Agricultural Intensification and Craft Specialization: A Nonrecursive Model

Abstract

The purpose of this paper is to examine four distinct theoretical positions which offer some implicit or explicit causal account of the empirically observed positive association between agricultural intensity and the division of labor into non? agricultural craft specialties. This empirical relationship has long been noted and discussed. In cross-cultural studies, for example, Hobhouse, Wheeler, and Ginsberg (1914) used number of craft specialties to help classify agricultural societies into incipient, pure, and advanced types, and Gouldner and Peterson's (1962) factor analysis of 71 societies scored on a wide range of traits uncovered a technology factor which shows high positive loadings on such traits as presence of pottery, weaving, basketry, metalwork, and agriculture. Similarly, Murdock and Provost's (1973a) more recent cross-cultural study ofthe division of labor by sex reports an increased specialization and intensification of agriculture factor which also loads highly on these traits. Archaeologists con? cerned with the causes and consequences of food production systems have also noted craft specialization as a key element in the cultural elaboration accompanying agricultural change (Childe 1946; Smith 197 5). And, finally, discussion of this relationship has also been embedded within wider theoretical studies ofthe evolution of social stratification and the emergence ofthe state (Fried 1967; Service 1975). While there is widespread agreement that this relationship is important, there is, however, little consensus concerning the nature and relative importance of the causal linkages underlying the observed association. In the following section, four causal hypotheses concerning this relationship are briefly reviewed. Since one of these posits reciprocal causal effects between agricultural intensification and craft specialization, a simultaneous equations model which explicitly allows for feedback effects between these variables is proposed. This model also allows for the simultaneous testing of all four hypotheses. Estimation strategies for the pro? posed model are then noted, and a full-information estimation procedure (three-stage least squares) is employed in the analysis of a world-wide sample of 131 preindustrial societies, all of which practice some form of agriculture. The analysis is replicated on three regional subsamples. The strikingly similar results for the world-wide and regional analyses provide support for the hypothesis of reciprocal effects.