Identification of academic peer effects in college: Does data aggregation matter?

Abstract

This study exploited random roommate assignments in a small Chinese college to estimate the causal effects of roommates’ scores on the national College Entrance Test (CET) on first-year students’ Grade Point Average (GPA). Analyzing data on an entire cohort of enrolled students, we found that the level of aggregation, for both the peer-ability measure and one’s own academic-outcome measure, matters for the identification of academic peer effects. Specifically, while roommates’ average CET score has a barely significant impact, the highest-scoring roommate’s CET score has a strong positive impact. Peer effects are also larger for one’s GPA for required courses than that of elective courses. Finally, peer effects in both types of courses decline over time while the effects of one’s own CET score increase over time, suggesting that students in this college tend to substitute their own ability for peer ability as they become more academically independent.