Analyzing Crime Displacement with a Simulation Approach

Abstract

Crime tends to cluster in small areas. Police have taken advantage of this by identifying these 'hotspots' of crime and concentrating their resources in these locations. This practice has shown evidence of reducing crime in the hotspot area. While it is possible that the benefits of such reduction might be diffused to the surrounding areas, one criticism of this practice is that the crime in the hotspot may also be displaced to the surrounding areas. A number of empirical studies have investigated the spatial pattern of crime displacement. However, few have attempted to uncover the mechanisms that lead to the displacement of crime. In this paper we present a theory-driven approach that applies agent-based modeling to simulate the mechanism of crime and policing. The processes that drive possible displacement of crime are investigated through experiments in a computational laboratory-SPACES. Our results reveal that crime cannot be displaced easily because opportunities for crime are limited in low-crime areas and offenders are often attached to the area where they perform their routine activities.

Non-Homogeneous Diffusion of Residential Crime in Urban China

“…The near-repeat effect does not fully explain criminal behaviors, as crime distribution is characterized not only by diffusion, but also by displacement [22][23][24][25][26]. Displacement is the change of crime from one place, time, target or tactic to another as a result of some crime prevention initiative or of enhancing vigilance of community residents or policing efforts.…”

“…This may occur among adjacent sub-districts, communities, and cities [31]. Some scholars do not agree with situational interventions, believing that offenses are inevitable due to social, economic, biological and psychological factors [23]. In their view, situational interventions contribute little to crime reduction, but simply result in the change of crime to another time, place or tactic.…”

Deter and protect: crime modeling with multi-agent learning

“…However, the work did not use any learning for guardian agents and they solely moved randomly in the simulated environment. Subsequent work by Wang et al [43] attempted to address the lack of guardian agent learning by having guardians seek the "hottest" crime locations to represent "hot spot" patrolling.…”

“…The novelty of using belief learning is that agents observe opposing agents' actions without regard for the payoffs that drove the opposing agents to make those choices, form beliefs using these observations, and then make future choices using these beliefs. Prior research by Wang [42] and Wang et al [43] used the reinforcement learning method Q-learning. In their approach, rewards and costs are accumulated in the environment where offenders seek optimal reward and targets seek least cost (due to potential of being victimized).…”

“…In this study, at each stage of the game, offender agents can attempt a crime, but only if their motivation outweighs the guardianship present. This study uses a variation of Eck's likelihood for a crime event [12] as used by Wang et al [43]. The formula in Eq.…”