Using household data from urban Ethiopia, we provide an empirical test of the economic theory of Rotating Savings and Credit Associations (ROSCAs) and identify the impact of ROSCAs and member characteristics on participation and volume of saving. Unlike other studies, we account for the endogeneity of all ROSCA variables. Muslims, individuals who live in richer households, the self-employed, private sector employees and households with large numbers of women are more likely to join ROSCAs. A robust finding is that the savings are significantly affected by the characteristics of the members but not that of the informal saving institution. Some policy implications are discussed.