Statistical discrimination with peer effects: Can integration eliminate negative stereotypes?

We introduce peer effects in the costs of human capital acquisition into a model of statistical discrimination in labour markets. This creates a link between the level of segregation in social networks and racial disparities in job assignment and wages. We show that this relationship is characterized by discontinuities: there is a threshold level of segregation below which negative stereotypes become unsustainable, and steady-state skill levels can change dramatically. This change can work in either direction: skill levels may either rise or fall in both groups. Which of these outcomes arises depends on the population share of the disadvantaged group and on the distribution of the costs of human capital investments. We also examine the effects of affirmative action policies in the presence of peer effects and provide conditions under which such policies eliminate negative stereotypes.