Utility of Hotspot Mapping for Predicting Spatial Patterns of Crime

• NCJ Number: 221770
• Journal: Security Journal Volume: 21 Issue: 1-2 Dated: 2008 Pages: 4-28
• Author(s): Spencer Chainey; Lisa Tompson; Sebastian Uhlig
• Date Published: 2008
• Length: 25 pages
• Publication Format: Article
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America

Annotation

This British study used crime data for a past time period in order to test the accuracy of four methods for generating hotspot maps (thematic mapping of census output areas, spatial ellipses, grid thematic mapping, and kernel density estimation) that predicted the occurrence of burglary, street crime, theft from vehicles, and theft of vehicles.

Abstract

Study findings show that the accuracy of crime hotspot mapping differed among the four mapping techniques as well as by crime type. Kernel density estimation (KDE) consistently provided more accurate predictions than the other techniques, and street-crime hotspot maps generated by all four techniques were consistently more accurate than predictions for the other crime types. The KDE method of hotspot mapping aggregates offense data within a user-specified search radius and calculates a continuous surface that represents the density or volume of crime events across the specified area. A smooth surface map is produced to show the variation of the point/crime density across the study area without having to conform to geometric shapes such as ellipses. There is flexibility in setting parameters such as the grid cell size and bandwidth (search radius); however, despite many useful recommendations, there is no universal agreement on how to set these and in what circumstances. Examples of the use of KDE are now widespread, with popular crime-mapping texts showing many examples of its use (See Harries, 1999; Goldsmith et al., 2000; Chainey and Ratcliffe, 2005; and Eck et al., 2005). Hotspot mapping for street crime yielded the most accurate predictions, possibly because it was concentrated in areas of retail and entertainment sites; whereas, the other crimes tended to be geographically transient. Geocoded crime point data were provided by the London Metropolitan Police for a 2-year period (January 1, 2002 to December 31, 2003). 8 tables, 5 figures, and 44 references