The Paradox of Segregation and Cohesion

Residential segregation, whether by race or class or some other characteristic, is often viewed as a problem. Decades of sociological research has documented the causes and (typically detrimental) consequences of residential segregation, often with the implicit goal of reducing segregation and promoting the formation of integrated neighborhoods. At the same time, neighborhood cohesion or what is sometimes called a “sense of community” is often viewed as a key to strong communities that can work together to solve local issues. Again, decades of research has examined what makes some neighborhoods more cohesive than others, often with the implicit goal of promoting the formation of cohesive neighborhoods characterized by strong social bonds among its residents.

But, there is a paradox here. Desegregated communities often have lower levels of cohesion, and tend to be characterized by more fragmented social networks. Likewise, the most cohesive neighborhoods tend to be the most segregated and homogenous. Dr. Zachary Neal (sociology) and Dr. Jennifer Neal (psychology) are using agent-based models, implemented in NetLogo, to understand why.

Their simulation models begin by creating hypothetical neighborhoods with varying levels of segregation. Within each of these hypothetical neighborhoods, the residents form relationships with one another according to three simple rules. First, homophily: the formation of a friendship between two people depends on their similarity; typically friendships form between similar individuals. Second, proximity: the formation of a friendship between two people depends on the physical distance between them; typically friendships form between nearby individuals. Third, transitivity: the formation of a friendship depends on the overlap in their existing social circles; typically friendships form between people who already have mutual friends. Various combinations of these three rules (e.g. strong preference for homophily, weak preference against proximity, and no preference for or against transitivity) define different social network formation scenarios.

The goal of this project is to understand what network formation scenarios make it possible to simultaneously minimize residential segregation and maximize neighborhood cohesion. Of particular interest is the likelihood of such scenarios, that is, is the goal of reducing segregation and increasing cohesion realistic given the way people typically form relationships. Although the simulations themselves are relatively simple, grounded in small hypothetical neighborhoods populated by agents following very basic rules, this type of analysis is not possible without the resources available from HPCC. Network statistics are computationally intensive, and understanding the relationship between segregation and cohesion requires considering the full range of combinations of segregation, homophily, proximity, and transitivity. Capitalizing on the large number of parallel processors available through HPCC has allowed this research to move from the conceptual stage focused on actual neighborhoods, to a more comprehensive analysis of all possible neighborhoods.