Abstract:
There is abundant evidence that having high-achieving peers generate positive gains on individual performance. We know little, however, about the mechanisms that operate when interacting with others. We implement a randomized experiment in 171 schools in Mexico that allows us to isolate the effect of changes in the mean and variance of peers on students' academic performance and behavior. We find that tracking and mixing (i.e., bimodal groups) generate similar average gains (0.07-0.08SD) without hurting low-achieving students. Both concerns about the relative ranking as well as peer support seem to play a role in explaining peer effects.