The Philadelphia Predictive Policing Experiment was a collaboration between Temple University’s Center for Security and Crime Science (housed in the Department of Criminal Justice) and the Philadelphia Police Department. This National Institute of Justice funded research project has been the first place-based, randomized experiment to study the impact of different police strategies on violent and property crime in predicted criminal activity areas.

Predictive policing is an emerging tactic relying in part on software predicting the likely locations of criminal events. Predictive policing, while sometimes applied to offenders, is most frequently applied to high crime places. In this context, it involves ‘the use of historical data to create a spatiotemporal forecast of areas of criminality or crime hot spots that will be the basis for police resource allocation decisions with the expectation that having officers at the proposed place and time will deter or detect criminal activity’ [Ratcliffe, J. H. (2014). “What is the future… of predictive policing?” Translational Criminology, 2014 (Spring): 4-5 (definition on page 4)].

At present, the law enforcement field lacks robust evidence to suggest the appropriate policing tactic in predicted areas. That has been the subject of this timely study. The aim has been to answer the question of whether different varieties of theoretically informed, but also operationally realistic, police responses to crime predictions estimated by a predictive policing software program can reduce crime.

The research team from Temple University and the Research and Analysis section of the Philadelphia Police Department randomly assigned 20 Philadelphia Police Department (PPD) districts into one of four experimental conditions. Five districts acted as controls, with a business-as-usual patrol strategy. In five districts, officers were made aware of the predicted high crime activity area at roll call and asked to concentrate there when able (a simple awareness model). Five districts received the awareness model treatment as well as an additional patrol car solely dedicated to the predicted crime area. Finally, five districts received an intelligence-led, investigative response with an unmarked unit dedicated to the predicted area. A three-page pdf has a fuller description of the research methodology.


When examining both predicted high-crime grid cells and the grids cells immediately surrounding them, the marked car patrols resulted in a 31% reduction in property crime counts, or a 36% reduction in the number of cells experiencing at least one crime. This translates to a reduction in three crimes over three months for an average city district patrolling around three grids. There were also signs of a temporal diffusion of benefits to the eight hours following the property crime marked car patrols. While the percentages were substantial, the results were not statistically significant due to floor effects.

There were no crime reduction benefits associated with the violent phase of the experiment, nor were there any benefits with the property crime awareness or unmarked car interventions. In summary, it appears that marked police cars dedicated to predictive policing areas were effective at reducing property crime. Unmarked cars and efforts to combat violence were not shown to be effective in the Philadelphia Predictive Policing Experiment.

A two-page pdf has a more detailed description of the experimental results.

The predictive policing software employed was the HunchLab program designed by Azavea. HunchLab is a web-based predictive policing system that accesses real-time Philadelphia Police data to produce crime forecasts for the city. It incorporates statistical modeling that considers seasonality, risk terrain modeling, near repeats, and collective efficacy. Azavea adapted the software at the request of the Philadelphia Police Department and researchers from Temple University to generate three predicted 500 feet square grids per district per shift. They also included a slight randomization component to reduce the possibility that the same grid cells were predicted every day. It is important to note therefore that the experiment artificially reduced the efficiency of the software, because it forced the software to choose grids in low crime districts, and limited the number of grids it could assign in high crime districts.


The software was able to predict twice as much crime as we would expect if crime were spread uniformly across the districts, even when artificially constrained by our experiment to be less effective than designed. A two-page pdf has more details of the software efficacy.