To address the mounting concern about the validity of student self-reported data, this study relies on official matriculation records to gauge the effect of classroom diversity on student dropout risk. Using panel data to track the 4-year dropout risk of a cohort of new first-year students (N = 3545) at a public research university, we employ a discrete-time-to-event history model to estimate the marginal effect of classroom ethnic-racial composition while controlling for both time-variant and time-invariant student-level and classroom-level factors on demographics, precollege preparation, college academic experience, campus living arrangement, and financial aid support. The study finds: 1) The effect of classroom ethnic-racial composition during a student’s enrollment spell varies across student ethnic-racial identity, first-generation status, and level of academic preparation of classroom peers; exposure to Asian classmates is associated with a lower dropout risk for Black students, while exposure to underrepresented minority classmates (excluding Asians) is associated with lower dropout risk for Hispanic, Native American, multi-ethic, and first-generation students. 2) Semester-to-semester rise in exposure to Asian classmates is associated with a lower dropout risk for Black students. 3) Observed effects of classroom ethnic-racial composition do not vary significantly with enrollment in diversity-focused courses. 4) Estimated effect sizes of ethnic-racial classroom composition are very small in comparison to effects of student academic engagement and success. Thus, the nexus between diversity and academic persistence is moderated by a host of factors, both time variant and invariant, and is difficult to leverage operationally due to observed small effects and student discretionary behavior.