Innovations in policing often involve isolating the behaviors of the best officers, distilling those behaviors into a code, and teaching that code to other officers so that they can be equally successful at meeting whatever metric police managers hope to improve. Such is the case with the concept of “de-escalation.” De-escalation is not one set of techniques or policies. Rather, it is the substitution of space, time, and cover in place of force to help officers gain control of a potentially violent confrontation. Many competent officers already do this, including members of most SWAT teams, but no comprehensive national training curriculum exists to refine and spread best practices. At their core, the techniques and policies that fall under the umbrella of de-escalation are about the proportional use of force and the sanctity of human lives, both of officers and the people they contact.

De-escalation has emerged near the top of Congress’s efforts to improve policing practices across the country. The Law Enforcement De-escalation Training Act of 2022 (LEDTA), introduced by Sen. John Cornyn (R-TX), authorized the appropriation to the Justice Department of $34 million to develop a de-escalation curriculum, and another $90 million earmarked for local police departments to participate in de-escalation programs. The size of the authorized investment reflects the potential of de-escalation rather than evidence of its prior success. Though many agencies already engage in various well-regarded de-escalation training regimes, law enforcement de-escalation is  only minimally tested. Still, de-escalation tactics are promising, and the diversity of practices is a feature, not a flaw, providing an opportunity to learn from many experiments and some notable success. This research roundup summarizes the best existing research on de-escalation1.

Prior to the 2014 death of Michael Brown at the hands of police in Ferguson, Missouri, de-escalation was a concept applied almost exclusively by nurses and psychiatric professionals. In those contexts, de-escalation has mostly referred to preventing harm to staff rather than to the members of the public with whom staff members interact. Training focuses on the motivations for violence among patients, reactive strategies, and self-defense. 

Before 2014, police employed some techniques similar to those used by healthcare professionals to de-escalate violent situations. These include conflict resolution and verbal judo, a technique developed in 1983 that combines rhetoric and empathy to help police gain voluntary compliance. Another prominent model, the Crisis Intervention Team program (CIT), was developed in Memphis in 1988 and applied de-escalation tactics specifically to people suffering from acute mental health or substance abuse episodes. Evaluations of CIT have shown mixed support for the program, though many of its elements were incorporated into LEDTA. 

Public outcry following high-profile incidents of police violence spurred the creation of the President’s Task Force on 21st-Century Policing, which released a set of recommendations for police practices in 2015. Prominent among these recommendations was that police agencies adopt de-escalation tactics. This recommendation was more of a bet than a sure thing. A systematic review conducted by Robin Engel and colleagues found that of 64 evaluations of de-escalation training programs published between 1976 and 2016, none were in a law enforcement context. Moreover, almost all of these evaluations employed relatively weak research designs. This initial lack of an evidence base led some critics to argue that de-escalation would threaten the lives of officers with little payoff to public safety. After all, officers are often trained to act decisively to gain control of a situation and quell the chance of violence. By instructing officers to withhold action and attempt to take charge through non-physical means, they reasoned, de-escalation training would create more chances for violent criminals to gain the advantage. Now, a small but growing evidence base contradicts these claims. 

The Camden County (NJ) Police Department was among the earliest adopters of de-escalation training for officers. Reorganized from the defunct Camden Police Department due to longstanding allegations of corruption and abuse, it began retraining officers and revising policies in 2013. Still, by 2015 it had among the highest use-of-force rates of any large police jurisdiction in the state. Between 2015-2016, Camden police developed a de-escalation training model based on a curriculum promoted by the Police Executive Research Foundation (PERF). Li Sian Goh evaluated the effects of this training on serious use-of-force incidents over  five years.

The first stage of evaluation happened at the officer level: Goh was interested in whether Camden police officers who had received training were more likely to engage in use-of-force than other Camden police officers who had yet to receive training. She found insignificant differences between the two groups. Suspecting that this null finding was driven by informal cultural changes by which lessons learned from the training were transmitted to untrained officers by their trained peers, Goh conducted a difference-in-differences analysis in which she compared the rates of use-of-force in Camden to a synthetic control created from the 35 other largest municipalities in New Jersey. She found that Camden saw a significant reduction in the use of force compared to the weighted aggregate of other large jurisdictions. The difference was marked even when these numbers were standardized by arrest, suggesting that use of force did not fall simply because officers made fewer arrests or initiated fewer encounters. Specifically, she found that after training, the rate of serious use of force incidents per 1,000 arrests dropped by about 50 per quarter, a decrease of 40 percent. What’s more, this drop was achieved without a serious increase in violent crime.  

At the time of the Camden evaluation, no randomized controlled trials of de-escalation training had been attempted. That changed in 2017, when the Tempe (AZ) Police Department launched the Tempe Smart Policing Initiative (SPI), a collaboration between police, Arizona State researchers, and the Bureau of Justice Assistance. The Tempe SPI developed a custom curriculum for de-escalation built around lessons drawn from dozens of other de-escalation training programs, which officers attended and reviewed on behalf of the department. These experiences were compiled by a committee of stakeholders into a curriculum that was gradually introduced over an 18-month period. A group of researchers led by Michael White randomized officers to treatment and control groups by squad to prevent the potential for spillover of treatment effects, as had happened in Camden. 

As an initial evaluation strategy, researchers analyzed administrative data from six months before and after the implementation of training among the treatment group. Administrative data, collected as part of standard department practice, are ideal for RCTs. Unfortunately, the study period coincided with the COVID-19 pandemic, which sharply reduced the number of officer-citizen interactions, artificially depressing the number of use-of-force incidents recorded. For this reason, researchers analyzed body-worn camera (BWC) footage collected during the same period to evaluate changes in officer behavior that were not apparent in the administrative data. They found that officers who had received training were significantly more likely to employ de-escalation tactics after training and that these tactics were associated with a 58 percent reduction in civilian injuries. Though the findings from BWC footage are more sensitive than administrative data to researcher choices, the randomization of officers into treatment and control groups makes this study more likely to produce accurate estimates of the effect of the de-escalation curriculum.  

The most convincing study to date is another RCT, conducted by the Louisville Metro Police Department and researchers at the University of Cincinatti’s Center for Police Research and Policy. The LMPD adopted a de-escalation curriculum based on the principles developed by PERF beginning in February 2019. As with the Tempe evaluation, the study period in Louisville coincided with the COVID-19 pandemic. Fortunately, an element of the study design allowed researchers to estimate a treatment effect for the program. The evaluators used a stepped-wedge design, in which the training was gradually extended to sequential cohorts of officers. Though the research design allowed the possibility of spillover, such a threat would only attenuate the results rather than magnify them. 

The research team found that over about 18 months, the rate of use-of-force per month fell consistently with each sequential cohort of officers trained. The baseline period was 2015-2018. The study period concluded in April 2020, so the effects of the pandemic on the results are minimal. The results were most pronounced in serious uses-of-force, such as TASER use, hand strikes, and take-downs. Overall, de-escalation was associated with a 28 percent reduction in uses-of-force, a 26 percent reduction in citizen injuries, and a 36 percent reduction in officer injuries. These results were not sensitive to changes in arrest patterns or other things that varied with time. 

Congress’s bipartisan push to promote de-escalation is a welcome addition to the growing federal effort to improve policing practices. The results of rigorous research suggest that de-escalation can meaningfully change officer behavior, yield safer working conditions and more respect for civil rights and human dignity. Still, several open questions remain. The ability of officers to leverage space, time, and cover may be situationally dependent or a function of the ratio of officers to civilians involved in a given confrontation. Severely outnumbered officers may find it difficult to safely employ de-escalation tactics, and situational assessments are necessary for sound implementation.The research is also hampered by the relatively rare occurrence of use-of-force incidents, which makes accurate evaluation difficult. Much of the existing research focuses on changes in officer attitudes rather than behavior. Such research provides a helpful backstory but is difficult to translate into policy recommendations.  

De-escalation is not an inexpensive endeavor. As Robin Engel pointed out at a recent National Institute of Justice research conference, the Louisville de-escalation program required frequent follow-up training sessions with officers to be successful. This need for perpetual improvement makes identifying the proper dosage of education difficult. Translating a curriculum into practice was also a challenge, and the research team and police managers had to devote significant time to shaping the training, even after the conclusion of the study period. Still, the cost of department-wide de-escalation training is probably lower than that of even one unjustified fatal or near-fatal use of force, which can cost departments tens of millions of dollars directly, in addition to broader costs borne by society

The existing research on de-escalation strongly suggests that regard for officer safety and respect for civilian lives are compatible, even in potentially violent situations. That lesson provides a great start. Now, LEDTA has the potential to expand on this lesson and change policing practices nationwide by turning the things the best officers and agencies already do to avoid unnecessary violence into standard operating procedures.

  1. For the purposes of making policy recommendations, the Niskanen Center prioritizes studies that employ methods that isolate the effects of policies or practices from other confounding factors in order to make causal inferences. Randomized controlled trials (RCTs) are the gold standard for this purpose, but other, so-called quasi-experimental methods can approximate the conditions of an experiment. Non-experimental methods, though useful for descriptive purposes, are often inadequate for the purpose of making policy recommendations and are excluded from this research roundup.