Forecasting the Future of Predictive Crime Mapping (From Analysis for Crime Prevention, P 29-57, 2002, Nick Tilley, ed. -- See NCJ-194015)

• NCJ Number: 194018
• Author(s): Elizabeth R. Groff; Nancy G. La Vigne
• Date Published: 2002
• Length: 29 pages
• Sale Source: Criminal Justice Press/Willow Tree Press
• Publication Format: Book (Softbound)
• Publication Type: Report (Technical Assistance)
• Language: English
• Country: United States of America
• Note: *This document is currently unavailable from NCJRS.

Annotation

This paper discusses the methods used to develop predictive models of crime based on crime mapping and examines issues related to their use and accuracy in predicting future crime hot spots.

Abstract

Most applications of crime mapping have been retrospective. However, the real promise of crime mapping rests on its ability to determine early warning signs across time and space to inform proactive policing and crime prevention. Police agencies’ most common method of forecasting crime is simply to assume that past hot spots are future hot spots. The crime prevention benefits of focusing on repeat victims rather than the high-crime area is well established and raises the question of whether repeat victimizations of individuals and places can predict not just hot dots, but also hot spots. A variety of univariate methods can predict crime with a minimum of data collection. These methods range from simple random walk and naïve lag 12 to more sophisticated models that include both seasonality and time trends. Multivariate methods based on leading indicators have great promise, because they are the only method with the ability to predict pattern changes. However, their use requires significant expertise. Additional methods include the point process model, a method based on the concept of artificial neural networks, and methods that combine polygon grid cells and raster-based geographic information systems. The most sophisticated approaches are in the development stages and have not received testing by end users. In addition, the variability of the methods precludes direct comparisons of their accuracy. Technology has improved the ability to create, maintain, and manipulate data, but much work must take place before effective forecasts of crime trends will be possible. The analysis concludes that more complicated methods are not always better predictors and that further research needs to focus on the choices made in sophisticated models and the input variables in multivariate models. Chart, notes, and 61 references