Article dans une revue: Using a quasi-random allocation of students to classes in a French university, we are able to estimate peer effects and teacher effects, with a specific attention to non-linear peer effects. We find that teacher effects are strong, as found at other levels of the education system, but that peer effects have very limited impact. This implies that restricting student access to some universities is of no benefit to remaining students in terms of academic performance. In contrast, attention to teacher performance should be strong at the higher education level.