Social norms and the fertility transition

Abstract

This paper provides a norm-based explanation for two features of the fertility transition that have been observed in many different settings: the slow response to external interventions and the wide variation in the response to the same intervention. Most societies have traditionally put norms into place to regulate fertility. When the economic environment changes, individuals gradually learn through their social interactions about the new reproductive equilibrium that will emerge in their community. This characterization of the fertility transition as a process of changing social norms is applied to rural Bangladesh, where norms are organized at the level of the religious group and interactions rarely cross religious boundaries. Consistent with the view that changing social norms are driving changes in reproductive behavior in these communities, we find that the individual's contraception decision responds strongly to changes in contraceptive prevalence in her own religious group within the village whereas cross-religion effects are entirely absent. Local changes in reproductive behavior occur independently across religious groups despite the fact that all individuals in the village have access to the same family planning inputs.

Introduction

Countries at similar levels of economic development are often seen to display very different patterns of fertility behavior (Bongaarts and Watkins, 1996). Although fertility rates have declined throughout the world over the past decades, long delays and wide differentials in the response to family planning programs have also been frequently observed, both across countries as well as within countries (see, for instance, Bulatao, 1998; Cleland et al., 1994; NRC, 1993). One explanation for these stylized facts is based on the idea that many aspects of individual behavior, including fertility, are socially regulated in a traditional economy. While such social regulation has advantages of its own, the drawback is that it may prevent individuals from responding immediately to new economic opportunities. Social norms are typically seen to emerge in environments characterized by multiple equilibria, to keep the community in a preferred equilibrium (Kandori, 1992). Changes in the economic environment, such as the availability of modern contraceptives, could reopen the possibility for such multiple equilibria, which would explain the slow response to external interventions, as well as the differential response to the same external stimulus, as each community gradually converges to a new reproductive equilibrium.¹

The setting for this study is the fertility transition in rural Bangladesh. The International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) launched a Maternal Child Health–Family Planning (MCH-FP) project in 1978, covering seventy villages in Matlab thana, Comilla district. The MCH-FP project is quite possibly the most intensive family planning program ever put in place: all households in the intervention area have been visited by a Community Health Worker (CHW) once every 2 weeks since the inception of the project in 1978, and contraceptives are provided to them free of cost. Despite these economic incentives, and the sustained pressure on the households to change their behavior, we still see long delays in the adoption of contraceptives. The family planning program was already well established in the intervention area by the time our data began in 1983. Nevertheless, contraception levels continued to increase steadily over the sample period (1983–93), from 40% in 1983 to 63% in 1993, with an accompanying decline in total fertility rates from 4.5 children per woman to 2.9 children over that period. Wide variation in long-run contraceptive prevalence is also observed across villages in the intervention area.

Most societies have traditionally put norms into place to regulate fertility. In rural Bangladesh, the traditional norm was characterized by early and universal marriage, followed by immediate and continuous child-bearing. Religious authority provided legitimacy and enforced the rules that sustained this equilibrium. In such a social environment, the unexpected availability of modern contraceptives through the family planning program would have opened up the possibility for new equilibria, in which a sufficient fraction of the women in the village ignored the religious sanctions and began to regulate fertility. The point of departure for our simple model of fertility change, following the exogenous economic intervention, is a social uncertainty: the individual does not know what level of contraceptive prevalence will ultimately be sustained in her community. This uncertainty is slowly resolved over time as women in the village interact sequentially with each other from one period to the next, which explains the gradual change in contraceptive prevalence that we see in the data, as well as the convergence to different levels of contraceptive use across communities.

While this characterization of social change as a learning process explains the broad stylized facts that we described above, it also allows us to endogenously derive the individual's decision rule during the transition from the traditional equilibrium to the modern equilibrium: The contraception decision in any time period is determined by the individual's lagged decision and the lagged level of contraceptive prevalence in the community. However, this prediction by itself has little bite since it is well known that a spurious correlation between the individual's decision and her neighbors' past decisions could be obtained when unobserved determinants of the contraception decision are correlated across neighbors and over time (Manski, 1993). For example, neighbors' decisions could simply proxy for changes in economic opportunities or the effectiveness of the MCH-FP project itself.

Our strategy in this paper to provide additional support for the view that changes in contraceptive prevalence were driven by changes in underlying social norms takes advantage of the institutional background that we will provide in Section 2. Female mobility in Bangladesh has traditionally been severely restricted by the institution of purdah. Young married women will rarely leave the homestead (bari), and when they do, it will typically be to visit extended family or kin. While the two major religious groups in rural Bangladesh, Hindus (who constitute 18% of the population in our villages) and Muslims, share a common language and a common Bengali culture, female interactions almost never cross religious boundaries even within the village. Changes in social norms must thus occur independently across religious groups within the village.

We test these implications of the model with a unique data set, which includes contraceptive use information as well as demographic and socioeconomic characteristics for all the women residing in the 70 villages in the intervention area over an 11-year period (1983–93). Consistent with the preceding prediction, we present the striking result in Section 5 that while individuals respond strongly to contraceptive prevalence within their own religious group in the village, cross-religion effects are entirely absent in the data. In contrast, when we partition the village by other variables, such as age or education, we consistently observe large and significant cross-group effects. We will also show in Section 5 that omitted determinants of the individual's contraceptive decision must be completely uncorrelated across religious groups within the village to spuriously generate the within-religion and cross-religion patterns that we just described. We will argue in that section that standard omitted variables, such as unobserved program effects or economic change, which complicated the interpretation of the estimated contraception decision rule above, are unlikely to satisfy this condition. For example, while health inputs and information signals supplied by the MCH-FP project may have varied across religious groups within the village, it is difficult to imagine that they were uncorrelated across these groups. After all, it is the same agency, and the same CHW, that is providing these inputs.²

The paper is organized in six sections. Section 2 describes the institutional setting, paying special attention to the social restrictions that prevented the immediate adoption of contraceptives in the intervention area. Section 3 describes the village level data: we see a gradual change in contraceptive prevalence over time as well as sorting across villages to different long-run contraceptive prevalence levels. Section 4 presents a simple model of social change that is consistent with these stylized facts. The individual's (optimal) decision rule is also derived in this section. Section 5 subsequently presents the data and the estimation results that support the view that religion-specific social interactions gave rise to the changes in reproductive behavior that we see in the data. Section 6 concludes the paper.