[Excerpt] We present new theory and the first empirical test of promotion discrimination models based on job assignment signaling. In our theory, promotions serve as signals of worker ability, and job hierarchies differ in the degree to which tasks vary across hierarchical levels. When tasks differ substantially across levels, the opportunity cost (in terms of foregone output) of not promoting qualified workers from a disadvantaged group (e.g. racial minorities or females) is large, so employers are less likely to (inefficiently) retain such workers in lower-level jobs. Thus, given performance in the pre-promotion job, the extent to which disadvantaged workers have lower promotion probabilities than advantaged workers should decrease when tasks vary more across hierarchical levels. Also, the difference between the favored and disfavored groups in the wage increase attached to promotion should diminish when tasks vary more across hierarchical levels. We test these implications empirically for the case of racial discrimination in promotions, using personnel data from a large U.S. firm and also data from the National Compensation Survey. We find strong empirical support for the theoretical model’s predictions concerning promotion probabilities, whereas empirical support is mixed for the model’s predictions concerning the wage growth attached to promotions.