We analyze the relationship between spatial location and segregation between Estonian speaking and Russian speaking population in Estonia based on mobile phone records. Empirically, we are thus able to estimate the geographic segregation of each individual's place of residence, the social segregation of each individual's social network, and migrations
This paper makes three primary contributions. First, we document a strong correlation between the ethnic composition of an individual's physical surroundings and the ethnic composition of his or her social network. This effect is attenuated among migrants, who are more likely to interact with co-ethnics, but who are less influenced by their immediate physical surroundings.
Second, we develop a structural model to interpret the empirical results. The results suggest that both ethnic and geographic distance weight prominently in the tie formation, and that being of the same ethnicity is roughly equivalent to a 10-fold reduction in physical distance.
Finally, and more speculatively, we use the structural model to simulate the counterfactual segregation patterns in Estonia. These simulations show that in a context of cities, the ethnic preferences are the main determinant of segregation while geographic location dominates for more distant ties. We conclude by suggesting how policies that promote physical integration might affect social integration.
This is work with Josh Blumenstock.