Introduction

As the survival of any society hinges on (1) the technological adaptation to a given environment, (2) the demographic balance between births, deaths, and migration streams, and (3) the economic principles governing production and distribution, it is proper that this book on the various reproductive regimes of sub-Saharan Africa begins with the links between "production and reproduction." This is a classic starting point: Thomas Malthus integrated population dynamics into the context of the moral economies of peasant societies 200 years ago. His notions of preventive and positive checks still seem to have maintained their relevance, and although most of his applications dealt with the Western European system of the eighteenth century, it is by now evident that all populations, from the hunter-gatherer stage onwards, have generated arrangements through which the reproductive process is regulated. Moreover, these arrangements reflect the basic institutional setup that governs the functioning of the social system as a whole.

The links between the organizing principles of a society and the specific features of its reproductive regime, that is, the parameters of the starting, spacing, and stopping patterns of fertility, are not only of importance for gaining insights into demographic variations at a particular point in time, but they are of even greater value for the understanding of the various paths followed in the course of social change. More specifically, as changes in the spheres of political organization, division of labor, social stratification, economic exchange, cultural differentiation, and demographic regulation are not likely to proceed in a synchronized fashion, a clearer view of the relations between "production and reproduction" is essential to understand individual and group strategies, or, as McNicoll specified (1978, p. 89): "to

locate emerging contradictions and to assess possibilities for their resolution."

This institutional approach to reproductive regimes calls for a comparative methodology. The classic multivariate statistical techniques will continue to occupy an important place, but in the institutional approach the accent shifts from the treatment of strictly individual-level data to a multilevel approach. For instance, it does not suffice to introduce the attribute "ethnic group" into a regression equation with individual-level characteristics such as schooling level or age at marriage. For one, ethnicity stands for organizational and cultural traits of a collectivity, and these traits require further specification. One way of clarifying the situation is to study the relationships between individual-level variables separately for each ethnic context, comparing the outcomes across such contexts. For instance, the impact of a rise in female education (a microvariable) on the length of postpartum abstinence may vary across ethnic groups or regions. This variation may then be explained in terms of characteristics measured for the group as a whole, such as religious composition (a macro- or contextual variable). This approach is not free from problems.

First, as just indicated, the meaning of any broad contextual variable such as ethnic group, region, language . . . , requires specification in terms of more precise and explicit cultural and organizational dimensions such as the type of religion, level of technology, pattern of division of labor, structure of the domestic group, form of wealth concentration or circulation, authority structure, and so forth. At this point choices with respect to the presumed importance of each of these underlying features cannot be avoided. Such choices are of necessity reductionist or subjective in nature, not only because of the merging of complex organizational forms into a few manageable dimensions, but also because of the subsumed causal ordering projected by any author onto the network of cultural and organizational features.

Second, the statistical tools and models used in measuring associations and effects require slight adaptations to fit multilevel theory, that is, theory operating on the levels of both individuals and aggregates. As a consequence, we shall adopt the basic design of "contextual analysis" by establishing the separate impact of individual-level variables (holding the context constant), of aggregate-level variables (fixing individual variable values), and, above all, of interactions between these two. These interaction terms most frequently address the question of how the relationships between individual-level variables vary in shape and strength across the various contexts. In the statistical parlance of multiple regression, for instance, the question is whether slopes and intercepts characterizing the relations between individual-level variables are themselves a function of characteristics of contexts (see Boyd and Iversen, 1979). This obviously leads into a question of comparative sociology: why are individual level elasticities differentiated with respect to contexts?

From this outline, it is clear that we cannot restrict ourselves to the variables measured in the recent round of African fertility surveys, but that additional source materials have to be integrated as well. In order not to lose the advantage of subsequent statistical testing, such qualitative materials have to be operationalized, but this involves approximation and some loss of validity. More specifically for this chapter, we shall attempt to re-express ethnicity as a few basic organizational dimensions on the basis of anthropological monographs and G. P. Murdock's (1967) classification of societies according to a set of organizational characteristics. Additional survey information on variables, such as female involvement in commerce, or on the shift from animism to Islam or Christianity, will also be of use. In short, in the next sections a choice of organizational dimensions believed to be both at the core of a society's functioning and relevant for its reproductive regime will be made. This leads into the construction of an operational typology useful for further work.