This paper examines the role that racial discrimination plays in the decision to retain or release an employee and demonstrates the implications for estimating pay gaps. Our empirical setting, professional American football players (NFL), allows us to separate the retention decision from the wage decision. For the first four years of a player’s career, wages are mechanically determined and players are under a restricted ‘rookie’ contract, during which they can be released without cost. Players who survive in the league beyond four years receive a large uptick in their remuneration upon signing their first ‘free-agency’ contract. Consequently, marginal decisions over employment retention during the rookie contract have substantial implications for earnings realised over a player’s career. We find subtle but significant differences in retention rates between Black and White players (approximately 3 percentage points) that can’t be explained by a comprehensive set of individual characteristics including their productivity. We also show that traditional wage gap estimates, which appear to show equal earnings between Black and White players conditional upon playing position and productivity, mask underlying
disparities in career earnings that become apparent when adjusting for these unequal retention rates.