Abstract
This paper investigates the effects of city size, shape, and form, and neighborhood size and shape in agent-based models of residential segregation. We find that, in many key respects, model-generated segregation outcomes are not influenced in important ways by variation in these factors. For example, the expression of segregation based on agent preferences for coethnic contact in agent-based models does not vary with city size, city shape, city form, or the shape of neighborhoods involved in agent vision, or the use of distance-decay functions for evaluating neighbors. These findings indicate that results obtained from model-based segregation studies are likely to be robust relative to choices regarding these aspects of model specification. We do find important effects of the size or scale of neighborhoods involved in agent vision: model-generated segregation outcomes vary in complex ways with agent vision. Significantly, however, the effect of neighborhood size or scale does not appear to vary in important ways with neighborhood shape. Thus, what is important is the scale of agent vision—that is, the number of neighbors they ‘see’—not the particular spatial arrangement of those neighbors. With the exception of the effect of the spatial scale of agent vision, our results suggest that researchers can generally presume that their findings regarding how model-generated segregation outcomes vary with substantive factors, such as agent preferences for coethnic contact or the ethnic demography of the city, are not contingent on choices regarding model implementation of city size, city shape, city form, and neighborhood shape. These findings are welcome because they suggest that simulation studies can devote less attention to technical specification choices and more attention to assessing substantive questions regarding the effects of social dynamics and sociodemographic distributions in the context of model systems.
Get full access to this article
View all access options for this article.