Cleansing My Abuse: A Reparative Response Model of Perpetrating Abusive Supervisor Behavior

Abstract

Research on abusive supervision has predominantly focused on the consequences for victims while overlooking how leaders respond to their own abusive behavior. Drawing from the literature on moral cleansing, we posit that supervisors who engage in abusive behavior may paradoxically engage in more constructive leadership behaviors subsequently as a result of feeling guilty and perceiving loss of moral credits. Results from two experience sampling studies show that, within leaders on a daily basis, perpetrating abusive supervisor behavior led to an increase in experienced guilt and perceived loss of moral credits, which in turn motivated leaders to engage in more constructive person-oriented (consideration) and task-oriented (initiating structure) leadership behaviors. In addition, leader moral attentiveness and moral courage strengthen these indirect effects by amplifying leaders’ awareness of their immoral behavior and their willingness and determination to make reparations for such behavior. Our research contributes to the theoretical understanding of leaders’ responses toward their own abusive supervisor behavior and provides insights into how and when destructive leadership behaviors may, paradoxically, trigger more constructive behaviors.

Full text

1
Cleansing My Abuse:
A Reparative Response Model of Perpetrating Abusive Supervisor Behavior
ZHENYU LIAO
Olin Business School
Washington University in St. Louis
Email: z.liao@wustl.edu
KAI CHI YAM
NUS Business School
National University of Singapore
Email: bizykc@nus.edu.sg
RUSSELL E. JOHNSON
Eli Broad College of Business
Michigan State University
Email: johnsonr@broad.msu.edu
WU LIU
Faculty of Business
The Hong Kong Polytechnic University
Email: wu.liu@polyu.edu.hk
ZHAOLI SONG
NUS Business School
National University of Singapore
Email: songzl@nus.edu.sg
Authors’ Notes: We thank Herman Aguinis, Kristopher Preacher, and Sean Lane for helpful suggestions
on statistics. We also thank participants of the Organizational Behavior seminar series in the Olin
Business School at the Washington University in St. Louis for helpful feedback on an earlier draft of this
manuscript. An earlier version of this manuscript was presented at the 77th Annual Meeting of the
Academy of Management, Atlanta.
Grant Acknowledgement: This research was supported by Singapore Ministry of Education Research
Grant R-317-000-110-112 and the Research Grants Council of the Hong Kong Special Administrative
Region, China under Grant RGC NO. PolyU155006/15B.
Correspondence concerning this article should be addressed to Zhenyu Liao, Washington University in St.
Louis, Olin Business School, Campus Box 1156, One Brookings Drive, Saint Louis, MO 63130-4899.
Email: z.liao@wustl.edu

2
Cleansing My Abuse:
A Reparative Response Model of Perpetrating Abusive Supervisor Behavior
Abstract
Research on abusive supervision has predominantly focused on the consequences for victims
while overlooking how leaders respond to their own abusive behavior. Drawing from the
literature on moral cleansing, we posit that supervisors who engage in abusive behavior may
paradoxically engage in more constructive leadership behaviors subsequently as a result of
feeling guilty and perceiving loss of moral credits. Results from two experience sampling studies
show that, within leaders on a daily basis, perpetrating abusive supervisor behavior led to an
increase in experienced guilt and perceived loss of moral credits, which in turn motivated leaders
to engage in more constructive person-oriented (consideration) and task-oriented (initiating
structure) leadership behaviors. In addition, leader moral attentiveness and moral courage
strengthen these indirect effects by amplifying leaders’ awareness of their immoral behavior and
their willingness and determination to make reparations for such behavior. Our research
contributes to the theoretical understanding of leaders’ responses toward their own abusive
supervisor behavior and provides insights into how and when destructive leadership behaviors
may, paradoxically, trigger more constructive behaviors.
Keywords: Leadership; Abusive supervisor behavior; Moral attentiveness; Moral courage;
Moral credits; Guilt

3
Research on abusive supervision has held an important place in the leadership literature
over the past two decades (Mackey, Frieder, Brees, & Martinko, 2015; Tepper, 2007). Following
Tepper’s (2000, p.178) seminal work that defines abusive supervision as leaders’ “sustained
display of hostile verbal and nonverbal behavior, excluding physical contact,” much of extant
research has taken a between-person static approach and suggested that some leaders are
generally more abusive than others and that abusive supervision has deleterious effects on
various follower outcomes, including psychological well-being, task performance, and helping
behaviors (for meta-analytic reviews, see Mackey et al., 2015; Zhang & Liao, 2015). Extending
this research, leadership scholars have recently investigated abusive supervisor behavior as a
within-person dynamic phenomenon that fluctuates on a daily basis (e.g., Barnes, Lucianetti,
Bhave, & Christian, 2015; Courtright, Gardner, Smith, McCormick, & Colbert, 2016; Johnson,
Venus, Lanaj, Mao, & Chang, 2012). This stream of research establishes a more holistic
understanding regarding why and how abusive supervisor behavior occurs from one to the next
day and how it shapes subordinates’ momentary work states and behaviors.
Although prior studies have provided valuable insights into how abusive supervision
negatively affects subordinates from both between- and within-person perspectives, research to
date has overlooked how displaying abusive supervisor behavior affects the leaders themselves
and their subsequent behaviors toward followers. Leader behaviors do not occur in a vacuum;
rather, how leaders behave at one moment can shape the way in which they behave later (Dinh,
Lord, Gardner, Meuser, Liden, & Hu, 2014; Lin, Ma, & Johnson, 2016). For example, following
displays of abusive behavior, some leaders may be motivated to make amends by exhibiting
more constructive behavior toward followers. This possibility is hinted at by Ilies, Peng, Savani,
and Dimotakis (2013), who found that when employees are made aware of their prior

4
counterproductive work behaviors, they are subsequently more likely to engage in organizational
citizenship behavior. Owing to the lack of attention paid to the consequences of abusive
supervisor behavior for leaders, it is not clear how leaders react to their own abusive behavior
and what mechanisms drive those reactions.
In this research, we diverge from the “victim-centric perspective” of abusive supervision
and instead adopt a “perpetrator-centric perspective” to explore how leaders respond to their
own momentary display of abusive supervisor behavior. To do so, we draw from moral cleansing
theory and propose that after perpetrating abusive supervisor behavior toward specific followers,
leaders emotionally feel guilty and cognitively perceive a deficit of moral credits, because
behaving abusively is a moral transgression that violates moral norms and jeopardizes leaders’
moral self-concept (Miller & Effron, 2010; Sachdeva, Iliev, & Medin, 2009; Zhong, Liljenquist,
& Cain, 2009). According to moral cleansing theory, such negative emotional and cognitive
experiences may motivate leaders to subsequently engage in reparative actions toward abused
followers as a way to compensate for prior wrongdoings (Zhong et al., 2009).
What reparative actions will leaders perform to the followers in the aftermath of abusive
supervisor behavior? Moral cleansing theory suggests that people engage in reparative actions
that compensate victims and counteract the threatened moral self-concept (West & Zhong, 2015).
Demonstrating the ethics of care and justice are two core components of people’s moral concerns
(Gilligan, 1982; Kohlberg, 1971), yet displaying abusive supervisor behavior undermines
followers’ well-being and justice experience, and thus runs counter to these fundamental moral
concerns (Brown & Mitchell, 2010). Consequently, leaders may engage in person-oriented
behaviors, which include consideration leadership acts such as showing concern and support, to
improve followers’ well-being and reestablish the moral concept of care (Judge, Piccolo, & Ilies,

5
2004). Leaders may also engage in task-oriented behaviors, which include initiating structure
acts such as providing accurate guidance and feedback, to reestablish the moral concept of
justice. We therefore focus on consideration and initiating structure behaviors as fundamental
reparative actions that leaders may perform in response to experienced guilt and perceived loss
of moral credits arising from perpetrating abusive supervisor behavior.
In addition to proposing these reparative effects, we suggest that moral cleansing
responses are contingent on leaders’ tendencies to reflect on morality in their life experiences
(i.e., moral attentiveness; Reynolds, 2008) and willpower to uphold moral principles (i.e., moral
courage; Sekerka, Bagozzi, & Charnigo, 2009). As the behavioral ethics literature suggests,
individuals high in moral attentiveness are more aware of the moral elements of their behaviors
(Reynolds, 2008) and those high in moral courage are more likely to feel responsible and
motivated to resolve moral challenges by overcoming internal and external oppositions (Sekerka
et al., 2009). We propose that in the aftermath of abusive supervisor behavior, morally attentive
leaders are more likely to experience guilt and loss of moral credits, and morally courageous
leaders are more likely to act on their experienced guilt and perceived loss of moral credits by
taking reparative actions (see Figure 1 for our theoretical model). To capture the within-leader
dynamic behavioral shift from destructive to constructive behaviors, we conducted two field
studies using experience-sampling methodology (ESM; Bolger, Davis, & Rafaeli, 2003).
Our research makes three theoretical contributions. First, we extend abusive supervision
research by shifting the focus from a victim-centric to a perpetrator-centric perspective. By
investigating why and when perpetrating abusive supervisor behavior leads to more constructive
leadership behavior, we provide a finer-grained understanding of the consequences of abusive
supervision. Second, we contribute to the leadership literature more generally by exploring how

6
different types of leadership behaviors are interrelated, even those that ostensibly contradict one
another (e.g., abusive supervision and consideration). We highlight that seemingly incompatible
leadership behaviors can exist within the same leaders, shedding light on paradoxical leadership
patterns (Denison, Hooijberg, & Quinn, 1995; Smith & Lewis, 2011). Additionally, by
examining within-leader variation in leadership behaviors on a momentary basis, we respond to
calls for taking a more dynamic process perspective in leadership research (Dinh et al., 2014).
Third, we complement theoretical knowledge and empirical evidence of moral cleansing theory
by using a more granular approach to examine momentary emotional and cognitive experiences
as underlying mechanisms and leader moral attentiveness and courage as boundary conditions of
the moral cleansing process (Zhong et al., 2009).
Theoretical Development and Hypotheses
Abusive Supervisor Behavior as an Exemplar of Moral Transgression
Being moral is an important aspect of a person’s self-concept (Aquino & Reed, 2002) and
people thus strive to act congruently with moral norms and principles (Treviño, Weaver, &
Reynolds, 2006). Although people’s understanding of morality varies markedly and the content
of the moral domain includes canonical yet disparate elements, scholars have identified care and
justice as two core components of moral concerns (Gilligan, 1982; Kohlberg, 1971). In a similar
vein, management researchers have generally considered “organizational ethics in terms of
fairness (avoiding harm to and caring for others)” (Weaver, Reynolds, & Brown, 2014, p.113).
To enhance the caring facet of the morality, leaders are encouraged to be considerate and helpful
to followers, and show concern for followers’ personal welfare and psychological well-being
(Fehr, Yam, & Dang, 2015; Stogdill, Goode, & Day, 1962). To boost the just facet of morality,
leaders should establish transparent communication channels, maintain uniform performance

7
standards, and provide comprehensive and timely feedback to followers (Fehr et al., 2015). Such
caring and just behaviors facilitate interpersonal interaction and group functioning, helping
leaders maintain a positive sense of morality within organizations (Aquino & Reed, 2002).
Nevertheless, due to conflicts of interest or breakdowns in self-regulation, leaders may
occasionally exhibit abusive behaviors that violate moral norms and principles (Barnes et al.,
2015; Courtright et al., 2016; Lin et al., 2016). As exemplars of moral transgressions (Brown &
Mitchell, 2010), the daily display of verbal and nonverbal interpersonal mistreatment exerts
negative consequences on followers and threatens leaders’ moral self-concept (Aquino & Reed,
2002; West & Zhong, 2015). Perpetrating abusive supervisor behavior, such as deriding
followers in front of others, badgering followers about their past mistakes, or making fun of
followers, harms subordinates’ psychological well-being (e.g., increasing followers’ anxiety and
depression and decreasing their job and life satisfaction; Mackey et al., 2015). Such actions are
in direct contradiction to leadership behaviors that signify moral caring and thus threaten the
caring facet of leaders’ moral self. Additionally, various examples of abusive supervisor
behavior, such as not giving followers credit for difficult jobs requiring a great deal of effort,
telling followers their suggestions are stupid, or ignoring or ostracizing followers, elicit
perceptions of injustice in followers (Tepper, 2007), and thus detract from the just facet of
leaders’ moral self. Taken together, displaying abusive supervisor behavior is a form of moral
transgression that threatens the care and justice cores that comprise leaders’ moral self-concept.
According to moral cleansing theory, people monitor fluctuations in their sense of
morality and endeavor to maintain a favourable moral self-concept by balancing their moral and
immoral actions (Nisan, 1990; Zhong et al., 2009). Behaving in immoral and norm-violating
ways, such as abusive supervisor behavior, dampens leaders’ moral self-concept, prompting

8
them to subsequently engage in compensatory behaviors that help “cleanse” the emotional and
cognitive feelings of immorality and restore their moral self-concept (Jordan, Mullen, &
Murnighan, 2011; Sachdeva et al., 2009). In the following sections, we first develop our
hypotheses regarding the parallel mediating roles of experienced guilt and perceived loss of
moral credits in the relationships between perpetrating abusive supervisor behavior and
reparative actions. Then, we articulate how leader moral attentiveness and courage moderate
these mediating effects at different stages.
Emotional and Cognitive Responses to Perpetrating Abusive Supervisor Behavior
Moral cleansing theory suggests that there are emotional and cognitive ramifications for
perpetrating abusive supervisor behavior (Zhong et al., 2009). Guilt is a self-conscious moral
emotion that arises when individuals perceive that they have violated moral norms and standards
(Baumeister, Stillwell, & Heatherton, 1994; Tangney, 1990). The elicitation of guilt involves a
negative moral self-appraisal process in which individuals realize that their preceding moral
transgressions result in other people’s negative consequences and attribute the occurrence of
moral transgressions to factors under their control (Tangney, 1990; Tracy & Robins, 2006).
Displaying abusive supervisor behavior results in deleterious effects on followers’ well-being
and justice perceptions, dampening the care and justice components of leaders’ sense of morality
(Brown & Mitchell, 2010). Moreover, although followers’ attributes, work behaviors, and job
performance may trigger supervisory abuse (e.g., Tepper, Moss, & Duffy, 2011), leaders, as
perpetrators of abusive supervisor behavior, ultimately control whether they respond
destructively by perpetrating interpersonal abuse or constructively by providing useful feedback
and coaching. Thus, leaders are likely to perceive that they are primarily responsible for the

9
abusive behavior enacted toward followers during the workday (i.e., internal causal attribution).
Such a negative self-evaluation process will induce a momentary feeling of guilt in leaders.
Hypothesis 1: Within leaders, perpetrating abusive supervisor behavior is positively
related to an increase in experienced guilt.
In addition to feeling guilty, moral cleansing theory also suggests that individuals
cognitively perceive a deficit of moral credits in the aftermath of moral transgressions (Miller &
Effron, 2010; Zhong et al., 2009). Akin to a moral bank account, individuals accumulate moral
credits for their good deeds, which individuals can then withdraw to ‘pay’ for future bad deeds
(Lin et al., 2016; Yam, Klotz, He, & Reynolds, 2017). Conversely, individuals may feel morally
deficient following moral transgressions that cause a shortfall of moral credits. Such feelings
prompt subsequent moral behaviors to restore their moral credits (Jordan et al., 2011; Sachdeva
et al., 2009). Behavioral ethics scholars have acknowledged moral credits as a quantitative
hallmark of moral self-concept: the number of moral credits a person has at a given moment
represents how he/she perceives his/her levels of moral self-concept (Miller & Effron, 2010;
Zhong et al., 2009). Thus, the fluctuation of moral credits in response to one’s moral and
immoral behaviors captures the waxing and waning of individuals’ moral self-concept.
We propose that perpetrating abusive supervisor behavior evokes leaders’ momentary
perception of a loss of moral credits. As we discussed earlier, engaging in abusive supervisor
behavior is a blatant and obvious exemplar of moral transgression, jeopardizing leaders’ moral
self-concept. Such deprived moral self-concept echoes leaders’ cognitively perceived loss of
moral credits. Supporting our proposition, a handful of empirical studies demonstrate that prior
moral transgressions create shortfalls in moral credits. Zhong and colleagues (2010), for
example, suggested that people experience morally deficit after making unethical choice.
Similarly, Gino and colleagues (2015) showed that lying to others makes individuals feel

10
immorally. Taken together, we propose that leaders may perceive a loss of moral credits in
response to their abusive supervisor behavior.
Hypothesis 2: Within leaders, perpetrating abusive supervisor behavior is positively
related to an increase in perceived loss of moral credits.
Behavioral Responses to Experienced Guilt and Perceived Loss of Moral Credits
What consequences await leaders who feel guilty and perceive a deficit of moral credits
after perpetrating abusive behavior? Moral cleansing theory suggests that leaders who
emotionally and cognitively feel immoral are motivated to perform reparative actions in order to
compensate the abused followers and thus re-establish a favourable moral self-concept (West &
Zhong, 2015; Zhong et al., 2009). Considering that the major role and responsibility of leaders is
to provide support and guidance to followers as they complete work tasks, the most accessible
and role-appropriate reparative actions for leaders to engage in after abusive supervisor behavior
are consideration and initiating structure behaviors.
Consideration refers to leadership behavior that involves showing concern and respect
for followers, praising and recognizing followers, and looking out for the welfare of followers.
Initiating structure refers to leadership behavior that involves clarifying work goals and role
expectations, maintaining open channels of communication about tasks and procedures, and
providing guidance and feedback to help followers excel (Fleishman & Harris, 1962; Judge et
al., 2004). Consideration and initiating structure are two fundamental dimensions of leadership
behavior that provide the most direct and leadership-specific channels for leaders to give support
and guidance to followers. These behaviors are “among the most robust of leadership concepts”
(Fleishman, 1995, p.51) and they contribute to many paramount criteria of leadership
effectiveness, including the psychological well-being, justice perceptions, and job performance
of followers (Judge et al., 2004). Given these constructive consequences for followers, leaders

11
who experience guilt and perceive a deficit of moral credits may be particularly motivated to
display these reparative behaviors as a way of cleansing the emotional and cognitive fallout to
their own immoral behavior.
We suggest that consideration and initiating structure behaviors help restore leaders’ care
and justice facets of morality, respectively. Consideration behavior is a clear behavioral
manifestation of care, as it emphasizes a deep concern for followers and focuses on expressing
rapport and providing help needed by followers (Fleishman & Harris, 1962). Initiating structure
behavior is oriented toward providing followers with sufficient guidance, communicating openly
on role expectations, and imparting feedback to help followers improve (Stogdill et al., 1962),
which enhance followers’ perceptions of openness, transparency, and justice (Judge et al., 2004).
Prior research on morality and leadership together with our above discussion suggest that
behaviors reflective of consideration and initiating structure are indicators of leaders’ morality in
terms of care and justice (Fehr et al., 2015; Weaver et al., 2014). Consequently, leaders can
leverage consideration and initiating structure behaviors as ways to lessen any perceived threats
to the caring and just components of their moral self-concepts. We therefore propose experienced
guilt and perceived loss of moral credits as parallel underlying mechanisms that explain the
positive links between perpetrating abusive supervisor behavior and reparative actions.
Hypothesis 3: Perpetrating abusive supervisor behavior has positive indirect
relationships with leaders’ subsequent (a) consideration and (b) initiating structure
behaviors via experienced guilt.
Hypothesis 4: Perpetrating abusive supervisor behavior has positive indirect
relationships with leaders’ subsequent (a) consideration and (b) initiating structure
behaviors via perceived loss of moral credits.
The Moderating Effects of Moral Attentiveness and Moral Courage

12
Our theoretical model thus far highlights the roles of experienced guilt and perceived loss
of moral credits in the links between daily abusive supervision and reparative actions. It is
important to recognize that these parallel mediation effects vary across leaders depending on
their moral character (Mullen & Monin, 2016). Drawing on behavioral ethics research, we
identify leader moral attentiveness and courage as two theoretically relevant dimensions of moral
character. These dimensions are relevant because they impact the extent to which individuals
attend to the morality of their behaviors and fluctuations in their moral self-concept (moral
attentiveness), and the extent to which they are willing to uphold moral principles by taking
action in response to their immoral acts (moral courage).
Moral attentiveness is defined as “the extent to which an individual chronically perceives
and considers morality and moral elements in his or her experiences” (Reynolds, 2008, p. 1028).
It captures an innate sensitivity in recognizing moral issues. Moral attentiveness includes “a
perceptual aspect in which information is automatically colored as it is encountered and a more
intentional reflective aspect by which the individual uses morality to reflect on and examine
experiences” (p.1028). Perceptual moral attentiveness focuses on the recognition of moral issues
now, whereas reflective moral attentiveness involves a reflective cognitive process of
considering and examining past moral experiences. Prior research has demonstrated that
reflective moral attentiveness is particularly critical to processing preceding moral transgressions
and it affects subsequent moral experiences and behaviors (Reynolds, 2008). Given that we are
interested in how leaders respond to their prior abusive supervisor behavior, we thus focus our
theorizing on the moderating role of reflective moral attentiveness.
We argue that leaders high in reflective moral attentiveness will experience higher levels
of experienced guilt and perceived loss of moral credits, compared to those low in trait moral

13
attentiveness. Morally attentive leaders tend to think about ethics in general and the moral
aspects of their decisions and behaviors at work in particular (Reynolds, 2008). Such leaders are
more cognizant of the moral content of their behaviors. In other words, leaders high in moral
attentiveness are more likely to re-examine what they had done incorrectly to their followers
during work interactions. Reflecting on their prior wrongdoings, morally attentive leaders are
more likely to vividly imagine how their behaviors harm the abused followers (Whitaker &
Godwin, 2013). Such reflections are likely to result in a stronger deprived moral self-concept, as
abusive supervisor behaviors are blatant violations of the ethics of care and justice, which
translates into higher levels of experienced guilt and perceived loss of moral credits. In contrast,
leaders low in moral attentiveness are less aware of and thus might more easily forget their daily
abusive supervisor behavior. Research indeed suggests that memories of unethical behavior can
be easily forgotten, compared to amoral memories (Kouchaki & Gino, 2016). Even if they do
recall their abusive supervisor behavior, morally inattentive leaders are unlikely to spend time
and reflect on how their behaviors might have actually harmed the abused followers (Reynolds,
2008), leading to less intense feelings of guilt and perceived loss of moral credits. We thus posit
the following first-stage moderation hypothesis.
Hypothesis 5: Leader moral attentiveness moderates the positive relationships of
perpetrating abusive supervisor behavior with (a) experienced guilt and (b) perceived
loss of moral credits, such that these relationships are stronger when leader moral
attentiveness is high (vs. low).
Our second moderator is leader moral courage, broadly defined as a moral character
strength that provides the fortitude to engage in moral actions by overcoming internal resistance
and external threats when facing moral dilemmas, including those caused by one’s own immoral
behaviors (Kidder, 2003; Sekerka et al., 2009). The core of moral courage is an inner willingness
that drives a powerful impetus for translating moral intentions into moral behaviors and

14
maintaining moral principles (Hannah, Avolio, & May, 2011). Morally courageous individuals
are especially motivated to make amends when they feel immoral owing to their own
transgressions. In short, moral courage enables leaders to translate moral intentions to moral
behaviors, especially when perceiving immorality of their behaviors (Sekerka et al., 2009).
We expect leader moral courage to strengthen the relationships of experienced guilt and
perceived loss of moral credits with constructive leadership behaviors. Although guilt and
perceived moral credits usually drive leaders’ direct compensatory actions toward abused
followers (Baumeister et al., 1994; Zhong et al., 2010), many organizational and individual
factors may prevent leaders from readily enacting such actions. For example, by exhibiting
compensatory behaviors that are contradictory to prior abusive supervisor behaviors, it is an
implicit and public signal of leaders admitting their wrongdoing, which can threaten their
perceived effectiveness as a leader. Therefore, any reparative action puts leaders at risk for losing
power and status within the workgroup and they may be in fear of being deprived respect and
authority from followers (Kidder, 2003; Sekerka et al., 2009). Given these adverse
consequences, leaders might find it more expedient to overlook their feelings of guilt and their
perceived loss of moral credits by refraining from any sort of reparative action. Even if leaders
do not overlook their feelings of guilt and perceived loss of moral credits, they might find it
easier to assuage feelings of guilt and perceived loss of moral credits by engaging in
compensatory behaviors outside of the organization (e.g., toward friends or family; Jordan et al.,
2011; Sachdeva et al., 2009), thus bypassing any potential threats to their status, power, or
authority at work. Alternatively, leaders may utilize self-serving cognitive reframing to justify or
excuse their prior abusive behaviors (e.g., as a form of negative feedback and to motivate higher
performance; Shalvi, Gino, Barkan, & Ayal, 2015).

15
However, high moral courage provides leaders with a fortitude to overcome these
perceived threats and challenges. Leaders with high moral courage will not be deterred by the
risks that come from admitting to being abusive toward followers nor will they opt to justify,
defend, or overlook their moral misdeeds. Rather, such leaders will directly confront their
immoral deeds by acknowledging and admitting responsibility for them (Sekerka et al., 2009).
Leaders with high moral courage will therefore be more likely to initiate more direct and
effortful compensatory actions as the primary strategy to assuage their felt guilt and perceived
loss of moral credits. We posit the following second-stage moderation hypothesis.
Hypothesis 6: Leader moral courage moderates the positive relationships of experienced
guilt with (a) consideration and (b) initiating structure behaviors, such that these
relationships are stronger when leader moral courage is high (vs. low).
Hypothesis 7: Leader moral courage moderates the positive relationships of perceived
loss of moral credits with (a) consideration and (b) initiating structure behaviors, such
that these relationships are stronger when leader moral courage is high (vs. low).
Integrating our theoretical arguments presented in Hypotheses 5 through 7, we further
propose the following dual-stage moderated mediation hypotheses.
Hypothesis 8: The positive indirect effects of perpetrating abusive supervisor behavior
on (a) consideration and (b) initiating structure behaviors via experienced guilt are the
strongest when leader moral attentiveness and moral courage are both high.
Hypothesis 9: The positive indirect effects of perpetrating abusive supervisor behavior
on (a) consideration and (b) initiating structure behaviors via perceived loss of moral
credits are the strongest when leader moral attentiveness and moral courage are both
high.
Research Overview
We conducted two field studies using ESM and multisource data to test our hypotheses.
In Study 1, we tested the mediating effects of experienced guilt and the second-stage moderating
effects of leader moral courage (i.e., H1, H3, and H6) by sampling 31 leaders and 72 direct
followers over two weeks. Participants completed a daily survey at the end of each workday that

16
assesses study variables. Building on this initial effort, in Study 2 we tested our full theoretical
model (i.e., H1 to H9) by surveying 68 leader-follower dyads. We administrated two surveys per
day with the independent and mediating variables measured at midday and the outcome variables
measured at the end of workday. Doing so allowed for temporal precedence across study
variables and enabled us to test the proposed causal effects more rigorously. Taken together, the
two studies provide a complementary investigation of how leaders respond after perpetrating
abusive supervisor behavior.
Study 1 Method
Sample and Procedure
Our sample initially comprised 34 mid-level managers and 85 of their immediate
followers working for a real estate company in Southwest China.
1
Using the personal network of
one of the authors, we contacted the company’s human resources (HR) department to introduce
our research purpose and procedure and ask for assistance in recruiting research participants.
Leaders and one to three of their followers were invited to participate in our study based on their
expressed interest (leaders had four direct followers on average). Participants performed different
types of work across departments, including administration, marketing, agency, or property
management. This organization had an interactive climate in which managers and employees
work together closely. Additionally, this company emphasized maintaining good working
relationships with colleagues and encouraged employees to regularly reflect on and correct
inappropriate work behaviors. For example, the company organized weekly meetings in which
managers and employees discussed their work conduct and solutions for misbehavior. Hence,
this company provided an appropriate research site for studying leaders’ moral cleansing
1
Study 1 were part of a broader research project, which was reviewed and approved by the Ethics Committee at
National University of Singapore (A-16-007: Leader-member Interaction Experiences).

17
process. Prior to the commencement of the data collection, we assured participants of the
voluntary and confidential nature of their participation.
We received usable data from 31 leaders and 72 followers (we excluded leader-follower
dyads with three or fewer daily surveys; Ilies, Wagner, & Morgeson, 2007; Johnson et al., 2012),
yielding a 91% response rate for leaders and 85% for followers. Of the 31 leaders, 58% were
male, 74.2% had college educations or above, they had an average age of 37 years (SD = 9.1),
and their organizational tenure averaged 2.6 years (SD = 3.2). Of the 72 followers, 52% were
male, 56.9% had college educations or above, they had an average age of 32 years (SD = 8.5),
and their organizational tenure averaged 1.0 years (SD = 1.1). The average relationship tenure
between the leaders and followers was 9.3 months (SD = 13.1).
The data collection included two parts. First, during an initial study briefing session, we
introduced the purpose of the research but withheld the specific hypotheses. At the end of this
introduction, participants completed a paper-and-pencil entry survey that assessed demographic
information. Leaders additionally reported their level of moral courage. Second, one week after
the initial one-time survey, participants began completing daily online surveys sent via email
over a period of two weeks (10 consecutive workdays, Monday to Friday). Leaders and followers
completed questionnaires assessing affective states at the start of work each morning around 8:30
AM. In the afternoon around 5:30 PM, leader-follower dyads completed questionnaires
measuring study constructs. Given that leaders and followers might not interact directly every
workday, leader-follower dyads were directed to complete afternoon surveys only on days when
they had direct work interactions, including face-to-face communications, phone calls, emails,
and teleconferences (Barnes et al., 2015; Bolger et al., 2003). Leaders assessed their abusive
supervisor behavior and experienced guilt toward each follower that day. Followers reported

18
leaders’ consideration and initiating structure behaviors toward themselves. We used electronic
timestamps to verify that daily responses were submitted at scheduled time points and to pair
leader and follower responses on the same interaction day. We received a total of 568 dyadic
morning-afternoon matched responses from leaders and followers as our final sample. At the end
of the study, leaders and followers estimated the number of workdays when they had direct
interactions during the study period. We divided the average of numbers of interaction days
captured by the average of numbers of interaction days estimated by leaders and followers,
yielding an overall response rate of 86%.
Measures
All measures used in this study were originally developed in English and translated into
Mandarin Chinese following standard translation-back translation procedures to ensure
equivalence in meaning (Brislin, 1980). One of the authors first translated the survey items into
Chinese and then an independent bilingual scholar who was blind to the research hypotheses
translated the Chinese items back into English. As a final step, we conducted a pretest of the
items on 25 Chinese working adults and found no major problems regarding the item meaning.
Abusive supervisor behavior. We measured daily abusive supervisor behavior via a five-
item scale used by Johnson and colleagues (2012). These five items were adapted because they
show higher daily variance compared to other items on the full abusive supervision scale (Barnes
et al., 2015). As in previous abusive supervision research using an experience sampling approach
(e.g., Courtright et al., 2016; Lin et al., 2016), leaders reported their own abusive supervisor
behavior by indicating the frequency that they exhibited each behavior toward the focal follower
on that day on a scale ranging from 1 (never) to 5 (four or more times). A sample items is “I

19
yelled or swore at this subordinate.” The scale showed high between- (RKRN = .90) and within-
dyad (RCN = .78) reliability.
2
Collecting self-reports (as opposed to other-reports) of abusive supervisor behavior in
this study was deemed appropriate for two reasons. First, self-reports are the best way to capture
leaders’ own perceptions of perpetrating abusive behaviors toward the specific followers and
thus is more congruent with our perpetrator-centric investigation focus. Second, a meta-analysis
by Berry and colleagues (2012) documented relatively high correlations between self- and other-
ratings of counterproductive work behaviors and similar relational patterns of self- and other-
reports of counterproductive work behaviors with common correlates.
Guilt. Experienced guilt was assessed using Tangney, Miller, Flicker, and Barlow’s
(1996) three-item scale. Leaders indicated the extent to which each emotional adjective
(“repentant,” “guilty,” “blameworthy”) captured how they felt during their interactions with the
focal follower on that day. The scale had high between- (RKRN = .91) and within-dyad (RCN = .78)
reliability.
Reparative behaviors. Consideration behavior was assessed with Lanaj, Johnson, and
Lee’s (2016) four-item scale, which was adapted from Euwema et al. (2007) and Stogdill et al.
(1962). Initiating structure behavior was measured with a five-item scale adapted from Stogdill
et al. (1962). Followers reported the frequency with which their leader displayed these behaviors
toward them on that day using a scale ranging from 1 (never) to 5 (four or more times). Sample
2
We followed Shrout and Lane’s (2012) approach to compute the between- and within-dyad reliability for daily
measures. Between-dyad reliability is indicative of the internal consistency of measures taken over the study period
across all participants; within-dyad reliability is indicative of the reliability of change captured by the measurement
within particular participants throughout the study (Cranford, Shrout, Iida, Rafaeli, Yip, & Bolger, 2006). We first
estimated the variance components using VARCOMP (ANOVA) method. Based on the framework of
generalizability theory, we then used the obtained variance components to compute the between- and within-dyad
reliability of daily variables. Given that participants might have different numbers of daily observations over the
study period in both two studies, we calculate RKRN and RCN as the coefficients for the between- and within-dyad
reliability, respectively (Shrout & Lane, 2012).

20
items are “My supervisor looked out for my personal welfare” (consideration) and “My
supervisor told me what to do and how to do it” (initiating structure). The two scales had high
between- and within-dyad reliability (RKRN = .97 and .96, RCN = .77 and .87 for consideration and
initiating structure scales, respectively).
Moral courage. We measured leader moral courage with a four-item scale adapted from
Hannah and Avolio (2010) and Gibbs et al. (1986). Leaders indicated their level of agreement
with each item using a scale ranging from 1 (strongly disagree) to 5 (strongly agree). A sample
item is “I will defend someone who is being taunted or talked about unfairly, even if the victim is
only an acquaintance.” The coefficient alpha for this scale was .74.
Control variables. We controlled for participants’ positive and negative affect at the start
of work because previous research has suggested that affective states before work may influence
emotional states at work and displays of leadership behaviors (Lanaj et al., 2016). We assessed
positive and negative affect with Song and colleagues’ (2008) 10-item scale, which was adapted
from the PANAS (Watson & Clark, 1994). Participants indicated the extent to which each
emotional adjective (e.g., “enthusiastic” for positive affect, “upset” and for negative affect)
captured how they felt at that moment on a scale ranging from 1 (very slightly or not at all) to 5
(extremely). We also controlled for the mediator and outcome variables on Day t-1 in all
analyses, enabling us to capture changes in felt guilt and consideration and initiating structure
behaviors. Additionally, we controlled for demographics of leader gender, age, education, and
leader-follower dyadic tenure. We conducted analyses both with and without control variables
and obtained comparable results that did not alter our conclusions.
Analytic Strategy

21
Given the nested structure of data (daily variables were nested within leader-follower
dyads, which were in turn nested within leaders), we followed Preacher, Zyphur, and Zhang’s
(2010) recommendations to conduct three-level path analyses within the framework of multilevel
structural equation modeling (MSEM) using Mplus 7.0 (Muthén & Muthén, 1998-2014). Doing
so enabled us to test our hypotheses simultaneously rather than in a causal sequence and
piecemeal approach. Specifically, we first estimated an MESM model (M1) with random slopes
to test the main and mediation effects (cross-level moderators were not included in M1).
Mediation effects were tested with a Monte Carlo simulation procedure using the online software
R. We then estimated an MESM model (M2) that included leader moral courage (Level 3) as a
predictor of within-dyad random slopes of experienced guilt with consideration and initiating
structure behaviors (Preacher, Zhang, & Zyphur, 2016). We group-mean centered daily variables
and grand-mean centered between-leader level variable (i.e., moral courage) in our analyses
(Hofmann, Griffin, & Gavin, 2000). We examined the hypothesized indirect effects with a 95%
confidence interval (Preacher et al., 2010). We estimated values of pseudo-R2 to calculate effect
seizes assessing amount of within-dyad variance in mediators and outcome variables explained
by the study variables (Hofmann et al., 2000).
Study 1 Results
We first estimated null models in hierarchical linear modeling (HLM, Hofmann et al.,
2000) to partition the total variance in within-dyad variables into components at within-,
between-dyad, and leader levels. Results of null models revealed that all variables had significant
within-dyad variances: 52.2% for abusive supervisor behavior, 40.8% for experienced guilt,
27.0% for consideration behavior, and 29.9% for initiating structure behavior, suggesting that

22
leader behaviors and experienced guilt vary significantly from one day to the next.
3
Table 1
presents the means, standard deviations, and inter-correlations among study variables.
Table 2 reports results of MSEM models examining Hypotheses 1, 3, and 6. Results
showed that after we controlled for experienced guilt on Day t – 1, abusive supervisor behavior
was positively related to an increase in experienced guilt on Day t (γ = .58, p < .01),
demonstrating support for Hypothesis 1. We found that 20% of the within-dyad variance in
experienced guilt was explained by abusive supervisor behavior. Hypothesis 3 posited that
experienced guilt mediates the effects of perpetrating abusive behavior on subsequent
consideration (H3a) and initiating structure (H3b) behaviors. Results showed that after we
controlled for consideration and initiating structure behaviors on Day t – 1 respectively,
experienced guilt on was positively related to increases in consideration (γ = .28, p < .01) and
initiating structure behaviors (γ = .33, p < .01) on Day t. The mediation effect test based on a
Monte Carlo simulation with 20,000 replications indicated that the indirect effects of abusive
supervisor behavior on consideration and initiating structure behaviors through experienced guilt
were positive and significant (ab = .18, 95% CIs [.03, .34] for consideration behavior; ac = .19,
95% CIs [.04, .33] for initiating structure behavior), providing support for Hypotheses 3a and 3b.
Taken together, the mediation model explained 12% of the within-dyad variance in consideration
behavior and 12% in initiating structure behavior.
Hypothesis 6 posited that leader moral courage strengthens the within-dyad relationships
of experienced guilt with consideration behavior (H6a) and initiating structure behavior (H6b).
Results showed that leader moral courage was positively related to the within-dyad random slope
3
Results of intraclass correlations (ICCs) analyses were as follow: ICC(1) = .48 and ICC(2) = .88 for abusive
supervisor behavior; ICC(1) = .59 and ICC(2) = .89 for experienced guilt; ICC(1) = .73 and ICC(2) = .95 for
consideration behavior; ICC(1) = .71 and ICC(2) = .95 for initiating structure behavior.

23
between experienced guilt and consideration behavior (γ = .32, p < .01) but not significantly
related to that between experienced guilt and initiating structure behavior (γ = .09, p = .52).
Using Cohen and colleagues’ (2003) procedure, we plotted the interaction effect on
consideration behavior at two conditional values of the moderator (i.e., 1 SD above and below
the mean) in Figure 2. Simple slope analyses
4
(Preacher, Curran, & Bauer, 2006) showed that
experienced guilt and consideration behavior had a stronger relationship when moral courage
was high (simple slope = .41, p < .01) versus low (simple slope = .10, p = .39). These results
provide support for H6a but not for H6b. We further tested the indirect effect of abusive
supervisor behavior on consideration behavior via experienced guilt with the second-stage
moderation of leader moral courage. Results showed that the indirect effect was positive and
significant for leaders high in moral courage (estimate = .24, 95% CI [.10, .39]) but not
significant for those low in moral courage (estimate = .04, 95% CI [-.14, .21]). The difference
between these two conditional indirect effects was significant (estimate = .21, 95% CIs
[.04, .38]), suggesting that leader moral courage strengthens the indirect effect of abusive
supervisor behavior on consideration behavior via experienced guilt. The moderation mediation
model explained 15% of the within-dyad variance in consideration behavior.
Study 1 Discussion
Results from Study 1 provide initial support for our proposed model by demonstrating the
emotional path of the leader moral cleansing process and the second-stage moderating effect of
leader moral courage. Despite these encouraging findings, Study 1 is limited in a few respects.
First, all daily variables were reported at the same time (end of workday), prohibiting us from
teasing apart the temporal ordering among the variables (Podsakoff, MacKenzie, Lee, &
4
We used the online calculator at http://www.quantpsy.org/interact/mlr2.htm to examine the simple slopes for
moderation effects.

24
Podsakoff, 2003). Although our supplementary analyses
5
suggest that our theoretical model is
superior to alternative explanations, a more stringent approach to examine the hypothesized
causal relationships warrants a time-lag design. Second, Study 1 involved a relatively small
sample of leaders, which might reduce the statistical power in detecting the moderating effect of
leader moral courage. Third, although we assessed leader affective states before work as
controls, we only measured guilt at the end of work, resulting in potential common method
biases owing to measurement priming effects (Podsakoff et al., 2013). Fourth, we did not
measure the cognitive mechanism underlying moral cleansing (i.e., perceived loss of moral
credits) and the first-stage self-awareness moderator (i.e., moral attentiveness) in Study 1. To
address these limitations, we conducted another ESM study with a time-lag design, allowing for
temporal sequence between the predictor and outcome variables. Additionally, we surveyed 68
leaders and their immediate follower to increase statistical power and examined both pathways –
emotional and cognitive – of our proposed leader moral cleansing model.
Study 2 Method
Sample and Procedure
Using a similar recruitment strategy as in Study 1, we invited 72 first-line managers and
their immediate follower (each leader was paired with one follower) working for a footwear
manufacturing company in Southeast China to participate in Study 2.
6
During workdays,
managers had constant interactions with their followers regarding daily job arrangements,
product quality inspection, and production line maintenance, creating ample opportunities for
5
We conducted analyses testing the reversed indirect effects (i.e., reparative actions  experienced guilt  abusive
supervisor behavior). Results showed that the indirect effects of both consideration (estimate = .03, p = .47) and
initiating structure (estimate = .03, p = .19) behaviors on abusive supervisor behavior via experienced guilt on the
same workday were not significant.
6
Study 2 was reviewed and approved by the Human Subjects Ethics Committee at Hong Kong Polytechnic
University (HSEARS20141222002: Leader's Display of Anger).

25
managers to potentially exhibit positive and negative leadership behaviors toward their followers
each day. We received usable data from 68 leaders and followers, yielding a 94% response rates.
Of the 68 leaders, 68% were male, 24% had college educations or above, they had an average
age of 33 years (SD = 8.1), and their organizational tenure averaged 3.1 years (SD = 2.1). Of the
68 followers, 53% were male, 25% had college educations or above, they had an average age of
30 years (SD = 7.8), and their organizational tenure averaged 2.8 years (SD = 2.0). The average
relationship tenure for leaders and followers was 1.9 years (SD = 1.7).
The data collection included an initial one-time survey followed by two weeks of daily
surveys. The initial survey (paper-and-pencil based) assessed participants’ demographic
information and leader moral attentiveness and moral courage. One week later, participants
began completing daily online surveys for a period of 10 consecutive workdays (Monday to
Friday). We collected our daily surveys at two time points with the predictor and mediating
variables measured at noon and the outcome variables measured at the end of workday. In the
noon surveys, leaders completed questionnaires (sent at 11: 30 AM) assessing their abusive
behaviors, experienced guilt toward the follower, perceived loss of moral credits, and general
affective states during work in the morning. Followers reported their general affective states at
work that morning. At the end of workday, followers completed afternoon surveys (sent at 5 PM)
reporting leaders’ consideration and initiating structure behaviors that afternoon. As in Study 1,
leader-follower dyads were directed to complete noon or afternoon surveys only when they had
direct interactions at work that morning or afternoon.
7
All daily surveys were time-stamped,
7
In Study 2, we added a screener question ensuring that leaders and followers had direct interactions in the
morning/afternoon before they completed the surveys. In particular, we asked participants to answer the question
“How much contact did you have with your direct leader/ follower in the morning/afternoon?” using a 5 point Likert
scale in which 1=none, 2=few, 3=a moderate amount, 4=quite a bit, and 5=a high amount of interactions. As
recommended by Barnes et al. (2015), we included responses in which participants reported 3 or greater on the
scales. Doing so helped ensure participants answering noon/afternoon surveys with sufficient work information.

26
allowing us to verify that each survey was completed on the scheduled time point. Participants
received 5 RMB (approximately $0.74 USD) for each of their noon/afternoon survey response as
a token of appreciation for their participation. We received a total of 664 dyadic noon-afternoon
paired responses from leaders and their follower, yielding an overall response rate of 98% (out of
a total possible of 680). The company sponsorship and the usage of work time to administer
questionnaires facilitated our high response rates.
Measures
We used the same items and response scales as in Study 1 to measure abusive supervisor
behavior, experienced guilt, reparative behaviors, and moral courage. We instructed participants
to provide their responses based on the work interactions with their follower (or leader) in that
morning (or afternoon). Leaders indicated the frequency that they displayed abusive behaviors
and the extent to which the emotional adjectives of guilt captured how they felt toward their
follower in that morning. Followers reported the frequency with which their leader engaged in
reparative behaviors toward them in that afternoon. The between-dyad reliability estimates
(RKRN) were .96, .93, .95, and .95 and the within-dyad reliability estimates (RCN)
were .65, .82, .88, .83, and .85 for the scales of abusive supervisor behavior, experienced guilt,
consideration behavior, and initiating structure behavior, respectively. The coefficient alpha for
moral courage scale was .75. We translated other measures used in Study 2 into Mandarin
Chinese following standard translation-back translation procedures (Brislin, 1980) and conducted
a pretest as in Study 1.
Perceived loss of moral credits. We measured leader perceived loss of moral credits with
a five-item scale adapted from Lin et al. (2016). Leaders indicated their level of agreement with
each item using a scale ranging from 1 (strongly disagree) to 5 (strongly agree). The items are “I

27
lost moral credits for performing an immoral behavior in this morning,” “My morning bad deeds
forfeited me credits as a moral person,” “Behaving badly in this morning decreased my account
of moral credits,” “Each bad deed I performed in this morning reduced my moral credits,” and
“Acting in an unethical manner in this morning yielded a deficit of moral credits.” The scale
showed high between- (RKRN = .95) and within-dyad (RCN = .88) reliability.
Moral attentiveness. Leader moral attentiveness was assessed with Reynolds’ (2008)
five-item reflective moral attentiveness scale. Leaders rated their levels of agreement using a
scale ranging from 1 (strongly disagree) to 5 (strongly agree). A sample item is “I often reflect
on the moral aspects of my decisions/behaviors.” The coefficient alpha for this scale was .79.
Control variables. We controlled for both leaders’ and followers’ positive and negative
affect at work in the morning. Participant positive and negative affect were assessed with the
same items and response scale as in Study 1. We also controlled for leader demographics and
mediation and outcome variables on Day t-1 to capture the change in these variables across
workdays. We conducted analyses with and without control variables and both sets of analyses
yielded virtually identical results.
Analytic Strategy
Our analytic strategies were similar to Study 1. Considering the multilevel structure of
Study 2 data (i.e., daily variables were nested within leader-follower dyads), we conducted two-
level path analyses using Mplus 7.0 (Muthén & Muthén, 1998-2014). We estimated MSEM
models with random slopes to test the main, parallel mediation, and moderation effects. To
examine cross-level interaction effects, we included Level 2 leader moral attentiveness (a first-
stage moderator) and moral courage (a second-stage moderator) as predictors of within-dyad
random slopes of abusive supervisor behavior with two mediators and between mediators and

28
outcome variables, respectively (Preacher et al., 2016). We also computed values of pseudo-R2
as estimates of effect sizes to assess the amount of within-dyad variance in mediators and
outcome variables explained by the study variables.
Study 2 Results
We again estimated null models in HLM to partition the total variance of daily variables
into components at within- and between-dyad levels. Results showed that all daily variables had
significant within-dyad variances: 32.3% for abusive supervisor behavior, 37.2% for perceived
loss of moral credits, 42.7% for experienced guilt, 32.8% for consideration behavior, and 34.1%
for initiating structure behavior, revealing that leaders’ abusive and reparative behaviors and the
relevant cognitive and emotional experiences vary significantly from one day to the next.
8
Table
3 presents the means, standard deviations, and inter-correlations among study variables.
Table 4 reports MSEM results testing hypothesized main and moderation effects. Results
showed that abusive supervisor behavior was positively related to increases in perceived loss of
moral credits (γ = .29, p < .01) and experienced guilt (γ = .64, p < .01) in Day t morning, after we
controlled for perceived loss of moral credits and experienced guilt on Day t-1 respectively,
supporting Hypotheses 1 and 2. We found that 15% of the within-dyad variance in experienced
guilt and 10% in perceived loss of moral credits were explained by abusive supervisor behavior.
Hypothesis 3 proposed that experienced guilt mediates the effects of perpetrating abusive
supervisor behavior on consideration (H3a) and initiating structure (H3b) behaviors. Consistent
with Study 1 results, after we controlled for consideration and initiating structure behaviors on
8
Results of ICCs analyses were as follow: ICC(1) = .60 and ICC(2) = .94 for abusive supervisor behavior; ICC(1)
= .58 and ICC(2) = .93 for experienced guilt; ICC(1) = .61 and ICC(2) = .94 for perceived loss of moral credits;
ICC(1) = .57 and ICC(2) = .93 for consideration behavior; ICC(1) = .57 and ICC(2) = .93 for initiating structure
behavior.

29
Day t-1, experienced guilt on Day t morning was positively related to increases in consideration
(γ = .29, p < .01) and initiating structure (γ = .25, p < .05) behaviors that afternoon. Results based
on a Monte Carlo simulation indicated that the indirect effects of abusive supervisor behavior on
consideration and initiating structure behaviors through experienced guilt were significant (a1b1
= .19, 95% CIs [.06, .31] for consideration behavior; a1c1 = .16, 95% CIs [.03, .28] for initiating
structure behavior), demonstrating support for H3a and H3b. Likewise, results showed that
perceived loss of moral credits on Day t morning was positively related to increases in
consideration (γ = .44, p < .05) and initiating structure (γ = .41, p < .05) behaviors that afternoon.
Results of mediating effect tests revealed that the indirect effects of abusive supervisor behavior
on consideration and initiating structure behaviors through perceived loss of moral credits were
significant (a1b1 = .13, 95% CI [.00, .26] for consideration behavior; a1c1 = .12, 95% CI
[.00, .29] for initiating structure behavior), providing support for H4a and H4b. Taken together,
the parallel mediation model explained 14% of the within-dyad variance in consideration
behavior and 11% in initiating structure behavior.
Hypothesis 5 proposed that leader moral attentiveness strengthens the positive effects of
perpetrating abusive supervisor behavior on experienced guilt (H5a) and perceived loss of moral
credits (H5b). Results showed that leader moral attentiveness was positively related to the
within-person random slopes of abusive supervisor behavior with experienced guilt (γ = .26, p
< .05) and perceived loss of moral credits (γ = .30, p < .05). We plotted the interaction effects on
experienced guilt (Figure 3) and perceived loss of moral credits (Figure 4) at two conditional
values of leader moral attentiveness. Simple slope analyses (Preacher et al., 2006) showed that
perpetrating abusive supervisor behavior was more strongly related to experienced guilt and
perceived loss of moral credits when leader moral attentiveness was high (simple slope = 1.07, p

30
< .001 for experienced guilt; simple slope = .45, p < .01 for perceived loss of moral credits)
versus low (simple slope = .53, p < .05 for experienced guilt; simple slope = .02, p = .81 for
perceived loss of moral credits). These results are in line with H5a and H5b.
Hypothesis 6 proposed that leader moral courage strengthens the effects of experienced
guilt on consideration (H6a) and initiating structure (H6b) behaviors. Results showed that leader
moral courage was positively related to the within-person simple slope between experienced
guilt and consideration behavior (γ = .25, p < .05) but not significantly related to the within-
person simple slope between experienced guilt and initiating structure behavior (γ = .08, p = .44).
The interaction effect on consideration behavior at two conditional values of the moderator was
plotted in Figure 5. Simple slope analyses showed that experienced guilt had a stronger
relationship with consideration behavior when leader moral courage was high (simple slope
= .53, p < .01) versus low (simple slope = .05, p = .72). These results provided support for H6a
but not for H6b.
Hypothesis 7 posited that leader moral courage strengthens the effects of perceived loss
of moral credits on consideration (H7a) and initiating structure (H7b) behaviors. Results showed
that leader moral courage was positively related to the within-person simple slope between
perceived loss of moral credits and consideration behavior (γ = .25, p < .05) but not significantly
related to the within-person simple slope between perceived loss of moral credits and initiating
structure behavior (γ = .06, p = .69). We plotted the interaction effect on consideration behavior
at two conditional values of the moderator in Figure 6. Simple slope analyses showed that
perceived loss of moral credits had a stronger relationship with consideration behavior when
leader moral courage was high (simple slope = .35, p < .05) versus low (simple slope = -.00, p
=.99). These results demonstrate support for H7a but not for H7b.

31
Table 5 presents results examining the indirect effects of perpetrating abusive supervisor
on reparative leadership behaviors at different conditions of leader moral attentiveness (i.e., first-
stage moderator) and moral courage (i.e., second-stage moderator). Results showed that the
indirect effect of abusive supervisor behavior on consideration behavior via experienced guilt
was the strongest when leaders were high in both moral attentiveness and courage (estimate
= .36, 95% CI [.14, .58]), compared to other combinations of moral attentiveness and courage in
the remaining three conditions (estimates = .08, .19, and .04, 95% CI [-.15, .30], [.03, .34], and
[-.08, .15] respectively). Moreover, the indirect effect under the condition of high leader moral
attentiveness and courage was also significantly different from those under the other three
conditions (difference estimates = .28, .17, and .32, 95% CI [.01, .56], [.00, .35], and [.09, .55]
respectively). However, results showed that the indirect effect on initiating structure behavior via
experienced guilt was not significantly different from those under the other three conditions.
Thus, Hypothesis 8a was supported but Hypothesis 8b was not supported by the results.
Results also showed that the positive indirect effect of abusive supervisor behavior on
consideration behavior via perceived loss of moral credits was only significant when leaders
were high in both moral attentiveness and courage (estimate = .20, 95% CI [.01, .39]).
Additionally, the indirect effect under this condition was significantly stronger than that under
each of the other three conditions (difference estimates = .19, .16, and .20, 95% CI [-.00, .39],
[-.01, .35], and [.01, .38] respectively). We acknowledge that 95% CI of the differences of
Condition 1 with Conditions 2 and 3 included 0 but 90% CI of these differences did not include 0
(90% CI = [.02, .36] and [.02, .31], respectively). According to Preacher et al. (2010) and studies
with similar multilevel design in the management literature (e.g., Uy, Lin, & Ilies, 2017), 90% CI
are justifiable when testing multilevel moderated mediation models. Therefore, results supported

32
Hypothesis 9a. Likewise, results revealed that the indirect effect on initiating structure behavior
via perceived loss of moral credits was the strongest (estimate = .15, 90% CI [.01, .30]) under the
condition of high leader moral attentiveness and courage. However, this indirect effect was only
significantly stronger than that under the condition of low leader moral attentiveness and high
leader moral courage (difference estimate = .12, 90% CI = [.004, .26]) and that under the
condition of low leader moral attentiveness and courage (difference estimate = .12, 90% CI =
[.01, .28]). Thus, Hypothesis 9b received partial support from our results. The moderated
mediation model explained 15% of the within-dyad variance in consideration behavior and 12%
in initiating structure behavior.
9
General Discussion
We integrated the literatures on abusive supervision and moral cleansing to develop and
investigate a reparative response model that identifies psychological mechanisms and boundary
conditions under which engaging in abusive supervisor behavior affects how leaders
subsequently behave toward their followers. Across two ESM studies, we found that daily
displays of abusive supervisor behavior evoke leaders’ momentary experience of guilt and
perceptions of a loss of moral credits, which in turn prompted reparative actions. Moreover, such
psychological and behavioral responses were contingent on leaders being morally attentive and
courageous. Below, we discuss the theoretical and practical implications of our findings.
9
Although moral attentiveness and moral courage were hypothesized as first- and second-stage moderators,
respectively, as part of supplemental analyses we examined whether they serve as moderators at different stages of
our model. In Study 1, we found that moral courage moderated the first-stage relation of abusive supervisor behavior
with experienced guilt (γ = .50, p =.05). In Study 2, in contrast, moral courage did not moderate the first-stage
relations of abusive supervisor behavior with experienced guilt (γ = .18, p =.28) nor loss of moral credits (γ = .14, p
=.16). Similarly, moral attentiveness did not moderate the second-stage relations of experienced guilt with
consideration behavior (γ = .01, p =.66) nor loss of moral credits with consideration behavior (γ = .02, p =.89) in
Study 2. These results, together with our theoretical arguments, provide further support for our moderation
hypotheses, suggesting that moral attentiveness and moral courage are most appropriately modeled as first- and
second-stage moderators, respectively, in our moderated mediation model.

33
Theoretical Implications
Our focus on the consequences of abusive supervision with a perpetrator-centric
perspective advances understanding of leaders’ psychological and behavioral responses toward
their own abusive behaviors. Perpetrating abusive supervisor behavior, as an exemplar of moral
transgressions (Brown & Mitchell, 2010), may have a strong impact on leaders and thus their
subsequent leadership behaviors. To date, however, the overwhelming majority of work on
abusive supervision has focused on the effects of abusive supervision on followers and
organizations (Mackey et al., 2015), resulting in little empirical evidence supporting this
theoretical argument (for two recent exceptions, see Foulk, Lanaj, Tu, Erez, & Archambeau,
2017, and Qin, Huang, Johnson, Ju, & Hu, 2017). We applied moral cleansing theory to abusive
supervision and found that leaders feel emotionally and cognitively immoral and thus perform
reparative actions to cleanse themselves. Our research suggests that the negative psychological
experiences of abusive supervision are not only limited to followers but can also extend to
leaders, since engaging in such behaviors violates widely endorsed moral norms and jeopardizes
leaders’ care and justice components of moral self-concept (Aquino & Reed, 2002). Nonetheless,
unlike followers’ adverse behavioral responses, leaders subsequently display behaviors that are
constructive and beneficial to followers in order to cleanse their emotional and cognitive feelings
of immorality. In other words, from a perpetrator-centric perspective, the distal outcomes of
abusive supervision become paradoxical (i.e., constructive leadership behaviors). These findings
diverge from typical research on the consequences of abusive supervision, which revolves
around negative reciprocity, retaliation, and impaired regulation (Tepper, 2007). Our research,
therefore, represents an important complement to the abusive supervision literature.

34
Our findings also suggest that leaders’ moral cleansing process occurs through both
emotional and cognitive mechanisms. Although the mechanisms via which bad deeds lead to
subsequent positive behavior have been heavily discussed in the moral cleansing literature, they
have rarely been empirically demonstrated. More importantly, while past research has mainly
theorized a cognitive approach to moral cleansing (Gino et al., 2015; Zhong et al., 2010), we
reveal that moral emotions, such as guilt, also play a critical role in this process. By capturing
both emotional- and cognitive-based mechanisms, our research adds to the moral cleansing
literature regarding why preceding moral transgressions engender subsequent reparative actions.
Moreover, not only are we one of the first to adopt a perpetrator-centric perspective and
explore the paradoxical effects of abusive supervisor behavior, we also provide insights on the
boundary conditions of leaders’ moral cleansing process. Our results suggest that morally
attentive leaders are more likely to recognize and recall their prior abusive behavior and reflect
on how such behaviors can negatively affect followers (Reynolds, 2008). These lead to more
intense feelings of guilt and perceived loss of moral credits. In addition, morally courageous
leaders are able to translate these emotional and cognitive mechanisms into reparative actions
(Sekerka et al., 2009). In sum, we provide a more complete understanding of the types of leaders
most likely to respond constructively after engaging in abusive supervisor behavior.
Surprisingly, our results consistently showed that moral courage strengthens only the
indirect effect of perpetrating abusive supervisor behavior on person-oriented reparative action
(i.e., consideration) but not task-oriented reparative action (i.e., initiating structure). This may be
due to the interpersonal nature of abusive supervisor behavior. Abusive supervision is inherently
a destructive interpersonal act, damaging followers’ psychological well-being and causing
negative interpersonal outcomes (Mackey et al., 2015). Hence, the most direct way for leaders to

35
make amends is to perform interpersonal reparative actions. Compared to initiating structure,
consideration is more strongly related to positive interpersonal outcomes (Judge et al., 2004). It
is more likely to help rebuild interpersonal relationships, and reduce the hurt caused by
preceding abusive behavior. Leaders high in moral courage, hence, are more likely to engage in
consideration behavior as a more effective way to replenish their threatened moral self-concept.
Additionally, according to the interpersonal conflict literature (Baumeister, Stillwell, & Wotman,
1990), compared to initiating a task-oriented interaction, leaders are more likely to feel
embarrassed and reluctant to initiate an interpersonal-oriented interaction following conflict with
their followers. Leaders thus need a stronger conation to engage in such an interaction with the
previously abused followers but not for task-oriented behaviors such as initiating structure.
By demonstrating the dynamic nature of leadership behavior, our research also makes
important contributions to the broader leadership literature. Our findings highlight that leadership
goes beyond static leadership behavioral style and attributes and instead it is a dynamic process
that consists of considerable within-person fluctuations regarding specific leadership behaviors
(Dinh et al., 2014). By documenting such within-person variations, our research appropriately
responds to scholars’ recent call that leadership research should examine leadership behaviors
using a process perspective to explore how leaders shift their behaviors from time to time across
different situations (Dinh et al., 2014).
Our research further sheds light on our growing knowledge on paradoxical leadership
patterns by revealing the display of different leadership behaviors that are seemingly
incompatible within the same leaders (Denison et al., 1995; Smith & Lewis, 2011). Using a more
granular approach, we found that destructive leadership behavior has positive effects on
constructive leadership behavior, providing an ontologically accurate description of leadership

36
phenomenon. Our findings complement a recent finding by Lin et al. (2016), which shows that
constructive leader behaviors (e.g., ethical behavior) can trigger destructive behaviors (e.g.,
abusive supervision) via a licensing effect. Destructive and constructive leader behaviors
therefore appear to be reciprocally related via the moral cleansing and licensing processes. Taken
together, these findings implicitly suggest that leaders are neither ‘good’ nor absolutely ‘bad,’
but rather their behavior is either good or bad from one moment to the next. Hence, it is not
accurate to simply label people as either constructive or destructive leaders, suggesting that
leadership behavior is perhaps more paradoxical and complex than previously assumed. Due to
individual and/or situational factors, they may engage in a variety of leadership behaviors that
are seemingly competing and contradicted yet interconnected over time (Denison et al., 1995).
Finally, we extend the moral cleansing literature by capturing within-person dynamic
fluctuations of morality. By taking an initial step toward investigating leaders’ reparative
responses to their own moral transgressions during workdays, our research suggests that moral
and immoral deeds wax and wane within the same individuals across time, complementing the
empirical evidence of moral cleansing (Zhong et al., 2009). Our findings on the moderating roles
of moral attentiveness and moral courage also add to the moral cleaning literature by identifying
individual differences in moral character as contingent factors of cleansing responses.
Practical Implications
Our findings also yield important implications for managerial practice. Our research
demonstrates the idea that engaging in abusive supervisor behavior also yields negative
psychological experiences for leaders (Zhong et al., 2009). To relieve emotional and cognitive
experiences of immorality, leaders must perform constructive leadership behaviors at a level
above and beyond what they normally do subsequently. Although these reparative behaviors may

37
be beneficial and helpful toward followers eventually, they sap the time, attention, and energy
that leaders need for completing other managerial work, decreasing leader effectiveness in
general (Tepper, 2007). Consequently, organizations should implement training programs to help
managers improve their leadership and interpersonal skills and curb their abusive leadership
behavior in the first place. Managers, however, sometimes perpetrate interpersonal mistreatment
toward followers impulsively because of breakdowns in self-regulation (Barnes et al., 2015;
Yam, Fehr, Keng-Highberger, Klotz, & Reynolds, 2016). After such impulsive mistreatment,
they should proactively make amends through providing followers with additional support and
guidance. In doing so, managers could better regulate themselves from negative psychological
experiences and potentially reduce the detrimental effects of abusive behavior on subordinates.
Considering the moderating effects that we observed in the moral cleansing process, the
likelihood that leaders exhibit reparative behavior toward followers could be increased by
enhancing leaders’ moral attentiveness and/or moral courage. To do this, organizations could
provide training programs to bolster employees’ moral attentiveness and moral courage. For
example, ethics training regularly requires leaders to reflect on their past behaviors, which might
foster higher moral attentiveness (Reynolds, 2008). Implementing a strong ethical culture
centered on ethics would likely also increase employees’ moral self-attentiveness and their
willingness to step in and redress immoral behavior (Yam, Reynolds, & Hirsh, 2014).
Organizations can also establish norms for apologies and forgiveness by instituting policies that
encourage employees to be morally courageous and remove the risks associated with implicit
admission of abusive supervision (Sekerka et al., 2009).
By understanding that leaders may perform compensatory actions after behaving
abusively, employees can become better aware of leadership paradoxes and develop a more

38
balanced view toward leaders who occasionally display abusive supervisor behavior. Leaders’
behaviors fluctuate from time to time and those who are abusive at this moment may be helpful
afterward. Accordingly, employees should learn to cope with occasional instances of abusive
leader behaviors. Rather than withdrawing from work after being mistreated, subordinates could
seize the opportunity to establish more constructive interactions with their leaders and encourage
more supportive and task directive behaviors from their leaders in the future.
Limitations and Future Research Directions
Despite the use of multiple within-person studies to support the proposed relationships,
our research has some limitations that can be redressed by future research. One limitation
concerns leaders’ moral cleansing behaviors toward followers who were not the recipients of
abusive behavior. As an initial effort to explore how leaders behave in the aftermath of abusive
supervisor behavior, we focused only on the reparative actions targeted at the abused followers.
The moral cleansing literature suggests that leaders may also display constructive leadership
behaviors toward other (non-abused) followers or engage in metaphorical cleansing actions as
approaches to cleansing their sense of immorality (West & Zhong, 2015). Our results of Study 1
supplementary analyses provide insight into this perspective. Specifically, we conducted
analyses testing whether experienced guilt toward the focal follower also leads to constructive
leadership behaviors toward other followers by aggregating consideration and initiating structure
behaviors reported by other followers under the same leaders on the same workday (excluding
the focal follower). Results showed that experienced guilt was not positively associated with
leaders’ consideration (γ = .09, p = .44) and initiation structure (γ = .02, p = .88) behaviors
towards other followers that day. Taken together, our findings suggest that leaders are more
likely to engage in compensatory behaviors towards the abused followers. Given that scholars

39
have revealed cross-boundary moral cleansing effects, we therefore encourage future research to
examine when leaders might engage in compensatory behavior towards other followers.
Second, although our research shows the linear effects of perpetrating abusive supervisor
behavior on immoral emotion and cognition and subsequent reparative actions, it is possible that
abusive supervisor behavior may have curvilinear relationships with these consequences. That is,
the levels of immoral feelings associated with abusive supervisor behavior may depend on the
amount of abusive supervisor behavior engaged that day or in general by leaders (Mullen &
Monin, 2016). For example, perpetrating abusive supervisor numerous times may have
exponentially increasing effects on feelings of immorality. However, we did not observe any
curvilinear effects in our two within-person studies. Nevertheless, we invite future research to
investigate these potential non-linear effects of perpetrating abusive supervisor behavior on
leaders’ emotional, cognitive, and behavioral consequences.
Third, both of our studies were conducted in China, raising concerns about the
generalizability of our findings. We note that an initial pilot study conducted in the United States
with 61 full-time employees enrolled in a part-time MBA program of a large university provided
support for the link between daily abusive supervision and experienced guilt.
10
That said, we
encourage future studies to extend our full theoretical model in different cultural settings.
Finally, although we find that leaders try to compensate after abusing their followers, we
have little knowledge about how followers react to leaders’ such behavioral shift. Duffy,
Ganster, and Pagon (2002) found that social support from leaders would even exacerbate the
10
The data collection included an initial one-time survey followed by 10 daily surveys. The initial survey assessed
moral courage and demographic information. One week later, participants began completing daily online surveys at
the end of workday for a period of 10 workdays to assess enacted abusive supervisor behavior and experienced guilt
on that day. Results showed that after controlling for Day t-1 experienced guilt, abusive supervisor behavior was
positively related to an increase in experienced guilt (γ = .35, p < .01). We omitted this pilot study for brevity.
Details are available from the first author.

40
negative effects of leader social undermining on follower outcomes due to followers’ cognitive
dissonance. However, this research finding was based on a between-person retrospective design.
We therefore encourage future research to explore the potential effects on followers of leaders’
transient behavioral shift using a within-person approach. Additionally, future research could
delve into the role of the intentionality of abusive supervision in shaping leaders’ subsequent
responses. Whereas some leaders might engage in abuse as a result of an occasional blip, prior
research has suggested that other leaders may intentionally perpetrate in abusive supervisor
behavior from time to time to as a strategy means (Tepper et al., 2011). Consequently, such
leaders may less likely experience immorality and perform subsequent reparative actions. We
invite future research to investigate leaders’ motivations of daily abusive supervisor behavior and
explore their effects on leaders’ following experiences.
Conclusion
By adopting a perpetrator-centric perspective of abusive supervision, we examined why
and when perpetrating abusive supervisor behavior gives rise to leaders’ reparative actions. We
found that leaders felt guilty and perceived a loss of moral credits following their acts of abusive
behavior toward followers, especially leaders who have high moral attentiveness. The feeling of
guilt and loss of moral credits, in turn, motivated leaders to take reparative actions by providing
greater task and interpersonal support to followers, especially leaders who have the moral
courage to accept responsibility for their immoral acts. Although our results shed light on the
affective, cognitive, and behavioral consequences of perpetrating abusive supervisor behavior,
we recognize that we have only taken an initial step in exploring these phenomena. We therefore
encourage future research to extend scholarly understanding of the consequences of perpetrating
abusive supervisor behavior.

41
References
Aquino, K., & Reed II, A. (2002). The self-importance of moral identity. Journal of Personality
and Social Psychology, 83, 1423-1440. doi: 10.1037//0022-3514.83.6.1423
Barnes, C. M., Lucianetti, L., Bhave, D. P., & Christian, M. S. (2015). “You Wouldn’t Like Me
When I’m Sleepy”: Leaders’ Sleep, Daily Abusive Supervision, and Work Unit
Engagement. Academy of Management Journal, 58, 1419-1437. doi:
10.5465/amj.2013.1063
Baumeister, R. F., Stillwell, A. M., & Heatherton, T. F. (1994). Guilt: an interpersonal approach.
Psychological Bulletin, 115, 243-267. doi: 00006823-199403000-00005
Baumeister, R. F., Stillwell, A., & Wotman, S. R. (1990). Victim and perpetrator accounts of
interpersonal conflict: autobiographical narratives about anger. Journal of Personality
and Social Psychology, 59, 994-1005. doi: 10.1037/0022-3514.59.5.994
Berry, C. M., Carpenter, N. C., & Barratt, C. L. (2012). Do other-reports of counterproductive
work behavior provide an incremental contribution over self-reports? A meta-analytic
comparison. Journal of Applied Psychology, 97, 613-636. doi: 10.1037/a0026739
Bolger, N., Davis, A., & Rafaeli, E. (2003). Diary methods: Capturing life as it is lived. Annual
Review of Psychology, 54, 579-616. doi: 10.1146/annurev.psych.54.101601.145030
Brown, M. E., & Mitchell, M. S. (2010). Ethical and unethical leadership: Exploring new
avenues for future research. Business Ethics Quarterly, 20, 583-616.
Brislin, R. W. (1980). Translation and content analysis of oral and written material. In H. C.
Triandis & J. W. Berry (Eds.), Handbook of Cross Cultural Psychology (Vol. 2, pp. 349-
444). Boston, MA: Allyn & Bacon.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum.
Courtright, S., Gardner, R., Smith, T., McCormick, B., & Colbert, A. (2016). My family made
me do it: A cross-domain, self-regulatory perspective on antecedents to abusive
supervision. Academy of Management Journal, 59, 1630-1652.
Cranford, J. A., Shrout, P. E., Iida, M., Rafaeli, E., Yip, T., & Bolger, N. (2006). A procedure for
evaluating sensitivity to within-person change: Can mood measures in diary studies
detect change reliably? Personality and Social Psychology Bulletin, 32(7), 917-929.
Denison, D. R., Hooijberg, R., & Quinn, R. E. (1995). Paradox and performance: Toward a
theory of behavioral complexity in managerial leadership. Organization Science, 6, 524-
540. doi: 10.1287/orsc.6.5.524
Dinh, J. E., Lord, R. G., Gardner, W. L., Meuser, J. D., Liden, R. C., & Hu, J. (2014). Leadership
theory and research in the new millennium: Current theoretical trends and changing
perspectives. The Leadership Quarterly, 25, 36-62. doi: 10.1016/j.leaqua.2013.11.005
Euwema, M. C., Wendt, H., & van Emmerik, H. (2007). Leadership styles and group
organizational citizenship behavior across cultures. Journal of Organizational Behavior,
28, 1035–1057. http://dx.doi.org/10.1002/job .496
Fehr, R., Yam, K. C. S., & Dang, C. (2015). Moralized leadership: The construction and
consequences of ethical leader perceptions. Academy of Management Review, 40, 182-
209. doi: 10.5465/amr.2013.0358
Fleishman, E. A. (1995). Consideration and structure: Another look at their role in leadership
research. In F. Dansereau & F. J. Yammarino (Eds.), Leadership: The multiple-level
approaches (pp. 51–60). Stamford, CT: JAI Press.

42
Fleishman, E. A., & Harris, E. F. (1962). Patterns of leadership behavior related to employee
grievances and turnover. Personnel Psychology, 15, 43-56.
Foulk, T., Lanaj, K., Tu, M. H., Erez, A., & Archambeau, L. (2017). Heavy is the head that
wears the crown: An actor-centric approach to daily psychological power, abusive leader
behavior, and perceived incivility. Academy of Management Journal. Advance online
publication.
Gibbs, J. C., Clark, P. M., Joseph, J. A., Green, J. L., Goodrick, T. S., & Makowski, D. G.
(1986). Relations between moral judgment, moral courage, and field independence. Child
Development, 57, 185-193. doi: 10.2307/1130650
Gilligan, C. (1982). In a different voice. Cambridge, MA: Harvard University Press.
Gino, F., Kouchaki, M., & Galinsky, A. D. (2015). The moral virtue of authenticity: How
inauthenticity produces feelings of immorality and impurity. Psychological Science, 26,
983-996. doi: 10.1177/0956797615575277
Hannah, S. T., Avolio, B. J., & May, D. R. (2011). Moral maturation and moral conation: A
capacity approach to explaining moral thought and action. Academy of Management
Review, 36, 663-685. doi: 10.5465/amr.2010.0128
Hofmann, S. A., Griffin, M. A., & Gavin, M. B. (2000). The application of hierarchical linear
modeling to organizational research. In K. Klein & S. Kozlowski (Eds.), Multilevel
theory, research, and methods in organizations: Foundations, extensions, and new
directions (pp. 75–170). San Francisco, CA: Jossey-Bass
Ilies, R., Peng, A. C., Savani, K., & Dimotakis, N. (2013). Guilty and helpful: An emotion-based
reparatory model of voluntary work behavior. Journal of Applied Psychology, 98, 1051-
1059. doi: 10.1037/a0034162
Johnson, R. E., Venus, M., Lanaj, K., Mao, C., & Chang, C. H. (2012). Leader identity as an
antecedent of the frequency and consistency of transformational, consideration, and
abusive leadership behaviors. Journal of Applied Psychology, 97, 1262-1272. doi:
10.1037/a002904.
Jordan, J., Mullen, E., & Murnighan, J. K. (2011). Striving for the moral self: The effects of
recalling past moral actions on future moral behavior. Personality and Social Psychology
Bulletin, 37, 701-713. doi: 10.1177/0146167211400208
Judge, T. A., Piccolo, R. F., & Ilies, R. (2004). The forgotten ones? The validity of consideration
and initiating structure in leadership research. Journal of Applied Psychology, 89, 36-51.
doi: 10.1037/0021-9010.89.1.36
Kidder, R. M. (2003). Moral courage. New York: William Morrow.
Kohlberg, L. (1971). From is to ought: How to commit the naturalistic fallacy and get away with
it in the study of moral development. In T. Mischel (Ed.), Cognitive development and
epistemology: 151–235. New York, NY: Academic Press.
Kouchaki, M., & Gino, F. (2016). Memories of unethical actions become obfuscated over time.
Proceedings of the National Academy of Sciences, 113, 6166-6171. doi:
10.1073/pnas.1523586113
Lanaj, K., Johnson, R. E., & Lee, S. M. (2016). Benefits of transformational behaviors for
leaders: A daily investigation of leader behaviors and need fulfillment. Journal of
Applied Psychology, 101, 237-251. doi: 10.1037/apl000005
Lin, S. H. Ma, J., & Johnson, R. E (2016). When ethical leader behavior breaks bad: How ethical
leader behavior can turn abusive via ego depletion and moral licensing. Journal of
Applied Psychology, 101, 815-830. doi: 10.1037/apl0000098

43
Mackey, J. D., Frieder, R. E., Brees, J. R., & Martinko, M. J. (2015). Abusive Supervision: A
Meta-Analysis and Empirical Review. Journal of Management. Advance online
publication. doi: 0149206315573997
Miller, D. T., & Effron, D. A. (2010). Psychological license: When it is needed and how it
functions. In M. P. Zanna & J. Olson (Eds.), Advances in experimental social psychology
(Vol. 43, pp. 115–155). San Diego, CA: Elsevier.
Mullen, E., & Monin, B. (2016). Consistency versus licensing effects of past moral behavior.
Annual Review of Psychology, 67, 363-385. Doi: 10.1146/annurev-psych-010213-115120
Muthén, L. K., & Muthén, B. O. (1998-2014). Mplus 7.11 [Computer software]. Los Angeles,
CA: Muthén & Muthén.
Nisan, M. (1990). Moral balance: A model of how people arrive at moral decisions. In T. E.
Wren (Ed.), The moral domain: 283–314. Cambridge, MA: MIT Press.
Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003). Common method
biases in behavioral research: a critical review of the literature and recommended
remedies. Journal of Applied Psychology, 88, 879-903. doi: 10.1037/0021-9010.88.5.879
Preacher, K. J., Curran, P. J., & Bauer, D. J. (2006). Computational tools for probing interactions
in multiple linear regression, multilevel modeling, and latent curve analysis. Journal of
Educational and Behavioral Statistics, 31, 437-448. doi: 10.3102/10769986031004437
Preacher, K. J., Zhang, Z., & Zyphur, M. J. (2016). Multilevel structural equation models for
assessing moderation within and across levels of analysis. Psychological Methods, 21,
189-205. doi: 10.1037/met0000052
Preacher, K. J., Zyphur, M. J., & Zhang, Z. (2010). A general multilevel SEM framework for
assessing multilevel mediation. Psychological Methods, 15, 209-233. doi: 10.1037/a0020141
Qin, X., Huang, M., Johnson, R., Hu, Q., & Ju, D. (2017). The short-lived benefits of abusive
supervisory behavior for actors: An investigation of recovery and work engagement.
Academy of Management Journal. Advance online publication.
Reynolds, S. J. (2008). Moral attentiveness: Who pays attention to the moral aspects of life?.
Journal of Applied Psychology, 93, 1027-1041. doi: 10.1037/0021-9010.93.5.1027
Sachdeva, S., Iliev, R., & Medin, D. L. (2009). Sinning saints and saintly sinners the paradox of
moral self-regulation. Psychological Science, 20, 523-528. doi: 10.1111/j.1467-
9280.2009.02326.x
Sekerka, L. E., Bagozzi, R. P., & Charnigo, R. (2009). Facing ethical challenges in the
workplace: Conceptualizing and measuring professional moral courage. Journal of
Business Ethics, 89, 565-579. doi: 10.1007/s10551-008-0017-5
Shalvi, S., Gino, F., Barkan, R., & Ayal, S. (2015). Self-serving justifications doing wrong and
feeling moral. Current Directions in Psychological Science, 24, 125-130.
Shrout, P. E., & Lane, S. P. (2012). Psychometrics. In M. R. Mehl & T. S. Conner (Eds.).
Handbook of Research Methods for Studying Daily Life (pp. 302-320). New York: The
Guilford Press.
Smith, W. K., & Lewis, M. W. (2011). Toward a theory of paradox: A dynamic equilibrium
model of organizing. Academy of Management Review, 36, 381-403. doi:
10.5465/amr.2009.0223
Song, Z., Foo, M. D., & Uy, M. A. (2008). Mood spillover and crossover among dual-earner
couples: a cell phone event sampling study. Journal of Applied Psychology, 93, 443-452.
Stogdill, R. M., Goode, O. S., & Day, D. R. (1962). New leader behavior description subscales.
The Journal of Psychology, 54, 259-269. doi: 10.1080/00223980.1962.9713117

44
Tangney, J. P. (1990). Assessing individual differences in proneness to shame and guilt:
development of the Self-Conscious Affect and Attribution Inventory. Journal of
Personality and Social Psychology, 59, 102-111. doi: 10.1037/0022-3514.59.1.102
Tangney, J. P., Miller, R. S., Flicker, L., & Barlow, D. H. (1996). Are shame, guilt, and
embarrassment distinct emotions?. Journal of Personality and Social Psychology, 70,
1256-1269. doi: 10.1037/0022-3514.70.6.1256
Tepper, B. J. (2000). Consequences of abusive supervision. Academy of Management Journal,
43, 178-190. doi: 10.2307/1556375
Tepper, B. J. (2007). Abusive supervision in work organizations: Review, synthesis, and research
agenda. Journal of Management, 33, 261-289. doi: 10.1177/0149206307300812
Tepper, B. J., Moss, S. E., & Duffy, M. K. (2011). Predictors of abusive supervision: Supervisor
perceptions of deep-level dissimilarity, relationship conflict, and subordinate
performance. Academy of Management Journal, 54, 279-294.
Tracy, J. L., & Robins, R. W. (2006). Appraisal antecedents of shame and guilt: Support for a
theoretical model. Personality and Social Psychology Bulletin, 32, 1339-1351.
Treviño, L. K., Weaver, G. R., & Reynolds, S. J. (2006). Behavioral ethics in organizations: A
review. Journal of Management, 32, 951-990. doi: 10.1177/0149206306294258
Uy, M. A., Lin, K. J., & Ilies, R. (2017). Is it better to give or receive? The role of help in
buffering the depleting effects of surface acting. Academy of Management Journal, 60(4),
1442-1461. doi: 10.5465/amj.2015.0611
Watson D, Clark LA. (1994). Manual for the positive and negative affect schedule (expanded
form). Unpublished manuscript, University of Iowa, Iowa City.
Weaver, G. R., Reynolds, S. J., & Brown, M. E. (2014). Moral intuition connecting current
knowledge to future organizational research and practice. Journal of Management, 40,
100-129. doi: 10.1177/0149206313511272
West, C., & Zhong, C. B. (2015). Moral cleansing. Current Opinion in Psychology, 6, 221-225.
doi: 10.1016/j.copsyc.2015.09.022
Whitaker, B. G., & Godwin, L. N. (2013). The antecedents of moral imagination in the
workplace: A social cognitive theory perspective. Journal of Business Ethics, 114(1), 61-
73. doi: 10.1007/s10551-012-1327-1
Yam, K. C., Fehr, R., Keng-Highberger, F. T., Klotz, A. C., & Reynolds, S. J. (2016). Out of
control: A self-control perspective on the link between surface acting and abusive
supervision. Journal of Applied Psychology, 101, 292-301. doi: 10.1037/apl0000043
Yam, K. C., Klotz, A. C., He, W., & Reynolds, S. J. (2017). From good soldiers to
psychologically entitled: Examining when and why citizenship behavior leads to
deviance. Academy of Management Journal, 60, 373-396. doi: 10.5465/amj.2014.0234
Yam, K. C., Reynolds, S. J., & Hirsh, J. (2014). The hungry thief: Physiological deprivation and
its effects on unethical behavior. Organizational Behavior and Human Decision
Processes, 125, 123-133. doi: 10.1016/j.obhdp.2014.07.002
Zhang, Y., & Liao, Z. (2015). Consequences of abusive supervision: A meta-analytic review.
Asia Pacific Journal of Management, 32, 959-987. doi: 10.1007/s10490-015-9425-0
Zhong, C. B., Liljenquist, K., & Cain, D. M. (2009). Moral self-regulation: Licensing &
compensation. In D. De Cremer (Ed.), Psychological perspectives on ethical behavior
and decision making (pp. 75–89). Charlotte, NC: Information Age Publishing.
Zhong, C. B., Ku, G., Lount, R. B., & Murnighan, J. K. (2010). Compensatory ethics. Journal of
Business Ethics, 92, 323-339. Doi: 10.1007/s10551-009-0161-6

45
Table 1
Means, Standard Deviations, and Correlations among Study 1 Variables
Variables
Mean
Between-
dyad (leader)
S.D.
Within-
dyad
S.D.
1
2
3
4
5
6
7
Within-dyad level
1. Leader before-work positive affect
3.91
.65
.52
2. Leader before-work negative affect
1.52
.51
.50
-.30**
3. Follower before-work positive affect
3.92
.68
.49
.02
.08
4. Follower before-work negative affect
1.42
.54
.36
.13**
.03
-.28**
5. Abusive supervisor behavior
1.25
.36
.30
-.02
.12
-.05
.17**
6. Experienced guilt
1.37
.48
.38
-.07
.12
.02
.01
.38**
7. Consideration behavior
2.95
.97
.52
-.01
-.02
.21**
-.09
.10*
.24**
8. Initiating structure behavior
3.29
.95
.54
.03
-.01
.09
-.08
.11**
.24**
.59**
Between-dyad level
1. Dyadic tenure (month)
9.33
13.09
Leader level
1. Leader gender
.58
.50
2. Leader age (year)
36.19
9.07
.14
3. Leader education
2.81
.91
-.26
-.17
4. Leader moral courage
4.12
.58
-.37*
-.06
-.20
Note: Within-dyad level, N = 568; Between-dyad level, N = 72; Leader level, N = 31. Within-dyad correlations are based on within-dyad scores.
Gender was coded as 1 = male, 0 = female. Leader education was coded as 1 = middle school, 2 = high school, 3 = polytechnic, 4 = bachelor’s
degree, and 5 = master’s degree and above.
*p < .05, **p < .01 (two-tailed).

46
Table 2
Unstandardized Coefficients of MSEMs for Testing Main, Mediation and Moderation Effects in Study 1
Variables
Experienced guilt
(Day t afternoon)
Consideration behavior
(Day t afternoon)
Initiating structure behavior
(Day t afternoon)
MSEM 1
MSEM 1
MSEM 2
MSEM 1
MSEM 2
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Variables on Day t-1
Experienced guilt
.03
.11
Consideration behavior
.01
.04
.01
.04
Initiating structure behavior
.05
.04
.05
.04
Variables on Day t
Leader before-work positive affect
-.03
.04
-.03
.05
-.03
.05
.03
.06
.02
.10
Leader before-work negative affect
.01
.03
-.07
.05
-.07
.06
-.03
.05
-.03
.08
Follower before-work positive affect
.01
.07
.12*
.05
.13*
.05
.04
.06
.04
.09
Follower before-work negative affect
-.02
.06
-.04
.05
-.05
.05
-.07
.07
-.07
.20
Abusive supervisor behavior
.58**
.21
.07
.09
.08
.11
.06
.10
.07
.17
Experienced guilt
.28**
.11
.25*
.10
.33**
.11
.29**
.11
Variables at the leader level
Leader moral courage (MC)
-.39
.41
-.03
.39
Cross-level interactions
Experienced guilt × MC
.32**
.12
.09
.14
Note: Within-dyad level, N = 528; Between-dyad level, N = 72; Leader level, N = 31. MSEM = multilevel structural equation modeling; SE =
standard error. We omitted estimates of demographics for brevity.
*p < .05, **p < .01 (two-tailed).

47
Table 3
Means, Standard Deviations, and Correlations among Study 2 Variables
Variables
Mean
Between-
dyad
S.D.
Within-
dyad
S.D.
1
2
3
4
5
6
7
8
Within-dyad level
1. Leader morning positive affect
3.92
0.75
0.61
2. Leader morning negative affect
1.45
0.60
0.41
-.14**
3. Follower morning positive affect
3.87
0.77
0.69
.03
.11**
4. Follower morning negative affect
1.32
0.42
0.43
-.00
.02
.00
5. Abusive supervisor behavior
1.26
0.38
0.24
.03
.10**
.02
-.11**
6. Experienced guilt
1.38
0.59
0.47
.02
.13**
.02
.02
.34**
7. Perceived loss of moral credits
1.27
0.50
0.36
.02
.05
.10**
.02
.25**
.28**
8. Consideration behavior
2.60
1.04
0.67
-.01
-.02
.02
-.00
.14**
.20**
.19**
9. Initiating structure behavior
2.89
1.00
0.66
-.02
.02
-.01
.02
.14**
.19**
.20**
.49**
Between-dyad level
1. Leader gender
0.68
0.47
2. Leader age (year)
33.24
8.06
-.02
3. Leader education
1.25
0.47
-.03
.16
4. Dyadic tenure (year)
1.92
1.70
.25*
-.01
-.02
5. Leader moral attentiveness
4.01
0.77
-.23
.02
-.03
-.19
6. Leader moral courage
4.16
0.70
.13
.07
-.01
-.14
.35**
Note: Within-dyad level, N = 664; Between-dyad level, N = 68. Within-dyad correlations are based on within-dyad scores.
Gender was coded as 1 = male, 0 = female. Leader education was coded as 1 = middle school, 2 = high school, 3 = polytechnic, 4 = bachelor’s
degree, and 5 = master’s degree and above.
*p < .05, **p < .01 (two-tailed).

48
Table 4
Unstandardized Coefficients of MSEMs for Testing Main, Mediation and Moderation Effects in Study 2
Variables
Experienced guilt
(Day t noon)
Perceived loss of moral
credits (Day t noon)
Consideration behavior
(Day t afternoon)
Initiating structure
behavior
(Day t afternoon)
MSEM 1
MSEM 2
MSEM 1
MSEM 2
MSEM 1
MSEM 2
MSEM 3
MSEM 4
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Est.
SE
Variables on Day t-1
Experienced guilt
-.03
.03
-.04
.07
Perceived loss of moral credits
.03
.03
.06
.06
Consideration behavior
.08
.05
.08
.05
Initiating structure behavior
.10*
.04
.10*
.04
Variables on Day t
Leader morning positive affect
.05*
.02
.05*
.02
.00
.01
.00
.01
-.01
.04
-.00
.04
-.01
.04
-.01
.04
Leader morning negative affect
.19**
.07
.19**
.07
.04
.03
.02
.03
-.08
.06
-.08
.06
-.03
.06
-.03
.09
Follower morning positive affect
-.02
.02
-.01
.02
.02
.02
.02
.02
.05
.03
.06
.03
.03
.04
.03
.05
Follower morning negative affect
-.07
.05
.07
.04
.08
.04
.09
.04
-.00
.05
.01
.04
.02
.05
.03
.06
Abusive supervisor behavior
.64**
.12
.62**
.12
.29**
.09
.33**
.08
.16
.13
.12
.13
.20*
.09
.14
.16
Experienced guilt
.29**
.10
.27*
.11
.25*
.10
.25*
.12
Perceived loss of moral credits
.44*
.20
.18
.11
.41*
.19
.22
.15
Variables at the between-dyad level
Moral attentiveness (MA)
-.22
.13
-.29*
.14
.01
.15
.10
.24
Moral courage (MC)
-.63*
.29
-.22
.26
Cross-level interactions
Abusive supervisor behavior × MA
.26*
.12
.30*
.13
.03
.08
-.06
.19
Experienced guilt × MC
.25*
.12
.08
.11
Perceived loss of moral credits × MC
.25*
.12
.06
.14
Note: Within-dyad level, N = 664; Between-dyad level, N = 68. MSEM = multilevel structural equation modeling; Est. = estimate; SE = standard
error. We omitted estimates of demographics for brevity.
*p < .05, **p < .01 (two-tailed).

49
Table 5
Summary of Indirect Effects of Abusive Supervisor Behavior on Outcomes at Different Conditions of Moderators in Study 2
Conditions
Abusive supervisor behavior
 Consideration behavior
Abusive supervisor behavior
 Initiating structure behavior
Via experienced guilt
Via perceived loss of
moral credits
Via experienced guilt
Via perceived loss of
moral credits
Estimate
95% CI
Estimate
95% CI
Estimate
95% CI
Estimate
95% CI
1. High moral attentiveness, High moral courage
.36**
[.14, .58]
.20*
[.01, .39]
.25*
[.00, .50]
.15†
[-.01, .28]
2. High moral attentiveness, Low moral courage
.08
[-.15, .30]
.01
[-.02, .02]
.16
[-.03, .34]
.10
[-.13, .34]
3. Low moral attentiveness, High moral courage
.19*
[.03, .34]
.04
[-.11, .13]
.13
[-.03, .29]
.03
[-.06, .11]
4. Low moral attentiveness, Low moral courage
.04
[-.08, .15]
.00
[-.07, .14]
.08
[-.03, .19]
.02
[-.05, .09]
Indirect effect differences
Difference between Conditions 1 and 2
.28*
[.01, .56]
.19†
[-.00, .39]
.09
[-.15, .34]
.04
[-.17, .28]
Difference between Conditions 1 and 3
.17*
[.00, .35]
.16†
[-.01, .34]
.12
[-.03, .27]
.12†
[-.02, .26]
Difference between Conditions 1 and 4
.32**
[.09, .55]
.20*
[.01, .38]
.17
[-.06, .41]
.13†
[-.02, .28]
Note: Within-dyad level, N = 664; Between-dyad level, N = 68. CI = confidence interval.
90% CI = [.02, .36] for the difference between Conditions 1 and 2 and 90% CI = [.02, .31] for the difference between Conditions 1 and 3 of the
indirect effect from abusive supervisor behavior on consideration behavior via perceived loss of moral credits. 90% CI = [.01, .30] for the indirect
effect of abusive supervisor behavior on initiating structure behavior via perceived loss of moral credits under Condition 1. 90% CI = [.004, .26]
for difference between Conditions 1 and 3 and 90% CI = [.01, .28] for the difference between Conditions 1 and 4 of the indirect effect from
abusive supervisor behavior on initiating structure behavior via perceived loss of moral credits.
† p < .10, *p < .05, **p < .01 (two-tailed).

50
Figure 1. A reparative response model of perpetrating abusive supervisor behavior
Note: All paths in the model are positive. The dashed box indicates that leader moral courage
was tested at Level 3 in Study 1 but at Level 2 in Study 2 as a second-stage moderator.
Between-dyad (or leader) level
Increases in
Consideration & Initiating
Structure Behaviors
Moral Courage
(At Level 3 in Study 1)
Perpetrating in Abusive
Supervisor Behavior
Perceived Loss of
Moral Credits
Experienced Guilt
Moral attentiveness
Within-dyad level

51
Figure 2. Cross-level moderating effects of leader moral courage on the relationship of
experienced guilt with consideration behavior (Study 1)
Note: Using Preacher et al.’s (2006) approach, we tested the significance of two simple
slopes, and the t statistical results are t = .88 (γ = .10, p = .39) for the low moral courage
condition and t = 4.14 (γ = .41, p < .01) for the high moral courage condition.
2.9
2.95
3
3.05
3.1
3.15
Low Experienced Guilt High Experienced Guilt
Consideration behavior
Low Moral Courage High Moral Courage

52
Figure 3. Cross-level moderating effects of leader moral attentiveness on the relationship of
perpetrating abusive supervisor behavior with experienced guilt (Study 2)
Note: Using Preacher et al.’s (2006) approach, we tested the significance of two simple slopes,
and the t statistical results are t = 3.82 (γ = .53, p < .05) for the low moral attentiveness condition
and t = 4.59 (γ = 1.07, p < .001) for the high moral attentiveness condition.
1.2
1.25
1.3
1.35
1.4
1.45
1.5
1.55
1.6
Low Abusive Supervisor Behavior High Abusive Supervisor Behavior
Experieced guilt
Low Moral Attentiveness High Moral Attentiveness

53
Figure 4. Cross-level moderating effects of leader moral attentiveness on the relationship of
perpetrating abusive supervisor behavior with perceived loss of moral credits (Study 2)
Note: Using Preacher et al.’s (2006) approach, we tested the significance of two simple slopes,
and the t statistical results are t = 0.24 (γ = .02, p = .81) for the low moral attentiveness condition
and t = 4.28 (γ = .45, p < .01) for the high moral attentiveness condition.
1.1
1.15
1.2
1.25
1.3
1.35
1.4
Low Abusive Supervisor Behavior High Abusive Supervisor Behavior
Perceived loss of moral credits
Low Moral Attentiveness High Moral Attentiveness

54
Figure 5. Cross-level moderating effects of leader moral courage on the relationship of
experienced guilt with consideration behavior (Study 2)
Note: Using Preacher et al.’s (2006) approach, we tested the significance of two simple slopes,
and the t statistical results are t = .37 (γ = .05, p = .72) for the low moral courage condition and t
= 3.85 (γ = .53, p < .01) for the high moral courage condition.
2.35
2.4
2.45
2.5
2.55
2.6
2.65
2.7
2.75
2.8
Low Experienced Guilt High Experineced Guilt
Consideration beahvior
Low Moral Courage High Moral Courage

55
Figure 6. Cross-level moderating effects of leader moral courage on the relationship of perceived
loss of moral credits with consideration behavior (Study 2)
Note: Using Preacher et al.’s (2006) approach, we tested the significance of two simple slopes,
and the t statistical results are t = -.01 (γ = -.00, p = .99) for the low moral courage condition and
t = 2.57 (γ = 0.35, p < .05) for the high moral courage condition.
2.48
2.5
2.52
2.54
2.56
2.58
2.6
2.62
2.64
2.66
2.68
Low Perceived Loss of Moral Credits High Perceived Loss of Moral Credits
Consideration beahvior
Low Moral Courage High Moral Courage