| Using administrative panel data from the University of Maryland and results from a new student survey, this dissertation explores peer effects and the effects of friends at a large public university. The first chapter describes the contributions of the dissertation to economics, and introduces the new administrative data set used throughout the following two chapters. In the second chapter, I exploit the fact that a subset of students are randomly assigned to housing, and therefore do not self-select into peer groups. I show that there is little evidence of robust residential peer effects in this context. I then examine the impact of friends' characteristics on student achievement. Because friends are endogenously determined, I instrument for friends' characteristics using freshman-year randomized peers' characteristics. The results indicate that friends do not impact student performance more than randomized peers, as would be expected---in fact, the opposite is true. This casts doubt on the widely-held, if rarely made explicit, belief that friendship is the route to peer effects. In the third chapter, I use the same dataset as well as the results of a new survey of University of Maryland undergraduates to explore alternative peer group effect mechanisms, and to study students' friendship formation processes. I find no strong empirical evidence to support any of these peer effect mechanisms suggested in other disciplines. Additionally, I find that undergraduates self-report sorting primarily by attributes other than ability. |
