Beyond Identification: A Problem-Oriented Approach to Diagnosing Racial Disparities in Policing

Abstract Objectives This study developed a framework for diagnosing some of the theoretical mechanisms underlying racial disparities in policing to inform the design and implementation of problem-oriented police reforms. Methods We sample over 1.5 million traffic stops conducted between 2019 and 2023 in a state police agency and use the Veil-of-Darkness (VOD) method on a subset of 299,767 stops within the intertwilight period to demonstrate the utility of the diagnostic framework. This method utilizes weighted logistic regression with daylight as the primary independent variable and driver race as the dependent variable. Results As part of demonstrating the framework, we found that disparities were largely diffuse across the agency, with some concentration among officers and patrol regions. As such, it would be best to design multi-pronged trainings and interventions that reflect these distinct patterns observed in the agency. Conclusions The diagnostic framework provides a data-driven tool for researchers and practitioners to begin to understand where different types of racial disparity originate within police agencies. Future research building on this framework should explore the utility of incorporating alternative data sources and theoretical mechanisms, as well as assessing other metrics used to assess different types of racial disparities.