ABSTRACT

The current study analyzes police use of force as a series of time-bound transactions between officers, civilians, and bystanders. The research begins with a systematic social observation of use of force events recorded on police body-worn cameras in Newark, NJ. Researchers measure the occurrence and time stamps for numerous participant physical and verbal behaviors. Data are converted into a longitudinal panel format measuring all observed behaviors in five-second intervals. Panel logistic regression models estimate the effect of each behavior on use of force in immediate and subsequent temporal periods. Findings indicate certain variables influence use of force at a distinct point in time whereas others exert influence on use of force across multiple time periods. The most influential variables relate to authority maintenance theoretical constructs. This finding supports prior perspectives arguing that police use of force largely results from officer attempts to maintain constant authority over civilians during face-to-face encounters. Nonetheless, a range of additional variables reflecting procedural justice, civilian resistance, and bystander presence significantly affect when police use force during civilian encounters. Results provide nuance to theoretical frameworks considering use of force as resulting from the interplay between officer and civilian actions and reactions.

KEYWORDS

body-worn cameras, panel regression, police use of force, systematic social observation, video data analysis

CITATION

Piza, E.L., Connealy, N.T., Sytsma, V.A., and Chillar, V.F. (2022) Situational Factors and Police Use of Force Across Micro-Time Intervals: A Video Systematic Social Observation and Panel Regression Analysis. Criminology. https://doi.org/10.1111/1745-9125.12323

1 | INTRODUCTION

Use of force is arguably the most critical issue in contemporary policing, with recent high-profile use of force events leading to wide-spread calls for reform (Bennell et al., 2021) and street-level activism focused on changing the role of police in public safety (Cobbina, 2019). Colloquial discussions of police use of force typically revolve around whether an incident violates legal standards. Legal standards alone, however, provide an incomplete picture (Stoughton, Noble, and Alpert, 2020). For example, the U.S. Supreme Court declined to regulate police officer perceptions of imminent danger as it relates to the use of deadly force on two separate occasions, first in Grahm v. Connor (1989) and then in Kisela v. Hughes (2018) (Sherman, 2020). In other words, situational factors that influence an officer’s assessment of the need for force, as well as the officer’s tactical responses to such factors, are not guided by court rulings (Garrett & Stoughton, 2018).

While agency policy and legislation constrain use of force to particular types of situations, police officers enjoy a great deal of discretion in most instances (Reiss 1980). Situational characteristics of police-civilian encounters are widely considered the most salient influence on police use of force (Bennell et al., 2021; Bolger, 2015). Prior research has consistently found that police-citizen encounters involving force typically extend across fairly lengthy time periods, which contradicts common narratives that police use of force decisions must be made in a “split-second” manner (Stoughton, 2014). During these extended periods, however, police officers must rapidly assess a series of dynamic situational factors, requiring the use of cognitive shortcuts that can influence later use of force decisions (Mears et al., 2017). As such, use of force is not a single decision, but the result of a series of interactions between an officer and a subject (Stoughton et al., 2020). In this regard, prior research has conceptualized use of force as a transactional event resulting from the interplay of officer and civilian actions (Terrill, 2005).

Situational factors, and the cognitive shortcuts they motivate, can ebb and flow throughout a police-civilian encounter, differentially influencing the likelihood of force at various points. In this regard, prior research has conceptualized use of force as a transactional event resulting from the interplay of officer and civilian actions (Binder and Scharf, 1980). Common data sources typically only allow for the cross-sectional analysis of use of force events, precluding the use of longitudinal methods that can sufficiently measure transactional events. With both police discretion and data limitations presenting challenges, researchers and practitioners have struggled to understand the situational factors that drive use of force events (Sherman, 2018). This shortcoming has continued to enable controversial use of force events that strain police and community relations.

The current study reports on a Systematic Social Observation (SSO) of body-worn camera (BWC) footage and panel regression analysis in Newark, NJ. This analysis incorporates a unique dataset containing many novel variables and predictors of force. The unique structure of the dataset allows for a new way of examining the situational nature of police-citizen encounters, which to our knowledge has yet to be done in use of force research. Rather than analyze encounters as cross-sectional moments in time, we parsed out the entirety of each individual use of force event into a series of five-second time intervals to examine encounters longitudinally. Each use of force event was then broken down into a chronological sequence of intervals spanning the duration of the encounter. Data on the actions of police officers, suspects, and bystanders, which were collected during the SSO, were scored into their appropriate corresponding time interval(s) based on the start and/or end time of the given action or behavior. We believe both the study methodology and the findings have important implications for research and policy on police use of force.

2 | REVIEW OF RELEVANT LITERATURE

2.1 | Police Use of Force as a Transactional Event

Binder and Scharf’s (1980, 111) classic framework considers use of force as a “developmental process in which successive decisions and behaviors by either [a] police officer or civilian, or both, make the violent outcome more likely.” Within this context, empirical evaluations require measuring and assessing what happened in the seconds or minutes that preceded force as well as how the actions and behaviors of officers (Stoughton et al., 2020) and the civilians they engage with (Bennell et al., 2021) influence the probability of force being used. Binder and Scharf (1980) contended that police use of force involves four distinct, sequential phases: anticipation, entry, information exchange, and final frame decision. The anticipation phase begins when officers realize they will be interacting with a civilian and start gathering information about the location, involved individuals, and further context about the interaction. The entry phase marks when the encounter begins, with officers deciding how to approach the civilian. The information exchange phase involves officers and civilians interacting with one another and reacting to verbal and nonverbal cues. The final frame is the culmination of the prior phases and involves the officer’s decision to use force as well as what type of force.1

The transactional nature of use of force is further informed by psychological perspectives that consider how humans process information. Scholars have noted the importance of the observe, orient, decide, and act (OODA) loop in understanding police use of force (Garrett and Stoughton, 2017; Stoughton et al., 2020). The OODA loop demonstrates that individuals actively and passively gather information (observe) before processing and interpreting that information to draw conclusions (orient). Individuals then determine how they will react to these conclusions (decide) before putting their decisions into motion (act). The OODA framework has been incorporated by researchers to analyze various aspects of legal analysis (Chestek, 2020), inclusive of police tactics during use of force events (Garrett and Stoughton, 2017).

The OODA loop complements prior perspectives by conceptualizing how situational factors may accelerate or decelerate the movement of police-civilian encounters from the information exchange phase to the final frame. For example, a driver quickly reaching for their glove box during a traffic stop forces an officer to speed up their decision making to quickly ascertain whether such behavior constitutes a significant threat worthy of force (Stoughton et al., 2020). The OODA loop can also be strategically applied in a manner that influences the decision making of other involved parties (Chestek, 2020). Officers can slow down the OODA loop by “creating time” in a manner that affords them the ability to consider a larger range of information before making a final decision (Garrett and Stoughton, 2017) or by informing civilians of BWC presence to potentially generate a civilizing effect (Patterson and White, 2021). Situational factors tangential to police and civilian behaviors can also influence the OODA loop, like the arrival of additional officers on scene which can generate a civilizing effect that may alter the subsequent behaviors of both the civilians and officers present (Bayley and Garofalo, 1989; Terrill, 2005).

Criminological theories on police use of force reflect the transactional nature suggested by these psychological frameworks. Civilian resistance has featured prominently in this body of scholarship. Sykes and Clark (1975) developed the theory of deference exchange, building upon Goffman’s (1961) work on social interactions. Sykes and Clark (1975) argued that when the police encounter a civilian there is an expectation—on the part of officers and the general public—of deference. Civilians refusing to show deference openly reject the principles of the legal and moral community. Conversely, as the higher-status actor in the interaction, officers are not obliged to show the same deference to the civilians, which may escalate initial civilian resistance. Sykes and Brent (1983) demonstrate that responses of officers and civilians are congruent to their status, with each party’s actions directly influencing the other’s behavior.

Congruent to civilian resistance, several studies have considered use of force to primarily emerge in response to officer attempts to maintain charge throughout the situation. In their conceptualization of “taking charge” Sykes and Brent (1980, 186-187) presented three “regulation sequences” that officers use in response to various “disturbance types”: definitional regulation, imperative regulation, and coercive regulation. Definitional regulation refers to a cognitive approach to taking charge and includes officer actions such as asking questions of the civilian or making an accusation against the civilian. Imperative regulation refers to directly intervening in civilian actions by giving a command/directive or making a non-violent threat. Coercive regulation is use of force or threat of use of force. Sykes and Brent (1980) theorized that the use of various approaches to regulating an interaction will proportionally depend on the type of disturbance the officer has encountered and how the civilian responds during a particular regulation sequence. For instance, if the officer’s goal is to collect information and the civilian repeatedly refuses to provide the requested information, the officer may escalate from definitional (questions) to imperative (commands) regulation.

Alpert and Dunham (2004) developed the authority maintenance theory, which extends the prior notion that control of the scene is the most important consideration for police officers. This theory acknowledges the exaggerated role that authority plays in police-civilian interactions, as well as the officer’s primary concern of maintaining their authoritative edge over the public. This theory postulates that police-civilian encounters unfold according to social rituals requiring each party to show respect and regard to one another (Goffman, 1959). Authority maintenance rituals are contingent on the principle of reciprocity, with police and civilians each able to anticipate each other’s expectations within an interaction. Alpert & Dunham (2004) considered police-civilian interactions as a unique type of social interaction because expectations and behaviors commonly violate the principle of reciprocity, owing to the officer’s inherent authority. When reciprocity breaks down, the likelihood of officer use of force and/or civilian resistance increases.

Empirical research consistently indicates a positive relationship between civilian resistance and use of force. Alpert and Dunham (1997) developed the Force Factor, a measure of the level of police use of force relative to the level of civilian resistance, measuring both officer and civilian actions on a four-part Likert scale. The scale was expanded to six parts by Alpert & Dunham (2004), who found that most officers began encounters by giving civilians verbal directives, often increased their level of force during the early stages of an encounter, and used a level of force that typically reflected the level of civilian resistance. Three particularly noteworthy and related studies emerged from the Project on Policing Neighborhoods (POPN), which involved in-person SSOs and interviews with patrol officers. Terrill and Mastrofski (2002) found that officers frequently use force in response to civilian resistance, but they were not more coercive toward disrespectful civilians. Terrill (2003) examined the extent and variation between police force and civilian resistance, with an emphasis on the interplay between the two measures. Civilians display some form of resistance in only 12% of encounters. Although, encounters that begin with some form of coercion result in a greater frequency of subsequent civilian resistance. In a follow-up study, Terrill (2005) focused on situational influences of use of force escalation. Officers escalate force in about one in five encounters involving nonresistant civilians, and de-escalated the level of force in three of four encounters involving resistant civilians. Regression models indicate a range of situational factors affected the manner in which officers apply force.

Close assessment of the methodologies used in prior research shows operationalizations of civilian resistance oftentimes speak to the authority maintenance and “taking charge” concepts in addition to resistance. Actions such as physical assault and aggravated resistance with a weapon appear at the high end of the force factor, with actions such as verbal noncompliance and passive resistance comprising the lower end of the scale. A similar type of classification has been used in other studies applying variations of the force factor (see e.g. Bazley et al., 2007; Terrill, 2001, 2005; Terrill and Zimmerman, 2021; Tillyer, 2022; Wolf et al., 2009) as well as the POPN studies that used alternative operationalizations of the resistance-force relationship (Terrill, 2003; Terrill and Mastrofski, 2002). The higher-end resistance arguably makes use of force situationally necessary, which reflects theoretical perspectives that consider resistance as a direct cause of force. The lower end resistance, conversely, provides officers at least some discretion in how to handle the situation. Force resulting from such cases reflect the authority maintenance and taking charge reasoning offered in prior research (Alpert & Dunham, 2004; Paoline, 2003; Sykes & Brent, 1980).

Although we argue that use of force is situational and transactional in nature, it is worth noting the numerous external factors which exert influence over the interaction at the ecological and organizational levels. Garner and colleagues (2002, p. 727) presented perhaps the most comprehensive list of use of force correlates, which they organize across five domains: type of location, nature of the offense, police mobilization (one example indicator of “police mobilization” is the number of officers on the scene), characteristics of the officer, and characteristics of the suspect. Predictors of both prevalence and severity of force most relevant to what we term “contextual variables” (see section 4.5) include day of the week (weekend compared with weekday), age of officer, and sex of both officer and suspect (Garner et al., 2002). That said, other work has found that officer characteristics have little importance in predicting use of force when compared with visibility of the incident (Friedrich, 1980). In addition, work by Nouri (2021) indicates that use of force is more probable in census tracts that are primarily designated for commercial land use, in census tracts that are characterized by concentrated disadvantage, and in census tracts with higher rates of violent crime. Similarly, early research by Reiss (1972) found that police were more likely to use excessive force in private settings compared with in public ones. These findings are echoed elsewhere, along with time of day and location: Use of force is more likely when the incident takes place during nighttime hours and indoors (Quinton et al., 2020; Sytsma et al., 2021a). Other ecological predictors of use of force include the number of officers on the scene (Bolger, 2015; Brandl & Stroshine, 2017; Garner et al., 2002; Terrill, 2005), whether a subject was arrested (Bolger, 2015), proximity (White & Ready, 2010), and lighting conditions (Garner et al., 2002; Sytsma et al., 2022).

At the organizational level, clearly articulated use of force policies, which are regularly reviewed and revised as needed, can reduce use of force (Prenzler, Porter, & Alpert, 2013). For example, Terrill and Paoline (2017) demonstrate that officers working in departments with restrictive use of force policies used less force versus those who worked in more permissive environments. Similarly, Nowacki (2015) found that more restrictive firearm policies (e.g., drawing a firearm and pointing it at civilians) were associated with fewer uses of deadly force by officers. In addition, accountability tools, such as use of force reports that require officers to articulate justification for use of force, have been found to decrease the number of complaints of excessive force and the number of citizen injuries (Prenzler et al., 2013). That said, organizational predictors of force can be complicated by police culture, which privileges officer safety from violence above all else (Sierra-Arévalo, 2021, 75). Termed by Sierra-Arévalo (2021) as the “danger imperative”, this cultural emphasis on the constant potential for violence can result in the inability of officers to carry out objective and/or accurate risk-assessments (Sierra-Arévalo, 2019; 2021), thus increasing the likelihood of use of force. The preoccupation of imminent death and danger is not a new phenomenon, however, and has been documented in the literature for more than five decades (Sierra-Arevalo, 2021; Skolnick, 1966).

The structure of an organization can also influence the behavior of its members (e.g., use of force by police officers). Merton (1940) posited that organizational factors and structure caused changes in the personality of individual members within an organization. Organizational theorists have referred to this phenomenon as organizational identification (Cheney, 1983; Dutton et al., 1994). According to Pratt (1998, p.172), the process of organizational identification “occurs when an individual’s beliefs about his or her organization become self-referential or self-defining”. March and Simon (1993) highlighted that larger organizations that yield distinguishable products produce members with a stronger sense of identification with the organization (e.g., police as crime fighters). Thus, officers who work in environments where the use of force is more permissible and is accompanied by a culture steeped in the “danger imperative”, may be more willing to use force during interactions with the public. In other words, violence is structured. Reiss (1980) noted that for police organizations to curtail the use of force, they must alter their structure in such a way that applies additional control over officer’s opportunities (e.g., policies and directives) and decision-making process to use force, including lethal force.

Tactical approaches that can diffuse and de-escalate police-civilian encounters have further informed contemporary use of force research. A systematic review conducted by Engel and colleagues (2020) found de-escalation training typically reduces the severity of incidents and the number of injuries suffered by both staff and clients across a range of professions. Specific to policing, Goh (2021) found the implementation of de-escalation training by the Camden, NJ Police Department led to a 40% reduction in use of force events compared with a synthetic control condition. Central to notions of de-escalation is officer adherence to procedural justice, commonly defined as “perceived fairness of the procedures involved in decision-making and implementation, and the treatment people receive from the authority” (Murphy et al., 2008, 139). Officers whose decisions reflect procedural justice are typically more legitimate in the eyes of civilians (Tyler and Fagan, 2008), which facilitates compliance and cooperation (McCluskey et al., 2019). Police officers can promote positive perceptions of procedural justice by consistently treating civilians with respect and dignity (Liang et al., 2021; Tyler and Jackson, 2014), providing civilians with accurate and useful information (Liang and Ma, 2021), and providing civilians opportunities for participation through asking questions during the encounter (McCluskey et al., 2019).

Findings of systematic reviews indicate procedural justice positively affects civilian perceptions of police legitimacy, civilian compliance and cooperation with police, and civilian satisfaction and confidence with rendered police services (Mazerolle et al., 2013; Walters and Bolger, 2019). Recent scholarship has indicated procedural justice may provide direct benefits for use of force. Wood and colleagues (2020) found a procedural justice training program implemented by the Chicago Police Department reduced police use of force against civilians. Observational use of force research has found officer behaviors aligning with procedural justice can generate benefits inclusive of reduced likelihood of force (Schafer et al., 2022), reduced risk of force escalation (Sytsma et al., 2021a), and increased time to physical force during police-civilian encounters (Piza & Sytsma, 2022).

2.2 | Body-Worn Cameras and Video Data Analysis in Use of Force Research

BWCs hold great promise as a data source for researchers wishing to contribute to the knowledge base on police-civilian encounters. Limitations of common data sources have led researchers to treat use of force as a unilateral event rather than as a process of a complex system emerging from the collective behavior of involved parties (Bennell et al., 2021). The analysis of BWC footage allows police-civilian encounters to be modeled as fluid processes, emphasizing the sequence of police-civilian interactions scholars consider paramount (Alpert and Dunham, 2004; Terrill and Zimmerman, 2021; TIllyer, 2022).

The use of BWC as a data source fits within the emerging video data analysis (VDA) framework (Nassauer and Legewie, 2019, 2021). Although some prior SSOs used recorded video as a data source, such footage is typically created explicitly for research purposes (e.g., Sampson and Raudenbush, 1999; Schnell et al., 2019; St. Jean, 2007). VDA instead involves the analysis of pre-existing footage to understand how phenomena of interest transpire within their native settings (Nassauer and Legewie, 2019, 2021). Researchers can overcome traditional limitations of qualitative research—such as recall error, the Hawthorne effect, and embellishment (or blatant dishonesty) on the part of research participants—by leveraging video footage in SSO (Chillar et al., 2021; Piza and Sytsma, 2016).

Several recent use of force studies have leveraged BWC footage. Todak and White (2019) reviewed BWC videos with officers nominated by their peers as skilled in de-escalation to explore their perceptions of the tactics that impact the effectiveness of de-escalation strategies. Officers identified a broad range of tactics that could bring calm to a situation and allow conflicts to be resolved through the least amount of force possible (also see commentary #3 by Michael White in Bennell et al., 2021). Mangels and colleagues (2022) showed use of force scenarios taken from BWC footage to officers identified as either expert or novice decision-makers. Expert officers more frequently identified the importance of force mitigation opportunities and used words associated with verbal de-escalation, whereas novices more frequently used words associated with physical control. McLean and colleagues (2022) measured officer perceptions of the civilian’s level of resistance, threat the civilian posed to officers, and the necessity of physical force after viewing confrontational police-civilian encounters captured with BWCs. McLean and colleagues (2022) found high levels of agreement regarding civilian resistance and threat levels, but significant heterogeneity on officer assessments of force necessity. Sytsma and colleagues’ (2021b) SSO of BWC footage measured officer adherence to procedural justice standards during use of force events. They found that a slim majority of use of force events were procedurally just, but certain standards (e.g., addressing civilian concerns and using respectful language) were observed in few cases.

Other studies have conducted SSOs of BWCs to analyze correlates of use of force outcomes. Willits and Makin (2018) used temporal sequencing of unedited BWC footage to identify incident characteristics that influence how quickly into an interaction officers used force, the duration of that force, and the type and severity of force used. They found that force is used more quickly against Black suspects and males, suspect resistance predicts both time to and duration of force, and police will likely to use greater levels of force in more time drawn-out interactions. Piza & Sytsma (2022) used similar survival analysis techniques to measure the effect of civilian resistance and police adherence to principles of informational and interpersonal justice on time to the use of physical force during police-civilian encounters. They found officer adherence to informational justice was significantly associated with a longer time until both first use of force and highest level of force. Although civilian resistance did not achieve statistical significance in any model, civilian weapon possession was significantly associated with shorter time until the highest level of force. Sytsma and colleagues (2021a) used data generated from BWC footage to conduct a script analysis of police officer force escalation from soft empty-hand control to more severe types of force. Sytsma and colleagues (2021a) found evidence that civilian impairment was an important contributor to increased risk of force escalation, procedurally just officer behavior was associated with the lowest risk of escalation, presence of a victim increased the risk of escalation, and the presence of non-antagonistic bystanders minimized the risk of escalation. Schafer and colleauges’ (2022) analysis of BWC footage found force occurred significantly more often when civilians actively resisted officers, and when the police-civilian encounter occurred in response to a call type commonly associated with violence and/or use of force.

3 | SCOPE OF THE CURRENT STUDY

The current study analyzes police use of force through the chronological sequencing of individual force events into a series of five-second intervals.2 While a recent VDA of CCTV footage capturing bystander intervention in violent conflicts similarly incorporated five-second panel data (Lindegaard et al. 2021), BWC footage has yet to be evaluated through such a technique. Our panel methodology has important implications for use of force research, as scholars have noted empirically measuring contemporary theoretical perspectives requires “data collection and analysis on the sequence of events unfolding in police-citizen interactions and on how the ordering of events affects the outcome of force” (Alpert and Dunham, 2004, 14). Panel models allow for police-civilian encounters to be analyzed in a manner that reflects the transactional nature of use of force as observed in prior research (Binder and Scharf, 1980; Stoughton et al., 2020; Tillyer, 2022). Prior research has effectively incorporated the sequential ordering of civilian resistance and police use of force (Alpert and Dunham, 2004; Terrill, 2005; Tillyer, 2022; Willits and Makin, 2018), but such studies typically measure how often civilian resistance occurred prior to the use of force. The precise nature of temporality (i.e., whether the effect of resistance is immediate or delayed) remains elusive in such research designs. Furthermore, such an approach does not measure the effect of situational factors other than the relative level of civilian resistance have on the likelihood of force. The current study builds upon such work by examining the influence of actions and behaviors the instant they occur, during the time of their ongoing occurrence, after they occur, and in their frequency of occurrence.

4 | DATA AND METHOD

4.1 | Data Collection

The study sample includes 91 use of force events recorded on BWCs in Newark, NJ between December 2017 and December 2018. This study period reflects the pilot phase of the NPD’s BWC deployment, with only one of the seven NPD precincts completing BWC deployment prior to the beginning of the study period (October 2017). The other six precincts began BWC deployment between May and November 2018, with deployment taking between one and four months to complete in each precinct. The 91 cases in our sample represent the use of force events recorded in their entirety during the study period.

The primary database for this study was created via SSO, a method of data collection developed by Reiss (1968, 1971) in which data collection is independent of the phenomena being observed. Prior to coding the video footage, we engaged in an in-depth review of the data during a five-day research team retreat at NPD headquarters, meeting about six hours each day. Through this data retreat, we determined the interactions and context key to the dimensions (Nassauer and Legewie, 2021) of police-civilian encounters. As a result of technological limitations preventing coders from viewing the footage remotely, a single coder oversaw data collection. Following coding, a series of tests for intra-coder reliability were conducted. The research team first coded ten full incidents together to determine how to best interpret and code the activities taking place on screen until meeting general saturation of potential uncertainties. The primary coder tasked with leading the SSO for the remainder of the project then spent a full day coding incidents on his own, with the principal investigators on hand to further discuss coding decisions as necessary. This team approach to coding was maintained throughout the project, with the primary coder and principal investigators consistently meeting (both remotely and in person) to discuss coding issues and to reach consensus. This approach is particularly helpful in maintaining coding consistency for cases that provide less clarity than the typical cases in the sample (Parry et al., 2021; Schafer et al., 2022).

Coding occurred at NPD headquarters across 8 months. The 91 use of force events were captured across 275 separate videos resulting from multiple responding officers wearing BWCs at recorded incidents. The multiple BWC videos enabled us to view police-civilian encounters from numerous vantage points, which facilitated observation and measurement of the variables of interest. It took approximately 300 hours to review and code all use of force events. A detailed coding instrument helped guide the SSO.3 Coding instruments take on increased importance in VDA research of BWC footage, given that prior research has found different viewers can interpret the same BWC footage differently, especially when they lack practical policing experience (Boivin et al., 2020). Strict adherence to a coding instrument, combined with the aforementioned regular team meetings, supported a reflexive process that minimizes the risk that any ontological or epistemological assumptions impact the data collection (Chillar et al., 2021). These processes help to maintain face and content validity of the analysis, which are key but easily overlooked considerations of research using video recordings as a data source (Parry et al., 2021). To test reliability, a total of 20 cases (21.9%) were randomly selected and recorded in the original wide version of the database 6 months after the original coding commenced. All kappa coefficients were >0.60, confirming reliability for this study (Landis and Koch, 1977).4

The unit of analysis is use of force events, which incorporates the entry, information exchange, and final frame decision phases conceptualized by Binder & Scharf (1980).5 Use of force events begin when an officer first engages an involved party at the scene and conclude when full civilian compliance is secured. We initially assumed encounter “end points” would be obvious, such as when an officer applied handcuffs on a civilian. We found that such events, however, did not always represent the end of the encounter, as civilians sometimes continued to physically resist arrest while handcuffed. We operationalized the end of each encounter as the time of resolution, defined as a natural break in the event upon which full civilian compliance was secured. The time of resolution took a several forms, including an arrest, officers placing the civilian within the patrol car, and officers leaving the scene (Chillar et al., 2021). Prior VDA research has similarly used natural breaks in event sequences to identify stages of civilian encounters (Sytsma and Piza, 2018).6

4.2 | Database Creation

We converted the structure of the original SSO dataset from ‘wide’ into ‘long’ format to enable longitudinal analysis. The present long format approach differs from traditional approaches in several ways. First, it allows for different numbers of intervals across cases to reflect the appropriate length of the encounter (a case-specific number of five-second intervals). Second, it allows for all recorded actions or behaviors within an incident to be reflected in separate columns based on the time, duration, and frequency of occurrence (when an action or behavior occurs, how long it lasted, and the number of times it occurred). Lastly, the present approach creates a single, aggregated final variable that cumulatively depicts when the action or behavior occurred during an encounter in the correct interval(s).7

4.3 | Independent Variable Conceptualization and Operationalization

In the present study we create new ‘time points’ for individual case observations by assessing the entirety of the interaction as a series of five-second intervals. All model covariates were categorized into theoretical constructs. Four different interval-based operationalizations measured how the independent variables influence the dependent variable across different time constructs during the police-civilian encounter.

4.3.1 | Theoretical Constructs

Four variables measure active civilian resistance, which prior research has consistently identified as a direct precipitator of force: civilian verbal antagonism, civilian physical antagonism, civilian flee attempt, and officer discovering a weapon on scene. Five variables measure authority maintenance8 activities of officers observed in prior research (Alpert and Dunham, 2004; Sykes and Clarke, 1975). Four authority maintenance variables measure activities of officers and subsequent civilian responses, including: calm-command issued, shout-command-issued, calm-command noncompliance, and shout-command noncompliance. The final authority maintenance variable, officer count, measures the presence of officers on scene, which increases each time additional officers arrive as back up in support of the primary officers. Five variables measure aspects of procedural justice. Three of the variables have been commonly used in prior research: officer provides the reason for responding to the scene, officer explains the reason for civilian detainment, and officer allows the civilian to speak to explain their side of the story. Officer verbal antagonism measures officer deviation from procedural justice, given the model’s emphasis on calm, respectful tones. BWC camera announcement is also included to reflect recent research that indicates announcing BWC presence significantly improves civilian perceptions of procedural justice and police legitimacy (Demir, 2021). Lastly, three variables measure bystander presence that may influence the OODA loop process of both officers and civilians: antagonistic bystanders count, neutral bystanders count, and uninvolved bystanders count.

4.3.2 | Instant variables

The first independent variable classification delineates the precise moment or time at which an action or behavior occurred. Termed instant measures, these variables have start and end times that occurred instantaneously and simultaneously.9 Ten variables were operationalized using the instant classification: (1) BWC announcement, (2) officer arrival, (3) uninvolved bystander arrival, (4) neutral bystander arrival, (5) antagonistic bystander arrival, (6) civilian physical antagonism, (7) calm command non-compliance, (8) shout command non-compliance, (9) weapon discovery, and (10) civilian flee attempt. Each dichotomous, instant variable was assigned a score of “1” at each time point it was recorded to have occurred.

The civilian non-compliance variables were conceptualized in juxtaposition to police issued calm and shout commands (the active predictor variables described later). Instances of civilian non-compliance were abrupt, resulting in immediate interruption or defiance. Relatedly, the discovery of a weapon constituted a single point in time at which an officer discovered a weapon. Instances in which civilians were observed to throw a punch or shove an officer (civilian physical antagonism) often took place immediately and were also immediately concluded. Similarly, civilian attempts to flee were often immediately stopped or met with some form of resistance that ended the “attempt.” The announcement of an officer’s BWC was also a brief, momentary action with clear language following NPD protocols (as opposed to longer speech-based instances recorded in the active class described below). The arrival variables also fit the instant classification and were recorded and used to examine the precise time point an additional officer or bystander appeared on scene and the immediate impact on the use of force.

4.3.3 | Active variables

The second independent variable classification is the active operationalization, which marks all of the consecutive interval(s) at which an action was actively occurring. The dataset delineates an ongoing action as occurring by dichotomously indicating a ‘1’ value for an action as it transpires and a ‘0’ value when it is not currently occurring. Seven variables, each representing a speech pattern by an involved party, met the active variable criteria: (1) officer providing the reason for response to the scene, (2) officer explaining detainment, (3) civilian allowed to speak for the purpose of expressing their views, (4) officer verbal antagonism, (5) calm command issued, (6) shout command issued, (7) civilian verbal antagonism. All active measure start times, end times, and total durations are specific to each individually recorded occurrence.

4.3.4 | Post-occurrence variables

The next classification of variables measure the post-occurrence impact of both independent variable types (instant and active). The post-occurrence variables for every instant and active variable were generated by assigning the next six intervals (30 seconds) after a recorded instance of an action or behavior a value of ‘1’ to explore their potentially delayed effect.10 Psychology research suggests that due to the recency effect (see Davellar, 2005), most information is only stored as a short-term memory for 30 seconds (Cowan, 2008). New information has also been found to cognitively displace old information (Koene and Hasselmo, 2007), thereby limiting lasting effects of events as they unfold.

4.3.5 | Lasting variables

The final independent variable operationalization accounts for both the potentially lasting impact of each of the instant and active variables, and the potential impact of frequency of occurrence. Lasting measures capture the number of times an action or behavior occurs, and the influence of its occurrence on the remaining duration of the event. The lasting operationalization is analogous to the ‘ringing of a bell’, which suggests that once completed, an action or behavior cannot be undone. Lasting variables were scored as ‘0’ until the interval at which the action started, and they were then filled down through the following intervals with a value of ‘1’ to indicate that the action or behavior already took place. If the action or behavior occurred again, the ‘1’ values were switched to ‘2’ and were filled down through remaining intervals. This process was repeated for every instance of an instant or active variable. Alternatively, the count-focused arrival variables reflected the recorded count of people on scene matching that criterion through all subsequent intervals until a change was observed. Table 1 displays a hypothetical depiction of each of the aforementioned independent variable operationalizations in the dataset.

Table 1. Independent Variable Operationalization

4.4 | Dependent Variable

The construction of the dependent variable included all use of force types and applications recorded during the SSO, including soft, hard, blunt, chemical, and threats of lethal force.11 The dependent variable is a binary measure indicating whether use of force occurred within a given time interval (1) or not (0). To account for the restructuring of the data and the aggregation into intervals, all force event instances were scored as occurring in one additional consecutive interval as a sensitivity consideration (the same procedure applied to the instant independent variables). Scoring force events as a two-interval event results in a maximum time allotment of ten seconds, mirroring the often-rapid nature of decisions to enact force and the following force application. There were 187 unique instances of force in the dataset. These instances occurred within 184 unique intervals as several threats of elevated force were coupled with an additional type of force within the same interval. Adding an additional sensitivity interval to each of the 184 force intervals resulted in 360 total intervals in which force was recorded as occurring.

4.5 | Contextual Variables

Use of force events were linked to disparate data sources provided by the NPD as well as the American Community Survey data from the United States Census Bureau. As discussed later in the analytical approach section, the current study used a fixed-effects panel regression model that controls for observed and unobserved characteristics of the use of force events. This approach precludes the need for static, incident-level control variables. The triangulated data, however, provide important contextual insights into the study sample, which motivates us to present their descriptive findings. Calls for service data provided the initial call type, the hour and day of week the call occurred, and the reported location of the event. Arrest data provided information on arrestee demographics, the arrest location, time of the arrest, and the crime(s) charged. NPD personnel records provide the age, sex, race, rank, and number of years employed for each officer. Each case was successfully geocoded and assigned the concentrated disadvantage index and general land use information of its of its surrounding block group.12

5 | ANALYTIC APPROACH

5.1 | Logistic regression

We use fixed-effects panel data logistic regression to examine the situational determinants of police use of force. Fixed-effects models automatically control for observed and unobserved heterogeneity of the analytical units (e.g. use of force events in the current study) by considering only within-unit variation in the dependent variable (Bruderl and Ludwig, 2015; Lindegaard et al., 2021). The regression analysis tests all of the independent (instant/active, post-occurrence, and lasting) predictors in a single model. This approach allows for the differing time-effects to control for one another while discerning the temporal dynamic exhibiting the strongest relationship with force. Each of the 91 use of force events was operationalized as a unique longitudinal panel using a series of five-second intervals spanning the duration of the event (totaling in aggregate 10,212 intervals). The longitudinal nature of the data was ordered by both panel identifiers (unique case ID’s) and time (sequential five-second intervals) to control for incident length in the parameters of the model.13

5.2 | Wald tests for coefficient comparison

Following the logistic regression, we incorporated seemingly unrelated estimation post-hoc tests to determine whether the independent variables maintained significance, similar coefficient and effect sizes, and overall influence across each specification of time (instant/active, post-occurrence, lasting). Global Wald tests indicate whether any significant differences exist in the magnitude of the coefficients across any of the variable operationalizations that demonstrate effects in the same direction (positive/negative). Local Wald pairwise comparisons were then conducted following all global Wald tests that achieved statistical significance, or when necessary to compare two variable types. This post-hoc analysis tests each combination of coefficients for a given independent variable to determine whether differences exist across the coefficients.

5.3 | Dominance analysis

Use of dominance analysis (Azen and Budescu, 2003; Budescu, 1993) helped to determine the influence of each unique, significant variable identified in the regression model. General dominance statistics are calculated through weighting techniques that determine the relative importance of each significant predictor in reducing prediction error in the dependent variable outcome. The results of the dominance analysis provide rankings of the significant variables tested in the post-regression subset from the most to least predictive. The approach involves testing every possible combination of identified predictors in a subset iterative regression model to identify the predictor(s) with the greatest predictive capacity relative to one another.

6. | RESULTS

6.1 | Descriptive statistics

Table 2 presents descriptive statistics that contextualize the nature of the police-civilian encounters observed in the SSO. Violent crime is the most frequently observed call classification (n=40; 46.96%), civilians are predominately men (n=73; 80.22%) and Black (n=71; 78.02%), with nearly three-quarters (n=68) not exhibiting evidence of being under the influence of drugs or alcohol. Officers are predominately men (n=85; 93.41%), Latino (n=57;62.64%) and have a non-supervisory position (i.e., rank below a sergeant) (n=81; 89.01%). Use of force events are almost evenly split between the weekday (n=47; 51.65%) and the weekend (n=44; 48.35%). Slightly over half (n=46) of the events occurred during the third patrol shift (4pm-midnight) with seventy-two (79.12%) events occurring in public settings.

Table 2. Descriptive Statistics: Contextual Variables (Force Events. N = 91)

Table 3 presents descriptive statistics for the dependent variable across the 10,212 five-second time intervals comprising the 91 use of force events. One hundred and eighty-seven force applications appear within the database across 184 unique time intervals. Soft empty-hand force is the most common type (n=109; 58%) followed by hard empty-hand force (n=63; 33.69%). As previously discussed, the 184 intervals involving force were extended into one additional subsequent interval as a sensitivity measure for the five-second aggregation. Accounting for these measures, use of force was coded in 360 of the 10,212 intervals (3.53%). Events last 9.33 minutes (112 intervals) on average with a median of 5.75 minutes (69 intervals).

Table 3. Descriptive Statistics. Dependent Variable (Total Intervals. N =10,212)

a The 187 force applications occurred in 184 unique intervals. There were a few instances where a threat of lethal force was coupled with a force application in the same interval.

b The 187 force events were extended into one additional subsequent interval as a sensitivity measure for the five-second aggregation and to allow for force to be applied in a ten second time allotment in the absence of recorded end points. There were several instances where a different application of force was recorded in the next consecutive interval, which is why the 184 single intervals are not precisely doubled (360 instead of 368).

Table 4 presents descriptive statistics of the independent variables across the 10,212 five-second time intervals. The instant/active and post­-occurrence measures refer to individual time intervals in which the variables were coded whereas the lasting measure notes the variable range. As an example, police officer calm and shout commands were observed in 900 (8.81%) and 386 (3.78%) intervals in the active operationalization, respectively. In the post-occurrence period, calm commands and shout commands were coded in 716 (7.01%) intervals and 348 (3.41%) intervals, respectively. The frequency of calm commands provided in a given event ranges from zero to five whereas shout commands exhibit a range of zero to three.

6.2 | Logistic regression and Wald coefficient comparison tests

Table 5 presents results of the regression model and Wald coefficient comparison tests simultaneously. Significant findings are observed for two of the four civilian resistance variables. Civilian physical antagonism is significantly associated with an increased likelihood of force in the instant and post-occurrence operationalizations. The instant variable indicates use of force is more than 4 times more likely to occur immediately after a civilian is physically antagonistic (OR=4.06; p.<0.001). The post-occurrence measure indicates use of force is more than twice as likely to occur within 30-seconds of civilian physical antagonism (OR=2.44; p.<0.01). Wald tests indicate the instant and post-occurrence variables significantly differ in magnitude, with the influence of physical antagonism being most salient the instant it happens. Civilian attempts to flee increase force likelihood by greater than three times in both the instant (OR=3.63; p.<0.01) and lasting (OR=3.72; p.<0.01) operationalizations, respectively. Wald tests confirm the lasting effect yields the most salient influence on use of force. No operationalizations achieved significance for the suspect verbal antagonism or weapon discovery variables.

Three of the four authority maintenance variables demonstrate significant effects on the use of force. Active officer calm commands and shout commands increase the likelihood of force by more than three times (OR= 3.29; p.<0.001) and more than four times (OR=4.14; p.<0.001), respectively. Officer calm commands nearly double the likelihood of force in the post-occurrence period (OR=1.85; p.<0.05). Active calm commands demonstrate a significantly greater influence on use of force than the post-occurrence variable. Shout commands are associated with a nearly 70% decreased likelihood of use of force in the lasting operationalization (OR=0.31; p.<0.01). The lasting effect of shout commands was significantly different and greater than the post-occurrence delineation, suggesting that continued use of shout commands had the most salient impact on reducing use of force. The likelihood of force increases more than two times immediately after suspect non-compliance to calm commands (OR=2.59; p.<0.01). Wald tests indicate active non-compliance to calm commands was significantly greater than the post-occurrence operationalizations. All three variable operationalizations are statistically significant for officer arrival/count. The instant (OR=2.98; p.<0.001) and post-occurrence (OR=2.12; p.<0.001) variables indicate force is nearly three times more likely to occur immediately and more than twice as likely to occur within 30-seconds of when additional officers arrive on scene, respectively. Wald tests indicate the instant and post occurrence variables are significantly different in magnitude, with the instant variable being the most salient indicator of the impact of officer arrival on use of force. The lasting variable suggests, however, that officer arrival results in a 20% reduction of force likelihood (OR=0.80; p.<0.01).14 No operationalizations achieved statistical significance for shout command noncompliance.

Four of the five procedural justice variables demonstrate significant effects. All three operationalizations achieved significance for officer verbal antagonism. Use of force is more than 10 times higher when an officer is actively verbally antagonistic (OR=10.44; p.<0.001) and nearly four times higher in the subsequent 30-second period (OR=3.73; p.<0.01). Wald tests indicate the active variable has a significantly greater effect on use of force. Conversely, the lasting variable is associated with a greater than 80% decreased likelihood of use of force (OR=0.18; p.<0.001). Two operationalizations were significant for the officer reason for response variable. Officers providing their reason for responding to the incident was bi-directional and was associated with a greater than 70% decreased likelihood of force in the post-occurrence period (OR=0.27; p.<0.05) and a nearly four-fold increase in the likelihood of force in the lasting effect (OR=3.94; p.<0.001). Despite significance in the model, the post-occurrence operationalization of officers providing a reason for response did not demonstrate a significantly greater magnitude of effect on the use of force compared with the active operationalization. Civilian speaking also has bi-directional effects on the use of force. Civilians speaking is associated with nearly a 50% decreased likelihood of force in the active period (OR=0.55; p.<0.05) and a 78% increased likelihood of force in the lasting period (OR=1.78, p.<0.01). The lasting effect of civilian speaking is significantly greater in its effect on use of force compared with the post occurrence coefficient in the Wald tests. Officer explaining detainment more than doubles force likelihood in the post-occurrence period (OR=2.29; p.<0.01). The post-occurrence operationalization had a significantly greater impact on the use of force than the lasting effect. No operationalizations achieved statistical significance for BWC announced.

In terms of bystander presence, neutral bystander arrival increases force likelihood by 88% within 30-seconds of additional bystander arrival on scene (OR=1.88; p.<0.05) and 77% in the lasting effect (OR=1.77; p.<0.001). Wald tests indicate the post-occurrence operationalization has a more salient impact on use of force compared to the lasting effect. No operationalizations achieved significance for antagonistic bystander arrival or uninvolved bystander arrival.

6.3 | Dominance analysis

Table 6 displays the results of the dominance analysis. Seventeen of the 51 independent variables included in the logistic regression model are significantly predictive of time intervals when use of force occurred. The top six predicators are authority maintenance variables, and they cumulatively account for greater than 65% of the logistic regression model’s predictive capacity, as observed through standardized dominance statistics (proportions): active calm commands (0.163), post-occurrence officer count (0.124), post-occurrence calm command (0.100), active shout command (0.096), instant officer count (0.092), and instant calm command noncompliance (0.078). The lasting effect of neutral bystander arrival, the post-occurrence officer verbal antagonism variable, and the lasting effect of officer reason for response add the least predictive capacity to the model, with each variable having standardized dominance statistics of 0.001 (less than 1%).

Five of the 51 independent variables are predictive of decreased use of force likelihood. The lasting operationalization of officer count accounts for approximately 61% of the model’s predictive capacity (0.612), by far the largest value observed in the dominance analysis. Lasting officer verbal antagonism (0.152) and lasting shout command issued (0.138) account for approximately 15% and 14% of the predictive capacity, respectively. Active civilian speaking had the lowest predictive capacity for decreasing use of force likelihood (0.026).

Table 6: Dominance Analysis - Significant Predictors Increasing the Use of Force

Note. I: instant, A: active, PO: post-occurrence, L: lasting

N=10,212

7 | DISCUSSION AND CONCLUSION

Results indicate that even though various independent variables significantly predict use of force, the temporal influence of variables greatly differs. Calm command noncompliance increases the likelihood of force in the instant it occurs, whereas officer explaining detainment increases the likelihood of force in the post-occurrence period. All other statistically significant variables impact force likelihood across multiple time periods, although Wald tests confirmed stronger effects of certain within variable operationalizations. For example, civilian physical antagonism increases force likelihood in both the instant and the post-occurrence time periods, with a stronger effect observed for the post-occurrence operationalization. Interestingly, divergent, bi-directional effects were observed for five variables. For example, officer verbal antagonism increases force while actively ongoing and decreases force through its frequency and reoccurrence over time, as reflected in the lasting variable. The officer arrival/count variable was significant in increasing force in both the instant and the post-occurrence periods, while significantly decreasing force in the lasting operationalization. Such findings further demonstrate the transactional nature of use of force. These findings suggest that officers arriving on scene may immediately be involved in controlling the scene, they may have been requested for back-up after escalation, and/or the need for force may have been delayed until back up arrived. The influence of different time operationalizations is often specific to each variable, highlighting the need for continued longitudinal evaluation of encounters.

The dominance analysis indicates authority maintenance was the most influential theoretical construct, as the top six predictors of force are authority maintenance measures. This finding supports prior theoretical perspectives arguing that police use of force largely results from officer attempts to maintain authority over civilians during face-to-face encounters (Alpert and Dunham, 2004; Sykes and Brent, 1983; Sykes and Clark, 1975). Nonetheless, a fair proportion of variables achieve significance within each of the four theoretical constructs, showing how police use of force is influenced by a range of situational factors. Several procedural justice measures exhibited significant relationships with force. Officers providing a reason for response to the scene and allowing civilians to speak were associated with decreased likelihood of force in the post-occurrence and active periods, respectively. Interestingly, though, both measures were associated with increased force likelihood in the lasting time period. Given force is used in all cases in the current study, this finding may reflect the fact that officers would repeatedly apply procedural justice methods throughout an encounter before resorting to force, which fits into the basic force continuum framework (Terrill, 2001).

Officer verbal antagonism significantly increases force in the instant operationalization. This finding suggests that officers engaging with civilians in a negative verbal capacity may immediately invoke them to use force or may elicit suspect actions that directly require force responses. Although force was used in all cases included in the current study, the fact that procedural justice mechanisms significantly impact use of force suggests procedural justice may be a worthwhile component of de-escalation strategies. Civilian resistance also significantly influences when force occurs, echoing findings of prior research (Terrill, 2003; Terrill & Mastrofski, 2002). For example, in the present study, civilian physical antagonism and civilian flee attempts both increase force likelihood instantly as well as in subsequent time periods.

This study’s results contribute nuance to Binder and Scharf’s (1980) transactional framework. The “information exchange phase” of Binder and Scharf’s framework involves officers and civilians interacting with one another and reacting to the verbal and nonverbal cues emitted. Our data suggest the nature of the shift from the information exchange phase to the “final frame” can vary widely. Understanding that information refers to overt or subtle behaviors, some forms of information being exchanged can immediately trigger (the shift to) the decision to use force—like civilian physical antagonism. Although the objective of the present research is not to test the Binder and Scharf framework, nor the OODA loop, we wish to introduce the idea that the OODA loop may be an important step of the information exchange phase. Civilian physical antagonism immediately triggers a shift; however some other forms of information exchange either can result in either an immediate shift to the final frame or may build to a tipping point before a shift is triggered—depending on various situational factors. During this build, actors may be repeatedly cycling through the OODA loop, each time coming to the conclusion not to use force until the last OODA loop that takes place at the point of shift to the final frame phase of the Binder and Scharf framework. Future research can conceptually and empirically interrogate these ideas further—particularly the role of the OODA loop in each phase of the Binder and Scharf framework. In addition, dispatch may relay important information (e.g., potential injuries sustained by victim) to officers before they arrive on scene. Although technical aspects of the BWCs used by NPD prevented us from including the anticipation period in this study, future research may seek to unpack how the entry phase of the OODA loop is differentially impacted by the type of information received by officers prior to arrival on scene.

Although we believe this study makes several important contributions to the literature, we acknowledge specific limitations readers should be aware of. As already discussed, technological limitations prevented us from coding the first 30-seconds of each video or including multiple coders in the SSO. Even though the nature of our data collection protocol and active management of the SSO helped maintain coding reliability, a BWC system equipped with video streaming capabilities (meaning videos can be viewed remotely, rather than exclusively on-site at the NPD) would have offered several methodological benefits (Chillar et al., 2021). BWCs themselves present certain inherent limitations for research, such as capturing incomplete information on the facial expressions and non-verbal cues emitted by police officers, and being obstructed when officers are in close proximity to civilians, such as when applying handcuffs or conducting searches (Schafer et al., 2022; Terrill and Zimmerman, 2021). Although having multiple videos per event provides different viewpoints of each police-civilian encounter, which helped navigate obstructions and allow for direct observation of officers on-scene, the current study nonetheless suffers from these shortcomings. This project was supported by a relatively modest grant, which limited the number of use of force events we could realistically view and code during the project period. Our sample, however, is comparable to prior VDAs of BWC footage (Friis et al., 2020; Willits and Makin, 2018) and larger than VDAs of CCTV footage (Mosselman et al., 2018; Piza and Sytsma, 2016). Furthermore, the conversion of 91 incidents into 10,212 time-intervals provided sufficient statistical power for the analysis. Nonetheless, we advocate that future work maximize sample size to increase generalizability.

While acknowledging these limitations, we believe this study makes a positive contribution to the rich literature on police use of force. The data structure used in this study advances prior work on the transactional nature of police-civilian encounters. By considering use of force as a series of time-bound transactions, we identify situational factors that influence opportunities for physical force. Our approach also presents a framework for analyzing police-civilian encounters in which violence was averted through officer actions (i.e. de-escalation), a topic that some scholars have begun to leverage BWC footage to explore (see e.g. Schafer et al., 2021; Terrill and Zimmerman, 2021). The study takes an important step forward by bridging past data limitations to better understand the situational factors driving police use of force decision-making (Sherman, 2018).

REFERENCES

Alpert, G. P., & Dunham, R. G. (1997). The force factor: Measuring police use of force relative to suspect resistance. Washington, D.C.

Alpert, G. P., & Dunham, R. G. (2004). Understanding police use of force: Officers, suspects, and reciprocity. Cambridge, UK: Cambridge University Press.

Azen, R., & Budescu, D. V. (2003). The dominance analysis approach for comparing predictors in multiple regression. Psychological Methods, 8(2), 129–148. https://doi.org/10.1037/1082-989X.8.2.129

Bayley, D. H., & Garofalo, J. (1989). The management of violence by police patrol officers. Criminology, 27(1), 1–26. https://doi.org/10.1111/j.1745-9125.1989.tb00861.x

Bazley, T. D., Lersch, K. M., & Mieczkowski, T. (2007). Officer force versus suspect resistance: A gendered analysis of patrol officers in an urban police department. Journal of Criminal Justice, 35(2), 183–192. https://doi.org/10.1016/j.jcrimjus.2007.01.005

Bennell, C., Alpert, G., Andersen, J. P., Arpaia, J., Huhta, J. M., Kahn, K. B., … White, M. D. (2021). Advancing police use of force research and practice: Urgent issues and prospects. Legal and Criminological Psychology, 26(2), 121–144. https://doi.org/10.1111/lcrp.12191

Binder, A., & Scharf, P. (1980). The violent police-citizen encounter. The Annals of the American Academy of Political and Social Science, 452(1), 111–121.

Boivin, R., Faubert, C., Gendron, A., & Poulin, B. (2020). Explaining the body-worn camera perspective bias. Journal of Qualitative Criminal Justice & Criminology, 9(1), 1–28. https://doi.org/10.21428/88de04a1.bc4fdeda

Bolger, P. C. (2015). Just following orders: A meta-analysis of the correlates of American police officer use of force decisions. American Journal of Criminal Justice, 40(3), 466–492. https://doi.org/10.1007/s12103-014-9278-y

Brandl, S.G., & Stroshine, M.S. (2017). Oleoresin capsicum spray and TASERs: a comparison of factors predicting use and effectiveness. Criminal Justice Policy Review, 28(3), 279-306. https://doi.org/10.1177/0887403415578732

Bruderl J. & Ludwig, V. (2015). Fixed-effects panel regression. In H. Best & C. Wolf (eds.) The SAGE Handbook of Regression Analysis and Causal Inference. Sage Publications Ltd. London. 

Budescu, D. V. (1993). Dominance analysis: A new approach to the problem of relative importance of predictors in multiple regression. Psychological Bulletin, 114(3), 542–551.

Chestek, K. D. (2020). Thinking in circles: Using OODA to sharpen legal analysis. Wyoming Law Review, 20(2), 320–345.

Cheney, G. (1983). On the various and changing meanings of organizational membership: A field study of organizational identification. Communications Monographs, 50(4), 342-362.

Chillar, V., Piza, E., & Sytsma, V. (2021). Conducting a systematic social observation of body-camera footage: Methodological and practical insights. Journal of Qualitative Criminal Justice & Criminology, 10(4), 1–30.

Cobbina, J. (2019). Hands up don’t shoot: Why the protests in Ferguson and Baltimore matter, and how they changed America. New York, NY: New York University Press.

Cowan, N. (2008). What are the differences between long-term, short-term, and working memory? Progress in Brain Research, 169, 323–338.

Davelaar, E. J., Goshen-Gottstein, Y., Ashkenazi, A., Haarmann, H. J., & Usher, M. (2005). The demise of short-term memory revisited: Empirical and computational investigations of recency effects. Psychological Review, 112(1), 3-42.

Demir, M. (2021). Effect of awareness and notification of body-worn cameras on procedural justice, police legitimacy, cooperation, and compliance: Findings from a randomized controlled trial. Journal of Experimental Criminology, (0123456789). https://doi.org/10.1007/s11292-021-09487-6

Dutton, J. E., Dukerich, J. M., & Harquail, C. V. (1994). Organizational images and member identification. Administrative Science Quarterly, 239-263.

Engel, R. S., McManus, H. D., & Herold, T. D. (2020). Does de-escalation training work?: A systematic review and call for evidence in police use-of-force reform. Criminology and Public Policy, 721–759. https://doi.org/10.1111/1745-9133.12467

Friedrich, R.J. (1980). Police use of force: Individuals, situations, and organizations. Annals of the American Academy of Political and Social Science, 452, 82-97.

Friis, C. B., Liebst, L. S., Philpot, R., & Lindegaard, M. R. (2020). Ticket inspectors in action: Body-worn camera analysis of aggressive and nonaggressive passenger encounters. Psychology of Violence, 10(5), 483–492.

Garner, J.H., Maxwell, C.D., & Heraux, C.G. (2002). Characteristics associated with the prevalence and severity of force used by the police. Justice Quarterly, 19(4), 705–746

Garrett, B., & Stoughton, S. (2017). A tactical fourth amendment. Virginia Law Review, 103(2), 211–307.

Goffman, E. (1959). Presentation of self in everyday life. New York, NY: Anchor.

Goffman, E. (1961). Encounters: Two studies in the sociology of interaction. New York, NY: Macmillan.

Goh, L. S. (2021). Did de-escalation successfully reduce serious use of force in Camden County, New Jersey? A synthetic control analysis of force outcomes. Criminology & Public Policy, 20(2), 207–241.

Koene, R. A., & Hasselmo, M. E. (2007). First-in-first-out item replacement in a model of short-term memory based on persistent spiking. Cerebral Cortex, 17(8), 1766–1781.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159–174.

Liang, J., Chen, X., Li, T., & Wang, Y. (2021). Beyond justice perceptions: The role of interpersonal justice trajectories and social class in perceived legitimacy of authority figures. Frontiers in Psychology, 12(February), 1–10. https://doi.org/10.3389/fpsyg.2021.595731

Liang, J., & Ma, H. (2021). Interpersonal injustice and perceived legitimacy of authority: The role of institutional trust and informational justice. Journal of Community & Applied Social Psychology, 31(2), 184–197.

Lindegaard, M. R., Liebst, L. S., Philpot, R., Levine, M., & Bernasco, W. (2021). Does danger level affect bystander intervention in real-life conflicts? Evidence from CCTV footage. Social Psychological and Personality Science. https://doi.org/10.1177/19485506211042683

Mangels, L., Suss, J., & Lande, B. (2020). Police expertise and use of force: Using a mixed-methods approach to model expert and novice use-of-force decision-making. Journal of Police and Criminal Psychology, 35(3), 294–303. https://doi.org/10.1007/s11896-020-09364-4

March, J. G., & Simon, H. A. (1993). Organizations. John Wiley & sons.

Mazerolle, L., Bennett, S., Davis, J., Sargeant, E., & Manning, M. (2013). Procedural justice and police legitimacy: A systematic review of the research evidence. Journal of Experimental Criminology, 9(3), 245–274. https://doi.org/10.1007/s11292-013-9175-2

McCluskey, J. D., Uchida, C. D., Solomon, S. E., Wooditch, A., Connor, C., & Revier, L. (2019). Assessing the effects of body-worn cameras on procedural justice in the Los Angeles Police Department. Criminology, 57(2), 208–236. https://doi.org/10.1111/1745-9125.12201

McLean, K., Alikhan, A., & Alpert, G. P. (2022). Re-examining the Use of Force Continuum: Why Resistance is Not the Only Driver of Use of Force Decisions. Police Quarterly, 0(0), 1–26. https://doi.org/10.1177/10986111211066353

Mears, D. P., Craig, M. O., Stewart, E. A., & Warren, P. Y. (2017). Thinking fast, not slow: How cognitive biases may contribute to racial disparities in the use of force in police-citizen encounters. Journal of Criminal Justice, 53(September), 12–24. https://doi.org/10.1016/j.jcrimjus.2017.09.001

Merton, R. K. (1940). Bureaucratic structure and personality. Social Forces, 18, 560-568.

Mosselman, F., Weenink, D., & Lindegaard, M. R. (2018). Weapons, body postures, and the quest for dominance in robberies: A qualitative analysis of video footage. Journal of Research in Crime and Delinquency, 55(1), 3–26. https://doi.org/10.1177/0022427817706525

Murphy, K., Hinds, L., & Fleming, J. (2008). Encouraging public cooperation and support for police. Policing and Society, 18(2), 136–155.

Nassauer, A., & Legewie, N. M. (2019). Analyzing 21st century video data on situational dynamics-issues and challenges in video data analysis. Social Sciences, 8(3), 1–21. https://doi.org/10.3390/socsci8030100

Nassauer, A., & Legewie, N. M. (2021). Video data analysis: A methodological frame for a novel research trend. Sociological Methods and Research, 50(1), 135–174. https://doi.org/10.1177/0049124118769093

Nouri, S. (2021). Police use of force at street segments: Do street-level characteristics matter?. Journal of Criminal Justice, 77. https://doi.org/10.1016/j.jcrimjus.2021.101862

Nowacki, J. S. (2015). Organizational-level police discretion: An application for police use of lethal force. Crime & Delinquency, 61(5), 643-668.

Paoline, E.A. (2003). Taking stock: Toward a richer understanding of police culture. Journal of Criminal Justice, 31, 199-214. doi:10.1016/S0047-2352(03)00002-3

Parry, M. M., Dario, L. M., & Moule, R. K. (2021). Is seeing believing?: On the content validity of video recordings for use in criminological research. Deviant Behavior, 42(7), 807–820. https://doi.org/10.1080/01639625.2019.1699259

Patterson, Q., & White, M. D. (2021). Is there a civilizing effect on citizens? Testing the pre-conditions for body worn camera-induced behavior change. Police Quarterly. https://doi.org/10.1177/1098611121996716

Piza, E. L., & Sytsma, V. A. (2022). The impact of suspect resistance, informational justice, and interpersonal justice on time until police use of physical force: A survival analysis. Crime & Delinquency. DOI: 10.1177/00111287221106947.

Piza, E. L., & Sytsma, V. A. (2016). Exploring the defensive actions of drug sellers in open-air markets: A systematic social observation. Journal of Research in Crime and Delinquency, 53(1), 36–65. https://doi.org/10.1177/0022427815592451

Pratt, M. G. (1998). To be or Not to Be? Central questions in organizational identification. In Whetten, D. and Godfrey, P. (Eds.) Identity in Organizations, 171-207. New York: Sage.

Prenzler, T., Porter, L., & Alpert, G.P. (2013). Reducing police use of force: Case studies and prospects. Aggression and Violent Behavior, 18, 343-356. http://dx.doi.org/10.1016/j.avb.2012.12.004

Quinton, P., Dymond, A., Boyd, K., & Teers, R. (2020). Police use of force: Tactics, assaults and safety: Exploratory analysis of police recorded data 2017/18. London: College of Policing Limited. Retrieved from https://whatworks.college.police.uk/Research/Documents/Use_of_force_report.pdf

Reiss, A. J. (1968). Stuff and nonsense about social surveys and observation. In H. Becker, B. Geer, D. Riesman, & R. Weiss (Eds.), Institutions and the Person: Papers Presented to C. Everett Hughes. (pp. 351–367). Chicago, IL: Aldine.

Reiss, A. J. (1971). Systematic observation of natural social phenomena. Sociological Methodology, 3(3–33).

Reiss, A.J., (1972). Police Brutality. In Radzinowicz, L. and Wolfgang, M. (Eds.) Crime and Justice, Vol. 2: The Criminal in the Arms, 293-308.

Reiss, A. J. (1980). Controlling police use of deadly force. Annals of the American Academy of Political and Social Sciences, 452(1), 122–134.

Sampson, R. J., & Raudenbush, S. W. (1999). Systematic social observation of public spaces: A new look at disorder in urban neighborhoods. American Journal of Sociology, 105(3), 603–651.

Schafer, J., Hibdon, J., & Kyle, M. (2022). Studying rare events in policing: The allure and limitations of using body-worn camera video using body-worn camera video. Journal of Crime and Justice, 00(00), 1–16. https://doi.org/10.1080/0735648X.2022.2062036

Schnell, C., Grossman, L., & Braga, A. A. (2019). The routine activities of violent crime places: A retrospective case-control study of crime opportunities on street segments. Journal of Criminal Justice, 60, 140–153. https://doi.org/10.1016/j.jcrimjus.2018.10.002

Sherman, L. W. (2020). Evidence-based policing and fatal police shootings: Promise, problems, and prospects. The Annals of the American Academy of Political and Social Science, 687(1), 8–26. https://doi.org/10.1177/0002716220902073

Sherman, L. (2018). Reducing fatal police shootings as system crashes: Research, theory, and practice. Annual Review of Criminology, 1, 421-449.

Sierra-Arévalo, M. (2021). American policing and the danger imperative. Law & Society Review, 55(1), 70-103. https://doi.org/10.1111/lasr.12526

Sierra‐Arévalo, M. (2019). The commemoration of death, organizational memory, and police culture. Criminology, 57(4), 632-658.

Skolnick, J. H. (1966). Justice without trial: Law enforcement in democratic society. Quid pro books.

St. Jean, P. (2007). Pockets of crime: Broken windows, collective efficacy, and the criminal point of view. Chicago, IL: The University of Chicago Press.

Stoughton, S. W. (2014). Policing facts. Tulane Law Review, 88(5), 847-898.

Stoughton, S. W., Noble, J. J., & Alpert, G. P. (2020). Evaluating police uses of force. New York, NY: New York University Press.

Sykes, R., & Brent, E. (1983). Policing: A social behaviorist perspective. New Brunswick, NJ: Rutgers University Press.

Sykes, R., & Clark, J. (1975). A theory of deference exchange in police citizen encounters. American Journal of Sociology, 81(3), 587–600.

Sykes, R. E., & Brent, E. E. (1980). The regulation of interaction by police: A systems view of taking charge. Criminology, 18(2), 182–197.

Sytsma, V. A., Chillar, V. F., & Piza, E. L. (2021a). Scripting police escalation of use of force through conjunctive analysis of body-worn camera footage: A systematic social observational pilot study. Journal of Criminal Justice, 74(101776). https://doi.org/10.1016/j.jcrimjus.2020.101776

Sytsma, V. A., Piza, E. L., Chillar, V. F., & Grossman, L. S. (2021b). Measuring procedural justice policy adherence during use of force events: The body-worn camera as a performance monitoring tool. Criminal Justice Policy Review. https://doi.org/10.1177/08874034211021894

Sytsma, V.A., Laming, E., & Pohl, E. (2022). Situational and ecological predictors of conducted energy weapon application severity. Canadian Journal of Criminology and Criminal Justice, 6(1), 99-126. https://doi.org/10.3138/cjccj.2021-0019

Sytsma, V. A., & Piza, E. L. (2018). Script analysis of open-air drug selling: A systematic social observation of CCTV footage. Journal of Research in Crime and Delinquency, 55(1), 78–102. https://doi.org/10.1177/0022427817709502

Terrill, W. (2001). Police coercion: Application of the force continuum. New York.

Terrill, W. (2003). Police use Of force and suspect resistance: The micro process of the police-suspect encounter. Police Quarterly, 6(1), 51–83. https://doi.org/10.1177/1098611102250584

Terrill, W. (2005). Police use of force: A transactional approach. Justice Quarterly, 22(1), 107–138.

Terrill, W., & Mastrofski, S. D. (2002). Situational and officer-based determinants of police coercion. Justice Quarterly, 19(2), 215–248. https://doi.org/10.1080/07418820200095221

Terrill, W., & Paoline III, E. A. (2017). Police use of less lethal force: Does administrative policy matter?. Justice Quarterly, 34(2), 193-216.

Terrill, W., & Zimmerman, L. (2021). Police use of force escalation and de-escalation: The use of systematic social observation with video footage. Police Quarterly, 0(0), 1–23. https://doi.org/10.1177/10986111211049145

Tillyer, R. (2022). Unpacking sequential actions within use of force incidents. Police Quarterly, 0(0), 109861112110495. https://doi.org/10.1177/10986111211049549

Todak, N., & White, M. D. (2019). Expert officer perceptions of de-escalation in policing. Policing: An International Journal, 42(5), 832–846. https://doi.org/10.1108/PIJPSM-12-2018-0185

Tyler, T. R., & Fagan, J. (2008). Legitimacy and cooperation: Why do people help the police fight crime in their communities? Ohio State Journal of Criminal Law, 6(1), 231–275.

Tyler, T. R., & Jackson, J. (2014). Popular legitimacy and the exercise of legal authority: Motivating compliance, cooperation and engagement. Psychology, Public Policy, and Law, 20(1), 78–95.

Walters, G. D., & Bolger, P. C. (2019). Procedural justice perceptions, legitimacy beliefs, and compliance with the law: a meta-analysis. Journal of Experimental Criminology, 15(3), 341–372. https://doi.org/10.1007/s11292-018-9338-2

White, M.D., & Ready, J. (2010). The impact of the Taser on suspect resistance: Identifying predictors of effectiveness. Crime & Delinquency, 56(1), 70-102. https://doi.org/10.1177/0011128707308099

Willits, D. W., & Makin, D. A. (2018). Show me what happened: Analyzing use of force through analysis of body-worn camera footage. Journal of Research in Crime and Delinquency, 55(1), 51–77. https://doi.org/10.1177/0022427817701257

Wolf, R., Mesloh, C., Henych, M., & Thompson, L. F. (2009). Police use of force and the cumulative force factor. Policing: An International Journal of Police Strategies & Management, 32(4), 739–757. https://doi.org/10.1108/13639510911000795

Wood, G., Tyler, T. R., & Papachristos, A. V. (2020). Procedural justice training reduces police use of force and complaints against officer. Proceedings of the National Academy of Sciences, 117(18), 9815–9821.

Court Cases Cited

Graham v. Connor, 490 U.S. 386 (1989).

Kisela v. Hughes, 138 S. Ct. 1148 (2018).