Racial gaps in teacher performance ratings have emerged nationwide across newly implemented educator evaluation systems. Using Chicago Public Schools data, we quantify the magnitude of the race gap in teachers' classroom observation scores, examine its determinants, and describe the potential implications for teacher diversity. Between-school differences explain most of the race gap and within-school classroom-level differences--poverty, incoming achievement, and prior-year misconduct of a teacher's students--explain the remainder of the race gap. Teachers' value-added scores explain none of the race gap. Leveraging within-teacher variation in the teacher--evaluator race match, we find that racial mismatch does not influence observation scores. Adjusting observation scores for classroom and school context will generate more equitable ratings of teacher performance and mitigate potential adverse consequences for teacher diversity.