Using data from a representative 1989 household survey for Italy we estimate random-coefficient earnings frontiers by gender, marital status and location. These estimates are used to calculate discrimination indices. Our results show that fixed coefficients can be rejected in all cases. A wide range on the estimated coefficients indicates a high degree of variation in the quality of the observed human-capital variables as well as different degrees of ability as perceived by the employer. We find reverse discrimination for single females in the South and the North. For married females there is evidence of discrimination, particularly in the South. We isolate the effects of tenure and education on discrimination and find that these reduce discrimination for Southern-married females.