Journal of Applied Psychology
2017, Vol. 102, No. 11, 1528 –1544
© 2017 American Psychological Association
0021-9010/17/$12.00 http://dx.doi.org/10.1037/apl0000249
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
When Fellow Customers Behave Badly: Witness Reactions to Employee
Mistreatment by Customers
M. Sandy Hershcovis
Namita Bhatnagar
University of Calgary
University of Manitoba
In 3 experiments, we examined how customers react after witnessing a fellow customer mistreat an
employee. Drawing on the deontic model of justice, we argue that customer mistreatment of employees
leads witnesses (i.e., other customers) to leave larger tips, engage in supportive employee-directed
behaviors, and evaluate employees more positively (Studies 1 and 2). We also theorize that witnesses
develop less positive treatment intentions and more negative retaliatory intentions toward perpetrators,
with anger and empathy acting as parallel mediators of our perpetrator- and target-directed outcomes,
respectively. In Study 1, we conducted a field experiment that examined real customers’ target-directed
reactions to witnessed mistreatment in the context of a fast-food restaurant. In Study 2, we replicated
Study 1 findings in an online vignette experiment, and extended it by examining more severe mistreatment and perpetrator-directed responses. In Study 3, we demonstrated that employees who respond to
mistreatment uncivilly are significantly less likely to receive the positive outcomes found in Studies 1 and
2 than those who respond neutrally. We discuss the implications of our findings for theory and practice.
Keywords: deonance theory, third-party injustice, witnesses, workplace aggression, workplace incivility
is negative behavior that may have an ambiguous (e.g., incivility, Andersson & Pearson, 1999) or clear (e.g., aggression,
Neuman & Baron, 2005) intent to harm the target. Workplace
mistreatment does not go unnoticed. Research has argued and
found (Porath, MacInnis, & Folkes, 2010) that witnesses react
negatively to the mistreatment of others. Witnessing mistreatment
between employees can lead insiders to feel emotionally drained
(Totterdell, Hershcovis, Niven, Reich, & Stride, 2012) and outsiders (e.g., customers) to form negative impressions of the firm and
its employees (Porath et al., 2010). These reactions may ultimately
harm business, which makes witness reactions to employee mistreatment a particularly important line of inquiry.
But, how do customers react upon witnessing fellow customers
mistreat employees? We contend that customers who witness this
mistreatment may act to support and compensate employees, and
punish perpetrators. According to deontic theory, people do not
like to see others being mistreated, and feel the urge to rectify such
injustices (Folger, 2001). Building on this theory, we argue that
customers who witness fellow customers mistreat employees not
only punish perpetrators, but also support targets and try to make
amends.
This research makes four key contributions. First, research to
date has largely focused on witnessed reactions to mistreatment
between employees. Drawing on deontic theory, we examine how
customers (i.e., the witnesses) react toward employees (Study 1
and 2) and perpetrators (Study 2) when fellow customers mistreat
the employees. Employee mistreatment by customers is important
to consider for two reasons. First, organizations impose asymmetric constraints on employees in their dealings with customers.
“Service with a smile” is a well-known edict for service employees, and display rules limit the types of emotions that employees
are permitted to express during customer interactions. For instance, Grandey, Rafaeli, Ravid, Wirtz, and Steiner (2010) found
“Customers are always right” is a mantra that many organizations espouse despite the deplorable ways in which some customers treat front-line service workers. Grandey, Kern, and Frone
(2007) found that mistreatment by customers occurs more often
than mistreatment by coworkers/supervisors, and that such mistreatment negatively affects employee well-being above mistreatment by insiders (i.e., other employees). Further, mistreatment
negatively effects target well-being, attitudes, and performance
(see Hershcovis & Barling, 2010). The service sector accounts for
120.64 million jobs in the U.S. (United States Department of
Labor, 2014), which underscores the physiological and psychological damage that mistreatment can leave in its trail.
Workplace mistreatment, which may include verbal abuse (e.g.,
Grandey et al., 2007), incivility (e.g., Cortina, Magley, Williams,
& Langhout, 2001), or aggression, (e.g., Neuman & Baron, 2005)
This article was published Online First July 27, 2017.
M. Sandy Hershcovis, Haskayne School of Business, University of Calgary; Namita Bhatnagar, Asper School of Business, University of Manitoba.
This research was supported by a grant from the Social Sciences and
Humanities Research Council of Canada. We thank Steve Granger, Lukas
Neville, Valerie Shan, Nick Turner, and Justin Weinhardt for providing
constructive comments on earlier drafts of this article. We thank Sarah
McAmmond for her research assistance on the field experiment. We also
thank Kevin Kelloway, Nicolas Roulin, and Yannick Griep for their
assistance/advice with data collection and/or analysis. Earlier versions of
this article were presented at the Academy of Management and Frontiers in
Service conferences.
Correspondence concerning this article should be addressed to M. Sandy
Hershcovis, Haskayne School of Business, University of Calgary, 2500 University Drive NW, Calgary, Alberta, T2N 1N4. E-mail: m.hershcovis@
ucalgary.ca
1528
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
that while employees may express anger toward coworkers, and to
a lesser degree toward supervisors, they may not display anger
toward customers. In contrast, customers do not face such constraints and may mistreat employees without sanctions. Second,
preexisting relationships (e.g., those rooted in power dynamics or
friendships) between insiders may inhibit or encourage interventionist reactions among witnesses when mistreatment occurs internally. On the other hand, customers that witness other customers
mistreat service employees are not bound by preexisting relationships and have greater freedom to act. These asymmetric constraints on employee and customer behaviors leave witnessing
customers with a potentially large role to play in the justice
restoration process of targeted employees. We aim to understand
this role and in particular, the extent to which witnessing customers attempt to restore justice.
Second, in Study 2 we build on deontic theory (Folger, 2001) by
examining a target-focused path to justice restoration. Research to
date has focused almost exclusively on perpetrator-directed reactions to witnessed mistreatment (see Reich & Hershcovis, 2015
and Mitchell, Vogel, & Folger, 2015 for exceptions). Deonance
theory argues that witnesses of an injustice become angry, and
strive to restore justice by punishing perpetrators. We suggest
another approach to justice restoration via target-aiding. In particular, we examine three target-directed behaviors— cognitive, affective, and behavioral—to determine whether witnesses, in addition to punishing perpetrators, also help targets (O’Reilly &
Aquino, 2011). This is an important avenue of research because,
while previous research has examined perpetrator-directed punishment as a method for redressing injustice, target-aiding is a more
constructive approach to justice restoration. Further, we introduce
empathy, an other-focused emotion produced when people perceive that others are suffering (Batson, Fultz, & Schoenrade,
1987), as the deontic mechanism underlying target-directed responses. Whereas anger is expected to explain perpetrator-directed
deontic action, we posit that empathy explains target-directed
deontic action. This study thus adds to deontic theory by considering a second—targeted-directed—pathway to justice restoration
through empathy.
Third, we make an empirical contribution by examining whether
the magnitude of the mistreatment increases the magnitude of the
response (Study 2). That is, when the mistreatment is strong, do
witnesses engage in stronger target-aiding and perpetratorpunishing behaviors than when the mistreatment is weak? Research on workplace mistreatment has argued that different forms
of mistreatment (e.g., incivility vs. aggression) are conceptually
different. Incivility is low in intensity and ambiguous in intent
(Andersson & Pearson, 1999), whereas aggression is intense and
unambiguous (Neuman & Baron, 2005). However, meta-analytic
research (Hershcovis, 2011) has found that regardless of the level
of mistreatment, the magnitude of the relationship between mistreatment and its outcomes is the same. In a customer service
context, it is important from an employee welfare standpoint to
understand whether customer support would originate in all instances of mistreatment, or whether this support gets dampened in
the face of ambiguity (e.g., incivility) or personal safety considerations (e.g., aggressive behavior).
Finally, we examine whether target responses influence witness
reactions (Study 3). The “customer is always right” mantra assumes that employees must remain respectful in the face of mis-
1529
treatment. We test this assumption empirically by comparing witness reactions when employees respond uncivilly versus neutrally
to mistreatment from customers to understand the constraints on
employee reactions. Given the inclination for employees to respond in kind (Gouldner, 1960), this study investigates whether
employees risk losing support from witnesses if they become
discourteous during an unpleasant customer interaction. Understanding this is important because, on the one hand, if employees
are able to respond in kind to rude customers without harming their
image in the eyes of witnesses, this may alleviate the need to
engage in depleting emotional labor (Grandey et al., 2007; Sliter,
Jex, & Wolford, 2010). On the other hand, if employees who
respond in kind are perceived more negatively by witnesses, this
highlights an important boundary condition for employee reactions
in the face of customer mistreatment.
In the next section we discuss deonance theory and develop
hypotheses surrounding witness reactions toward employees who
experience mistreatment from other customers.
Deonance Theory and Customer Reactions
to Mistreatment
People care about justice not only out of self-interest, but also
because it is the “right thing to do” (Folger, 2001). According to
deonance theory, people have a morally rooted sense of how others
ought to behave; they hold themselves and others accountable for
moral behavior. Deontic reactions are morally based reactions,
such as righteousness or indignation in response to perceptions of
unfair treatment (Folger, 2001). In the present study, we consider
workplace mistreatment to be a form of unfair treatment because it
violates workplace norms (Andersson & Pearson, 1999) and contravenes interpersonal justice, or the degree to which people are
treated with respect and dignity at work (Bies & Moag, 1986). We
examine two forms of workplace mistreatment: incivility (Andersson
& Pearson, 1999; Porath & Erez, 2007) and aggression (Neuman &
Baron, 2005).
Deonance theory argues that people have bounded autonomy in
that moral norms restrict us from acting as we please (e.g., Folger,
2001; Folger, Ganegoda, Rice, Taylor, & Wo, 2013). Folger et al.
(2013) described such norms as “informal behavioral control
mechanisms that dictate the appropriateness of certain types of
behavior in organizational settings” (p. 909). A violation of moral
norms, as conceptualized here, suggests that someone behaves in a
way that they ought not to behave (Folger, 2001). Given its nature,
workplace mistreatment arguably violates this norm; people ought
not to behave in uncivil/aggressive ways toward others. Indeed, the
definition of incivility includes that it is a violation of organization
norms, and the definition of aggression states that targets are
motivated to avoid it. Thus, both behaviors should be constrained
by bounded autonomy. Therefore, people that witness customers
mistreating service employees are expected to turn into deontic
agents, or those who act in the name of fairness (Beugré, 2010).
Folger (2001) conceptualized deonance theory as the motivational consequences that are instigated when people observe others
who violate moral norms. This theory underpins a growing body of
research that shows that third-parties react negatively to injustices,
including workplace mistreatment (e.g., Porath & Erez, 2009;
Reich & Hershcovis, 2015; Turillo, Folger, Lavelle, Umphress, &
Gee, 2002). For example, Turillo et al. (2002) found that witnesses
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1530
HERSHCOVIS AND BHATNAGAR
engaged in retributive actions against perpetrators, even at personal cost to themselves (i.e., sacrificing their own money). Similarly, Reich and Hershcovis (2015) found that witnesses evaluated
mistreatment perpetrators more negatively, and punished them by
assigning them more undesirable work than their more civil counterparts. The overall findings from this literature suggest that
witnesses are perpetrator-focused and punitive in their responses.
Though research has examined insider (e.g., Mitchell et al.,
2015) and outsider reactions (e.g., Porath et al., 2010) to mistreatment between employees, we know little about how customers
react when other customers mistreat employees. We know that
customers form more negative impressions of the organization
when employees mistreat each other because such behavior is
unprofessional (Porath et al., 2010). However, customer reactions
to customer mistreatment is qualitatively different because when
mistreatment originates from the customer, the violation is no
longer one of professionalism, but one of fair treatment.
Although deontic theory emphasizes perpetrator punishment as
the primary deontic reaction, O’Reilly and Aquino (2011) argued
that witnesses can also respond by aiding the target. In Study 1, we
focus on target-aiding behaviors in customer service settings because customers are less able to punish other customers. Indeed,
doing so may itself be a norm violation (i.e., that one should stay
out of others’ business). Customers interact primarily with service
providers, and have greater opportunity to engage in targetdirected aiding behaviors.
We examine three types of target aiding: cognitive, affective,
and behavioral. In terms of cognitive aiding, we examined customer evaluation of the employee’s service. This form of aiding
provides long-term benefits to employees, especially when incentive programs used to motivate employees emphasize customer
satisfaction (Hauser, Simester, & Wernerfelt, 1994). For our affective measure, we examined customer affective support. Affective support gives recipients the feeling that they are valued and
cared for (Ducharme & Martin, 2000), and typically occurs soon
after they experience an injustice. Research in the mistreatment
literature (e.g., Sakurai & Jex, 2012) has found that support behaviors from others can buffer its negative effects on targets.
Finally, we examine customer tipping as a behavioral measure.
This form of aiding helps employees from a short-term financial
perspective, and serves as a compensatory reward. With this type
of aiding behavior, the customer sacrifices his or her own money
to offset the bad behavior of a fellow customer. Research (e.g.,
Kahneman, Knetsch, & Thaler, 1986) has shown that when participants perceive an injustice, they are willing to sacrifice some of
their own financial gain to punish the perpetrator and support the
victim. In the present study, we build on this work to consider
whether witnesses are willing to compensate the victim without the
chance to simultaneously punish the perpetrator. That is, do witnesses try to restore justice through compensatory tipping? Based
on the deonance arguments above, we expect that:
Hypothesis 1 (H1): Customers who witness another customer
mistreat a service provider will be more likely to: support the
target (H1a), evaluate the target positively (H1b), and leave
higher tips (H1c) than customers who do not.
In the next sections, we describe our three studies. Study 1
examines target-directed witness reactions in the context of a field
experiment. Building on Study 1, Study 2 examines the extent to
which anger and empathy mediate the relationship between mistreatment and perpetrator- and target-directed reactions, respectively. We also investigate the magnitude of the mistreatment to
determine whether more severe mistreatment has a stronger effect
on witness reactions. Finally, in Study 3, we investigate employee
reactions to mistreatment (uncivil vs. neutral), to understand the
potential constraints on employee reactions to mistreatment.
Study 1
Method
Design and participants. We carried out a field experiment
involving customer reactions to witnessing the purported neutral
versus uncivil mistreatment1 of a service employee by another
customer (both confederates). Participants were 85 actual customers (42% female, M age ⫽ 31.59 years, SD ⫽ 9.1, age range ⫽
19 – 65) of a local fast food restaurant in a midsized North American city. We obtained access to this research site through a
graduate student who worked at the restaurant. We discussed the
details and rationale for the study, our expected findings, and
potential risks with the owner, who provided written consent to
conduct the study. The participants were not initially aware that
they were part of a study, but were informed at the end when we
thoroughly debriefed them, obtained consent, and thanked them
with a $10 gift certificate to the restaurant. This procedure is
consistent with institutional review board ethics guidelines that
allow researchers to obtain consent after-the-fact under special
circumstances. Obtaining consent prior to the study would have
primed participants that the interaction they observed was false,
and we would not have been able to test our research questions.
We stopped one experiment as the participant began to yell at our
confederate actor,2 and we removed a second case because the
participant expressed suspicion that the interaction was contrived.
Ethics was granted by the Joint Faculty Research Ethics Board at
the University of Manitoba under the title “Witness Reactions to
Customer Rudeness” (protocol number J2014-072).
Procedure and manipulations. In this experiment, real customers (i.e., the participants) at a fast food restaurant witnessed a
fellow customer (who was a confederate actor) behave either
neutrally toward an employee (who was also a confederate) or
mistreat the employee. We also placed a coder (blind to hypotheses) at a table near the front counter. The coder recorded the
number of times the participant said “please” and “thank you,” and
subjectively evaluated the support shown by the participant toward
the target following the confederate interaction. The coder also
1
Our control condition may be interpreted as civil rather than neutral
because the customer used pleasantries when addressing the server. We
therefore conducted a post hoc script comparison between a potential
neutral and civil script, by removing the civil language. For example,
instead of “no thanks” in response to whether the customer would like
guacamole, the customer replies “No.” In a sample of undergraduate
students, we assessed participants’ (n ⫽ 101) perceptions of incivility, F(1,
99) ⫽ .75, p ⫽ .39, and lack of justice, F(1, 99) ⫽ .53, p ⫽ .47, using the
manipulation checks used in Study 2. There were no significant differences
between the civil and neutral conditions.
2
We thoroughly debriefed the customer to his satisfaction and gave him
a gift certificate.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
recorded all verbal statements made by the witness related to the
confederate interaction. The perpetrator and target remained the
same throughout the study and both were female to control for
possible gender effects (a cross-gender interaction might induce a
power dimension).3 The coder did not code the interaction between
confederates across the two conditions because this may have
biased their subsequent coding of the dependent variables.
We conducted the experiment during nonpeak restaurant hours
because we wanted to ensure that only one customer witnessed the
interaction at a time. We were concerned that multiple customers
might lead to diffusion of responsibility on the part of the customer
(Darley & Latane, 1968). A large window allowed the confederates to see when a customer was approaching. At this point, the
actor playing the customer positioned herself in front of the counter, and pretended to read the menu until the real customer (hereafter the participant) stood behind her. The actor (hereafter the
perpetrator) then began either a neutral or uncivil interaction with
the confederate employee (hereafter the target). The condition was
determined randomly before the participant entered the restaurant,
by drawing either a “neutral” or “mistreatment” slip of paper. The
target responded in the same neutral way regardless of condition.
The script is the same across each of the neutral and mistreatment oriented interactions (see Appendix A), with minor verbal,
intonation, and nonverbal changes (e.g., arms crossed) in the
mistreatment condition. We drew on existing mistreatment scales
(e.g., Baron & Neuman, 1996; Bennett & Robinson, 2000; Cortina
et al., 2001) to develop our script for the mistreatment interaction.
Our script reflects mistreatment in several ways. For instance, the
perpetrator ignores the target several times, is on her iPhone during
the interaction, and walks out of the restaurant without saying
“thank you.” Further, she is condescending (e.g., “This is a pretty
simple order”) and sarcastic (e.g., “I said ‘to go,’ so obviously”) to
the target. These examples map onto items such as “giving someone the silent treatment”, (Baron & Neuman, 1996), “acted rudely
towards someone,” (Bennett & Robinson, 2000), and “put you
down or was condescending to you,” (Cortina et al., 2001). In the
neutral interaction, the perpetrator uses a similar script, but does
not look at her iPhone, does not use the sarcastic tone, and
responds to inquiries in a neutral manner.
Following the interaction, the perpetrator leaves the restaurant
and the target (i.e., employee) moves on to take the order of the
participant (i.e., the real customer). After the participant pays, the
target asks the customer whether he or she is willing to complete
a short customer service survey while waiting for the order. All
participants agreed to complete the survey, which they do privately
at another counter, and then place it into a locked survey box. On
the survey, participants evaluate the employee’s deservingness for
an employee of the month award (see below) in addition to
demographics and open-ended comments.
Measures.
Target support. We assessed target support using three measures (two quantitative and one qualitative). First, the coder
counted the number of times the participant used the words
“please” and “thank you” during each service interaction. Second,
the coder evaluated the participant’s friendliness (1 ⫽ not at all
friendly, 5 ⫽ extremely friendly) toward the target. Finally, the
coder recorded all comments that the participant made to the
target. These responses were coded for target support and we
report the qualitative findings.
1531
Target evaluation. We assessed participants’ evaluation of the
target’s service on the customer service survey by asking “Does
your server today deserve the employee of the month award?”
(1 ⫽ definitely not, 10 ⫽ definitely).
Tip percentage. We assessed tips by leaving a tip jar with
precounted change at the front counter.4 After each condition, we
emptied the tip jar and recorded the difference between the precounted cash and any additional tip. Similarly, when customers
paid by debit/credit card, we recorded the tip amount. Tips were
calculated as a percentage as tips vary based on the total bill.
Manipulation check. Due to the nature of the study, we could
not conduct a direct manipulation check during the experiment
because it would have primed participants that they were involved
in a study (Webb, Campbell, Schwartz, & Sechrest, 2000). However, coders recorded the participants’ comments to the target
following the interaction—these comments were indicative of perceived target mistreatment (reported below).
We also conducted a manipulation check on the scripts used in
our experiment (see Appendix A). We randomly assigned undergraduate participants at a midsize North American university (N ⫽
98) to read either the neutral or the mistreatment script. We then
asked participants, on a scale from 1 to 7 (1 ⫽ strongly disagree,
7 ⫽ strongly agree) the extent to which they agreed with five items
(e.g., “the customer insulted the employee,” “the customer belittled the employee,” Cronbach’s alpha ⫽ .98). These items are
adapted from Cortina et al.’s (2001, 2013) measures, which include for example: “made insulting or disrespectful remarks about
you,” “accused you of incompetence,” and a range of items capturing rudeness (e.g., “interrupted or spoke over you,” “doubted
your judgement . . .”). Responses to the five items were averaged
to form a customer mistreatment index. Results of an analysis of
variance with the incivility condition as the independent variable
and the mistreatment manipulation check as the dependent variable
showed that participants in the mistreatment condition (n ⫽ 49)
were significantly more likely to rate the customer as rude (M ⫽
5.95, SD ⫽ .80) than those in the neutral condition (n ⫽ 49, M ⫽
2.93, SD ⫽ 1.68), F(1, 96) ⫽ 128.71, p ⬍ .001, 2 ⫽ .57.
Results
Test of hypotheses.
Quantitative results. Means, standard deviations, and intercorrelations appear in Table 1. To investigate our hypotheses, we
conducted a multivariate analysis of variance (MANOVA) using
mistreatment condition (mistreatment vs. neutral) as our independent variable and target support, (please and thank you counts,
friendliness), tip percent, and target evaluations as our dependent
variables. The overall test was significant, F(5, 79) ⫽ 4.82, p ⫽
.001, 2 ⫽ .24, and we proceeded to examine the between-subjects
3
Coders and actors were trained for 15 hr prior to conducting the
experiment. Two coders traded off shifts, but were trained together. After
each practice experiment, we examined coder agreement, which was 100%
for objective support (counting “please” and “thank you”) and about 80%
for the subjective friendliness measure. In our analyses, there were no
significant differences on the friendliness measure between coders. We
also trained the server to react in the same neutral manner toward both the
actor, and the real customer, regardless of condition.
4
Change was consistent across all participants and the confederate
customer did not leave a tip.
HERSHCOVIS AND BHATNAGAR
1532
Table 1
Means, Standard Deviations, and Intercorrelations (Study 1)
Variable
1.
2.
3.
4.
5.
Mistreatment condition
Target support: Friendliness
Target support: Please/thank you
Target evaluation
Tip percent
Note. N ⫽ 85
p ⬍ .10. ⴱ p ⱕ .05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
†
ⴱⴱ
M
SD
1
2
3
4
—
2.65
4.24
8.95
8.00
—
.96
2.56
1.65
11.00
—
.42ⴱⴱ
.21ⴱ
.30ⴱ
.19†
.44ⴱⴱ
.15
.18†
⫺.01
⫺.08
.09
p ⬍ .001.
effects. Given that our hypotheses are theory-driven and directional, we conducted one-tailed tests (Jones, 1954). In support of
H1a, participants were more supportive of targets of mistreatment
(M ⫽ 4.71, SD ⫽ 2.50, 90% CI [4.10, 5.34]) than nontargets (M ⫽
3.75, SD ⫽ 2.55, 90% CI [3.09, 4.40]), using the “please and thank
you” count, F(1, 83) ⫽ 3.07, p ⫽ .04, 2 ⫽ .04, Cohen’s d ⫽ .38.
Similarly, using the coder’s evaluation of target support (friendliness), we found the same pattern of results, F(1, 83) ⫽ 14.32, p ⬍
.001, 2 ⫽ .14, Cohen’s d ⫽ .83, in the mistreatment (M ⫽ 2.96,
SD ⫽ .80, 90% CI [2.77, 3.31]) and neutral (M ⫽ 2.35, SD ⫽ .66,
90% CI [2.16, 2.54]) conditions. In support of H1b, we found a
significant effect of mistreatment condition on target evaluation
F(1, 83) ⫽ 8.51, p ⫽ .005, 2 ⫽ .09, Cohen’s d ⫽ .63, such that
customers gave higher evaluations to targets in the mistreatment
condition (M ⫽ 9.64, SD ⫽ .89, 90% CI [9.16, 10.13]) than the
neutral condition (M ⫽ 8.40, SD ⫽ 2.12, 90% CI [7.88, 8.92]). We
also found a significant effect of mistreatment condition on tip
percent, F(1, 83) ⫽ 3.21, p ⫽ .039, 2 ⫽ .04, Cohen’s d ⫽ .40. In
support of H1c, customers gave a higher percentage of tips to
targets in the mistreatment (M ⫽ 11%, SD ⫽ 14%, 90% CI [.08,
.13]) than the neutral (M ⫽ 6%, SD ⫽ 7%, 90% CI [.03, .09])
condition.5
Qualitative results. The participant intervened in only five of
the 45 mistreatment interactions (11%, three males and two females). In three instances, the participant spoke directly to the
actor by stating, for example “that was very rude.” In two of the
instances, the participant confronted the perpetrator using body
language. For example, in one instance the participant stared at the
perpetrator and got into her personal space to intimidate her into
stopping.
Although most of the participants failed to intervene, in 33 of
the 45 uncivil interactions, the participant expressed support toward the target. These expressions of support took three forms: (a)
empathy for the target, (b) anger at the perpetrator, and (c) mocking the perpetrator; these were sometimes expressed simultaneously (e.g., “some people should be embarrassed of themselves!”
[anger] and “I am very sorry” [empathy]. In 14 instances (41%),
the participant expressed empathy with comments such as “Sorry,
that was terrible! I wanted to stop her.” In 10 instances (30%), the
participant mocked the perpetrator, for example by joking “Don’t
worry, I’m not in a hurry,” referencing the perpetrator’s snide
opening comment (see script). Finally, participants expressed overt
anger in 10 instances (30%; e.g., “What a bitch!”).
Discussion
Study 1 demonstrates that employees benefit both relationally
and financially when customers are uncivil to them. Consistent
with deonance theory, when employees were mistreated, witnessing customers provided verbal and nonverbal support, were friendlier, evaluated them more positively, and left them 83% higher tips
than when employees were treated neutrally. Customers exhibited
enhanced manners in the form of verbalized “please and thank
you” and other supportive comments to the mistreated employee
(e.g., expressions of empathy for the employee, and anger and
mockery directed at the perpetrator). Despite the unpleasant experience of being mistreated by customers, employees may take
some comfort from the support offered by other customers. That is,
witnessing customers appear to compensate for the bad behavior of
other customers. It is particularly interesting that employees did
nothing, except maintain the often prescribed “the customer is
always right” stance (Grandey, Dickter, & Sin, 2004), to gain these
benefits. That is, by virtue of using the same neutral demeanor
used within neutral interactions with customers, witnesses rewarded employees who were mistreated.
Study 1 has a number of strengths. It is a naturalistic field
experiment, with concomitant external validity, in which we observed how real customers responded after witnessing other customers mistreat employees. Moreover, we collected objective dependent variables and qualitative customer statements to assess our
hypotheses.
Despite these strengths, a number of questions remain from
Study 1. First, we could not assess mediating mechanisms without
revealing to customers that they were participating in an experiment. This is problematic because one might argue that equity
theory (Adams, 1963) and not deonance theory explains our findings. That is, Study 1 showed that employees who coped with
mistreatment by behaving neutrally to rude customers were rewarded for their efforts. This might mean that customers reward
those who work harder, which is consistent with equity theory.
Examining explanatory mechanisms would help determine
whether deonance theory best explains our findings. Second, in
Study 1, we examined low-intensity mistreatment (i.e., incivility)
in comparison with a neutral condition, but due to ethical constraints, could not examine whether more intense and less ambiguous mistreatment yields stronger effects. Third, we focused only
on target-directed outcomes to the exclusion of perpetrator5
A statistically significant difference between the means can exist even
if the confidence intervals about the means overlap somewhat. For this
reason, it is generally recommended that one not judge statistical significance by examining the overlap between confidence intervals as this results
in an overly conservative test whether the null hypothesis is false (Austin
& Hux, 2002). Greenland et al. (2016) caution that confidence intervals are
only a first step and that p values should be used to assess statistical
significance.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
directed outcomes because it would have been difficult to observe
indirect forms of retaliation toward the perpetrator, especially
because we constrained our study to one customer at a time. This
prevented customers from talking to each other about the perpetrator. Fourth, because we used one actor to play the perpetrator
and another to play the target, there is a risk that our effects are
based on actor idiosyncrasies. Fifth, we could not conduct a direct
check of our experimental manipulations but instead relied on post
hoc script tests and qualitative reactions. Sixth, we held gender
constant due to the potential power differences implied by gender.
That is, women are historically seen as less powerful than men
(Cortina, 2008), therefore, it is possible that participants perceive
mistreatment directed from a male toward a female as more
negative than the reverse.
To address the first five limitations, we conducted a second
study aimed at replicating Study 1 findings, and extending them
by: (a) examining deontic mechanisms that explain witness reactions; (b) comparing neutral, uncivil, and aggressive conditions to
determine whether the magnitude of mistreatment influences outcomes; and (c) introducing perpetrator-directed outcomes. Perpetrator treatment and retaliatory intent is therefore assessed in the
vignette based Study 2 where participants can report on their
reactions to witnessed mistreatment. Moreover, the use of a vignette prevents any actor idiosyncrasies from influencing the outcomes and allows us to conduct a direct manipulation check. To
address the gender limitation, we conducted a pilot test to assess
whether gender influences customer perceptions of justice and
incivility, and found that it does not;6 therefore, our vignette
descriptions omit gender in Study 2.
Study 2
Moral Anger and Perpetrator-Directed Action
As described earlier, deontic theory argues that certain actions
violate moral norms, and that such violations are perceived to be
unjust (Folger, 2001). Folger and colleagues (e.g., Folger, 2001;
Folger et al., 2013) theorized that perceived injustice drives witnesses’ subsequent emotions and actions. The primary deontic
emotion proposed by deontic theory is moral anger (Folger,
2001)—a temporary discrete negative emotion that arises in response to a perceived injustice (O’Reilly & Aquino, 2011). When
people witness an injustice, such as workplace mistreatment, they
intuit that something is wrong, and moral anger is roused. Anger is
the driver of the deontic agents’ desire to act to restore justice.
Thus, the focus of justice restoration in deontic theory is on the
perpetrator. Folger (2001) argued that witnesses do not want
violators to get away with unjust behaviors, and hold violators
accountable by punishing them. Thus, this theory predicts that
moral anger explains the relationship between witnessed mistreatment and perpetrator-targeted punitive outcomes. Research has
found that moral anger mediates the relationship between witnessed mistreatment and perpetrator outcomes when perpetrators
are employees (e.g., Mitchell et al., 2015; Porath et al., 2010;
Reich & Hershcovis, 2015). However, in a service environment,
perpetrator-directed outcomes may be less straightforward because
customers have little occasion to interact with other customers.
Understanding whether employee mistreatment influences intentions to punish offending customers is important as such reactions
1533
may adversely affect business. That is, both customers may leave
the service interaction with a negative experience whether or not
intentions to punish the perpetrator are followed through by actions. According to literature on customer-to-customer interactions
(CCI), other customers influence the service experience of fellow
customers within service contexts (Eiglier & Langeard, 1977). For
instance, CCI has been found to influence customer satisfaction
(Harris, Davies, & Baron, 1997) and loyalty (Moore, Moore, &
Capella, 2005), with negative CCI being correlated with dissatisfaction (Wu, 2007).
Thus, in the present study, we examine two perpetrator-directed
outcomes: perpetrator positive treatment and retaliatory intentions.
Deontic action can take one of several forms (do nothing, aid the
target, or directly/indirectly punish the perpetrator; O’Reilly &
Aquino, 2011). We examine worsened perpetrator positive treatment intentions as a form of indirect punishment because it involves a withdrawal of behavior (i.e., the choice to refrain from
engaging in friendly and respectful interaction). We also examine
retaliatory intentions as a more direct form of punishment, which
corresponds with the notion of retributive justice. In deontic theory, retributive justice is a way to restore justice by punishing the
perpetrator in a fair or symmetrical way (Folger, 2001). Given that
witnesses in the present study observe mistreatment of the server,
customers may think that retaliatory mistreatment of the perpetrator is a fair punishment. Thus we predict that:
Hypothesis 2 (H2): Customers who witness customer mistreatment toward a service provider are less likely to hold
positive treatment intentions (H2a) and more likely to hold
retaliatory behavioral intentions (H2b) toward the perpetrator
than customers who witness neutral treatment.
Hypothesis 3 (H3): Moral anger will mediate the relationship
between witnessed mistreatment and perpetrator-directed positive treatment intentions (H3a), and retaliatory intentions
(H3b).
Empathy and Target-Directed Action
The crux of deontic theory is about witness reactions to a
perceived injustice, and in particular, the altruistic desire to restore
justice. Although the theory focuses on perpetrator-targeted emotion and punishment, researchers have suggested that witnesses
may also aid the target (Mitchell et al., 2015; O’Reilly & Aquino,
2011). Indeed, O’Reilly and Aquino (2011) argue that one way to
aid a victim is to lend him or her a sympathetic ear, which implies
that witnesses may not only experience anger for the perpetrator,
but also empathy for the target. Similarly, Mitchell et al. (2015)
posited that “it may also be possible for third parties to experience
sympathy, compassion, or empathy for the abused [target]” (p.
1050) and they suggest that future research should explore such
6
We conducted a 2 (customer gender: male vs. female) ⫻ 2 (employee
gender: male vs. female) between subjects experiment using the same
aggression vignette in Study 2, except with gender pronouns throughout.
97% of participants (n ⫽ 203) correctly recalled the gender of both the
perpetrator and target. There were no significant main or interaction effects
for lack of perceived justice or mistreatment F(2, 198) ⫽ .18, p ⫽ .84.
There were also no significant effects for any of the mediators or dependent
variables.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1534
HERSHCOVIS AND BHATNAGAR
target-directed emotions. Furthermore, our own qualitative findings from Study 1 showed that witnesses experienced empathy
(42%) more often than anger (30%).
Empathy is defined as an other-focused emotion produced by
witnessing another person’s suffering (Batson et al., 1987). Batson
and colleagues (Batson et al., 1987, 1988) found that witnesses,
upon perceiving others’ distress, experience vicarious emotional
arousal in the form of empathy. This empathy motivates them to
altruistically reduce this distress by helping the distressed individual. In other words, Batson et al. argued that, consistent with
deonance theory, a key motivational rationale for witness intervention following an empathic reaction is altruism rather than
self-interest.
Integrating deonance theory, which argues that witnessing an
injustice will evoke an angry response, with Batson et al. (1987)
theory that witnessing another’s distress will evoke an empathetic
response, we argue that witnesses are likely to experience empathy
toward the target in addition to anger toward the perpetrator. As
noted above, perpetrator-directed anger is likely to drive
perpetrator-directed intentions; whereas, we expect target-directed
empathy to drive target-directed reactions. Although customers
may possess negative behavioral intentions toward perpetrators,
this alone may not restore justice. The next best thing to intervention (which is rare as evidenced in Study 1 and in the literature on
bystander intervention; Darley & Latane, 1968) would be for
witnesses of mistreatment in a service context to support targets.
Unlike internal mistreatment between coworkers where employees
have the means to punish perpetrators, customers exit the service
interaction at the end of the transaction, and are unlikely to see
fellow customers again. Therefore, to restore justice, they are
likely to also turn their attention to the target. Thus, we hypothesize the following.
Hypothesis 4 (H4): Empathy will mediate the relationship
between witnessed mistreatment and target-directed support
(H4a), target evaluations (H4b), and tips (H4c).
Severity of the Mistreatment
Although mistreatment in not the norm, when it occurs it can
range in severity, especially in the service industry where customers are not governed by organizational policies and reprimands in
the same way employees would be. That is, whereas most organizations have policies regarding respectful workplace treatment
that might help to curb internal employee aggression, customers do
not face the same constraints. Therefore, employees may be victim
to minor forms of mistreatment, such as incivility (Andersson &
Pearson, 1999), or more intense forms of mistreatment such as
workplace aggression (Neuman & Baron, 2005).
Researchers have argued for the conceptual differences between
incivility and more severe forms of mistreatment. Incivility, unlike
other forms of mistreatment, is low in intensity and ambiguous in
intent (Andersson & Pearson, 1999). In contrast, workplace aggression has clear intent and is more severe. However, in a metaanalysis comparing workplace incivility with workplace bullying,
Hershcovis (2011) found that the magnitude of the relationship
between different forms of mistreatment and its outcomes are the
same. Hershcovis (2011) argued that more severe forms of mistreatment may have more severe outcomes than workplace inci-
vility, but due to measurement overlap between these two constructs, the data do not reflect these differences.
Given that incivility is more frequent than more severe forms of
workplace aggression, we were interested in whether the magnitude of witness reactions differed depending on the nature of the
mistreatment. Based on the conceptualization of incivility as a less
intense and more ambiguous form of mistreatment, we expect that
witnesses will be less likely to engage in positive perpetratordirected action, and more likely to engage in positive targetdirected action, when they witness incivility versus workplace
aggression, respectively.
Hypothesis 5 (H5): Customers who witness workplace aggression will be significantly less likely to hold positive treatment
intentions (H5a) and more likely to hold retaliatory behavioral
intentions (H5b) toward the perpetrator than customers who
witness incivility.
Hypothesis 6 (H6): Customers who witness workplace aggression will be significantly more likely to support the target
(H6a), more likely to evaluate the target positively (H6b), and
leave higher tips (H6c) than customers who witness incivility.
Method
Participants. Participants were 183 people (50% female, M
age ⫽ 34.87 years, SD ⫽ 10.92, age range ⫽ 18 – 68) recruited
through an online panel provider (MTurk). Online panels have
been described as reliable sources for access to diverse types of
samples (e.g., Landers & Behrend, 2015; Roulin, 2015). Participants that resided in North America were invited to take part in a
15-min study in exchange for $2. Participants worked on average
37.62 (SD ⫽ 9.05) hr per week, and their mean tenure was 6.13
years (SD ⫽ 5.49). They held a wide range of positions such as:
teacher, data analyst, construction supervisor, accountant, and engineer.
Materials and procedure. We conducted an online vignette
experiment in which we manipulated the neutral, uncivil, and
aggressive treatment of an employee by a customer. Participants
read a vignette in which they were asked to assume the role of a
customer in a fictional coffee shop called Java Time. In this
vignette, participants read about a neutral, uncivil, or aggressive
interaction that they witnessed between the customer in front of
them and the server, and then they answered questions aimed to
assess the mediators (i.e., anger and empathy), the perpetratordirected dependent variables (i.e., customer treatment and customer retaliation), and target-directed dependent variables (i.e.,
target support, target evaluation, and tip percent; see Appendix B
for the vignette scripts). Ethics was granted by the Joint Faculty
Research Ethics Board at the University of Manitoba under an
amendment to the title “Witness Reactions to Customer Rudeness”
(protocol number J2014-072).
Perpetrator positive treatment intentions. We assessed perpetrator positive treatment intentions with three items. An example
item is “I would treat the customer in a friendly manner” (1 ⫽
strongly disagree, 7 ⫽ strongly agree).
Perpetrator retaliatory intentions. We assessed perpetrator
retaliatory intentions with three items based on constructs from the
broad aggression literature (e.g., Cortina et al., 2001; Neuman &
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
Baron, 2005). An example item is “I would be rude to the customer” (1 ⫽ strongly disagree, 7 ⫽ strongly agree).
Target support. We assessed target support using four items
adapted from Ducharme and Martin (2000). We chose this scale
because it mapped onto our Study 1 item about friendliness, and it
was also consistent with the kinds of supportive comments that
witnesses made toward the target (e.g., “You handled that very
well”). The items were: “I would make the employee feel appreciated,” “I would let the employee know they are doing a good
job,” “I would be friendly towards the employee,” and I would
take a personal interest in the employee” (1 ⫽ strongly disagree,
7 ⫽ strongly agree).
Target evaluation. We used the same measure used in Study 1.
Tip percent. Following Study 1, we assessed tips by asking
what percentage tip the employee would leave (from 0% to 50% in
5% increments). Tipping norms in North American coffee shops
suggest that this range would adequately capture the possible range
of tipping percentages plus allow the witness the opportunity to
compensate.
Anger. We asked participants three items used by O’Reilly,
Aquino, and Skarlicki (2016) following the stem “To what extent
did the interaction between the customer and the employee make
you feel the following . . .” The response items were “angry at the
customer,” “upset with the customer,” “and “irritated with the
customer” (Cronbach’s alpha ⫽ .96).
Empathy. We measured empathy using five items adapted
from Batson et al. (1988). Participants were asked “To what extent
did the interaction between the customer and the employee make
you feel the following towards the employee . . .” (1 ⫽ not at all
to 7 ⫽ very much so): “empathy for the employee,” “pity for the
employee,” “concern for the employee,” “protective towards the
employee,” and “sorry for the employee.”
Analytic strategy. As with Study 1, we tested our main effects with one-tailed tests using MANOVA. To test our mediation
effects we conducted structural equation modeling using Bayesian
estimation, which is appropriate for dealing with small sample
sizes (see Kruschke, Aguinis, & Joo, 2012). We conducted a full
and partial mediation model and determined which model fit the
data better using the Deviance Information Criterion (DIC;
Spiegelhalter, Best, Carlin, & van der Linde, 2002) and Bayesian
Information Criterion (BIC; Aiken & West, 1991). The DIC represents the likelihood of the model as a function of the actual
number of parameters, whereas BIC represents the balance between the parameters (i.e., model complexity) and the fit of the
1535
model to the data. In both cases, the lowest value indicates the best
fitting model.
The mediation effects were tested using the product-ofcoefficients approach and we used 10,000 Markov Monte Carlo
iterations to be 90% confident that fifth and 95th percentiles are
estimated with an error smaller than .0005 (Hox, 2010). We linked
the regression coefficients of incivility and aggression to target
support, evaluation, and tip percent via empathy, and to
perpetrator-directed positive treatment and retaliation via anger.
We tested both mediating mechanisms simultaneously in the same
mediation model. Note however that due to the categorical nature
of our independent variable, we were required to run the analyses
on two conditions at a time. That is, we conducted two separate
structural models on the: (a) neutral and uncivil condition, and (b)
neutral and aggressive condition.
Results
Manipulation checks. We conducted two manipulations
checks in Study 2. First, we used the same mistreatment manipulation check as we used in Study 1 (Cronbach’s alpha ⫽ .97).
Second, given that we base our hypotheses on deontic theory, we
checked participants’ justice perceptions about the mistreatment
interaction. We measured interpersonal justice (Bies & Moag,
1986) using Colquitt’s (2001) interpersonal justice measure. Participants were presented with the stem “In this scenario, the customer . . .” followed by four items: “treated the employee in a
polite manner,” “treated the employee with dignity,” “was disrespectful towards the employee,” and “made inappropriate remarks/
comments to the employee” (1 ⫽ strongly disagree to 7 ⫽ strongly
agree). Cronbach’s alpha is .90. Manipulation checks were significant in the expected direction F(2, 180) ⫽ 156.42, p ⬍ .001, and
mean differences are presented in Table 2.
Measurement model. Confirmatory factor analysis shows
that the hypothesized model (Model 1) fits the data very well,
2(151) ⫽ 295.01, p ⬍ .001, CFI ⫽ .97, TLI ⫽ .96, RMSEA ⫽
.07, SRMR ⫽ .07. We compared the seven-factor structure to two
alternative structures. The hypothesized model was significantly
better than Model 2, which combines empathy and anger into one
latent variable, ⌬ 2(6) ⫽ 226.71, p ⬍ .001 [2(157) ⫽ 521.73,
CFI ⫽ .91, TLI ⫽ .89, RMSEA ⫽ .11, SRMR ⫽ .07]. The
hypothesized model is also better than Model 3, which combines
customer positive treatment and retaliation, ⌬ 2(6) ⫽ 357.08, p ⬍
Table 2
Manipulation Checks and Main Effects for Study 2 Outcomes
Neutral
Uncivil
Aggressive
Variable
M
SD
90% CI
M
SD
90% CI
M
SD
90% CI
Target support
Target evaluation
Tip percent
Perpetrator positive treatment intent
Perpetrator retaliation intent
Rudeness manipulation check
Lack of interpersonal justice manipulation check
5.14
5.62
17.58
5.99
1.87
1.93
2.64
1.06
2.10
12.87
.89
1.21
1.24
1.24
[4.93, 5.35]
[5.19, 6.04]
[15.00, 20.16]
[5.60, 6.40]
[1.51, 2.24]
[1.65, 2.22]
[2.35, 2.94]
5.71
6.32
20.40
5.02
2.72
4.76
5.44
1.02
2.01
11.68
1.54
1.41
1.24
1.07
[5.51, 5.92]
[5.91, 6.74]
[17.87, 22.94]
[4.64, 5.40]
[2.35, 3.08]
[4.48, 5.05]
[5.15, 5.73]
6.12
6.97
25.58
4.44
3.39
6.47
6.17
1.02
1.84
12.04
1.91
1.67
.88
1.16
[5.92, 6.33]
[6.55, 7.40]
[23.00, 28.16]
[4.06, 4.82]
[3.02, 3.75]
[6.18, 6.75]
[5.88, 6.47]
Note.
Neutral n ⫽ 60; Uncivil n ⫽ 61; Aggressive n ⫽ 61.
HERSHCOVIS AND BHATNAGAR
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1536
.001 [2(157) ⫽ 652.04, CFI ⫽ .88, TLI ⫽ .85, RMSEA ⫽ .13,
SRMR ⫽ .10].
Test of hypotheses. Means, standard deviations, and intercorrelations appear in Table 3.
Main effects for target-directed outcomes. To investigate
H1a to H1c, we conducted a MANOVA using mistreatment conditions (neutral, uncivil, aggressive) as our independent variable
and target support, target evaluation, and tip percentage as our
dependent variables. The overall test was significant, F(3, 179) ⫽
9.27, p ⬍ .001, 2 ⫽ .13; therefore, we proceeded to examine the
between-subjects effects. Means, standard deviations, and confidence intervals are reported in Table 2.
Target support. In support of H1a, participants were more
likely to support the target in the uncivil (p ⫽ .002) and aggression
(p ⬍ .001) conditions than the neutral condition, F(2, 180) ⫽
12.10, p ⬍ .001, 2 ⫽ .12, Cohen’s d ⫽ .88 and .55 for the
difference between the aggression and neutral mistreatment conditions and the uncivil and neutral conditions, respectively. Moreover, in support of H6a, the difference between the aggression and
uncivil condition was significant (p ⫽ .01), Cohen’s d ⫽ .34.
Target evaluation. In support of H1b, participants were more
likely to positively evaluate the target in the aggression (p ⬍ .001)
and in the uncivil (p ⫽ .05) than in the neutral condition, F(2,
180) ⫽ 6.52, p ⫽ .001, 2 ⫽ .07, Cohen’s d ⫽ .66 and .34 for the
difference between aggression and neutral and uncivil and neutral
conditions, respectively. Moreover, in support of H6b, the difference between the aggression and uncivil condition was significant
(p ⫽ .04), Cohen’s d ⫽ .31.
Tip percent. We found a significant effect of mistreatment
condition on tip percentage, F(2, 180) ⫽ 5.98, p ⫽ .002, 2 ⫽ .06.
Participants were more likely to endorse a higher tip percentage for
employees in the aggression than in the neutral (p ⬍ .001) condition, but there was no significant difference between the neutral
and uncivil (p ⫽ .10) condition. Cohen’s d ⫽ .61 and .23 for the
difference between aggression and neutral and uncivil and neutral
conditions, respectively. Thus, H1c was partially supported. Further, in support of H6c, the difference between the aggression and
uncivil condition was significant (p ⫽ .02), Cohen’s d ⫽ .40.
Main effects for perpetrator-directed outcomes. To investigate H2a and H2b, we conducted a MANOVA using mistreatment conditions (neutral, uncivil, aggressive) as our independent
variable and perpetrator positive treatment and retaliation intentions as our dependent variables. The overall test was significant,
F(2, 180) ⫽ 21.95, p ⬍ .001, 2 ⫽ .20; therefore, we proceeded
to examine the between-subjects effects.
Perpetrator positive treatment intentions. In support of H2a,
we found that participants were less likely to endorse positive
perpetrator treatment in the aggression (p ⬍ .001) and uncivil (p ⬍
.001) conditions than in the neutral condition, F(2, 180) ⫽ 16.38,
p ⬍ .001, 2 ⫽ .15, Cohen’s d ⫽ 1.05 and .78 for the difference
between the aggression and neutral and uncivil and neutral conditions, respectively. Moreover, in support of H5a, there was a
significant difference between the aggressive and uncivil condition
(p ⫽ .03), Cohen’s d ⫽ .33.
Perpetrator retaliation intentions. In support of H2b, we
found that participants were more likely to endorse perpetrator
retaliation intentions in the aggression (p ⬍ .001) and uncivil (p ⫽
.001) conditions than in the neutral condition, F(2, 180) ⫽ 16.75,
p ⬍ .001, 2 ⫽ .16, Cohen’s d ⫽ 1.04 and .65 for the difference
between the aggression and neutral and uncivil and neutral conditions, respectively. Moreover, in support of H5b, the difference
between the aggressive and uncivil condition was significant (p ⫽
.001), Cohen’s d ⫽ .43.
Mediation analyses. To investigate the indirect effect of witnessing mistreatment on perpetrator-directed outcomes through
anger, we relied on a structural equation model using Bayesian
estimation. Both incivility and workplace aggression used the
control (neutral) group as its referent. We examined anger and
empathy as parallel meditators. Before presenting the mediation
results for all our mediation hypotheses, we set out to determine
whether a full or partial mediation model offered the best fit to the
data. When comparing the BIC and the DIC values, we noticed
that the full mediation model fit the data best compared to a partial
mediation model for all mediations. The BIC and DIC values were
lower in the full mediation model when examining the uncivil and
neutral conditions at 7984.58 and 7766.04, respectively, compared
with 8002.46 and 7772.21 in the partial mediation model. Similarly, the BIC and DIC values were lower in the full mediation
model when examining the aggressive and neutral conditions at
7858.07 and 7644.34, respectively compared with 7878.54 and
7646.14 for partial mediation. Thus, our model suggests full me-
Table 3
Means, Standard Deviations, and Intercorrelations (Studies 2 and 3)
Variable
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
MStudy 2/
MStudy 3
SDStudy 2/
SDStudy 3
1
Mistreatment condition
—
—
—
Anger
3.91/5.88
2.19/1.35
.79ⴱⴱ
Empathy
4.40/5.68
2.11/1.10
.76ⴱⴱ
Perpetrator positive treatment intent 5.15/4.47
1.63/1.75 ⫺.39ⴱⴱ
Perpetrator retaliation intent
2.66/3.16
1.56/1.55
.40ⴱⴱ
Target evaluation
6.29/5.26
2.04/2.36
.26ⴱⴱ
Target support
5.64/5.47
1.09/1.20
.34ⴱⴱ
Tip percent
21.07/21.43 12.53/14.32
.25ⴱⴱ
Gender
—
—
.03
Age
35.87/33.30 10.92/9.59
.04
2
3
ⴱ
⫺.23
.96/.96
.87ⴱⴱ
⫺.47ⴱⴱ
.47ⴱⴱ
.35ⴱⴱ
.44ⴱⴱ
.34ⴱⴱ
⫺.01
⫺.04
4
ⴱⴱ
⫺.40
.55ⴱ
.96/.89
⫺.36ⴱⴱ
.34ⴱⴱ
.42ⴱⴱ
.51ⴱⴱ
.33ⴱⴱ
⫺.02
.02
.09
.28ⴱⴱ
⫺.30
.97/.98
⫺.58ⴱⴱ
⫺.06
⫺.06
⫺.23ⴱⴱ
.01
.09
5
⫺.04
.14
.23ⴱ
⫺.60ⴱⴱ
.93/.88
.16ⴱ
.16ⴱ
.17ⴱ
.01
⫺.15ⴱ
6
7
ⴱⴱ
8
ⴱⴱ
9
ⴱⴱ
⫺.45 ⫺.31 ⫺.28
⫺.06
.21ⴱ
.50ⴱⴱ
.26ⴱⴱ
.24ⴱ
.47ⴱⴱ
.71ⴱⴱ
.40ⴱⴱ
.19ⴱ
⫺.13 ⫺.20ⴱ ⫺.21ⴱ ⫺.09
.20ⴱ
.28ⴱⴱ
.35ⴱⴱ ⫺.06
—
.45ⴱⴱ
.46ⴱⴱ
.05
.50ⴱⴱ .82/.84
.49ⴱⴱ
.17
.47ⴱⴱ
.43ⴱⴱ
—
.002
⫺.07
.06 ⫺.02
—
⫺.11
.07
.00
.06
10
⫺.03
.05
⫺.11
.13
⫺.17
⫺.06
⫺.07
⫺.09
.14
—
Note. N ⫽ 182. Alpha coefficients appear along the diagonal with Study 2 appearing first and Study 3 appearing second. Study 2 correlations appear below
the diagonal and Study 3 correlations appear above the diagonal. Alpha coefficients appear on the diagonal in italics.
ⴱ
p ⱕ .05. ⴱⴱ p ⬍ .001.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
diation is a better fit to the data. All results reported below are
standardized.
Anger. In support of H3a and H3b, witnessed uncivil mistreatment was positively related to anger (b ⫽ .72, p ⬍ .001), and anger
was negatively related to perpetrator positive treatment intentions
(b ⫽ ⫺0.27, p ⬍ .002), and positively related to perpetratordirected retaliation intentions (b ⫽ 0.25, p ⫽ .001). In addition,
anger mediated the relationship between witnessing uncivil mistreatment and perpetrator-directed positive treatment intentions
(PE7 ⫽ ⫺.19, SE ⫽ .08, 90% CI [⫺.33, ⫺.07]), and perpetratordirected retaliatory intentions (PE ⫽ .18, SE ⫽ .08, 90% CI [.07,
0.32]).
Further, anger fully mediated the relationship between witnessing
aggressive mistreatment and perpetrator-directed positive treatment
intentions (PE ⫽ ⫺.78, SD ⫽ .19, 90% CI [⫺1.11, ⫺.50]) and
perpetrator-directed retaliation intentions (PE ⫽ 0.76, SD ⫽ .19, 90%
CI [.48, 1.08]). Witnessed aggressive mistreatment was positively
related to anger (b ⫽ 2.88, p ⬍ .001), and anger was negatively
related to perpetrator positive treatment intentions (b ⫽ ⫺0.28, p ⬍
.001) and positively related to perpetrator-directed retaliation intentions (b ⫽ 0.27, p ⬍ .001).
Empathy. In support of H4a, H4b, and H4c, witnessed uncivil
mistreatment was positively related to empathy (b ⫽ .39, p ⬍
.001), and empathy was positively related to target support (b ⫽
0.60, p ⬍ .001), target evaluation (b ⫽ 0.98, p ⬍ .001), and tip
percent (b ⫽ 4.07, p ⬍ .001). In addition, empathy mediated the
relationship between witnessing uncivil mistreatment and target
support (PE ⫽ .23, SD ⫽ .07, 90% CI [.13, 0.35]), target evaluation (PE ⫽ .38, SD ⫽ .11, 90% CI [.21, 0.59]), and tip percent
(PE ⫽ 1.52, SD ⫽ .60, 90% CI [.72, 2.66]).
Further, empathy also mediated the relationship between witnessing aggressive mistreatment and target support (PE ⫽ 0.68,
SD ⫽ .17, 90% CI [.41, .96]), target evaluation (PE ⫽ 1.11, SD ⫽
.31, 90% CI [.61, 1.62]), and tip percent (PE ⫽ 5.06, SD ⫽ 1.86,
90% CI [2.17, 8.25). Witnessed aggressive mistreatment was positively related to empathy (b ⫽ 3.12, p ⬍ .001), empathy was
positively related to target support (b ⫽ .22, p ⬍ .001), target
evaluation (b ⫽ .36, p ⬍ .001), and tip percent (b ⫽ 1.64, p ⫽
.002).
Discussion
Study 2 replicated Study 1 by demonstrating that customers who
witness employee-directed mistreatment from other customers,
were more likely to compensate for their fellow customers’ bad
behavior by being more supportive toward the employee, rating
them higher on customer service, and leaving higher tips. Study 2
also extended Study 1 in a number of ways. First, we found that
witnesses develop negative behavioral intentions toward perpetrators,
consistent with research on witness reactions to insider workplace
mistreatment (e.g., Mitchell et al., 2015; Reich & Hershcovis, 2015).
Specifically, witnesses were less likely to endorse positive treatment intentions and more likely to endorse retaliatory intentions
against perpetrators. This may have negative implications for
customer experiences, because research shows that other customers can influence a customer’s satisfaction (Wu, 2007). Second, we
found that anger mediated this relationship. That is, consistent with
deontic theory, witnessing another customer mistreat a server gave
rise to anger in response to the moral violation, and led witnesses
1537
to develop negative behavioral intentions toward the perpetrator.
This extends prior research that has found that anger mediates the
relationship between insider mistreatment and perpetrator-directed
outcomes (e.g., Mitchell et al., 2015; Reich & Hershcovis, 2015),
by showing that unfair interactions anger other organizational
patrons, and this anger could translate to negative reactions by
customers toward other customers. This finding is important because it has negative implications for the service experience of
both customers.
Third, we examined a new deontic mediator, target-directed
empathy, and found that it mediated the relationship between
witnessing mistreatment and target-directed outcomes. Research to
date has largely focused on both consequences and emotions
directed at the perpetrator. This study extends the deontic relationship to the target, and acknowledges that benevolent affect directed
at the target yields positive target-directed outcomes. That is, there
are two possible means of redressing unfair treatment, one can
punish the perpetrator or one can aid the target, and different
affective mechanisms explain each of these outcomes. Moreover,
this finding helps to show that our results are not likely to be
explained by equity theory. Witnesses are not simply reducing
perpetrator and restoring target outcomes out of anger or inequity,
but also out of an other-oriented positive empathetic emotion
directed at targets.
Finally, we found that an increase in magnitude of the mistreatment generally corresponds to an increase in the magnitude of the
outcome. That is, witnesses of both incivility and aggression were
more likely than witnesses of a neutral situation to develop negative behavioral intentions toward the perpetrator and positive
intentions toward the target (except for the case of tipping where
a significantly higher tip percentage resulted from aggressive but
not uncivil misbehavior). Moreover, there were also significant
differences between the aggression and uncivil conditions such
that aggression yielded stronger negative responses toward perpetrators and supportive responses toward targets among witnesses.
Research has argued that there are conceptual differences between
incivility and aggression, but that these differences do not bear out
in the current measurement (Hershcovis, 2011). The present study
shows that with an increase in the intensity and intent of mistreatment, and a decrease in ambiguity, witness reactions are stronger.
However, it is important to note that even minor forms of mistreatment yielded significant negative consequences for perpetrators and positive outcomes for targets (in the form of supportive
behavioral intentions and target evaluations).
Although Studies 1 and 2 collectively demonstrate that customers punish perpetrators and compensate targets on behalf of badly
behaved fellow customers, a limitation of both studies is that we do
not vary the response of the target. In both studies, targets respond
neutrally regardless of customer treatment. Our results thus assume
that employees must maintain “service with a smile” to accrue the
“benefits” from customer mistreatment. In our final study, we test
this assumption by examining how witnesses react when targets
reciprocate uncivil treatment.
Reciprocity is a universally held social norm (Gouldner, 1960),
and negative reciprocal notions of “tit for tat” (Andersson &
Pearson, 1999; Cropanzano & Mitchell, 2005) or an “eye for an
7
PE stands for point estimate.
HERSHCOVIS AND BHATNAGAR
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1538
eye” personify many social exchanges. In the domain of customer
service interactions, van Jaarsveld, Walker, and Skarlicki (2010)
argue that mistreatment from customers can beget mistreatment
from employees. Perceptions of unfairness can instigate targets to
respond by also treating perpetrators disrespectfully. Customerdirected mistreatment can harm service provision and outcomes
such as customer retention (e.g., Schneider, Ehrhart, Mayer, Saltz,
& Niles-Jolly, 2005). Moreover, empathy is a deontic emotion that
explains why witnesses react more positively toward targets. If
targets take justice restoration into their own hands by behaving
rudely, we expect that witnesses will be less empathetic toward
targets and thus, less likely to offer support and compensation to
employees mistreated at the hands of other customers found in the
previous studies. Therefore, we hypothesized that:
Hypothesis 7 (H7): Customers who witness customer mistreatment toward a service provider will: be less likely to support the
target (H7a), less likely to evaluate the target positively (H7b),
and more likely to leave lower tips (H7c) when employees
respond uncivilly than when they respond neutrally.
Hypothesis 8 (H8): Target-directed empathy will mediate the
above relationships such that customers who witness customer
mistreatment toward a service provider will be less empathetic
and thus less likely to support the target (H8a), evaluate the target
positively (H8b), and leave higher tips (H8c) when employees
respond uncivilly than when they respond neutrally.
Moreover, we also expect that when employees fail to maintain
the “customer is always right” mantra, and instead behave uncivilly back to the customer, the employee has effectively punished
the perpetrator him- or herself. Therefore, moral anger is diminished and the deontic desire to hold the perpetrator accountable for
his or her misdeeds is likely to diminish, and we should see a
reduction in the witness’s negative behavioral intentions toward
the perpetrator.
Hypothesis 9 (H9): Customers who witness customer mistreatment toward a service provider will: be more likely to develop
positive treatment intentions (H9a), and less likely to develop
retaliatory intentions toward the perpetrator (H9b) when employees respond uncivilly than when they respond neutrally.
Hypothesis 10 (H10): Perpetrator-targeted anger will mediate
the above relationships such that customers who witness customer mistreatment toward a service provider will be less
angry and thus more likely to develop positive treatment
intentions (H10a), and less likely to develop retaliatory intentions toward the perpetrator (H10b) when employees respond
uncivilly than when they respond neutrally.
Study 3
Method
Participants. Following Study 2, participants were 119 people
residing in North America (51.4% female, M age ⫽ 34.7 years, SD ⫽
11.97, age range ⫽ 18 –73) recruited through the same online panel
provider (MTurk) in exchange for $2. Participants who completed
Study 2 were excluded from also being allowed to complete Study 3.
Participants worked on average 38.91 (SD ⫽ 8.56) hr per week, and
their mean tenure was 6.54 years (SD ⫽ 6.03). They held a wide range
of job titles such as social worker, occupational therapist, taxi dispatcher, electrician, and security consultant.
Materials and procedure. We conducted an online vignette
experiment in which we held the mistreatment condition constant, and
manipulated the employee response to be neutral versus uncivil.
Participants read a similar vignette as the vignette used in Study 2,
except we manipulated the server’s response. In this vignette, participants read about a customer’s aggressive treatment of a server at a
coffee shop, and either a neutral or uncivil response to this treatment
by the employee. We chose the aggressive condition because it would
be the most stringent test of our hypotheses as employees who
respond uncivilly in response to aggression may be more understandable than employees who respond uncivilly to a more minor form of
mistreatment. Participants then answered a series of questions aimed
to assess our dependent variables (see Appendix C for the vignette
scripts). We assessed perpetrator-directed positive treatment intentions, perpetrator-directed retaliatory intentions, target-directed support, target evaluation, tip percent, anger, and empathy using the same
measures as those used in Study 2. Ethics was granted by the Joint
Faculty Research Ethics Board at the University of Manitoba under an
amendment to the title “Witness Reactions to Customer Rudeness”
(protocol number J2014-072).
Results
Manipulation check. To test whether our manipulations
worked, we used the same two checks used in Study 2, except the
referent was the employee instead of the customer. We also had to
eliminate one item from the incivility manipulation check (i.e., “the
employee made the customer feel incompetent”), because this item
did not make sense with the customer as the frame of reference (␣ ⫽
.95 for incivility and .89 for lack of justice). We conducted a
MANOVA with employee incivility as our independent variable and
both manipulation checks as our dependent variable. The overall test
was significant, F(2, 116) ⫽ 60.08, p ⬍ .001, 2 ⫽ .51; therefore, we
proceeded to examine the between-subjects effects (see Table 4).
Table 4
Manipulation Checks and Main Effects for Target-Directed Outcomes (Study 3)
Target support
Condition
Neutral
Uncivil
n
M
SD
90% CI
Target evaluation
M
SD
90% CI
M
Tip percent
Rudeness manipulation
check
Lack of interpersonal
justice manipulation
check
SD
M
M
90% CI
SD
90% CI
SD
90% CI
59 5.84 .98 [5.60, 6.09] 6.34 1.90 [5.88, 6.80] 25.51 13.89 [22.53, 28.49] 1.22 .63 [.97, 1.46] 1.60 .82 [1.37, 1.83]
60 5.10 1.29 [4.85, 5.34] 4.20 2.31 [3.75, 4.65] 17.42 14.46 [13.89, 20.40] 2.98 1.49 [2.73, 3.23] 3.74 1.27 [3.57, 3.97]
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
Test of hypotheses.
Main effects for target-directed outcomes. Means, standard
deviations, and intercorrelations appear in Table 3. To investigate
Hypotheses 7a–7c, we conducted a MANOVA using employee response condition (uncivil vs. neutral) as our independent variable and
tip percentage, target support, and target evaluation as our dependent
variables. The overall test was significant, F(3, 115) ⫽ 11.01, p ⬍
.001, 2 ⫽ .22; therefore, we examined the between-subjects effects.
As in Studies 1 and 2, we conducted one-tailed tests. Means, standard
deviations, and confidence intervals are presented in Table 4.
Target support. In support of H7a, we found a significant
effect of the employee response condition on target support F(1,
118) ⫽ 12.64, p ⬍ .001, 2 ⫽ .10, Cohen’s d ⫽ .65.
Target evaluation. In support of H7b, we found a significant
effect of the employee response condition on target evaluation F(1,
118) ⫽ 30.46, p ⬍ .001, 2 ⫽ .04, Cohen’s d ⫽ 1.01.
Tip percent. In support of H7c, we found a significant effect
of the employee response condition on tip percent, F(1, 118) ⫽
10.24, p ⫽ .001, 2 ⫽ .08, Cohen’s d ⫽ .59.
Main effects for perpetrator-directed outcomes. To investigate H9a and H9b, we conducted a MANOVA using employee
response condition (uncivil vs. neutral) as our independent variable
and perpetrator positive treatment and retaliatory intentions as our
dependent variables. Contrary to our hypotheses, the overall test
was nonsignificant F(2, 116) ⫽ 0.52, p ⫽ .60. It therefore appears
that even when the employee reacts to customer aggression uncivilly, witnesses, despite tempering the support they express to the
targeted employee, retain their treatment and retaliatory intentions
toward the perpetrator. That is, although the injustice has been
rectified and equity has been restored by the employee, the deontic
emotion toward the perpetrator remains. We further test this assumption by examining the extent to which anger mediates our
relationships.
Mediation analyses. To test mediation we used Preacher and
Hayes process macros (Hayes, 2013, Model 4). In support of H8a,
empathy mediated the relationship between employee response
and employee-directed support (PE ⫽ ⫺.59, SE ⫽ .15, 90% CI
[⫺.88, ⫺.37]). Employee rudeness negatively related to empathy
(b ⫽ ⫺.87, p ⬍ .001) and empathy positively related to employee
treatment (b ⫽ .68, p ⬍ .001). In support of H8b, empathy
mediated the relationship between employee rudeness and
employee-directed evaluation (PE ⫽ ⫺.70, SE ⫽ .23, 90% CI
[⫺1.13, ⫺.38]). Employee rudeness negatively related to empathy
(b ⫽ ⫺.87, p ⬍ .001) and empathy positively related to employee
evaluation (b ⫽ .81, p ⬍ .001). In support of H8c, empathy
mediated the relationship between employee rudeness and employee tips (PE ⫽ ⫺3.55, SE ⫽ 1.33, 90% CI [⫺5.82, ⫺1.40]).
Employee rudeness negatively related to empathy (b ⫽ ⫺.87, p ⬍
.001) and empathy positively related to employee tips (b ⫽ 4.09,
p ⬍ .001). We failed to support H10a and H10b as anger did not
mediate the relationship between employee response and
perpetrator-directed outcomes.
Discussion
While Study 1 and Study 2 demonstrated that customers engage
in compensatory reactions to targets, and hold punishment intentions toward perpetrators, Study 3 examined the added effects of
employee reactions to the mistreatment. It is human nature to
1539
reciprocate negative treatment in kind (Cropanzano & Mitchell,
2005). Results of this study demonstrate that when targets are rude
in response to rude customers, witnessing customers are less
willing to support targets. These results demonstrate that people
feel less empathy toward victimized employees who are rude in
kind. Rather, they are more likely to support targets who are
unable/unwilling to defend themselves. This finding is consistent
with prior research which shows that witnesses are more likely to
act on vicarious injustice (e.g., help the target) when the target is
helpful/likable (De Cremer & Van Hiel, 2006). Thus, the behaviors
and reactions of a target appears to influence whether or not a
witness will offer support.
In addition, our results showed that when employees were
uncivil in response to aggressive treatment, witness empathy goes
down. That is, compared with the neutral response condition,
witnesses were less empathetic toward employees and were therefore less likely to engage in target-directed supportive and compensatory behaviors. This finding lends further support to the
deontic argument over the equity theory argument, because
empathy is a deontic emotion focused on the well-being of the
target.
Interestingly, however, the server’s uncivil response did not
affect witness negative reactions toward the misbehaving customer. That is, although the target’s uncivil reaction was
enough to reduce target-directed supporting behaviors, it was
not enough to reduce perpetrator-directed punishing behaviors.
Moreover, anger did not mediate the relationship between employee incivility and perpetrator-directed outcomes. These
combined findings suggest that target compensation is not sufficient to reduce anger and restore justice. That is, while customers may have felt less of a need to support and compensate
targets for their mistreatment when targets punished perpetrators directly, the target’s reactions were not enough to diminish
their anger and their desire to punish the poor behavior of a
perpetrator.
General Discussion
Previous research has focused on mistreatment between employees and found that witnesses tend to punish perpetrators (e.g.,
Mitchell et al., 2015; Skarlicki & Rupp, 2010; Turillo et al., 2002).
Research in the service domain has also examined mistreatment
between employees and found that when customers witness mistreatment, they evaluate the organization and its employees negatively (Porath et al., 2010). The present research is the first to
examine how customers react toward targeted employees and
perpetrators when they witness other customers mistreat these
employees. This research significantly contributes to the literature
in several ways.
First, our studies demonstrate a key role for customers in the
restoration of justice when a fellow customer mistreats an employee.
Whereas mistreatment between employees causes customers to perceive the organization more negatively, mistreatment by customers
toward employees appears to elicit greater positive affect toward
employees in the form of empathy, and negative affect toward the
perpetrating customer. Our results show that customers are willing to
help restore justice, both by rewarding and supporting the target, and
by punishing (or intending to punish) the perpetrator.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1540
HERSHCOVIS AND BHATNAGAR
Second, we make a theoretical contribution to the literature by
examining a target-directed pathway to justice restoration. Deonance theory argues that people ought to behave in certain ways,
and when they do not follow prescribed norms, witnesses will
become morally affronted (angry) and want to take action in the
form of perpetrator punishment. Thus, research on witnessing
mistreatment has focused on witness reactions toward perpetrators; and the few studies that have examined witness reactions
toward targets (Mitchell et al., 2015; Reich & Hershcovis, 2015)
have found mixed results. Mitchell et al. (2015) found that witnesses sometimes supported targets of supervisor mistreatment,
but only if they felt the mistreatment was unjustified. Reich and
Hershcovis (2015) found that witnesses did not support targets, but
instead punished perpetrators. However, Reich and Hershcovis
(2015) examined only anger as an explanatory mechanism. Consistent with deontic arguments, we show that witnesses do indeed
experience anger in response to transgressions and this anger
translates into perpetrator-directed reactions (e.g., less positive and
more retaliatory treatment). This may have negative implications
for the customer’s experience because negative CCI can adversely
affect customer satisfaction (Wu, 2007). However, we also introduce a second deontic emotion: target empathy. This emotion is
consistent with deontic reasoning because it motivates altruistic
action to reduce target distress. Thus, whereas moral anger yields
punitive perpetrator-directed action, moral empathy yields targetdirected action in the form of affective, cognitive, and behavioral
support aimed at redressing injustice. We contribute to deonance
theory by showing that deontic action may also take a positive
form (i.e., target support), and that empathy plays a role in explaining the relationship between an injustice and target-aiding.
Moreover, the pathway from witnessed mistreatment to targetand perpetrator-directed outcomes via empathy and anger, respectively, demonstrate that this is not an equity argument. Witnesses
are not simply rewarding mistreated employees for working harder
to deal with difficult customers. Rather, customer affective responses in the form of empathy and anger suggest an otheroriented process. The very fact that witnesses were angry and
empathetic about a fairness issue that had nothing to do with them
supports a deontic explanation for the findings.
Third, our results contribute empirically by showing that even
minor forms of mistreatment such as incivility trigger witness reactions. That is, even when mistreatment was low in intensity and
ambiguous in nature, witnesses supported targets, and evaluated them
more positively. These findings were even stronger when targets
experienced aggression. Targets in the aggressive group received
more affective support, higher evaluations, and higher tips than those
in the neutral or uncivil group. This finding contributes to the debate
in the literature that questions whether different forms of mistreatment
yield different outcomes. Hershcovis (2011) argued that although
mistreatment constructs are conceptually different, they are operationally similar, yielding the same outcomes. The present study shows
that when we operationalize the differences such that conceptual
definitions match operationalizations, the outcomes do indeed
differ. This finding suggests that researchers need to pay more
attention to how workplace mistreatment is being operationalized.
Distinctions between different forms of mistreatment are not being
captured in existing measurement, which has led to potentially
inaccurate results.
Fourth, our research makes a practical contribution by demonstrating that targets may unwittingly benefit from being victims of
mistreatment, but only if they maintain a neutral demeanor. Organizations reinforce the message that “customers are always right,”
which requires employees faced with rude customers to plaster a
smile on their face and take the abuse. Service sector organizations
often regulate employees’ externalized emotions via “display
rules” to enhance the customer experience (Rafaeli & Sutton,
1987), which can lead to employee burnout (Grandey, 2003). The
present research shows that, if employees are able to engage in
emotional labor by being neutral to uncivil customers, they may be
rewarded by other customers who witness the mistreatment; however, these rewards are tempered if they fail to regulate their
emotions and respond to rude customers in kind. It is possible that
the knowledge of these benefits (disseminated through training)
may make it easier for employees to endure unpleasant interactions—future research is needed to assess this possibility.
In addition, our research contributes by examining our questions
in the customer service context. Few studies have examined customer reactions to mistreatment, and those that have, considered
reactions to ostensibly justifiable employee-on-employee incivility
(e.g., delayed service, Porath et al., 2010) and found that customers
evaluate the organization and its employees negatively. Others
such as Mattila, Hanks, and Wang (2014) examined the impact of
observing other customers’ service failure on witnessing customers’ reactions and found that customer return intentions were lower
when they witnessed other customers’ poor past service experience
and recovery. Fairness and justice perceptions (Porath et al., 2010)
as well as co-customers’ demonstrated behaviors (e.g., Bitner,
Booms, & Tetreault, 1990; Kim & Lee, 2012) are key in customer
service evaluations. Miao and Mattila (2013) show that negative
customer behaviors (e.g., not keeping kids in line) negatively
impacts other customers’ emotional responses. Building on this
research, we contribute to this conversation by showing that customers punish fellow customers who mistreat service employees,
and compensate mistreated employees to make up for the injustice.
That is, the negative behaviors of fellow customers impact the
customer’s experience enough that they are willing to take action.
More broadly, this research has the potential to contribute to
research in the workplace mistreatment literature by considering the
role of empathy as a pathway to constructive intervention. Previous
research has found little or no relationship between witnessed mistreatment and target-directed positive action (e.g., Reich & Hershcovis, 2015), and Mitchell et al. (2015) even found that witnesses may
exclude targets from their scope of justice. Our findings demonstrate
that witnesses may engage in target-aiding behavior when they feel
empathy toward targets. Thus, the next step is to investigate what
organizations can do to encourage empathetic responses to mistreatment, and how to translate empathy into constructive responses that
support targets and discourage perpetrators.
Strengths and Limitations
There are a number of strengths to the study. First, to our knowledge, Study 1 provides the first field experiment that studies workplace mistreatment. This design allows us to draw strong inferences
about causality, and is likely to be generalizable to most service
interactions. Our findings with respect to tips were particularly interesting because the service interaction occurred in a fast-food restau-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
EMPLOYEE MISTREATMENT
rant where tipping is not the norm. These results may underestimate
the tipping behaviors that would occur in a full service restaurant.
Moreover, the results may extend to other service environments (e.g.,
retail), albeit such environments do not yield the same financial
rewards (i.e., tips). It would be interesting to investigate whether—in
nontip contexts— customers are more likely to make purchases from
employees who are mistreated by other customers. In addition, we
replicated our findings in a vignette design, and extended them by
examining affective mechanisms. Finally, we showed a boundary
condition to our effects; that is, customers only reward targets if they
maintain the “customer is always right” mantra. Regardless of employee reaction however, customers maintained their intent to punish
the perpetrator.
These results are not without limitations, most of which we discussed previously. In Study 1, we restricted our examination to
interactions where one customer was present (to minimize diffusion
of responsibility); however, this choice might affect generalizability
because other customers are often present and may influence each
other’s responses either positively or negatively. Study 2 and Study 3
used service vignettes that involved the presence of other customers
(there was a line-up), and compensatory target outcomes and punitive
perpetrator outcomes were found. Vignette designs are limited in
terms of their ecological validity because they ask participants to
imagine what they would do. In a real setting where there are strong
normative constraints on how customers react, it is possible that
witnesses would react in a more constrained manner. On the other
hand, real settings may in fact enhance a witness’s response because
people are likely to experience stronger emotional reactions to real
than imagined scenarios. Future research in busier retail settings in the
field can confirm whether witnesses would react the same way when
other customers are present, intensify these behaviors as a consequence of contagion or mob effects, or whether diffusion of responsibility would hinder reactions to employee mistreatment. It is also
possible that in very busy stores, where employees are doing their best
to cope with heavy workloads, customers feel more empathetic toward mistreated employees and angry at the offending customers.
Future research that manipulates social presence in the form of no
other customers, some other customers, and many other customers
would shed light on these relationships. Further, these studies examined only two possible employee reactions, neutral or uncivil. However, the server might instead call a supervisor for help, refuse service,
or apologize to the customer to appease him/her. Future research
should examine how different reactions influence witness responses.
And, we manipulated injustice but our manipulation check measured
interpersonal justice rather than injustice. Our incivility manipulation
check confirmed that participants perceived a negative interaction, but
follow-up research should specifically use a measure of injustice.
Finally, services research demonstrates that the service scape
(Schmidt & Sapsford, 1995) and store atmospherics (Moore et al.,
2005) may influence the types of customer-to-customer interactions
that ensue. Additional research can examine the design of appropriate
service scapes. For instance, the presence versus absence of in-store
signage that highlights respectful treatment of service employees and
the denial of service in the face of mistreatment may positively impact
witnessing customers’ willingness to actually intervene and buffer
employees against lowered support when they retaliate against mistreatment in kind.
1541
Conclusion
This is the first study to link customer mistreatment of
employees with witnessing customer reactions toward victimized employees. Our findings show that customers support
mistreated employees affectively, cognitively, and behaviorally, suggesting that although customers are not always right,
treating them as if they are may have hidden benefits.
References
Adams, J. S. (1963). Toward an understanding of inequity. Journal of
Abnormal and Social Psychology, 67, 422– 436. http://dx.doi.org/10
.1037/h0040968
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Thousand Oaks, CA: Sage.
Andersson, L. M., & Pearson, C. M. (1999). Tit for tat? The spiraling effect
of incivility in the workplace. Academy of Management Review, 24,
452– 471.
Austin, P. C., & Hux, J. E. (2002). A brief note on overlapping confidence
intervals. Journal of Vascular Surgery, 36, 194 –195.
Baron, R. A., & Neuman, J. H. (1996). Workplace violence and workplace
aggression: Evidence of their relative frequency and potential causes.
Aggressive Behavior, 22, 161–173. http://dx.doi.org/10.1002/
(SICI)1098-2337(1996)22:3⬍161::AID-AB1⬎3.0.CO;2-Q
Batson, C. D., Dyck, J. L., Brandt, J. R., Batson, J. G., Powell, A. L.,
McMaster, M. R., & Griffitt, C. (1988). Five studies testing two new
egoistic alternatives to the empathy-altruism hypothesis. Journal of
Personality and Social Psychology, 55, 52–77. http://dx.doi.org/10
.1037/0022-3514.55.1.52
Batson, C. D., Fultz, J., & Schoenrade, P. A. (1987). Distress and empathy:
Two qualitatively distinct vicarious emotions with different motivational
consequences. Journal of Personality, 55, 19 –39. http://dx.doi.org/10
.1111/j.1467-6494.1987.tb00426.x
Bennett, R. J., & Robinson, S. L. (2000). Development of a measure of
workplace deviance. Journal of Applied Psychology, 85, 349.
Beugré, C. D. (2010). Resistance to socialization into organizational corruption: A model of deontic justice. Journal of Business and Psychology,
25, 533–541.
Bies, R. J., & Moag, J. S. (1986). Interactional justice: Communication
criteria of fairness. Research on negotiation in organizations, 1, 43–55.
Bitner, M. J., Booms, B. H., & Tetreault, M. S. (1990). The service
encounter: Diagnosing favorable and unfavorable incidents. Journal of
Marketing, 54, 71– 84. http://dx.doi.org/10.2307/1252174
Colquitt, J. A. (2001). On the dimensionality of organizational justice: A
construct validation of a measure. Journal of Applied Psychology, 86,
386 – 400. http://dx.doi.org/10.1037/0021-9010.86.3.386
Cortina, L. M. (2008). Unseen injustice: Incivility as modern discrimination in organizations. Academy of Management Review, 33, 55–75.
http://dx.doi.org/10.5465/AMR.2008.27745097
Cortina, L. M., Kabat-Farr, D., Leskinen, E. A., Huerta, M., & Magley,
V. J. (2013). Selective incivility as modern discrimination in organizations evidence and impact. Journal of Management, 39, 1579 –1605.
http://dx.doi.org/10.1177/0149206311418835
Cortina, L. M., Magley, V. J., Williams, J. H., & Langhout, R. D. (2001).
Incivility in the workplace: Incidence and impact. Journal of Occupational Health Psychology, 6, 64 – 80. http://dx.doi.org/10.1037/10768998.6.1.64
Cropanzano, R., & Mitchell, M. S. (2005). Social exchange theory: An
interdisciplinary review. Journal of Management, 31, 874 –900. http://
dx.doi.org/10.1177/0149206305279602
Darley, J. M., & Latané, B. (1968). Bystander intervention in emergencies:
Diffusion of responsibility. Journal of Personality and Social Psychology, 8, 377–383. http://dx.doi.org/10.1037/h0025589
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
1542
HERSHCOVIS AND BHATNAGAR
De Cremer, D., & Van Hiel, A. (2006). Effects of another person’s fair
treatment on one’s own emotions and behaviors: The moderating role of
how much the other cares for you. Organizational Behavior and Human
Decision Processes, 100, 231–249.
Ducharme, L. J. & Martin, J. K. (2000). Unrewarding work, coworker
support, and job satisfaction. Work and Occupations, 27, 223–243.
Eiglier, P., & Langeard, E. (1977). Services as systems: Marketing implications. In P. Eiglier (Ed.), Marketing consumer services: New insights
(pp. 83–103). Cambridge, MA: Marketing Science.
Folger, R. (2001). Fairness as deonance. In S. W. Gilliland, D. D. Steiner,
& D. P. Skarlicki (Eds.), Research in social issues in management (pp.
3–31). Charlotte, NC: Information Age.
Folger, R., Ganegoda, D. B., Rice, D. B., Taylor, R., & Wo, D. X. H.
(2013). Bounded autonomy and behavioral ethics: Deonance and reactance as competing motives. Human Relations, 66, 905–924. http://dx
.doi.org/10.1177/0018726713482013
Gouldner, A. W. (1960). The norm of reciprocity: A preliminary statement.
American Sociological Review, 25, 161–178. http://dx.doi.org/10.2307/
2092623
Grandey, A. A. (2003). When “the show must go on:” Surface acting and
deep acting as determinants of emotional exhaustion and peer-rated
service delivery. Academy of Management Journal, 46, 86 –96. http://
dx.doi.org/10.2307/30040678
Grandey, A. A., Dickter, D. N., & Sin, H. P. (2004). The customer is not
always right: Customer aggression and emotion regulation of service
employees. Journal of Organizational Behavior, 25, 397– 418. http://dx
.doi.org/10.1002/job.252
Grandey, A. A., Kern, J. H., & Frone, M. R. (2007). Verbal abuse from
outsiders versus ...
Purchase answer to see full
attachment