This paper presents an economic model of group formation with an application to data collected from an agricultural credit program in western Honduras. We formulate a simple theory of group formation using the concept of centers of gravity to explain why individuals join a group. According to our theory, prospective members join based on the potential benefits and costs of group membership, and based on their perception of social distance between themselves and other group members. Social distance is unobservable by outsiders but known by the individual: if you are in then you know who has blue hair. Thus, we argue that social distance helps explain preferences for group formation. To test our theory we analyze data collected from members and non-members of PRODERT, a program that has helped create 188 “Cajas Rurales” (CRs). Using conjoint analysis we test for differences in preferences between members and non-members for the main attributes of the CR. We find that members and non-members exhibit similar preferences for the attributes of the CR; therefore non-membership is not related to supply factors. Using information gathered by executing field experiments, we estimate a proxy for social distance. We use this proxy to run a group formation equation and find that it explains, along with individual characteristics, participation in the CR. Finally we offer suggestions on how to balance performance and coverage in programs in which beneficiaries decide who joins. Small cohesive groups may show exceptional performance at the cost of low coverage, and the opposite may be true.