This proposal outlines a plan of research to better understand the causes of racial and economic residential segregation in American cities. We propose to develop data-based models of household residential mobility and to use these models as a basis for a realistic, agent-based model of neighborhood formation and residential segregation. Current methods used to study residential segregation have examined specific single processes linked to segregation. These methods usually rely on intuitive judgments to draw conclusions about how these processes aggregate to form segregated neighborhoods. This is problematic because stylized simulation models show feedback and constraints in residential systems produce relations of inputs and outcomes that are highly interactive and non-linear. Our alternative approach is based on discrete choice models of destination selection in residential relocation and an agent-based simulation. We estimate the discrete choice models using the data on mobility from the Panel Study of Income Dynamics matched with data from the decennial censuses. Our proposed basic model incorporates race and neighborhood racial composition, income and housing cost, neighborhood income composition, and tenure (owner or renter). We also propose an extension to the model that incorporate wealth, housing market discrimination, and impacts of neighborhood change in the residential neighborhood and adjacent neighborhoods. The base model we propose could incorporate other future refinements toward greater realism and usefulness. The model has two uses. The first use is to address basic science questions regarding the causes of segregation. The second use is to evaluate the effects of spatially targeted housing policies on race and income segregation. We use the model to evaluate the effects of the shift from traditional fixed-site public housing to housing vouchers and mixed-income housing on race and income segregation.