2

1

Article

Exploring the Potential for Body-Worn
Cameras to Reduce Violence in
Police–Citizen Encounters
Michael D. White*, Janne E. Gaub** and Natalie Todak***

Abstract One of the most compelling perceived benefits of body-worn cameras (BWCs) involves the potential for
reductions in citizen complaints and police use of force. A handful of early studies reported significant reductions in

both outcomes following BWC adoption, but several recent studies have failed to document such effects. The current

study explores this question using data from a randomized controlled trial conducted in the Spokane (WA) Police

Department. Approximately half of patrol officers (n = 82) were assigned BWCs in May 2015, while the other half
(n = 67) received their BWCs 6 months later (November 2015). The study explores the effects of BWCs on use of force,
complaints against officers, and officer injuries, using more than three years of official department data pre- and post-

BWC deployment. The outcomes of interest are rare in Spokane, which limited both statistical power and the results

from significance testing. However, the within-group trends are consistent with a positive effect, particularly for

percent change. Following BWC deployment, the percentage of officers with a complaint in each group declined by

50% and 78% (Control and Treatment, respectively); the percentage of officers with a use of force declined notably

(39%) for one group only. The reductions disappeared after 6 months for the Treatment group. There was no

relationship between BWCs and officer injuries. The authors discuss the implications of the findings for the ongoing

dialogue on BWCs.

Introduction
that occurred during the 1960s (Kerner Commis-

A persistent undercurrent of racial tension has been sion, 1968, p. 157). Fifty years later, the final report

one of the defining features of American law en- of the President’s Task Force on 21st Century

forcement over much of the last century (White Policing (2015, p. 5) again pointed to low reserves

and Fradella, 2016). In 1968, the National of trust and police legitimacy in minority commu-

Advisory Commission on Civil Disorders con- nities as the explanation for civil unrest following

cluded ‘deep hostility between police and ghetto police killings of citizens in Ferguson (MO),

communities’ was a primary cause of the riots Baltimore (MD), and other cities across the US:

*Arizona State University, School of Criminology and Criminal Justice, Phoenix, AZ, USA. E-mail: [email protected]
**Arizona State University, Center for Violence Prevention and Community Safety, Phoenix, AZ, USA
***University of Alabama, Birmingham, Department of Justice Sciences, Birmingham, AL, USA

Policing, pp. 1–11
doi:10.1093/police/paw057
© The Author 2017. Published by Oxford University Press. All rights reserved.
For permissions please e-mail: [email protected]

Deleted Text: "

Deleted Text: "

mailto:[email protected]

mailto:[email protected]

2 Policing Article M. D. White et al.

‘In establishing the task force, the President spoke

of the distrust that exists between too many police

departments and too many communities—the

sense that in a country where our basic principle

is equality under the law, too many individuals,

particularly young people of color, do not feel as

if they are being treated fairly.’

The Task Force (2015) final report identified

nearly 60 recommendations for building trust be­

tween police and citizens, and body-worn cameras

(BWCs) are highlighted as a tool for achieving that

objective. Since 2015, the White House and the US

Department of Justice have strongly promoted the

adoption of BWCs by police, as evidenced by the

creation of a National Body-Worn Camera Toolkit

(Bureau of Justice Assistance, 2016a), a federal

funding program that has provided $40 million to

more than 175 law enforcement agencies for the

purchase of BWCs (Department of Justice, 2016),

and a training and technical assistance mechanism

that facilitates BWC adoption and program man­

agement (Bureau of Justice Assistance, 2016b).

Proponents of BWCs have made numerous claims

regarding the benefits of the technology, including

that BWCs can reduce violence during police-citizen

encounters (White, 2014).
1

A number of early studies

reported significant reductions in citizen complaints

against officers and police use of force following de­

ployment of BWCs, suggesting the technology can

produce measurable change in these two important

outcomes. An evaluation of BWCs in the Rialto (CA)

Police Department documented a nearly 90% drop

in citizen complaints against police, and a 60% de­

cline in use of force by officers (Ariel et al., 2015).

Similarly, positive results have emerged from studies

in Mesa (AZ; Mesa Police Department, 2013),

Orlando (FL; Jennings et al., 2015), and Tampa
(FL; Sullivan and Marrero, 2016).

2
Hedberg and col­

leagues (2016) estimated the effect of BWCs on citi­

zen complaints in Phoenix (AZ) and concluded ‘if

BWCs are employed as prescribed [i.e., 100 percent

activation compliance], a majority of complaints

against officers would be eliminated’ (p. 16).

However, several recent studies have failed to

document positive effects on citizen complaints

and use of force. A study by the Edmonton Police

Service (2015) concluded BWCs had no measurable

impact on either outcome. Grossmith et al. (2015)
found a statistically significant decline in citizen

complaints in only two of the ten London police

boroughs examined.
3

Ariel and colleagues (2016c,

p. 2) described findings from ten BWC studies and

concluded the technology ‘had no effect on use of

force’ overall, but the null finding was explained by

mixed results across studies.
4

Ariel et al. (2016c)
also found a troubling link between BWCs and

increased rates of assaults on officers.
5

Ariel et al.
(2016b) tied patterns in use of force to officer de­

cisions on BWC activation. That is, when officers

followed policy—they activated the BWC at the

start of citizen encounters and advised citizens of

the BWC—use of force declined by 37%. When

officers did not follow policy, use of force actually

increased by 71%.

The mixed findings on BWCs and violence in

police–citizen encounters suggest the dynamics at

play may be considerably more complex than ori­

ginally described by advocates of the technology.
6

Research has explored a number of other potential benefits and limitations associated with BWCs. Due to space constraints,
the authors focus on the outcomes most relevant for the current study: use of force, complaints, and officer injuries. See
White (2014) and the National Body-Worn Camera Toolkit (Bureau of Justice Assistance, 2016a) for additional discussion of
other benefits and challenges associated with BWCs.
2

For additional studies reporting reductions in complaints and use of force see: Goodall (2007); Ellis et al. (2015).
3

Six of the ten boroughs did experience fewer complaints but only two reached statistical significance, and the effect across all
boroughs was also not significant (Grossmith et al., 2015).
4

Researchers randomized shifts rather than officers (Ariel et al., 2016c).
5

To our knowledge, Ariel et al. (2016c) is the only published study to examine this important outcome.
6

We use ‘violence’ as a general term that captures aggression and combativeness in police–citizen encounters, and we treat
use of force, complaints against officers, and officer injuries as indicators of violence.

1

Deleted Text: "

Deleted Text: "

Deleted Text: percent

Deleted Text: percent

Deleted Text: "

Deleted Text: "

Deleted Text: g

Deleted Text: "

Deleted Text: "

Deleted Text: —

Deleted Text: —

Deleted Text: –

3 Potential for Body-Worn Cameras Article Policing

The inconsistent results are especially troubling

given the rapid diffusion of BWCs in law enforce­

ment and the potentially severe, longstanding con­

sequences of violence in police–citizen encounters

(Fyfe, 1988; President’s Task Force on 21st Century

Policing, 2015; White and Klinger, 2012).

Moreover, the potential link between BWCs and

increased assaults on officers (Ariel et al., 2016c)
warrants immediate attention from researchers.

The current study explores these questions using

data from a randomized controlled trial in

Spokane, Washington.

Methods and data

The current study is part of a larger project exam­

ining the impact and consequences of BWCs. In

early 2015, the Spokane Police Department (SPD)

leadership devised a plan for a staggered rollout of

BWCs to all patrol officers in two phases (n = 149).7

The leadership worked with the authors to randomize

the process by which officers were selected for the first

(May 2015; Treatment group [n = 82]) and second
(November 2015; Control group [n = 67]) phases of
the deployment.

8
The officers in each group received

the TASER Axon Body 1 camera
9

on a rolling sched­

ule, as groups of officers were trained on consecutive

Fridays during the two deployment months.
10

The

authors compared both groups across officer demo­

graphics, rank, years of service, and pre-RCT rates of

use of force and complaints. No differences reached

statistical significance (Table 1).

The SPD policy directs officers to record any law

enforcement activity, including self-initiated citi­

zen contacts, and to continue recording until the

interaction or activity concludes.
11

Officers are

given discretion to not record if doing so would

jeopardize safety and/or the ability to perform

7
SPD implemented a small-scale pilot study of BWCs in fall 2014, involving approximately 20 volunteer officers. This pilot

study occurred prior to the authors’ collaboration with the agency, though most of the pilot study officers continued to wear
their BWCs up to the start of the RCT. Those volunteers who opted to stop wearing the BWC largely cited their state’s law
governing public records requests (i.e. very liberal with regard to access), and the lack of a clear department policy at the time,
as reasons for withdrawing from the pilot.
8

SPD provided the authors with a complete list of all officers, corporals, and sergeants assigned to patrol. Officers were
randomly assigned by the authors to either the Treatment or Control group using the random number generator in Microsoft
Excel. There were 12 departures from random assignment (8% departure rate). Eight of the randomization departures
involved officers who participated in the department’s BWC pilot study prior to the phased rollout, were randomly assigned
to the Control group, but asked to keep their BWCs. The department leadership and authors agreed to re-assign those officers
to the Treatment group. The remaining departures occurred as a result of officers missing their assigned BWC training
because of injury, family leave, vacation, or similar reasons. Four officers were removed from the study because they retired or
transferred to a non-patrol assignment during the RCT period (and were no longer assigned a BWC).
9

This model features a 30 second ‘buffer’, wherein the camera continually records video (without audio) for the 30 s prior to
camera activation.
10

The BWC training was included as part of an 8-h required use of force report writing training. Training was completed by
the agency’s academy instructors, and occurred in two parts. The first part consisted of classroom-based instruction, which
focused on laws and policy governing use of the cameras. Officers were also trained in BWC operational use. The second part
consisted of scenario-based training, in which officers participated in mock citizen interactions and use of force scenarios
while wearing the BWC. Following the mock scenarios, officers learned to complete reports incorporating the video evidence.
One of the authors observed the BWC training on several occasions. For Treatment officers, the RCT began on the day they
received a camera in the training course in May 2015. All activity that occurred prior to that training day is considered
pre-RCT activity. The same principle applies for the Control officers who received their BWC on a rolling schedule in
November 2015.
11

When a video is recorded, officers are instructed to label it using their mobile devices. At the end of each shift, they are
required to dock their cameras, during which time all videos are automatically uploaded to TASER’s cloud-based storage
system, Evidence.com. Officers do not have access to manipulate the video in any way on either their mobile devices or on
Evidence.com. All activity (viewing, tagging, notations, etc.) is documented in the audit trail on Evidence.com and cannot be
altered.

Deleted Text: –

Deleted Text: White & Klinger, 2012

Deleted Text: D

Deleted Text: see

http:Evidence.com

http:Evidence.com

http:Evidence.com

http:concludes.11

http:months.10

4 Policing Article M. D. White et al.

Table 1: Characteristics of study officers, by group

Treatment group (n = 82) Control group (n = 67) Total (n = 149)

Male (%) 91.5 83.6 87.9

White (%) 93.9 95.5 94.6

Rank (%)

Officer 80.5 76.1 78.5

Corporal 9.8 10.4 10.1

Sergeant 9.8 13.4 11.4

Years of service Mean=12.77 SD = 7.87 Mean= 14.30 SD = 7.03 Mean=13.46 SD = 7.52

Monthly use of forcea Mean=0.97 SD = 0.44 Mean=1.06 SD = 0.49 Mean=1.02 SD=0.46

Monthly complaintsa Mean=0.32 SD = 0.38 Mean=0.40 SD = 0.43 Mean=0.36 SD = 0.41

a
Monthly use of force and complaints were standardized per 1,000 calls for service, per group.

their law enforcement duties.
12

The body camera

does not visually indicate to citizens that they are

being recorded, and the agency’s policy does not

require that officers notify citizens that they are

being recorded. In a separate study, the authors

interviewed 249 Spokane citizens who had BWC-

recorded encounters with police officers, and only

28.5% were aware they had been recorded (White et
al., forthcoming).

The SPD provided officer-level measures of use

of force, internal complaints, citizen complaints,

and officer injuries, from 1 January 2013 through

30 April 2016 for all 149 officers in the study.
13

This

period includes 28 months pre-RCT (January

2013–April 2015), 6 months of the RCT (May

2015–October 2015), and 6 months post-RCT

(where both groups have BWCs; November 2015–

April 2016)—for a total study period of 40 months.

The authors calculated standardized monthly rates

of each measure by dividing the monthly outcome

total for each officer group by the number of calls

responded to by each group, and multiplying by

1,000 (e.g. for each group, monthly rate of

force = [# force incidents per month/# calls per

month] * 1,000). Call activity includes both citizen
and officer-initiated calls. Several sets of analyses

were carried out. First, the authors descriptively

examine long-term trends in each outcome meas­

ure with the two officer groups to assess general

change over time. We also employ difference-in­

difference (DID) estimations to test variation in

outcomes among the groups over time (DID pro­

vides a fixed-effect estimation of the intervention’s

impact on both groups). The authors then con­

ducted a more focused analysis of monthly

change by comparing outcomes during 6-month

intervals, beginning in January 2013 and including

the pre-RCT (11/14 – 4/15), RCT (5/15 – 10/15),

and post-RCT (11/15 – 4/16) periods. Independent

and paired-sample t-tests are employed to compare

within- and between-group change across the 6­

month time periods.
14

The authors also examine

change in the percent of each group recording an

event (force, complaint) during the pre-RCT, RCT,

and post-RCT periods. The authors conducted

power analysis with GPower and the results

Data regarding activation compliance is not yet available. The authors are currently working with SPD to gather meta-data
from Evidence.com, which will be analysed in conjunction with call data. The department’s CAD/RMS system and
Evidence.com are not integrated, and as a consequence, the activation compliance analysis is very labor intensive.
13

Use of force and complaint data were provided by the department’s Internal Affairs unit. The officer injury variable was
captured from the department’s official use of force reports, which are publicly available on the SPD website. In short, the
authors applied the department’s official definitions of use force, complaints, and officer injuries.
14

All of the analyses are conducted with the 12 randomization departures remaining in their final group. The authors also
conducted the analysis with the 12 departures removed from the study (n = 137). Those analyses are not presented here given
space constraints, but there were no differences in the findings using this alternate approach.

12

Deleted Text: percent

Deleted Text: , Todak, and Gaub

Deleted Text: 30,

Deleted Text:

Deleted Text:

Deleted Text: six

Deleted Text:

Deleted Text:

Deleted Text: six

Deleted Text:

Deleted Text:

Deleted Text: —

Deleted Text: ,

Deleted Text:

Deleted Text: s

Deleted Text: six

Deleted Text: six

http:Evidence.com

http:Evidence.com

http:periods.14

http:study.13

http:duties.12

http:Mean=0.36

http:Mean=0.40

http:Mean=0.32

http:Mean=1.02

http:Mean=1.06

http:Mean=0.97

http:Mean=13.46

http:Mean=12.77

5 Potential for Body-Worn Cameras Article Policing

Figure 1: Use of force rates by officer group, January 2013–April 2016.

indicate weak statistical power across the outcomes

(force [0.28], complaints [0.07], officer injuries

[0.15]) because of low base rates.

Results

Use of force

Figure 1 shows monthly use of force rates by officer

group, standardized by call activity, with vertical

lines representing the start of the first (May 2015)

and second phases (November 2015) of BWC roll-

out. Use of force by police is an uncommon event

(both groups average about one use of force inci­

dent per month, per 1,000 calls). The standardized

trend over the entire study period is relatively flat

for both officer groups, ranging between 0.5 and 2.0

incidents per month. Table 2 shows the DID esti­

mates were not significant. Table 2 also shows mean

use of force rates between groups during 6-month

intervals periods, with a specific focus on the pre-

RCT (11/14 – 4/15), RCT (5/15 – 10/15), and post-

RCT (11/15 – 4/16) periods. None of the within- or

between-group differences reach statistical signifi­

cance (e.g. t-test results), and patterns in use of

force are inconsistent over time (e.g. a notable

spike in 5/13 – 10/13). There are some interesting

trends in the pre-RCT, RCT, and post-RCT periods.

Use of force by the Treatment group declined by 8%

following BWC deployment (0.91 to 0.84), despite

stable call activity.
15

During that same time, use of

force among the Control group increased by 17%.

Once the Control group was assigned BWCs, their

use of force declined by nearly 50% (1.07 to 0.60).
16

Finally, the decline in use of force was temporary for

the Treatment group, as their post-RCT use of force

increased by 27 percent (from 0.84 to 1.18). Table 3

shows these trends in terms of the percent of each

group with a use of force during the pre-RCT, RCT,

and post-RCT periods. For example, the percentage

of the Treatment group with a use of force remained

relatively flat over all three periods (from 24.4 percent

to 28.0 percent), but the percentage of the Control

group with a use of force (pre-post BWC deploy­

ment) declined by 39%, from 26.9% to 16.4%.

Complaints

Figure 2 shows the standardized rates of complaints

against officers.
17

Complaints rose steadily in the

15
The number of calls for the Treatment group declined by one percent from pre-RCT (38,270) to RCT (37,891) periods.

16
During this time, call activity declined by 12 percent for the Control group (from 30,332 RCT to 26,762 post-RCT).

17
Given the very low rate of complaints per month, the authors merged citizen and internal complaints into one measure.

Deleted Text: F

Deleted Text: six

Deleted Text: ,

Deleted Text: ,

Deleted Text: percent

Deleted Text: percent

Deleted Text: percent

Deleted Text: percent

Deleted Text: percent

Deleted Text: percent

http:officers.17

http:0.60).16

http:activity.15

6 Policing Article M. D. White et al.

Table 2: Mean outcomes and difference-in-difference estimations by officer group

Pre-RCT RCT Post-RCT Difference-in­
1/13 � 5/13 � 11/13 � 5/14 � 11/14 � 5/15 � 11/15 � Difference
4/13 10/13 4/14 10/14 4/15 10/15 4/16 Coeff. (SE)**
Rate (n) Rate (n) Rate (n) Rate (n) Rate (n)* Rate (n)* Rate (n)*

Use of force

Control 0.84 (17) 1.54 (54) 1.17 (37) 0.77 (26) 0.92 (28) 1.07 (33) 0.60 (16) 0.27

Treatment 0.96 (23) 1.29 (49) 0.70 (26) 1.01 (41) 0.91 (36) 0.84 (33) 1.18 (42) (0.26)

Complaints

Control 0.59 (12) 0.95 (33) 0.21 (7) 0.08 (3) 0.23 (7) 0.28 (8) 0.15 (4) -0.004

Treatment 0.26 (6) 0.67 (26) 0.24 (9) 0.15 (6) 0.24 (9) 0.05 (2) 0.19 (7) (0.13)

Officer injuries

Control 0.10 (2) 0.17 (6) 0.03 (1) 0.00 (0) 0.10 (3) 0.04 (1) 0.07 (2) -0.02

Treatment 0.08 (2) 0.14 (5) 0.07 (3) 0.15 (6) 0.02 (1) 0.05 (2) 0.05 (2) (0.05)

* None of the within- and between-group mean differences reach statistical significance (P < 0.05). ** None of the difference-in-difference estimations reach statistical significance (P < 0.05). Table 3: Group percentages of complaints and use of force Pre-RCT RCT Post-RCT 11/14 � 4/15 % (n) 5/15 � 10/15 % (n) 11/15 � 4/16 % (n) Use of force Control 28.4 (19) 26.9 (18) 16.4 (11) Treatment 24.4 (20) 25.6 (21) 28.0 (23) Complaints Control 10.4 (7) 9.0 (6) 4.5 (3) Treatment 11.0 (9) 2.4 (2) 6.1 (5) first part of 2013 for both officer groups before dropping substantially and remaining at a low rate throughout the rest of the study period. The rate of complaints becomes near-zero after the start of the RCT and rarely moves above 0.5 for either group. The DID estimates were not significant (see Table 2). The within- and between-group differ­ ences in the 6-month intervals are not statistically significant—though again there are post-BWC de­ clines for both groups. For the Treatment group, Table 2 shows a nearly 80% drop in complaints, from 0.24 pre-RCT to 0.05 RCT (from 9 to 2). Post-BWC deployment, complaints for the Control group drop by nearly 50%, from 0.28 (8) to 0.15 (4). Again, there is an increase in complaints among the Treatment group during the post-RCT period (from 0.05 [2] to 0.19 [7]). Table 3 shows the percentage of the Treatment group with a com­ plaint decreased from 11.0% (pre-RCT) to 2.4% (RCT) – or a 78 percent decline. After the Control group was assigned BWCs, the percentage of the group with a complaint dropped from 9.0% to 4.5% (percent change = 50%). Officer injuries Figure 3 shows officer injuries that occurred during use of force incidents, again standardized by monthly group call activity. 18 Officer injuries are 18 The injury data only includes incidents involving the 149 officers in the study. Injuries of officers who are not in the study are excluded, as are injuries that did not result from a police–citizen encounter (e.g. off-duty; traffic accident). Deleted Text: Deleted Text: six Deleted Text: -- Deleted Text: percent Deleted Text: percent Deleted Text: percent Deleted Text: percent Deleted Text: percent Deleted Text: percent Deleted Text: percent Deleted Text: I http:activity.18 7 Potential for Body-Worn Cameras Article Policing Figure 2: Complaint rates by officer group, January 2013 – April 2016. Figure 3: Officer injury rates by officer group, January 2013–April 2016. extremely rare, and their prevalence is not affected by BWC deployment. Table 2 shows no statistically significant changes in officer injuries over time. Discussion Several important themes emerged from the cur­ rent study’s results. First, the outcomes of interest are rare. Both officer groups averaged about one use of force per month per 1,000 calls. Complaints and injuries were even less common. Use of force and complaints against officers are typical outcomes in police research, and low base rates are a common issue. For example, prior research has consistently shown that police use of force occurs in less than 2% of all police citizen encounters (Hickman et al., 2008). Moreover, many of the most influential BWC studies have examined small police Deleted Text: two percent 8 Policing Article M. D. White et al. departments (Rialto, CA) or large departments with limited BWC deployment (Phoenix, AZ), which tends to compound the low base rate issue. 19 The second theme involves the noteworthy though nonsignificant declines in outcomes follow­ ing BWC deployment. Though statistical power was weak, the outcomes clearly trended in a positive direction. For example, complaints declined when each officer group was assigned BWCs. Use of force followed a similar pattern after BWC deployment for the Control group. The positive trends are per­ haps best captured in the percentage of each group with an event, before and after BWC deployment: the percentage of the Treatment and Control groups with a complaint declined by 78 percent and 50%, respectively; and the percentage of the Control group with a use of force declined by 39% (Table 3). Statistical significance aside, one could make a persuasive argument about the prac­ tical significance of the findings. In fact, SPD lead­ ership was quite pleased with the reductions in force and complaints when briefed about the re­ sults, and unconcerned about statistical significance. Moreover, the positive trends for the Treatment group were temporary. Use of force and citizen complaints increased during the post-RCT period, and the uptick for the Treatment group occurred during the same time the Control group (with their newly assigned BWCs) posted declines in those same outcomes. The reason for this trend in the SPD remains unclear. Part of the trend may be ex­ plained by a policy change. Beginning in January 2016, the department implemented a new use of force policy that required officers to record a larger universe of behaviours as reportable force. The policy change may have influenced our find­ ings. Given that there are only 3 months of data after the policy change, it is not possible to fully explore the effect of the change on use of force prevalence. As the authors collect additional data over time, they will be able to more formally inves­ tigate the policy change. Alternatively, perhaps the Treatment officers became more cautious or re­ strained in the months after they were assigned BWCs because of the novelty of the technology, or because they were concerned about how super­ visors might review their behavior. Prior research on officer perceptions of BWCs has identified supervisor review as a common concern (Gaub et al., 2016). But as time passed, officers may have become more comfortable with the technology and the potential for supervisory review. As a result, their BWC-generated restraint dissipated over time and they returned to their normal pre- BWC behavior. 20 Of course, it is important to bear in mind that study officers’ ‘normal pre-BWC be­ havior’ rarely involved force or produced citizen complaints. Nevertheless, the potential for BWC- generated benefits to wane over time warrants add­ itional research attention. Thirdly, recent studies conducted by Ariel and colleagues raise important questions about the impact and consequences of BWCs. For example, Ariel et al. (2016c) reported a higher rate of assaults on BWC officers compared to officers without cam­ eras. The authors offered several potential explan­ ations for the connection, including changes in officer reporting patterns and increased vulnerabil­ ity to assault as officers became less assertive. The finding has received significant media attention and was recently cited by the Boston police union in their lawsuit seeking an injunction to stop the de­ partment from creating a BWC program (Levenson and Allen, 2016). Ariel et al. (2016c, p. 10) note ‘the question about the reason for the increased assaults is not something that can be left to debate and must be [scrutinized] empirically’. 19 The Rialto study involved all 54 patrol officers in the department. The studies in Mesa (Mesa Police Department, 2013), Phoenix (Katz et al., 2014), and …