Safety in Numbers? Evidence on the Relationship Between Crime and Mobility from American Cities During the COVID-19 Pandemic

Abstract

Traditional theories of crime present seemingly contradictory predictions concerning how criminal activity should respond to changes in mobility. In this paper, I draw on these existing frameworks to explore the impact of mobility upon crime by analyzing cellular phone data during the fall of 2019 through the onset of COVID-19 in five large American cities (Atlanta, Boston, Chicago, Philadelphia and San Francisco). Using panel data, with observations collected at the census block group level, I find that mobility overall is positively related to crime; this pattern exists both when I use an ordinary least squares regression with fixed effects and when I run a two-stage least squares regression, instrumenting for mobility. Further, I introduce a couple difference-in-differences regressions which study the effects of the relaxation of lockdown policies upon crime. I find a \emph{negative} average treatment effect from these reopening policies, suggesting a non-linearity in the relationship between crime and mobility.