Proactive monitoring and operator discretion: A systematic social observation of CCTV control room operations

Abstract:

Background: Research on police discretion indicates a host of legal and extra-legal factors can impact officer decision making. The emergence of video surveillance technologies has made certain police interactions with the public more remote in nature. Decisions to enforce the law, consequently, now commonly begin outside of the context of face-to-face police/citizen interactions. This study explores police decision-making in the context of proactive video surveillance operations.

Methods: Data were generated from a systematic social observation of CCTV operator activity during the Newark CCTV Directed Patrol Experiment. Regression models tested how covariates affect the duration of CCTV operator targeted surveillances and CCTV operator decisions to report incidents providing reasonable suspicion and/or probable cause for a police response.

Results: A range of model covariates significantly influenced CCTV operator decision making, inclusive of surveillance targets identified as prior suspects, the CCTV site being within a commercial area, visible obstructions impeding camera view, CCTV operator rank, and CCTV operator gender.

Conclusions: Findings suggest that organizational culture, CCTV operator characteristics, and land usage of target areas may foster differential surveillance behavior across CCTV operators. As remote strategies for policing continue to expand internationally, the identification of factors that impact discretionary practices is critical.

Keywords:

Systematic social observation; police discretion; police decision-making; closed-circuit television (CCTV)

Citation:

Piza, E. and Moton, L. (2023). Proactive monitoring and operator discretion: A systematic social observation of CCTV control room operations. Journal of Criminal Justice. https://doi.org/10.1016/j.jcrimjus.2023.102071

1. Introduction

Policing’s reliance on technology has steadily increased throughout history (Koper & Lum, 2019). Technological advancements have recently allowed seamless integration of a range of surveillance technologies (Skogan, 2019), making video surveillance a core component of daily police operations around the world (Thomas et al., 2022). While the scientific evidence on technology is generally less developed than patrol- and investigation-based strategies (Lum & Koper, 2017), a robust literature has emerged for closed-circuit television (CCTV), with enough evaluation studies conducted to allow for three separate systematic

While the increase in evaluation research has provided insight into crime control outcomes associated with CCTV, many procedural and contextual aspects of video surveillance strategies remain underexplored (Lum & Koper, 2017). Of particular importance is the lack of understanding of the human factors that drive surveillance interventions, and how decision-making processes influence the manner in which video surveillance translates to enforcement actions in the field (Donald, 2010; Salvemini et al., 2015). Outside of some noteworthy exceptions (Heebels & van Aalst, 2020; Lomell, 2004; Loveday & Gill, 2004; Norris & Armstrong, 1999; Norris & McCahill, 2006), research has not analyzed how CCTV operators select which persons to observe or the factors that lead operators to report observed behavior to law enforcement. Furthermore, research focusing on such CCTV operator actions are mostly descriptive in nature, lacking the type of inferential analysis needed to identify significant and practically meaningful correlates of surveillance behavior.

The current study seeks to contribute to the literature through a systematic social observation (SSO) of CCTV operator activity during the CCTV Directed Patrol Experiment in Newark, NJ (Piza et al., 2015). Data from this SSO was previously used in a brief process evaluation of the CCTV Directed Patrol Experiment to report the number of incidents of concern reported by CCTV operators, as well as the proportion of reported events leading to a police action. We substantially build upon the process evaluation by coding field notes created by researchers during the SSO and creating a database that allows for a statistical analysis of each targeted surveillance—an observation of an individual or group lasting one minute or more (Norris & Armstrong, 1999; Norris & McCahill, 2006)—conducted during the CCTV Directed Patrol Experiment. Regression models test the effect of a range of factors on (1) the duration of targeted surveillances, (2) whether an incident of concern was observed by the CCTV operator, and (3) whether the CCTV operator reported the incident of concern to police. Study results have theoretical and practical implications for CCTV strategies and police discretion.

2. Review of Relevant Literature

The discretionary use of police authority is an inevitable (and central) component of policing. Due to the highly-individualized nature of police discretion, it may be problematic to assume that this is always an objective process (Charman & Williams, 2022). Factors such as judgement, interpretation, and previous experience illustrate the subjective nature of police discretion (Gelsthorpe & Padfield, 2003; Klockars, 1985). These factors particularly come into play when considering policing stops and arrest disparities across demographics (Myhill & Johnson, 2016).

Various internal and external characteristics are argued to influence discretionary police practices, such as neighborhood and socioeconomic makeup, education level and training of the officer, officer’s race and gender, law enforcement policies, police agency size, characteristics of the police organization, political structure of the community, situational influences of the perceived suspicious individual, police culture, and job satisfaction (Skogan & Frydl, 2004; Worden & McLean, 2014). Further, factors like environment and cultural context are important to consider when examining how discretion is used (Gelsthorpe & Padfield, 2003). According to Charman and Williams (2021), there are two factors within police culture that may impact police discretion. First, officers frequently depend on their personal set of internalized guidelines to inform their actions (Charman, 2017; O’Neill, 2016). These guidelines are developed through the norms associated with their social identities and knowledge gained from previous experiences in various contexts (Chan, 1997). The second element questions how the guidelines that have been internalized by police officers compare against the universal understanding of policing (i.e., officers positioned as enforcers of the law) (Manning, 1979). A police culture of cynicism, suspicion, and pessimism further complicate this juxtaposition (Reiner, 2010).

Use of personal discretion is not restricted to police officers, but is employed by other actors during the reporting and response to calls for service (CFS). The pathway from a citizen reporting an incident to the deployment of police rests on a number of judgements. First, the individual who has opted to call 911 must determine that what they have witnessed is deserving of police intervention. The call-taker must then rely on the description from the caller to determine if what they are describing is worthy of police response. The call taker also assigns the call a risk level. In the event a police response is deemed necessary, a dispatcher radios a police unit with background information on the incident and instructions to respond to the scene.1 Depending on the agency’s call volume, this process can be highly interpretive and discretionary on the part of the dispatcher (Simpson, 2021; Simpson & Orosco, 2021). Such decisions made by call takers and dispatchers can influence enforcement actions in the field by priming officer perceptions of the potential danger at the scene (Gillooly, 2020, 2022; Taylor, 2020). This body of research disputes the classification of call-takers and dispatchers as solely ‘agents of information transfer’, and recognizes the key influence these agents exert on the delivery of police services (Gillooly, 2022). Indeed, Simpson (2021) argues that their role as intermediaries between crime reporting and police response positions call takers and dispatchers—not police officers— as the true gatekeepers of the criminal justice system.

The increased role of technology in policing creates a situation where remotely positioned agents increasingly influence street-level activities of police officers (Tyler, 2016). This phenomenon is perhaps most evident in video surveillance policing strategies. Police personnel assigned to CCTV units are tasked with monitoring cameras strategically placed in geographic locations to alert police officers of suspicious or criminal behavior (Norris & Armstrong, 1999). CCTV operators must determine whether observed behavior provides reasonable suspicion or probable cause necessary for police deployment2, a discretionary process that parallels the decision-making of dispatchers given the remote proximity from the crime scene.

While 911 CFS represent a larger proportion of incidents leading to police operations than CCTV detections, they do make up a reasonable proportion of overall incidents resulting in police enforcement. For example, Piza and colleagues (2014) collected CFS and CCTV activity data within CCTV target areas in Newark, NJ over a three-year period (N=8,115). CCTV detections accounted for 1,385 incidents or 17% of the overall sample. It is important to understand the discretionary processes of the individuals charged with observing, assigning a priority level, and deploying police to almost 20% of incidents that require police intervention.

CCTV operators often employ targeted surveillances on incidents of interest, which is an example of proactive policing. Lum et al. (2020) examined proactive policing (police-initiated events) and reactive policing (911 or citizen-initiated events) by analyzing 1,965,867 computer assisted dispatch (CAD) records from four police agencies. Analysis of these data were supplemented through a systematic social observation of 84 officers for 180 hours across two of the agencies. They found that across the four agencies between 35% and 42% of CAD events were police-initiated. Given this figure is comprised of all police-initiated events, CCTV operators conducting surveillance that results in the deployment of police units are included in that figure. Therefore, CCTV surveillance and resulting discretionary actions taken by operators may make up a relatively significant portion of operations resulting in police enforcement.

CCTV operators use certain indicators to determine reasonable suspicion that is based on non-normative or unusual behavior (Piza et al., 2017). A number of other characteristics such as, time, location, age, race, gender, and clothing impact CCTV operator decisions on which cameras to direct their attention (Norris, 2003). Similar factors are in line with what street-patrol officers use to determine probable cause or reasonable suspicion. Alpert and colleagues (2005), for example, found that there were several characteristics associated with the likelihood an officer would make a stop: (1) appearance of an individual or vehicle, like distinctive clothing, perceived class status, or vehicle condition, (2) behavior, like overt inappropriate or bizarre actions, (3) time and place, and (4) information relayed by a dispatcher to the responding officer. No matter the location—in the field or remotely— policing practices rely on a great deal of discretion in order to be facilitated successfully (Alpert et al, 2004).

Norris & Armstrong (1999) used the term ‘targeted surveillance’ in reference to the act of a CCTV operator observing an individual or group of individuals for a period of one minute or longer. CCTV operators typically engage in targeted surveillances in search of suspicious behavior that may be indicative of (in-progress or forthcoming) criminal activity (Norris & Armstrong, 1999). An important consideration pertains to how operators select targets for surveillance and respond to perceived instances of crime and disorder.

Similar to street-level police-citizen interactions, age, race, and gender have been noted as significant factors for CCTV operators when employing targeted surveillance (Norris & Armstrong, 1999; Loveday & Gill, 2004, Arts et al., 2013). Norris & Armstrong (1999) found young teenaged males who exhibited minority racial and subcultural characteristics accounted for 39% of individuals targeted. CCTV operators deployed police to the scene in 23% of cases involving racial minorities, and only 18% of racial minority surveillance targets were subject to arrest. Norris & Armstrong (1999) identified these types of motivations for targeted surveillance as discriminatory practices that reinforce stereotypes. Similarly, Lomell (2004) found operators routinely targeted individuals who were poorly dressed, looked to have poor hygiene, or exhibited an appearance of a substance abuser. Other studies found that erratic behavior such as walking quicker than normal, consistent looking over one’s shoulder, avoiding cameras that are in view, abnormal body language, running down busy streets, and hyper awareness of other people alerts operators to employ targeted surveillance (Arts et al., 2013; Loveday & Gill, 2004; Grant & Williams, 2011).

Research has also found discretionary surveillance activity may be influenced by characteristics of individual CCTV operators. Pertini and colleagues (2014) found that operators with more job experience are more efficient in predicting harmful behavior. However, while experience is important to identify criminal offenses, common conditions such as low hourly wages and subsequent job dissatisfaction can contribute to high attrition rates within control rooms (Gill et al., 2005). Consistent turnover rates result in novice control room operators who have difficulty accurately perceiving intent to harm. This could result in deploying police when they are not needed, or conversely, not deploying law enforcement when it is a necessary action.

Less is known about the direct or indirect impacts other aspects of operator characteristics like gender, race, age, etc. have on discretion. For example, Norris and Armstrong (1999) briefly discuss the disparities of police deployment during domestic violence altercations involving men and women versus altercations with only men. There may be interesting and important impacts that operator gender, for example, may have on these types of incidents. Women operators may deploy police or respond to an incident of perceived gender-based violence differently than their male colleagues. These dynamics require a deeper examination.

Loveday and Gill (2004) found that operator suspicion and subsequent targeted surveillance did not commonly lead to the witnessing of theft in a chain of grocery stores. Further, operators who conducted the least number of targeted surveillances were actually the most likely to deploy management/staff or police to the area. The authors attribute this finding to the potential of these operators being more experienced and familiar with the local faces to be able to target known offenders (Loveday & Gill, 2004). More recently, Heebels & van Aalst (2020) argued that not only do CCTV operators rely on established rules, protocols, and their own judgement to determine suspicion, they also make decisions based on other’s opinions and actions in the control room. One operator expressed that they like to use the cameras as if they are walking on the sidewalk themselves, another operator opted to use a ‘bird’s eye view,’ some operators gave each camera equal attention, others utilized cameras in known crime ‘hot spots,’ and some operators simply zoomed in on groups of people that possessed certain demographics, like a group of young boys (Heebels & van Aalst, 2020). This indicates that there were a number of strategies when searching for criminal activity that appear to be dictated by the individual operator as opposed to a uniform protocol.

3. Scope of the Current Study

The current study analyzes discretionary CCTV operator actions through observational data collected during the CCTV Directed Patrol Experiment in Newark, NJ (Piza et al., 2015). The CCTV Directed Patrol Experiment was the culmination of an applied research partnership between the Newark Police Department and a team of academic researchers from across the New York City metro area. The City of Newark began installing a public network of CCTV video cameras beginning in 2007. A newly created Video Surveillance Unit (VSU) conducted CCTV surveillance activities, with CCTV camera feeds monitored from a central control room. During all tours of duty, two CCTV operators (a mix of civilian personnel and sworn police officers) under the supervision of a police Sergeant monitored camera feeds and reported any observed infractions via the department’s computer-aided dispatch system.

Throughout the CCTV system rollout, the aforementioned research team generated findings on the surveillance program’s crime prevention effects. Caplan and colleagues (2011) evaluated the effect of the first 73 cameras installed, finding that motor vehicle theft was the only crime type to experience a decline. Piza's (2018) evaluation of the fully deployed CCTV system similarly found potential benefits were exclusively restricted to motor vehicle theft, though the reduction only approached statistical significance (i.e., p.<0.10). Newark police officials considered the findings disappointing, given their primary goal of violence reduction.

Through interviews with surveillance operators, content analysis of computer-aided dispatch call logs, and statistical analysis of CCTV operator behavior, researchers identified two broad ‘surveillance barriers’ that likely mitigated the effect of the surveillance program. CCTV operators predominately stated they only report a minority of suspected crime events observed on camera, specifically owing to typically long-time frames between incident reporting and officer dispatch (Piza et al., 2017). Quantitative analyses found that a rapidly expanding CCTV system resulted in decreased surveillance activity. Each camera installation phase was associated with up to a 47% reduction in weekly proactive surveillance activity, with the system growing too large for CCTV operators to effectively monitor (Piza et al, 2014).

The CCTV Directed Patrol Experiment was conducted for the purpose of testing how the removal of ‘surveillance barriers’ impacted the crime control effect of CCTV. The experimental strategy was tested through a randomized controlled trial (RCT) conducted from 7/20/2011 – 10/1/2011. During all experimental tours of duty an additional CCTV operator was stationed within the control room, strictly monitoring cameras in the RCT treatment areas. Two unmarked patrol cars were dedicated to the experiment for the purpose of responding to crime incidents detected on treatment cameras, with incidents relayed by the experimental operator directly to the experimental patrol supervisor. The experiment ran Wednesday through Saturday from 8pm to 12am to coincide the days and times the targeted crime categories were at their peak occurrence during the pre-intervention period. The RCT found the experimental strategy generated sizable reductions in violent crime and social disorder (Piza et al., 2015) and was largely cost-effective for the Newark Police Department (Piza et al., 2016).  

Basic operator activity from the CCTV Directed Patrol Experiment was previously reported by Piza and colleagues (2015, p. 56-57) in a brief process evaluation. CCTV operators were documented to have reported a total of 72 incidents of concern to the experimental patrol officers, with 55 (76.4%) of these incidents classified as narcotics related. Sixty-four of the reported incidents resulted in officers conducting an enforcement action. This represented a drastic increase in CCTV operator activity, as the entire 146-camera CCTV system generated weekly averages of between 0.36 and 2.18 over the 52 weeks preceding the experiment. The current study builds upon the directed patrol process evaluation through an analysis of CCTV operator surveillance and reporting behaviors, as documented by researchers stationed in the Newark Police Department control room during the RCT.

4. Methodology

4.1 Unit of Analysis

SSO was the primary data collection method for this study. SSO involves the observation of social phenomena in a systematic, replicable manner, involving a means of observation that is independent of the phenomena being observed (Reiss, 1968, 1971). While police observational research can take many different forms, SSOs take a more structured approach that combines aspects of ethnography and survey research, which facilitates the enumeration of sizable data and potential study replication (Brunson & Miller, 2023). SSO has a rich history in policing research, generating a great deal of knowledge on police actions and behaviors within natural field settings (Brunson & Miller, 2023) and being widely considered the primary factor pioneering the systematic study of police behavior (McCluskey et al., 2019, p. 215). The current study was approached with the mindset that observational research methods can provide similar insight into low-visibility surveillance activity of CCTV operators.

The unit of analysis was the targeted surveillances conducted by the CCTV operators during the CCTV Directed Patrol Experiment. The research team used Norris & Armstrong's (1999) definition of targeted surveillance: an operator observation that lasted more than one minute on an individual or group of individuals. Researchers identified targeted surveillances by using stopwatch applications on their cellular phones. Once the time on the stopwatch reached 1 minute, the incident was considered a targeted surveillance and the SSO data collection commenced.  

During all experimental patrol shifts, the lead author and two research assistants observed the activity of the CCTV operators and actions of those being surveilled from within the CCTV control room. CCTV camera feeds were displayed on large monitors mounted on the control room walls, allowing the research team to easily view all activity. The CCTV control room had 3-tiered stadium setting, with CCTV operators conducting their surveillance activities from the top tier. All research team members were seated on the first tier, away from the CCTV operators. The position of the research team was meant to provide researchers the opportunity to observe the surveillance operation without unnecessarily impeding CCTV operator workflows. It was further anticipated that positioning the research team out of the direct line-of-sight of CCTV operators would minimize the likelihood that observer presence would influence productivity, a well-noted validity threat observed in prior labor studies (Wickström & Bendix, 2000).

The CCTV operators worked on the Directed Patrol Experiment on an overtime basis. While Newark’s standard CCTV operation involved both civilian and sworn police officers, the RCT included only sworn police officers as CCTV operators. The Newark Police Department’s Outside Employment Unit (OEU) managed the scheduling and payment of CCTV operators for the project, given it was funded by a grant awarded to an outside academic institution. Prior to scheduling a shift, the OEU supervising officer explained the nature of the RCT and provided an informed consent form to CCTV operators. CCTV operators were informed that they could opt-out of participation prior to or during their tour of duty by informing either the OEU commanding officer or the on-duty VSU Sergeant, respectively. Although, none of the operators opted out of their shifts at any point during the RCT. Six CCTV operators worked across the 39 shifts comprising the CCTV Directed Patrol Experiment.

Control room observations occurred for approximately 176 hours over the 11-week experiment period. A total of 237 individual targeted surveillances occurred. In 12 targeted surveillances, the experimental patrol units were out of service addressing an in-progress event previously reported by the CCTV operators. Given our study focused on operator decision-making during an active patrol initiative (i.e., when the operators have the ability to dispatch patrol units) we excluded these 12 targeted surveillances from the analysis, leaving a final sample of 225. Researchers used a standardized code sheet to document each targeted surveillance. Code sheets were digitized into spreadsheet format for the current study.

4.2. Dependent Variables

The current study focuses on three dependent variables relating to discretionary CCTV operator activity. Survey minutes measures the total length of the targeted surveillance. This variable was calculated by subtracting the start time of the targeted surveillance from the end time. The start, end, and difference time variables were originally in hh:mm:ss format. The difference time was converted into a continuous measure reflecting the length of the surveillance in total minutes.

Reasonable suspicion or probable cause observed is a binary measure reflecting whether any behaviors justifying a police response appeared in the CCTV footage during the targeted surveillance. This variable was measured different ways. First, CCTV operators working on the experiment oftentimes communicated with the other police personnel in the room (i.e., the CCTV operators monitoring the non-experiment cameras and the on-duty sergeant supervising the CCTV operation). These communications oftentimes amounted to affirmative declarations of (e.g., “Look at this person committing a crime”) or questions about (e.g. “Do you think this person is committing a crime?”) potentially criminal behavior observed on camera. Both operator statements and other personnel responses that affirmed suspicion of criminal behavior were coded as reasonable suspicion or probable cause observed. When researchers noted a CCTV operator observing a potentially relevant action not articulated by the operator, the lead author recorded the time of occurrence and camera on which the behavior was observed. At the end of the tour of duty, the lead author sat with the experimental operator, watched the behavior(s) in question with the CCTV operator in the recorded footage, and asked the operator if they believed the behavior amounted to reasonable suspicion or probable cause necessary for a police response. Affirmative responses were coded as ‘1’.3

The final dependent variable was Incident Reported to the Police, operationalized as a binary measure. This variable denotes whether the CCTV operator reported any observed instances of reasonable suspicion or probable cause to the patrol units. Communication between CCTV operators and patrol supervisors occurred via two-way radio, which was audible throughout the CCTV control room via a radio transponder. The incident reported to police variable was recorded as ‘1’ for each targeted surveillance in which the CCTV operator was heard reporting the incident to the field patrol units.

4.3. Independent Variables

In total, eight independent variables were measured for the analysis. Variance inflation factors indicate the absence of multicollinearity between the independent variables. Known suspect is a binary measure that captures whether the person primarily monitored during the targeted surveillance was previously known to the CCTV operator. This was determined a number of ways. First, if the person was arrested, researchers looked up the name of the arrestee and cross referenced all prior arrest reports associated with the intervention to determine if the suspect was previously arrested during the intervention. Given that only 39 (16.46%) incidents resulted in arrest, researchers further relied on observational methods to code the known suspect variable. Similar to the reasonable suspicion or probable cause observed dependent variable, researchers noted when CCTV operators communicated that subjects observed on camera were previously seen engaged in potentially criminal behaviors during the RCT. When a researcher believed a person on camera may be considered a known suspect, the lead author noted the camera and time the person was viewed and inquired with the CCTV operator at the conclusion of the shift.

Researchers visually determined the gender, race, and age of the person primarily targeted during each targeted surveillance. Similar to a prior SSO of CCTV footage (Piza & Sytsma, 2016), researchers classified persons according to five age ranges: late teens – early 20s; mid 20s-early 30s; mid 30s-early 40s; mid 40s-early 50s, and; older than 50. For the purpose of the current study, all persons in the late teens – early 20s and mid 20s-early 30s categories were considered young aged.4 We also measured the total number of individuals monitored during each targeted surveillance. The gender and rank (i.e., Officer, Sergeant, Lieutenant, or Captain) of the CCTV Operators were provided by the Newark Police Department’s OEU.5

The primary land usage of the camera’s surrounding area was classified as either commercial or mixed-residential, as no cameras were installed in purely residential neighborhoods. Instances in which a targeted surveillance was temporarily disrupted by a visible obstruction (e.g., a light pole, bush, or bus shelter) were noted by the researchers during the SSO.

5. Analytical Approach

Each dependent variable was tested through a separate regression analysis. A generalized linear regression model tested the effect of covariates on Surveillance Minutes, with a Gaussian specification to accommodate the continuous dependent variable. Such a generalized model is more robust to normality assumption violations than traditional ordinary least square regression models. Given the nature of the dependent variable, model coefficients are interpreted as the number of minutes the total targeted surveillance increases/decreases with each one-unit change in the independent variable. Logistic regression models tested the effect of covariates on Reasonable Suspicion or Probable Cause Observed and Incident Reported to the Police. Model coefficients were calculated as odds ratios, reflecting the percent of increased or decreased likelihood of the affirmative dependent variable value (i.e., ‘1’) occurring with each one-unit change in the covariate. Reasonable Suspicion and Surveillance Minutes were included as covariates alongside the aforementioned independent variables in their respective regression models given the possible relationship between these two measures. Surveillance Minutes was also included as a covariate in the Incident Reported to the Police logistic regression model. The Surveillance Minutes generalized linear model and Reasonable Suspicion or Probable Cause Observed logistic regression model included all targeted surveillances as observations.6 The Incident Reported to the Police logistic regression model included only the targeted surveillances during which an instance of reasonable suspicion/probable cause was observed. Standard errors were calculated at the CCTV operator level (n=6) in all of the regression models to account for unobserved operator characteristics that may influence discretionary surveillance activities.

6. Results

Table 1 displays descriptive statistics for all dependent and independent variables. The average targeted surveillance lasted 16.52 minutes, with a standard deviation of 15.85 minutes. The shortest targeted surveillance lasted 1 minute while the longest lasted 82 minutes. An instance of reasonable suspicion or probable cause was observed in 104 (46.22%) cases. Of these 104 cases, the CCTV operator reported the event to the patrol units in 72 instances (69.23%). On average, targeted surveillances involved the monitoring of approximately 4 individuals with a standard deviation of 2.96. The minimum and maximum values for targeted persons was 1 and 23, respectively. Fifty-five (24.44%) targeted surveillances involved a known suspect. Seventy-two (32.00%) targeted surveillances used a primary camera within a commercial area. Targeted surveillances were temporarily disrupted due to visible obstructions in 32 (14.22%) instances. Sixty-one (27.11%) targeted surveillances were conducted by a female CCTV operator. Twenty-six (11.56%) targeted surveillances were conducted by CCTV operators with a supervisory rank (Sergeant or above). One-hundred-thirty-one (58.22%) targeted surveillances involved primary targets that were young-aged (from late teens to early 30’s). Fifteen (6.67%) of the targeted surveillances focused on White persons with 202 (89.79%) targeting Black persons. It should be noted that the 35 census tracts intersecting the viewshed of a CCTV camera in the RCT treatment condition had a Black residential population of 68.75% as reflected in 2010 decennial census data.7 While the CCTV-covered neighborhoods are majority Black, the number of Black persons targeted is arguably higher than expected given the neighborhood composition. The fact that such a high proportion of surveilled individuals were Black may explain why race did not achieve statistical significance in any of the regression models, as discussed subsequently.

Table 1. Descriptive Statistics

Table 2 displays the results of the generalized linear regression analysis of Surveillance Minutes. Four independent variables significantly influence the length of targeted surveillances. The presence of a known suspect increased the targeted surveillance by over 7 minutes (β=7.61, p.<0.01). Targeted surveillances involving actions providing reasonable suspicion or probable cause for a police response were over 16 minutes longer than incidents not involving such actions (β=16.73, p.<0.01). Visible obstructions and operators with a supervisor rank were associated with over 9-minute increases (β=9.42; p.=0.01) and over 3-minute decreases (β=-3.13; p.<0.01) in targeted surveillance duration, respectively.

Table 2: Generalized Linear Regression Results for Surveillance Minutes

Table 3 displays the results of the logistic regression analysis of Reasonable Suspicion or Probable Cause Observed. Five independent variables significantly influence observation of reasonable suspicion/probable cause during the targeted surveillance. Each one-minute increase in the surveillance minutes increased the likelihood that an action providing reasonable suspicion/probable cause was observed by 15% (OR = 1.15, p.<0.01). When the targeted surveillance was conducted on a camera in a commercial area the likelihood of reasonable suspicion/probable cause being observed increased nearly four-fold (OR = 3.98, p.<0.01). Targeted surveillances conducted by female CCTV operators were associated with a 43% (OR =1.43, p.=0.03) increased likelihood of reasonable suspicion/probable caused being observed. CCTV operators with supervisory ranks were associated with an over two-fold increased likelihood of reasonable suspicion/probable cause being observed (OR=2.33, p.=0.03). The known suspect variable was associated with a 49% decreased likelihood of reasonable suspicion/probable cause being observed (OR= 0.51; p.=0.03).

Table 3: Logistic Regression Results for Reasonable Suspicion or Probable Cause Observed

Table 4 displays the results of the logistic regression analysis of Incident Reported to the Police. Four independent variables significantly influence the likelihood that a targeted surveillance was reported to the experimental patrol units. Incidents of reasonable suspicion/probable cause observed on a commercial-area camera were nearly 4 times more likely to be reported (OR = 3.89; p.=0.01). Visible obstructions were associated with an over four-fold increase in reporting likelihood (OR = 4.25; p.=0.04). Group size was negatively associated with the dependent variable, with each additional individual added to surveilled groups generating a 14% reduction in the likelihood of the incident being reported (OR = 0.86; p.<0.01). Female CCTV operators were over twice as likely to report observed reasonable suspicion/probable cause to the patrol units (OR = 2.49; p.<0.01).

Table 4: Logistic Regression Results for Incident Reported to Police

7. Discussion and Conclusion

Police discretion has been predominantly analyzed in the context of street-level enforcement decisions. The current study contributes to this rich body of literature by examining how police decision-making unfolds from afar. Study results indicate that a range of factors impact the duration of CCTV operator targeted surveillances and CCTV operator decision-making when determining whether to report criminal infractions.

Fifty-five (24.44%) targeted surveillances observed a known suspect which may be credited to the focused nature of the intervention, as CCTV operators monitored the same subset of cameras each tour of duty. Targeted surveillances of known suspects were over 7 times as long as surveillances of persons unknown to the CCTV operators, but 49% less likely to involve incidents of reasonable suspicion or probable cause. Operators may focus on a given incident for a prolonged amount of time to ensure a police response is warranted. This finding is especially relevant when considering jurisdictions that are short staffed or have a simultaneous incident that demands more police presence (Loveday & Gill, 2004). However, the decreased likelihood of reasonable suspicion or probable cause in targeted surveillances of known suspects suggests somewhat of a diminishing effect, with CCTV operators perhaps focusing on such individuals longer in the hopes of observing criminal infractions that never emerge. In such cases, CCTV operator activity arguably may be better spent observing other situations where criminogenic conditions include aspects other than the individuals on screen. However, this is probably easier said than done in practice, with CCTV operators focusing their attention on individuals their prior experience indicates are worthy of close surveillance. This suggests the need for agency policy to guide CCTV operators on when to focus their surveillance activity on specific situations—and when to focus their attention elsewhere when incidents of concern do not emerge.

Another contributing factor to CCTV operator discretionary decision making is officer rank. Operators in supervisory positions have more experience, thus may require less time to identify and determine reasonable suspicion or probable cause. Operators with a supervisor rank were associated with over 3-minute decreases in a targeted surveillance length, but a two-fold increase in reasonable suspicion/probable cause observed. This is similar to prior literature indicating that a CCTV operators’ ability to evaluate an individual’s intent to harm was largely more efficient in operators with more experience. This may be due to experienced operators being more familiar with the local faces and behavior to be able to target known offenders (Loveday & Gill, 2004; Petrini et al., 2014).

Land use of the targeted surveillance was also meaningful when examining CCTV operator discretion. When a targeted surveillance was conducted on a camera in a commercial area versus mixed-residential, the likelihood of reasonable suspicion/probable cause being observed increased more than four-fold, and the likelihood of reporting increased over three-fold. This finding reflects the general literature on land use and crime (Anderson et al., 2013; Browning et al., 2010; Roncek & Maier, 1991; Stucky & Ottensmann, 2009). All studies found that commercial neighborhoods were associated with higher rates of criminal activity. CCTV operators may have found it easier to detect reasonable suspicion/probable cause because of the higher proportion of outsiders within the area who are coming in for the commercial businesses. This may create more opportunities for both property and violent crime to occur and be subsequently detected on CCTV.

The physical environment may also influence CCTV operator actions by impeding the view of cameras. Visible obstructions were significantly associated with increases in surveillance minutes and the likelihood that incidents of reasonable suspicion/probable cause were reported to the patrol units. In terms of the survey minutes finding, visible obstructions may have forced CCTV operators to view persons longer in order to determine the nature of the on-screen activity, specifically whether it amounted to potentially criminal behavior. Regarding the reporting finding, CCTV operators may have considered the visual obstruction of surveilled individuals as inherently suspicious behavior. Prior research on public open-air drug markets has found that market participants typically use natural settings to obstruct the view of their activity. Sellers and buyers may meet in alleys or behind a bus stop to engage in the sale, for example (Piza & Sytsma, 2016). Such observations are particularly relevant to the current study findings, given over 76% of incidents reported to the patrol units were narcotics related.

Female CCTV operators were 40% more likely than male operators to observe incidents of reasonable suspicion/probable cause and over 4 times more likely to report observed reasonable suspicion/probable cause to the patrol units. Despite assumptions made about women in policing being less likely to engage in more punitive behavior on the job, prior literature echoes the current study’s findings. In an SSO of police officers analyzing gender and arrest decisions, Novak and colleagues (2011) found that gender has little impact on an officer’s decision to arrest. However, female officers were significantly more likely to make an arrest when observed by supervisors. This may be due to women officers occupying a marginalized identity in policing and feeling the need to outperform their colleagues in front of higher ranked officers. Given the visible nature of CCTV operator behavior in the current study, with other operators and supervisors present within the control room, such dynamics may help explain this finding. Furthermore, research examining gender differences in police use of controlling and supporting behaviors in police-citizen encounters concluded that male officers were more likely to employ supporting behaviors than female officers, which may indicate male officers using less punitive discretionary practices than women (Rabe-Hemp, 2008). Perhaps due to the operator room employing both supervisory and novice officers, women CCTV operators may feel more pressure to detect criminal offenses and behavior, and “prove” that they carry out their job the same as their male counterparts.8

Despite the implications of the observed results, this study, like most, suffers from certain limitations that should be acknowledged. Given the remote nature of the observations, researchers used visual cues to determine the race, gender, and age-range of the persons observed on the CCTV video. While the researcher measures matched official arrest reports in each case such reports were generated, we acknowledge that an undetermined number of cases may have been misclassified. Race in particular may be at heightened risk of misclassification. A recent study by Laniyonu & Donahue (2023) compared police officers’ racial categorization of stopped persons with those same persons’ racial self-identification, finding substantial evidence of misclassification, with the classification of Hispanics as non-Hispanic White the most common form. However, given reliance on observational methods in the current study, researchers classified suspects as either White or Black, and did not make any attempts to determine Latino ethnicity. Nonetheless, we acknowledge the potential threat of miscalculation in our data, specifically since we are unaware of any prior research that has empirically estimated rates of gender and age misclassification.

We also acknowledge that while the SSO focused on video footage, it did not enjoy certain benefits identified in prior SSOs of video footage. While researchers have noted a main benefit of video SSOs as the ability to conduct tests of inter-rater reliability (Lindegaard & Bernasco, 2018), the deployment of the RCT did not allow the current study to leverage such benefits. The current study incorporated SSO of real-time (rather than pre-recorded) CCTV footage, meaning researchers were not able to retroactively re-code incidents for reliability test purposes. The use of real-time footage also precluded researchers from pausing, rewinding, and watching footage in slow-motion, which originators of video data analysis methods highlight as a main benefit of the technique (Nassauer & Legewie, 2021, 2022). As such, the SSO in the current study has more in common with traditional SSOs than with studies comprising the emerging video data analysis framework. While this is by no means a fatal flaw, given the rich history and substantial contributions to policing research generated by in-person SSOs (Mastrofski, Parks, & McCluskey, 2010; Worden & McLean, 2014), we acknowledge our approach does not leverage the oft-noted benefits of video data analysis. However, the Newark CCTV Directed Patrol Experiment generated video of the CCTV target areas, inclusive of the behaviors of surveilled persons and any police responses to such. There was no video recordings of the CCTV control room activities. Given that the research team relied on CCTV operator dialogue to code a number of variables, the in-person SSO best allowed the research team the opportunity to directly observe phenomenon relevant to the research question, a hallmark of observational policing research (Brunson & Miller, 2023). As such, our study should be considered as having many of the oft-observed benefits (e.g., qualitative depth, increased structure) and limitations (e.g., reflexivity, perspective bias, participant recall error) as traditional SSOs (Brunson & Miller, 2023; Chillar et al., 2021).

We further acknowledge the post-hoc nature of the current study, leveraging data created from a different research process. The conceptualization and operationalization of study variables may have differed had the initial SSO been designed explicitly to answer the research questions explored in the current study. We recommend that future research replicate our SSO with CCTV operator behavior as the primary, rather than tangential, focus of the study.

The current study sought to examine CCTV operator discretion by utilizing observational data collected during the CCTV Directed Patrol Experiment in Newark, NJ (Piza et al., 2015). Findings suggest that organizational culture, CCTV operator characteristics, and land usage of target areas may foster differential surveillance behavior across CCTV operators. This may have collateral consequences for police and community relations, especially given how CCTV operator activity can impact on-the-ground enforcement activities of police. Viewing from remote cameras allows operators to have a larger scope of the events occurring on the ground. For example, Piza and colleagues (2014) describe an exchange between a lieutenant at the Newark Police Departments’ Narcotics Division, who was monitoring CCTV cameras, and undercover officers in the field. The lieutenant was able to point the officers to the exact location, the alleged crime type (drug sale), and descriptions of the suspect’s physical attributes like, gender, clothing, and vehicle type (Piza et al., 2014). This event shows how remote surveillance can impact officer decision-making. As remote strategies for policing continue to expand internationally, the identification of factors that impact discretionary practices is critical.

Funding Statement

The Newark CCTV Directed Patrol Experiment and systematic social observation activities that collected data for the current study were funded by the National Institute of Justice (grant number 2010-IJ-CX-0026).

References

Alpert, G. P., MacDonald, J. M., & Dunham, R. G. (2005). Police suspicion and discretionary decision making during citizen stops. Criminology, 43(2), 407-434.

Anderson, J. M., MacDonald, J. M., Bluthenthal, R., & Ashwood, J. S. (2013). Reducing crime by shaping the built environment with zoning: An empirical study of Los Angeles. University of Pennsylvania Law Review, 699-756

Arts, K., Van Den Berg, S., Lodders, A., Wagenaar, P., De Wit, D. (2013). Making sense behind the lens: Closed circuit television and the Dutch customs office. Wroclaw Review of Law, Administration and Economics. Vol. 3:2.

Balakrishnan, J. D. (1998). Measures and interpretations of vigilance performance: Evidence against the detection criterion. Human Factors, 40(4), 601-623.

Browning, C. R., Byron, R. A., Calder, C. A., Krivo, L. J., Kwan, M. P., Lee, J. Y., & Peterson, R. D. (2010). Commercial density, residential concentration, and crime: Land use patterns and violence in neighborhood context. Journal of Research in Crime and Delinquency, 47(3), 329-357.

Brunson, R.K. & Miller, A. (2023). Police observational research in the twenty-first century. Annual Review of Criminology, 6, 205-218.

Caplan, J. M., Kennedy, L. W., & Petrossian, G. (2011). Police-monitored CCTV cameras in Newark, NJ: A quasi-experimental test of crime deterrence. Journal of Experimental Criminology, 7(3), 255–274. https://doi.org/10.1007/s11292-011-9125-9

Chan, J. (1997) Changing police culture: Policing in a multicultural society. Cambridge: Cambridge University Press.

Charman, S. (2017) Police socialisation, identity and culture: Becoming blue. London: Palgrave.

Charman, S., & Williams, E. (2022). Accessing justice: The impact of discretion,‘deservedness’ and distributive justice on the equitable allocation of policing resources. Criminology & Criminal Justice, 22(3), 404-422.

Chillar, V., Piza, E., and Sytsma, V. (2021). Conducting a Systematic Social Observation of Body-Camera Footage: Methodological and Practical Insights. Journal of Qualitative Criminal Justice and Criminology, 10(4), 1-30.

Donald, F. M. (2010). A model of CCTV surveillance operator performance. Ergonomics SA: Journal of the Ergonomics Society of South Africa, 22(2), 2–13.

Feise, R. (2002). Do multiple outcome measures require p-value adjustment? BMC Medical Research Methodology, 2(8), 8-11.

Garcia, L. (2004). Escaping the Bonferroni iron claw in ecological studies. Oikos, 105(3), 657–663.

Gaub, J. E., & Koen, M. C. (2021). Cameras and police dataveillance: A new era in policing. In B. A. Arrigo & B. Sellers (Eds.), The pre-crime society: Crime, culture and control in the ultramodern age. Bristol University Press.

Gelsthorpe L and Padfield N (2003) Exercising discretion: Decision-making in the criminal justice system and beyond. Devon: Willan Publishing.

Gill, M., Spriggs, A., Allen, J., Hemming, M., Jessiman, P., Kara, D., ... & Swain, D. (2005). Control room operation: findings from control room observations. London, UK: Home Office.

Gillolly, J.A. (2022). “Lights and sirens”: Variation in 911 call-taker risk appraisal and its effects on police officer perceptions at the scene. Journal of Policy Analysis and Management, 41(3), 762-786.

Gillolly, J.A. (2020). How 911 callers and call-takers impact police encounters with the public: The case of the Henry Louis Gates Jr. arrest. Criminology & Public Policy, 19(3), 787-803

Grant, D., & Williams, D. (2011). The importance of perceiving social contexts when predicting crime and antisocial behaviour in CCTV images. Legal and Criminological Psychology. Vol. 16. 307-322.

Harris, D. A. (1994). Factors for reasonable suspicion: When black and poor means stopped and frisked. Indiana Law Journal, 69(3), 659-688.

Heebels, B., & van Aalst, I. (2020). Surveillance in practice: Operators’ collective interpretation of cctv images. Surveillance and Society, 18(3), 312–327. https://doi.org/10.24908/ss.v18i3.13916

Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 6(2), 65-70.

Klockars, .C.B. (1985) The Idea of the Police. Thousand Oaks, CA: SAGE Publications.

Koper, C. S., & Lum, C. M. (2019). The limits of police technology. In D. L. Weisburd & A. A. Braga (Eds.), Police innovation. Contrasting perspectives (2nd ed., pp. 571–543). Cambridge, UK: Cambridge University Press.

Laniyonu, A. & Donahue, S. (2023). Effect of racial misclassification in police data on estimates of racial disparities. Criminology. DOI: 10.1111/1745-9125.12329.

Lindegaard, M. R., & Bernasco, W. (2018). Lessons learned from crime caught on camera. Journal of Research in Crime and Delinquency, 55(1), 155-186.

Lomell, H. M. (2004). Targeting the unwanted: Video surveillance and categorical exclusion in Oslo, Norway. Surveillance & Society, 2(2/3), 346–360.

Loveday, K., & Gill, M. (2004). The impact of monitored CCTV in a retail environment: What CCTV operators do and why. Crime Prevention and Community Safety, 6, 43–55.

Lum, C., Koper, C., Wu, X., Johnson, W., & Stolz, M. (2020). Examining the empirical realities of proactive policing through systeamtic observations and computer-aided dispatch data. Police Quarterly, 23(3), 283-310.

Lum, C. M., & Koper, C. S. (2017). Evidence-based policing: Translating research into practice. Oxford, United Kingdom: Oxford University Press.

Manning, P. (1979) The social control of police work. In: Holdaway, S. (Ed.) The British Police. London: Edward Arnold, pp. 41–65.

Mastrofski, S. D., Parks, R. B., & McCluskey, J. D. (2010). Systematic social observation in criminology. In Piquero, A. & Weisburd, D. (Eds.) Handbook of Quantitative Criminology (pp. 225-247). Springer, New York, NY.

McCahill, M., & Norris, C. (2003). Estimating the extent, sophistication and legality of CCTV in London. CCTV, 51-66.

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.

Myhill, A. and Johnson, K. (2016) Police use of discretion in response to domestic violence. Criminology & Criminal Justice, 16(1), 3–20.

Nassauer​, ​​A.​, & ​​​Legewie​, ​​N. M.​ (​​2022​). Video data analysis: How to use 21^st century vidoe in the social sciences. Sage: New York.  

​​​Nassauer​, ​​A.​, & ​​​Legewie​, ​​N. M.​ (​​2021​). ​​Video data analysis: A methodological frame for a novel research trend​. ​​Sociological Methods & Research​, ​​50​(​​1​), ​​135​–​​174​.

Neusteter, S. R., O’Toole, M., Khogali, M., Rad, A., Wunschel, F., Scaffidi, S., Sinkewicz, M., Mapolski, M., DeGrandis, P., Bodah, D., & Pineda, H. (2020). Understanding police enforcement: Multicity 911 analysis. Vera Institute of Justice. Available at https://www.vera.org/downloads/publications/understanding-police-enforcement-911-analysis.pdf

New York City Council. (2016). Criminal Justice Reform Act.

Norris, C., & Armstrong, G. (1999). CCTV and the Social Structuring of Surveillance. Surveillance of Public Space, 10, 157–178.

Norris, C., & McCahill, M. (2006). CCTV: Beyond penal modernism? British Journal of Criminology, 46(1), 97–118. https://doi.org/10.1093/bjc/azi047

Novak, K. J., Brown, R. A., & Frank, J. (2011). Women on patrol: An analysis of differences in officer arrest behavior. Policing: An International Journal of Police Strategies and Management, 34(4), 565–587.

Olejnik, S., Li, J., Supattathum, S., & Huberty, C. (1997). Multiple testing and statistical power with modified Bonferroni procedures. Journal of Education and Behavioral Statistics, 22,389–406.

O’Neill M (2016) Revisiting the classics: Janet Chan and the legacy of ‘changing police culture’. Policing and Society 26(4): 475–480.

Petrini, K., McAleer, P., Neery, C., Gillard, J., & Pollick, F. E. (2014). Experience in judging intent to harm modulates parahippocampal activity: An fMRI study with experienced CCTV operators. Cortex. VOL. 57. pp. 74-91.

Piza, E. L. (2018). The crime prevention effect of CCTV in public places: A propensity score analysis. Journal of Crime and Justice, 41(1), 14–30. https://doi.org/10.1080/0735648X.2016.1226931

Piza, E., Caplan, J., & Kennedy, L. (2017). CCTV as a tool for early police intervention: Preliminary lessons from nine case studies. Security Journal, 30(1): 247-265.

Piza, E., Caplan, J., & Kennedy, L. (2014). Is the punishment more certain? An analysis of CCTV detections and enforcement. Justice Quarterly, 31(6): 1015-1043.

Piza, E. & Sytsma, V. (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.

Piza, E. L., Caplan, J. M., Kennedy, L. W., & Gilchrist, A. M. (2015). The effects of merging proactive CCTV monitoring with directed police patrol: A randomized controlled trial. Journal of Experimental Criminology, 11(1), 43–69. https://doi.org/10.1007/s11292-014-9211-x

Piza, E. L., Gilchrist, A. M., Caplan, J. M., Kennedy, L. W., & O’Hara, B. A. (2016). The financial implications of merging proactive CCTV monitoring and directed police patrol: A cost–benefit analysis. Journal of Experimental Criminology, 12(3), 403–429. https://doi.org/10.1007/s11292-016-9267-x

Piza, E. L., Welsh, B. C., Farrington, D. P., & Thomas, A. L. (2019). CCTV surveillance for crime prevention: A 40-year systematic review with meta-analysis. Criminology & Public Policy, 18(1), 135–159. https://doi.org/10.1111/1745-9133.12419

Rabe-Hemp, C. E. (2008). Female officers and the ethic of care: Does officer gender impact police behaviors?. Journal of Criminal Justice, 36(5), 426-434.

Ratcliffe, J. (2006). Video surveillance of public places. Problem-Oriented Guides for Police. Response Guide Series. Guide No. 4. US Department of Justice Office of Community Oriented Policing Services. Center for Problem-Oriented Policing.

Reiner R (2010) The Politics of the police, 4th edn. Oxford: Oxford University Press.

Roncek, D. W., & Maier, P. A. (1991). Bars, blocks, and crimes revisited: Linking the theory of routine activities to the empiricism of “hot spots”. Criminology, 29(4), 725-753.

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).

Salvemini, A. V., Piza, E. L., Carter, J. G., Grommon, E. L., & Merritt, N. (2015). Integrating human factors engineering and information processing approaches to facilitate evaluations in criminal justice technology research. Evaluation Review, 39(3), 308–338. https://doi.org/10.1177/0193841X15583404

Sankoh, A., Huque, M., & Dubey, S. (1997). Some comments on frequently used multiple endpoint adjustment methods in clinical trials. Statistics in Medicine, 16(22), 2529–3542.

Simpson, R. & Orosco, C. (2021). Re-assessing measurement error in police calls for service: Classifications of events by dispatchers and officers. PLOS One, 16(2), e0260365.

Simpson, R. (2021). Calling the police: Dispatchers as important interpreters and manufacturers of calls for service data. Policing: A Journal of Policy and Practice, 15(2), 1537-1545.

Skogan, W. G. (2019). The future of CCTV. Criminology & Public Policy, 18(1), 161–166. https://doi.org/10.1111/1745-9133.12422

Skogan, W., & Frydl, K. (Eds.) (2004). Fairness and Effectiveness in Policing: The Evidence. Washington, D.C.: The National Academies Press.

Smith, G. (2004). Behind the screens: Examining constructions of deviance and informal practices among CCTV control room operators in the UK. Surveillance & Society, (2/3), 376-395.

Stucky, T. D., & Ottensmann, J. R. (2009). Land use and violent crime. Criminology, 47(4), 1223-1264.

Thomas, A. L., Piza, E. L., Welsh, B. C., & Farrington, D. P. (2022). The internationalisation of cctv surveillance: Effects on crime and implications for emerging technologies. International Journal of Comparative and Applied Criminal Justice, 46(1), 81–102. https://doi.org/10.1080/01924036.2021.1879885

Tillyer, R., Smith, M., & Lloyd, C. D. (2019). Another piece of the puzzle: The importance of officer characteristics and group processes in understanding post-stop outcomes. Journal of Research in Crime and Delinquency, 56(5), 736-779.

Taylor, P. L. (2020). Dispatch priming and the police decision to use deadly force. Police Quarterly, 23(3), 311–332.

Tyler, T. (1990) Why People Obey the Law. New Haven, CT: Yale University Press.

Tyler, T. (2016). Police discretion in the 21st century surveillance state. University of Chicago Legal Forum, 2016(1), 14.

Welsh, B. C., & Farrington, D. P. (2002). Crime prevention effects of closed circuit television: A systematic review. Home Office (Vol. Research S). https://doi.org/10.4073/csr.2008.17

Welsh, B. C., & Farrington, D. P. (2009). Public area CCTV and crime prevention: An updated systematic review and meta‐analysis. Justice Quarterly, 26(4), 716–745. https://doi.org/10.1080/07418820802506206

Wickström, G. & Bendix, T. (2000). The “Hawthorne Effect” — what did the original Hawthorne studies actually show? Scandinavian Journal of Work, Environment & Health, 26(4), 363-367.

Williams, D. (2007). Effective CCTV and the challenge of constructing legitimate suspicion using remote visual images. Journal of Investigative Psychology and Offender Profiling, 4(2), 97-107.

Williams, E. & Stanko, B. (2016) Researching sexual violence. In: Brunger M, Tong, S. & Martin D. (Eds.) Introduction to Policing Research: Taking Lessons from Practice. Oxford: Routledge, pp. 121–136.

Wilson, D., & Sutton, A. (2003). Open-Street CCTV in Australia: A comparative study of establishment and operation. Report No. 26/01-02. Australia: Criminology Research Council.

Worden R.E., McLean S.J. (2014). Police Discretion in Law Enforcement. In: Bruinsma G. & Weisburd D. (Eds.) Encyclopedia of Criminology and Criminal Justice. Springer: New York, NY.