Class assignment and peer effects: evidence from brazilian primary schools

I propose a novel way of identifying peer group effects by employing a regressiondiscontinuity design that makes use of the assignment mechanism of students into classes in Brazilian primary schools. I estimate a significant negative effect for students close to the cut-off point from being in the older class in the size of about 0.5 standard deviations in math achievement. The paper provides evidence that differences in the variation of age and ability between classes lead to changes of teaching practices and lead to behavioural changes of students and make educational production less efficient in more heterogeneous classes.