Assessing the association between street view–derived perceived liveliness and street thefts

Crime geographers are attentive to how the spatial characteristics of the environment affect criminal behaviors. Recently, scholars have found that aside from physical and social attributes, the perception of the environment, such as safety perception, is also associated with crime occurrence. However, perceived liveliness, a visual perceptual proxy for potential routine activity settings that are closely linked to routine activity theory and crime pattern theory, is rarely explored in prior research. This study examines the association between perceived liveliness and street thefts in the City of Cincinnati, the United States. The liveliness perception scores are derived from street view imagery using a support vector machine model trained on the MIT Place Pulse V2 crowdsourcing image survey. Negative binomial regression models are used to test the relationships between perceived liveliness and street theft, with social and physical environmental attributes controlled. Results show that perceived liveliness significantly improves model performance and is positively associated with street theft. This study provides the first evidence that perceived liveliness helps explain crimes. The positive relationship between perceived liveliness of the built environment and crime adds a new dimension to the routine activity theory and the crime pattern theory. It also offers a scalable and cost-effective measure that can assist urban planners and law enforcement in identifying potential theft hotspots using readily available street view imagery.