The Spatial Extent of Criminogenic Places: A Changepoint Regression of Violence around Bars

Crime scientists have long known that crime clusters near certain places such as drinking establishments, although the spatial parameters of that clustering are less established. This article proposes a methodology to estimate a distance beyond which there is significantly less evidence of a correlation between locations and concentrations of crime. The technique uses changepoints derived from a segmented regression applied to spatial buffers emanating from around particular crime-generating land uses. Geographic information system techniques are used to create a series of buffers to determine the density of crime around sites. A changepoint Poisson regression of the buffer midpoints is used to estimate the distance beyond which crime densities do not appear to decline significantly with increasing distance. A case study of violent crime around 1,282 bars in Philadelphia, Pennsylvania, for 2008 reveals that violence is highly clustered within 279?m (85 feet) then dissipates rapidly, a pattern that is not replicated using control sites (fire stations). This is an estimate of the spatial extent of violence around bars, and the technique could be used to estimate the extent of other crimes around a variety of crime-generating locations.