Linking Theory to Practice: Testing Geospatial Predictive Policing in a Medium-Sized Police Agency

The purpose of this applied research project is fourfold: to link theory and appropriate data/measures to the practice of predictive policing; to determine the accuracy of various predictive algorithms, including traditional hotspot analyses, regression-based analyses, and data mining algorithms; to implement the identified/best performing algorithm into a predictive policing process in Columbia (SC) Police Department; and, to evaluate the effectiveness of the predictive policing process using a randomized control trial and regression-discontinuity design. Additionally, this research will generate data report cards for all used data sources that identify: the availability of the data, the process to obtain the data, how often the data is updated, quality of the data (including information on incomplete/blank/inaccurate fields), which theory-relevant predictor variables can be obtained from the data, amount of time to recode/clean the data, and univariate variable descriptions of the data. This information will be coupled with measures of the predictive ability of the theory-relevant variables to the most accurate predictive algorithm. The intent of these report cards is to eliminate data sources that consume an inordinate amount of time/work in relations to the predictive power they provide a model. ca/ncf