The immense literature on discrimination treats outcomes as relative: One group suffers compared to another. But does a difference arise because agents discriminate against others - are exophobic - or because they favor their own kind - are endophilic? This difference matters, as the relative importance of the types of discrimination and their inter-relation affect market outcomes. Using a field experiment in which graders at one university were randomly assigned students' exams that did or did not contain the students' names, on average we find favoritism but no discrimination by nationality, and neither favoritism nor discrimination by gender, findings that are robust to a wide variety of potential concerns. We observe heterogeneity in both discrimination and favoritism by nationality and by gender in the distributions of graders' preferences. We show that a changing correlation between endophilia and exophobia can generate perverse predictions for observed market discrimination.
favoritism discrimination field experiment wage differentials economics of education