Regional residential segregation
Numerous measures of residential segregation are aligned with the various dimensions of segregation. We explored the relationship between a measure of concentration calculated from a Gaussian kernel smoothing of ethnoracial composition at the census tract level within counties. Analyses were stratified by census regions (Figure 2 ). Residential segregation measures varied substantially by subgroup, with the smallest range of values for Asian people and the largest range for white people. In New England, we observed a roughly monotonic increase in the number of CDDs with higher concentrations of Asian and Latino people. The relationship was non-monotonic for Black people, with an increase until an inflection point, and then fewer CDDs in the highest concentration tracts. In the mid-Atlantic, higher concentrations of Black and Latino people were associated with higher CDDs. In the South Atlantic, a non-monotonic relationship was apparent for Black people, with concentrations of 25%–75% Black having higher CDDs, and confidence intervals >90% crossing zero, meaning that the most segregated areas for Black people were no different than their county average. A positive association was roughly monotonic for Latino people. While non-monotonic in the South Atlantic, negative associations were apparent in all three regions with high concentrations of white people. There was some heterogeneity in the shape of the associations based on analyses by state (Figures S2–S4 ).