Using administrative data collected from 17 US law enforcement agencies, spread across the US, covering over 2348 UOF incidents, this study tested the effect of less-lethal weapons on civilian injury and injury severity, conditional on a robust set of officer, civilian, and situational characteristics. Our findings indicate that the use of chemical agents significantly reduced the likelihood of hospitalization or death for civilians involved in UOF encounters, relative to other less-lethal weapons. In contrast, the use of police canines led to significantly larger increases in both the likelihood of civilian injury and the severity of civilian injury, relative to other less-lethal weapons.

Our findings are largely consistent with prior work on less-lethal weapons [6, 11, 15, 17], suggesting that—on average—chemical agents and ECW/CEDs pose the lowest risk to civilians involved in UOF encounters while police canines pose the highest risk. However, the fact that chemical agents reduced the risk of hospitalization or death significantly more than ECW/CEDs in our study should not be understated. Chemical agents and ECW/CEDs are two of the most common less-lethal weapons employed by police agencies. Recent estimates suggest that over 90% of police agencies authorize the use of both weapon types [14]. However, prior research has indicated that there is little consistency as to where these weapon types exist on UOF continuums or force option models, with some agencies considering both to be comparable levels of force [24, 25]. Despite this, chemical agents and ECW/CEDs have distinctly different mechanisms of effect and potential pathways to injury. Whereas chemical agents incapacitate suspects by causing pain and irritation to the eyes and skin [19], ECW/CEDs produce an electric current that causes involuntary muscle contraction [18]. The risk of severe injury for both weapons has generally been linked to interaction with pre-existing health conditions; however, ECW/CEDs also carry the secondary potential for injury resulting from a subsequent fall, or in rare cases, cardiac disturbances related to the electrical shock [37].

Though we cannot pinpoint how civilian injuries occurred, our results suggest that—for civilians directly involved in individual UOF encounters—chemical agents may be less severe forms of force than ECW/CEDs. Our data indicates, however, that the use of ECW/CEDs is currently much more common than the use of chemical agents during UOF encounters involving less lethal weapons. Further, there is substantial variation in the use of ECW/CEDs and chemical agents across agencies and officers, reflecting differences in policies, training, and discretionary practices regarding the use of these weapons. In our sample, the use of ECW/CEDs ranged from a low of 29% to a high of 98% of cases per agency, while the use of chemical agents ranged from a low of 0% to a high of 48% of cases per agency.

An important complementary question is whether there exists a tradeoff between civilian and officer safety when using less-lethal weapons. Though officer injuries in our data were rare (less than 10% of incidents), we note limited support for this hypothesis. In supplementary regression models (see Supplementary Material), we find that police canines significantly reduced the odds of officer injury, but that no significant differences existed between ECW/CEDs, chemical agents, and impact weapons after controlling for officer, civilian, and situational characteristics. In combination with our main findings, these results suggest that increased use of chemical agents—in place of ECW/CEDs or impact weapons—may reduce the risk of civilian injury without jeopardizing officer safety. However, we caution against making global recommendations from these findings for several reasons.

First, we cannot fully account for the context of less-lethal weapon use across the incidents included in our sample. Recent studies have suggested that injury risk and severity from less-lethal weapons—particularly chemical agents and kinetic impact projectiles—may be greater in crowded settings where larger numbers of civilians are exposed [4, 21, 22]. In these contexts, airborne agents and projectiles may strike unintended targets or linger in urban environments [21, 23, 38]. The incidents analyzed in this study primarily represent encounters between police and individual civilians. As such, we cannot speak to the differential risk of injury when less-lethal weapons are deployed as crowd control weapons or in the vicinity of other civilians. This is particularly salient given the widespread use of chemical agents and impact weapons during recent protests across US cities and college campuses [3, 4]. Second, recent studies have pointed to potential long-term or delayed side effects associated with exposure to chemical agents, which are unlikely to be captured in our data [4]. Though it is possible that these secondary effects are more likely following the use of area-wide chemical agents (e.g., tear gas) rather than handheld OC/pepper spray, we cannot definitively test this hypothesis given the current data. Thus, while chemical agents are associated with less severe injuries, it is unclear whether they lead to greater numbers of injuries secondary to the immediate UOF encounter. This potential complicates broad policy conclusions, and more research is needed.

Similar caution should be exercised with regard to police canines, given that their use corresponds to both an increased risk of civilian injury and a decreased risk of officer injury. In our sample, the use of police canines ranged from a low of 0% to a high of 46% of cases per agency, but we lack the agency-level data needed to understand when and why police canines were used. In general, canines are intended for highly specified situations where other types of less-lethal weapons may not be applicable, such as the detection of suspects at night and/or situations in which the perceived threat to officer or public safety is high [39]. As a result of this, some have argued that police canines deescalate encounters and prevent the use of deadly force in situations where it would otherwise be necessary [39, 40]. This is not a hypothesis that we can directly test in our data, and thus, we can only confidently suggest that—in an effort to reduce civilian injury—the use of police canines should be reserved for situations in which they are truly necessary to resolve an encounter.

One of the most important implications of our findings concerns the distinction between injury prevalence and severity. Prior efforts of this kind are often limited to binary measures of injury [6, 15, 20, 23, 41], which may under or overestimate the risks associated with less-lethal weapons. Our results, for example, indicate that hospitalization data alone may severely underestimate the prevalence of civilian injury, and that for certain weapon types, injury may be more likely than not. In contrast, we also find that the majority of injuries associated with certain weapon types (e.g., chemical agents and ECW/CEDs) appear minor, suggesting the importance of considering both outcomes when structuring UOF policies.

Injury prevalence and severity are also important given the potential for police UOF to perpetuate racial and ethnic health inequities. Though we do not find a significant direct effect of race/ethnicity on injury risk, the majority of the civilians included in our sample were non-White, and our results point to aggregate racial/ethnic differences in the frequency with which certain less-lethal weapons are used. Taken together, these results suggest disproportionate consequences of less-lethal weapon usage for certain people and places. UOF policy and practice should take into consideration the immediate risk of less-lethal weapon usage for the individuals involved, but also the broader application of less-lethal weapons across sociodemographic groups and implications for the communities affected [42].

These findings are not without limitations. First, our sample is not nationally representative and intentionally excludes smaller police agencies. Second, we observed significant variation in injury rates across agencies, possibly pointing to inconsistency in the way that agencies defined and recorded these outcomes (despite our efforts at harmonizing the data). Third, our analyses focused on the most severe form of force used in each encounter, and thus, incidents involving multiple forms of force received only one weapon-type code. Though these incidents were rare, and our analyses accounted for the number of weapons used in each encounter, we were unable to identify which weapon was responsible for civilian injury when multiple were used. Our data are also inherently non-experimental and subject to omitted variable bias. It is possible that additional unobserved factors may confound our relationships of interest.

Finally, our study focuses only on the use of less-lethal weapons and their efficacy for reducing civilian injury. As such, we do not consider other forms of force such as hands-on tactics, which are used in the majority of UOF encounters and may account for the majority of civilian injuries [11]. We also do not consider the failure rate of less-lethal weapons or their effectiveness in successfully apprehending civilians. This is an important consideration for law enforcement practitioners, and additional research comparing both injury risk and effectiveness across weapon types is needed. Nonetheless, multi-agency data on civilian injury in UOF encounters is difficult to collect, and this study represents one of the largest and most current analyses of these relationships to date.