Force in 0.05% of calls is one call in 2,000 [9][1]. If that share and the agitation share rest on the same denominator, then even among the roughly 2.7% of encounters that arrive agitated, force shows up in fewer than two of every hundred [8][2]. The rarity is why the counting exercise was worth three years. It is also what makes the study's second claim hard to close out.
Capt. James Pittman told FSU that moving the rate from 0.05% to 0.04% would justify every dollar and all three years [10]. That is a 20% relative reduction [3], and it is not a change any single agency can see in its own annual totals: at those base rates, a trained group and an untrained group would each need on the order of 700,000 calls before a 20% difference cleared conventional significance at 80% power [4]. Which is why the workable outcome variable for anyone adopting this is technique counts rather than force itself.
There the association reported by the researchers, that deputies who used more techniques were more likely to resolve encounters without force [11], can run in either direction. A subject who is already settling gives an officer room to ask questions, hand over a voice in what happens next, and move them away from whatever set them off [11]. Second-by-second coding across multiple blind passes [6] is exactly the instrument that could break that tie, by establishing whether the technique preceded the subject's shift or followed it, and FSU's account does not say whether the analysis ordered them that way [16].
The account also omits an inter-rater reliability figure for the ten coders, which is the number blind passes exist to produce [6][16], and it gives no size or design for the in-study training trial that preceded the agencywide rollout [12][16]. The population behind the percentages is unstated too: 0.05% of 1,000 recordings is half an event [5], so the force rate has to come from a wider set of call records than the footage that was coded.
What survives all of that is the first half of the work: an operational definition of de-escalation as any action, statement or consideration that lowers the likelihood or severity of force [5], a coding scheme ten people can apply to unredacted footage [6], and base rates for calm, agitated and force outcomes [7][8][9]. Use-of-force reporting cannot produce those numbers by construction, because it records only the encounters where force happened. Part-funding from the Justice Department's Office of Community Oriented Policing Services [4] gives the manual a plausible route to other agencies. On the evidence released so far, I would call de-escalation measurable and the measurement exportable; the stronger word in FSU's summary, teachable [15], is carried by the training trial, and that trial's numbers are in the manuscripts Sheppard and Pittman are still writing [13].
