https://orcid.org/0000-0002-3899-0172
![Sample our Social Sciences journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110822/TOC-Subject-Sample-2021-Social-Sciences.jpg"})
Abstract
Although researchers have generated many studies related to police use of force, with an increasing focus on de-escalation, none have sought to systematically assess escalation, related factors, and the extent to which force usage may be considered appropriate from an objectively reasonable framework. Using video-based data (N = 540) from two agencies (Dallas and Smith County, TX) this study employs a Systematic Social Observation (SSO) methodology and incorporates an modified analytical tool (i.e., Resistance Force Comparative Scheme) to assesses de-escalation/escalation. The findings demonstrate that officers most commonly use proportional and incremental force during suspect encounters, neither de-escalating nor escalating. However, when officers do deviate, they are much more likely to escalate as opposed to de-escalate. Further, officers are more apt to use escalated force during proactive encounters, as well as when interacting with suspects who are male and impaired. Additionally, officers with a more negative demeanor (i.e., disrespectful) are more likely to use escalated force. Conversely, officers are less likely to use escalated force when backup officers are on scene, when there is more evidence of a crime present, and when interacting with older suspects. We discuss the training, policy, and research implications of this work, as well as the importance of conducting methodologically rigorous use of force studies within the framework of potential appropriateness.
Disclosure Statement
No potential conflict of interest was reported by the authors.
Notes
1 There are a wide range of studies that have been completed illustrating the utility and realistic nature of employing a SSO methodology, including the President’s Commission Study (Reiss, Citation1968), Midwest City Study (Sykes & Brent, Citation1983), Police Service Study (Ostrom et al., Citation1978), Project on Policing Neighborhoods (Parks et al., Citation1999), Policing in Cincinnati Study (Frank et al., Citation2001), as well as a number of smaller scale studies in Denver (CO), Miami (FL), New York City (NY), Richmond (VA), and Savannah (GA) (see Worden & McLean, Citation2014).
2 As stated by McLaughlin (Citation1992, p. 65) “[a] force continuum is a guideline representing the appropriate amount of force that should be utilized by a law enforcement officer in generic situations. It should provide a means for escalating force when the subject shows noncompliance and a means for de-escalating force when the subject complies.”
3 In his recent iterative examination of force reports in two agencies, Tillyer (Citation2022) also found evidence of escalation on the part of the police. In over 40 percent of exchanges that began with an officer threatening to use or deploying less lethal weapons, hard hand tactics, or deadly force, they were responding to a citizen whose initial resistance was at the defensive level or less (i.e., verbal or passive resistance).
4 For exceptions, see Alpert and Dunham (Citation2004) and Hine et al. (Citation2018b).
5 The Cincinnati, OH study (1997-1998) has been the largest (see Brown & Frank, Citation2006), with smaller studies in Clermont County, OH (1999-2000), Flint, MI (2010-2011), and Spokane, WA (2016) (see Liederbach & Frank, Citation2003; Terrill et al., Citation2014; Todak & James, Citation2018).
6 The current study also incorporates many of these tactics.
7 Taken together, the results have been mixed. For instance, while some have found that the presence of a BWCs led to reductions in use of force instances and citizen complaints (see Ariel et al., Citation2015, Citation2018; Braga et al., Citation2017; Jennings et al., Citation2015; Peterson & Lawrence, Citation2021; White et al., Citation2017), others have not produced such positive findings (see Ariel et al., Citation2016a, Citation2016b, Citation2018; Koslicki et al., Citation2020).
8 Initially there was a total of 530 videos. Ultimately, 92 of them were excluded due to video/audio issues, language barriers, or instances where there was no suspect present. Further, out of the 438 videos that were coded, 22 were of the same incident but from a different officer. In these cases, the research team used all the available video to code the incident, but still treated it as a single encounter. They also coded the collective actions of all officers on scene rather than just the officer who was the original source of the footage.
9 After coding whether an officer proactively initiated the contact, the research team also attempted to identify the initial problem or underlying reason for why the encounter took place. Through supplementary analysis (not tabled), it was found that over 77 percent of proactive encounters were traffic related, seven percent of which were for DUIs. While the research team was not able to discern an initial problem in six percent of these instances, the next highest category for proactive encounters was for suspicious persons/circumstances (5%). As for reactive encounters, the modal categories for the initial problem were for assault/battery (32%), domestic disputes (11%), and medical assistance (11%). These findings were further parsed out to assess the differences between initial problems and the video source (i.e., dashcam versus BWCs). For instance, a higher percentage of initial problems related to non-dui traffic stops (55% to 42%), and domestic assaults (7.7% to 3.4%) were captured on dashcams compared to BWCs. On the other hand, BWCs made up a greater proportion of initial problems concerning public nuisances (9.0% to 0.3%) and mental health crises (8.0% to 2.0%). Important to note is that the differences in initial problems are also likely to be a function of the jurisdictions these video sources were deployed in. It is probable that officers in Dallas (wearing BWCs) who work in a major urban area will face a higher number of public nuisances and mental health crises when contrasted to their rural counterparts in Smith County (using dashcams).
10 All the measures come directly from the videos as opposed to the use of official police records/data. Similar to the approach used in prior SSO research (see Mastrofski et al., Citation1998), observers based their assessments on what they observed and heard. Importantly, while we had hoped to measure suspect race in a more precise manner, we discovered that it was too difficult to determine race beyond white/nonwhite in any sort of reliable or valid manner. More specifically, we found that during the training sessions inter-reliability checks demonstrated that observers often (∼40%) did not agree on a more specific suspect race/ethnicity designation (i.e., distinguish with respect to race/ethnicity at a more specific level). This was a result of challenges such as camera angels, lighting at night/darkness, suspects not always fully appearing on camera or with limited camera exposure (e.g., in a house or car during traffic stop), and the lack of officers explicitly denoting the suspect’s race (e.g., running their identification information through dispatch and verbalizing). Hence, the decision was made to use a white/non-white designation in an attempt to reduce error/misclassification.
11 All measures with a temporal element were coded to ensure they occurred prior to the dependent variable.
12 Note that officer demographics such as race and gender are not included given the group process of managing encounters with several officers (e.g., white male and black female) on the scene during 29 percent of the encounters.
13 Reasoning for use of a sliding scale falls within the realm that force should not only be proportional, but also incremental as argued in prior work (see Terrill, Citation2005; Terrill et al., Citation2003).
14 The following scenarios offer examples of commensurate, escalated, and de-escalated calculations. Commensurate: An officer makes a traffic stop and the suspect complies throughout the encounter and the officer never uses any force. In this case there is one coded sequence (no suspect resistance/officer force command) with the encounter calculated as commensurate. Escalation: An officer questions an initially compliant suspect and commands him to sit down on the sidewalk. The suspect verbally resists and the officer uses physical force to take him to the ground. In this case there are two coded sequences (no suspect resistance/officer force command, verbal suspect resistance/officer takedown) with the encounter calculated as escalation. De-escalation: An officer stops a person believed to have a pending warrant and commands him to sit tight until he runs a warrant check. After the check comes back positive, the officer handcuffs the suspect but when trying to place him into the squad car he physically resists and the officer verbally commands the suspect to get in the car, rather than using physical force to seek compliance. In this case there are three coded sequences (no suspect resistance/officer force command, no suspect resistance/officer force cuff (part of arrest), defensive physical resistance/officer command) with the encounter calculated as de-escalation.
15 Nonetheless, one must be careful not to overstate the relative infrequency in which suspects do resist and pose a threat, as such cases can potentially end with severe injuries and even death.
16 It is important to recall that such force occurred while also accounting within the coding structure for whether the suspect had a weapon, warrant, or was in the processing of being arrested. Thus, while the force used was minimal in nature, there was no discernable reason detected by the observers for such force usage. However, there may have been aspects known or perceived by the officers that were not captured in the video.
17 Escalation in response to resistance could be immediate as well as after repeated suspect resistance. Further, of the 117 escalated force encounters, there was more than one escalation sequence in 18 encounters. When this occurred, the first sequence served as the cut point for the dependent measure (i.e., meaning all the independent variables occurring prior to this escalation were assessed to ensure proper temporal ordering for predictive purposes).
18 Unfortunately, we do not have any data on agency culture or other agency specific priorities that might account for differences across sites.
19 Some of the video and audio limitations were more pronounced based on the source of the footage (i.e., BWC vs. dashcam). For instance, dashcam video was often superior to BWCs when coding encounters that took place in front of the patrol vehicle. In these situations, the stationary dashcam was often able to capture the entirety of the interaction rather than losing detail if an officer wearing a BWC were to shift their stance or leave the encounter to return to their vehicle. On the other hand, dashcam video created limitations when attempting to code interactions that took place away from the patrol vehicle (i.e., on the side of the vehicle, indoors, etc.).
20 DPD commanders expressed concerns about the release of potentially sensitive videos of critical incidents involving officer involved shootings currently under investigation. Such incidents are protected from public release by the internal affairs unit given the evidentiary status during internal investigations.