Police De-Escalation Training: What Is It Good For?
Concrete Evidence (January 26, 2026)
“De-escalation training” has been trendy in the police-reform world for a long time. If we teach cops how to defuse tense situations, the thinking goes, maybe they’ll need to use force less often. The vast majority of police academies include it in their teaching.
Does it work? Sometimes, in some ways.
A new study in the Journal of Criminal Justice reviews previous work, reports a brand-new experiment conducted within the Virginia Beach Police Department, and offers some theories to explain the confusing mix of results that research into this topic has produced.
Previous studies are mixed on the core question of whether de-escalation training reduces the use of force. Advocates can point to reductions of about a fifth in Indianapolis and a quarter or so in Louisville. Skeptics might highlight the disappointing findings of this 2020 study.
Further, as the new study’s authors note, when the training fails to prevent violence, it can be unclear why. Did the training fail to change cops’ behavior at key moments in the field, or was the cops’ behavior not the problem to begin with? Uses of force stem from interactions between cops and civilians, after all, and cops don’t have full control over how things unfold.
For their new experiment, the authors randomly assigned half the department’s officers to receive de-escalation training. (The other half were waitlisted to the next year.) The training, designed by Polis Solutions, sought, in the words of the lead trainer, “to break down communication in the same way we break down how to fire a gun.” The researchers monitored not only the officers’ use-of-force rates, but also used body-cam footage to monitor how cops behaved when responding to domestic-violence and intoxicated-person calls.
The upshot: The training did change cops’ behavior a bit on the margins. After receiving the treatment, officers became more likely to attempt conversation and less likely to repeat commands. They were also a bit more empathetic (in the judgment of those trained to code the videos), though this difference is borderline in terms of statistical significance. When trained cops attempted conversation, they followed the trainers’ advice that they make connections based on things like hobbies and relationships.
Unfortunately, this didn’t translate to reduced force. Trends were the same for officers who did and didn’t receive the training.
What gives? The authors note that “the research has yet to support a link between the intermediary behaviors [targeted by the training] and fewer uses of force, but has established a link between some training, increased use of alternatives to arrest, and fewer uses of force.”
The authors posit that these types of training might reduce force only when they also reduce arrests. As they concede, however, whether that’s a desirable outcome “is a philosophical question better left to the realm of police chiefs, mayors, city councils, and individual communities, rather than the pages of an academic journal.”
Other Work of Note
How many Americans do shrooms?
Law schools have scaled back their use of racial preferences.
The CBO assesses the accuracy of its projections for Fiscal Year 2025.
Some fascinating modeling of how human migration patterns work.
If you’ve seen the claim that Americans are paying 96% of the tariffs, you can check out the underlying analysis from the Kiel Institute here.
We’ve talked about holding kids back a grade in this space before, and this new study notes some difficulties in studying the effects. Flagging kids for retention seems to modestly boost their performance, comparing kids who were just barely flagged with kids slightly on the other side of the cutoff. However, many flagged kids are not actually held back, and even these kids get extra support, so it’s harder to tell what the impact of actual retention is.
How valuable are BAs granted by community colleges, as compared with traditional BAs and associate’s degrees?
Researchers managed to cajole some Swiss moms into working more, by giving them “objective information about the long-run costs of reduced labor supply.”
How couples with different education levels split up breadwinning and caretaking duties.
Why are women more skeptical about AI than are men?
Sports gambling may ruin lives, but at least it boosts tax revenue.
Since about 1990, economists have concerned themselves less with efficiency and more with equity.
Quantifying how work requirements affect food-stamp denials.
Timely suggestions for policing protests.
This study suggests there’s only a modest connection between declining fentanyl prevalence (in drug seizures) and declining opioid mortality. But see also last week’s column.
Till next week.