Including Preference and Social Distance Dynamics in Multi-Factor Theories of Segregation

All agree that many factors contribute to ethnic segregation, but controversy continues regarding the relevance of particular factors. In my paper I acknowledge the importance of discrimination, but I direct attention to the role of ethnic preferences and social distance dynamics. I do so because many dismiss the relevance of these factors for contemporary segregation without engaging formal theories and analytic models which suggest that dynamics associated with in-group attraction and out-group avoidance may take on increasing importance as past forms of discrimination slowly fade. Agent-based modeling shows promise for exploring the issue of whether prejudice against out-groups and affinity toward in-groups can build and sustain segregation in the absence of discrimination. Research drawing on this approach is at an early stage of development but is significant for highlighting two things. One is that the implications of preferences for segregation are strongly conditioned by the ethnic demography of the city. Another, closely tied to the first, is that different views about the implications of preferences for segregation often hinge on inconsistencies in how notions of integration and segregation are applied in discussions of individual location choices, the ethnic mix of single neighborhoods, and the ethnic distributions for all neighborhoods in a city. Critiques of agent-based models of Schelling-style preference effects will carry more force when they outline models indicating how location decisions guided by preferences documented in surveys can produce or sustain integration.

Keywords:

• agent-based model
• simulation
• ethnic segregation
• urban ecology
• prejudice
• ethnic preferences
• social distance

Notes

¹Goering states that my definition of discrimination fails to include some aspects of discrimination. Since my paper does not report simulations that model discrimination dynamics, it is unclear to me why this is relevant. When my simulation program is used to model discrimination, it can produce very high levels of segregation based on a handful of discrimination processes (e.g., exclusion, steering, and differential treatment in loan qualifying). I am unaware of a crucial substantive point about discrimination that cannot be illustrated using my model.

²Tipping is not “handled” in my simulations, it occurs.

³Macy praises the model on this count. This appears to highlight a basic difference in Goering's and Macy's views about theory and research strategy. I share many of Macy's views.

⁴It is a simple matter to demonstrate this by activating and deactivating discrimination dynamics in the simulation exercises similar to those I report in my paper. When discrimination dynamics are activated, segregation increases to even higher levels. When they are deactivated, segregation drops back to lower but still high levels.