A Conditional Ordinal Stereotype Model to Estimate Police Officers’ Propensity to Escalate Force

This article introduces a conditional ordinal stereotype model for estimating a police officer’s latent propensity to escalate to higher-severity force options during an encounter. Propensity to escalate is the likelihood of selecting a more serious force category than peer officers confronting the same circumstances, not necessarily implying that such force is excessive or contrary to policy. The associated conditional likelihood depends solely on data from the times and places where multiple officers use various forms of force, precisely the types of situations that police departments are most likely to document. By design, all shared situational and environmental features cancel out of the conditional likelihood, eliminating the need to observe or model them explicitly. The article proves that officer-level inference independent of shared environmental features can arise only from the proposed conditional ordinal stereotype model, establishing it as the unique model with this invariance property. On simulated examples, the conditional ordinal stereotype model recovers which officers escalate force the most, potentially risking unnecessary injuries or fatalities, and which officers may be too reluctant, potentially endangering themselves, colleagues, or the public. Applying the approach to use-of-force data from Seattle finds nine officers with a high probability of being local outliers in their use-of-force networks. By quantifying officer-level variation while accounting for situational context, the conditional ordinal stereotype model expands the analytic toolkit available to police executives and oversight bodies for proactive risk management. Supplementary materials for this article are available online, including a standardized description of the materials available for reproducing the work.