There has been significant discussion and some media reports regarding a recent research project by Brian Landers, a former police officer and current chair of a college criminal justice department.
Lander’s graduate capstone, “An Analysis of a Nation-Wide Use of Force De-Escalation Policy and the Impact on Officer Safety,” takes aim at some of the Police Executive Research Forum’s (PERF) de-escalation language in PERF’s Guiding Principles on Use of Force report and its inclusion in law enforcement policy.
Landers’ capstone project’s stated purpose is to test the hypothesis that “…officer safety is increased when force de-escalation policies are in effect, and therefore officer deaths and injuries should be reduced.”
While the report’s narrative indicates the hypothesis is disproven, some media reports and social media discussion boards have skewed the results to a finding that de-escalation policies increase officer injury and death. I do not believe either to be true based on the data or the method of evaluation.
I spoke with Landers, and it was evident he is concerned about the direction of use of force policy language – as we all should be. Don’t misinterpret my critique of Landers study as being for or against de-escalation training. Various de-escalation techniques are important, but a policy or belief that “thou shalt de-escalate” every situation is also problematic. Regardless, it is important to know that Landers agreed that his research does not empirically prove de-escalation policies cause increased officer injuries and death. However, we both agree there is a need for future study and that innovative law enforcement agencies must take the lead in evidence-based policing.
Understanding research findings on de-escalation
To understand why Landers’ project does not empirically prove de-escalation policy places officers in danger, let’s first review Joel Shults’ article on interpreting research findings. Placing some of Shults’ comments in context, Landers’ capstone project does not conduct the inferential statistical evaluation necessary to indicate correlation or causation.
A preferable manner to establish causation through experimental research is by gathering initial data (e.g., officer killed/injured), introducing a treatment (e.g., de-escalation policy) and then determining if a statistically significant change occurred.
Control variables (e.g., de-escalation techniques used or not) should be included to ensure the “treatment” caused the change, if it exists.
Landers’ capstone project provides data on law enforcement officers injured and killed in several cities in the United States. A portion of the cities implemented de-escalation language at some point during the evaluation range.
Part of Landers’ narrative presents descriptive data as frequencies (e.g., officers killed) that are shown before and after the policy change.
Questions we must ask include:
- Were all the injuries related to suspect interaction?
- Were de-escalation tactics used?
- Were any changes significant?
We also want to know if crime rates, calls for service or proactive policing increased or decreased during the same time.
None of these questions are answered in Landers’ capstone project. Answering them requires access to departmental data and intensive research. However, it is possible to restructure Landers’ data in a more realistic way.
To do so, I averaged the total number of officers killed and injured (all years provided) for each department that implemented de-escalation policies and compared the average to the number killed/injured after de-escalation policies were in effect. If the de-escalation policy was implemented after mid-year (e.g., June or beyond), I only counted the following full years post-implementation.
Two points were immediately clear:
1. There is not a lot of pre/post data to compare;
2. Some data should be removed all together based on its relevance. For instance, the four Dallas PD officers killed by an active shooter are removed from my analysis as those deaths cannot be attributed to de-escalation policy in this context.
Let me be clear, there are significant limitations with this method and I am only using it to present a more reasonable evaluation of change pre/post de-escalation policy. The results of my analysis using Landers’ data show agencies with de-escalation policy had decreases in officers injured and killed from pre-policy averages to post-policy yearly totals. However, control variables such as police proactivity, crime rates and de-escalation specific situations are not included, making this analysis severely limited. My results do not show that de-escalation policy is associated with the reduction in any way; it simply shows a trend that could be attributable to many things including de-escalation policy. It should be noted that trends in some of the agencies without de-escalation policy also showed significant reductions of officers injured and killed in the same time period (e.g., Chicago, Milwaukee and Tucson).
*The only officers killed in the line of duty at Dallas PD were the victim of an active shooter and are excluded as they cannot reasonably be attributed to de-escalation language in use of force policy.
What is de-escalation?
Landers’ project presents two important issues regarding de-escalation and associated language placed in policy:
1. The failure to define de-escalation or de-escalation training adequately.
Landers’ research indicates de-escalation has two common themes: slow the incident down and reduce or avoid force. While this may be a good start, de-escalation training can consist of mental illness recognition, communication techniques, active listening, tactics (time/distance), resilience, or a combination of all the above. One cannot research de-escalation without defining its parameters and objectives to ensure they are present in the evaluation.
2. A need to determine the outcomes we expect from de-escalation techniques to measure them.
Landers’ project indicates officer safety is an important measurable outcome and I agree. What about injuries to suspects or the necessity to arrest vs. mental health assistance? Do law enforcements agencies implementing de-escalation concepts know whether this meets objectives?
To that end, my own research attempting to answer this question resulted in finding a lack of empirical study on the effectiveness of de-escalation or associated platforms.
For instance, in their article “Mental Illness, Police Use of Force, and Citizen Injury,” Michael Rossler and William Terrill state empirical evidence of the benefit is limited regarding techniques such as verbal mediation to lower citizen resistance. Sema Taheri systematically reviewed the empirical evidence on crisis intervention training (Memphis Model) finding that, “CIT has no effect on outcomes of arrest, nor on officer use of force… these results raise some concern about the widespread implementation of CITs.”
Do other industries have evidence of the benefit of de-escalation practices and training?
The health care industry, especially psychiatric health, has been studying de-escalation techniques since the 1980s, seeing similar results to law enforcement. For instance, a study from the health care industry synthesized previous research on de-escalation training and determined there were a “lack of trials conducted under rigorous experimental conditions.” Another study on de-escalation training in psychiatric health care concluded the training had no effect. Lastly, a recent study stated, “While a number of theoretical models have been proposed, the lack of advances made in developing a robust evidence base for the efficacy of de-escalation is striking and must, at least in part, be credited to the lack of a clear conceptualization of the term.”
Conclusion
Knowing law enforcement as a former practitioner, I’m sure I’ve angered folks on both sides of the de-escalation topic – my former peers will certainly recognize the personality trait. However, this is intended to be an unbiased review of both sides of the debate. It’s time to put emotion, social rhetoric, politics and opinion aside to make way for evidence-based policing.
Landers and I both share a concern for some of the Police Executive Research Forum (PERF) recommended policy language and training being implemented across the nation. My primary concern is the absence of peer-reviewed research on its effects (officer and public safety). This is not to be taken that I think all the PERF concepts in the report are unwise for implementation. I have been and always will be a proponent of slowing down, creating distance, gathering information and deploying resources when the subject’s actions allow for it. However, there is a dire need for realistic, generalizable, peer-reviewed research in law enforcement that speaks to policy, practice, training and tactics in this regard.
The way we’ve always done it has been criticized as faulty thinking but I opine change without solid evidence is also faulty thinking. I ask law enforcement to take a proactive position in collecting data and collaborating with academia to conduct research.
Law enforcement agencies hold the keys to the internal statistics that can be used to answer questions regarding many contemporary issues. Academia holds the keys to making sure the research is valid and generalizable for implementation. A partnership between the two ensures the appropriate contextual variables are present and field testing and evaluation of new ideas is conducted properly. Only through such a partnership can evidence-based policing practices come to fruition.
Be safe. Be vigilant.
