Capturing Resource Deployment Variation During a True Experiment in Micro-Policing and Property Crime

• In Event: Crime and Place: Philadelphia Predictive Policing Experiment: An Update

Abstract

Given variations in resource constraints across a large police department, and distinct police district cultures, true policing experiments have the potential to go awry. Most importantly, are resources delivered evenly across conditions, controlling for variations over time and organization units? We investigate this question using assigned officer reports. Multiple items reflecting perceived time spent in assigned grids, equipment problems, and being called away generated a latent deployment variable. Multilevel models confirmed deployment varied significantly across the districts, and that deployment declined later in the treatment period. Controlling for particular districts and time, however, deployment levels were equivalent across uniform and unmarked patrol conditions. Next steps include creating a broader deployment variable that reflects supervisors’ as well as patrol officers’ responses, dealing with missing data, and cross referencing deployment with officers’ reactions and observers’ reports. A deployment variable may prove useful to help pack differences in treatment effectiveness across districts within an experimental condition, should those variations surface.

Authors

• Ralph Brecken Taylor, Temple University

• Jerry Ratcliffe, Temple University

• Mackenzie Tyler, Temple University

• Christopher Naimo, Temple University