A dynamic discrete choice model is set up to estimate the effects of grade retention in high school, both in the short- (end-of-year evaluation) and long-run (drop-out and delay). In contrast to regression discontinuity designs, this approach captures treatment heterogeneity and controls for...

While the current empirical literature on peer group effects in schools highlights that credible causal peer effects cannot be estimated unless parental sorting is taken into account, the present paper highlights that causal peer effects might be conditional on the learning environment in which...