Diversity and popularity in social networks

Abstract: We present a new model to understand the nature and structure of homophily in social networks. We introduce heterogeneity to the study of stochastic network formation and show that meeting friends of friends tends to reduce social segregation. This allows us to derive sharp theoretical predictions on homophily patterns. Our main result is that individuals with more connections tend to have more diverse networks. Our theoretical predictions are well supported empirically when looking at friendships between boys and girls in U.S. high schools. We derive some welfare implications and show that our analysis can help identify the source of homophily