What causes racially segregated residential communities? Testing agent-based and institutional explanations for the locational decisions of homebuyers
Nelson, Michael Henry
Neighborhoods across the United States are still segregated along racial lines. This dissertation contains three linked analyses that test the most prominent explanations for contemporary racial residential segregation. The analyses use spatial hierarchical linear modeling of home loan application data in Nashville to look at the relative strengths of ethnocentric preference, socioeconomic, and institutional racial bias explanations of segregation. The first analysis tests agent-based ethnocentric preference models of segregation by focusing on home loan applications by race and their connection to neighborhood racial composition (NRC). Previous research, based on mathematical models, has shown that contemporary segregation could be fully explained by slight ethnocentric preferences among households. In contrast to this, the results suggest that mixed and minority neighborhoods would become whiter, while the whitest neighborhoods are predicted to have in-movers that were less white than the existing NRC. The second analysis examined the role of lending institutions and whether loan approval decisions had an impact on segregation. While lending decisions did not have a consistent impact on the racial composition of in-movers in tracts across time, there was a non-linear statistical association between neighborhood percent white and the extent that lenders favored white applicants. The third analysis looked at whether lenders show racial bias in their decisions to make home loans. It tested three forms of racial bias: discrimination based on the race of applicant, discrimination based on NRC (“redlining”), and discrimination based on the interaction between race of applicant and NRC (“geographically contingent lending”). White applicants were consistently favored as were loans for mixed race neighborhoods. There was mixed evidence of geographically contingent lending. All three analyses control for socioeconomic factors. The results showed that preference theories do not provide an adequate explanation of the aggregate behavior of homebuyers in Nashville; white applicants are more likely to have loans approved when compared to non-whites; NRC has an impact on lending decisions, but it is not linear; and, socioeconomic factors do not seem to driving racial residential segregation.