We provide the first experimental estimates of the long-term impacts of a voucher to attend private school by linking data from a privately sponsored voucher initiative in New York City, which awarded the scholarships by lottery to low-income families, to administrative records on college enrollment and degree attainment. We find no significant effects on college enrollment or four-year degree attainment of the offer of a voucher. However, we find substantial, marginally significant impacts for minority students and large, significant impacts for the children of women born in the United States. Negative point estimates for the children of non-minority and foreign-born mothers are not statistically significant at conventional levels. The information needed to match students to administrative data on postsecondary outcomes was available for 99% of the sample. We find evidence of substantial bias due to attrition in the original evaluation, which relied on data collected at follow-up sessions.