The Role of Ethical Leadership in Shaping | My Assignment Tutor

Leaders Matter Morally: The Role of Ethical Leadership in ShapingEmployee Moral Cognition and MisconductCelia MooreBocconi UniversityDavid M. MayerUniversity of MichiganFlora F. T. ChiangChina Europe International Business SchoolCraig CrossleyUniversity of Central FloridaMatthew J. KarleskySuffolk UniversityThomas A. BirtchUniversity of South AustraliaThere has long been interest in how leaders influence the unethical behavior of those who they lead.However, research in this area has tended to focus on leaders’ direct influence over subordinate behavior,such as through role modeling or eliciting positive social exchange. We extend this research byexamining how ethical leaders affect how employees construe morally problematic decisions, ultimatelyinfluencing their behavior. Across four studies, diverse in methods (lab and field) and national context(the United States and China), we find that ethical leadership decreases employees’ propensity to morallydisengage, with ultimate effects on employees’ unethical decisions and deviant behavior. Further,employee moral identity moderates this mediated effect. However, the form of this moderation is notconsistent. In Studies 2 and 4, we find that ethical leaders have the largest positive influence overindividuals with a weak moral identity (providing a “saving grace”), whereas in Study 3, we find thatethical leaders have the largest positive influence over individuals with a strong moral identity (catalyzing a “virtuous synergy”). We use these findings to speculate about when ethical leaders mightfunction as a “saving grace” versus a “virtuous synergy.” Together, our results suggest that employeemisconduct stems from a complex interaction between employees, their leaders, and the context in whichthis relationship takes place, specifically via leaders’ influence over employees’ moral cognition.Keywords: ethical leadership, moral disengagement, moral identity, unethical behavior, employeedevianceLeadership is often singled out as a critical driver of both ethicaland unethical behavior inside organizations (Brown & Mitchell,2010; Brown & Treviño, 2006; Ng & Feldman, 2015). Priorresearch has found that ethical leadership—that is, “the demonstration of normatively appropriate conduct through personal actions and interpersonal relationships and the promotion of suchconduct to followers through two-way communication, reinforcement, and decision making” (Brown, Treviño, & Harrison, 2005,p. 120)—predicts a wide range of subordinate outcomes. Mostresearch on ethical leadership has focused on positive outcomes,from organizational commitment (Hannah, Jennings, Bluhm,Peng, & Schaubroeck, 2014) and job satisfaction (Avey, Wernsing, & Palanski, 2012), to prosocial helping behaviors (Mayer,Kuenzi, Greenbaum, Bardes, & Salvador, 2009) and voice (Tu &Lu, 2013), to task and job performance (Piccolo, Greenbaum, denHartog, & Folger, 2010). Research has also related ethical leadership to employee deviance and unethical behavior (Mayer,Aquino, Greenbaum, & Kuenzi, 2012; Mayer et al., 2009).Researchers have been slower to explore the mechanismsthrough which ethical leadership elicits these outcomes (Brown &Treviño, 2006). In keeping with the roots of ethical leadership’stheoretical grounding in social learning (Bandura, 1977, 1986) andsocial exchange (Blau, 1964) traditions, the limited number ofstudies about how ethical leadership elicits positive outcomes havefocused predominantly on mechanisms implied in the definition ofethical leadership itself. To date, the three main categories ofThis article was published Online First September 17, 2018.Celia Moore, Department of Management and Technology, BocconiUniversity; David M. Mayer, Department of Management and Organizations, Ross School of Business, University of Michigan; Flora F. T.Chiang, Organizational Behavior and Human Resource Management,China Europe International Business School (CEIBS); Craig Crossley,Department of Management, University of Central Florida; Matthew J.Karlesky, Department of Management and Entrepreneurship, Suffolk University; Thomas A. Birtch, School of Management, UniSA BusinessSchool, University of South Australia.Matthew J. Karlesky is now at the Department of Management, University of South Florida.Correspondence concerning this article should be addressed to CeliaMoore, Department of Management and Technology, Bocconi University,Via Roentgen, 1, Milan 20136, Italy. E-mail: [email protected] document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.Journal of Applied Psychology© 2018 American Psychological Association 2019, Vol. 104, No. 1, 123–1450021-9010/19/$12.00 that have been explored are (a) “demonstrating normatively appropriate conduct,” implying mediators such as rolemodeling and ethical climate (Mayer, Kuenzi, & Greenbaum,2010; Neubert, Carlson, Kacmar, Roberts, & Chonko, 2009); (b)“interpersonal relationships,” implying mediators such as trust (Ng& Feldman, 2015); and (c) “two-way communication [and] reinforcement,” implying mediators such as the desire to reciprocatethe positive social exchanges ethical leaders initiate (Piccolo et al.,2010). As some researchers have noted (Antonakis, 2017), focusing on mechanisms implied by the definition of the construct canlead to circular theorizing and runs the risk of testing explanationswhere the mediating mechanisms are actually captured by theindependent variable itself.Extending prior research, we focus on how ethical leadersinfluence morally problematic subordinate behaviors by looking athow ethical leaders shape the way employees construe decisionswith moral import. Research has identified the way an individualconstrues a decision with moral import as a key determinant ofwhether the decision ultimately made is ethical or not (Moore &Gino, 2015). How an individual construes moral choices has beenstudied in terms of how the decision is framed (Kern & Chugh,2009; Tenbrunsel & Messick, 1999), whether we are appropriatelyaware of our choice’s moral implications (Reynolds, 2006), andwhether we are attentive to those implications (Reynolds, 2008;van Gils, Van Quaquebeke, van Knippenberg, van Dijke, & DeCremer, 2015).Another way to think about how an individual construes moralchoices is in terms of that individual’s propensity to morallydisengage (Moore, Detert, Treviño, Baker, & Mayer, 2012). Bandura (1991, 1999a) originally described moral disengagement as aset of eight cognitive mechanisms—moral justification, euphemistic labeling, advantageous comparison, diffusion, displacement ofresponsibility, distorting consequences, dehumanization, and attributing blame to others—that individuals use to facilitate behavior that contravenes moral standards without feeling distress. Theextent to which individuals agree with statements that reflectmorally disengaged thinking (e.g., believing that it’s acceptable todo something if everyone else is doing it) is a strong predictor oftheir likelihood of engaging in deviant and unethical behaviors(Bandura, Barbaranelli, Caprara, & Pastorelli, 1996; Detert,Treviño, & Sweitzer, 2008; Moore et al., 2012).Offering moral disengagement as a mechanism to explain howethical leadership is related to subordinate behavior goes beyondmodeling, affective, and exchange-based explanations of ethicalleadership’s effects, and suggests that ethical leaders influencehow employees construe moral choices (their cognitions). Showing that ethical leaders can affect how their subordinates construedecisions with moral import represents an important extension ofethical leadership theory and tests a mechanism for its effects thatis more distal from previously described mediators. Further, because ethical leaders may not have uniformly profound effects ontheir subordinates, it is also important to examine subordinatecharacteristics that might govern how leaders’ behavior is associated with how their employees construe moral choices. Thus, wealso explore how employee moral identity (Aquino & Reed, 2002)may interact with ethical leadership to predict employee moraldisengagement and, ultimately, misconduct.We aim to make a number of contributions to the literature. Ourforemost contribution is to the literature on ethical leadership.First, offering moral disengagement as a mechanism to explainhow ethical leadership influences subordinate behavior goes beyond observational learning and modeling explanations of ethicalleadership’s effects, and suggests that ethical leadership influencesemployees on a cognitive as well as a behavioral level. Second, wecontribute to the literature on moral disengagement by showingleaders behavior is related to how employees construe decisionswith moral import, ultimately affecting their unethical behaviorthrough this influence over employees’ cognitions. This is a finding of central importance to organizations. If leaders influence theextent to which their employees morally disengage, organizationsneed to be more mindful of whom they promote, since leaders willshape their subordinate’s moral cognitions, ultimately affecting theextent to which they engage in deviant and unethical behavior.Third, we contribute to the moral identity literature, by showinghow the extent to which an employee’s moral identity interactswith the ethicality of his or her leader, which can ultimately resultin more virtuous outcomes, or more vicious ones.Our perspective is consistent with the view that employee misconduct is an outcome of the interaction between individuals andtheir situational context, rather than a case of a few bad apples(Martin, Kish-Gephart, & Detert, 2014; Moore & Gino, 2013;Treviño, 1986). Though we have long understood unethical behavior to be a function of the interaction between individual andsituational characteristics (Treviño, 1986), the complex ways inwhich such factors interact to produce outcomes are less oftenstudied. In this article, we provide a more comprehensive examination of how leader behavior interacts with subordinate characteristics to influence employee deviance and unethical behavior.We test our moderated mediation model in four studies spanninglab and field contexts, including multiple measures of employeemisconduct rated by different sources, data from a diverse set ofcompanies across different countries, and controlling for alternative mechanisms.Theoretical BackgroundThe role of an organizational leader is to influence how his orher subordinates are motivated and act (Bass, 1960). Employees’immediate supervisors exert a strong influence over the way employees make sense of their job and its responsibilities, determining what they consider appropriate and inappropriate behavior(Ashforth & Anand, 2003), and influencing their attitudes andperceptions (Podsakoff, Bommer, Podsakoff, & MacKenzie,2006). In other words, they create the context within which employees act (Ashforth & Anand, 2003).Ethical Leadership and Employee Moral DisengagementEthical leadership theory is grounded in social learning theory(Bandura, 1977), a fundamental premise of which is that individuals look to role models for behavioral cues and guidance. Thepower and status of leaders make them salient role models to theiremployees (Bandura, 1986). As a result, much of the early work onethical leadership focused on how the presence of ethical rolemodels is positively associated with engaging in more positivebehaviors (Treviño, Brown, & Hartman, 2003; Treviño, Hartman,& Brown, 2000). A fundamental underpinning of ethical leadership theory is that leaders have a substantial effect on their subThis document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.124 MOORE ET AL.ordinates, because they represent key models for normative andappropriate behavior, as well control over the way their subordinates’ behavior is penalized or reinforced, through both praise(intangible benefits) and rewards (tangible benefits).However, a social– cognitive perspective emphasizes that truelearning requires behavior to be integrated into an individual’sown cognitive functioning (Bandura, 1986, 1991). From this perspective, if subordinates truly learn from their leaders, ethicalleaders should do more than make subordinates more aware of andattendant to ethical concerns (Reynolds, 2008; Treviño et al.,2003) and actually affect how employees understand ethicallymeaningful decisions. Accordingly, we examine whether leaderbehavior is related to how employees construe moral decisions,through affecting their propensity to morally disengage (Moore etal., 2012).Social– cognitive theory proposes that, even though one’s moralconduct remains strongly influenced by one’s context, selfregulation ultimately governs moral behavior (Bandura, 1991).Childhood socialization helps individuals develop internallydriven self-sanctions against behaving unethically. These selfsanctions are reinforced socially as well as by one’s own reasoningabout the consequences of potential actions to oneself and others.When internal controls are working properly, when an individualanticipates a potentially morally problematic action, they becomeactivated. Self-regulatory processes then kick in to inhibit theseimmoral choices. Yet, as Bandura explained, our contexts canhabituate us to use mechanisms of moral disengagement that shift,over time, how we construe potentially immoral choices:Disengagement practices will not instantly transform considerate persons into cruel ones. Rather, the change is achieved by gradualdisengagement of self-censure. People may not even recognize thechanges they are undergoing. Initially, their self-reproof has beendiminished through repeated enactments . . . until eventually . . .inhumane practices become thoughtlessly routinized. (Bandura,1999a, p. 203)In other words, individuals can become accustomed to usingmechanisms of moral disengagement and the extent to which anindividual has become habituated in their use will affect thebehavior he or she believes is justifiable or morally acceptable.We argue that leaders are a key influence on the extent to whichtheir subordinates become habituated in the use of morally disengaged cognitions in the workplace. Leaders play a central role inemployees’ sense-making processes and are instrumental in framing their employees’ environments (Hoffman, Bynum, Piccolo, &Sutton, 2011; Piccolo & Colquitt, 2006). Over time, employeesshift their own ethical decision-making frameworks to align withthose their supervisors use (Schminke, Wells, Peyrefitte, &Sebora, 2002). Our argument builds on this work, and proposesthat leaders not only influence the formal ethical frameworks theirsubordinates use, but also the extent to which they morally disengage.There are two ways to think about how ethical leadership mightinfluence subordinates’ moral disengagement. First, highly ethicalleaders might dampen subordinates’ propensity to morally disengage. For example, a leader who sets an example of how to dothings the right way in terms of ethics makes it harder for employees to displace responsibility for their actions onto their supervisoror diffuse responsibility to others using a belief that “everyone elseis doing it.” Leaders who listen to what others say, have others’best interests in mind, and make fair and balanced decisions likelythwart the ease with which employees might disregard others’feelings and attribute the blame for their own suspect actions ontotheir victims.Pope Francis provides some concrete examples of ethical leadership behaviors that would likely dampen his followers’ propensities to morally disengage. During his first year as Pope, Francisabandoned the traditional practice of bending for a “token swipe atthe feet of twelve selected priests” on Holy Thursday and, instead,literally washed and kissed the feet of 12 juvenile delinquents,including two Muslims and two females (Carroll, 2013). Seeingthe Pope engage with marginalized people in this way would makeit more difficult for other religious leaders to dehumanize thedisenfranchised. Pope Francis also abandoned the Apostolic Palacefor a two-room apartment and replaced the papal Mercedes with aFord Focus, which likely curtailed the opportunity for other prieststo displace onto their leader responsibility for their own potentialextravagances. The Pope cannot change Church doctrine, yet hisbehaviors have shifted the ways in which that doctrine is presentedand as a senior Jesuit in the papal headquarters recently stated,“The way we practice our faith affects how we believe” (Carroll,2013).Second, leaders with low-ethical standards might amplify theirsubordinates’ propensities to morally disengage. For instance, ifone has a leader who defines success in terms of results rather thanthe means through which they are obtained, subordinates couldmore easily morally justify short-changing customers in the interest of profit (and the greater good of company performance).Leaders who demand obedience or absolute loyalty from theirsubordinates facilitate their subordinates’ displacement of responsibility, such that individual employees can shift their moralagency onto their domineering leader (Milgram, 1974). A leaderwho refers to clients or competition as a resource to be exploitedor an enemy to be overcome can support euphemistic labeling ofharmful practices or the attribution of blame for mistreatment ontothose whom they are harming. Similarly, a leader’s exclusive focuson profits, emphasizing meeting targets, goals, and operationalefficiencies can create a situation in which unethical behavior hasno apparent victim (Beu & Buckley, 2004).An example of how leaders with low ethical standards canfacilitate moral disengagement comes from the trial of KwekuAdoboli, the rogue trader from UBS ultimately sentenced to 7years in prison for illegal activity that cost the bank $2.3 billion.During his testimony, Adoboli revealed that his supervisor had toldhim, “You don’t know you are pushing the boundaries hardenough until you get a slap on the wrist” (Fortado, 2012). Evenafter having been caught numerous times for violating internalcontrols, Adoboli received only one warning from his supervisors,and his actions were never restricted, suspended, or further scrutinized (Swiss Financial Market Supervisory Authority [FINMA],2012). In fact, after earning $6 million on a trade that exceeded hisdaily limit by $100 million, his supervisor congratulated ratherthan penalized him (FINMA, 2012). This treatment from hisleaders communicated that what Adoboli was doing was perfectlyacceptable. Later testimony recounted that “the mantra comingfrom above was revenue, revenue, revenue” (Fortado & Moshinsky, 2012). When Adoboli later said that his actions had all been“for the benefit of the bank” (UBS ‘rogue trader’ sobs: ‘I only triedThis document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 125my best’, 2012), it suggests that these leadership messages facilitated his belief that his actions were in the service of the bank’sgreater good (i.e., moral justification).Moral disengagement theory states that the self-regulatorymechanisms that govern our moral conduct do not come into playunless they are activated, and the fact that self-regulatory mechanisms can be activated selectively helps explain why individualswith the same moral standards can behave differently in differentcircumstances (Bandura, 1999a). Our argument here is that leadership as a key contextual contingency that can influence theextent to which subordinates depend on morally disengaged cognitions in the course of their work, ultimately explaining whenself-regulatory mechanisms are activated (mitigating the likelihood of behavior that counters moral standards) or not (amplifyingthis behavior). A highly ethical leader may remind employees oftheir own internal sanctions against immoral behavior, whereas aleader with low ethical standards will provide no such activation.This argument is consistent with the idea of moral disengagement as a “dynamic disposition” (Bandura, 1999b) that can shift asa function of a wide range of situational contingencies. For example, research has documented that individuals’ tendencies to morally disengage are heightened in environments in which individuals feel coerced or pressured (Hodge & Lonsdale, 2011), cannotparticipate in setting their own performance goals (Barsky, 2011),where there is a wide berth to behave in self-serving ways withoutnoticeable consequences (Kish-Gephart, Detert, Treviño, Baker, &Martin, 2014; Shu, Gino, & Bazerman, 2011), or when one feelsparticularly connected to someone who is engaging in unethicalbehavior themselves (Gino & Galinsky, 2012).Together, the work showing the importance of leaders in determining employee decisions and behaviors, coupled with the understanding of moral disengagement as a “dynamic disposition”susceptible to contextual influence, suggests that ethical leadershipwill dampen employees’ tendencies to morally disengage.Hypothesis 1: Ethical leadership will be negatively related toemployee moral disengagement.Moral Disengagement and Employees’ Deviant BehaviorHigh levels of moral disengagement predict a wide range ofgenerally undesirable behaviors, such as unethical work behavior(Detert et al., 2008; Moore et al., 2012), social undermining(Duffy, Scott, Shaw, Tepper, & Aquino, 2012), sexual harassment(Claybourn, 2011), and computer hacking (Rogers, 2001). Themagnitude of the effects these studies report is also impressive. Inone recent study, the effect size for the role of moral disengagement in unethical behavior was .36 (aggregating across fourstudies and 857 individuals, see Moore et al., 2012). The nextlargest effect size we could find for an individual dispositionaffecting unethical behavior was .27 for Machiavellianism(aggregating across 11 studies and 2,290 individuals, see KishGephart, Harrison, & Treviño, 2010). If ethical leadership is associated with the extent to which their subordinates morally disengage, this will ultimately increase the likelihood that thosesubordinates will make morally problematic decisions and engagein unethical behavior. We intentionally link employee moral disengagement with a broad set of morally undesirable outcomes,consistent with extensive research documenting the numerousproblematic behaviors resulting from the propensity to morallydisengage (for a review, see Moore, 2015).Hypothesis 2: Employee moral disengagement will be positively related to employee deviance and unethical decisionmaking.Hypothesis 3: Moral disengagement will mediate the relationship between ethical leadership and employee deviance andunethical decision making.The Moderating Role of Moral IdentityAccording to social– cognitive theory, moral functioning is explicitly interactive, an outcome of the interplay between personaland social influences (Bandura, 1991, 2002). The salience ofethical leadership on an individual’s level of moral disengagementwill thus depend on the individual characteristics they bring to thesituation. Moral identity, an individual trait referring to the extentto which one’s self-conception, is strongly rooted in moral characteristics (Aquino & Reed, 2002). Moral identity is a powerfulmotivator of pro-social behavior (Reed & Aquino, 2003; Reynolds& Ceranic, 2007) as well as an important inhibitor of antisocialbehavior (Aquino & Becker, 2005). The motivational potency ofmoral identity arises from the psychological desire to strive forself-consistency (Blasi, 1984), because acting immorally wouldelicit feelings of inconsistencies with the moral self that is centralto one’s identity.Aquino and Reed (2002) identified two main facets of moralidentity: symbolization and internalization. The symbolization dimension taps the degree to which one’s actions reflect one’s moralself. It is associated with less consistent outcomes than the internalization dimension (Jennings, Mitchell, & Hannah, 2014). Moreover, because this aspect of moral identity is reflected in one’sactions, it is more often associated with pro-social behavior (ratherthan the motivation to resist unethical behavior), particularly whenindividuals are publically recognized or rewarded for their moralactions (Ormiston & Wong, 2013; Winterich, Aquino, Mittal, &Swartz, 2013).In contrast, the internalization dimension taps how central morality is in an individual’s working self-concept. It is identifiedmore consistently as the facet of moral identity that denotes “thestrength to act morally” (Reynolds & Ceranic, 2007, p. 1621),helping one self-regulate against temptations to engage in unethical behavior (Gino, Schweitzer, Mead, & Ariely, 2011; Mayer etal., 2012; O’Fallon & Butterfield, 2011; Perugini & Leone, 2009).Research has also documented a negative relationship betweenmoral identity internalization and moral disengagement (Aquino,Reed, Thau, & Freeman, 2007; Detert et al., 2008; McFerran,Aquino, & Duffy, 2010; Moore et al., 2012). Thus, we focus onmoral identity internalization as a potential moderator of oureffects (and is what we refer to hereafter when using the termmoral identity).We represent the possible patterns that a moderated relationshipcould take in Figure 1. Panel A represents what the relationshipbetween ethical leadership and moral disengagement would looklike if moral identity does not moderate it. In this case, we wouldexpect to observe two independent effects. Moral identity shouldhave a dampening effect on moral disengagement (as, indeed,several studies have found; see Aquino et al., 2007; Detert et al.,This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.126 MOORE ET AL.2008; Hardy, Bean, & Olsen, 2015; Moore et al., 2012) as shouldethical leadership (our prediction in Hypothesis 1). If there is nomoderation, the two lines should be parallel: Those with a weakmoral identity (dashed bar in Figure 1) will be more morallydisengaged than those with a stronger moral identity (solid bar inFigure 1), and the two bars will both slope downward and behigher for those with less ethical leaders (the left side of Figure 1),and lower for those with highly ethical leaders (the right side ofFigure 1).If there is a moderating relationship, however, the influence ofethical leadership on moral disengagement will differ as a functionof the strength of the employee’s moral identity, and the two lineswill deviate from parallel. Moral identity has been recognized asan important moderator of the effect of several contextual variables on individual attitudes and behavior (Aquino, McFerran, &Laven, 2011; Aquino et al., 2007; Caldwell & Moberg, 2007;Reed, Aquino, & Levy, 2007). How this moderation plays out,however, seems to depend largely on the context in which moralidentity is activated and on the outcome it influences. Extantresearch provides evidence for two possible patterns the moderation might take, which we represent in Panels B and C of Figure1, and theorize, in the following text, as competing hypotheses.Figure 1, Panel B: Ethical leaders as providing a “savinggrace.” One perspective is that ethical leaders will have amore substantial influence on subordinates with a weak moralidentity, whereas subordinates with a strong moral identity willbe largely immune to it. The argument for this pattern hinges onthe fact that the moral self is chronically accessible to individuals with a strong moral identity as part of their workingself-concept (Markus & Kunda, 1986). The drive for selfconsistency will require these individuals to behave in alignment with their moral self-concept regardless of the contextualtriggers their local environment presents, suggesting they willneither be led astray by less ethical leaders, nor overly influenced by highly ethical ones. Rather, their moral identity willbe consistently associated with low levels of moral disengagement and morally problematic behavior, regardless of theirleader.A number of studies support the idea that individuals with astrong moral identity will be more immune to contextual triggers that may tempt one to engage in unethical behavior. Forexample, individuals with strong moral identities were able toresist the temptation to cheat after their self-regulatory resources were depleted, while individuals with weak moral identities succumbed (Gino et al., 2011). In another study, theauthors found that a highly central moral identity made one lesssusceptible to the cognitive protections that moral disengagement affords (Aquino et al., 2007). A strong moral identity, inother words, may provide a vaccination against morally disengaged reasoning.Conversely, the moral self is less chronically accessible forindividuals low in moral identity. As a result, such individualsmay be more affected by contextual cues, in particular theirleader’s ethical conduct. O’Fallon and Butterfield (2011) foundthat individuals without a strong moral identity were particularly influenced by the ethicality of their peers: Those withunethical peers were more likely to engage in unethical behavior, if their moral identity was not central to their self-concept.Caldwell and Moberg (2007) also found results consistent withthis idea in their study exploring the role of ethical culture inactivating moral imagination: Ethical culture only amplifiedindividuals’ moral imagination for individuals without a strongmoral identity; the moral imagination of those with a strongmoral identity was consistently high, regardless of culture. In asimilar way, highly ethical leaders may function as a “savinggrace” for individuals with weak moral identities, providing acontextual cue that dampens their moral disengagement tendencies. Taken together, these results lead to the following hypothesis:Hypothesis 4a: The mediated effect of ethical leadership onemployee deviance and unethical decision making via moraldisengagement will be moderated by moral identity, such thatFigure 1. Graphic representations of the competing predictions of Hypothesis 4. Panel (A) The relationship between ethical leadership and moraldisengagement is not moderated by moral identity. Panel (B) Ethicalleaders as a “saving grace” (Hypothesis 4a): Ethical leaders will influencethose low in moral identity, whereas individuals high in moral identity willhave stable, low levels of moral disengagement. Panel (C) Ethical leadersas catalyzing a “virtuous synergy” (Hypothesis 4b): Ethical leaders willinfluence those high in moral identity, while individuals low in moralidentity will have stable, high levels of moral disengagement.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 127the positive effect of ethical leaders will be stronger for thoselow in moral identity than for those high in moral identity.Figure 1, Panel C: Ethical leaders as catalyzing a “virtuoussynergy.” A second perspective on moral identity leads to theopposite prediction: that ethical leaders will have a more substantial influence on subordinates with a strong moral identity, whereassubordinates with a weak moral identity will be largely immune totheir influence. This view focuses on moral identity as beingsubject to different degrees of “activation potential” (Higgins &Brendl, 1995), depending on its centrality in one’s self-concept. Inthis view, when an individual’s morality is highly central in his orher self-concept, it is more easily accessible, and thus more easilyactivated by situational cues. As Aquino and colleagues argue, “ifa situational factor increases the current accessibility [of moralidentity] within the working self-concept, then it strengthens themotivation to act morally” (Aquino, Freeman, Reed, Felps, & Lim,2009, p. 123).From this perspective, individuals with highly central moralidentities will be more susceptible to the positive situational influences of highly ethical leaders because they will assign morerelevance and weight to the ethically positive aspects of theircontext. Evidence in support of this point of view can be found inAquino and colleagues’ study that found that moral priming hadmore durable positive effects on cooperative behavior for individuals with highly central moral identities (Aquino et al., 2009). Theauthors argue for the importance of salient situational cues inreinforcing individuals’ moral identities, particularly when otheravailable situational cues may dampen the importance of moralitywithin one’s working self-concept.Leaders represent a particularly effective and available contextual cue determining appropriate behavior. In fact, Aquino andcolleagues explicitly discuss the importance of studying exemplarsin terms of how they can positively prime more virtuous behavior,particularly among those with highly central moral identities(Aquino et al., 2009). Reynolds and colleagues made a similarargument about how implicit assumptions interact with cues in theenvironment in shaping moral behavior (Reynolds, Leavitt, &DeCelles, 2010). Although merely recalling the moral behavior ofgeneral “others” does not seem to affect one’s moral self-view orunethical behavior (Jordan, Mullen, & Murnighan, 2011), leadersare particularly well situated to be a moral exemplar to theiremployees (Moberg, 2000).This second argument focuses on the “virtuous synergy” thatmight be created when an employee with a highly central moralidentity is paired with a highly ethical leader. Empirical support forthis argument is provided in an article that examined the interaction between moral identity centrality and moral elevation ininspiring charitable behaviors. Individuals watched a moral elevation video, a video designed to elicit a positive mood, or a controlvideo and were then asked to donate to a charitable cause. Individuals who watched the moral elevation video were more likely todonate, particularly so if they had highly central moral identities(Aquino et al., 2011). The authors concluded that individuals withstrong moral identities are positively predisposed to engage inmore ethical behavior as a function of ethical environmental cues,whereas individuals with weak moral identities will be moreindifferent to these virtuous environmental triggers. This perspective leads to the following hypothesis:Hypothesis 4b: The mediated effect of ethical leadership onemployee deviance and unethical decision making via moraldisengagement will be moderated by moral identity, such thatthe positive effect of ethical leaders will be stronger for thosehigh in moral identity than for those low in moral identity.Overview of StudiesWe test our hypotheses in four studies using both field andlaboratory data. Our goal in using this multimethod and multicontext design was to ensure that our conclusions are robust andrepresent “full cycle” organizational research (Chatman & Flynn,2005). In Study 1, we examine the mediating role of moral disengagement in explaining the relationship between ethical leadershipand deviant and unethical workplace behaviors and compare moraldisengagement to two other potential mediators (Machiavellianismand moral attentiveness) that also tap differences in how individuals approach moral decisions. In Study 2, we manipulate ethicalleadership in an experimental design to investigate its causal effecton employee moral disengagement, and subsequent unethical behavior. We also test the role of the centrality of one’s moralidentity as a potential moderator of this mediating effect, and findthe mediated relationship holds for individuals low (but not high)in moral identity. In Study 3, we use a three-wave field survey of208 employees and their supervisors in China to test our proposedmoderated mediation a second time. Though we again find thatmoral identity moderates the mediated relationship between ethicalleadership and employee behavior through moral disengagement,in this case we find the mediated relationship holds for individualshigh (but not low) in moral identity. A fourth study attempts toshed light on why the moderated relationship takes different forms.The results show that moral identity moderates the mediated effectbut replicates the pattern found in Study 2, which suggests thatsomething in the broader cultural context (China vs. the UnitedStates) or something unique to the organizational context of Study3 are the most likely culprits driving these different results.Study 1Study 1 provides an initial test of the role of moral disengagement as a mediator of the relationship between ethical leadershipand unethical employee behavior, in a broad-based sample ofemployees, their supervisors and coworkers. We also use Study 1as an opportunity to compare the explanatory roles of moraldisengagement, Machiavellianism and moral attentiveness intranslating ethical leadership’s effects on negative employee behaviors. These alternative mechanisms represent additional measures of how individuals construe decisions with ethical import,and affect how potential moral choices are understood. Testingthem simultaneously represents a useful test of the explanatorypower of moral disengagement.MethodParticipants and ProcedureWe extended an invitation to participate in this study to 625undergraduate students attending a large university in the UnitedStates. A total of 432 students volunteered to take part; theyThis document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.128 MOORE ET AL.received extra credit for their involvement. Each student was givena packet that included three instruction sheets: one for a focalemployee, one for the employees’ immediate supervisor, and onefor a coworker of the focal employee. The student then returned tothe researchers the personal e-mail address of the focal employee,supervisor, and coworker whom they solicited. Each of these 1,296individuals received an individual link via their personal e-mailaddress to the version of the survey appropriate to them (employee,supervisor, or coworker). This snowball sampling method (e.g.,Morgeson & Humphrey, 2006; Skarlicki & Folger, 1997) has beenused to reduce potential error caused by single-source bias. We haddata available for all relevant measures for 252 focal employees (a58% response rate) and were able to match 193 employeesupervisor dyads and 202 employee-co-worker dyads on all therelevant measures.The 252 focal employees were employed in a variety of industries, including finance (19%), health care and social assistance(15%), manufacturing (13%), professional services (10%), education (9%), real estate (4%), and retail trade (4%). Of the focalemployees, 43% were male, and 92% were full-time employees.Their average age was 44.0 years (SD 12.3). Nine percent hadbeen at their employer for less than 1 year, 31% had an organizational tenure of between 1 and 5 years, 18% had between 6 and 10years, and 43% had 11 or more years. Two thirds (68%) were inmanagement positions, and 32% were in nonmanagerial roles.Seventy-eight percent of the respondents were White, 3% Hispanicor Latino/a, 4% African American, 11% Asian American, and 3%indicated “other.”Of the 193 supervisors we were able to match back to employeeresponses, 68% were male and 98% were full-time employees (2%worked part-time). Their average age was 49.5 years (SD 9.1).Nineteen percent had between 0 and 5 years organizational tenure,20% had 6 to 10 years tenure, 12% had 11 to 15 years tenure, and49% had 16 or more years of organizational tenure. Eighty-threepercent of the respondents were White, 2% Hispanic or Latino/a,2% African American, 10% Asian American, and 3% indicated“other.” Of the 202 coworkers we were able to match back toemployee responses, 43% were male and 91% were full-timeemployees (9% worked part-time). Their average age was 42.0years (SD 11.7). Nine percent had been at their employer forless than 1 year, 34% had between 1 and 5 years organizationaltenure, 22% had 6 to 10 years tenure, and 35% had 11 or moreyears of organizational tenure. Eighty-two percent of the respondents were White, 5% Hispanic or Latino/a, 2% African American,10% Asian American, and 1% indicated “other.”The University of Michigan’s Institutional Review Board approved this research protocol (HUM00107807) under the application “Field Survey–MO 300/302.” These data are part of a broaderdata collection effort, and this study is the first publication fromthat effort.MeasuresAll response scales ranged from 1 (strongly disagree) to 7(strongly agree), unless otherwise noted.Ethical leadership. Focal employees completed the 10-itemethical leadership scale developed by Brown and colleagues(2005). Sample items include, “My supervisor disciplines employees who violate ethical standards” and “My supervisor discussesbusiness ethics or values with employees” ( .93).Moral disengagement. Focal employees completed an eightitem measure of moral disengagement developed by Moore andcolleagues (2012). The stem for the items was “At work,” and asample item is, “taking personal credit for ideas that were not yourown is no big deal” ( .85).Machiavellianism. Focal employees completed the five-itemMach version (Rauthmann, 2013) of the traditional Mach-IVscale (Christie, 1970). This measure was developed using itemresponse theory to create a more parsimonious measure by identifying the items from the original scale that provide the mostinformation and best measurement precision (e.g., “Anyone whocompletely trusts anyone is asking for trouble”; .68).Moral attentiveness. Seven items (e.g., “In a typical day, Iface several ethical dilemmas”) measured focal employees’ perceptual moral attentiveness, the extent to which “the individualrecognizes moral aspects in everyday experiences” ( .74), andfive items (e.g., “I regularly think about the ethical implications ofmy decisions”) measured reflective moral attentiveness, the extentto which “the individual regularly considers moral matters” ( .84; Reynolds, 2008, p. 1038).Employee organizational deviance. We measured organizational deviance using Bennett and Robinson’s (2000) 12-itemmeasure (e.g., “This person has taken property from work withoutpermission”), reported both by the focal employees’ supervisor aswell as a coworker ( .88, .90, respectively).Employee unethical behavior. We also asked both supervisors and coworkers to report on the focal employee’s unethicalbehavior, with Akaah’s (1996) 17-item measure of observed unethical behavior in the workplace, using a scale that ranged from1 (not at all), to 4 (occasionally), to 7 (frequently). Raters wereprovided the stem, “To what extent does this employee [yoursubordinate/your coworker]” followed by practices such as “usecompany services for personal use,” “falsify time/quality/quantityreports,” “divulge confidential information,” and “fail to reportothers’ violations of company policies” ( .92 for supervisorsand .87 for coworkers).Control variables. On the basis of previous findings (Barling,Dupré, & Kelloway, 2009; Mitchell & Ambrose, 2007), we alsoasked employees to report their gender, age, organizational tenure(1 less than a year, 2 1–5 years, 3 6–10 years, 4 11–15years, 5 16–20 years, 6 21–25 years, and 7 25 years),and whether they had managerial responsibilities.Results and DiscussionTo examine the distinctiveness of the measured variables, weconducted a confirmatory factor analysis (CFA; maximum likelihood, Lisrel 8.8; Jo¨reskog & So¨rbom, 2006). The measurementmodel consisted of nine factors: ethical leadership, moral disengagement, Machiavellianism, perceptual moral attentiveness,reflective moral attentiveness, coworker-rated organizationaldeviance, coworker-rated unethical behavior, supervisor-ratedorganizational deviance, supervisor-rated unethical behavior.Because of the large number of constructs and items relative to thesample size, we used the balanced item parceling technique described in Little, Rhemtulla, Gibson, and Schoemann (2013), withthree parcels per construct. The measurement model providedThis document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 129good fit to the data: 2(288) 393.94 p .01, (root mean squareerror of approximation [RMSEA] .049, confirmatory fit index[CFI] .97, standardized root mean square residual [SRMR] .053. We compared this model with a five additional models: Aseven-factor model that combined deviance and unethical behaviors into respective supervisor and coworker factors, 2(303) 728.27 p .01, (RMSEA .095, CFI .90, SRMR .068)provided a significantly worse fit to the data, 2(15) 334.33p .01, as did a six-factor model that combined deviance andunethical behaviors into a single factor, 2(309) 1288.40 p .01, (RMSEA .143, CFI .80, SRMR .120), 2(21) 894.46 p .01, a model that combined independent variables intoone factor and dependent variables onto a second factor, 2(323) 2255.99 p .01, (RMSEA .196, CFI .58, SRMR .170),2(35) 1862.05 p .01, and a single factor model, 2(324) 2569.31 p .01, (RMSEA .211, CFI .52, SRMR .180),2(36) 2175.37 p .01. Because the hypothesized measurement model demonstrated a superior fit to the data compared to allcomparison models, we were confident that our measures wereadequate and distinct from one another.Table 1 presents means, standard deviations, and correlationsamong the study variables, and Table 2 presents the results ofmodels testing our predicted mediation. We ran these models usingthe PROCESS macro for SPSS, that is designed to test multiplemediators simultaneously, as well as provide bootstrapped estimates of the significance of indirect (mediated) effects (Hayes,2013). As control variables did not impact the patterns of results orsignificance of findings, we excluded them from tables for purposes of clarity and parsimony (Carlson & Wu, 2012); however,results including the control variables are available from the authors.Under the heading “Mediator models” in Table 2, we present theresults of eight regression models, demonstrating how ethicalleadership predicts each of our four potential mediator variables(moral disengagement, Machiavellianism, perceptual moral attentiveness, and reflective moral attentiveness) for both supervisorreported outcomes and coworker-reported outcomes. Ethical leadership is significantly related to both moral disengagement(b .18, SE .06, p .006 in the models predictingsupervisor-reported outcomes; b .17, SE .06, p .005 inthe models predicting coworker-reported outcomes) and Machiavellianism (b .24, SE .07, p .002 in the models predictingsupervisor-reported outcomes; b .16, SE .07, p .018, inthe models predicting coworker-reported outcomes), providingsupport for Hypothesis 1. Of the two dimensions of moral attentiveness, ethical leadership is marginally positively related to reflective moral attentiveness and only in the models for supervisorreported outcomes (b .19, SE .11, p .060). These modelspresent what is typically noted as the “A” paths in mediator modelsand suggest that ethical leadership is related to moral disengagement and Machiavellianism, but not moral attentiveness.Under the heading “Dependent variable models” in Table 2, wepresent the results of the four regression models predicting ouroutcomes of interest, when all of our mediators are included in themodels, as well as our independent variable (ethical leadership).Moral disengagement is the only mediator that consistently predicts the outcome across the four dependent variables, albeit in oneof the models this relationship is only marginally significant. Theonly other potential mediator that significantly predicts an outcome is perceptual moral attentiveness, which is significantlynegatively related to supervisor-reported unethical employee behavior (b .09, SE .04, p .025). These models present whatis typically noted as the B paths in mediator models, and suggestthat, controlling for the independent variable (ethical leadership)and the other potential mediators, moral disengagement is the mostconsistent predictor of unethical employee outcomes, providingsupport for Hypothesis 2.Under the heading “Indirect effects” in Table 2, we present theestimates (from 5,000 bootstrap samples) of the indirect effects ofethical leadership on unethical employee outcomes via the potenTable 1Study 1: Means, Standard Deviations, and Correlations for Relevant VariablesVariable M SD 1 2 3 4 5 6 7 8 9 10 11 12 131. Sex (male 1, female 0) .43 .502. Age 44.0 12.3 .033. Tenure 3.45 1.8 .12† .554. Management (yes 1, no 0) .68 .47 .04 .30 .225. Ethical leadership 5.85 .90 .03 .03 .03 .10 (.93)6. Moral disengagement 1.76 .78 .16 .17 .11† .10 .20 (.85)7. Machiavellianism 2.71 .88 .04 .07 .04 .11† .19 .38 (.68)8. Perceptual moral attentiveness 3.48 1.01 .09 .09 .05 .08 .01 .04 .11 (.74)9. Reflective moral attentiveness 4.59 1.23 .00 .01 .01 .05 .11† .12† .01 .54 (.84)10. Supervisor-reported organizational deviance11. Supervisor-reportedunethical behavior1.†.30 .21.03.03 (.88) .22 .09.02.71 (.92) 12. Coworker-reported organizational deviance13. Coworker-reportedunethical behavior1. . . .28 (.90)1.32.45 . .08.09.15 .28 .26 .55 (.87) Note. Depending on pairwise availability of data, sample sizes range from 143 (supervisor– coworker correlations) to 252 (focal employee–focal employeecorrelations). Variables in rows 1 through 9 are reported by the focal employee. Tenure was measured categorically (1 less than 1 year, 2 1–5 years,3 6–10 years, 4 11–15 years, 5 16–20 years, 6 21–25 years, 7 25 years). Alpha reliabilities are on the diagonal in parentheses.†p .10. p .05. p .01.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.130 MOORE ET AL.tial mediators (the AB paths in mediation models) and as the 95%confidence intervals. Our results show that the indirect effects ofethical leadership through employee moral disengagement weresignificant for supervisor-reported deviance (b .035, 95% CI[ .106, .006]) and for both supervisor-reported and coworkerreported unethical employee behaviors (b .034, 95% CI[ .098, .008] and b .022, 95% CI [ .060, .004], respectively), providing support for Hypothesis 3. Moral disengagementdid not mediate the relationship between ethical leadership andcoworker reported organizational deviance, though we note thatnone of the alternative mediators we tested significantly mediatedthis relationship either.Thus, our results provide broad support for the argument thatemployee moral disengagement mediates the relationship betweenethical leadership and employee deviance and unethical behavior.Even when testing multiple alternative mediators simultaneously,none of them show the same mediating effect as moral disengagement does in this relationship.Study 2Although results from Study 1 suggest that ethical leadershiphas a negative effect on employee deviance and unethical behaviorthrough moral disengagement, the correlational nature of the studyprecludes causal claims. To provide better evidence that ethicalleadership affects the extent to which their employees morallydisengage, in Study 2 we use an experimental design that uses amanagerial decision-making simulation in which we manipulatedtheir supervisor’s level of ethical leadership. Random assignmentinto high and low-ethical leadership conditions also helps rule outalternative explanations for our findings in Study 1, such as selection or attraction. Methodologically, this type of decisionmaking simulation provides participants with more a realisticorganizational task than scenarios (Treviño, 1992), and has beenused successfully to study unethical behavioral outcomes (Reynolds et al., 2010; Treviño & Youngblood, 1990). We also useStudy 2 to provide a first test of the moderating role of the moralidentity.MethodParticipantsTwo hundred 45 individuals participated online through Amazon Mechanical Turk (MTurk). MTurk is an online labor market,where “requesters” can post short tasks for “workers” to completefor a small fee. Data collected through MTurk is of comparablequality to data collected through more traditional methods(Buhrmester, Kwang, & Gosling, 2011; Horton, Rand, & Zeckhauser, 2011; Paolacci, Chandler, & Ipeirotis, 2010). Participantswere 54% male, and their mean age was 32.6 years old (SD 8.7).Participants were paid $3 for participating and were informed thatthey would earn a bonus of up to $1 extra if their performanceTable 2Study 1: Regression Coefficients and Standard Errors for Model Testing Whether the Effect of Ethical Leadership on EmployeeDeviance and Unethical Behavior is Mediated by Moral DisengagementFor supervisor-reported outcomes For coworker-reported outcomesMediator model Coefficient SE R2 Coefficient SE R2Model 1: Constant 2.81 .38 2.77 .36Ethical leadership ¡ Moral disengagement .18 .06 .04 .17 .06 .04Model 2: Constant 4.14 .44 3.62 .39Ethical leadership ¡ Machiavellianism .24 .07 .05 .16 .07 .03Model 3: Constant 3.51 .01 3.26 .45Ethical leadership ¡ Perceptual moral attentiveness .01 .09 .00 .04 .08 .00Model 4: Constant 3.47 .59 3.72 .58Ethical leadership ¡ Reflective moral attentiveness .19† .11 .02† .15 .09 .01Dependent variable (DV) modelOrganizationaldevianceUnethicalbehaviorOrganizationaldevianceUnethicalbehaviorConstant 1.07 .38 1.53 .32 1.53 .32 1.44 .26Moral disengagement .20 .06 .19 .05 .10† .05 .13 .04Machiavellianism .08 .05 .05 .04 .03 .05 .01 .04Perceptual moral attentiveness .00 .05 .09 .04 .00 .05 .03 .04Reflective moral attentiveness .00 .04 .05 .04 .00 .04 .03 .03Ethical leadership .04 .05 .09 .04 .07† .04 .02 .03R2 .11 .16 .06 .07Indirect effects (Ethical Leadership ¡ DV) Effect [95% CI] Effect [95% CI] Effect [95% CI] Effect [95% CI]Via moral disengagement .035 [.106, .006] .034 [.098, .008] .016 [ .056, .002] .022 [.060, .004]Via machiavellianism .018 [ .055, .004] .012 [ .038, .004] .006 [ .036, .012] .001 [ .014, .021]Via perceptual moral attentiveness .000 [ .011, .010] .001 [ .017, .022] .000 [ .012, .006] .001 [ .016, .001]Via reflective moral attentiveness .001 [ .015, .027] .010 [ .001, .041] .000 [ .017, .011] .004 [ .032, .002]Note. N 193 for supervisor reported outcomes, and N 202 for coworkers reported outcomes. We report the 95% confidence intervals (CIs) calculatedusing 5,000 bootstrap samples, with lower and upper limits in brackets. Statistically significant indirect effects are in bold face text.†p .10. p .05. p .01.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 131warranted it. In fact, this bonus was paid to all participants, but wewanted to ensure that participants were taking the choices theymade seriously and created a more realistic simulation of an actualmanagerial decision-making task.Design and ProcedureWe used a between-subjects design and included an experimental manipulation of supervisor ethical leadership. Measures werecollected in the following order: (1) Before the participants enteredthe main part of the study, they completed a measure of moralidentity centrality (Aquino & Reed, 2002). (2) Participants thenentered the simulation and told that they had recently been promoted to the position of insurance claims manager at the managedcare company where they worked. In this new role, they wouldevaluate the more complex or controversial claims for reimbursement and had the authority to decide whether to approve or denythem. They were informed that they would receive a series ofemails which would contain queries about contested claims, andthey would have to decide whether to approve or deny each ofthem. We intentionally chose this context for our study as it offersparticipants the potential to make decisions that are realistic andhave been identified as morally problematic (Chan, 2002; Rimler& Morrison, 1993). The participants then (3) read the experimentalmanipulation, and (4) responded to a set of four items designed tomeasure temporally activated moral disengagement as our mediator. We then (5) asked the attention check question and (6)engaged the participants in five decisions that comprise the dependent variable. After the main measures were collected, we (7)included a manipulation check and (8) collected demographics.The first e-mail participants read was a welcome e-mail fromtheir new supervisor. It contained our ethical leadership manipulation, which we based directly on behaviors Brown and colleagues (2005) identified as central to the ethical leadership construct. The e-mail read as follows:Dear [insert name],Congratulations on the new position! As an insurance claims supervisor, you need to do your best to manage client relationships whileensuring our costs are kept under control.[High ethical leadership] I will measure your performance both byyour results and how those results are obtained. I want you to focus onmaking claims decisions that are fair and balanced. When you aredoing so, you need to ask yourself, “What is the right thing to do”?You can always question others, including me, during the course ofyour work.It’s important to do what’s best for the business, but making decisionsthat violate your values are not worth it. I will discipline employees whoviolate standards of fair treatment of our clients. We want to ensure thebest practices of the insurance industry for our clients. They are justpeople like us, wanting the best care for themselves and their families.[Low ethical leadership] I measure your performance by your results.When you are doing processing claims, you need to ask yourself,“What is best for the bottom line”? Frankly, I am not that interestedin other people’s views. I’m the supervisor, and your job is to followmy orders. You’re not here to ask questions.It’s important to do what’s best for the business. Bringing in your ownvalues here just complicates things. I can overlook decisions that push theboundaries as long as they meet our minimum responsibility to clients.We want to ensure our clients receive the most common practices of theinsurance industry. They will otherwise try to get away with anything,attempting to squeeze everything they can out of us.I look forward to working with you in this new capacity.Best regards,Preston Manfred, Senior VP, Claims AssessmentAt this point in the experiment, we included an attention checkdesigned to eliminate participants who may not have been attending carefully enough to the materials. Immediately after readingthis e-mail and paging forward [participants could not go backward in the survey], but before beginning their main task ofprocessing claims, they were asked to recall the name of thesupervisor whose e-mail they just read. Participants who wereunable to recall the supervisor’s name (N 45) were directed tothe end of the study and did not complete any of the othermeasures. We report our results for the 200 participants whopassed the attention check and completed the decision-making taskthat comprised our dependent variable. The London BusinessSchool Ethics Review Board approved this research protocol (No.REC322) under the application “Ethical Leadership and DevianceThrough Moral Disengagement.”MeasuresAll response scales ranged from 1 (strongly disagree) to 7(strongly agree) unless otherwise noted.Moral identity centrality. We measured the extent to whichmoral identity was central to participants’ identities using Aquinoand Reed’s (2002) internalization subscale of their measure of theself-importance of moral identity. Respondents were presentedwith a set of nine adjectives (e.g., caring, compassionate, fair,honest) and informed that the adjectives represent “some characteristics that might describe a person.” Respondents assessed thedegree to which they agreed with five statements about the characteristics represented an important part of their identity. A sampleitem was, “Being someone who has these characteristics is animportant part of who I am” ( .80).Moral disengagement. In this study, we were focused on thecausal role of ethical leadership on a specific set of ethical decisionsvia moral disengagement. As others have both noted and empiricallydemonstrated (Kish-Gephart et al., 2014; White, Bandura, & Bero,2009), different situations evoke the use of specific moral disengagement mechanisms rather than all of them equally. In experimentalsettings, moral disengagement is typically measured with items thatreflect the moral disengagement mechanisms most relevant to theexperimental context (Gino & Galinsky, 2012; Shu et al., 2011). Theexpectation in these studies is that the context will shift in-themoment morally disengaged cognitions, not dispositional moral disengagement. Thus, we focused on measuring aspects of moral disengagement of greatest relevance to our experimental context.Making the decision to deny health care insurance coverage topatients causes harm to them. As Bandura noted, in organizational(collective) contexts, self-exoneration for harm is often accomplishedthrough displacement and diffusion of responsibility, as well asthrough disparaging the victims (White et al., 2009). We thus focusedon those mechanisms in measuring moral disengagement in thisThis document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.132 MOORE ET Items were adapted from the Moore et al. (2012) measure(adaptations are italicized, and original wording is in brackets). Weused two items to measure displacement and diffusion of responsibility: “People shouldn’t be held accountable for doing questionablethings when they were just doing what an authority figure told themto do,” “People can’t be blamed for doing things that are technicallywrong when everyone else is [all their friends are] doing it too.”Blaming customers for their role in the harm the corporation causedhas been a longstanding strategy in industries that directly damagecustomers’ health, such as the lead industry (White et al., 2009). Thus,we also asked questions to measure attribution of blame: “People whodon’t get what they want [get mistreated] have usually done something to bring it on themselves” and “Sometimes in professionalcontexts you need to ensure that you are not taken advantage of byindividuals who don’t deserve it” ( .61).Unethical decisions. Participants had to make decisions aboutfive cases, each drawn from contemporary news accounts abouthealth care insurance companies that denied customers’ reimbursement requests for ethically questionable reasons. One decision, takendirectly from the news (Rosenthal, 2014), involves whether to denycoverage for additional services a patient received while under generalanesthetic, when he or she was, therefore, without the ability toconsent to the extra costs. Another involves a request to reimburse adoctor’s charge, when all evidence in advance of incurring the costsuggested the doctor was in the managed care network and thuscovered by the policy (Rosenthal, 2014).We intentionally designed the experiment to reduce demandeffects that can unduly influence participants to respond in whatthey perceive to be socially desirable ways (Zizzo, 2010). Using adesign that requires participants to make decisions that includeoptions that are widely acknowledged as morally problematic butare not so clear cut that they are likely to trigger simplistic sociallydesirable responding allows us to be best able to see how subtleleadership messages affect the choices their subordinates make.Our outcome measure was a count of the number of insuranceclaims that were denied for morally problematic reasons. Thisnumber could range from 0 to 5 (M 2.74, SD 1.18).Ethical leadership. At the end of the study, as a manipulationcheck we asked participants to characterize the supervisor who hade-mailed them at the beginning of the simulation by responding tothe 10 items on Brown, Treviño, and Harrison’s (2005) ethicalleadership scale ( .97).Results and DiscussionWe examined the distinctiveness of our measured variables usingthe same approach as Study 1 (CFA, maximum likelihood, Lisrel 8.8).The measurement model consisted of two factors, moral identity andmoral disengagement, because in this study ethical leadership wasmanipulated and our dependent variable was a count. The measurement model provided good fit to the data: 2(8) 19.06 p .01,(RMSEA .080, CFI .97, SRMR .069). We compared thismodel with a single factor model, 2(9) 96.15 p .01, (RMSEA .220, CFI .78, SRMR .140), which provided a significantlyworse fit to the data, 2(1) 77.09 p .01, supporting thedistinctiveness of our measures.Participants in the high-ethical leadership condition characterized their supervisor as significantly more ethical (M 4.10,SD .58) than participants in the low-ethical leadership condition(M 2.19, SD .77), t(177) 18.78, p .001, suggesting thatour manipulation of ethical leadership was effective.1Consistent with Hypothesis 1, individuals in the high-ethical leadership condition (M 3.25, SD .96) reported significantly lowerlevels of moral disengagement than participants in the low-ethicalleadership condition (M 3.54, SD .92, t(198) 2.13, p .034).In addition, individuals in the high-ethical leadership condition (M 2.60, SD 1.21) were marginally less likely than participants in thelow-ethical leadership condition (M 2.89, SD 1.13, t(198) 1.78, p .077) to deny patients’ insurance claims.Hypothesis 4 predicted that the mediated relationship of ethicalleadership on insurance claims denials through moral disengagement would be moderated by the subordinate’s moral identity,though we had competing predictions about whether this mediatedrelationship would be strongest for individuals with a weak moralidentity (Hypothesis 4a) or a strong moral identity (Hypothesis4b). We tested these competing hypotheses using Model 8 of thePROCESS macro (Hayes, 2013).2 Specifically, we tested whetherthe direct effect of ethical leadership on employee unethical decisions is moderated by moral identity, as well as whether theindirect effect of ethical leadership on employee unethical decisions [via moral disengagement] is moderated by moral identity.We present the results in Table 3. This model also provided a fewother results worth noting. Consistent with prior research (Detert etal., 2008; Moore et al., 2012), moral identity centrality protects onefrom the temptation to morally disengage (b .15, SE .06,p .014). Consistent with Hypothesis 1, ethical leadership alsohas a main, negative effect on moral disengagement (b .29,SE .13, p .025). Consistent with Hypothesis 2, there is a(marginally) positive effect of moral disengagement on insuranceclaims denials (b .17, SE .09, p .053).With regard to Hypothesis 4, in the mediator (M) model, ethicalleadership affects moral disengagement only for those at low oraverage levels of moral identity; individuals with a strong moralidentity are unaffected by the ethicality of their supervisor. Thisresult (see Figure 2, Panel A) is consistent with Hypothesis 4a, the“saving grace” hypothesis (see Figure 1, Panel B) rather than withHypothesis 4b (see Figure 1, Panel C). The dependent variable (Y)model confirms that the conditional indirect effect of ethical leadership on unethical decisions (via moral disengagement) holdswhen moral identification is low or average but not when it is high.Although this finding is supportive of Hypothesis 4, the 95%confidence interval around index of moderated mediation (Hayes,2013) just straddles zero ( .003 to .138). Thus, although this studydoes offer evidence of conditional indirect effects that are significantly different from zero, we cannot offer conclusive evidence ofmoderated mediation in this case. We can merely infer that twospecific paths in the mediation model are moderated (the indirect1 There were fewer degrees of freedom in this analysis compared withthe primary analysis as there was some respondent attrition after completing the decision-making part of the task (our outcome variable).2 We use Model 8 to test the moderated mediation hypotheses in each ofStudies 2 through 4, to provide a clean estimate of moral identity’s role inmoderating the mediating relationship over and above any direct moderatingeffect it may have in the relationship between ethical leadership and deviance,because this is what the literature would predict (Mitchell, 2008; O’Fallon &Butterfield, 2011). In none of the models are the conditional direct effectssignificant, and thus, for simplicity, they are not reported in the tables, but areavailable upon request from the first author.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 133effects for those low or average in terms of their moral identity).This provides some support for Hypothesis 4 and motivates furtherexamination.Study 3We conducted Study 2 to provide causal evidence of a linkbetween ethical leadership and employee moral disengagementand to run a first test of whether the mediated relationship betweenethical leadership and unethical employee outcomes via moraldisengagement would be moderated by moral identity. Though ourexperimental design was useful in allowing us to make causalclaims about the relationship between ethical leadership and moraldisengagement, our arguments about our complete theoreticalmodel would be more robust if we could replicate the moderatedmediation in the field. In Study 3, we return to the field andattempt to replicate the effects from Study 2 in an organizationalcontext. We also undertook the study in a non-U.S. setting, aidinggeneralizability implications. Study 3 also has the advantage ofcollecting data at three points in time, reducing concerns associated with cross-sectional data, and providing additional support forthe directional effects we propose.MethodParticipants and ProcedureRespondents were 513 subordinates and their corresponding 513supervisors (one supervisor rated one subordinate), identified froma workforce of around eight thousand employees from a largemanufacturing company group located in a northern city of China.Questionnaires were administered to the random sample of thesubordinates and supervisors with help from the HR department.Respondents were informed that the purpose of the survey was toexamine human resource practices. Confidentiality was assured.Respondents placed completed surveys in sealed envelopes andreturned them to researchers.Three waves of data collection were conducted over a 10-monthperiod. This procedure was undertaken to help minimize commonmethod bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).During Wave 1 (T1), questionnaires were administered to 513subordinates. Respondents were asked to provide demographicinformation (e.g., age, gender, position, and tenure) and theirperceptions about moral identity and ethical leadership. A total of375 usable responses were obtained, representing a response rateof 73.1%. Five months later, we conducted Wave 2 (T2) toevaluate moral disengagement. Because eight subordinates left thecompany between T1 and T2, the potential sample size was reduced to 367. A total of 286 usable surveys were collected,representing a response rate of 77.9%. Five months later, wedistributed Wave 3 (T3) questionnaires were distributed to supervisors (who supervised the 286 subordinates), asking them toprovide ratings of their subordinates’ organizational deviance.3The preceding procedure yielded a total sample of 208 matchedmanager–subordinate dyads with complete data across all threewaves, representing an ultimate response rate of 40.5%. For the208 subordinates, 59.6% were male, average age was 32.12 years(SD 7.52), and average organizational tenure was 6.40 years(SD 4.51). In terms of position, 49.5% were lower level employees, 39.9% were lower level managers, and 11.6% were middle level managers. China Europe International Business School(CEIBS) Ethics Committee approved the use of these data(CEIBS-OBHRM-01282018-1A) under the application “Understanding Ethical Leadership, Moral Values, and Workplace Deviance.”3 Unfortunately, we do not know which leaders left the organization overthe data collection period, if any. That said, annual bonus structures withconsiderably high pay outs disincentivize off-cycle turnover at the organization, and thus very few leaders are likely to have changed over the datacollection period. Nevertheless, we would expect leader turnover to attenuate our effects, and as such our results may offer overly conservativefindings.Table 3Study 2: Regression Coefficients and Conditional Indirect Effect EstimatesIndependent variablesMoral disengagement (M) Unethical decisions (Y)Path Coefficient SE LLCI ULCI p Path Coefficient SE LLCI ULCI pConstant 3.39 .07 3.264 3.523 .001 2.14 .31 1.527 2.768 .001Ethical leadership (X) a .29 .13 .554 .036 .025 c= .24 .17 .565 .092 .157Moral identity (W) .15 .06 .269 .030 .014 .14 .07 .012 .291 .071X W .21 .12 .027 .449 .082 .17 .15 .467 .135 .279Moral disengagement (M) b .17 .09 .003 .350 .053R2 .07 R2 .05Moderator Conditional effect of X on M Conditional indirect effect of X on Y via MMoral identity1 SD .52 .19 .891 .158 .005 .09 .06 .257 .005M .29 .13 .554 .036 .026 .05 .03 .146 .0031 SD .07 .18 .433 .286 .686 .01 .03 .101 .042Note. N 200; LLCI Lower limit confident interval; ULCI Upper limit confident interval. Coefficients are centered in all models. EL is a dummyvariable that indicates the experimental condition, where 1 high-ethical leadership and 0 low-ethical leadership. For the conditional indirect effects,we report the bias-corrected and accelerated 95% confidence intervals (CIs) calculated using 5,000 bootstrap samples. Conditional effects that arestatistically significant at the p .05 level are in bold.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.134 MOORE ET AL.MeasuresUnless otherwise indicated, all variables were measured using afive-point Likert-type scale, ranging from 1 (strongly disagree) to5 (strongly agree). Because all scales were originally written inEnglish, translation and back-translation was performed to ensureequivalence of meaning (Brislin, 1980). Scales were first translated from English into Chinese by a management professor andthen back-translated into English by another professor. Finally,one bilingual management scholar verified and crosschecked theEnglish and Chinese versions of the survey instrument and mademodifications to resolve any minor discrepancies.Ethical leadership. Participants completed the 10-item ethical leadership scale developed by Brown et al. (2005), which wasused in Study 1 ( .93).Moral identity centrality. We measured moral identity centrality using the same five-item measure (Aquino & Reed, 2002)that was used in Study 2 ( .87).Moral disengagement. In this study, moral disengagementwas measured with four items from Bandura, Barbaranelli, Caprara, and Pastorelli’s (1996) measure of moral disengagement thattap moral justification.4 The items were adapted slightly to refer to“colleagues” or “work” where the original items referred to“friends” or “family.” A sample item was “It is alright to fightwhen your work group’s honor is threatened” ( .82).Employee organizational deviance. To measure workplacedeviance, we used Stewart, Bing, Davison, Woehr, and McIntyre’s(2009) minor adaptations of Bennett and Robinson’s (2000) employee deviance scale used in Study 1. The items this scale uses tomeasure property and production deviance are the same as thoseBennett and Robinson used to measure organizational deviance.To make our outcome measure as similar as possible to that ofStudy 1, we aggregated the eight items from the Stewart measurethat are also included in Bennett and Robinson’s organizationaldeviance measure to evaluate organizational deviance ( .91).Control variables. As in Study 1, we collected data on sex,age, tenure, and level, which are four demographics variables thatprevious research has shown are associated with workplace deviance.Results and DiscussionWe examined the distinctiveness of the measured variablesusing the same approach as the prior two studies (CFA, maximumlikelihood, Lisrel 8.8). The measurement model consisted of fourfactors: ethical leadership, moral disengagement, moral identity,and organizational deviance. The measurement model providedgood fit to the data, 2(48) 44.36 p .01, (RMSEA .004,CFI 1.0, SRMR .025). We compared this model to a threefactor model where ethical leadership and moral disengagementwere combined into the same factor: 2(51) 348.18 p .01,(RMSEA .168, CFI .87, SRMR .160), which provided asignificantly worse fit to the data, 2(3) 303.82 p .01, as dida model that combined independent variables into one factor anddependent variables onto a second factor, 2(53) 871.77 p .01, (RMSEA .273, CFI .58, SRMR .230), 2(5) 827.41 p .01, and as did a single factor model, 2(54) 1047.17p .01, (RMSEA .298, CFI .45, SRMR .22), 2(6) 1002.81 p .01. Because the hypothesized measurement modeldemonstrated fit superior to comparison models, we were confident that our measures were adequate and distinct.Table 4 presents the means, standard deviations, and correlations for all key variables, and Table 5 reports the results of thePROCESS models used to test our hypotheses. Consistent with4 The eight-item measure of moral disengagement used in Studies 1 and4 (Moore et al., 2012) had not been published when these data werecollected. Constraints on survey length limited the number of Bandura’s(1986) original 32 items that could be included.Figure 2. How the effect of ethical leadership on moral disengagement ismoderated by moral identity centrality, Studies 2 through 4. Panel A: Study2, experiment (American sample, employed adults, simulated health insurance context). Panel B: Study 3, multisource, time-lagged survey (Chinesesample, manufacturing plant employees). Panel C: Study 4, multisourcesurvey (American sample, employed adults across several industries).This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 135Study 1, we present models without the control variables, thoughresults remain the same whether or not we include them (data areavailable on request). As predicted in Hypotheses 1 and 2, ethicalleadership at T1 has a main negative effect on employees’ moraldisengagement at T2 (b .31, SE .07, p .001), and moraldisengagement at T2 has a positive main effect on supervisorreported organizational deviance at T3 (b .28, SE .05, p .001).Consistent with Study 2, results show that the effect ethicalleadership on organizational deviance via moral disengagement ismoderated by employees’ moral identity, and in this case the 95%confidence interval for the index of moderated mediation excludeszero ( .178 to .042), so we can say definitely that the mediatingrelationship of ethical leadership on deviance through moral disengagement depends on employees’ moral identity. However, thepattern of this moderation differs from Study 2. As Table 5 shows,in Study 3, ethical leadership affects moral disengagement only forthose at average or high levels of moral identity; individuals witha weak moral identity are unaffected by their supervisor’s ethicality. This result (see Figure 2, Panel B) is consistent with Hypothesis 4b, the “virtuous synergy hypothesis” (see Figure 1, Panel C),rather than with Hypothesis 4a (see Figure 1, Panel B). Table 5confirms that the conditional indirect effect of ethical leadershipon unethical decision making (via moral disengagement) holdswhen moral identification is average or high, but not low. In thismodel, the 95% confidence interval surrounding the index ofmoderated mediation excludes zero ( .184 to .049), providingmore support for Hypothesis 4b.Study 4The results of Studies 2 and 3 suggest that moral identity is animportant moderator of the relationship between ethical leadershipand morally problematic behaviors via moral disengagement.However, the studies found that ethical leaders influenced moraldisengagement for different groups of subordinates. In Study 2,ethical leaders provided a “saving grace,” which dampened moraldisengagement tendencies for employees low on moral identity buthad no influence over those with strong moral identities; whereasin Study 3, ethical leaders catalyzed a “virtuous synergy,” whichdampened moral disengagement tendencies for employees withTable 4Study 3: Means, Standard Deviations, and Correlations for Relevant VariablesVariable M SD 1 2 3 4 5 6 7 81. Sex (male 1, female 0) .59 .49 —2. Age 32.12 7.52 .27 —3. Organizational tenure 6.40 4.51 .22 .65 —4. Organizational level 1.61 .67 .03 .25 .18 —5. Ethical leadership 3.64 .74 .16 .15 .10 .01 (.93)6. Moral identity 3.93 .79 .14 .08 .12† .10 .06 (.87)7. Moral disengagement 2.72 .88 .08 .06 .14† .00 .30 .27 (.82)8. Organizational deviance 1.88 .68 .03 .13† .23 .03 .20 .22 .42 (.91)Note. N 208. Rows 1 through 6 are reported by the focal employee at T1, moral disengagement was reported by the focal employee at T2, andorganizational deviance was reported by the focal employee’s supervisor at T3. Organizational level: 1 frontline employees, 2 frontline managers, 3 middle managers. Alpha reliabilities are on the diagonal in parentheses.†p .10. p .05. p .01.Table 5Study 3: Regression Coefficients and Conditional Indirect Effect EstimatesIndependent variablesMoral disengagement (M) Organizational deviance (Y)Path Coefficient SE LLCI ULCI p Path Coefficient SE LLCI ULCI pConstant 2.73 .05 2.629 2.85 .001 1.11 .31 .798 1.41 .001Ethical leadership (X) a .31 .07 .46 .163 .001 c= .08 .06 .195 .044 .214Moral identity (W) .25 .07 .39 .113 .001 .11 .06 .217 .006 .063X W .36 .10 .56 .167 .001 .00 .08 .467 .135 .976Moral disengagement (M) b .28 .05 .174 .391 .001R2 .21 R2 .20Moderator Conditional effect of X on M Conditional indirect effect of X on Y via MMoral identity1 SD .02 .11 .248 .201 .835 .07 .03 .066 .056M .31 .07 .456 .163 .000 .08 .03 .151 .0441 SD .60 .10 .797 .395 .000 .17 .04 .269 .095Note. N 208; LLCI Lower limit confidence interval; ULCI Upper limit confidence interval. Coefficients are centered in all models. For theconditional indirect effects, we report the bias– corrected and accelerated 95% confidence intervals (CIs) calculated using 5,000 bootstrap samples.Conditional effects that are statistically significant at the p .05 level are in bold.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.136 MOORE ET AL.strong moral identities but had no influence over employees withweak moral identities.There are several differences between Studies 2 and 3 that couldexplain these different patterns. First, the studies used differentempirical paradigms: Study 2 used an experimental design,whereas Study 3 was a multiwave and multisource field survey.Second, one of the implications of these different paradigms is thatthe studies operationalized the key independent variable differently: Ethical leadership was manipulated in Study 2 and measuredin Study 3. Though designed to reflect an absence of the behaviorsdescribed in definitions of ethical leadership, the low-ethical leadership condition in Study 2 may have reflected unethical leadership instead of low levels of ethical leadership. Third, the studiesoperationalized the key dependent variable differently: Study 2used the number of times the participant denied patients insurancecoverage for technically allowable but morally problematic reasons (a measure of unethical decision making), whereas Study 3used supervisor-reported employee deviance. Fourth, the studiesused different organizational contexts: a simulation of a health careinsurance company in Study 2 and a Chinese manufacturing firmin Study 3. Finally, the studies were conducted in different culturalcontexts: Study 2 uses a U.S.-based sample, whereas Study 3 usesa Chinese sample.We designed a fourth study—a second field study in theUnited States—in an effort to rule out some of these alternativeexplanations while providing a better understanding of why thepattern of the moderation differed across the two studies. Wedesigned the field study to mimic several aspects of Study 3,measuring ethical leadership and employee deviance in thesame way, but by using a U.S.-sample of participants from abroad array of organizations. If this study finds a “saving grace”pattern to the moderation (similar to Study 2), we could rule out(1) the experimental paradigm, different operationalizations ofthe (2) independent and (3) outcome variables, and (4) the useof the insurance industry as the organizational context as reasons for the different patterns of the moderation observed inStudies 2 and 3. Indeed, finding this pattern of results wouldsuggest that the different pattern might be a function of thecultural context (China vs. United States) or that it might beattributable to something idiosyncratic about the specific organization that provided the setting for Study 3. If, however, thisstudy finds a pattern of the moderation reflecting a “virtuoussynergy” (similar to Study 3), we could rule out the culturalcontext (China vs. the United States) or something unique to theorganization sampled in Study 3 as the reasons for the differentpattern. This pattern of results would suggest instead that thepattern of the moderation is more likely a function of theexperimental paradigm or measures used in Study 2.MethodParticipants and ProcedureThe procedure of this study followed the snowball samplingprocedure used in Study 1. A total of 322 students volunteered totake part in a study using a snowball sampling procedure; theyreceived extra credit for their involvement. Each student was givena packet that included an instruction sheet for a focal employee andfor a coworker of that employee. The student then returned to theresearchers the personal e-mail address of these two employees,who received an individual link via their personal e-mail address tothe appropriate version of the survey. We had data available for allrelevant measures for 272 focal employees (an 84% response rate)and were able to match 175 employee-co-worker dyads on all therelevant measures.Of the 272 focal employees who completed the survey, 41%were male, and 89% were full-time employees (11% worked parttime). Their average age was 43.2 years (SD 13.2). Theyrepresented several industries, including health care and socialassistance (15%), manufacturing (14%), education (13%), professional services (12%), finance (10%), real estate (6%), and retailtrade (3%). Eleven percent had an organizational tenure of lessthan 1 year, 36% had an organizational tenure between 1 and 5years, 17% had an organizational tenure between 6 and 10 years,and 36% had an organizational tenure of 11 or more years. Twothirds (67%) were in management positions (33% were in nonmanagerial roles). Eighty-one percent of the respondents wereWhite, 2% Hispanic or Latino/a, 2% African American, 13%Asian American, and 2% indicated “other.”Of the 175 coworkers we were able to match back to employeeresponses, 41% were male, and 91% were full-time employees(9% worked part-time). Their average age was 42.0 years (SD 12.3). Eleven percent had an organizational tenure of less than 1year, 40% had an organizational tenure between 1 and 5 yearsorganizational tenure, 18% had an organizational tenure between 6and 10 years, and 31% had an organizational tenure of 11 or moreyears. Seventy-five percent of the respondents were White, 7%Hispanic or Latino/a, 3% African American, 12% Asian American,and 2% indicated “other.”The University of Michigan’s Institutional Review Board approved this research protocol (HUM00107807), under the application “Field Survey–MO 300/302.” These data are part of abroader data collection effort; this study is the first publicationfrom that effort.MeasuresAll response scales ranged from 1 (strongly disagree) to 7(strongly agree) unless otherwise noted.Ethical leadership. Focal employees reported their perceptions of their supervisor’s ethical leadership using same the 10-item ethical leadership scale (Brown et al., 2005) as all otherstudies ( .94).Moral disengagement. Focal employees reported moral disengagement using same the eight-item measure (Moore et al.,2012) as used in Study 1 ( .89).Moral identity centrality. Focal employees reported theirmoral identity centrality using the same five-item measure(Aquino & Reed, 2002) as used Studies 2 and 3 ( .88).Organizational deviance. Coworkers reported the focal employee’s organizational deviance using the 12-item measure (Bennett & Robinson, 2000) as used in Study 1, ( .94).Employee unethical behavior. Coworkers also reported onthe focal employee’s unethical behavior, using the 17-item measure (Akaah, 1996) as used in Study 1 ( .92).Control variables. As in Studies 1 and 3, we collected information about employee’s sex, age, tenure, and level to use ascontrol variables.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.LEADERS MATTER MORALLY 137Results and DiscussionAs with the prior studies, we examined the distinctiveness of themeasured variables via CFA (maximum likelihood, Lisrel 8.8, balanced parcels). The measurement model consisted of five factors:ethical leadership, moral disengagement, moral identity, organizational deviance, and unethical behavior, 2(80) 122.58 p .01,(RMSEA .055, CFI .98, SRMR .050). We compared thismodel to a four-factor model where deviance and unethical behaviorswere combined: 2(84) 201.18 p .01, RMSEA .090, CFI .96, SRMR .053, which provided a significantly worse fit to thedata, 2(4) 78.60 p .01, as did a two-factor model thatcombined IVs into one factor and DVs onto a second factor, 2(89) 966.69 p .01, (RMSEA .238, CFI .72, SRMR .180),2(9) 844.11 p .01, as did a single factor model, 2(90) 2179.00 p .01, (RMSEA .365, CFI .40, SRMR .30),2(10) 2056.42 p .01. Because the hypothesized measurementmodel demonstrated a superior fit to the data than all comparisonmodels, we were confident that our measures were adequate anddistinct from one another.Table 6 presents the means, standard deviations, and correlations among the study variables; Table 7 presents the results of theregression models designed to test our predicted moderated mediation. Consistent with Studies 1 and 3, we present models withoutthe control variables, though results remain the same whether ornot we include them (data available on request). Consistent with allthe other studies, ethical leadership was negatively related toemployees’ moral disengagement (b .10, SE .06, p .088).Moral disengagement was in turn positively related to coworkerreported organizational deviance (b .11, SE .05, p .033) aswell as coworker reported unethical behavior (b .15, SE .06,p .009).In this study, the pattern of the moderation matched Study 2rather than Study 3: Ethical leadership is related to moral disengagement only for those at low levels of moral identity rather thanfor individuals with a strong moral identity. In addition, as inStudy 3, the 95% confidence interval for the index of moderatedmediation excludes zero for both outcome variables (.004 to .101for organizational deviance and .003 to .110 for unethical behavior), providing definitive evidence that the mediating relationshipof ethical leadership on deviance and unethical behavior throughmoral disengagement depends on employees’ moral identity. Thisresult (see Figure 2, Panel C) is consistent with Hypothesis 4a (seeFigure 1, Panel B) and Study 2—the “saving grace” hypothesis—rather with Hypothesis 4b (see Figure 1, Panel C) and Study3—the “virtuous synergy” hypothesis. Table 7 confirms that theconditional indirect effect of ethical leadership on coworker reported organizational deviance (via moral disengagement) holdswhen moral identification is low, but not high. The same patternwas observed for coworker reported unethical behavior (via moraldisengagement): the relationship was present when moral identification is low but not when it was high.This pattern of results helps rule out several explanations for thedifferent pattern of the moderation observed in Studies 2 and 3.First, replicating the same pattern of results across experimentaland field studies reduces concerns that features of the experimentcreated an artificial effect. Specific features of concern were (1)the empirical paradigm used (a scenario with a fictitious boss andno long-term job implications for performing poorly on the task orfor actually harming others with one’s decisions), (2) the independent variable (the experiment potentially tapped a different rangeof ethical leadership than the other studies, with the manipulationcreating a condition that represented unethical leadership ratherthan low levels of ethical leadership), (3) the dependent variable(operationalizing as a count of specific decisions compared tomore general deviant behaviors), and (4) the organizational contextused in Study 2 (an insurance company and role of claims adjustervs. a variety of industries and job roles). Instead, the pattern of theresults did not replicate the Chinese field study. This implies thatsomething to do with the cultural context or a unique feature of theorganization likely explains why we found a different pattern ofthe moderation across Studies 2 and 3. We explore this in greaterdetail in the General Discussion.General DiscussionThree primary findings emerge from these four studies. First, weconsistently find a negative relationship between ethical leadershipand employee moral disengagement. This supports our primaryhypothesis: leader behavior is associated with how employeesconstrue decisions with ethical import. Our manipulation of ethicalleadership and its resulting effects provide confidence that ethicalTable 6Study 4: Means, Standard Deviations, and Correlations for Relevant VariablesVariable M SD 1 2 3 4 5 6 7 8 91. Sex (male 1, female 0) .36 .482. Age 42.70 12.72 .043. Organizational tenure 3.15 1.72 .05 .484. Organizational level 4.63 1.28 .38 .26 .115. Ethical leadership 5.87 .82 .08 .03 .00 .06 (.94)6. Moral identity 6.41 .69 .12 .01 .03 .06 .24 (.88)7. Moral disengagement 1.61 .75 .17 .11 .05 .05 .25 .52 (.89)8. Coworker-reported organizational deviance 1.20 .42 .08 .04 .01 .16 .08 .00 .14† (.94)9. Coworker-reported unethical behavior 1.34 .46 .15† .07 .00 .10 .02 .02 .17 .80 (.92)Note. N 165–175 (10 individuals did not complete certain demographic characteristics). Tenure (1 less than 1 year, 2 1–5 years, 3 6–10 years,4 11–15 years, 5 16–20 years, 6 21–25 years, 7 25 years) and level (1 CEO/owner, 2 top management team, 3 vice–president, 4 middle management, 5 supervising employee, 6 low ranking employee) were measured categorically. Alpha reliabilities are on the diagonal inparentheses.†p .10. p .05. p .01.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.138 MOORE ET AL.leadership has a direct causal influence over employee moraldisengagement. In addition, this finding was consistent in bothAmerican and Chinese work contexts, suggesting the effect is notculturally bound. Second, we also found evidence across all fourstudies that moral disengagement functions as a mechanism toexplain the relationship between ethical leadership and employeeunethical decisions and behaviors. Again, this result was consistentacross time- and respondent-separated field studies and an experiment, in American and Chinese organizations, and using differentmeasures of our primary constructs, providing important assuranceof the generalizability of our findings and bolstering our confidence that moral disengagement as an important, unique, androbust mechanism to explain ethical leaders’ positive effectswithin their organizations.Finally, we found persistent evidence that the centrality of anemployee’s moral identity plays a key role in the relationshipbetween ethical leadership and employee unethical decisions andbehavior (through moral disengagement). However, the nature ofthis moderated relationship varied across studies. In the experiment (Study 2) and the U.S. field study (Study 4), the form of theinteraction supported the idea that ethical leadership has a “savinggrace” effect, wherein ethical leaders dampen the propensity ofindividuals with a weak moral identity to morally disengage. In theChinese manufacturing organization (Study 3), this interactiontook the form of a “virtuous synergy,” wherein ethical leaders havethe strongest effect on individuals with a strong moral identity,dampening their propensity to morally disengage specifically.These studies make three important contributions, despite somewhat conflicting results. First, they lend empirical support to theidea that ethical leadership is related to how employees construedecisions with moral import. Second, through this cognitive mechanism, ethical leaders can limit their employees’ unethical decisions and deviant behavior. Third, an employee’s moral identityinfluences the mediated relationship between ethical leadership onemployee deviance and unethical behavior through moral disengagement. However, the nature of this relationship is complex,with ethical leaders sometimes providing a saving grace for employees low in moral identity or sometimes a virtuous synergy foremployees high in moral identity. The findings suggest that thepattern that occurs depends less on methodology (lab vs. field, orthe operationalization of independent and dependent measures)and more on context (whether cultural or organizational). Belowwe elaborate on the opportunities for theorizing these divergentresults provide.Theoretical ImplicationsImplications for ethical leadership theory. Our primarycontribution rests in testing a novel cognitive mechanism to explain the relationship between ethical leadership and employeebehavior. Although a growing body of work enumerates the consequences of ethical leaders, researchers have historically devotedless effort toward identifying the various pathways through whichthese effects may emerge (Brown & Treviño, 2006). Moreover,many of the mechanisms that have been tested to date are actuallycaptured by the independent variable itself (Antonakis, 2017),raising concerns about circular theorizing. Showing that ethicalleaders can affect how their subordinate construe decisions withmoral import represents an important extension of ethical leadership theory, and it offers a more distal mechanism from thosepreviously described in the literature.In addition, as work on ethical leadership rapidly advances, it iscritical to examine boundary conditions of its effects on employeecognition and behavior. Our findings highlight moral identity as acritical individual difference moderator of ethical leadership’seffects, but the different pattern of results we found across thestudies suggests that the way this individual difference interactswith ethical leadership likely differs across contexts, highlightingthe need for continued research on boundary conditions of therelationship between ethical leadership and employee cognitionand conduct.We also note, as have others (Brown & Mitchell, 2010), that therelationship between ethical and unethical leadership requires further theoretical and empirical clarification. We primarily exploredthe role of ethical (rather than unethical) leadership in mitigatingemployee unethical behavior and deviance via moral disengageTable 7Study 4: Regression Coefficients and Conditional Indirect Effect EstimatesIndependent variablesMoraldisengagement (M)Coworker reported organizationaldeviance (Y)Coworker reported unethicalbehavior (Y)Path Coefficient SE LLCI ULCI p Path Coefficient SE LLCI ULCI p Coefficient SE LLCI ULCI pConstant 1.57 .04 1.481 1.666 .001 1.02 .09 .844 1.193 .001 1.06 .10 .913 1.299 .001Ethical leadership (X) a .10 .06 .216 .015 .088 c= .03 .04 .114 .046 .399 .01 .04 .081 .095 .872Moral identity (W) .45 .07 .590 .307 .001 .06 .05 .043 .172 .235 .05 .06 .067 .170 .393X W .29 .07 .150 .427 .001 .03 .05 .131 .068 .534 .06 .06 .167 .053 .310Moral disengagement (M) b .11 .05 .009 .216 .033 .15 .06 .038 .266 .009R2 .35 R2 .03 R2 .04Moderator Conditional effect of X on M Conditional indirect effect of X on Y via MMoral identity1 SD .30 .07 .446 .155


Leave a Reply

Your email address will not be published. Required fields are marked *