Mohler, G. O., Short, M. B., Malinowski, S., Johnson, M., Tita, G. E., Bertozzi, A. L., & Brantingham, P. J. (2015). Randomized controlled field trials of predictive policing. Journal of the American statistical association, 110(512), 1399-1411.
Location in the Matrix and Methodological Rigor:
Micro places, General, Proactive; Very Rigorous; Effective
What police practice or strategy was examined?
This study compares the use of hot spots policing patrols that use predictive policing models versus those that use crime analysts. The predictive policing models (epidemic-type aftershock sequence, ETAS) take into account both the long-term fixed environmental conditions and the short-term dynamic changes in risk. The ETAS model was automated to produce treatment missions using crime data from the most recent 365 days available in the live records management system. Each day, twenty 150×150 meter-sized geographic boxes where crime was likely to occur were generated using the ETAS model for each 12-hour shift in three divisions in Los Angeles Police Department (LAPD). Police officers were encouraged to use available time to “get in the box” and use their discretion to select appropriate field tactics upon entering predicted locations. The control condition used predictions derived from the best practices of trained crime analysts. Specifically in the control condition, the analysts emphasized small clusters of recent crimes as signaling emerging problems in need of short-term response as part of the COMPSTAT operation. The treatment lasted for five to eight months in each of the divisions.
How was the intervention evaluated?
A single-blind experimental design was used in which days were randomly assigned to treatment or control conditions. No direction of specific intervention was given to either group. Thus, control and treatment conditions contrasted directed patrol in space and time (micro areas where crime is likely to occur are predicted using the ETAS model during treatment days, and using the existing best practice during control days), rather than differences in field tactics. A total of 62 control days and 62 treatment days in Foothill, 82 control days and 70 treatment days in North Hollywood, and 177 control and 117 treatment days in Southwest were randomly assigned. Target crime types included burglary, car theft, and burglary-theft from vehicles in Los Angeles. Daily crime volume in all three divisions was regressed on the police patrol time across all active prediction boxes for the treatment and control conditions separately.
What were the key findings?
The researchers found the ETAS predicted crime more accurately than traditional crime analysis approaches. Patrol time across all active prediction boxes significantly reduced daily crime volume in the treatment condition, but not in the control condition. The impact translated to 4.3 fewer crimes per week at mean patrol levels, or an average 7.4% reduction in crime volume as a function of patrol time.
What were the implications for law enforcement?
The authors suggest that dynamic police patrol in response to ETAS crime forecasts can disrupt opportunities for crime and lead to real crime reductions.
Where can I find more information about this intervention, similar types of intervention, or related studies?