In this paper we provide an empirical analysis of the social composition and performance of microfinance groups, known as Self‐Help Groups, based on an original census we carried out in a poor area of Northern India. We examine whether traditionally disadvantaged villagers, such as scheduled tribes and landless farmers, are as likely to draw benefits from these SHGs as other villagers. While the initial participation in the groups closely reflects the social composition of the village, we find evidence of a selective attrition process against scheduled tribes who are less likely to remain members. Their expected access to bank loans ‐ which is the primary aim of those groups ‐ is also much more limited. By contrast, landless farmers are over‐represented in these groups. As a result, even though they are more likely to leave the groups, they tend to benefit disproportionately from the SHGs. In expected terms, they receive more than two times the amount of bank loans given to other farmers. Overall, the program has therefore non trivial but important distributional implications.