Although sociology has a long tradition of research documenting patterns of segregation in American cities, the causes of segregation are not well understood. The literature identifies three potential explanations for segregation: (1) individuals' preferences to live among their own group or avoid other ethnic groups; (2) economic inequalities among ethnic groups coupled with housing cost differences and affordability constraints; and (3) institutional discrimination on the part of real-estate agents and lenders. Empirical research documents survey respondents' level of tolerance for integrated neighborhoods, economic inequalities among ethnic groups, and differential treatment of minorities on the part of real-estate agents and lenders. But these studies do not demonstrate how and to what extent these factors contribute to existing patterns of segregation.

Residential mobility and residential segregation is one instance of a larger class of social problems in which individuals both respond to some population statistic (in this case, neighborhood composition and housing prices) and contribute to that statistic. For example, every white person who moves out of a neighborhood because she cannot tolerate its proportion black leaves that neighborhood with a slightly higher proportion black (and slightly increases the proportion white of the neighborhood she moves into). Thus, there is feedback between the behavior of individuals and features of their environment. When such feedback exists, there is no linear or obvious relationship between the preferences or behavior of individuals and aggregate population processes. So one cannot assume, for example, that a given level of ethnic tolerance will generate a corresponding level of segregation; this relationship must be modeled directly.

This dissertation uses statistical models to examine the role of ethnic and economic factors in residential mobility decisions. It then incorporates these models into an agent-based model of residential turnover to show how these decisions affect segregation dynamics. My research shows that residential mobility based on ethnic factors alone can generate high or low levels of segregation depending on the spatial context. Sorting on economic factors alone tends to result in low levels of segregation. When individuals sort on both ethnic and economic characteristics, they have offsetting effects. The magnitude of these effects varies across ethnic groups. Most notably, sorting on economic factors appears to have the largest desegregating effects for blacks. I discuss the extent to which these results can explain segregation processes in real cities, and sugguest some avenues for future work.