Poisson Indices of Segregation

Existing indices of residential segregation are based on an arbitrary partition of the city in neighborhoods: given a spatial distribution of racial groups, the index provides different levels of segregation for different partitions. This paper proposes a method in which individual locations are mapped to aggregate levels of segregation, avoiding arbitrary partitions. Assuming a simple spatial process driving the locations of different racial groups, I define a location-specific segregation index and measure the city-level segregation as average of the individual index. After deriving several distributional results for this family of indices, I apply the idea to US Census data, using nonparametric estimation techniques. This approach provides different levels and rankings of cities' segregation than traditional indices. I show that high aggregate levels of spatial separation are the result of very few locations with extremely high local segregation. I replicate the study of Cutler and Glaeser (1997) showing that their results change when segregation is measured using my approach. These findings potentially challenge the robustness of previous studies about the impact of segregation on socioeconomic outcomes.