More Eyes, (No Guns,) Less Crime: Estimating the Effects of Unarmed Private Patrols on Crime Using a Bayesian Structural Time-Series Model

This work studies the effect of unarmed private security patrols on crime. We make use of a initiative, triggered by an arguably exogenous events, consisting in hiring unarmed private security agents to patrol, observe and report to ordinary police criminal activities within a well-defined city area. Our identification strategy capitalizes on the fact that the portions of the city outside the arbitrarily defined intervention area remain unaffected by the patrolling activity. To estimate the effects of the security patrols, we use both a difference-in-difference approach and we additionally propose the first application in Law & Economics of a Bayesian structural time-series model (Brodersen et al., 2015). This model overcomes some limitations and provides a generalization to the time-series setting of the standard difference-in-difference approach. Results show that unarmed private security patrolling decreases crime in the treated area by 30-43%. Our results suggest that a large share of the police-crime elasticity estimated by prior work is due to perceptual deterrence.