We exploit a policy designed to randomly allocate roommates in a large South African university to investigate whether interracial interaction affects stereotypes, attitudes and performance. Using implicit association tests, we find that living with a roommate of a different race reduces White students' negative stereotypes towards Black students and increases interracial friendships. Interaction also affects academic outcomes: Black students improve their GPA, pass more exams and have lower dropout rates. This effect is not driven by roommate's ability.

We propose a model to study when an intermediate action can serve as a stepping stone that enables the elimination of a harmful norm. While the intermediate action may facilitate the first “step,” it may also become a new norm. We derive intuitive conditions for stepping stones, which depend on the relative size of social penalties and intrinsic utility benefits. We propose an econometric approach to testing whether an intermediate action is a stepping stone, and apply it to original data on female genital cutting in Somalia. The analysis shows that the intermediate action may become the new norm.