Ethnic Preferences and Residential Segregation: Theoretical Explorations Beyond Detroit

We are strongly supportive of Fossett's theoretical approach and modeling methodology, which uses computational methods to perform thought experiments that generate compelling insights into the enigma of persistent residential segregation in the U.S. We also agree with his theoretical results, which challenge the prevailing view among demographers that institutional discrimination is the essential cause. However, we think he did not go far enough. Fossett limited his analysis to a narrow region of the parameter space that corresponded to conditions observed in one city at one time. This precludes generalization to other times and places and exploration of theoretically motivated “what if” scenarios that trespass beyond the Detroit city limits. When we extended the parameter space, we noticed two interesting results. First, Fossett's “paradox of weak minority preferences” requires qualification. Disproportionate in-group preferences among minorities are indeed segregation-promoting, not integration-promoting, but they generally have less impact on segregation than the in-group preferences of the majority. Second, not only are exclusionary practices and institutional discrimination not necessary for segregation (as Fossett demonstrates), we show that in certain regions of the parameter space they are not even sufficient.

Keywords:

• segregation
• discrimination
• agent-based modeling

Notes

¹Fossett does not explore what happens if search is constrained by distance. Assuming spatial correlation in ethnic distribution, local search should make it more difficult for agents to find more ethnically diverse places to live.

²Charles (2003: 185) observes how “preferences for integration decline as the number of out-group members increases.” Fossett does not explore that possibility in this paper, but Charles' empirical observation suggests an intriguing possibility for future theoretical research.

³According to Fossett (private communication), newer versions of SimSeg do have this option, but these are not publicly available.

⁴Fossett implements noise as “forced moves” induced by evictions, job changes, or the like. However, he did so to make the model more realistic, and did not consider the theoretical importance of forced moves for perturbing an unstable equilibrium so that more stable solutions might obtain. To that end, noise could also be implemented by replacing Fossett's deterministic rule for moving with a stochastic rule in which the probability of moving increases monotonically with the increase in utility.

⁵We introduced neighborhood heterogeneity using Voronoi diagrams (Flache & Hegselmann, 2001) instead of regular grids.

⁶We measured segregation using Fossett's modified index of dissimilarity, which is normed to be margin free.

⁷Fossett does point out, however, that these practices are not sufficient to explain segregation between ethnic minorities.