UCLA

This dissertation investigates the role of space and institutional structures in shaping criminal justice contact through three essays. In Chapter 1, co-authored with Keith Chen, Katherine Christensen, Elicia John, and Emily Owens, we use smartphone location data to track on-shift movement of police officers in the 21 largest US cities, enabling us to construct and examine police presence—what it means for an area to be “policed”—without relying on a department’s cooperation. We find that police spend significantly more time in non-white neighborhoods, a disparity that persists even after controlling for population density, socioeconomic factors, and crime rates. Disparities in police presence also predicts a large share of observed racial disparities in downstream police actions such as arrests and stops. Importantly, our data facilitates cross-city comparisons, revealing unique issues leading to disparities across cities.In Chapter 2, co-authored with Keith Chen and Emily Owens, we show how policing could be endogenous to place-based investments effective at reducing crime, using the smartphone-based measure of policing. Exploiting the variation in Quali- fied Census Tract (QCT) status due to administrative rules under the Low-Income Housing Tax Credit (LIHTC) program, this study finds that police patrols increase by 13.5% in QCTs compared to non-selected but similar tracts. These increases can more than explain observed investment-induced violent crime reductions. This research challenges the notion that place-based investments can significantly reduce crime without considering the broader equilibrium effects on policing patterns. In Chapter 3, also co-authored Keith Chen and Emily Owens, we train a con- volutional neural network on Google Street View images to explore the relationship between physical space, perceived safety, and actual crime rates. The study aims to identify specific urban features that influence safety perceptions and the discrep- ancies between perceived and actual safety. By integrating generative AI tools, this research provides a new framework that could help identify physical features that could help potentially mitigate perceived safety and crime, providing actionable in- sights for urban planners and policymakers.