Heterogeneity in peer effects in random dormitory assignment in a developing country

We study the effect of random dormitory assignment in a tertiary level educational institution in India on students’ subsequent academic achievements. We examine the importance of interactions between the characteristics of the student and his peers for educational outcomes, including non-linear peer-effects and the importance of different socio-economic and geographical backgrounds. We find that peer ability effects are around one-third the size of the effects of one's own ability, and students from non-urban and non-English backgrounds do particularly better when assigned to higher-ability peers. We find that all groups of ability students gain from being matched to high-ability peers, but that this gain is highest for students who are themselves of higher-ability. Our results suggest peer effects are stronger in the first year in dorm. In terms of mechanisms, we find no evidence for effects of peers via mental health, life satisfaction, or risk attitudes. We observe that a roommate's study times is highly correlated with a student's own study times, but we see only a weak positive association between study habits and grades.