Do disadvantaged students benefit from attending classes with more skilled colleagues? : evidence from a top university in Brazil

Peers play an essential role in cognitive and non-cognitive skills formation. Ordinal rank may also change incentives and environment, impacting students' efforts. Using two rich administrative data sets and a rule of admission at one top university in Brazil, we apply a regression discontinuity design to study the effect of class allocation on academic performance and labour market outcomes. The rule creates two potential effects on students: peer and ranking effects. The last student of the first class will have higher-ability peers, but at the same time will have a lower ordinal rank than the first student of the second class. These effects usually play in different directions. The rich data also allows conducting all the analyses separately for affirmative action students and regular students. The main results suggest that affirmative action students, a group of disadvantaged Black and mixed-race students from low-income families and with lower levels of education, are the most negatively impacted by being the last student of the first class. They have a lower GPA in the first semester, in the first year, and when they graduate. They also have a higher number of failures. Students in STEM majors drive results. However, the class allocation has no impact on employment and income.