contraceptive methods. Although the Easterlin- Cnmmins model is usually taken to be a model of lifetime fertility or of the behavior of ending or limiting reproduction, and their socioeconomic deter 1Easterlin and Crimmins (1985) introduce a second concept of demand, having to do with the desired number of births in the absence of mortality risk, that assumes away all monetary, time, or psychic costs of contraception. As discussed in Montgomery (1987), this hypothetical or notional concept of demand plays no essential part in the synthesis model. Hence, the concept of demand employed in the material that follows is conventional, in that mortality risks and contraceptive costs are hypothesized to affect the desired number of births. 221

OCR for page 221
222 FACTORS AFFECTING CONTRACEPTIVE USE minants, the model can be used equally well to understand the demand for spacing births. As discussed below, the spacing-limiting distinction has to do with the length of the decision period envisioned in the synthesis model. Potential supply and demand may be compared over a period as short as the monthly reproductive cycle or as long as the reproductive span itself. In addition to its connections to birth spacing and limiting, the length of deci- sion periods is also an issue in the model's treatment of proximate determi- nants other than contraception. In effect, the synthesis perspective divides the full range of proximate determinants of fertility (e.g., contraception, involvement in sexual union, postpartum nonsusceptibility, and induced abortion) into three mutually ex- clusive groups. First, there are the exogenous proximate determinants such as individual fecundability, which set constraints on reproductive capacity. By exogenous we mean those determinants that are not under the control of the couple during the decision period under consideration. This category of determinants defines the couple's potential supply of births. The second group of proximate determinants comprises the effective and less effective methods of contraception; these are clearly endogenous in being, at least in part, the means by which the couple puts into effect its reproductive de- mands.2 In the third category of dete~-~inants fall those behaviors that cannot be neatly classified as either exogenous or endogenous in nature. One could include breastfeeding intensity and duration in this third cat- egory, together with postpartum sexual abstinence, participation in a repro- ductive union, and spousal separation. The degree of erogeneity displayed by such behaviors will depend on the length of the decision period in ques- tion and on the specific character of the economic, social, and cultural organization. By distinguishing among categories of proximate determinants, direct- ing attention to the category of contraceptive use, and providing an organ~z- ing framework that shows how reproductive constraints and demands are expressed in contraceptive use, the synthesis perspective has much to rec- ommend it. Its usefulness in this regard was recognized by an earlier National Research Council report on determinants of fertility in developing countries (Bulatao and Lee, 1983~. 2At this point we should remark on the place of induced abortion in the conceptual frame- work developed below. It seems that by all accounts, the psychic and access costs of abortion in sub-Saharan Africa are of such a magnitude that it is rarely employed in marriage, although recent evidence indicates that abortion is increasing (Coeytaux, 1988). We regard these costs as being grouped with other factors that implicitly determine the demand for contraception. There is little need to single out the costs of abortion as a distinct motivation for contraception, unless (1) these costs, or components of the costs, can be measured; and either (2) the costs vary significantly across socioeconomic or demographic groups; or (3) the costs are expected to change in the future, perhaps as a result of policy.

OCR for page 221
APPENDIX A 223 In what follows we make a few brief comments, primarily of a theoreti- cal nature, on the elements of the analytic framework as applied to Africa. The bulk of the discussion concerns the relations between the concepts of supply and demand and their connections to contraceptive use, these factors being grouped in the lowermost portion of Figure 1-1. The ambiguities in these concepts are revealed perhaps more clearly in an application to sub- Saharan Africa than they are in other settings, and we attempt to clarify the issues as much as possible. DEMAND FOR BIRTHS As envisioned in the framework of Figure 1-1, the concept of demand for births incorporates demands both for spacing births and for limiting them. The distinction between spacing and limiting motivations is of course very important with respect to the longer-run implications of method use. However, apart from sterilization, use of a contraceptive method does not in itself give evidence as to the relative weight of the spacing and limiting motives. . Rather, birth spacing is accommodated in the Easterlin-Crimmins syn- thesis by the consideration of a relatively short decision period, wherein an individual attaches utility rankings to various probabilities of conception within the period in question. A person with-a strong motivation to delay the next birth would assign a utility penalty to behavioral choices that entail high probabilities of conception; so too would a person with a strong moti- vation to avert the next birth altogether. The birth spacer may be motiv fated by concern for child survival, the birth averter by a desire to limit family size; but each is engaged in behavior that has a contraceptive effect. The differences in these underlying motivations would emerge in the subsequent sequence of decision periods, as the utility penalty associated with concep- tion progressively declines for birth spacers while (presumably) remaining constant for birth averters. To put it differently, the spacing-lim~ting distinction has to do, in one respect, with the length of the decision period to which the analytic model is applied and, in another, with the linkages among such decision periods over the reproductive life span. We do not propose to address here the theoretical issues associated with birth spacing and limiting. A full treat- ment would require an explicitly dynamic decision framework with due allowance for uncertainty and for the interplay over the reproductive span between reproductive outcomes and contraceptive-related choices.3 Rather, 3For exploratory work in this area, see the volume edited by Tsui and Herbertson (1989); one article in the collection, Montgomery (1989), sets out the dynamic theory as applied to contraceptive use.

OCR for page 221
224 FACTORS AFFECTING CONTRACEPTIVE USE the framework of Figure 1-1 is drawn so as to give equal theoretical stand- ing to spacing and limiting motives for contraceptive use, and to suggest further that in sub-Saharan Africa, there need be no neat line of division separating them. Along with the intertemporal aspect of demand, the interpersonal aspect must also be recognized. We argue in Chapters 3 and 4 that a reproductive decision framework appropriate to sub-Saharan Africa must admit the pos- sibility of separate interests on the part of the male and the female in a reproductive pair. It follows that the concept of demand must itself be individual specific. Thus, in assigning utility rankings to decision options regarding method use, a woman would take into consideration her own present and expected future resources, resources from her husband on which she can expect to rely in the future, the possibility of future marital dissolu- tion or the entry of new wives into the union, and so on. The husband would have his own calculations in these matters, which would not neces- sarily yield rankings in agreement with those of his wife. Indeed, once the reproductive pair is split in this fashion, a gap is introduced between individual demands and contraceptive behavior. Whose demands and utility rankings are actually expressed in contraceptive use? The use of certain methods, including most of the coitus-dependent meth- ods, requires a measure of cooperation between spouses or, at minimum, an acquiescence on the part of one spouse. Other methods can be employed unilaterally and even surreptitiously, without requiring the knowledge of one's spouse. In some socioeconomic circumstances where one might ex- pect husband-wife conflicts, it might prove useful to distinguish among contraceptive methods according to the degree of spouse cooperation they entail. SUPPLY OF BIRTHS As indicated above, the supply of births that influences decisions about contraceptive use is affected by proximate determinants that are clearly exogenous, such as individual fecundability; those that can be treated as exogenous only in the short run, such as spouse separation; and possibly determinants such as breastfeeding and postpartum abstinence, whose exogeneity is in doubt. However, there is considerable empirical evidence that marriage, mari- tal dissolution' spouse separation, breastfeeding, and postpartum abstinence are correlated with the same socioeconomic factors (e.g., urbanization and female education) that influence contraceptive use. Moreover, in a number of African societies (Page and Lesthaeghe, 1981) the motivation for breastfeeding and abstinence is explicitly understood in contraceptive terms: as the traditional African means of spacing births.

OCR for page 221
APPENDIX A 225 (The spacing motive has to do with the health of the child and the mother, but this is not the important point.) How then can these behaviors be viewed as exogenous determinants of contraceptive use? A related question must be raised about the appropriateness of including women who are breastfeeding or postpartum abstinent in calculations of contraceptive prevalence. From the analytic point of view, it would be awkward to group postpar- tum behaviors such as breastfeeding, which is tied in a physiological sense to a previous birth, with contraceptive methods that in principle can be adopted at any point in the reproductive span. One simply does not take up breastfeeding six months into a birth interval; the prevalence of breastfeeding is therefore bound up with the incidence of births in a way that the preva- lence of the pill or the intrauterine device (IUD) is not. The case of postpartum sexual abstinence is less clear-cut because, unlike breastfeeding, abstinence could well be adopted midway in a birth interval. Here we must argue that the psychic costs associated with a given spell of abstinence are lower if the spell begins at the event of birth, rather than at some later point in the birth interval. This argument returns to the African emphasis given to birth spacing and the social sanctions that sur- round weaning and the resumption of sexual relations (Caldwell and Caldwell, 1981~. From this perspective, abstinence can be viewed as a postpartum behavior somewhat akin to breastfeeding, although its link to the previous birth is behavioral and social in origin rather than physiological. The analytic awkwardness associated with treating breastfeeding and postpartum abstinence on an equal footing with contraceptive methods con- sists in both the interpretation and the implications of contraceptive preva- lence rates. Consider a prevalence calculation in which the base comprises all nonpregnant women and nonuse is distinguished from use of various contraceptive methods. If we decide to group breastfeeding and abstinence with other conventional contraceptive measures, we must at the same time recognize their highly distinctive patterns of adoption and discontinuation, and for breastfeeding, a contraceptive failure rate that is strongly dependent on duration.4 To be correctly interpreted, contraceptive prevalence rates that incorpo- rate breastfeeding and abstinence would have to be calculated by holding constant the duration since last birth. Consider a comparison of distribu 4It has been hypothesized that the contraceptive effect of breastfeeding could be maintained indefinitely if feeding were to be sustained at maximum intensity. But the usual pattern is for breastfeeding to decline in intensity as food supplements are introduced into the infant's diet, and this process may be accelerated by the physiological strain or fatigue on the part of the mother associated with long durations of full breastfeeding. Hence, it is safe to assume that the use-effectiveness of breastfeeding, as it were, will tend to erode with duration since last birth.

OCR for page 221
226 FACTORS AFFECTING CONTRACEPTIVE USE lions of nonuse and use by method among nonpregnant women at two dura- tions, d and d + 1, since last birth. In moving from duration d to duration d ~ 1, one would expect to see a systematic reduction in the percentage breastfeeding and postpartum abstinent, and a systematic inflation in the sum of nonuse and use of conventional methods. One would then have to devise corrections to the prevalence rates that would net out such system- atic patterns, in order for more interesting behavioral regularities to be discerned in the data. Moreover, owing to the systematic decline in the contraceptive failure rate of breastfeeding with duration since birth, the greater the duration d is, the larger is the fraction of women breastfeeding and not pregnant at d who then conceive and are removed from consideration at duration d + 1. So far, as we are aware, no other conventional means of contraception displays such a pronounced duration dependence in its failure rate. When contracep- tive methods are defined so as to include breastfeeding therefore, the impli- cations of a given method mix for the degree of contraceptive protection will depend on the duration since last birth. The second part of the rationale for treating breastfeeding and postpar- tum abstinence differently from contraceptive use has to do with the contri- bution to understanding and to policy of statistical analyses based on a synthesis framework. The concern is a general one, perhaps most readily evident here in respect to postpartum behaviors, but also of relevance in marriage, spouse separation, and other choice-related behavior having an influence on the risks of conception. The issue can be framed in the follow- ing way: In what sense, if any, will analyses of contraceptive use defined as conditional on the status of other proximate determinants, tend to mis- lead? The central issue is that of statistical erogeneity, selectivity bias, and the importance of unmeasured variables.5 Perhaps a simple representation SEconomic demographers have given great emphasis to the concept of statistical erogeneity, as is evident in Schultz's (1986) critique of the Easterlin synthesis. It should be recognized that the key issue is in fact of a statistical nature, rather than being an issue of the appropriate theory. The concept of conditional demand functions, wherein the demand for one good is ex- pressed as being conditional on demands for other goods, is well accepted in economic theory. The analogy here would be to the demand for contraception conditional on breastfeeding status or conditional on the status of other proximate determinants. But economists have also recog- nized the difficulties in statistical applications of conditional demand theory, given that the theoretically appropriate conditional demand functions will typically include a number of vari- ables that are not subject to empirical measurement. Thus, an empirical application of condi- tional demands will be vulnerable to the charge of selectivity bias arising from omitted vari- ables. This contrast in perspectives is evident in the exchange between Easterlin (1986) and

OCR for page 221
APPENDIX A 227 of method use and postpartum abstinence can help to illuminate. Let C denote the use of a contraceptive method by individual ~ and let Ai indicate postpartum abstinence on the part of this individual. A set of variables Xi encompasses all measurable exogenous socioeconomic determinants of con- traceptive use, including program-related measures of contraceptive access and costs. In any empirical application, important components of the deter- m~nants of use will doubtless go unmeasured, and these unobserved influ- ences are summarized in the variable i, the value of which will vary across individuals. We assume that Xi and i are unco~Telated.6 If both contraceptive use, Ci, and postpartum abstinence, Al, are viewed as endogenous, a linear statistical representations of the determinants of C. and Ai could be set out as follows: Ci = Xi Dc + ci + Vi Ai = XipA + a Pi +Ui' r where the unmeasured background factors si exert an influence on both method use and postpartum abstinence. (We accommodate any additional unmeasured influences on abstinence in the disturbance terms vi and ui, which can be taken to be uncorrelated with i.) Because Xi and (si and vi) are uncorrelated, we have EtCilXi ~ = Xi Dc' and the coefficients pc of the contraceptive use equation can be consistently estimated by conventional methods. Consider now the alternative approach in which the equation for contra- ceptive use Cz is estimated as conditional on abstinence status Ai. For instance, the contraceptive use equation could be estimated on the subsample of women who are not postpartum abstinent, which we may represent as A = 0. Then in the subsample, Schultz (1986) concerning empirical applications of the synthesis model. Easterlin defends his empirical specification of the model in terms of conditional demand theory (although he does not use this language), whereas Schultz attacks the specification primarily on statistical grounds. 6That is, Xi is exogenous from the statistical point of view. 7We use a linear regression representation only to illustrate the key issues; clearly it is not the most appropriate framework for qualitative variables such as contraceptive method choice.

OCR for page 221
228 FACTORS AFFECTING CONT~CEPTIVE USE EtCilXi' Ai = 0] = Xi'pC + ErilAi = 0] = Xi TIC + Et~1i = ~-11a)(Xi PA + Ur)] ~ X'0 In short, when a common unmeasured variable Pi enters both the contracep- tion and the abstinence equations, the influence of socioeconomic and pro- gram determinants of demand Xi cannot be consistently determined in gen- eral through the conditional approach. The degree of bias inherent in this approach is an empirical matter, and it would be very difficult to know, a priori, whether the bias would be small enough to ignore for policy or other purposes. The conditional approach can be defended in two ways. First, one could argue that i does not in fact appear in the abstinence equation (i.e., a = 0), which amounts to an assertion that unobserved characteristics af- fecting abstinence do not have any influence on method use a strong as- sertion indeed given the documented associations between the observed socioeconomic factors Xi and abstinence. Second, one could assert that when the data are grouped according to specific socioeconomic or cultural criteria, so little individual-specific variation remains in Pi that it is, in effect, absorbed into the constant terms of the regression equations, which is the essence of the argument that proceeds from the assertion that postpar- tum abstinence and breastfeeding are "culturally determined," to the con- clusion that one can examine contraceptive use on a conditional basis.8 We wish to underscore this point: To justify using the conditional approach, one or the other of these arguments must be applied to each of the other proximate determinants, including marital status and spouse separa- tion. These analytic and statistical complications surrounding the synthesis framework are not beyond resolution, but to address the issues in full in this report would carry us into new and possibly controversial terrain with re- gard to' methodology. In addition, we believe that the payoff in terms of lessons for policy could be small. For both physiological and behavioral reasons, breastfeeding and postpartum abstinence are necessarily limited behaviors. No conceivable policy could hope to extend the duration of breastfeeding or abstinence much beyond three years, and even reaching three years would be doubtful in the case of abstinence, whereas policies in sub-Saharan Africa could well encourage durations of modern contraceptive The group-specific nature of Pi wounds then be captured in dummy variables indicating region, ethnicity, and the like.

OCR for page 221
APPENDIX A 229 use of this length or longer. With regard to the other proximate determi- nants such as marital status, it seems unlikely that program interventions directed to contraceptive use could have important spillover influences on selection into marriage. To sum up, our approach in this volume proceeds as follows: Our judgment is that marital status is probably not sufficiently endogenous to cause concern. Thus, we calculate measures of contraceptive prevalence for currently married women. Breastfeeding and abstinence are treated as po- tentially endogenous; we do not calculate contraceptive prevalence mea- sures conditional on breastfeeding and abstinence status. But neither do we merge breastfeeding and abstinence with other contraceptive methods. Our attention is restricted to modern method use. This focus on a subset of the endogenous variables may well be ad- equate for an exploration of contraceptive use, but given the possibilities for substitution between breastfeeding/abstinence and modern contraceptive methods, it could yield misleading predictions about fertility. This issue is discussed in Chapters 2 and 7 in connection with the two-phased fertility transition in Africa.