An Evaluation of Crisis-Intervention Team Training

Police officers in the United States are often the first responders to mental health crises, despite growing concerns about whether traditional policing is well-suited to these encounters. One response has been crisis-intervention team (CIT) training for police. Unlike alternatives such as unarmed responders or co-responder models, CIT seeks to improve outcomes by training officers to de-escalate mental health crises themselves. This paper provides causal evidence on whether CIT training reduces police use of force and arrests during mental health incidents. I construct a comprehensive administrative dataset linking calls for service, police reports, use-of-force records, officer demographics, and detailed training records from the New Orleans Police Department from 2017 through 2023. To estimate the causal effect of CIT training, I use a difference-in-differences framework that exploits variation in the timing of training across officers. Specifically, I compare changes in propensity to use force and make an arrest for officers before and after they receive training to those of officers who are not-yet-trained but will be trained in the future. I find no evidence that CIT training reduces officers’ use of force or likelihood of making an arrest in mental health incidents. I also find no spillover effects on officer behavior in other types of calls. Importantly, officers who select into training are officers who are already less likely to use force even prior to training, indicating strong positive selection. Taken together, these results suggest that voluntary training programs, as currently implemented, may not meaningfully change officer behavior and instead primarily attract officers who are already less prone to use force.