Major “Combination-Patterns” of Residential Segregation Based on Five Dimensions of Segregation: Latent Profile Analysis

Han Soo Woo, Johns Hopkins University

Literature on residential segregation has traditionally focused on using just two domains of segregation, evenness and exposure, while making less use of concentration, centralization, and clustering. When studies do invoke all five dimensions of segregation, discussions have typically centered on “hypersegregation”--a pattern in which all five measures of segregation domains are simultaneously high. Nevertheless, studies demonstrate distinct consequences (e.g., crime rate, mortality, etc) associated with a large array of non-hypersegregated combination-patterns of residential segregation. This paper attempts to get a clearer purchase on what combination-patterns of segregation exist in the U.S. cities and what meanings we are to make of these patterns. Based on 380 metropolitan areas (US Census 2000), I use Latent Profile Analysis—a class of finite mixture modeling—to reduce 380 unique combination-patterns of segregation into an essential few for blacks, Hispanics, and Asians. The study shows that all integrated cities are alike but that segregated cities are segregated in their own but few dominant ways.