This paper estimates the effect of students' background characteristics on the academic outcomes of their college roommates. It uses data from four classes of students at Wellesley College, where roommate assignment is credibly random, conditional on student responses to a housing preference form. In linear specifications, there are no peer effects on students' grade point averages. There is some evidence that students' SAT scores have nonlinear effect on their roommates' achievement, but the results are not robust. We conclude that roommate peer effects might exist among small groups of students, but they are not a key determinant of students' average academic outcomes at Wellesley College.