How people are exposed to neighborhoods racially different from their own

Abstract

In US cities, neighborhoods have long been racially segregated. However, people do not spend all their time in their neighborhoods, and the consequences of residential segregation may be tempered by the contact people have with other racial groups as they traverse the city daily. We examine the extent to which people's regular travel throughout the city is to places "beyond their comfort zone" (BCZ), i.e., to neighborhoods of racial composition different from their own-and why. Based on travel patterns observed in more than 7.2 million devices in the 100 largest US cities, we find that the average trip is to a neighborhood less than half as racially different from the home neighborhood as it could have been given the city. Travel to grocery stores is least likely to be BCZ; travel to gyms and parks, most likely; however, differences are greatest across cities. For the first ~10 km people travel from home, neighborhoods become increasingly more BCZ for every km traveled; beyond that point, whether neighborhoods do so depends strongly on the city. Patterns are substantively similar before and after COVID-19. Our findings suggest that policies encouraging more 15-min travel-that is, to amenities closer to the home-may inadvertently discourage BCZ movement. In addition, promoting use of certain "third places" such as restaurants, bars, and gyms, may help temper the effects of residential segregation, though how much it might do so depends on city-specific conditions.

Keywords: cell-phone data; diversity; everyday mobility; racial segregation; third places.

Fig. 1.

Most visits to POIs are to those at the top of the distribution. Median visits per device (blue) and POIs per capita (orange) for the top 16 across cities. Restaurants, shops, and malls account for most traffic. Data for 2019, 100 largest US cities.

Fig. 2.

Which urban places bring people “BCZ” depends on the city. BCZ scores for visits to POIs, for each category and city (gray circles). Median values across cities in color. Differences across POIs for the median city are small; differences within POIs across cities are large. Data for 2019, 100 largest US cities.

Fig. 3.

The farther the POI, the more BCZ its neighborhood is—for the first ~10 km. Each line represents one city. Median value across cities in black. To reduce noise, bins in each city’s top decile and distances greater than 50 km are dropped. For the first ~10 km, BCZ score increases with distance, leveling off on average thereafter. However, beyond ~10 km, heterogeneity across cities increases dramatically. For example, four cities with similar residential segregation levels—Long Beach (0.20), Buffalo (0.35), Dallas (0.35), and Louisville (0.28)—exhibit widely different BCZ scores beyond 10 km. Data for 2019, 100 largest US cities.

Fig. 4.

Regardless of POI, people visit POIs in neighborhoods less dissimilar than they could have visited. Racial dissimilarity measured on 0 to 1 scale. Separate models by POI; each model includes only neighborhoods with at least one focal POI. Results based on models with and without controls for number of residents, median age, median household income (log), proportion of owners, proportion of married families, proportion of residents with a bachelor’s degree or greater, and distance (in km)—in both visiting (i) and visited (j) neighborhoods—with city fixed effects. Bars are 95% CI from city clustered SE. Regardless of POI, visits exhibit preference for similar neighborhoods. Observed preference for similarity is strongest when visiting grocery stores and weakest when visiting gyms. Data for 2019, 100 largest US cities.

Fig. 5.

Top. Do people visit POIs in dissimilar neighborhoods without encountering dissimilar people? (A) Neighborhood i visits similar POI k in dissimilar neighborhood j. (B) Neighborhood i visits dissimilar POI k in dissimilar neighborhood j. Travel of type (B) is more accurately described as beyond the comfort zone. Bottom. The more dissimilar (from home) the POI’s neighborhood is the more dissimilar the POI itself is. In both axes, racial dissimilarity is measured in comparison to the composition of home neighborhood i. The x axis compares i to neighborhood j where POI k is located; the y axis compares i to k itself, wherein the racial composition of k is a weighted average of that of the neighborhoods where other visitors to k live. See text for details. Data for 2019, 100 largest US cities.