Article dans une revue: In this article, we use data from the European Community Household Panel to evaluate the impact of a French guaranteed income program, the RMI, on the hazard out of unemployment. Self-selection into the program is corrected using a multivariate duration model developed by [Abbring, J.H., van den Berg, G.J., 2003. The non-parametric identification of treatment effects in duration models. Econometrica 71 (5), 1491- 1517]. We find that RMI receipt has a strong negative impact during the first months of program participation, but that this disincentive effect quickly falls to insignificant levels after six months. Household structure also appears to be an important determinant of the importance of the adverse effect of program participation.