Pedestrian volume and street crime: a comparative analysis of tourist and residential areas in Manhattan

Purpose This study investigates the impact of pedestrian volume on street crime by distinguishing between tourist-visited and residential areas using the distribution of tourist points of interest (POIs). It empirically assesses how this spatial classification moderates the pedestrian activity – crime incidence relationship. Design/methodology/approach This study used street larceny and violent crime as dependent variables and measured pedestrian volume using Google Street View images processed with the YOLOv8 object detection model. It combined pedestrian and POI data to capture contextual effects, and applied negative binomial regression models. Findings Pedestrian volume was positively associated with both dependent variables. For larceny, the relationship was consistently linear across residential and tourist-visited areas. However, for violent crime, the effect was nonlinear: beyond a certain threshold, increased pedestrian flows were linked to a slower rate of increase in crime. This attenuation of the marginal effect was stronger in residential neighborhoods than in tourist-visited areas, highlighting the role of social composition and community ties in shaping crime outcomes. Practical implications The study recommends tailoring Crime Prevention Through Environmental Design strategies to both spatial context and crime type, as tourist-visited areas and residential neighborhoods exhibit distinct pedestrian–crime dynamics. Originality/value Rather than directly identifying pedestrians’ social attributes, this study extends the literature by incorporating the spatial context of tourist concentration into empirical crime analysis. By using tourism POIs as a proxy to estimate tourist presence, it provides a practical approach to explore whether and how pedestrian–crime dynamics vary by urban setting.