Research

Abstract: Extensive research finds that place-based investment reduces crime, leading practi- tioners to propose it as an alternative to police-centered policies. We explore another channel linking local investment to crime—that police patrol is endogenous to the built environment-using smartphone location data. Exploiting quasi-experimental variation in HUD rules designating Qualified Census Tracts (QCTs), we find police increase patrol in QCTs enough to explain all the observed violent crime reduction. Police increase patrol more in neighborhoods with more Black residents and fewer recently-built units. Our findings highlight the importance of understanding police response to local development before framing it as a substitute for policing.

Abstract: Research on policing has focused on documented actions such as stops and arrests—less is known about patrols and presence. We map the neighborhood movement of nearly ten thousand officers across 21 of America’s largest cities using anonymized smartphone data. Police spend 0.36% more time in neighborhoods for each percentage point increase in Black residents. This neighborhood-level disparity persists after controlling for density, socioeconomic, and crime-driven demand for policing, and may be lower in cities with more Black police super- visors (but not officers). Patterns of police presence statistically explain 57% of the higher arrest rate in more Black neighborhoods.

Abstract: Using newly available smartphone mobility data, several recent studies document partisan gaps in compliance with COVID-19 social distancing orders. By relying on cross-county variation in election results and mobility measures, these studies may fail to isolate partisan effects from contemporaneous COVID trends due to spatial confounds: mobility and voting data are aggregated to the county-level, and nine of the ten earliest-hit counties were heavily Democratic. We extend these analyses by merging individual-level geolocation data for ten million smartphones with nationwide precinct-level vote counts, allowing us to estimate partisan differences in compliance with COVID restrictions by comparing Democratic and Republican voters living in close proximity and within the same county. By May 2020—when ninety percent of Americans were under a stay-at-home order—likely Clinton voters reduced travel outside the home by twice as much as their Trump-voting neighbors, and were significantly more likely to remain at home throughout the day. This gap survives aggressive geographic controls and remains one of the strongest predictors of stay-at-home compliance, confirming that strong partisan effects are not the result of spatio-temporal covariates.