Although qualitative research suggests that school choice and other interventions are more beneficial for moderately disadvantaged than severely deprived students, the subject has barely been explored by quantitative studies with either observational or experimental designs. We estimate experimentally the impact of a voucher offer on college attainment of poor minority students by household income and parental education. Estimates are obtained from a 1997 private, lottery-based voucher intervention in New York City. National Student Clearinghouse provided 2017 postsecondary outcomes. Positive impacts on moderately disadvantaged students do not extend to the severely deprived. [Additional support for this research was provided by the Hubbard Family Foundation and the Kovner Foundation.]