Studying rare events in policing: the allure and limitations of using body-worn camera video

ABSTRACT

Many intriguing and socially significant policing research questions center on relatively rare events. Scholars may find few viable options to studying those rare events, limiting the development of research to support scientific knowledge and policy responses. The proliferation of body-worn cameras (BWCs) has produced a unique research opportunity through the creation of video archives documenting how officers interact with citizens and police communities. Researchers have access to aspects of police work and behavior previously available only through resource- and time-intensive methodologies, such as field observation. The allure of using video footage is understandably strong, but the limitations of this research methodology need to be understood and ameliorated. This article examines the use of video to support research, focusing on the methodological implications of this emerging research approach. We offer a case study examining police use of force in an east coast department to illustrate the potential and limitations of analyzing video content.

KEYWORDS:

• observational methods
• body-worn camera
• video data analysis

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Citizens report experiencing force greater than a threat or the application of handcuffs in less than 1.2% of encounters with police (Davis, Whyde, and Langton 2018).

2. Within the context of coding from BWC footage, several specific methodologies can be used to guide researchers. Makin, Willits, and Brooks (2020) propose Systematic Social Event Modeling as a methodology for analyzing BWC footage. Chillar, Piza, and Sytsma (2021) suggest the use of a Video Data Analysis framework.

3. NNPD employs over 400 sworn officers serving a diverse population of almost 200,000 residents (approximately 43% white and 42% African American). The agency began to implement BWCs in 2013, reaching full implementation in 2016.

4. Video for a sixtieth reported force event was unavailable because the involved officer had not yet been issued a BWC.

5. The unreported force was identified upon viewing and coding the video. In all 17 events, the force was clearly below NNPD’s reporting threshold, such as the use of open-handed control techniques.

6. Time was recorded by the hour, minute, second indicated in the timestamp of video footage onto coding sheets by the coders. When data were entered into the database for the study, time variables were converted to a number that represented minutes. The seconds were converted to the nearest fraction of the minute in 6 second increments. The coding key was as follows: 0–6 seconds = .0; 7–12 seconds = .1, 13–18 seconds = .2, 19–24 seconds = .3, etc. Thus, an event that was 1:05:14 seconds translated to 65.2 minutes.

7. Prior studies have not been able to account for the duration of force events, consequently there is no clear definition of when force events begin or end. Observers were instructed that force events ended when no officer was in physical contact with the target citizen for 5 or more seconds.

8. Volatile events were initial call types meeting two criteria: (1) higher likelihood of unpredictability at the scene for an officer, and (2) events codes commonly associated with violence and/or use of force. These included calls for domestic disturbances, intoxicated individuals, mentally ill subjects, shootings, assaults, fights/riots, and individuals armed with guns.

9. Observers noted whether they observed each of the following common forms of citizen resistance/non-compliance (See Rojek, Alpert, and Smith 2012) toward the officer(s): ignoring the officer, verbal resistance, physical resistance, and (attempted) fleeing. Resistance is the sum of these observations (range 0–4).

10. For each interaction, observers assessed whether officers or citizens changed their behavior during the encounter because of awareness of a functioning BWC. When a ‘yes’ response was provided, the observer provided a narrative explanation of that justification (see Table 4).

11. Observers recorded whether they observed officers using each of either common types of de-escalation techniques (e.g., Todak and James 2018).

12. Overall, the metrics revealed that there was a high degree of agreement among the three raters for almost all questions on the final coding instrument. Many of the items in the instrument were modeled after the POPN study (e.g., dichotomous yes/no response options). The questions that had more variation in reliability scores were those that asked coders to rate environmental conditions, audio quality, and obstructions to video footage. Additionally, the metrics allowed us to see that one rater had consistently different marking that the other two raters on the items where there was less agreement.

13. Worden and McLean (2017) suggest ways to integrate measures of audio and video quality, subject identifiability, and observer level of confidence into video coding schemes (pp. 204–208).

14. Our observers coded a call of a suicidal subject that included video from three officers, each of whom arrived and departed at different times, independently moved about the citizen’s residence, and had different interactions with both the citizen and their spouse. At times, each officer assumed primary responsibility for the call, handling negotiations with the citizen in varying ways during a lengthy encounter. Repeated viewing of multiple video files was necessary to develop an understanding of the ordering and nuances of events.