Studies in Family Planning

Revising the proximate determinants of fertility framework: what have we learned in the past 20 years?

In the mid-1950s, Davis and Blake elaborated a framework of the factors affecting fertility that recognized both indirect and direct determinants of fertility (Davis and Blake, 1956). In 1978, Bongaarts developed these ideas into a framework for analyzing the proximate determinants of fertility that explained the fertility-inhibiting effects of the key direct determinants (Bongaarts, 1978 and 1982). Bongaarts's work represented a significant advance over previous attempts in that it presented a simple model that could be readily applied using available data. Since the publication of Bongaarts' first paper on this framework, more than 100 publications have appeared describing applications to different country and regional settings. The framework has been used for a variety of purposes, including: (1) decomposing the contribution of each of the proximate determinants to the realization of the current level of the total fertility rate; (2) analyzing the contribution of changes in the proximate determinants to changes in the total fertility rate over time; (3) comparing the differences in fertility between two countries or regions on the basis of differences in the proximate determinants; (4) estimating total abortion rates as a residual after the effects of all other proximate determinants have been removed; and (5) projecting future levels of contraceptive use that would be required to achieve fertility goals given expected changes in the other proximate determinants or future levels of fertility given expected or desired changes in contraceptive use.

This model is by no means the only approach to analyzing the factors that determine human fertility. Individual-level approaches have also been developed that examine a similar set of factors but describe the effects of these factors on individual fertility (Hobcraft and Little, 1984). Wood has developed a dynamic model of the proximate determinants of natural fertility (excluding contraception and induced abortion) that is also based on data for individuals (Wood, 1994). All these approaches have their uses. This article focuses on Bongaarts' aggregate-level approach and its use in explaining differences in total fertility over time and across different populations. Although some criticisms of Bongaarts' approach have been advanced (Reinis, 1992; Wood, 1994), his model remains one of the most widely used tools for analyzing fertility and fertility change.

Since the development of the proximate determinants framework more than 20 years ago, a wealth of new data has become available. More than 60 Demographic and Health Surveys have provided detailed information on fertility and the proximate determinants. A number of similar fertility and health surveys have been conducted by the Centers for Disease Control and Prevention as have various country surveys that have not been part of either of these two international data-collection efforts.

Given the importance of the Bongaarts framework and the amount of new data that has become available since its initial development, an examination of available information is appropriate to see what implications can be drawn for the application of the framework. This article suggests how some of these new data might be integrated into the proximate determinants model.

The Bongaarts Model

Bongaarts's original model included four proximate determinants: marriage, postpartum infecundability, abortion, and contraception. In a later paper, Bongaarts added a fifth determinant, pathological sterility (Bongaarts et al., 1984). The basic model is:

TFR = Cm * Ci * Ca * Cp * Cc * TF,

where Cm is the index of proportion married, Ci is the index of lactational infecundability, Ca is the index of abortion, Cp is the index of pathological sterility, Cc is the index of contraception, and TF is total fecundity. Although this aggregate version of the model is the most widely used, an age-specific version is also used that calculates the effects separately for each five-year age group from 15-19 to 45-49 (Bongaarts and Stover, 1986).

Index of Marriage

The index of marriage is intended to express the reduction in fertility caused by women's not being sexually active throughout their entire reproductive period. The index is calculated as the sum of age-specific proportions married, m(a), times age-specific marital fertility rates, g(a), divided by the sum of age-specific marital fertility rates:

Cm = {[Sigma] m(a) * g(a)}/[Sigma] g(a).

The index is often approximated by the proportion of women aged 15-49 who are married.

The intention of this index is to represent the effect of periods during which a woman is not sexually active. Because data on sexual activity were scarce in 1978, marriage (formal or informal) was used as a proxy. Today, data on recent sexual activity are available for a number of countries. Using sexual activity to define the index rather than marriage will increase the index to the extent that some unmarried women are sexually active and will decrease it to the extent that some married women are not sexually active.

Various definitions of sexual activity are possible depending on the reference period (for example, active in the last week, last month, last two months, last year, ever sexually active). Table 1 presents information on the proportion of women aged 15-49 who have ever been sexually active by time since last intercourse. The majority of these women report activity within the last month and nearly 80 percent report activity within the last two months. For the three surveys in Table 1 that reported monthly frequency of intercourse (Burundi, Kenya, and Uganda), an average of 11 percent of sexually active women reported having had intercourse only once in the last month (not shown). This figure is similar to those reporting activity in the last two months but not the last month. The best solution to the problem of defining sexual activity would be to use the data on frequency of intercourse and include this factor in the proximate determinants model by relating coital frequency to fecundability. Rutenberg has suggested a methodology to do this (Rutenberg, 1993). However, too few surveys collect such information to make this practical for most applications. Wood has analyzed the effect of variations in coital frequency on the fecundability of couples at the same age (Wood, 1994). He concluded that coital frequency does not explain a major portion of the variation in fecundability. Therefore, the approach adopted here is to use the proportion active in the last month as the best indicator of sexual activity for the purposes of the proximate determinants model. To this are added women who are not now sexually active but who are currently pregnant or abstaining postpartum, because clearly they have been exposed to the risk of pregnancy recently.

Figure 1 compares the proportions married or in union with the proportions who are sexually active according to this definition for 42 DHS surveys conducted between 1986 and 1995. For four surveys, the proportion [TABULAR DATA FOR TABLE 1 OMITTED] of women who are unmarried and sexually active is larger than the proportion of women who are married but not sexually active. For 23 other countries, the reverse is true. In 14 countries the two groups are about equal. For all 42 countries, the average proportion married or in union is 64 percent, while the average proportion sexually active is 60 percent.

Sexual activity is a more direct measure of exposure to pregnancy than marriage and should be used where such data are available. Data will not be available in all countries, however, because many Asian surveys only include ever-married women. In order to avoid confusion about which definition is being used, this index will be referred to as Cx when based on the proportion sexually active and Cm when based on the proportion married or in union.

Index of Postpartum Infecundability

The index of postpartum infecundability is intended to describe the effects on fertility of extended periods of postpartum amenorrhea. The index is calculated as the average birth interval in the absence of breastfeeding, divided by the average length of the interval when breastfeeding takes place:

Ci = 20/(18. …

Log in to your account to read this article – and millions more.