Do neighborhoods have boundaries? A novel empirical test for a historic question

Abstract

Do neighborhoods have boundaries? Scholars have debated how neighborhoods should be operationalized for decades. While recent scholarship has de-emphasized boundaries, I argue that boundaries are focal to understanding what neighborhoods are and why they are so segregated. Relying on everyday mobility patterns data from a panel of 45 million nationally representative devices, I demonstrate that divisions between contiguous census block groups in terms of everyday mobility patterns align with divisions in race, educational attainment, occupation, and age. Employing a novel clustering procedure, I further demonstrate that sets of census block groups can be easily separated in terms of differences in mobility patterns, indicating that meaningful clusters and boundaries within cities do exist. Subsequent results indicate these clusters are uniquely segregated in terms of race, educational attainment, and age, highlighting how other spatial aggregations can underestimate true segregation. Additional results suggest that unique social processes divide these clusters from one another, as indicated by exceptional variation in both COVID-19 case incidence and criminal offense timing. While I do not believe these clusters represent objective “neighborhoods,” I do assert that they can serve as a useful geographical unit for social analyses. These clusters may also be useful for measuring segregation in mobility patterns as well as for studying mobility network resiliency.