UCLA

Neighborhood racial composition preferences have the potential to produce extreme residential segregation between racial groups. Whites' preferences to avoid Blacks, and Blacks' preferences to maintain residential contact with their own group can induce high levels of White-Black segregation. Similar preferences can lead to segregation for Asians and Latinos. These theoretical facts have lead some observers to attribute continued residential segregation in the United States to preferences. These observers imply that elimination of housing market discrimination, socioeconomic differences, and other discrepancies between groups in metropolitan housing markets, would do little to lessen residential segregation by race. Most support for this perspective 1) relies on caricatures of empirically observed preferences, or 2) fails to examine whether preferences align with actual housing choices. These limitations are problematic. A failure to include realistic assessments of preferences in simulation models can lead to an overstatement of the potential for preferences to generate segregation, and a corresponding understatement of the potential for housing market discrimination and other race-related factors to induce and sustain segregation. This dissertation addresses the limitations of previous attempts to vet the importance of preferences for segregation. It develops an appropriate statistical model for comparing stated racial composition preferences to actual neighborhood attainments. It uses this model to assess whether people migrate in ways that agree with their preferences, or whether some groups are systematically frustrated in matching their neighborhood attainments to their preferences. Finally, it combines results from these empirical assessments in simulation models of segregation. When some groups are stymied in migrating according to their stated preferences, does this tend to result in higher levels of segregation? Using data from Los Angeles County, this dissertation shows that Latinos, and to a lesser degree, Blacks, are disadvantaged relative to Whites in matching their residential preferences to their attainments. In simulation, these frustrated preferences yield excess levels of racial segregation: If people were able to migrate according to their stated preferences, levels of residential segregation would be approximately 20% lower. These results illustrate that preferences set a floor for levels of segregation, but do not fully explain extant segregation. Room remains for explanations based on discrimination and social networks.