How has the microcredit movement managed to push financial frontiers? Theory shows that if borrowers vary in unobservable risk, then group‐based, joint liability contracts price for risk more accurately than individual contracts, provided that borrowers match with others of similar project riskiness (Ghatak (1999, 2000)). This more accurate risk‐pricing can attract safer borrowers and rouse an otherwise dormant credit market. We extend the theory to include correlated risk, and show that borrowers will match with partners exposed to similar shocks to lower their chances of facing liability for their partners. We use unique data on Thai microcredit borrowing groups to test for homogeneous matching by project riskiness and type of risk exposure. Evidence supports the theory, in that groups are more homogeneous in riskiness but less diversified in type of risk exposure than they would be under random matching. The results suggest that group lending is improving risk‐pricing by embedding a discount for safe borrowers, and can thus explain part of the unprecedented rise in financial intermediation among the world's poor; but that a potential pitfall of voluntary group formation is antidiversification, which points to strategies for lender intervention.