The Roads Between Nodes: How Travel Paths Affect Burglary Location Choices

Crime pattern theory stresses the importance of offender activity spaces for understanding where crime occurs. Activity spaces comprise both frequently visited places (nodes) and the routinely travelled paths that connect them. Although previous research showed that offenders indeed often commit crimes near their activity nodes, little is known about the extent to which the paths between these nodes feature in their crime location choices. This study examines whether offenders favour targets that are close to these paths, as well as those close to activity nodes. This burglary location choice study analyses 450 burglaries committed by 271 burglars in the Greater Hague area in the period 2006–2009. For each incident, we examine the location of the offence, along with the locations of likely activity nodes for the offender—their current and previous homes, those of their family members, and their previous offence locations. The study makes progress in two ways. First, it uses street segments (N = 28,221) as its spatial unit of analysis. Second, the plausible routes offenders could have used to travel between their activity nodes are identified by applying graph theoretic heuristics, and these are used to estimate the wider awareness spaces built up along the way. We use discrete spatial choice models to test several hypotheses derived from crime pattern theory. Burglars have greatest preference for targets that lie on the same street segment as one of their activity nodes, and the relative preference decreases with each additional degree of separation over the street network. They also tend to target houses along the paths between their nodes, although effects are substantially smaller and the decay pattern as the distance from the path increases is less pronounced. This study provides evidence that offenders preferentially target locations that lie on the paths between their activity nodes; a key proposition of crime pattern theory that had not previously been tested directly. Information on only a few offender-specific activity nodes, enriched with graph theory metrics for plausible paths that connect them, results in a high predictive accuracy for where burglars commit crime.