Abstract
This paper investigates the effects of social network structure on the diffusion of agricultural technologies using household-level panel data from Ethiopia. We correct for possible biases due to the endogeneity of social networks using a social experiment in which we provide mobile phones to randomly selected households. We find that the diffusion of information on a complex technology is not promoted by simply knowing an extension agent but by knowing an agent that a particular household can rely on and by clustered networks in which most friends of the household are friends of each other.