Predicting Violent Offenders’ Spatial Behavior: Does Public Transport Accessibility improve Model Performance?

• In Event: Crime and Place: Offender Target Selection and Location Choice

Abstract

Using discrete choice models, previous research has shown that, at least in certain settings, minimum traveling time (estimated using public transport system data) outperforms geographic distance as a predictor of violent offenders’ spatial behavior. The present research compares the performance of two predictive models of violent offender spatial behavior – each using one of these journey-to-crime proxies. Models are calibrated using training data and tested using a second independent data set. External validation is achieved through the use of various metrics, such as the minimum area that must be searched (based on ranked calculated probabilities) until the crime site is reached. The implications of the findings for the prioritization of suspects in serious crime investigations are discussed. Specifically, we argue that our findings may inform methods of generating offender residence probability areas around newly reported offense locations.

Authors

• Lucía Summers, Texas State University

• Shane Johnson, University College London