Proposition 209 banned race-based affirmative action at California public universities in 1998. Using a difference-in-difference research design and a newly-constructed longitudinal database linking all 1994-2002 University of California applicants to their educational experiences and wages, I show that ending affirmative action caused underrepresented minority (URM) freshman applicants to cascade into lower-quality colleges. The “Mismatch Hypothesis” implies that this cascade would provide net educational benefits to URM applicants, but their degree attainment declined overall and in STEM fields, especially among less academically qualified applicants. URM applicants’ average wages in their 20s and 30s subsequently declined, driven by declines among Hispanic applicants. These declines are not explained by URM students’ performance or persistence in STEM course sequences, which were unchanged after Prop 209. Ending affirmative action also deterred thousands of qualified URM students from applying to any UC campus. Complementary regression discontinuity and institutional value-added analyses suggest that affirmative action’s net educational and wage benefits for URM applicants exceed its net costs for on-the-margin white and Asian applicants.