Greg Ridgeway, Ph.D. Department of Criminology, University of Pennsylvania Department of Statistics and Data Science, University of Pennsylvania

In the aftermath of a problematic police use-of-force incident, investigators often look for warning signs, such as whether the involved officers had prior excessive force complaints or a history of using serious force. However, this retrospective analysis is often insufficient for managing future risks. In this seminar, I introduce a conditional ordinal stereotype model designed to assess an individual officer’s propensity to escalate to more serious forms of force. This model is based on a conditional likelihood version of Anderson's ordinal stereotype model. By matching officers who were present at the same use-of-force incident, the model accounts for shared environmental factors, such as location, subject characteristics, lighting, and the presence of an armed offender. This approach addresses a long-standing challenge of environmental confounding in police use-of-force analyses. When applied to use-of-force data from Seattle, the model flags 16 officers with a high probability of being outliers in their local use-of-force networks. The proposed approach expands the limited but growing set of tools available for police executives and oversight authorities to monitor high-risk police activities.