This project has developed technology that will depict and predict current and future crime potential. In doing so, it operationalizes two grounded theoretical approaches to understanding localized spatial crime patterns. This combination of a long-term crime potential map surface with a short-term crime spike surface creates the opportunity for law enforcement and other criminal justice agencies to apply a theoretical understanding to the business of crime prediction. The resulting map of predicted crime will enable police departments to take a proactive approach to crime prevention, disruption and reduction, and will provide a foundation for predictive policing and crime prevention.
The project team of researchers from Temple University’s Center for Security and Crime Science (housed in the Department of Criminal Justice) and technical experts from Azavea (a Philadelphia-based company that specializes in the creation of geographic web and mobile software, as well as geospatial analysis services to enhance decision-making) has created a custom software tool. This program will enable police departments and other agencies across the country to use their geocoded crime data in combination with freely-available census data and create micro-spatial estimates of future criminal activity at the local level.
As part of this ongoing project, one piece of research has been published already. We asked the question: Do fundamental demographic correlates of crime, proven important in community criminology, link to next year’s crime levels, even after controlling for this year’s crime levels? If they do, it would imply that shifting ecologies of crime apparent after a year are driven in part by dynamics emerging from structural differentials.
For Philadelphia (PA) census block groups, 2005 to 2009 data from the American Community Survey and 2009 crime counts were used to predict spatially smoothed 2010 crime counts in three different models: crime only, demographics only, and crime plus demographics. Models were tested for major personal (murder, rape-aggravated assault, and robbery) and property (burglary and motor vehicle theft) crimes.
We found that for all crime types investigated except rape and homicide, crime plus demographics resulted in the best combination of prediction/simplicity based on the Bayesian Information Criterion. Socioeconomic status (SES) and racial composition linked as expected theoretically to crime changes.
We concluded that intercommunity structural differences in power relationships, as reflected in SES and racial composition, link to later crime shifts at the same time that ongoing crime continuities link current and future crime levels. The main practical implication is that crime analysts tasked with long-term, one-year-look-ahead forecasting may benefit by considering demographic structure as well as current crime.
This research has been published here as: Taylor, R. B., Ratcliffe, J.H. and Perenzin, A. (2015) Can we predict long-term community crime problems? The estimation of ecological continuity to model risk heterogeneity. Journal of Research in Crime and Delinquency 52(5): 635-657.
This National Institute of Justice funded research project is ongoing, however the software is now available as of October 2016. Please visit the following website (you will leave the Temple University site) to access the software downloader, manual, and quick start guide.