Is Segregation Robust?

This paper studies the question of how well we understand segregation. The point of departure is Schelling’s spatial proximity model in one dimension. By introducing noise I show that segregation emerges as the long run prediction of neighborhood evolution, both when residents have Schelling-type threshold preferences and strict preferences for diversity. Analytical result are complemented with numerical simulations which show that within a reasonable time frame full segregation does not occur. When residents have a preference for diversity, I show that a natural perturbation away from the diversity monomorphism dramatically alters the long run prediction: integration is the unique long run prediction, even in the absence of noise.