SUMMARYThis paper analyzes allocations of agricultural extension services in India in terms of a model of villages as political interest groups. The empirical evidence is viewed in the light of two competing models of the distribution and quantity of public goods: the ‘efficiency’ model of Hayamiand Ruttan(1971), and the ‘interest group model’ of the newly developed literature on economic regulation. It is argued that in villages that are relatively remote from competitive markets, large landholders may be able to induce landless farmers to support the provision of extension services, because of the dependence of the landless farmers on the landowners for jobs and credit. In the absence of this ‘dependency structure’, the freerider problem would tend to prevent the village from acting collectively. This hypothesis is supported with evidence that villages with relatively large proportions of landless farmers are more likely to receive extension services,ceteris paribus; this relationship holds only in villages that are relatively far from links to competitive markets. It is also found that more highly educated farmers receive more extension services, despite evidence from previous studies that education and extension are substitutes in agricultural production. These results and other evidence support the interest group model over the efficiency model, which holds that efficiency considerations are the main determinants of governme