Abstract:Previous papers which examine the importance of peer effects using exogenous variation in college roommates have found only very limited evidence that a student's first year grade performance is influenced by the observable academic characteristics of his/her roommate. One possible explanation for this finding is that peer effects do not play a particularly important role in the higher education setting. However, another very plausible explanation for this finding is that peer effects are important in higher education but that these previous empirical efforts have simply not been "looking in the right place" to find the evidence of peer effects in this setting. Thus, while these papers have received considerable attention due to the general difficulty of finding credible exogenous variation in peer quality, they have difficulty answering the most fundamental question related to peer effects in this higher education -whether peer effects play an important role or not. This paper provides depth to the peer effects literature using unique new survey and administrative data.keywords: peer effects, education, educational attainment, higher education, poverty JEL codes: I2 -Education, J -Labor and Demographic Economics 1 The authors present a review of past theoretical and empirical work on peer effects. Given the thoroughness of the review, we have chosen not to repeat it here.2 Sacerdote (2001) finds no evidence that a student's first year grade point average is influenced by his/her roommate's score on an academic index created by the Dartmouth admissions office if this score is included in the specified grade regression in a linear fashion. In a different specification, having a roommate with an academic index score in the top 25% is found to increase a student's grade point average by .033 relative to having a roommate with a score in the bottom 25% and by .047 relative to having a roommate with a score in the middle 50%. A test of the null hypothesis that there is no difference between having a roommate in the top 25% and having a roommate in the bottom 25% is not rejected at standard significance levels. A test of the null hypothesis that there 1 Peer effects have the potential to play important roles in determining the impact of many current and potential education policies. Unfortunately, determining the nature and importance of peer effects in either lower or higher education is a difficult task. The empirical difficulty stems from the reality that a given student's classmates and friends are determined by a complex set of decisions made by the student, the student's parents, and/or school administrators and teachers. This non-randomness creates problems of inference because it implies that unobservable determinants of a particular student's academic outcomes may tend to be systematically related to the observable and unobservable characteristics of his/her friends or classmates.This empirical difficulty was emphasized recently by Sacerdote (2001) andZimmerman (2003) who reviewed past literature on...