Bridging Simulation and Inference Methods in the Study of Residential Segregation: A Case Study

Miruna Petrescu-Prahova, University of California, Irvine

In this paper we conduct a two-stage analysis of the Yuma, AZ, metropolitan area. First, we simulate a scenario with covariates from Census data, where we vary the parameter values to analyze the impact of various effects on segregation patterns, and obtain a range of possible Yuma residential configurations. Second, we estimate the relative contribution of effects such as homophily, xenophobia or push-pull between rent and income levels in determining the observed distribution of households across tracts. Exploring the range of alternate residential pattern configurations as well as the observed configuration allows us to understand how even slight changes in some parameters could affect the overall assignment of households to tracts and therefore segregation levels in this area. This combination of simulation and inference gives us a powerful tool for understanding residential segregation processes, while at the same time suggesting implications for public policy aimed at reducing segregation.