A Decade After Ferguson, We Still Barely Know Whether Police De-Escalation Training Works

Roll the clock back to 2020, when police departments across America were signing up for de-escalation courses as fast as trainers could deliver them, and ask a simple question: did any of it actually work? A team of researchers went looking for the answer that year. They combed the literature and found 64 separate evaluations of de-escalation training across the professions, most of them in nursing and psychiatry. The number involving actual police officers was zero.

That absence is at the heart of a new policy brief from Robin Engel and Jennifer Cherkauskas of the Niskanen Center. Years of reform momentum, hundreds of agencies retrained, eighteen states writing mandates into law, and almost nothing a scientist would call evidence.

De-escalation sounds like the most sensible idea in policing. Slow down the encounter. Talk instead of grab. Use time, distance and cover so an officer facing someone in crisis has more options than a trigger. The trouble is nobody quite agrees what the word means. Engel and Cherkauskas line up more than a dozen official definitions: some frame de-escalation as an outcome (the peaceful resolution), some as a set of techniques (active listening, tactical repositioning), some as both at once. When a term can mean that many things, measuring whether it works is hard indeed.

Still, the past five years have finally produced something to measure. Studies have begun to accumulate, and a handful are rigorous enough to trust.

Two Programmes Out of Six

Here’s what the strongest studies show. Of all the de-escalation curricula rolled out across the country, only six have ever been tested for their effect on officers’ actual use of force in the field, as opposed to how officers feel after a day in a classroom. Of those six, just two produced consistent reductions. One is ICAT, the Integrating Communications, Assessment and Tactics programme built by the Police Executive Research Forum. The other is Situational Decision-Making, developed by the University of Chicago Crime Lab.

The ICAT numbers, where they land well, land hard. In Louisville, trained officers used force 28% less often, injured the people they encountered 26% less often, and, tellingly, hurt themselves 36% less often. The Chicago programme cut use of force by roughly a quarter and officer injuries by nearly half.

This cuts against the oldest objection to de-escalation: that teaching officers to slow down will get them killed. Critics have argued since the mid-2010s that these tactics encourage hesitation and retreat, raising the risk to the officer rather than lowering it. The field data so far point the other way: the programmes that reduce force on the public tend to reduce injuries to police as well. It is a counterintuitive result, and is the single most useful thing to have emerged from a decade of study. The other four programmes tested, though, showed no reliable effect on use of force at all. Whatever de-escalation is, it’s plainly not a magic word.

And even the wins are fragile. Officers who leave training full of enthusiasm tend to drift back toward old habits within months, a pattern the researchers call training decay.

The thing that seems to hold the gains in place is not the course itself but what happens afterwards. In Indianapolis, officers whose sergeants actively reinforced the training were 1.7 times more likely to keep using de-escalation on the street. The catch, and it’s a big one, is that fewer than a quarter of supervisors reported doing that reinforcing regularly, even while insisting they believed in the skills.

Money Without a Map

None of this has stopped Washington from spending. The bipartisan Law Enforcement De-Escalation Training Act sailed through the Senate by unanimous consent in late 2022, and the Justice Department has since pushed more than $30 million out to the field, along with roughly $2 million to build national standards for approving de-escalation curricula. Engel and Cherkauskas are blunt about the risk in that sequence. Setting up an approval bureaucracy for a practice you can’t yet define or reliably measure, they warn, means “this federal investment prioritizes administrative infrastructure over evidence development.” Standards first, evidence later, is a strange way round.

Now add politics. Federal immigration enforcement has grown fast since late 2024. The government hired about 12,000 new agents and shortened their training. These agents usually get much less de-escalation and crisis training than officers in a mid-sized city police force would.

In January 2026, federal agents in Minneapolis shot and killed two American citizens. Those deaths turned the issue into a political flashpoint. At the same time, the Justice Department’s research arm has canceled research grants and stopped announcing new funding. The authors conclude that federal support for this training now looks “deeply unstable.”

That leaves the field in an odd place. The evidence is finally starting to look good, right as the funding and the appetite for it dry up. Engel and Cherkauskas think the way to save de-escalation is to stop selling it as a restraint on police and start selling it as something that keeps officers alive and out of the hospital, an argument the injury data increasingly supports.

Frequently Asked Questions