Divided Spaces and Divided Opinions: Modeling the Impact of Residential Segregation on Opinion Polarization

Today we observe widespread polarization of attitudes around controversial topics such as immigration, but we know little of how we got here. This dissertation addresses the questions: are polarized attitudes more likely to develop in ethnically segregated or non-segregated societies? And why?
Existing theories offer conflicting answers, and the available empirical evidence, limited to observational approaches, is mixed. To better understand the segregation-polarization link we need further theoretical refinement. I use computer simulation experiments to study the mechanisms through which attitudes polarize, and how segregation interacts with these mechanisms. This allows me to compare alternative theories and determine where and under what conditions they yield conflicting predictions.
Simulations reveal that the predicted sign of the segregation-polarization link hinges on the assumed behavioral model – that is, how we expect individuals to change their attitudes when influenced by others. Some models predict that segregation fosters polarization; and some predict the reverse. Generally, predicted effects are stronger when social influence is local (meaning that neighbors influence each other more than non-neighbors), and when ethnic membership shapes the interactions (meaning that individuals from the same ethnic group influence each other differently than individuals from different groups).
In developing adequate simulation experiments I encounter and address various methodological challenges – from the perils of formalizing behavioral theories initially expressed in natural language; to making the simulation models sufficiently realistic by seeding them with empirical data; to mitigating the computational bottleneck of high-dimensional, large-scale simulation experiments.