We estimate the returns to college using administrative data on college enrollment matched to administrative data on weekly earnings. Utilizing the fact that colleges dismiss low-performing students based on exact GPA cutoffs, we use a regression discontinuity design to estimate the earnings impacts of college. Dismissed students are permitted to apply for readmission, but since relatively few do so, these students end up completing fewer years of school and are approximately 10 percentage points less likely to graduate college. Our estimates suggest that low-performing students (on the margin of college dismissal) derive substantial earnings benefits from college.