17

In this paper, we try to evaluate a college value-added defined as a difference in performances. We propose a strategy of identification and an empirical application to student satisfaction in Paris area. Indeed, college concentration and zoning enrollment regulation that apply to high school graduates provide a quasi natural experiment. We show that ignoring selection on the unobservables systematically leads to biased estimations of the true value-added. As a consequence, evaluations relying only on usual observed student characteristics could be seriously misleading. In this case, public policies that would use college performances as a criterion for public fundings could raise important inequity problems.Classification JEL : I23, C21, C26