Although the potentially negative impacts of credit constraints on economic development have long been discussed conceptually, empirical evidence for Africa remains limited. This study uses a direct elicitation approach for a national sample of Rwandan rural households to assess empirically the extent and nature of credit rationing in the semi-formal sector and its impact using an endogenous sample separation between credit-constrained and unconstrained households. Being credit constrained reduces the likelihood of participating in off-farm self-employment activities by about 6.3 percent while making participation in low-return farm wage labour more likely. Even within agriculture, elimination of all types of credit constraints in the semi-formal sector could increase output by some 17 percent. Two suggestions for policy emerge from the findings. First, the estimates suggest that access to information (education, listening to the radio, and membership in a farm cooperative) has a major impact on reducing the incidence of credit constraints in the semi-formal credit sector. Expanding access to information in rural areas thus seems to be one of the most promising strategies to improve credit access in the short term. Second, making it easy to identify land owners and transfer land could also significantly reduce transaction costs associated with credit access.