Microfinance has been heralded as an effective way to address imperfections in credit markets. From a theoretical perspective, however, the success of microfinance contracts has puzzling elements. In particular, the group-based mechanisms often employed are vulnerable to free-riding and collusion, although they can also reduce moral hazard and improve selection. We created an experimental economics laboratory in a large urban market in Lima, Peru and over seven months conducted eleven different games that allow us to unpack microfinance mechanisms in a systematic way. We find that risk-taking broadly conforms to predicted patterns, but that behavior is safer than optimal. The results help to explain why pioneering microfinance institutions have been moving away from group-based contracts.