In this paper we estimate the short-term effects of a French retraining program that was intended to improve reemployment prospects of displaced workers. Our empirical analysis uses non-experimental data collected by the French Ministry of Labour. Transitions from unemployment to permanent vs. temporary jobs are assumed to be generated by a dependent competing risks duration model. Moreover we assume that program entry but also direct transitions from the program to employment are both affected by a selectivity process depending on observable and unobservable individual heterogeneity. We find mixed evidence regarding relevance of the program. For displaced workers who join the program, program participation increases the average probability to obtain a permanent job by 8 points (passing from 42 to 50%). However, for displaced workers who do not join the program, this average probability would have been increased by 28 points (passing from 43 to 71%) had they participated in the program. We conclude that, although the program has been globally effective, it has been insufficiently proposed to the displaced workers who would have benefited the most from it.