Pre-print, Working paper: This paper takes a network perspective t oinvestigate how rural households in developing countries form the links through which they provide and get economic support. I test the hypothesis that indirect contacts (e.g. friends of friends) matter for link formation. An estimation procedure of a network formation model à la Jackson and Wolinsky (1996) is proposed and applied to data on a single village in Tanzania. Results show that when agents evaluate the net advantage of forming a link they also consider the wealth and the position of indirect contacts. The network externalities from indirect contacts are negative, which suggests a mechanism of competition over scarce resources. This paper proposes the first structural estimation of an endogenous network formation model, and also contributes to the development literature by overcoming the dyadic regression approach and providing evidence that village-level network structure has an explanatory value disregarded by all previous studies.