Regarding the process evaluation, all treatment districts adhered to the prediction model as envisioned. Regarding impact, the evaluation found no statistical evidence that crime was reduced to a greater extent in the experimental district compared to the control districts. Suggested reasons for this are low statistical power of the predictive model, failure of program implementation, and flawed program theory. The statistical tests used had low statistical power given the small number of experimental and control districts, as well as low and widely varying crime counts per month and district in Shreveport. Based on these findings, its is unclear that a map that identifies "hot spots" based on a predictive model of crime is more useful than a traditional map that identifies "hot spots" based on prior crime locations. More definitive assessments of whether predictive maps can lead to greater crime reductions than traditional crime maps would require further evaluations that have a higher likelihood of detecting a meaningful effect. Conducted in 2012, the evaluation focused on the use of a predictive policing strategy to reduce property crimes in three districts. Three control groups continued with the usual policing approach for reducing property crimes. The evaluation used qualitative and quantitative approaches that addressed the processes, impacts, and costs of the predictive policing experiment. 12 figures, 22 tables, and 16 references
Downloads
Similar Publications
- Estimating the Number of Crimes Averted by Incapacitation: An Information Theoretic Approach
- Supporting Implementation of Universal Prevention Initiatives in K-12 Schools: Impacts on Fidelity through Organizational Readiness and Team Functioning in a Cluster-Randomized Trial
- Post-burn and Post-blast Rapid Detection of Trace and Bulk Energetics by 3D-printed Cone Spray Ionization Mass Spectrometry