A New Measure of Urban Racial Segregation

Abstract: https://journals.sagepub.com/doi/full/10.1177/0042098020978965#abstract

Racial residential segregation, the “linchpin of racial stratification,” has been a durable feature of American cities and a focus of much empirical inquiry since the early twentieth century. While research on racial segregation in cities has grown rapidly over the last several decades, its foundation remains the analysis of the neighbourhoods where people reside.                         

Contact between racial groups depends not merely on where people live, however, but also on where they travel over the course of everyday activities. The neighborhoods in which people live are not the sole sites of daily interactions—people work, shop, find entertainment, and participate in multiple activities in different neighbourhoods throughout a city. 

To capture this reality, our study introduces a mobility-based measure of racial segregation—the segregated mobility index (SMI)— which offers a novel perspective on segregation based on the everyday travels of city residents, as well as the structural connectedness of neighbourhoods that these daily flows produce. The dynamic index of segregated mobility that we propose provides new insights into the social organisation of cities beyond residential neighbourhoods. 

Capturing neighbourhood connectedness via movement first requires fine-grained location data on people’s neighbourhood of residence and their travel across neighbourhoods in large U.S. cities, which until recently have been unavailable. A number of data sources, at least in principle, could capture everyday movement, such as from social media posts, cell phone tracking, and largescale surveys that ask about travel activity. Our study relies on publicly available Twitter data in the form of hundreds of millions of geotagged tweets sent by over 375,000 Twitter users in the 50 largest U.S. cities from March 2013 to November 2015.

To measure how people travel through cities, we need to know where they live and which neighbourhoods they visit. To identify home locations, we use a machine learning algorithm to estimate the residential block group for each user. We then observe how users travel throughout the city by tracking where they tweet outside of their home neighbourhood. With this information, we build mobility networks for each city with neighbourhoods as the nodes and the travels of neighbourhoods’ residents between neighbourhoods as the ties.  We aggregate this information to calculate the level of movement between different types of neighbourhoods.

 In our study, the segregated mobility index (SMI) captures the extent to which neighbourhoods of given racial compositions are connected to other types of neighbourhoods in equal measure. The racial segregation of a city becomes the extent to which residents fail to travel to different types of neighbourhoods with varying racial/ethnic compositions, controlling for the racial composition of a city’s neighbourhoods.

We also illustrate, through empirical models, the utility of our measure in capturing a distinct element of racial segregation, one that it is related to, but not solely a function of, residential segregation. We find that a city’s racial composition also matters; minority group threat, especially in cities with large Black populations and a troubled legacy of racial conflict, appears to depress movement across neighbourhoods in ways that produce previously undocumented forms of racial segregation.

Collectively, the results from our study underscore the importance of viewing segregation as multidimensional and dynamic. Spatial inequality in cities permeates through multiple domains, reaching well beyond residential neighbourhoods by shaping residents’ lived experiences.

The SMI, which could be constructed using other data sources, such as cell phone records, expands the possibilities for studying dynamic forms of racial segregation including their effects and shifts over time. While our study categorises neighbourhoods by their majority racial/ethnic composition, future work that uses SMI could also consider different neighbourhood classifications (e.g., via socioeconomic factors). We encourage future research that incorporates SMI and provide detailed methodological guidance in the paper for doing so.

Read the accompanying article on Urban Studies OnlineFirst here.