A SOCIAL CONTEXT MODEL OF ENVY AND SOCIAL UNDERMINING

Page created by Floyd Campbell
 
CONTINUE READING
娀 Academy of Management Journal
2012, Vol. 55, No. 3, 643–666.
http://dx.doi.org/10.5465/amj.2009.0804

                         A SOCIAL CONTEXT MODEL OF ENVY AND SOCIAL
                                       UNDERMINING
                                                                       MICHELLE K. DUFFY
                                                                      University of Minnesota

                                                                          KRISTIN L. SCOTT
                                                                          Clemson University

                                                                         JASON D. SHAW
                                                                      University of Minnesota

                                                                       BENNETT J. TEPPER
                                                                      Georgia State University

                                                                        KARL AQUINO
                                                                 University of British Columbia

                          We integrate moral disengagement, social identification, and social norms theories to
                          develop, test, and replicate a model that explains how and when envy is associated
                          with social undermining. In Study 1, a two-wave study of hospital employees, results
                          support the prediction that the mediated effect of envy on social undermining behavior
                          through moral disengagement is stronger when employees have low social identifica-
                          tion with coworkers. Study 2, a four-wave, multilevel study of student teams, shows
                          that the indirect effect of envy on social undermining through moral disengagement is
                          stronger in teams with low team identification and high team undermining norms.

   Employees have numerous opportunities to ob-                                                   (Vidaillet, 2006) and is an unpleasant, painful state
serve and contemplate how the benefits and advan-                                                 (Lieberman & Eisenberger, 2009). Consequently, in-
tages they enjoy at work compare with those of                                                    dividuals should be highly motivated to reduce
their colleagues. Whether these opportunities pres-                                               feelings of envy. The psychological literature
ent themselves formally (e.g., through performance                                                shows that when they are unsuccessful at doing so,
appraisals) or informally (e.g., through friendship                                               and feelings of envy persist, envy can lead to a
networks), common experience and empirical re-                                                    variety of deleterious outcomes, including
search suggest that favorable social comparison in-                                               schadenfreude, aggression, and even crime (see
formation gives people pleasure, but unfavorable                                                  Smith and Kim [2007] for a review). Indeed, envy is
information can focus their attention on what they                                                often argued to be a “call to action” to engage in
lack relative to their colleagues (Hogg, 2000). In the                                            interpersonal harm doing, especially actions that
latter case, feelings of envy—the emotion that sur-                                               “reduce or, better yet, fully remove the envied per-
faces when one lacks and desires others’ superior                                                 son’s advantage” (Smith & Kim, 2007: 53). If this
qualities, achievements, or possessions (Parrott &                                                argument is correct, then envy should be a reliable
Smith, 1993)—may arise.                                                                           predictor of social undermining, or behavior in-
   Envy of others at work (referred to here as                                                    tended to hinder the ability of others to establish
“envy”) may have positive consequences if, for ex-                                                and maintain positive interpersonal relationships,
ample, it motivates a person to increase perfor-                                                  work-related successes, and favorable reputations
mance or attempt self-improvement (Duffy, Shaw,                                                   (Duffy, Ganster, & Pagon, 2002). Social undermin-
& Schaubroeck, 2008; Schaubroeck & Lam, 2004).                                                    ing behavior differs from other forms of antisocial
But envy also has a dark side. Envy poses threatens                                               behavior at the conceptual level because it com-
the core of an individual’s professional identity                                                 prises only intentional behavior and behavior de-
                                                                                                  signed to weaken its target gradually or by degrees
                                                                                                  (Duffy, Ganster, Shaw, Johnson, & Pagon, 2006).
  Editor’s note: The manuscript for this article was ac-                                            The study of envy and its work-related conse-
cepted for publication during the term of AMJ’s former                                            quences has been surprisingly neglected, even
editor-in-chief, R. Duane Ireland.                                                                though work environments include a surfeit of po-
                                                                                         643
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s express
written permission. Users may print, download, or email articles for individual use only.
644                                        Academy of Management Journal                                       June

tential envy-inducing situations (Duffy et al., 2008).       strongly motivated to do so. For example, to thrive
Indeed, research linking envy to harmful behaviors           in work contexts, individuals must develop social
at work, such as undermining, is rather sparse, and          capital, make high-quality connections with capa-
the relationship has not been firmly established             ble others, and maintain some positive standing in
(e.g., Cohen-Charash, 2009; Cohen-Charash & Muel-            the work environment (Dutton & Heaphy, 2003;
ler, 2007; Dunn & Schweitzer, 2006). We aim to               Exline & Zell, 2008). Responding to envy with un-
advance the theoretical and empirical literature in          dermining or other forms of aggression can place
two ways.                                                    these objectives at risk. Consequently, we maintain
   First, the literature lacks a guiding theoretical         that even when they have strong feelings of envy,
perspective on the process through which envy                organization members do not always respond with
relates to social undermining. A basic assumption            social undermining. The possibility that social and
of our model is that envy is an undesirable state            organizational forces are likely to inhibit causing
that most employees want to avoid. We contend                harm to others, even when the desire to do so is
that one way employees can attempt to manage                 strong, may explain why some studies have shown
envy is to thwart the success of their colleagues            inconsistent relationships between measures of
through social undermining, thereby reducing the             workplace envy and antisocial behavior (Cohen-
gap between the outcomes that might lead to dis-             Charash, 2009). Indeed, it may be that a common
advantageous social comparisons. We do not claim             response to feeling envious of others is for an indi-
that this is the only path envious employees can             vidual to “suffer in silence,” resigned to accepting
pursue in response to envy, but it is the one we             that others have things that he or she desires but
focus on here, because it requires employees to              cannot have.
overcome personal and social barriers to harming                Thus, to address the question of when envy will
others in workplace settings.                                actually result in social undermining, we introduce
   What theoretical mechanism explains why em-               two features of social context: social identification
ployees may choose to act on their envious feelings          with one’s colleagues and undermining norms. In
with social undermining behavior? We offer one               the first of two studies, we tested whether social
theoretical explanation by proposing that the link           identification with colleagues can either facilitate
between envy and social undermining is mediated              or inhibit the emergence of moral disengagement
by a disengagement of the self-regulatory mecha-             cognitions, a first-stage moderation effect in our
nisms that would otherwise constrain such behav-             model. Prior research has focused largely on the
ior. These cognitive maneuvers are referred to               notion that social identification (e.g., closeness,
broadly as “moral disengagement” (Bandura, 1986,             similarity) and envy are positively related because
1991). We theorize that envy increases moral dis-            identification creates “the expectation that the
engagement and allows envious employees to over-             other ought to experience similar outcomes”
come the cognitive barriers, or self-sanctions, that         (Schaubroeck & Lam, 2004: 34; see also Smith and
most people abide by, breaking them “free of a               Kim [2007] for a review). Here, we propose a novel,
prevailing submissive frame of mind” (Smith &                moderating role for social identification. Drawing
Kim, 2007: 53). This part of our model allows us to          on theories of moral exclusion (Opotow, 1990,
answer the question of why envy might motivate               1995) and self-construal (Bandura, 1986), we argue
social undermining. Thus, we extend current                  that enviers will morally disengage only when they
thinking on envy by proposing moral disengage-               feel less psychologically connected to others in
ment as the mechanism through which envy leads               their work environment (i.e., they experience low
to harmful interpersonal behaviors. Our approach             social identification). In contrast, high social iden-
also extends previous theories of envy by suggest-           tification makes it more difficult to convert envy
ing that envy resulting from making multiple rather          into cognitive rationalizations for harm-doing.
than specific (or episodic) social comparisons can           Thus, although prior research has focused on the
influence the willingness to undermine coworkers             role of social identification in generating envy, we
who may or may not be the objects of envy. The               investigate how social identification shapes re-
reason is that envy motivates moral disengagement,           sponses to envy by defusing the tendency to mor-
which disables self-sanctions against harm doing.            ally disengage. In our second study, we incorporate
By advancing this argument, we are suggesting a              theory concerning behavioral or etic norms to sug-
potential spillover effect of envy that leads a person       gest that once moral disengagement has occurred,
to undermine others in general.                              an envious employee may nevertheless eschew so-
   Second, in organizational settings, often strong          cial undermining. Following previous researchers
social and institutional pressures inhibit employ-           (e.g., Bamberger & Biron, 2007; Tepper, Henle,
ees from undermining others, even if they are                Lambert, Giacalone, & Duffy, 2008), we hypothe-
2012                                             Duffy, Scott, Shaw, Tepper, and Aquino                                   645

size that weak norms for undermining in the social                       uations, work and otherwise) and episodic envy (an
context can serve as a second defusing mechanism                         emotional reaction to a specific event). In describ-
that weakens the relationship between moral dis-                         ing workplace envy, Vecchio (2005) argued that
engagement and social undermining behavior, a                            individuals in work situations recognize differ-
second-stage moderation effect in our model.                             ences in social standing, performance, and treat-
Taken together, our model suggests that two condi-                       ment with multiple comparators concurrently and
tions are needed for envy to lead to social under-                       that these comparisons may not be to a specific
mining: the employee who feels envy must not                             transient episode, but rather to an existing state of
strongly identify with his or her coworkers, and                         multiple unflattering comparisons. This view is
norms discouraging undermining must be weak.                             also in line with the views of Wood, who argued
   Our theoretical model is presented in Figure 1.                       that social comparison processes include “thinking
We test the first part of our moderated-mediation                        about one or more other people in relation to the
theory with data from employees of a university                          self” (1996: 520). When making comparisons about
hospital. We then develop an integrative moderat-                        career advancement, for example, Wood (1996)
ed-mediation model involving both social context                         stated that individuals evaluate their standing with
moderators and report tests of the full model using                      multiple coworkers before making negative or pos-
a sample of student teams. We conclude by discuss-                       itive judgments. In the case of workplace envy, an
ing the implications of the results for theory and                       individual may lack and desire certain colleagues’
research.                                                                superior achievements (e.g., their superior pattern
                                                                         of academic publishing), while concurrently envy-
                                                                         ing the level of recognition still others have re-
             THEORETICAL BACKGROUND
                                                                         ceived for their accomplishments (e.g., a colleague
Conceptualizing Envy of Others at Work                                   who holds a lucrative endowed chair). In such a
                                                                         case, envy is high because of the desired patterns of
   The experience of envy has been conceptualized
                                                                         successes of multiple others in the environment.
in three related ways: as situational—that is, as a
general envy of others in an environment, typically
a work context or team, involving multiple refer-
                                                                         Envy and Social Undermining
ents or comparators (e.g., Duffy & Shaw, 2000; ); as
dispositional (e.g., Smith, Parrott, Diener, Hoyle, &                       Envy is an unpleasant emotion (Smith & Kim,
Kim, 1999); and as specific and episodic, involving                      2007). The perception that one lacks and desires
a specific individual as a referent (e.g., Cohen-Cha-                    another’s superior qualities, achievements, or pos-
rash, 2009). In this study, we take the first view:                      sessions is a significant threat to self-esteem.
that individuals can and do make invidious social                        Envy is so uncomfortable, in fact, that it has been
comparisons with others in their immediate envi-                         linked to the activation of neural circuitry that is
ronment (Vecchio, 1995, 2005). Conceptually, envy                        responsible for physical pain (see Lieberman &
of others in a work context is distinct from dispo-                      Eisenberger, 2009). An individual can alleviate
sitional envy (a tendency generalized across all sit-                    the unpleasantness associated with envy by ag-

                                                          FIGURE 1
                                                  Proposed Theoretical Modela

  a
      Relationships designated with solid lines were tested in Study 1. The full model was tested in Study 2.
646                                       Academy of Management Journal                                        June

grandizing the self at the expense of targeted              dura, 1986; Bandura, Barbaranelli, Caprara, & Pas-
individuals (Wert & Salovey, 2004). As an instru-           torelli, 1996) and others (e.g., Detert et al., 2008),
mental form of aggression, social undermining               we conceptualize moral disengagement as a single
behaviors (e.g., belittling, gossiping, withholding         overarching construct. Taken together, these di-
information, giving someone “the silent treat-              mensions represent an overall cognitive orientation
ment”) are powerful ways to bolster oneself at              that is the additive sum of the conceptual dimen-
others’ expense (Salmivalli, 2001).                         sions; “the lack of any single dimension will de-
   Envious feelings are also unpleasant because             flate, though not completely eliminate, the overall
they can be accompanied by frustration and hostil-          degree” (Spreitzer, 1995).
ity (Smith & Kim, 2007). Individuals are often mo-             How does envy trigger the moral disengagement
tivated to reduce envy and to thereby reduce “the           process? The first broad mechanism (devaluing the
envious [employee]’s frustration with feeling infe-         target) includes dehumanizing and attributing
rior” to others (Cohen-Charash & Mueller, 2007: 5;          blame to victims. Although envious individuals de-
see also Crossley, 2009; Dunn & Schweitzer, 2006).          sire the qualities, achievements, or possessions of
We therefore expect envy to be related to under-            others, research has shown they often believe that
mining behaviors because undermining is one pos-            envied individuals are unworthy of or cannot be
sible way that individuals can reduce others’ per-          trusted with their advantages (e.g., Dunn &
ceived superiority and raise their own relative             Schweitzer, 2004; Salovey & Rodin, 1984; Silver &
standing while also venting their frustration and           Sabini, 1978; Smith, 2004). Moral disengagement
hostility (Cohen-Charash & Mueller, 2007; Dunn &            may allow envious individuals to justify the nega-
Schweitzer, 2006).                                          tive implications of harmful actions directed
                                                            against others. Smith and Kim (2007) argued that
Mediating Role of Moral Disengagement                       targets can be devalued over time as individuals
                                                            begin to ruminate obsessively on what they per-
  We turn now to elucidating the moral disengage-
                                                            ceive to be the undeserved advantages of envied
ment process by which envy leads to individual
                                                            others while ignoring the role that they themselves
social undermining behavior. Grounded in social
                                                            play in creating the disparity. As this thinking takes
cognitive theory, moral disengagement refers to a
                                                            hold, individuals “might be able to convince them-
set of cognitive justifications (referred to as mech-
                                                            selves that they have an increasingly legitimate
anisms) that allow an individual to commit acts
                                                            cause for feeling hostile although they may still be
such as social undermining while avoiding the self-
                                                            wary of publicizing their feelings. Seemingly legit-
sanctions (e.g., self-condemnation, self-loathing)
that ordinarily deter such behavior (Bandura, Bar-          imate grievances may “free” envious people to ex-
baranelli, Caprara, Pastorelli, & Regalia, 2001; De-        ecute indirect acts of hostility (e.g., negative gossip
tert, Treviño, & Sweitzer, 2008; McFerran, Aquino,          and backbiting)” (Smith & Kim, 2007: 56). If this
& Duffy, 2010). Critical to the foundation of social        reasoning holds, individuals are likely to under-
cognitive theory is that the same behavior might be         mine because they believe others deserve to be
viewed as reprehensible conduct toward colleagues           undermined (Harris, Cikara, & Fiske, 2008).
in one situation and as acceptable or necessary in             The second broad category operates via a cogni-
another context (Bandura, 1986). The activation of          tive reconstrual of the conduct itself, which in-
moral disengagement mechanisms eliminates self-             cludes moral justification, use of euphemistic lan-
deterrents to harmful behavior and can encourage            guage, and advantageous comparison (Bandura,
self-approval for antisocial conduct (Brief, Buttram,       1986). For example, through the process of moral
& Dukerich, 2001). Moral disengagement has been             justification, harmful behavior becomes acceptable
conceptualized under three broad mechanisms and             because it is viewed as valued or righteous. In this
eight specific examples of disengagement. The first         way, envious individuals may begin to believe that
broad mechanism— devaluing the target—includes              social undermining behavior is not only condoned
dehumanizing or attributing blame to victims. The           but appropriate. Envious individuals may use eu-
second—reconstruing the conduct—involves such               phemistic language to sanitize what normally
specific mechanisms as moral justification, use of          would be considered antisocial behavior (e.g.,
euphemistic language or labeling, and advanta-              “making things right” or “making things fair”) and
geous comparison. The third category— obscuring             may also disengage mechanisms against antisocial
or distorting consequences—includes displace-               behavior through advantageous comparison—for
ment of responsibility, diffusion of responsibility,        example, by rationalizing that their deviant behav-
and minimizing behavioral consequences. Like                ior is minimal compared with what others have
Bandura and colleagues (Bandura et al., 2001; Ban-          done to gain advantages.
2012                                    Duffy, Scott, Shaw, Tepper, and Aquino                                    647

   The last set of disengagement practices operates            growing body of theoretical and empirical work on
by obscuring or distorting the effects of harmful              moral exclusion has shown that people’s sense of
behavior (Detert et al., 2008). Under this broad cat-          moral obligation appears to be stronger when di-
egory, individuals may displace or diffuse respon-             rected toward those who “are closer to us and
sibility for their behavior or minimize its conse-             weaker toward those who are psychologically dis-
quences. Social undermining behaviors such as                  tant” (Opotow, 1995: 351; Reed & Aquino, 2003;
gossiping and backbiting may be subtle, and their              Tepper, Moss, & Duffy, 2011). When people so-
negative effects may not be immediately obvious.               cially identify with other individuals, they are
These conditions make it easier to distort the con-            more likely to see the others as entitled to a share of
sequences of such action (Bandura, 1999). Be-                  community resources and to other forms of aid
cause the negative effects of undermining are                  (Clayton & Opotow, 2003). In contrast, any percep-
cumulative and often grow more deleterious over                tion of separation, including the simple perception
time (Duffy et al., 2002), it is reasonable to expect          that one is disconnected from others, can generate
that envious individuals will use this form of                 social differentiation and moral exclusion (e.g.,
moral disengagement.                                           Hogg & Abrams, 1988; Opotow, 1990; Reed &
   In sum, we propose that envy activates the cog-             Aquino, 2003).
nitive mechanisms associated with moral disen-                    These arguments lead us to advance an interac-
gagement. Taken together, our arguments suggest a              tive prediction regarding how responses to envy are
mediating framework in which envy leads to sub-                shaped by a person’s degree of identification with
sequent social undermining behaviors via moral                 colleagues. The envy, moral disengagement, and
disengagement. We expect, however, that individ-               social undermining sequence is enmeshed in a
uals’ identity-based construals of envied targets              more complex social reality: social identification
will influence whether this mediating process will             and its corresponding sense of moral obligation
hold. Specifically, we propose that the mediating              may affect the likelihood that envy triggers the
effect of moral disengagement will be weaker when              moral disengagement process that leads to social
social identification with colleagues is higher and            undermining (a first-stage moderator in the media-
that the mediating effect will be stronger when                tion model). In the case of high social identifica-
social identification is lower. The foundation for             tion, the translation of envy into moral disengage-
our moderated-indirect effect hypothesis is de-                ment may be inhibited; in the case of low social
scribed below.                                                 identification, the negative effects of envy may be
                                                               exacerbated. Recall that moral disengagement al-
                                                               lows an envier to undermine others to cope with
Moderating Role of Social Identification
                                                               threats to self-esteem and the discomfort associated
   As noted above, social identification has a prom-           with envy. If the moral exclusion framework is
inent role in the study of envy (Smith, 2004; Vid-             correct, high levels of social identification should
aillet, 2006), although its precise influence remains          weaken envy’s triggering effect on the moral disen-
murky (Alicke & Zell, 2008). Conceptualized in                 gagement process. High social identification
terms of the connection that exists between peo-               should reduce the likelihood that envy will trans-
ple—personal affiliation, closeness, or similarity             late into dehumanization of others in a work envi-
(e.g., values, gender, culture) (Schaubroeck & Lam,            ronment, moral justification of antisocial behavior,
2004; Smith, 2000; Smith & Kim, 2007; van Dijk,                and a belief that others deserve to be harmed (Clay-
Ouwerkerk, Goslinga, Nieweg, & Gallucci, 2006)—                ton & Opotow, 2003; Exline & Zell, 2008). Thus, we
high levels of social identification with others               argue that when envy occurs, high levels of social
make invidious feelings more likely and more in-               identification prevent individuals from disengag-
tense (Tesser, 1988; see also van Dijk et al., 2006).          ing their moral self-sanctions and inhibit social
But the evidence thins and the overall picture be-             undermining.
comes more complicated when one considers how                     In contrast, we propose that when social identi-
social identification influences the mediation pro-            fication is low, envy will trigger a disengagement of
cesses outlined above. No studies, to our knowl-               self-regulatory mechanisms that makes social un-
edge, have examined how social identification en-              dermining more likely. Among individuals who
hances or weakens the effects of envy on antisocial            fail to identify with their coworkers and colleagues,
behaviors such as social undermining. Why might                the esteem threats and hostility associated with
social identification (i.e., focusing on common                envy are more likely to be translated into social
bonds or connections) with colleagues weaken the               undermining behaviors through moral disengage-
mediating relationships among envy, moral disen-               ment. In these situations, potential targets are seen
gagement, and social undermining? For one thing, a             as less entitled to compassion and moral obligation
648                                       Academy of Management Journal                                       June

and more as eligible targets for harm doing (Brock-         others.” The items had seven Likert-type response
ner, 1990; Opotow, 1995, 2001; Staub, 1989). In             options (␣ ⫽ .71).
other words, a lack of identification with coworkers           Social identification (time 1). We followed
allows envy to be translated into moral disengage-          Schaubroeck and Lam (2004) and operationalized
ment and subsequent harm-doing.                             social identification as perceived interpersonal
   The preceding arguments produce a first-stage            similarity (see also Liviatan, Trope, & Liberman,
moderated-indirect effect model that incorporates           2008). Three items were used, two of which are “In
several related predictions among envy, moral dis-          general, others with whom I work are similar to
engagement, and social identification. Stated               me” and “Most others I work with have a back-
formally,                                                   ground similar to mine.” The items had five Likert-
                                                            type response options (␣ ⫽ .89).
  Hypothesis 1. The strength of the mediated
                                                               Moral disengagement (time 1). Moral disengage-
  relationship between envy and social under-
                                                            ment was assessed with a 15-item measure (␣ ⫽
  mining (via moral disengagement) varies de-
                                                            .91) from McFerran et al. (2010) based on Bandura
  pending on the extent of social identification;
                                                            et al.’s (1996) moral disengagement scale but
  the indirect effect of envy via moral disengage-
                                                            adapted to a work context. The items assess the
  ment on social undermining is stronger when
                                                            extent to which individuals construe injurious con-
  social identification is lower.
                                                            duct as serving a morally justified purpose, mask
                                                            censurable activities through euphemistic language
              STUDY 1: METHODS                              or advantageous comparison, disavow or displace
                                                            responsibility for harm, and blame and devalue
Sample and Procedures
                                                            targets of harmful conduct. Following Bandura et
   The participants were full-time employees at a           al. (1996), the scale items comprised diverse forms
university hospital in a midwestern city. Data for          of detrimental conduct. Sample items are “People
this study were collected at two times eight months         who are mistreated at work have usually done
apart. At time 1, a hospital administrator notified         something to deserve it” and “Making fun of your
employees of the study and its goals via e-mail one         coworkers doesn’t really hurt them.” The items had
week prior to the study. Members of the research            seven Likert-type response options. Following
team staffed a private conference room near the             prior researchers (e.g., Bandura et al., 1996; Detert
cafeteria and administered the questionnaire dur-           et al., 2008; McFerran et al., 2010), we conceptual-
ing three lunch shifts (11 a.m.–1 p.m., 7–9 p.m.,           ized moral disengagement as a single higher-order
and 1–3 a.m.) for five consecutive days. We col-            construct. We conducted a second-order confirma-
lected data during the lunch break because hospital         tory factor analysis with the 15 items loading onto
administrators conveyed that it was important that          the three first-order latent factors and the three
all hospital employees (on all three shifts) be pro-        first-order mechanisms loading onto a single sec-
vided the chance to participate. A total of 432 em-         ond-order latent variable. The results revealed ex-
ployees completed time 1 questionnaires (a 15 per-          cellent overall model fit (␹2 ⫽ 158.53, df ⫽ 87, ␹2/df
cent approximate response rate). Eight months after         ⫽ 1.82, AGFI ⫽ .93, CFI ⫽ .96, RMSEA ⫽ .04).
the initial data collection, time 2 surveys were ad-        Thus, we averaged responses to the 15 items to
ministered following the same lunch hour proce-             form an overall moral disengagement measure (␣
dure described above. A total of 464 employees              ⫽ .91).
completed time 2 questionnaires (a 17 percent ap-              Social undermining (time 2). We measured so-
proximate response rate); 177 of these individuals          cial undermining at time 2 using Duffy et al.’s
also participated in the first phase of data collec-        (2006) six-item measure (␣ ⫽ .93). Participants
tion. Missing data on key variables reduced the             were asked to report their level of undermining in
analysis sample size to 160. Women totaled 77               the months since the first survey was completed.
percent of the participants; the average age was 37         Sample items are “I sometimes talk bad about my
years; and the average tenure was 9 years.                  coworkers behind their backs” and “I sometimes
                                                            intentionally give my coworkers the ‘silent treat-
                                                            ment.’” The items had seven Likert-type response
Measures
                                                            options.
   Envy (time 1). Envy was assessed using five                 Control variables (time 1). Drawing on a review
items that Vecchio (1995, 1999) developed and val-          of the relevant literature, we controlled for several
idated. Sample items are “Most of my coworkers              variables. Age, gender, and tenure may be related to
have it better than I do” and “At work, I see myself        perceptions of social interactions, status, and social
as an underdog who isn’t taken as seriously as              comparisons, as well as to antisocial behavior (e.g.,
2012                                           Duffy, Scott, Shaw, Tepper, and Aquino                                              649

Duffy et al., 2006; Lakey & Cassidy, 1990; Tepper,                      ables— envy and social identification—and to rule
2000) and were therefore controlled. We also con-                       out the possibility that common method effects
trolled for negative affective disposition and proce-                   could explain the presence of the second-order fac-
dural justice, as they may relate to envious emo-                       tor. Following Spreitzer (1995), we included social
tions and social undermining behavior (Cohen-                           identification (⌬␹ 2 ⫽ 227.25, df ⫽ 64, p ⬍ .01) and
Charash & Mueller, 2007; Duffy et al., 2002).                           envy (⌬␹2 ⫽ 274.93, df ⫽ 79, p ⬍ .01) as separate
Negative affective disposition (␣ ⫽ .83) was opera-                     first-order factors in the moral disengagement sec-
tionalized using the negative markers from the Pos-                     ond-order model. A third model included both so-
itive and Negative Affect Schedule (PANAS; Wat-                         cial identification and envy as additional first-or-
son, Clark, & Tellegen, 1988) (␣ ⫽ .87). Procedural                     der factors (⌬␹2 ⫽ 537.19, df ⫽ 138, p ⬍ .01). In
justice was measured with four items from Niehoff                       each comparison model, the fit of the model was
and Moorman (1993) (␣ ⫽ .75).                                           reduced. We also estimated several additional
                                                                        models in which social identification and envy
                                                                        items loaded on first-order factors in the moral
                     STUDY 1: RESULTS
                                                                        disengagement model. In each case, model fit was
Response Bias Checks and Measurement Issues                             significantly better for the hypothesized model
                                                                        with moral disengagement as a second-order latent
   To address the potential for response and self-
                                                                        factor and social identification and envy items
selection biases, we coded longitudinal partici-
                                                                        loading on their respective first-order factors.
pants as 1 and coded those who participated at time
1 only as 0 and included this dichotomy as the
dependent variable in a logistic regression analysis
                                                                        Hypothesis Tests
with time 1 predictors (envy, social identification,
and moral disengagement) as well as several demo-                          The descriptive statistics and correlations among
graphic variables (age, gender, tenure, and salary                      the study variables are shown in Table 1. The re-
level). No variable in the equation was significant.                    gression analyses are shown in Table 2. We tested
We also compared longitudinal participants (coded                       the moderated-mediation hypothesis using the
1) with those participating only at time 2 (coded 0)                    nested-equations path analytic approach advocated
on the demographic variables as well as on time 2                       by Edwards and Lambert (2007), which expresses
coworker undermining, but again no significant                          the relationships as the integration of the family of
predictors were found. The participants’ age (37                        equations that comprise moderated-mediation
years on average) and gender (77 percent female)                        tests. This is accomplished by substituting the re-
profiles were very similar to estimates of the                          gression equation(s) for the mediating variable(s)
broader hospital population provided by hospital                        (moral disengagement, in this case) into the equa-
administration (⬇38 –39 years, 80 percent female).                      tion for a given dependent variable (here, social
Although our second-order factor analysis for the                       undermining). These reduced-form equations are
moral disengagement measure showed good model                           then used to derive direct, indirect, and total effects
fit, we conducted additional analyses to assess the                     of the independent variable (individual-level un-
adequacy of the measurement of our other key vari-                      dermining) across levels of the moderator variable.

                                                            TABLE 1
                                         Correlations and Descriptive Statistics, Study 1a
                 Variable             Mean    s.d.      1        2        3         4          5         6        7        8        9

1.       Age                          37.40   9.31
2.       Gender                        0.77   0.41   –.09
3.       Tenure                        9.04   7.02    .57**     .08
4.       Negative affectivity          2.12   0.58   –.16*      .01    –.19*      (.83)
5.       Procedural justice            3.72   1.29   –.13      –.12    –.13       –.08      (.75)
6.       Envy                          3.60   0.99   –.07       .10     .07        .20**    –.27**    (.71)
7.       Social identification         3.28   0.81   –.14       .07     .02       –.02       .23**    –.09      (.70)
8.       Moral disengagement           3.24   0.76    .06       .03     .22**      .16*     –.32**     .50**    –.03     (.87)
9.       Social undermining, time 2    2.90   0.91   –.07      –.03    –.14        .18*      .02       .13*     –.10*     .18*    (.85)

     n ⫽ 160. Variables were assessed at time 1 except where noted. Gender was coded 1 for “female” and 0 for “male.” Coefficient alpha
     a

reliabilities are reported on the main diagonal in parentheses.
      * p ⬍ .05
     ** p ⬍ .01
650                                                 Academy of Management Journal                                                 June

                                                             TABLE 2
                                                    Regression Results, Study 1a
                                               Moral Disengagement                              Social Undermining ⴝ Time 2

          Variables                        Step 1                  Step 2                     Step 1                     Step 2

Age                                         .00                     .00                        .00                        .00
Gender                                     –.06                    –.03                       –.24                       –.24
Tenure                                      .02                     .02                       –.01                       –.01
Negative affectivity                        .21*                    .20*                       .25*                       .23*
Procedural justice                         –.11*                   –.09*                       .05                        .08
Envy                                        .44**                   .49**                      .16*                       .06
Social identification                       .02                    –.00                       –.17*                      –.18**
Social identification envy                                         –.24**
Moral disengagement                                                                                                       .22**
Total R2                                    .40**                   .43**                      .14**                      .18**
⌬R2                                         .40**                   .03**                      .14**                      .04**

  a
      n ⫽ 160. Variables were assessed at time 1 except where noted. Gender was coded 1 for “female” and 0 for “male.”
       * p ⬍ .05
      ** p ⬍ .01

Our theoretical model is an example of a “first-                        icant (b ⫽ –.24, p ⬍ .01) and explained an addi-
stage” mediation hypothesis because the moderat-                        tional 3 percent of the variance. As predicted, the
ing effect of social identification operates on the                     relationship between envy and moral disengage-
first stage of the indirect relationship between envy                   ment was strongly positive when social identifica-
and social undermining behavior (Edwards & Lam-                         tion was low (blow SI ⫽ .73, p ⬍ .01), and was
bert, 2007).                                                            significant, but it was significantly weaker when
   We used path analysis conventions for describ-                       social identification was high (bhigh SI ⫽ .25, p
ing relationships in terms of the direct, indirect,                     ⬍ .01).
and total effects of envy on social undermining at                         The right side of Table 2 reports the regression
different levels of social identification. Thus, PMX                    results for social undermining. Envy positively (b
refers to the paths from X (envy) to M (moral dis-                      ⫽ .16, p ⬍ .05) and social identification negatively
engagement mediator); PYM is the path from M                            (b ⫽ –.17, p ⬍ .05) predicted social undermining
(moral disengagement) to Y (social undermining);                        assessed eight months later at time 2. At step 2,
PYX is the path from X to Y (that is, the direct effect                 moral disengagement positively predicted social
of envy on social undermining); PYM ⫻ PMX refers                        undermining (b ⫽ .22, p ⬍ .01), explaining 4 per-
to the indirect effects, and PYX ⫹ (PYM ⫻ PMX) is the                   cent of the variance in social undermining.
total effect of X on Y. Product terms such as indi-                        We used the information from the regression re-
rect effects are not normally distributed and, con-                     sults in Table 2 to conduct path-analytic tests at
sequently, the type 1 error rate may be inflated                        low and high levels of social identification. As Ta-
when such terms are tested for significance (Shrout                     ble 3 shows and Figure 2 illustrates, the effects of
& Bolger, 2002). Therefore, we followed Edwards                         envy on social undermining through moral disen-
and Lambert’s (2007) suggestions and estimated the                      gagement vary across levels of social identification.
sampling distributions of the product of regression                     When social identification was low, the indirect
coefficients using a bootstrap procedure with                           effects of envy on social undermining through
10,000 samples to construct confidence intervals                        moral disengagement (PYM ⫻ PMX ⫽ .16, p ⬍ .01)
for the significance tests of indirect and total                        and the total effects (PYX ⫹ [PYM ⫻ PMX] ⫽ .22, p ⬍
effects.                                                                .01) were significant. In contrast, the indirect (PYM
   As the regression results in the left-hand col-                      ⫻ PMX ⫽ .05, n.s.) and total effects (PYX ⫹ [PYM ⫻
umns of Table 2 show, envy is significantly related                     PMX] ⫽ .11, n.s.) of envy on social undermining
to moral disengagement in step 1 (b ⫽ .44, p ⬍ .01),                    were not significant when social identification was
explaining a unique 19 percent of the variance                          high. Thus, Hypothesis 1 was supported. Following
(semipartial R2), but the main effect of social iden-                   Edwards and Lambert (2007) and Duffy et al.
tification was not significant (b ⫽ .02, n.s.). In step                 (2006), we conducted several tests to compare our
2 of the moral disengagement equations, the inter-                      hypothesized first-stage moderated mediation
action of envy and social identification was signif-                    model with other possible models that could pro-
2012                                                Duffy, Scott, Shaw, Tepper, and Aquino                                           651

                                                           TABLE 3
               Path-Analytic Results, Study 1: Indirect and Total Effects of Envy (via Moral Disengagement) on
                            Social Undermining at Low and High Levels of Social Identificationa
                                                                              Direct Effects       Indirect Effects       Total Effects
                   Variables                         PMX          PYM             (PYX)             (PYM ⴛ PMX)       (PYX ⴙ [PYM ⴛ PMX])

Simple paths for low social identification           .73**       .22**             .06                  .16**                .22**
Simple paths for high social identification          .25**       .22**             .06                  .05                  .11

  a
      n ⫽ 160. Coefficients in bold are significantly different across social identification levels.
      ** p ⬍ .01

vide an alternative explanation for our findings.                                   STUDY 2: THEORETICAL EXTENSIONS
Specifically, we examined whether social identifi-                                            AND CONTEXT
cation moderated the direct relationship between
envy and social undermining and whether social                                  The findings from Study 1 support our theory of
identification moderated the second stage between                             the relationship between envy and social under-
moral disengagement and social undermining. Nei-                              mining and thereby make two novel theoretical
ther interaction term was significant. Hence, the                             contributions. First, Study 1 results elucidate a
hypothesized first-stage moderation model was the                             mechanism, moral disengagement, through which
best representation of the data.                                              envy of others at work relates to social undermin-

                                               FIGURE 2
  Moderated Indirect Effect of Envy on Social Undermining (via Moral Disengagement) at Low and High
                                 Levels of Social Identification, Study 1
652                                       Academy of Management Journal                                       June

ing behavior. Second, these results diverge from               We argue that norms for undermining in a social
current thinking by showing that envy relates to            context either strengthen or weaken the relation-
social undermining behavior (through moral disen-           ship between moral disengagement and individual
gagement) only when social identification with co-          social undermining behavior. We define undermin-
workers is low. When social identification is high,         ing norms in the team context as aggregate percep-
envious individuals do not disengage their self-            tions that individuals in the social context are un-
sanctions against harmful behavior. Thus, social            dermining others (e.g., Duffy et al., 2006; Tepper et
identification appears to be an antidote to the neg-        al., 2008). Our conceptual definition reflects the
ative interpersonal effects of envy.                        idea of a descriptive norm, or the “perceived prev-
   Study 2 was designed to extend Study 1 in two            alence or typicality of a given behavior” (Jacobson,
ways. First, we sought to constructively replicate          Mortensen, & Cialdini, 2011: 434). In essence, we
Study 1 by examining our model at the team level            argue that team norms for undermining behavior
and using an alternative operationalization of so-          play a role in conditioning or moderating the rela-
cial identification: degree of within-team identifi-        tionship between moral disengagement and indi-
cation. Drawing on social-identification-based the-         vidual social undermining (Bamberger & Biron,
ories, Brown, Condor, Mathews, Wade, and                    2007; Duffy et al., 2006).
Williams (1986) suggested that team identification             Although team undermining norms may directly
involves awareness of membership, an evaluation             influence individual undermining behavior, anti-
(positive or negative) of this membership, and af-          social behavior researchers (e.g., Duffy et al., 2006;
fective responses to membership. That is, teams             Robinson & O’Leary-Kelly, 1998) and social influ-
differ systematically in terms of whether members           ence researchers (e.g., Hackman, 1992) have long
define themselves as part of a team, view this mem-         recognized that such norms may also moderate the
bership positively, and have an affinity for other          relationship between individual factors and behav-
team members. Modeling identification as a team-            iors. We expect that the relationship between moral
level construct challenges the robustness of the            disengagement and individual undermining behav-
Study 1 findings across contexts and operational-           ior will be moderated by team undermining norms.
izations of key constructs, which improves the                 We believe that when team undermining norms
value of our replication (Schmidt, 2009).                   are strong—when the social environment of a team
   The second major objective of Study 2 was to             is rife with individuals undermining their col-
incorporate a broader conceptualization of social           leagues—the relationship between moral disen-
context, one including ambient undermining                  gagement and individual undermining behavior
norms, into our model. Specifically, in Study 2 we          will be stronger. Under these conditions, as moral
examined the notion that undermining norms pres-            disengagement increases, individuals will have not
ent in a team may serve as another antidote to the          only selectively disengaged self-sanctions against
effects of envy by shaping how individuals react to         antisocial behavior, but will also have been in a
moral disengagement. Social influence and identity          team social context that encourages such behav-
researchers (e.g., Turner, 1982, 1991) have argued          ior. The information contained in a strong de-
that implicit and informal pressures to conform to          scriptive norm for social undermining is partic-
social norms can influence behavior, even in the            ularly pertinent in this situation. First,
absence of explicit agreements or formal rules. So-         individuals use the information communicated
cial influence can be categorized as normative, in-         in descriptive norms as a heuristic for behaviors
formational, and referent informational (e.g., Bam-         that are likely to be immediately advantageous to
berger & Biron, 2007; Turner, 1991). Normative              themselves—for example, behaviors that are
influence involves pressures to conform that are            aligned with intrapersonal rather than interper-
based on individual desires to be approved of               sonal motives (Cialdini & Trost, 1998). Second,
and/or avoid rejection. That is, pressures to con-          because individuals use descriptive norms, or the
form are exerted because others in the social envi-         typical behavior exhibited by others in their so-
ronment have the means to reward or punish. In-             cial environment, as a short-cut for behavioral
formational     influence    involves   individuals         decision making in a given context, the effects of
searching for information that can be used to make          norms are likely to be the most powerful when
accurate behavioral choices (Turner, 1982). Refer-          self-regulatory capability is depleted (Berger &
ent informational forces are those that individuals         Rand, 2008). When team undermining norms are
use in attempts “to align their own beliefs and             strong, moral disengagement (that is, the deacti-
perceptions with those deemed characteristic of             vation of self-sanctions against harming others)
some salient social identity” (Bamberger & Biron,           should be a powerful predictor of social
2007: 184).                                                 undermining.
2012                                     Duffy, Scott, Shaw, Tepper, and Aquino                                  653

   When team undermining norms are weak—when                    students to teams randomly. The analysis sample
few individuals in a social environment undermine               size was 247. Data for this study were collected four
their colleagues—the relationship between moral dis-            times (at four- to eight-week intervals) during the
engagement and individual undermining behavior                  course of the semester. The first collection took
should be weaker. Here, the social environment ex-              place during the first week of class; the second
erts normative, informational, and referent informa-            collection was done approximately four weeks lat-
tional pressures to refrain from injurious behavior. As         er; the third data collection was during week 8 of
such, even though individual self-sanctions against             the term; and the final collection was done during
antisocial behavior may be deactivated (i.e., moral             the last week of class. We used data from all four
disengagement is high), the social environment                  waves in this study. Data for the control variables
should assuage the inclination to undermine others.             were collected at time 1; data for the envy and team
   These arguments can be integrated with our theory            identification variables were collected at time 2;
development in Study 1 to produce an elaborated                 moral disengagement and team social undermin-
process model that accounts for the mechanism by                ing, at time 3; and individual social undermining
which envy leads to social undermining (i.e., moral             behavior, at time 4. The classes included 408 stu-
disengagement) and antidotes to envy based on both              dents in 96 teams. Missing data over the four waves
social identity and social context. In the literature,          of data reduced the individual-level analysis sam-
envy of others is viewed as being a “call to action”            ple to 247, representing an overall 60 percent par-
(Smith & Kim, 2007: 54; Vidaillet, 2006) in of the form         ticipation rate. Members of the research team (not
of antisocial behavior; we qualify this line of reason-         course instructors) distributed and collected the
ing and theorize that this link, through moral disen-           study questionnaires. Participants were informed
gagement, will be observed under a narrow set of                that their responses would be kept confidential and
circumstances. We propose that the relationship be-             that their participation was voluntary. The average
tween envy and moral disengagement is stronger                  age of participants was 22.99 (s.d. ⫽ 3.02), and 56
when team identification is low and that the relation-          percent were female.
ship between moral disengagement and individual
social undermining is stronger when undermining
                                                                Measures: Independent Variables
norms are high. Stated in the terms that Edwards and
Lambert (2007) used, the indirect effect of envy on                Envy (time 2). Envy was again assessed with the
social undermining behavior will be strongest when              Vecchio (1995, 1999) measure, adapted to the team
team identification is low (a first-stage moderator)            context. The items had seven Likert-type response
and when team undermining norms are strong (the                 options (␣ ⫽ .83).
second-stage moderator). The integrative hypothesis                Team identification (time 2). Team identifica-
is stated below:                                                tion was assessed using a nine-item measure from
                                                                Wheeless, Wheeless, and Dickson-Markman
  Hypothesis 2. The strength of the mediated
                                                                (1978). Sample items are “My team is very close to
  relationship between envy and social under-
                                                                each other” and “My team members share a lot in
  mining (via moral disengagement) varies de-
                                                                common with one another.” The items had seven
  pending on the extent of social identification
                                                                Likert-type response options (␣ ⫽ .79).
  and team undermining norms; the indirect ef-
                                                                   Moral disengagement (time 3). As in Study 1,
  fect of envy on social undermining is stronger
                                                                moral disengagement was measured with the 15-
  when social identification is lower and team
                                                                item scale from McFerran et al. (2010), adapted to
  undermining norms are higher.
                                                                the context. The items had seven Likert-type re-
                                                                sponse options (␣ ⫽ .90). A second-order confirma-
               STUDY 2: METHODS                                 tory factor model revealed good overall model fit
                                                                (␹2 ⫽ 128.76, df ⫽ 87, ␹2/df ⫽ 1.48, AGFI ⫽ .93, CFI
Sample
                                                                ⫽ .97, RMSEA ⫽ .05). The coefficient alpha reli-
  Participants were enrolled in upper-division                  ability for the 15-item overall measure was .90.
business administration courses at a large midwest-                Team undermining norms (time 3). The 13-item
ern university. The study was designed to closely               social undermining measure from Duffy et al.
approximate the work done by groups in industrial               (2002) was used (␣ ⫽ .93). The items were adapted
settings by ensuring that (1) students were assigned            from a coworker perspective to a group context.
to groups that remained intact throughout the term,             The instructions directed participants to report
and (2) groups completed multiple assignments                   how often their team members intentionally en-
throughout the term (see Duffy, Shaw, and Stark                 gaged in each form of undermining. Sample items
[2000] for a similar approach). Instructors assigned            are “How often have your team members intention-
654                                       Academy of Management Journal                                       June

ally talked bad about other teammates behind their          team identification and envy items loaded on first-
backs?” and “How often have your team members               order factors in the moral disengagement model. In
intentionally given each other the “silent treat-           each case, model fit was significantly better for the
ment”?” The items had response options from 1               hypothesized model with moral disengagement as
(“never”) to 7 (“all the time”). Mean scores on the         a second-order latent factor and team identification
13 items were aggregated to the team level.                 and envy items loading on their respective first-
   Individual social undermining (time 4). We               order factors.
measured individual social-undermining behavior                We calculated rwg(j) (James, Demaree, & Wolf,
at time 4 using Duffy et al.’s (2002) 13-item measure       1984) to assess intrateam agreement and also cal-
(␣ ⫽ .95). We adapted the measure to reflect self-          culated ICC1 and ICC2 before aggregating the team
reports of one’s own undermining behavior during            identification and team undermining norms mea-
the term (e.g., “How often have you intentionally           sures to the team level. For team identification at
talked bad about your team members behind their             time 2, the mean rwg(j) was .85 (range ⫽ 0.24 –1.00),
backs?”). The items had response options that               and 86 percent of teams had agreement levels
ranged from 1 (“never”) to 7 (“all the time”).              higher than 0.70. ICC1 represents the reliability of a
   Control variables (time 1). We controlled for            single or individual assessment of the group mean,
age, gender, negative affectivity (using PANAS              whereas ICC2 reflects the reliability of the team
[Watson et al., 1988]), and procedural justice (six         means. These values were .21 and .55 for ICC1 and
items [Moorman, 1991]), following the arguments             ICC2, respectively. For team undermining norms at
for their inclusion in Study 1. Coefficient alpha           time 3, rwg(j) values ranged from 0.96 to 1.00, with
reliabilities for negative affectivity and procedural       an average value of .99. ICC values were .17 and .48
justice were .88 and .79, respectively. Tenure was          for ICC1 and ICC2, respectively. Supporting aggre-
not relevant in the context of student teams.               gation, the rwg(j) values indicated strong within-
                                                            team agreement, and ICC1 calculations revealed
                                                            significant between-team variance in the team
               STUDY 2: RESULTS                             identification and team undermining norms mea-
                                                            sures. The ICC2 values fell below standard bench-
Response Bias Checks and Measurement Issues
                                                            marks for team-mean reliability, which may result
   We used logistic regressions and comparisons             in underestimated relationships (Bliese, 1998). The
over time periods to assess potential response bias.        implications will be addressed in the Discussion.
First, we compared those who were eliminated be-
cause of missing data (n ⫽ 161) with participants in
                                                            Hypothesis Tests
the final analysis sample (n ⫽ 247) on a range of
demographic and expectation variables collected at             The tests of the integrative moderated-mediation
time 1. The variables were age, gender, grade point         hypothesis using team-level identification in-
average (GPA), class standing, number of prior              volved a “slopes as outcomes” or cross-level mod-
classes taken that involved teamwork, and efficacy          erator analysis in hierarchical linear modeling
expectations for the class; for example, “I am con-         (HLM; e.g., Raudenbush & Bryk, 2002). The analy-
fident that I can do well in this class.” We coded          ses for evaluating the replication model were par-
analysis sample participants as 1 and participants          allel to those in Study 1, although coefficients are
at time 1 only as 0 and included this dichotomy as          ␥’s in HLM, and the interaction of envy and team
the dependent variable in a logistic regression anal-       identification (XZ) is a cross-level moderation ef-
ysis with the predictors. Only one variable, gender,        fect. Tests of the theoretical extension involved
was significant in this analysis. Male participants         team undermining norms as a second stage moder-
were more likely to have missing data on one of the         ator. Thus, the nested equations for the full model
study variables. Gender was included as a control           incorporated this additional moderating variable.
in the multivariate analyses. We also compared              The tests of conditional indirect effects in the full
measurement models with team identification (⌬␹2            model examine whether the mediated effect of
⫽ 240.24, df ⫽ 64, p ⬍ .01) and envy (⌬␹2 ⫽ 242.75,         envy on individual social undermining behavior
df ⫽ 62, p ⬍ .01) included as separate first-order          through moral disengagement varies as a function
factors in the moral disengagement second-order             of team identification (the stage 1 moderation of
model. A third model included both social identi-           team identification on the envy-moral disengage-
fication and envy as additional first-order factors         ment relationship) and team undermining norms
(⌬␹2 ⫽ 489.48, df ⫽ 208, p ⬍ .01). In each compar-          (the stage 2 moderation of team undermining
ison model, the fit of the model was reduced. We            norms on the relationship between moral disen-
also estimated several additional models in which           gagement and individual social undermining).
2012                                                  Duffy, Scott, Shaw, Tepper, and Aquino                                                 655

   A null model test with moral disengagement as                               where ␶11 intercept-as-outcomes is the residual be-
the outcome variable revealed that 8 percent (p ⬍                              tween-group variance in slopes in a model without
.05) of the variance in moral disengagement resided                            the cross-level interaction term, and ␶11 slopes-as-
at the team level and 92 percent resided at the                                outcomes is the residual between-group variance in
individual level. When individual social under-                                slopes in a model with the interaction term. Using
mining was the outcome, the null model test re-                                this formula, we found that team identification ex-
vealed that 6 percent (p ⬍ .05) of the variance                                plains 20 percent of the available slope variance in
resided at the team level and 94 percent resided at                            the relationship between envy and moral
the individual level.                                                          disengagement.
   Descriptive statistics for and correlations among                              The right side of Table 5 includes the results
the study variables are presented in Table 4. The                              when individual social undermining (time 4) was
HLM results are shown in Table 5, and the path-                                the outcome variable. Envy was positively related
analytic estimates are shown in Table 6. As the                                to individual social undermining assessed at time 4
HLM results in Table 5 show, when moral disen-                                 (␥ ⫽ .07, p ⬍ .05), explaining 3 percent of the
gagement at time 2 is the outcome, envy is signifi-                            available variance at level 1, and team identifica-
cantly and positively related (␥ ⫽ .29, p ⬍ .01) in                            tion was also significantly related to individual
step 1, explaining 11 percent of the available level                           social undermining assessed at time 4 (␥ ⫽ –.07, p
1 variance in moral disengagement. Team identifi-                              ⬍ .05). In step 2, moral disengagement was signif-
cation, in contrast, is not a significant level 2 pre-                         icantly related to social undermining (␥ ⫽ .07, p ⬍
dictor of moral disengagement in step 1 (␥ ⫽ –.11,                             .01) and explained 14 percent of the level 1 vari-
n.s.). In step 2 of the moral disengagement equa-                              ance, while team undermining norms was also a
tions, the team identification cross-level interac-                            significant predictor (␥ ⫽ .12, p ⬍ .01), explaining
tion is significant (␥ ⫽ –.26, p ⬍ .05). When team                             70 percent of the available variance in intercepts at
identification was low, the relationship between                               level 2.
envy and moral disengagement was significant and                                  The information from the HLM results in Table 5
positive (␥low time I ⫽ .60, p ⬍ .01), but it was not                          was used to conduct path-analytic tests at low and
significant when team identification was high                                  high levels of team identification. These results are
(␥high time I ⫽ .02, n.s.). We computed a pseudo-R2
                                                                               shown in Table 6. The path estimates revealed that
for team identification as a level 2 moderator of the
                                                                               the effects of envy on individual social undermin-
relationship between envy and moral disengage-
                                                                               ing through moral disengagement varied across lev-
ment. Level 2 slopes as outcomes pseudo-R2 values
                                                                               els of team identification. When team identification
are calculated relative to the amount of between-
                                                                               was low, the indirect effects of envy on social un-
group variation in slopes (Hofmann, Griffin, &
                                                                               dermining (PYM ⫻ PMX ⫽ .04, p ⬍ .05) and the
Gavin, 2000), and the formula is:
                                                                               total effects of envy on social undermining were
                                                                               significant (PYX ⫹ [PYM ⫻ PMX] ⫽ .06, p ⬍ .01).
Pseudo-R2, level 2 slope model ⫽ (␶11 intercept-
                                                                               As expected, neither indirect nor total effects of
           as-outcomes ⫺ ␶11 slopes-as-outcomes) ⁄                             envy on social undermining were significant when
                              ␶11 intercept-as-outcomes,          (1)          team identification was high. In addition, the first-

                                                                 TABLE 4
                                              Correlations and Descriptive Statistics, Study 2a
                        Variables                       Mean    s.d.      1       2        3        4      5        6        7        8       9

1.   Age, time 1                                        22.99   3.02
2.   Gender, time 1                                      0.56   0.49    –.01
3.   Negative affectivity, time 1                        1.81   0.62     .00    .05   (.88)
4.   Procedural justice, time 3                          3.59   0.88    –.06   –.03    .15*       (.79)
5.   Envy, time 2                                        2.33   0.87     .06    .18**  .13*       –.05 (.83)
6.   Team identification, time 2, team level             3.81   1.2     –.02   –.05   –.08        –.07  .19** (.79)
7.   Moral disengagement, time 3                         2.86   0.86    –.15**  .28**  .24**      –.09  .32**  .08   (.90)
8.   Team undermining norms, time 3, team level          1.22   0.16     .04    .08    .07        –.10  .06   –.15*  –.20** (.93)
9.   Social undermining, time 4                          1.32   0.28     .01    .10    .12         .01  .16*   .23**  .08    .44** (.95)

     a
         n ⫽ 247. Coefficient alpha reliabilities are reported on the main diagonal in parentheses. Gender coded 1 for “female” and 0 for “male.”
          * p ⬍ .05
         ** p ⬍ .01
You can also read