A Picture Worth A Thousand Words: The Effect of Murals on Crime

• In Event: Using Computer Vision to Analyze Urban Images and Answer Crime & Place Questions

Abstract

Theories of social disorganization and crime as a response to neighborhood characteristics are unique among criminology's many proposed explanations for crime in both testability and empirical success. Many of the randomized control trials and quasi-random studies drawing from place-based models have proven not only effective in countering crime and violence but also relatively inexpensive, including street lighting, lot greening, business improvement districts, and vacant housing abatement. The installation of large exterior murals fits well within this framework as an affordable intervention that transforms communities by influencing social disorganization with several potential positive externalities. In Philadelphia, the staggered implementation of the Mural Arts Program's murals lends itself to a differences-in-differences quasi-experimental design that allows us to not only measure the impact of mural installation on crime but to identify the optimal location, placement, and subject matter to invest most efficiently if murals prove effective in reducing crime and violence. Additionally, by investigating non-crime outcomes in a similar framework using the mural installation date, we can also analyze the mechanisms that spur crime reductions, including increased traffic, commercial revitalization, and augmented social cohesion through reduced disorder. Importantly, we can additionally check for the potentially negative externalities of neighborhood beautification, including crime spillovers into adjoining neighborhoods and gentrification. This effort can be aided and made more efficient through the use of machine learning and computer vision, which can help classify and identify both mural elements and environmental elements that may determine the effectiveness of murals.

Author

• Maya Moritz, University of Pennsylvania