Abstract
Geographic offender profiling (GOP) is an investigative activity aimed at locating an offender's residence on the basis of where he or she commits offences. Current tools that assist in GOP assume that spatial opportunity structures for crime are uniform: they assume that potential targets are evenly distributed in space, and that potential targets are equally attractive everywhere. In addition, they assume that travel distance is the only criterion that offenders consider when choosing a crime site. This paper demonstrates that, with the help of an extended target selection model, GOP tools could be improved if they measured and utilized spatial variation in criminal opportunity structures. The results of a computer simulation study illustrate this finding.
Acknowledgments
Paul Nieuwbeerta, Henk Elffers and Jasper van der Kemp and two anonymous reviewers of this journal provided helpful comments on previous versions of the manuscript. The Royal Netherlands Academy of Sciences (KNAW) provided a grant for the translation of the original Dutch manuscript (translation by Erin Jackson).
Notes
1. CrimeStat II is not a standalone GOP tool, but a general purpose program for the analysis of spatial crime data. Geographic profiling algorithms are implemented in the journey-to-crime estimation subroutine. The output must be imported into a GIS system for the construction of a prioritization surface.
2. The simulation was carried with the statistical program STATA (StataCorp, Citation2003). The source code of the simulation program is available upon request from the author.
3. All three methods are implemented in a conditional logit model. Tool A uses the equation:
in which O represents the set of all 100 locations and therefore comprises both locations where potential targets actually exist and locations where they do not. Tool B uses the equation
in which J is the set of 25 potential targets. Tool C uses
which is equation (Equation2) in the main text.