Using data from a large U.S. retail firm, we examine how racial matches between managers and their employees affect rates of employee quits, dismissals, and promotions. We exploit changes in management at hundreds of stores to estimate hazard models with store fixed effects that control for all unobserved differences across store locations. We find a general pattern of own-race bias in that employees usually have better outcomes when they are the same race as their manager. But we do find anomalies in this pattern, particularly when the manager-employee match violates traditional racial hierarchies (for example, nonwhites managing whites). (Contains 7 tables, 1 figure and 27 footnotes.)