Apr 22, 2009 
This paper estimates the treatment effect of grade repetitions in junior high schools, using a rich panel of French students. Test scores are observed when students are entering and leaving junior high school. Two identification strategies are used: instrumental variables and matching estimators. With the two methods, value-added is taken as an outcome. The IV estimator uses birth quarter as an instrument to identify the local average treatment effect, and the matching estimator relies on a conditional independence assumption to recover a semi-parametric estimation of the average treatment on the treated. The two methods give a positive effect of grade repetitions, between 10 and 25% of the test-score gain’s standard deviation. A grade repetition improves the probability to graduate from junior high school by 2.5 probability points.
Comment