LAUSR

The burgeoning field of individual level crime location choice research has required increasingly large datasets to model complex relationships between the attributes of potential crime locations and offenders’ choices. This study tests methods of sampling aiming to overcome computational challenges involved in the use of such large datasets. Using police data on 38,120 residential and non-residential burglary, commercial and personal robbery and extra-familial sex offense locations and the offenders’ pre-offense activity locations (e.g., home, family members’ homes and prior crime locations), and in the context of the conditional logit formulation of the discrete spatial choice model, we tested a novel method for importance sampling of alternatives. The method over-samples potential crime locations near to offenders’ activity locations that are more likely to be chosen for crime. We compared variants of this method with simple random sampling. Importance sampling produced results more consistent with those produced without sampling compared with simple random sampling, and provided considerable computational savings. There were strong relationships between the locations of offenders’ prior criminal and non-criminal activities and their crime locations. Importance sampling from alternatives is a relatively simple and effective method that enables future studies to use larger datasets (e.g., with more variables, wider study areas, or more granular spatial or spatio-temporal units) to yield greater insights into crime location choice. By examining non-residential burglary and sexual offenses, in New Zealand, the substantive results represent a novel contribution to the growing literature on offenders’ spatial decision making.