The hypothesis of full risk sharing can be taken to data from low-income countries and evaluate formal and informal financial systems. In many contexts, idiosyncratic risks are high, so credit/insurance arrangements could be beneficial. Statistical tests reveal that households in southern India take advantage of these possibilities; villages in Cote d'Ivoire and countries in Thailand do not do as well. The paper includes an empirical description of the devices used to smooth consumption and a theoretical discussion of private information and incentives on ideal operating systems. The full information and mechanism design frameworks provide benchmarks for policy analysis. Copyright 1995 by American Economic Association.