Complexity in patterns of racial segregation

Abstract: Cities are complex systems, their complexity manifests itself through fractality of their spatial structures and by power law distributions (scaling) of multiple urban attributes. Here we report on the previously unreported manifestation of urban complexity -- scaling in patterns of residential racial segregation. A high-resolution racial grid of a city is segmented into racial enclaves which are patches of stationary racial composition. Empirical PDFs of patch areas and population counts in 41 US cities were analyzed to reveal that these variables have distributions which are either power laws or approximate power laws. Power law holds for a pool of all patches, for patches from individual cities, and patches restricted to specific racial types. The average value of the exponent is 1.64/1.68 for area/population in 1990 and 1.70/1.74 in 2010. The values of exponents for type-specific patches vary, but variations had decreased from 1990 to 2010. We have also performed a multifractal analysis of patterns formed by racial patches and found that these patterns are monofractal with average values of fractal dimensions in the 0.94-1.81 range depending on racial types and the year of analysis. Power law distribution of racial patch sizes and a fractal character of racial patterns present observable and quantifiable constraints on models of racial segregation. We argue that growth by preferential attachment is a plausible mechanism leading to observed patterns of segregation.