Journal Article Summary APA Format
3.ARIEL 3/12/2017 4:53 PM
0091-4169/16/10604-0729
THE JOURNAL OF CRIMINAL LAW & CRIMINOLOGY Vol. 106, No. 4
Copyright © 2017 by Barak Ariel Printed in U.S.A.
729
CRIMINOLOGY
POLICE BODY CAMERAS IN LARGE
POLICE DEPARTMENTS
BARAK ARIEL*
Body Worn Cameras are spreading worldwide, under the assumption
that police performance, conduct, accountability, and legitimacy, in the
eyes of the public, are enhanced as a result of using these devices. In
addition, suspects’ demeanor during police–public engagements is
hypothesized to change as a result of the video-recording of the encounter.
For both parties—officers and suspects—the theoretical mechanism that
underpins these behavioral changes is deterrence theory, self-awareness
theory, or both. Yet evidence on the efficacy of Body Worn Cameras
remains largely anecdotal, with only one rigorous study, from a small force
in Rialto, California, validating the hypotheses. How Body Worn Cameras
affect police–public interactions in large police departments remains
unknown, as does their effect on other outcomes, such as arrests. With one
Denver police district serving as the treatment area and five other districts
within a large metropolitan area serving as comparisons, we offer mixed
findings as in the Rialto Experiment, not least in terms of effect magnitudes.
Adjusted odds-ratios suggest a significant 35% lower odds for
citizens’ complaints against the police use of force, but 14% greater odds
for a complaint against misconduct, when Body Worn Cameras are used.
No discernable effect was detected on the odds of use of force at the
aggregate, compared to control conditions (OR=0.928; p>0.1). Finally,
arrest rates dropped significantly, with the odds of an arrest when Body
Worn Cameras not present is 18% higher than the odds under treatment
conditions. The outcomes are contextualized within the framework of
reactive emergency calls for service rather than proactive policing. We
further discuss officers’ decisions and the degree of the necessity of arrest
in policing more broadly, because the burden of proof for tangible evidence
necessary for making a legal arrest can be challenged with the evidence
produced by Body Worn Cameras: officers become “cautious” about
3. ARIEL 3/12/2017 4:53 PM
730 ARIEL [Vol. 10
6
arresting suspects when Body Worn Cameras are present. Limitations
associated with the lack of randomly assigned comparison units are
discussed, as well, with practical recommendations for future research on
Body Worn Cameras.
TABLE OF CONTENTS
INTRODUCTION ……………………………………………………………………………… 730
I. WHAT DO WE KNOW ABOUT POLICE BODY WORN CAMERAS? ……….. 734
II. TESTING THE EFFECT OF BWCS IN LARGE POLICE DEPARTMENTS ….. 737
III. METHODS AND DATA ……………………………………………………………….. 738
A. Experimental Design ……………………………………………………….. 738
B. Settings and Procedure …………………………………………………….. 738
C. Data Sources …………………………………………………………………… 739
D. Treatment and Comparison Geographic Sites ……………………… 740
E. Apparatus ……………………………………………………………………….. 741
IV. MEASURES ……………………………………………………………………………… 742
A. Use of Force …………………………………………………………………… 742
B. Citizen Complaints ………………………………………………………….. 743
C. Arrests …………………………………………………………………………… 745
D. Citizen-Initiated 911 Calls for Service ……………………………….. 747
V. STATISTICAL PROCEDURE ………………………………………………………….. 747
VI. QUALITATIVE ANALYSES: OFFICERS’ SURVEYS …………………………… 748
VII. RESULTS ……………………………………………………………………………….. 749
A. Descriptive Statistics ……………………………………………………….. 749
B. Treatment Outcomes ……………………………………………………….. 751
VIII. DISCUSSION AND CONCLUSIONS …………………………………………….. 754
A. Effect Of BWCs on Use of Force: Accountability and
Transparency ……………………………………………………………….. 756
B. Effect Of BWCs on Complaints: Conditional on
Complaint Type ……………………………………………………………. 760
C. Effect of BWCs on Arrest Decisions………………………………….. 762
D. A Cautionary Note on Nonexperimental Designs in Future
Studies on BWCs ………………………………………………………….. 766
CONCLUSION ………………………………………………………………………………… 767
INTRODUCTION
Police departments have begun using Body Worn Cameras (BWCs) in
daily operations all over the world, in increasing rates.1 BWCs are
1 See Uri Friedman, Do Police Body Cameras Actually Work?, THE ATLANTIC, Dec. 3,
2015, http://www.theatlantic.com/international/archive/2014/12/do-police-body-cameras-
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 731
hypothesized to minimize the use of force in police–public encounters,
reduce citizens’ complaints, and increase the accountability and the
legitimacy of the police.2 At the same time, the massive growth in
implementation of BWCs is not mirrored by research on their cost-
effectiveness or efficiency.3 Currently, there is a dearth of rigorous
evaluation on the efficacy of BWCs4 with much of the published work
concentrating on implementation processes,5 officers’ perceptions about the
use of BWCs on policing and their professional role,6 the extent to which
officers feel micromanaged in an era of digital surveillance,7 and legal
issues associated with privacy rights in the public domain.8 One noteworthy
work-ferguson/383323 (discussing the increase in federal funding for the technology).
2 See infra notes 20–44 and accompanying text.
3 See MICHAEL D. WHITE, POLICE OFFICER BODY-WORN CAMERAS: ASSESSING THE
EVIDENCE 32–34 (OJP Diagnostic Center 2014) (speaking generally on costs and time
incurred with storing camera data).
4 See, e.g., WHITE, supra note 3, at 29–32 (weighing costs and benefits); Barak Ariel et
al., Effect of Police Body-Worn Cameras on Use of Force and Citizens’ Complaints Against
the Police: A Randomized Controlled Trial, 31 J. QUANTITATIVE CRIMINOLOGY 509, 528–31
(2015) (weighing the costs and benefits of BWCs); Justin T. Ready & Jacob T.N. Young,
The Impact of On-Officer Video Cameras on Police–Citizen Contacts: Findings from a
Controlled Experiment in Mesa, Arizona, 11 J. EXPERIMENTAL CRIMINOLOGY 445, 446
(2015).
5 See, e.g., POLICE EXECUTIVE RESEARCH FORUM, U.S. DEP’T OF JUST., IMPLEMENTING A
BODY-WORN CAMERA PROGRAM [hereinafter “POLICE EXECUTIVE RESEARCH”] (2014)
(giving an overview of the BWCs, issues with current usage, and recommendations for
implementation moving forward); NAT’L INST. OF JUST., U.S. DEP’T OF JUST., A PRIMER ON
BODY-WORN CAMERAS FOR LAW ENFORCEMENT 7–10 (2012) (reviewing implementation
issues on when to use, policies and procedures, training, and data storage and management);
Paul Drover & Barak Ariel, Leading an Experiment in Police Body-Worn Video Cameras,
25 INT’L CRIM. JUST. REV. 80 (2015) (discussing the issues in implementing BWCs).
6 See Wesley G. Jennings et al., Cops and Cameras: Officer Perceptions of the Use of
Body-Worn Cameras in Law Enforcement, 42 J. CRIM. JUST. 549, 550 (2014).
7 See Neil Wain & Barak Ariel, Tracking of Police Patrol, 8 POLICING 274, 278, 281
(2014) (“The limited literature on the impact of tracking further raises a number of questions
on how tracking technology can affect officer behaviour or police subculture. For instance,
what will the impact of GPS, body-worn videos and other tracking devices be on officer’s
discretion? On the one hand, there have been recent voices in British policing calling for
increased discretion, including providing officers with the power to implement out of court
disposals. However, if ‘every’ decision is tracked and inspected, would it have an effect on
the willingness of officers to exercise this discretion? Similarly, are officers less inclined to
engage with suspects when they are aware that they are being watched? We believe that
these questions are indeed normative in nature, yet before introducing policies, basic science
must be applied in order to provide policymakers with the necessary evidence about the
scope of these phenomena from an empirical perspective.” (internal citations omitted)).
8 David A. Harris, Picture This: Body-Worn Video Devices (Head Cams) as Tools for
Ensuring Fourth Amendment Compliance by Police, 43 TEX. TECH. L. REV. 357 (2010). See
3. ARIEL 3/12/2017 4:53 PM
732 ARIEL [Vol. 106
study on the effectiveness of BWCs, in the specific area of use of force and
complaints, was conducted in Rialto, California.9 The “Rialto Experiment”
showed that the likelihood that police use force when officers do not use
BWCs was roughly twice that of when officers use BWCs and that the
number of complaints lodged against officers dropped from 0.7 complaints
per 1000 contacts to 0.07 per 1000 contacts.10
The Rialto Experiment was widely cited in recent cases of police use
of force as a method to reduce the likelihood of these incidents.11 The death
of Eric Gardner in Staten Island, New York after police put him in a
chokehold for several seconds, in contravention of the departmental
prohibition;12 the death of Michael Brown in Ferguson, Missouri, which
resulted in protests in 2014;13 and the shooting of Walter Scott by a white
North Charleston police officer14 are examples thereof. Some, including the
current White House Administration, have argued that BWCs could be used
as a technological advent that would revitalize police–public relations and
generally Helen Nissenbaum, Protecting Privacy in an Information Age: The Problem of
Privacy in Public, 17 L. & PHIL. 559 (1998) (discussing the legal issues and questions
arising from new privacy concerns).
9 See Ariel et al., supra note 4.
10 Id. at 510. See Barak Ariel et al.,’Contagious Accountability’: A Global Multi-Site
Randomized Controlled Trial on the Effect of Police Body-Worn Cameras on Citizen’s
Complaints against the Police, CRIM. J. & BEHAV. (forthcoming).
11 See EUGENE P. RAMIREZ, A REPORT ON BODY WORN CAMERAS 6–11 (2014),
https://www.bja.gov/bwc/pdfs/14-005_Report_BODY_WORN_CAMERAS ; Kelly
Gates, Body-Worn Devices and Police Media Labor, in THE ROUTLEDGE COMPANION TO
LABOR AND MEDIA 215 (2015); Fanny Coudert et al., Body-Worn Cameras for Police
Accountability: Opportunities and Risks, 31 COMPUTER L. & SECURITY REV. 749, 750–51
(2015); Paul Marks, Body Worn Cameras Mean Police are Always Watching You, 220 NEW
SCIENTIST 21, 21 (2013); Alexandria C. Meetscu & Alex Rosenblat, Police Body-Worn
Cameras, DATA & SOC’Y 17–21 (2015), http://www.datasociety.net/pubs/dcr/Police
BodyWornCameras . But see First Scientific Report Shows Police Body-Worn-Cameras
Can Prevent Unacceptable Use-of-Force, UNIV. OF CAMBRIDGE [hereinafter “UNIVERSITY OF
CAMBRIDGE”] (Dec. 23, 2014), http://www.cam.ac.uk/research/news/first-scientific-report-
shows-police-body-worn-cameras-can-prevent-unacceptable-use-of-force (questioning the
study).
12 See Conor Friedersdorf, Eric Garner and the NYPD’s History of Deadly Chokeholds,
THE ATLANTIC (Dec. 4, 2014), http://www.theatlantic.com/national/archive/2014/12/context-
for-the-punishment-free-killing-of-eric-garner/383413/.
13 See U.S. DEP’T OF JUST., REPORT REGARDING THE CRIMINAL INVESTIGATION INTO THE
SHOOTING DEATH OF MICHAEL BROWN BY FERGUSON, MISSOURI POLICE OFFICER DARREN
WILSON (2015).
14 See Michael S. Schmidt & Matt Apuzzo, South Carolina Officer Is Charged with
Murder of Walter Scott, N.Y. TIMES (Apr. 7, 2015), http://www.nytimes.com/2015/04/08/us/
south-carolina-officer-is-charged-with-murder-in-black-mans-death.html?_r=0.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 733
prevent these incidents.15 Had the officers involved in the incidents cited
above been issued BWCs, they may have dealt with these situations
differently. The Rialto Experiment was suggested as the necessary evidence
to support this contention, including a citation by the United States district
judge in the 2013 ruling against the New York Police Department (NYPD)
over stop and search.16 The Rialto Experiment has influenced policy
discussions over improvements in police conduct and legitimacy, including
the recent discussion by the President’s Task Force on 21st Century
Policing.17
Yet the Rialto Experiment is only one study; replications are urgently
required in order to show whether these findings represent an anomaly
attributed to the Rialto context, to the novelty of these devices in police
operations, or both.18 Perhaps as important, Rialto is a small department,
with about 50 frontline officers force-wide.19 It remains unknown whether
similar effects on police–public encounters would be detected—in both
directionality as well as magnitude—in large police departments, within
large metropolitan settings.
In the present study, we tested the effect of using BWCs in police
frontline operations on police use of force, complaints lodged against
officers and arrest, in one of the largest state and local law enforcement
agencies in the United States—the Denver Police Department—over a
period of six months. One police district out of six was assigned BWCs,
while all other districts served as comparison sites without BWCs. We
observed the effect of BWCs on these outcomes based on adjusted odds
ratios at the aggregated level: comparing the odds of use of force,
complaints, and arrest in the ‘treatment district’ compared to five other
districts, as a way to estimate the effect of using BWCs.
After reviewing the existing research on BWCs and examining the
15 See, e.g., PRESIDENT’S TASK FORCE ON 21ST CENTURY POLICING, FINAL REPORT OF THE
PRESIDENT’S TASK FORCE ON 21ST CENTURY POLICING 31–32 (2015) [hereinafter
“PRESIDENT’S TASK FORCE”] (“An increasing number of law enforcement agencies are
adopting BWC programs as a means to improve evidence collection, to strengthen officer
performance and accountability, and to enhance agency transparency.”).
16 See, e.g., Floyd v. City of New York, 959 F. Supp. 2d 540 (S.D.N.Y. 2013).
17 See PRESIDENT’S TASK FORCE, supra note 15, at 31–32.
18 One similar replication was recently conducted in Orlando. See Wesley G. Jennings et
al., Evaluating the Impact of Police Officer Body-Worn Cameras on Response-to-Resistance
and Serious External Complaints: Evidence from the Orlando Police Department
Experience Utilizing a Randomized Controlled Experiment, 43 J. CRIM. JUST. 480, 481
(2015).
19 See Ariel et al., supra note 4.
3. ARIEL 3/12/2017 4:53 PM
734 ARIEL [Vol. 106
theoretical mechanisms that underpin the effect of these devices on police–
public encounters, we next lay out the Denver Police Department
experiment and its design. The outcomes of the study are then presented,
broken down into different outcomes of interest. The practical implications,
with an emphasis on avenues for future research, are contemplated in the
discussion chapter, including the methodological limitations of the present
study.
I. WHAT DO WE KNOW ABOUT POLICE BODY WORN CAMERAS?
Ariel, Farrar, and Sutherland have recently reported the findings of
what is now commonly referred to as the Rialto Experiment.20 The study,
conducted in the small jurisdiction of Rialto, California, with just over fifty
frontline officers, compared nearly 500 police shifts in which all police–
public encounters were assigned to treatment conditions and an equal
number of police shifts to control conditions.21 During treatment shifts,
officers were asked to videotape all their encounters with members of the
public, to announce to the parties with whom they have engaged that the
encounter was videotaped, and to subsequently store evidence on a secured
cloud.22 In control shifts, the officers were tasked never to use the devices.23
Outcomes were then measured in terms of officially-recorded use of force
incidents and complaints lodged against Rialto police officers.24 Following
this twelve month experiment, Ariel, Farrar, and Sutherland reported a
relative reduction of roughly 50% in the total number of incidents of use of
force compared to control conditions and a 90% reduction in citizens’
complaints, compared to the twelve months prior to the experiment.25
The findings have generated heated debates worldwide, particularly
around the transferability of the findings to other jurisdictions, or to larger
police departments.26 Whether unique circumstances in Rialto jeopardized
20 See id.
21 Id. at 518–19.
22 Id. at 511, 521.
23 Id. at 523 (discussing experimental versus control shifts within a Poisson model).
24 Id.
25 Id. at 523–24.
26 See, e.g., POLICE EXECUTIVE RESEARCH, supra note 5, at 22–24; Kirk Johnson,
Today’s Police Put on a Gun and a Camera, N.Y. TIMES (Sept. 27, 2014), http://www.
nytimes.com/2014/09/28/us/todays-police-put-on-a-gun-and-a-camera.html; see also First
Scientific Report Shows Police Body-Worn-Cameras Can Prevent Unacceptable Use-of-
Force, PHYS.ORG (Dec. 24, 2014), http://phys.org/news/2014-12-scientific-police-body-
worn-cameras-unacceptable-use of force.html.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 735
the external validity of the test were also raised.27 Major metropolitan cities,
and with them large law enforcement agencies, operate on a different scale
to small or even medium sized forces.28 Larger forces can be exposed to
more diverse problems, including a nighttime economy of a different scale
than small-scale departments, an incomparable volume of calls for service,
and potentially more serious crimes than local agencies.29 Training,
interagency collaborative work and, perhaps, the expertise of officers and
how likely they are to use force are potentially different in large versus
small police departments, not to mention police cultures, promotional
processes, and budgets.30
Despite these discussions, additional research on BWCs is virtually
nonexistent. The most updated literature review when this study was
conducted has concluded that:
Independent research on body-worn camera technology is urgently needed. Most of
the claims made by advocates and critics of the technology remain untested . . . .
Researchers should examine all aspects of the implementation and impact of the
technology—from its perceived civilizing effect, evidentiary benefits, and impact on
citizen perceptions of police legitimacy to its consequences for privacy rights, the law
enforcement agency, and other outside stakeholders.31
At least in theory, the mechanism which underpins the effect of BWCs on
police officers and suspects alike is consistent, regardless of department
size. Deterrence and self-awareness theories suggest that people alter their
behavior once made aware that they are being observed.32 A rich body of
evidence on perceived social surveillance—self-awareness and socially-
desirable-responding—suggests that people adhere to social norms and
change their conduct because of their cognizance that someone else is
27 UNIVERSITY OF CAMBRIDGE, supra note 11.
28 E.g., Laure Weber Brooks & Nicole Leeper Piquero, Police Stress: Does Department
Size Matter?, 21 POLICING 600, 604 (1998); Gary W. Cordner, Police Agency Size and
Investigative Effectiveness, 17 J. CRIM. JUST. 145, 146 (1989); Robert M. Regoli et al.,
Police Cynicism, Job Satisfaction and Work Relations of Police Chiefs: An Assessment of the
Influence of Department Size, 22 SOC. FOCUS 161, 162–63 (1989).
29 See UNIVERSITY OF CAMBRIDGE, supra note 11.
30 Brooks & Piquero, supra note 28, at 602–15 (researching the different size models in
police stress and how they handle their duties); Kwabena Gyimah-Brempong, Economies of
Scale in Municipal Police Departments: The Case of Florida, 69 REV. ECON. & STAT. 352,
352–54 (1987); Regoli et al., supra note 28; see also Victoria M. Follette et al., Mental
Health and Law Enforcement Professionals: Trauma History, Psychological Symptoms, and
Impact of Providing Services to Child Sexual Abuse Surviors, 25 PROF. PSYCHOL. 275, 280–
81 (1994) (discussing impact of a person’s personal history on how they act professionally).
31 WHITE, supra note 3, at 10.
32 See, e.g., id. at 13; Ariel et al., supra note 4, at 516.
3. ARIEL 3/12/2017 4:53 PM
736 ARIEL [Vol. 106
watching.33 We experience public self-awareness, which affects our various
social cognitive processes, when we know with sufficient certainty our
behavior is observed or judged.34
The immediate application of this psychological mechanism is
manifested in deterrence theory. An extensive body of recent rigorous
research, across several categories of criminal behavior, has shown that
when certainty of apprehension for wrongdoing is “strong,” socially and
morally-unacceptable acts are dramatically less likely to occur.35
Particularly in terms of crime and disorder, when the consequences of
apprehension can be bleak (imprisonment, fines, etc.), people simply do not
want to get caught.36 In this framework, getting caught doing something
morally or socially wrong is often registered as behavior that can
potentially lead to negative consequences, which is an outcome that rational
individuals tend to avoid.37 Studies have, nevertheless, uncovered a
propensity to avoid negative outcomes, and findings generally agree that
individuals react compliantly to even the slightest cues indicating that
somebody may be watching.38
33 Delroy L. Paulhus, Balanced Inventory of Desirable Responding, in MEASURES OF
PERSONALITY AND SOCIAL PSYCHOLOGICAL ATTITUDES 37, 41–43 (1991); Kristen Munger &
Shelby J. Harris, Effects of an Observer on Hand Washing in a Public Restroom, 69
PERCEPTUAL AND MOTOR SKILLS 733, 734 (1989); Robert A. Wicklund, Objective Self-
Awareness, 8 ADVANCES IN EXPERIMENTAL SOC. PSYCHOL. 233, 261–67 (1975).
34 See Pat Barclay, Trustworthiness and Competitive Altruism Can Also Solve the
“Tragedy of the Commons,” 25 EVOLUTION AND HUM. BEHAV. 209, 217–19 (2004); Will M.
Gervais & Ara Norenzayan, Like a Camera in the Sky? Thinking about God Increases
Public Self-Awareness and Socially Desirable Responding, 48 J. EXPERIMENTAL SOC.
PSYCHOL. 298, 301–02 (2012); Manfred Milinski et al., Donors to Charity Gain in Both
Indirect Reciprocity and Political Reputation, 269 PROC. ROYAL SOC’Y LONDON 881, 881
(2002); Claus Wedekind & Victoria A. Braithwaite, The Long-Term Benefits of Human
Generosity in Indirect Reciprocity, 12 CURRENT BIOLOGY 1012, 1014 (2002).
35 ANDREW VON HIRSCH ET AL., CRIMINAL DETERRENCE AND SENTENCE SEVERITY 26–27
(1999); Travis C. Pratt et al., The Empirical Status of Deterrence Theory: A Meta-Analysis,
in TAKING STOCK: THE STATUS OF CRIMINOLOGICAL THEORY 367, 371–74 (2006); Daniel
Nagin, Deterrence in the Twenty-First Century, 42 CRIME & JUST. 199, 201–04 (2013);
Raymond Paternoster, How Much Do We Really Know About Criminal Deterrence, 100 J.
CRIM. L. & CRIMINOLOGY 765, 769, 771, 774–76 (2010).
36 Nagin, supra note 35.
37 Steven Klepper & Daniel Nagin, The Deterrent Effect of Perceived Certainty and
Severity of Punishment Revisited, 27 CRIMINOLOGY 721, 742–744 (1989); Nagin, supra note
35, at 251–52. See generally Daniel Nagin et al., Deterrence, Criminal Opportunities, and
Police, 53 CRIMINOLOGY 74 (2015).
38 See Melissa Bateson et al., Cues of Being Watched Enhance Cooperation in a Real-
World Setting, 2 BIOLOGY LETTERS 412, 413 (2006); Kevin J. Haley & Daniel M.T. Fessler,
Nobody’s Watching? Subtle Cues Affect Generosity in an Anonymous Economic Game,
26 EVOLUTION AND HUM. BEHAV. 245, 254 (2005). But cf. Ernst Fehr & Frederic Schneider,
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 737
Deterrence and self-awareness work equally on suspects who would
otherwise decide to commit crime and on police officers who might
otherwise break the rules of conduct.39 For this reason, BWCs are
hypothesized to work simultaneously on both actors in a police–public
encounter.40 From this follows a logical conclusion that when officers and
suspects are cognizant of the BWC, they are equally assumed to have no
preference for breaking the rules, as the risk of apprehension and conviction
by the evidence captured on videotape is overwhelming. BWCs—unlike
CCTV, dashboard cameras or bystanders’ mobile-phone cameras—are
viewed as “credible threats.”41 It is therefore logical to assume that both
parties in the interaction are conscious not only of the fact that they are
being watched, but also of the consequences associated with
noncompliance.42 “Getting-away” with rule breaking is thus far less
conceivable if one is being-videotaped and one is cognizant that the
behavior is in fact videotaped.43 The evidence from the Rialto Experiment
supports this model.44
II. TESTING THE EFFECT OF BWCS IN LARGE POLICE DEPARTMENTS
To test whether the Rialto Experiment findings are translatable to
large, metropolitan law enforcement agencies, we turned to Denver Police
Eyes Are on Us, but Nobody Cares: Are Eye Cues Relevant for Strong Reciprocity?, 277
PROC. ROYAL SOC’Y B 1315, 1321–22 (2010) (finding a “null effect” when looking at
implicit eye cues).
39 Ariel et al., supra note 4, at 516–18, 527–30.
40 Barak Ariel, Technology in Policing: The Case for Body-Worn Cameras and Digital
Evidence, 83 THE POLICE CHIEF 27 (August 2016); Joshua Young, Implementation of a
Randomized Controlled Trial in Ventura, California: A Body-Worn Video Camera
Experiment 45–46 (Dec. 2014) (unpublished master’s thesis, University of Cambridge) (on
file with the University of Cambridge Institute of Criminology Thesis Database).
41 See e.g., ROBERT JERVIS ET AL., PSYCHOLOGY AND DETERRENCE 153 (1989) (discussing
deterrence theory and the “role of credible threats in deterring potential aggressors”).
42 See Jennings et al., supra note 19, at 485.
43 Drover & Ariel, supra note 5 at 89 (“[T]here had been a number of instances where a
member of the public had changed their behavior from aggression to reluctant compliance
after being informed they were being videotaped.”).
44 Ariel et al., supra note 4, at 526. See also Barak Ariel, et al., Wearing Body Cameras
Increases Assaults Against Officers and Does Not Reduce Police Use of Force: Results from
a Global Multi-site Experiment, EUROPEAN J. OF CRIMINOLOGY (2016), currently available at
http://euc.sagepub.com/content/early/2016/05/17/1477370816643734.full +html; Barak
Ariel et al., Report: Increases in Police Use of Force in the Presence of Body-Worn
Cameras are Driven by Officer Discretion: a Protocol-Based Subgroup Analysis of Ten
Randomized Experiments, J. EXPERIMENTAL CRIMINOLOGY (2016), currently available at
http://link.springer.com/article/10.1007/s11292-016-9261-3.
3. ARIEL 3/12/2017 4:53 PM
738 ARIEL [Vol. 106
Department. All response police officers in one district were assigned
BWCs, while the other districts were not given BWCs at all. Officers were
tasked to conduct patrol “as they normally would,” in both treatment and
comparison areas, while the only planned difference between the two arms
was the deployment of BWCs on all frontline officers in one, but not in
other, districts. Official reports of complaints against the police and
specifically of incidents of use of force, as well as other police outputs,
were measured before deploying the BWCs, and then again during the six
months of the experiment, across the entire city. Data was analyzed at the
aggregated geographic level (district level) using adjusted odds ratios.
III. METHODS AND DATA
A. EXPERIMENTAL DESIGN
The methodological standard for evaluations of causal estimates is the
randomized controlled trial,45 but these are not always feasible, for practical
or political reasons.46 In this experiment, we were unable to randomly
allocate shifts, officers, cases, or vehicles. Instead, we were able to closely
observe, prospectively, the deployment of BWCs in one district in Denver,
and compare it to the other districts that served as controls. Thus, we were
able to apply a Level 3+ on the Maryland Scientific Methods Scale,47 which
can provide informative causal inferences with a fairly satisfactory degree
of internal validity—at least in comparison with before–after studies with
no control groups—about the effect of BWCs on policing.48
B. SETTINGS AND PROCEDURE
Denver, Colorado, is a city that covers approximately 153 square miles
and is home to over 650,000 residents.49 The local population is 46.2% non-
45 WILLIAM R. SHADISH ET AL., EXPERIMENTAL AND QUASI-EXPERIMENTAL DESIGNS FOR
GENERALIZED CAUSAL INFERENCE 13 (2002).
46 See Jack R. Greene, New Directions in Policing: Balancing Prediction and Meaning
in Police Research, 31 JUST. Q. 193, 194–95, 199 (2014) (discussing the limitations of
controlled experimentation in police research and possible alternative ways of collecting
data about police).
47 David P. Farrington et al., The Maryland Scientific Methods Scale, in EVIDENCE-
BASED CRIME PREVENTION 13–21 (2002); see, e.g., NAT’L INST. OF JUST., U.S. DEP’T OF
JUST., PREVENTING CRIME: WHAT WORKS, WHAT DOESN’T, WHAT’S PROMISING 4 (1998)
[hereinafter “PREVENTING CRIME”] (noting the Scale of Scientific Methods ranks studies
from 1 (lowest) to 5 (highest)).
48 See generally PREVENTING CRIME, supra note 47.
49 See U.S. CENSUS BUREAU, THE 2014 DENVER COUNTY QUICKFACTS, http://www.
census.gov/quickfacts/table/PST045215/0820000,08031,00.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 739
White residents (10.1% African-Americans).50 There are disproportionally
more persons living below the poverty line in the city as compared to the
rest of the state (18.3% vs. 12%); however, residents of Denver County
have a similar per-capita income as the rest of the state—about $34,000 per
annum.51 In terms of crime, Denver experiences normal crime patterns (e.g.,
4.7 homicides per 100,000).52 The police department is one the fifty largest
police departments, with nearly 1500 sworn officers working in six
geographic districts.53 As in most major cities, each district is then broken
down into many precincts, and each precinct is patrolled by up to two
officers in a police car.
BWCs were allocated to all frontline officers in one district (n
officers=119) for a period of six months (July 23, 2014–December 15,
2014), but not to any other frontline officers of the other five geographic
districts (n officers=513). The single geographic district was therefore the
treatment area, while each of the five other districts served as comparison
sites.
C. DATA SOURCES
For the purposes of this study, access was granted to a rich database of
administrative, as well as geospatial data on all outputs reported internally
by police officers, during the six months of the experiment and twelve
months prior to the experiment (974,240 code lines).54 The data was broken
down into six police districts, and within these districts, more granular
analysis of each outcome was then conducted. We were particularly
interested in use of force and complaints as dependent variables, but were
quick to learn that arrests—a dimension that was not covered in the original
Rialto Experiment—could be addressed as well.
For arrests, however, we were concerned that initiatives specific to
individual districts are unique and therefore create statistical noise that
would make it difficult to compare the treatment and control conditions.55
Thus, we did not take into account police-generated arrests resulting from
stop-and-account, street checks or unique operations. While our results
50 See id.
51 See http://www.census.gov/quickfacts/table/PST045215/08,48089,0820000,08031,00.
52 See http://www.city-data.com/crime/crime-Denver-Colorado.html (2014 data).
53 See http://www.bjs.gov/content/pub/pdf/lpd13ppp ; https://en.wikipedia.org/wiki/
Denver Police_Department.
54 We did not incorporate data prior to twelve months pretreatment as force district
boundaries were changed prior.
55 See generally Ariel et al., supra note 4.
3. ARIEL 3/12/2017 4:53 PM
740 ARIEL [Vol. 106
would therefore be limited to a subset of police–public interactions—
victim-generated and witnesses-reported crimes—this tradeoff enabled us
to make causal inferences with fewer assumptions that are more difficult to
defend.
Another concern was raised when we noticed that the baseline figures
are different for the treatment and control conditions. This baseline
inequality required statistical adjustments, in order to make the groups more
comparable. This comparability is discussed below in more detail. We have
taken into consideration the number of victim-generated emergency calls
for service (911 calls), as a way to stabilize the pre-treatment conditions.
These are the types of calls for service that are less susceptible to data-
recording manipulations by the field officers, as they are recorded by call
takers, rather than the officers—and could be used as controlling variables.
Overall, access was granted to eighteen months of data on 1184
recorded incidents of use of police force, 844 complaints against police
misconduct and 223 complaints against police use of force. The breakdown
of the effect of BWCs on complaint types is novel, as the number of
complaints post-treatment was too small in the Rialto Experiment for
meaningful analyses by complaint type (n=3). Finally, we observed 16,774
unique arrests associated with incidents generated by citizen calls for
service.
D. TREATMENT AND COMPARISON GEOGRAPHIC SITES
As noted, the primary methodological challenge in this study was
baseline imbalance. District 6, the treatment area, was discernibly different
from any other district in the city.56 Over the years there had been more
calls for service to the police, more arrests, more reported incidents of use
of force, and more complaints against officers in the treatment area than in
any other district. For this reason, the raw between-group measures were
insufficient and thus required adjustment to allow for a fair comparison.
One methodological option was to select a comparison district that was as
close as possible to the treatment district.57 However, there was no single
site, similar to the treatment district, in terms of all dependent variables (use
of force, complaints, or arrests). Similarly, a single comparison site might
also have been very different due to variability in terms of intra-district
organizational culture, the socio-demographic characteristics of the
neighborhoods that make up each district, and major policy changes in one
56 See infra Table 1.
57 See PREVENTING CRIME, supra note 47, at 4 (“Level 3. A comparison between two or
more comparable units of analysis, one with and one without the [treatment].”).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 741
area of the city versus another, that could intervene in the causal inference.
Given these comparability challenges—which were predominately due
to a lack of the rigorous controlled settings that characterize true
experiments (discussed below)—a more conservative test was selected, in
which the treatment district was compared to all other five districts in the
city, combined within a difference-in-differences test (DID).58 We were
particularly interested in the before–after variations in the outcome
variables while comparing these products between the intervention area and
comparison areas. Therefore, the use of BWCs was compared against the
mean scores of all other districts. This is likely to be referred to as a “Level
4 study” on the Maryland Scale, which is one level shy of the golden
standard of evaluation research: randomized controlled trials.59
E. APPARATUS
TASER Inc.60 provided Denver Police officers with BWCs. These
body-mounted cameras capture evidence from the officer’s perspective;
they were affixed to the collar, so were visible to those people who came
into contact with the police. Like the Rialto study, officers were not given
discretion as to when to use the BWCs, and every enforcement encounter
was required to be recorded.61 The officers were instructed to have the
58 See Jacqueline Cohen & Jens Ludwig, Policing Crime Guns, in EVALUATING GUN
POLICY: EFFECTS ON CRIME AND VIOLENCE (2003); Rafael Di Tella & Ernesto Schargrodsky,
Do Police Reduce Crime? Estimates Using the Allocations of Police Forces After a Terrorist
Attack, 94 AM. ECON. REV. 115 (2004); Stephen G. Donald & Kevin Lang, Inference with
Difference-in-Differences and Other Panel Data, 89 REV. ECON. & STAT. 221, 226 (2007).
59 See PREVENTING CRIME, supra note 47, at 5 (“Level 4. Comparison between multiple
units with and without the program, controlling for other factors, or using comparison units
that evidence only minor differences.”). See generally SHADISH et al., supra note 45, at 13;
David Weisburd et al., Randomized Experiments in Criminology and Criminal Justice, in
ENCYCLOPEDIA OF CRIMINOLOGY & CRIMINAL JUSTICE 4283, 4284 (2014) (“There are a wide
variety of competing methods in experimental research used to strengthen internal validity,
but randomized experiments are considered by many researchers to be the ‘gold standard.’”);
Barak Ariel, Deterrence and Moral Persuasion Effects on Corporate Tax Compliance:
Findings from a Randomized Controlled Trial, 50 CRIMINOLOGY 27, 28–29 (2012); Marcia
L. Meldrum, A Brief History of the Randomized Controlled Trial: From Oranges and
Lemons to the Gold Standard, 14 HEMATOLOGY/ONCOLOGY CLINICS N. AM. 745 (2000).
60 We Are TASER, TASER, INC., https://www.taser.com/company (last visited Dec. 30,
2015).
61 Ariel et al., supra note 4, at 520–21. If the officer applies discretion and deviates from
the protocol which dictates the videotaping of the encounter—for whatever reason—the
deterrent effect of the camera could be nil. It would convolute the study; lacking differences
between the experimental and control conditions (i.e., in both groups officers do not use
BWCs). In this case, officer discretion becomes an important issue. Indeed, “police work is
complex, . . . police use enormous discretion, discretion is at the core of police functioning.”
3. ARIEL 3/12/2017 4:53 PM
742 ARIEL [Vol. 106
cameras on during every shift and to announce to suspect, victims and
witnesses with whom they engaged that the interaction was videotaped. At
the end of every shift, officers in the treatment site were required to return
to their stations, and to download, tag, and register all the incidents during
which evidence was captured on tape.
Notably, once recording commenced, officers did not have discretion
nor the technical access to delete footage. The footage would automatically
be downloaded to a cloud-based server, with metadata tags for future
access. While officers have had access to their own recorded footage and
ability to input notes, they were not allowed to delete data.
IV. MEASURES
A. USE OF FORCE
The Denver Police Department ordinarily records use of force
incidents, like most U.S. police forces.62 This standardized tracking system
counted the reported incidents during the experimental period in all districts
in the city, during the experimental, as well as pretreatment, period. Similar
to the Rialto Experiment, we were less concerned with the types of force
responses, or the eligibility of the use of force incidents (excessive,
GEORGE L. KELLING, NAT’L INST. OF JUST., U.S. DEP’T OF JUST., “BROKEN WINDOWS” AND
POLICE DISCRETION 6 (1999) [hereinafter “BROKEN WINDOWS”].
Discretion is the essence that underpins policing by consent. Yet, to what extent we
should accord officers discretion when using BWCs is currently unclear, mainly because we
simply do not know when it is appropriate to use BWCs. Should BWCs be used in every
type of incident? When communicating with victims of domestic violence? What data
should be stored as evidence, and for how long should they be kept? At which point of the
interaction should the device be turned on? Therefore, at this stage of the evidence on
BWCs, discretion should not be given to officers. To illustrate, think of the following
scenario: a study on BWCs in which two police districts—one with and one without
BWCs—participate. Assume that the “treatment district” allows full discretion to its officers
on when or how to use the devices. Also assume that BWCs have a cooling off effect that
reduces the use of police force. It is not unlikely, particularly in critical incidents where the
devices would have an effect, officers will not turn the cameras on, or on time. There is risk
that activating a camera during a tense situation may serve to increase aggression in the
citizen/suspect (and thus the officer). Therefore, BWCs should be thought of as one
particular intervention for which on-scene discretion should be relinquished, and every
encounter save certain crime types should be videotaped from the very first moment of the
encounter. No-discretion is contained in police in-car and dash cams, as well as CCTV in
interrogation rooms.
62 See Geoffrey P. Alpert & Michael R. Smith, Police Use of Force Data: Where We Are
and Where We Should Be Going, 2 POLICE Q. 57, 57–58 (1999); see also Matthew J.
Hickman et al., Reliability of the Force Factor Method in Police Use of Force Research,
18 POLICE Q. 368, 369 (2015).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 743
disproportional, or illegal);63 we operationalized the “use of force”
dependent variable as whether force was used or not.64 We counted the
number of these use of force incidents in each district, during pretreatment
(baseline) as well as post-treatment periods.
B. CITIZEN COMPLAINTS
As in other forces, complaints filed against the police by citizens are
tracked by the Denver Police Department.65 These measures are often used
in policing studies to illustrate how officers adhere to internal rules of
conduct, as deviations from these regulations can potentially be construed
as signals of noncompliance.66 Here, as well, we counted the number of
complaints filed (as opposed to processed) against police officers in the
treatment, as well as comparison, districts, both before and after assignment
to treatment and control conditions.
To be sure, it is not entirely clear how to interpret a high or low
prevalence of complaints, and what, if anything, they represent.67 On the
one hand, complaints may be construed as signs of misbehavior. Whenever
63 Ariel et al., supra note 4, at 521–22.
64 Much can be said about the crudeness of this binary measure, which we are unable to
dissect in the limited scope of the present Article. The specific context and the causal
mechanisms in which force is used is still contested, including what party—officer or
suspect—initiated the forceful contact with the other party. See, e.g., Darren Henstock &
Barak Ariel, Testing the Effects of Body-Worn Video on Police Use-of-Force Beyond
Restraint in Arrests: A Randomised Controlled Trial (forthcoming 2016); see also Ariel et
al., supra note 4, at 521–22 (“[I]t is nearly always up to the individual officer to account for
those incidents where force was used. Given the subjectivity of this variable and the
measurement problems . . . we therefore relied on these official written reports, but not
without hesitation. Specifically in our study, our dependent variable only indicates whether
or not force was used, but it does not say ‘how much’ force was used. The ‘amount’ of force
used is also up to the officer to write down, as he or she recollects it.”). Indeed, BWCs may
provide some evidence, as the incident is recorded and can demonstrate, at least from the
officer’s perspective, who “threw the first punch,” why and under which conditions force is
more likely to occur in these encounters, however this qualitative assessment also goes
beyond the scope of the present article. Future research should look more closely at the
context of use of force in light of the use of BWCs, more granularly, by breaking down the
use of force variable into the different types of forces that can be used in police–public
contacts.
65 Andrew J. Goldsmith, What’s Wrong with Complaint Investigations? Dealing with
Difference Differently in Complaints Against Police, 15 CRIM. JUST. ETHICS 36 (1996).
66 Ariel et al., supra note 4, at 515; see, e.g., Anthony A. Braga et al., Pulling Levers
Focused Deterrence Strategies and the Prevention of Gun Homicide, 36 J. CRIM. JUST. 332,
335–40 (2008) (comparing incident reports of gun homicide from subject police department
with data from police departments in similar sized cities).
67 Ariel et al., supra note 4, at 515, 522.
3. ARIEL 3/12/2017 4:53 PM
744 ARIEL [Vol. 106
a party feels aggrieved, a way to restore these feelings of injustice is by
formally complaining about the conduct, aftermath, and overall attitude of
the officer, in light of what the complainant has perceived was unfair
treatment.68 In this respect, more complaints would be viewed as more
problems for the department to deal with, as a higher prevalence of reports
mirrors a higher prevalence of noncompliance with the rules that govern
officers’ behavior. On the other hand, more complaints could also be
interpreted as a “good” thing, because they suggest that citizens have not
“lost hope in the system” and accept that processes will be revisited,
officers reprimanded as necessary, and restitution will be made.69 While
this interpretation is not mutually exclusive of the first, it is nevertheless
difficult to ascertain what the equilibrium point would be, or the optimal
number of complaints that balances between the two.
There is, however, a third interpretation which must be recognized,
adding yet another layer of complexity: it is not too uncommon for some
members of the public to file a complaint against an officer in spite;
spuriously, frivolously, or, indeed, maliciously.70 Some complaints are filed
when the complainant is cognizant that the outcome of the interaction with
the police, the procedure in which it was conducted, or both, were in fact
legal, proportional, and “professional”; yet the citizen, nevertheless,
complains.71
Regardless of the motivation for lodging a complaint, and whatever
the interpretation of the court might be, it can be agreed that complaints are
costly;72 most departments would be only too happy to reduce them to a
68 See, e.g., MIKE MAGUIRE & CLAIRE CORBETT, A STUDY OF THE POLICE COMPLAINTS
SYSTEM 55–58 (1991) (discussing the various reasons and needs people have for
complaints); Ian Waters & Katie Brown, Police Complaints and the Complainants’
Experience, 40 BRIT. J. CRIMINOLOGY 617, 621–622 (2000) (discussing the Police
Complaints Commission and effects it had on alleviating the unfair feelings felt by
individuals).
69 Ariel et al., supra note 4, at 515; see MAGUIRE & CORBETT, supra note 68, at 57–58
(examining the reasons why people file police reports and how their complaint experience
impacts their perception of the police); see also Waters & Brown, supra note 68, at 633–35
(finding that the majority of complainants surveyed lost confidence in the police after filing
a complaint).
70 See Tim Prenzler et al., Complaints Against Police: The Complainants’ Experience,
1 J. CRIM. JUST. RES. 1 (2010); Waters & Brown, supra note 68, at 621–622.
71 John Liederbach et al., Is It an Inside Job? An Examination of Internal Affairs
Complaint Investigation Files and the Production of Nonsustained Findings, 18 CRIM. JUST.
POL’Y REV. 353, 370 (2007).
72 See Alan Ray Stafford, Lawsuits Against the Police: Reasons for the Proliferation of
Litigation in the Past Decade, 2 J. POL. & CRIM. PSYCHOL. 30 (1986); see also Goldsmith,
supra note 65.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 745
minimum.73 In the U.S. context especially, complaints can be quite
expensive, and, in large police departments, can engender multimillion-
dollar settlements, not to mention investigative, litigation, and
organizational expenses.74 Thus, any apparatus that can reduce not only the
number of complaints, per se, but also the reasons for lodging a complaint,
can be considered beneficial. Therefore, we used the data captured by the
department to count the number of complaints filed against police officers
as a main outcome of interest. We observed both the total number of
complaints for police misconduct and the total number of complaints filed
against use of force in the pre-treatment and post-treatment periods in all
districts.
C. ARRESTS
We observed the number of arrests (for any offense) made by the
officers in each district (again, before and after treatment assignment).
However, counting arrests is not a straightforward measure in respect to
how BWCs would affect its occurrence. First, arrest should not be
considered an outcome of police actions but, rather, an output, and as such
can change on the basis of local policies, investigative requirements, and
the perceived dangerousness of the suspect.75 The frequency of applying the
alternatives—verbal warnings, requests to meet the police at the station at a
later date, cautions, etc.—can, therefore, change as well based on local
customs, including intradistrict cultures.76
Second, it is not clear whether BWCs would reduce or increase arrest
records: on the one hand, in some instances where previously the police
officer would have applied alternatives to arrest, now the recording “of
everything” might increase an averseness to risk by the officer. Put
differently, officers who would otherwise close the case with a caution or a
warning might instead apply the most common denominator—arrest. Some
73 Id.
74 See, e.g., http://www.huffingtonpost.com/moneytips/largest-legal-settlements_b_
8122202.html.
75 See generally Douglas A. Smith & Christy A. Visher, Street-level Justice: Situational
Determinants of Police Arrest; Decisions, 29 SOC. PROBS. 167, 167–68, 173 (1981); Douglas
A. Smith et al., Equity and Discretionary Justice: The Influence of Race on Police Arrest
Decisions, 75 J. CRIM. L. & CRIMINOLOGY 234, 235 (1984).
76 William Terrill & Stephen D. Mastrofski, Situational and Officer-Based Determinants
of Police Coercion, JUST. Q. 215, 234–35 (2002) (describing a “control” independent factor
where the Indianapolis police “crackdowns and aggressive stops” differed from the St.
Petersburg “problem solving and community organizing” ways).
3. ARIEL 3/12/2017 4:53 PM
746 ARIEL [Vol. 106
refer to this phenomenon, albeit in a different context, as “net widening.”77
On the other hand, the use of BWCs could, instead, result in fewer
arrests. The need for arrest could potentially decrease, as a result of the
suspect presenting a less confrontational demeanor in response to being
videotaped. A major predictor of both use of force and, more broadly, arrest
is the way in which the suspect responds to the interaction with the
officer.78 Demeanor goes hand in hand with additional facilitators of the
decision to arrest: suspects’ resistance,79 legal cues,80 or disrespect.81
Therefore, if the suspect is non-confrontational, non-threatening, and
compliant, an arrest is less likely.82 BWCs are then hypothesized to have a
“cooling off” effect on the police–public interaction anyway,83 and,
therefore, one conceivable output is fewer arrests.
Furthermore, the decision to arrest may be altered when BWCs are
used as well. Officers’ administrative decisions to arrest often go
uncontested if executed properly. If the arrest follows guidelines, courts
would seldom interfere with the officer’s discretion and subsequently his or
her decision to bring a suspect into custody. In many ways and when
considering the alternatives, making an arrest is the easiest option when
dealing with suspects, especially when the “amount” of evidence necessary
to justify the arrest is actually very minimal.84 More importantly, when
77 Thomas G. Blomberg & Julie Mestre, Net-Widening, in I THE ENCYCLOPEDIA OF
THEORETICAL CRIMINOLOGY 573, 574–75 (J. Mitchell Miller ed., 2014).
78 See Michael D. Reisig et al., Suspect Disrespect Toward the Police, 21 JUST. Q. 241,
242, 262–64 (2004); Terrill & Mastrofski, supra note 76, at 219.
79 Terrill & Mastrofski, supra note 76, at 219.
80 Id. at 217.
81 Reisig et al., supra note 78. But cf. Robin Shepard Engel et al., Further Exploration of
the Demeanor Hypothesis: The Interaction Effects of Suspects’ Characteristics and
Demeanor on Police Behavior, 17 JUST. Q. 235, 237, 255–57 (2000) (finding no statistical
significance in demeanor).
82 Smith & Visher, supra note 75, at 169 (“The demeanor of the suspect also influences
police decisions to invoke the law. . . . [R]esearch indicates that antagonistic or hostile
suspects run a greater risk of being arrested”).
83 Ariel et al., supra note 4, at 517.
84 E.g., Lathers v. United States, 396 F.2d 524, 531 (5th Cir. 1968) (“An officer need not
be astronomically precise before making an arrest.”); Overstock Book Co. v. Barry, 305 F.
Supp. 842, 850 (E.D.N.Y. 1969) (“To arrest someone [sic], a police officer does not have to
determine first that the law under which he is acting will later be found to be constitutional
by a court.”); Joseph G. Cook, Probable Cause to Arrest, 24 VAND. L. REV. 317, 318–38
(1970) (laying out the various methods for proving probable cause to arrest in factual
scenarios); Caleb Foote, Fourth Amendment: Obstacle or Necessity in the Law of Arrest, 51
J. CRIM. L. CRIMINOLOGY & POLICE SCI. 402, 402–03 (1960); Wayne A. Logan, An
Exception Swallows a Rule: Police Authority to Search Incident to Arrest, 19 YALE L. &
POL’Y REV. 381, 391–400 (2001); Amanda L. Robinson & Meghan S. Chandek, The
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 747
aligning the word of the officer against the word of the suspect, we
axiomatically place more credence to the former rather than the latter.85 Can
this axiom now be challenged when BWCs are introduced? Would officers
think twice before arresting suspects and subsequently use this tool less
frequently when equipped with BWCs?
Ideally, we would observe all arrests during the experimental period.
However, as alluded to earlier, we observed arrest counts associated with
victim-generated calls for service, not arrests following proactive policing,
such as stop-and-account, crackdowns, or street stops. Arrests associated
with proactive policing are policy-dependent (see above), and the
intragroup variance—that is, differences between the conditions in the
treatment site and the comparison sites—was too great and cannot be
controlled statistically. On the other hand, we have made an assumption
that, broadly speaking, arrests following victim-generated calls for service
are less susceptible to proactive policies and the likelihood of an arrest
under no-treatment conditions are broadly stochastic. The decision to
exclude street checks and stop-and-accounts obviously dilutes the potential
magnitude of the effect but, nevertheless, clears out some of the statistical
noise associated with natural variations between the study conditions.
D. CITIZEN-INITIATED 911 CALLS FOR SERVICE
We measured the total calls for service initiated by the public (e.g.,
victims and witnesses) as opposed to events initiated by the police officers
themselves in each district, broken down into pretreatment and post-
treatment periods. This measure, however, was primarily used as a
stabilizing variable for the aggregated data: the number of 911 calls was not
the same across the districts, and this hinges on different workloads the
different geographic regions experienced. With this variable, we were able
to create more balanced groups with less heterogeneity between the groups
at baseline by measuring the outcome variables as a rate of the calls for
service in the district.
V. STATISTICAL PROCEDURE
We used adjusted odds ratios (OR) to assess the differences and to
compare the responses (Y: the outcome variables) according to the value of
Domestic Violence Arrest Decision: Examining Demographic, Attitudinal, and Situational
Variables, 46 CRIME & DELINQ. 18, 19, 32–33 (2000); William B. Waegel, Case
Routinization in Investigative Police Work, 28 SOC. PROBS. 263, 263 (1981).
85 Jennifer Groscup & Steven D. Penrod, Battle of the Standards for Experts in Criminal
Cases: Police v. Psychologists, 33 SETON HALL L. REV. 1141, 1148 n.32 (2002).
3. ARIEL 3/12/2017 4:53 PM
748 ARIEL [Vol. 106
the explanatory variable (X: BWCs or controls). With the aggregated data,
we had two binary variables that each had only two possible levels (counts
of observations at each level), displayed in a 2 x 2 contingency table. We
therefore used the total counts as the denominator (pretreatment + post-
treatment) and the number of events within the post-treatment period only
as the numerator, and computed the OR for each outcome variable.
However, as we were aware that the treatment group was not similar to the
control areas (it is ‘hotter’ than other places), our measures were stabilized
by incorporating the number of 911 calls for service each district
experienced and extracting the rate per 1,000,000 calls. We measured the
outcomes—use of force; complaints for “general” misconduct; complaints
against use of force; and arrests—for the pretreatment and post-treatment
periods for the treatment and control districts. We carried out this procedure
twice, once by comparing each comparison district with the treatment
district (five times) and again by comparing the pooled controls against the
treatment districts.86 We also measured 95% confidence intervals, in order
to assess the significance of these results.
Figure 1
87
/
/
/
/
VI. QUALITATIVE ANALYSES: OFFICERS’ SURVEYS
In order to qualitatively explore the effect of BWCs on police officers,
we conducted surveys in the treatment district. All 119 officers were
approached with a request to fill out an online questionnaire aimed to
understand how they viewed the use of BWCs, what impact the devices
may have on their sense of self-legitimacy, and whether or not they viewed
the devices positively or negatively. Notably, we did not conduct surveys
with officers in the non-treatment conditions, so the results of these surveys
are limited to one arm of the experiment only and, therefore, cannot be used
as measures of treatment outcomes. Still, the findings are informative.
86 We are unable to perform a meta-analysis on the data as combining all five
comparisons meta-analytically violates the assumption that the inter-group comparisons are
independent of one another (they cannot be as the treatment group is always the same). This
approach was not feasible; instead, the pooled sample of all controls was used to show the
overall ORs.
87 Per 1,000,000 calls.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 749
These main findings are more robust elsewhere;88 however, herein we
provide the responses of officers about two open-ended items. First, we
asked officers: “Is there something else you wish to say about your
expectations for BWCs in police work?” This open-ended question was
meant to allow officers to express their perceptions about how BWCs
would affect their performance and police work day-to-day. Second, we
asked officers to tell us whether “there [was] something else [they]
wish[ed] to say about [their] fears for BWCs in police work?” Here, as
well, we were keen to explore the officers’ perceptions about what they see
as threats to their work environment. Hereunder, we report the responses
associated with the effect of BWCs on officers’ decision to arrest.
VII. RESULTS
A. DESCRIPTIVE STATISTICS
Table 1 presents pretreatment and post-treatment raw figures on the
outcome variables. As shown, there were over 924,457 citizen-initiated 911
calls for service in the study period; however, the counts substantially
varied between the districts, with District 6 (treatment area) situated in the
upper-bound of this range. The area experienced more arrests—in absolute
terms—than any other area. Differences in these two indications—arrests
and calls for service—suggest different levels of activity that took place in
the treatment area compared to other areas and strengthen the need for
adjustments at baseline prior to any testing. Our two major outcomes of
interest—use of force and complaints against the police—were also at the
upper bound of the range of counts at baseline.
88 Barak Ariel & Justice Tankebe, Racial Stratification and Multiple Outcomes in Police
Stops and Searches, POLICING & SOCIETY (forthcoming 2016), currently available at
http://www.tandfonline.com/doi/full/10.1080/10439463.2016.1184270.
3. ARIEL 3/12/2017 4:53 PM
750 ARIEL [Vol. 106
Table 1
Before and After Raw Measures Between Treatment (District 6) and Five
Comparison Conditions
District
1
District
2
District
3
District
4
District
5
District
6
911 Calls for Service
Pre-treatment* 99,647 102,514 121,572 86,548 53,044 105,482
Post-treatment^ 56,471 60,650 76,848 55,393 34,889 71,399
Arrests
Pre-treatment 1,690 1,598 2,132 1,673 1,326 2,718
Post-treatment 854 782 1,053 995 679 1,274
Use of Force Incidents
Pre-treatment 101 118 135 157 58 210
Post-treatment 67 49 57 77 44 111
Misconduct Complaints
Pre-treatment 83 120 91 84 30 109
Post-treatment 44 63 64 41 33 82
Use of Force
Complaints
Pre-treatment 10 17 14 22 6 42
Post-treatment 18 18 23 11 8 34
^ 6 months of data (6/23/14–12/15/14); * Pre-Treatment (7/01/2013–6/23/2014).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 751
B. TREATMENT OUTCOMES
Use of Force
Looking at the pooled ORs in Table 2, the estimates suggest that the
odds of reporting of use of force by police officers in the treatment area did
not change significantly compared to controlled conditions (OR=0.93, 95%
CI 0.834–1.031). There is considerable variability between the different
districts—ranging between OR=0.67 in District 1 and OR=1.17 in District 3
(Q=98.98; p<0.001); yet, overall, the comparisons show no discernable
effect of BWCs in Denver on the likelihood of reporting of use of force.
Complaints Against the Police
When observing the overall fluctuations from the year prior to the
experimental period and the six months of the study, we can see that the
Table 2
Use of Force, Complaints, and Arrests: Treatment Area vs. Control Areas: Odds Ratios (OR)
Complaints –
Misconduct
Complaints – Use of
Force
Use of Force
Arrests
95% CI
95% CI
95% CI
95% CI
Groups
OR
Lower
Limit
Upper
Limit
OR
Lower
Limit
Upper
Limit
OR
Lower
Limit
Upper
Limit
OR
Lower
Limit
Upper
Limit
Dist. 1 1.188** 1.045 1.352 0.375*** 0.289 0.487 0.668*** 0.600 0.743 0.777*** 0.754 0.799
Dist. 2 1.253*** 1.113 1.410 0.668** 0.530 0.843 1.113† 0.995 1.244 0.837*** 0.812 0.863
Dist. 3 0.999 0.878 1.138 0.460*** 0.357 0.592 1.168* 1.042 1.310 0.886*** 0.861 0.913
Dist. 4 1.458*** 1.284 1.656 1.527*** 1.215 1.918 1.019 0.926 1.122 0.745*** 0.725 0.766
Dist. 5 0.665*** 0.582 0.760 0.590*** 0.454 0.767 0.677*** 0.610 0.752 0.889*** 0.866 0.914
Pooled
Controls
1.136* 1.001 1.288 0.650*** 0.512 0.827 0.928 0.834 1.031 0.820*** 0.797 0.844
* p ≤ .05; ** p ≤ .01; *** p ≤ .001; † =.061
3. ARIEL 3/12/2017 4:53 PM
752 ARIEL [Vol. 106
overall number of complaints (against both misconduct and use of force)
increased by 38%89—or 1105 at pretreatment compared to 1524 post-
treatment. These findings contradict the findings from the Rialto
Experiment, where complaints have declined by nearly 90%.90 Notably,
however, this result is not experimental, as it looks at the overall
fluctuations in complaints, including variations in the control districts. This
increase does not take into account the number of calls either.
When looking at between-groups outcomes, the odds of a complaint
for misconduct filed against officers in the control areas was 14% higher
than in the treatment area (OR=1.136, 95% CI 1.001–1.288). Variations
between the comparison groups emerged as well, with the odds of
complaint ranging from 0.67 in District 5 to 1.46 in District 4, to no
discernable difference between the odds for treatment District 6 and
District 3.
Overall, when looking at the specific subcategory of complaints filed
against the police for use of force, we find that the odds of a complaint in
control districts was 35% higher compared to the treatment district
(OR=0.650, 95% CI 0.512–0.827). The pattern was maintained in the
pairwise comparisons with Districts 1, 2, 3, and 5, but not with District 4,
where odds for complaints against use of force went up compared to the
treatment conditions (Table 2).
Arrests
The overall odds of an arrest was significantly reduced in treatment
conditions compared to controls (OR=0.820, 95% CI 0.797–0.844), and this
pattern emerged across all comparisons. The odds of an arrest was overall
about 18% higher in control conditions compared to treatment area.
Qualitative Outcomes
Officers expressed views that emphasized the causal estimates
described above. Two particular expectations shared by officers are
noteworthy. Speaking about how BWCs are perceived as mechanisms of
control over officers, one officer remarked that: “Big Brother eye in the sky
has just landed on our heads! . . . There is no more honor in this world, only
lawsuits.” Another shared this sentiment, further emphasizing that BWCs
89
We annualized the number of complaints and observed the ratio between the
pretreatment and post-treatment periods, or 1- ∑
2*npost
npre
.
90 Ariel et al., supra note 4, at 523–25.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 753
“will hinder officer initiated police work.”
More pronounced responses were expressed as fears that BWCs will
impede police work; many officers viewed BWCs as a means of taking
away their discretion and introducing enhanced liability for decisions they
would otherwise make more freely. For instance, one officer claimed that:
Arm chair quarterbacks will dissect every word and every action said and done in
milliseconds. Frame by Frame the officer will be criticized for acting, re acting or not
acting at all. The Youtube community, however, will have a grand ‘ol time bashing
police officers who are working hard to protect the civilian public they are sworn to
protect. Cops are human after all. What’s next, cameras in the cock pit because we
want to make sure the pilot doesn’t have a human moment. We put our lives in the
hands of pilots and doctors and I don’t see doctors or pilots wearing BWV’s, and if
they did, I would be a little disturbed about it. It’s all about private corporations
making money, period.
Another officer commented that: “I would like to be able to review the
footage of my interactions as many times as necessary for me without fear
of being asked to articulate why I did it.” And another officer wrote:
The current [BWCs] policy that has been written by this department gives the Internal
Affairs Bureau the ability to make arbitrary discretion . . . . [T]his seems like a useless
policy unless the goal of the department is to trap officers in false statements due to
differences of perception.
The ability to ‘forgo’ certain decisions was raised as well:
The standard for keeping/deleting video should be the same for police as it is for the
citizens: If anyone with a cell phone camera can delete footage they don’t like, so
should the police. Police should also have the option when to record their interactions
with the public (it shouldn’t be mandatory to record every contact/call).
Mistrust was another key dimension raised by officers: “I fear that the
use of this technology shows an increase in the trend of lack of trust in
police officers both by the public and by those in management. My word
and the fact that I already hold myself to a higher standard no longer
matter.”
Officers also expressed doubt about how footage would be viewed by
the public or line managers, stating “Police work can be messy. Having
actions reviewed from a seat after the fact can be too easily abused by the
public or administration,” and that “the general public may not understand
or like how officers at times have to talk or behave towards the criminal
element to gain their cooperation in following orders.”
3. ARIEL 3/12/2017 4:53 PM
754 ARIEL [Vol. 106
VIII. DISCUSSION AND CONCLUSIONS
The proliferation of the implementation of BWCs in the United States
continues.91 If it is of any indication, the share value of the leading
company in the production of these devices—TASER International Inc.—
has continued to soar since the results of the Rialto Experiment began to
surface.92 The political and public pressure to procure these devices will
continue unabated so long as they are pushed as a panacea to virtually all
issues with policing in the 21st century.93 However, even if they could be a
game-changer, the scope of rigorous evidence currently available is
insufficient to support such a claim. BWCs are promising, but under what
conditions they ‘work’ is still unclear. We must be cautious with our limited
public budgets, and we ought to be hesitant about great promises,
particularly when there is as much a likelihood of a backfiring effect to
using BWCs in police operations as there is to unmitigated success.94
Would BWCs increase the likelihood of use of force in some other forces?
Would police legitimacy be jeopardized? Will the testimony of officers be
conditional of recorded footage?95 In the current state of affairs, when
published evidence is severely lacking, we do not know. While there seems
to be great promise that BWCs can have a “cooling off effect” on police–
public engagements—owed specifically to deterrence theory96—more
research is needed, on as many issues as possible, with as many
jurisdictions as possible.
In this study, we aimed to expand our understanding of the effect of
BWCs in large police departments. The Denver Police Department is an
example of such a force. We complicated the Rialto Experiment findings
further, by adding an additional layers to the use of force and citizens’
91 Research on Body-Worn Cameras and Law Enforcement, NAT’L INST. JUST., DEP’T OF
JUST., http://www.nij.gov/topics/law-enforcement/technology/pages/body-worn-cameras.
aspx (last visited Dec. 30, 2015).
92 See TASER International, Inc. Stock Chart for Dec. 30, 2014 through Dec. 30, 2015,
NASDAQ, http://www.nasdaq.com/symbol/tasr/stock-chart (last visited Dec. 30, 2015). But
see id. (noting an overall decline in price since mid-July 2015).
93 Adam Bannister, Body-Worn Cameras: a Panacea for Policing Problems? Or a
Regulatory and Technological Minefield?, IFSEC GLOBAL.COM (June 27, 2014), http://
www.ifsecglobal.com/body-worn-camera-trials-can-police-surmount-regulatory-techno
logical-challenges/.
94 Ariel & Tankebe, supra note 88.
95 See Tim Cushing, Indiana Supreme Court Declares an Officer’s Testimony is More
Reliable Than Video Evidence, TECHDIRT (Apr. 10, 2014, 7:19 AM), https://www.techdirt.
com/articles/20140405/17142626817/indiana-supreme-court-declares-officers-testimony-is-
more-reliable-than-video-evidence.shtml.
96 See Ariel et al., supra note 4.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 755
complaints dimensions: first, we looked at additional outcomes—the
productivity of policing, measured by arrest counts. We believed there was
logic behind a causal mechanism which stipulates that variations in
productivity would follow the use of BWCs—except we did not know in
which direction: Would BWCs cause officers to “think twice” about the
need for an arrest (i.e., reduction in arrests), would BWCs cause officers to
terminate more encounters with an arrest, being a common denominator for
the risk-averse officer (increase in arrests), or will BWCs reduce the need to
make an arrest, as the interaction between the suspect and the officer would
be less aggressive, thus allowing the officer to continue with a criminal
justice outcome via alternative methods?
The second layer we explored was methodological in nature, and was
to change the unit of analysis from shift-based97 to large aggregated police
district level. This analytical approach looked at the adjusted odds ratio of
any one of these four outcomes (use of force, complaint against
misconduct, complaints on use of force, and arrests), in one district where
BWCs were deployed, in comparison with the odds of these events
occurring in the five other districts that served as controls.
Before we discuss the findings more robustly, we note that taking into
account the number of victim-generated calls for service in order to create
baseline equality between the groups is useful, but not perfect. This
procedure was necessary because the outcome variables at pretreatment
levels were inconsistent across the study conditions. We assumed that
victim-generated 911 calls are exogenous to police policies. As the volume
of calls dictates so much about policing, taking them into consideration was
assumed to take away some of intergroup variance—despite the fact that
variability was expected, given the natural differences that exist between
police districts in large metropolitan cities. Methodologically, this model
will always be susceptible to alternative explanations,98 as we can only rule
out alternative explanations to the causal inference asymptomatically.
Incorporating 911 calls may create some intergroup balance, but it is not
enough for making a relaxed assumption about baseline equilibrium. Future
studies might benefit from other statistical procedures, such as propensity
score matching techniques, to overcome baseline inequality. Another
97 Barak Ariel & William A. Farrar, Rialto Police Department Wearable Cameras
Experiment: A Criminological Protocol for Operating Randomized Trials, in UNIVERSITY OF
CAMBRIDGE RESEARCH: REGISTRY OF EXPERIMENTS IN POLICING STRATEGY AND TACTICS 5
(2012), http://www.crim.cam.ac.uk/research/experiments/rex-post/rialto .
98 PREVENTING CRIME, supra note 47; SHADISH ET AL., supra note 45; see Lawrence W.
Sherman & Dennis P. Rogan, Effects of Gun Seizures on Gun Violence: “Hot Spots” Patrol
in Kansas City, 12 JUST. Q. 689 (1995) (discussing the methodology).
3. ARIEL 3/12/2017 4:53 PM
756 ARIEL [Vol. 106
approach is to use street segments as the unit of analysis and measure the
treatment effect within comparable hotspots; hotspots are stable and
consistent, and can therefore be used to somewhat stabilize baseline
imbalances.99 More broadly, these design concerns illustrate the superiority
of randomized controlled trials (RCT) in the assessment of causal inference:
it is unlikely that a properly designed experiment, with sufficient statistical
power to detect significant effect sizes about the magnitude of the
difference between treatment and control conditions that were randomly
assigned, would face similar jeopardies to the internal validity of the test.100
Our Level 3, possibly 4 study, on the Maryland Scientific Scale faces
baseline imbalance and susceptibility to the effect of outliers (our degrees
of freedom are 5, if the unit of analysis is the geographic district). For this,
among other reasons, we strongly recommend future impact evaluations in
the area of BWCs to endorse RCT methodologies, as discussed more
thoroughly below.
Still, our findings are instructive. We discuss the implications of these
findings below, but begin with a summary of the findings. Overall, we
found that the likelihood of reporting a use of force incident is no different
when BWCs are in use or not; the odds of a misconduct complaint against
police officers when BWCs are present are 14% higher than controls, but
there are 35% greater odds to attract a complaint for the use of force when
BWCs are not present. Finally, the odds of any arrests went down by 18%
in the treatment area compared to control areas—which we sense is the
primary novel contribution this paper makes. We elaborate on these
findings in the following sections.
A. EFFECT OF BWCS ON USE OF FORCE: ACCOUNTABILITY AND
TRANSPARENCY
BWCs in Denver have had a non-significant overall effect on use of
force. However, compared to three out of the five comparison districts, the
odds of officers filing a formal report about the use of force rose
significantly by more than 15% when BWCs were used by officers in the
treatment conditions compared to control conditions. While these increases
are not uniform across all comparisons (odds of use of force went down
compared to two police districts in Denver, which created an overall non-
significant outcome), we find these seemingly counterintuitive increases
particularly important for future research. We interpret these findings
99 Barak Ariel, Increasing Cooperation with the Police Using Body-Worn Cameras, 19
POLICE Q. 326 (2016).
100 PREVENTING CRIME, supra note 47, at 4–5.
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 757
within the framework of accountability and transparency.
When officers use “force,” they are nearly always required to file an
official report of such incidents, even if only in their pocketbooks.101 The
challenge, however, is that not every physical action on the part of an
officer is considered force. The very definition can be subjective, memory-
prone, and generally unclear. Adams suggests that use of force occurs “at
least twice as often” as suggested by official reports,102 and it is likely to be
the case particularly in incidents of low-level force that do not amount to
anything the police officer feels he or she needs to account for. In addition,
some ethnographic work in this area suggests that what is construed as a
“reportable incident of force” and how much force is appropriate, is often
predicated by a police department’s organizational culture.103 For example,
police subculture in relation to the reporting of use of force plays a role in
accepting or allowing for force to be applied in certain circumstances.104
Researchers who study police organizations have been claiming for years
that use of force and subsequently its reporting are a function of police
officers’ attitudinal commitment to certain institutional or organizational
cultures around their roles in society and, more broadly, their view of
power.105 Certain institutional and subcultural codes make police agencies
particularly resistant to cultural changes and transparency requirements.106
Feelings of loyalty sustain this code of silence and make it particularly
difficult to investigate purported unnecessary, or excessive, use of force,
especially when it goes unrecorded.107 For example, placing one’s hand on
another’s shoulder in an authoritative way or using aggressive language
may be considered use of force in some instances and for some individuals,
whereas for others they may not.108 Measuring “injury” or “assault” is also
101 Henstock & Ariel, supra note 64, at 15.
102 Kenneth Adams, Measuring the Prevalence of Police Abuse of Force, in POLICE
VIOLENCE: UNDERSTANDING AND CONTROLLING POLICE ABUSE OF FORCE 52, 62 (William A.
Geller & Hans Toch eds., 1996).
103 Jennifer Hunt, Police Accounts of Normal Force, 13 J. CONTEMP. ETHNOGRAPHY 315
(1985); Jeff Rojek et al., Examining Officer and Citizen Accounts of Police Use-of-Force
Incidents, 58 CRIME & DELINQUENCY 301 (2010).
104 See Jerome H. Skolnick, Enduring Issues of Police Culture and Demographics, 18
POLICING & SOC’Y 35 (2008).
105 William Terrill et al., Neighborhood Context and Police Use of Force, 40 J. RES.
CRIME & DELINQ. 291, 292–93, 308–09 (2003).
106 See Skolnick, supra note 104, at 37 (“[The] unrecorded code [of silence] has been
noted as a feature of policing across continents, wherever commissions of inquiry have
investigated police corruption [or not].”).
107 See id.
108 See Henstock & Ariel, supra note 64.
3. ARIEL 3/12/2017 4:53 PM
758 ARIEL [Vol. 106
likely to be challenged in terms of definitional threshold, as it is open to
interpretation when there are no clear signs of physical contact. Taken
collectively, we see that what needs to be reported or not is not always as
clear-cut as it could be.
What does seem to be clear is that the reporting of use of force is
closely linked to police accountability and transparency. Sound reporting of
use of force is the cornerstone of police accountability. It is essential if
officers are to be held responsible for the actions, regardless of whether or
not their actions are justified.109 As reviewed by other scholars, police
accountability refers to taking responsibility for the actions of the
organization by tracking or measuring its outputs.110 This requirement
demands from the police to be accountable for its performance and to
amend it, when necessary.111 The police must act in the public’s interest,
and therefore is usually assumed to be held to a higher degree of
accountability112—especially given the wide powers they hold in modern
society. For this and other reasons, Samuel E. Walker contends that:
[T]he first accountability procedure to be considered involves the direction and
control officer use of police authority through formal agency policies. This approach,
generically known as administrative rulemaking, is a basic feature of modern police
management, if not all public and private sector organizations. Administrative
rulemaking consists of three elements: specifying approved and forbidden actions in
written policies; requiring officers to file written reports on specific actions; requiring
administrative review of officer reports.113
While deterrence theory would suggest that forceful encounters will be
minimized, it is also logical to see why, once BWCs are mandated in police
operations, reporting of the use of force could increase. Because the cost of
using force without reporting is invariably higher than the cost of the using
109 Stephen D. Mastrofski, The Romance of Police Leadership, in CRIME & SOCIAL
ORGANIZATION 153 (Elin J. Waring & David Weisburd eds., 2002); Prenzler et al., supra
note 70, at 15–16.
110 Mastrofski, supra note 109; Prenzler et al., supra note 70, at 15–16.
111 Wesley G. Skogan, Concern About Crime and Confidence in the Police Reassurance
or Accountability, 12 POLICE Q. 301, 308–13 (2009) (studying the different police and
confidence levels and concerns).
112 Philip Stenning, The Idea of the Political “Independence” of the Police: International
Interpretations and Experiences, in POLICE & GOV’T RELATIONS: WHO IS CALLING THE
SHOTS 183, 187–89, 191 (2007); Philip Stenning, Governance and Accountability in a Plural
Policing Environment—The Story So Far, 3 POLICING 22, 26 (2009).
113 SAMUEL E. WALKER & CAROL A. ARCHBOLD, THE NEW WORLD OF POLICE
ACCOUNTABILITY 94 (2014) (“Critical incidents are a crucial element of police
accountability tools, but if officers fail to complete required reports or do not provide
complete and accurate data the entire accountability system begins to collapse. . . . [T]here is
evidence that officers do not always file required reports.”).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 759
and subsequently reporting about “using force,” officers in the treatment
condition began filing these administrative reports at an increased rate
compared to some other districts, potentially recording “force” which they
otherwise would have not recorded: compliant handcuffs, hand-to-hand
techniques, and possibly, word commands. What once was left to ad hoc
explanations by officers who did not record “lesser” types of force114 can no
longer be hidden from the radar. BWCs caused officers to become more
accountable, because the odds of “getting caught” using force—now on
videotape—substantially increased, and by implication has caused officers
to file use-of-force records more frequently. Closer observations will be
required in the future about the types of forces that officers are now more
likely to report—data which we did not have access to in the present
study—however, we suspect these behaviors are situated in the lower bands
of the force continuum. Still, if our results are credible, they illustrate the
implication that BWCs have on police accountability and particularly
around the transparency and the reporting of use of force, which continues
to be a contentious area in policing.
Still, why has the Rialto Experiment reported a 50% reduction in the
reported use of force while the present study reports up to 17% increase in
the reported incidents of use of force against some comparison Districts?
We speculate that the discrepancies have to do with the research design and
our decision to exclude police-generated activities. In Rialto, the data
reflected all police actions that were recorded with BWCs, including stop
and search, street stops and various operations initiated by police
officers.115 When Rialto officers proactively engaged with suspects while
using BWCs, they have had far more control over the situation, as they
initiated the contact.116 On the other hand, victim-generated incidents are
more volatile and the severity of force is greater, as officers would reach
offenders in the aftermath of a crime, when the demeanor of suspects is
already adversarial or confrontational—compared to circumstances initiated
by the police.117 Given what we know about demeanor,118 it is likely that
114 Jeff Rojek et al., Policing Race: The Racial Stratification of Searches in Police
Traffic Stops, 50 CRIMINOLOGY 993, 995 (2012).
115 Barak Ariel & William Farrar, The Rialto Police Department Wearable Cameras
Experiment (Experimental Protocol), CRIM-PORT 1.0 (2009), http://www.crim.cam.ac.uk/
research/experiments/rex-post/rialto .
116 Ariel et al., supra note 4, 526–27.
117 Joel H. Garner et al., Characteristics Associated with the Prevalence and Severity of
Force Used by the Police, 19 JUST. Q. 705, 736 (2002) (“In comparison to routine
approaches, when an officer is responding to a priority call, more force is used.”).
118 John D. McCluskey et al., To Acquiesce or Rebel: Predicting Citizen Compliance
3. ARIEL 3/12/2017 4:53 PM
760 ARIEL [Vol. 106
officers in Denver applied force responses to already-heated incidents and
were now more compelled to file these incidents for the reasons reviewed
above. The officers were reactive to an already-existing situation, compared
to proactive policing when officers can manage and govern the interaction
more comprehensively. Future RCTs should review force responses more
closely in police-generated encounters and ascertain who instigates an
aggressive response against whom, whether any ecological factors are at
play in the exacerbation of force responses, such as large audiences (what
can be construed as “theater effects”),119 and to what extent alcohol and
drugs moderate the deterrent effect of BWCs on behavior. As for the latter,
we must be cognizant of the fact that deterrence theory relies heavily on
rational calculations and awareness;120 for this reason, intoxicated offenders
are unlikely to be responsive to deterrent messages or the credible threat of
punishment through their videotaped demeanor. Deterrence requires
rationality, and as intoxicated suspects are often chaotic, deterrence
embodied through BWCs is unlikely to work on them.
In this study, it was not possible to look at police-generated
incidents—stop-and-frisks, crackdowns, hotspot policing, etc.—and to
observe the effect of BWCs in these situations compared to the other
districts. The variations between the districts were too pronounced and
could not be controlled for statistically. A randomized controlled trial, with
random allocation of units into treatment and control conditions would have
created comparable groups in which these baseline differences in proactive
as well as reactive policing tactics can be controlled for.121 This is yet
another example of the advantages of using prospective RCTs compared to
any other methodology,122 and why experimental designs are stronger than
the alternative designs.
B. EFFECT OF BWCS ON COMPLAINTS: CONDITIONAL ON COMPLAINT
TYPE
As the number of complaints in Denver was substantially larger than
in Rialto,123 our experiment can go beyond the crude before–after analyses
with Police Requests, 2 POLICE Q. 389, 404–07 (1999) (discussing the impact demeanor and
citizen action has on the situation).
119 David Schweingruber, Mob Sociology and Escalated Force: Sociology’s
Contribution to Repressive Police Tactics, 41 SOC. Q. 371 (2000).
120 Ronald L. Akers, Rational Choice, Deterrence, and Social Learning Theory in
Criminology, 81 J. CRIM. L. & CRIMINOLOGY 653 (1990).
121 E.g., Ariel et al., supra note 4, at 520.
122 PREVENTING CRIME, supra note 47, at 3–5.
123 Ariel & Farrar, supra note 97, at 524 (noting there were only three recorded
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 761
conducted in the original experiment on the effect of BWCs on citizens’
complaints. We are able to report between-group analyses as well. We have
also complicated the story further by looking at the type of complaint filed
against Denver officers.
In the present study, citizens filed more complaints against misconduct
in the treatment district, which may include grievances against cursing,
disrespectful conduct, or what citizens might otherwise consider as police
maltreatment. One interpretation for this finding is that under no-camera
conditions suspects felt they cannot make a claim against officers (in other
words, “it is my word against his”), but with the evidence captured by the
cameras, it is reasonable to assume that citizens have felt they can
corroborate their claims against rude or uncivilized mannerism. Drawing
from the literature on accountability reviewed earlier, BWCs changed the
perceived degree of liability officers can now face and, perhaps, suspects
now feel that officers would be more accountable for their incivilities.
Future research should consider surveys with complainants in both
treatment and control conditions, in order to ascertain their motivation for
lodging a grievance report.
On the other hand, we observed a major reduction in the odds of a
complaint against the use of force, compared to control conditions. A lower
rate of complaints can be viewed as a marker of enhanced perceptions of
police legitimacy and satisfaction with police performance, and we
therefore interpret the significant reductions as increased legitimacy with
District 6 officers compared to control officers, in terms of the application
of force. Complaints allow researchers to assess the extent to which police
legitimacy is influenced by whether community members perceive police–
public encounters that they were treated fairly, with respect and dignity by
police officers.124 It is true that the link between the complaints and
satisfaction and police legitimacy overall is tenuous, but it is not
confounded, particularly when considering the instances in which
reductions in the odds of complaints were observed: use of police force.125
This conclusion deserves a closer look, particularly when we reflect on
the nondiscernable effect in official recordings of use of force made by
officers. One could argue that once BWCs were used, police officers
applied the same force (i.e., made officers not more and not less coercive
when responding to 911 calls for service) and that suspects complained less
complaints during the year of the study).
124 See, e.g., Anthony A. Braga et al., Losing Faith? Police, Black Churches, and the
Resurgence of Youth Violence in Boston, 6 OHIO ST. J. CRIM. L. 141, 172 (2008).
125 Id.
3. ARIEL 3/12/2017 4:53 PM
762 ARIEL [Vol. 106
about these incidents—however, we find this claim implausible. Instead, we
argue that police officers became more accountable for their use of force in
these instances, and likely in the lower manifestation of the force
continuum such as verbal (commands and threats) and lesser physical
restraint (pat downs or firm grips); and when coercion was applied, it was
both justified and proportional. The fact that the odds for a complaint for
the use of force were substantially lower when BWCs were used, leads us
to conclude that when coercion was applied, it was overall perceived as
more appropriate, otherwise the number of complaints for this category of
behavior would be indistinguishable from control conditions. This was not
the case; as others claim, BWCs serve a “cooling down” mechanism for the
use of force—which our data on complaints on the use of force suggest—
but also that these devices increase transparency and accountability.126 In
future studies, the type of force used should be observed as well in order to
see whether there is evidential merit in our stipulations. We can
nevertheless make a strong claim that the effect of BWCs on complaints is
conditional of the type of citizen complaint: an increase in complaints
against the use of force, and a decrease in complaints for misconduct. This
suggests to us that in future research we should not consider “citizen
complaints” as a homogenous signal of police performance and that more
granular analysis is required.
C. EFFECT OF BWCS ON ARREST DECISIONS
We found support for the claim that the use of BWCs had an effect on
arrests. Our findings suggest that the odds for an arrest were about 18%
higher under no-camera conditions. If these estimates are reliable, we can
conclude BWCs do not cause net-widening but rather a diversion of
encounters that might have led to arrest into alternatives to arrests. An
alternative possibility we entertained earlier was that risk-averse officers
who wear BWCs would be driven into the most-common denominator in
policing—arrest—rather than considering other means to close the case. We
raised the argument because arrests are “easier” than the alternative: they
are often valued positively by many officers as a way to bring offenders to
justice;127 they fulfill the need of many frontline officers’ vision about being
126 Ariel et al., supra note 4, at 517–18.
127 See Stephen Armeli et al., Perceived Organizational Support and Police
Performance: The Moderating Influence of Socioemotional Needs, 83 J. APPLIED PSYCHOL.
288, 293–95 (1998) (discussing similarities with employees helping the company and others
helping the police).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 763
crime-fighters;128 and they escalate the problem into the criminal justice
system, which is the classic progression from enforcement into prosecution.
However, our evidence did not support this position: arrests went down
when BWCs are in use, even though we excluded police-generated actions.
A discussion of the larger context of arrests is required in order to
better understand their link to BWCs. While variations in both law and
practice exist between different jurisdictions, there are only a very limited
number of scenarios in which a police officer can make an arrest: (a) when
he personally observes a crime; (b) when he possess “probable cause” to
believe the arrestee has committed a crime; (c) when he needs to subdue an
aggressive individual from hurting himself or others; and (d) when he has a
legal arrest warrant issued by the court.129 Clearly, an arrest cannot be made
because the officer has a vague hunch that the suspect is a criminal.130 Thus,
the officer must be able to justify the arrest, often by showing some
“tangible evidence” that led him to probable cause.131 If an arrestee believes
that the arrest was unjustified or incorrect, she may challenge it later, and if
warranted, bring a civil rights case.132 In practice, however, there is a low
burden of proof for police officers to justify their decisions to exercise the
power of arrest.133 Even the smallest subjective cues of resistance—
tampering with evidence, disrespecting the officer, and presenting as a
potentially dangerousness person—provide the necessary justification for
an arrest. This includes a broad list of crime categories that can justify the
legal requirements for probable cause and tangible evidence.134 Unlike
incidents of use of police force (Taser, batons, deadly force, etc.), which
has spurred a fairly elaborate body of case law,135 the decision to arrest—
128 JOAN C. BARKER, DANGER, DUTY, AND DISILLUSION: THE WORLDVIEW OF LOS
ANGELES POLICE OFFICERS 93 (1998).
129 See generally Logan, supra note 84.
130 BARKER, supra note 128, at 93.
131 Id. at 95.
132 See Donald A. Dripps, Criminal Procedure, Footnote Four, and the Theory of Public
Choice; or, Why Don’t Legislatures Give a Damn About the Rights of the Accused, 44
SYRACUSE L. REV. 1079, 1094–96 (1993).
133 Id.
134 Paul Bator & James Vorenberg, Arrest, Detention, Interrogation and the Right to
Counsel: Basic Problems and Possible Legislative Solutions, 66 COLUM. L. REV. 62, 64–67,
70–71 (1966); Myron Moskovitz, Road to Reason: Arizona v. Gant and the Search Incident
to Arrest Doctrine, 79 MISS. L.J.181, 186–88 (2009).
135 See Anthony G. Amsterdam, Supreme Court and the Rights of Suspects in Criminal
Cases, 45 N.Y.U. L. REV. 785 (1970); Mark S. Bruder, When Police Use Excessive Force:
Choosing a Constitutional Threshold of Liability in Justice v. Dennis, 62 ST. JOHN’S L. REV.
735, 741–42 (2012); Petter Gottschalk, Police Criminality and Neutralization: An Empirical
3. ARIEL 3/12/2017 4:53 PM
764 ARIEL [Vol. 106
although far more prevalent—is difficult to challenge, despite public
outcry. It is particularly the case when the distribution of arrests across
races and ethnicities is often argued to be uneven: not all races are arrested
equally.136
Given this framework, one possible interpretation is that BWCs have
an effect on police officers’ decisions to arrest suspects. With the
introduction of BWCs, officers became “cautious” about arresting suspects,
as their decision can more easily be criticized. When the camera records
what the officer views and hears, an arrest that does not pass the tangible
evidence test may be more easily detected. Self-consciousness of being
observed (by a BWC),137 coupled with the credible threat of apprehension
for violating rules and regulations associated with the wrong decision to
arrest,138 has significantly lowered the likelihood that officers would use
arrest.
If this mechanism sticks, what does it say about arrest decisions more
globally, or arrests made in control districts? One critical interpretation
might be that Denver police officers—and we suspect police officers in
general—use arrests far too frequently when many incidents could have
been handled through alternatives to arrest.139 When forced to “think twice”
about arresting suspects, officers are required to ask themselves: “Will I get
in trouble for my decision to arrest?” Under controlled conditions, the
decision to arrest is difficult to oppose and the decision to apply it often
goes uncontested. However, when BWCs are used, officers must provide a
more convincing level of evidence to corroborate their decision to arrest a
person.140 Merely stating “resisting arrest,” “violent demeanor,” or “known
suspect of aggressive behavior” is a necessary, but no longer sufficient,
condition to justifying an arrest. It is particularly the case for officers who
are “more forceful” in the arrest process than others. Collectively, it is
likely that officers in District 6 made fewer arrests because they should
have made fewer arrests in the first place and consider alternatives,
Study of Court Cases, 13 POLICE PRAC. & RES. 501 (2012).
136 Ariel & Tankebe, supra note 88.
137 Gervais & Norenzayan, supra note 34, at 299; Wicklund, supra note 33, at 237, 265.
138 JERVIS ET AL., supra note 41, at 3; Nagin, supra note 35, at 95.
139 Lawrence W. Sherman & Heather M. Harris, Increased Death Rates of Domestic
Violence Victims from Arresting vs. Warning Suspects in Milwaukee Domestic Violence
Experiment, 11 J. EXPERIMENTAL CRIMINOLOGY 1 (2014).
140 But cf. Robinson v. State, 5 N.E.3d 362, 367 (Ind. 2014) (holding the trial court’s
finding that an officer’s testimony was more reliable than video evidence was not an abuse
of discretion).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 765
arguably in a large number of cases.141 Future research should look more
closely at this possibility, potentially through surveys of officers about their
decision to make an arrest or not, in light of the use of BWCs.
There is, however, another plausible explanation, which looks more
closely at suspects’ demeanor and officers’ response to the behavior of the
suspect. Under this prism, arrests went down in the treatment district
compared to the control districts, as a result of using BWCs, because
suspects may have behaved differently when the cameras were pointing at
them, and that caused their behavioral modifications. Relying on a rather
convincing line of research on police–public encounters, a strong predictor
of arrest (especially in cases of a police request to cease misbehavior) is the
suspects’ demeanor.142 Therefore, if the suspect is nonconfrontational,
nonthreatening, and cooperative, an arrest is less likely. In similar ways, as
BWCs have a “cooling off” effect on police–public interactions anyway,143
not only is police force less likely to be needed, but the odds of an arrest
altogether dropped, too. The less the suspect aggressively confronts the
officer and resists, the less likely the officer needs to arrest the suspect. The
antecedent for the decision to arrests, therefore, lies within the suspect.
To be sure, we can only speculate on the reasons for the diminished
propensity for an arrest in the treatment group. We did not conduct
systematic observations or surveys with officers about their decision to
arrest in BWCs-present cases versus BWCs-not-present cases. However,
based on the qualitative data we have from the officers’ surveys, we are
drawn to conclude that the effect of BWCs is mostly on decisions. Officers
argued rather explicitly that they feared their choices would be audited and
reviewed if BWCs were introduced. The ability of line managers to analyze
every decision officers made concerned officers. While officers did not
disagree in principle with the idea of using BWCs in police operations, they
nevertheless expected that the footage would be used as an additional layer
of supervision and accountability. The immediate translation of these
perceptions was a behavioral modification, which manifested in fewer
arrests. Additional analyses and interviews with officers about these choices
seems like the next plausible step in this area of research.
141 See Lawrence W. Sherman & Heather M. Harris, Increased Homicide Victimization
of Suspects Arrested for Domestic Assault: A 23-Year Follow-Up of the Milwaukee Domestic
Violence Experiment, 9 J. EXPERIMENTAL CRIMINOLOGY 491, 510–11 (2013).
142 McCluskey et al., supra note 118, at 399–411 (discussing the relation to citizen
behavior in a police encounter).
143 Ariel et al., supra note 4, at 517–18.
3. ARIEL 3/12/2017 4:53 PM
766 ARIEL [Vol. 106
D. A CAUTIONARY NOTE ON NONEXPERIMENTAL DESIGNS IN FUTURE
STUDIES ON BWCS
This paper problematizes the kinds of routine activity and outcome
measures that other researchers will assemble to conduct similar studies in
the future. We suspect that in the next few years, the literature on BWCs
will increase dramatically, particularly when funding bodies have awarded
millions of dollars for both implementation and impact evaluation
studies.144 The projected dynamics alluded to in this study must be explored
more granularly in future research, which ought to pay closer attention to
design and method concerns. First, we alluded to the fact that police data
are not necessarily fit for the purpose, because reporting of police behavior
and police behavior per se are difficult to match, and can be viewed as a
form of a reflexivity effect. We rely on official statistics as outcomes and
output variables rather heavily, while some of these data points—recorded
use of force, decision to arrest, decision to prosecute, etc.—make any
causal study challenging to interpret: how should we understand these
interaction effects between behavior and reporting of behavior? What
additional measures are needed in order to accurately capture the effect of
BWCs on policing? Are “reductions” in reporting necessarily a beneficial
outcome, and what do they teach us about the complex and delicate
relationship between police and communities? These are difficult questions
to answer, which future studies will undoubtedly have to face when
interpreting their own results.
The second methodological challenge, beyond determining how to
appropriately measure results in policing studies, is identifying the
appropriate overall research design required to show the causal inference
between BWCs and various outcomes. Despite some recent critiques of
experimental methods,145 science simply does not have a better model for
causal inference. It is true that implementation of experimental methods
“narrow down” the scope of research and they are often quite difficult to
carry out, and thus alternative research methods are necessarily, by
definition, weaker designs for showing a causal effect. 146 Bluntly, when it
comes to studies of cause and effect, we have experiments, and then we
have everything else. Mixed methods should be ventured, particularly
144 See, e.g., NAT’L INST. OF JUST., supra note 91.
145 Greene, supra note 46; Robert J. Sampson, Gold Standard Myths: Observations on
the Experimental Turn in Quantitative Criminology, 26 J. QUANTITATIVE CRIMINOLOGY 489
(2010).
146 Lawrence W. Sherman, An Introduction to Experimental Criminology, in HANDBOOK
OF QUANTITATIVE CRIMINOLOGY 399–436 (Alex Piquero & David Weisburd eds., 2010).
3. ARIEL 3/12/2017 4:53 PM
2016] THE DENVER BWC EXPERIMENT 767
survey methodologies that can capture perceptions and decisionmaking
processes. However, the ideal scientific method we have for singling out a
cause of an effect is still by way of random assignment under controlled
conditions. Everything else is a compromise. We hope that empirical jurists
as well as social scientists become more aware of this important distinction
and seek to implement RCTs more frequently.
With this in mind, answering any question about the effect of BWCs
on any outcome will be particularly challenging without an RCT
methodology, as non-spurious relationships between independent and
dependent variables are difficult to attain. Without proper comparison
groups under controlled conditions, the BWCs’ treatment effect is easily
masked and susceptible to rival explanations. Only true experiments can
comfortably assume that baseline comparability between the intervention
arm of the study (whatever the intervention is) and the comparison arm, has
been achieved—which is the key scientific framework for observing the
causal estimate of the intervention.147 The multiple comparisons in the form
of several police districts in the present study strengthen the internal
validity of the present study, yet future BWCs studies should be cognizant
that the most credible findings will come from multiple or multisite trials,
ideally utilizing cluster randomized controlled trials. Otherwise, it will be
challenging to adequately evaluate the effect of BWCs deployment, and
especially whether in fact they increase desired outcomes, relative to their
cost. Failing that, our knowledge on these fast-entering devices will be
handicapped.
CONCLUSION
The Denver Police Department has shown that deploying BWCs in
one police district caused a significant 35% lower odds for citizens’
complaints against the police use of force, but 14% greater odds for a
complaint against misconduct, compared to other Denver districts that did
not deploy BWCs. The analyses further suggest no discernable effect on use
of force in the aggregate compared to control conditions but suggests
increases against some districts, which are contextualized as enhanced
transparency and accountability as a result of deploying BWCs. Finally, the
odds for arrest were 18% lower than the odds under control conditions,
suggesting that officers become “cautious” about arresting suspects when
BWCs are deployed. Still, methodological challenges of the present study
clearly suggest that more research on these outcomes is needed, using
147
SHADISH ET AL., supra note 45.
3. ARIEL 3/12/2017 4:53 PM
768 ARIEL [Vol. 106
randomized controlled trials.
Copyright of Journal of Criminal Law & Criminology is the property of Northwestern
University School of Law and its content may not be copied or emailed to multiple sites or
posted to a listserv without the copyright holder’s express written permission. However, users
may print, download, or email articles for individual use.