Our experiment shows that subjects incorporate irrelevant group information when evaluating others: Individuals from on average badly performing groups receive low evaluations, though it is known that they themselves perform well. This group-bias occurs in a gendered setup, where females form the on average worse performing group, as well as in a non-gendered setup.
The type of discrimination we identify is neither taste-based nor statistical: It is due to conservatism in updating beliefs, which is even more pronounced in females. Furthermore, self-confident males overvalue male performers. Using our data to simulate a job promotion ladder, females get driven out quickly.