22
Beyond the Individual Victim: Multilevel Consequences of Abusive Supervision in Teams Crystal I. C. Farh Michigan State University Zhijun Chen University of Western Australia We conceptualize a multilevel framework that examines the manifestation of abusive supervision in team settings and its implications for the team and individual members. Drawing on Hackman’s (1992) typology of ambient and discretionary team stimuli, our model features team-level abusive supervision (the average level of abuse reported by team members) and individual-level abusive supervision as simultaneous and interacting forces. We further draw on team-relevant theories of social influence to delineate two proximal outcomes of abuse—members’ organization-based self-esteem (OBSE) at the individual level and relationship conflict at the team level—that channel the independent and interactive effects of individual- and team-level abuse onto team members’ voice, team-role performance, and turnover intentions. Results from a field study and a scenario study provided support for these multilevel pathways. We conclude that abusive supervision in team settings holds toxic consequences for the team and individual, and offer practical implications as well as suggestions for future research on abusive supervision as a multilevel phenomenon. Keywords: abusive supervision, teamwork behavior, multilevel, teams Defined as “the extent to which supervisors engage in the sustained display of hostile verbal and nonverbal behaviors, ex- cluding physical contact” (Tepper, 2000, p. 178), abusive super- vision has received increasing attention from scholars and practi- tioners due to its prevalence and detrimental impact in the workplace. To date, research has linked abusive supervision to victims’ decreased organizational commitment, job performance, and citizenship behavior, as well as increased psychological dis- tress, counterproductive behaviors, deviance, and turnover (e.g., reviewed in Martinko, Harvey, Brees, & Mackey, 2013; Tepper, 2007). Although the majority of this work has focused on the consequences of abuse for targeted victims, more recent research has begun to examine the effects of abuse on individuals residing in interdependent team contexts. Particularly as modern organiza- tions are increasingly team based, there is a need to understand how abuse is manifested in team settings, its implications for the team, and its impact on members’ cognitions, attitudes, and be- haviors toward the team. Relevant to this question, two streams of research have inves- tigated the effects of abusive supervision in group settings. The first examines group-level abusive supervision as aggregated per- ceptions of abuse attributed to a common leader by members of the same group. Research has linked this aggregate construct to col- lective problem drinking, counterproductive behaviors, and devi- ance (Bamberger & Bacharach, 2006; Detert, Treviño, Burris, & Andiappan, 2007; Mawritz, Mayer, Hoobler, Wayne, & Marinova, 2012; Ogunfowora, 2013). A second approach examines the ex- perience of individual victims of abuse in the presence of group- level abuse. This research has shown that higher or lower levels of aggregated abuse in the group modify the way individuals interpret and respond to their own individualized experiences of abuse (e.g., Duffy, Ganster, Shaw, Johnson, & Pagon, 2006a; Duffy, Shaw, Scott, & Tepper, 2006; Hannah et al., 2013; Peng, Schaubroeck, & Li, 2013). Together, existing frameworks imply that abuse resides and exerts consequences at the individual and group levels of analysis simultaneously. To inform understanding of abusive supervision in team set- tings, however, existing approaches remain limited in three as- pects. First, much of prior research has relied on individual-level theories such as justice (Bies & Moag, 1986), social exchange (Blau, 1964), and stress (Lazarus & Folkman, 1984) to explain group-level consequences of abuse (e.g., Detert et al., 2007). While these theories may appropriately generalize to a group of independent members, they do not account for the heightened interdependence inherent to teams (Guzzo & Shea, 1992) or how abusive supervision occurring in team settings can produce shared ways of thinking and acting among its members (Hackman, 1992). There is thus a need to employ team-relevant theories and pro- cesses to adequately assess team-relevant consequences of abuse. Supporting this idea, a recent study (Priesemuth, Schminke, Am- brose, & Folger, 2013) showed preliminary evidence that shared perceptions of abuse can shape shared team cognitions like team efficacy and identification; however, research is needed to under- This article was published Online First August 11, 2014. Crystal I. C. Farh, College of Business, Michigan State University; Zhijun Chen, Business School, University of Western Australia. We thank Mo Wang for his valuable and constructive guidance in the review process. We also owe gratitude to Larry Farh, Russ Johnson, John Schaubroeck, Huiwen Lian, and Jian Liang for their developmental com- ments on earlier versions of this article. Finally, we thank Michael Howe for his assistance with data collection. Correspondence concerning this article should be addressed to Crystal I. C. Farh, College of Business, Michigan State University, 632 Bogue Street, N475, East Lansing, MI 48824. E-mail: [email protected] 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. Journal of Applied Psychology © 2014 American Psychological Association 2014, Vol. 99, No. 6, 1074 –1095 0021-9010/14/$12.00 http://dx.doi.org/10.1037/a0037636 1074

Multilevel Consequences

  • Upload
    hjkl

  • View
    22

  • Download
    1

Embed Size (px)

DESCRIPTION

multilevel consequences

Citation preview

  • Beyond the Individual Victim: Multilevel Consequences of AbusiveSupervision in Teams

    Crystal I. C. FarhMichigan State University

    Zhijun ChenUniversity of Western Australia

    We conceptualize a multilevel framework that examines the manifestation of abusive supervision in teamsettings and its implications for the team and individual members. Drawing on Hackmans (1992)typology of ambient and discretionary team stimuli, our model features team-level abusive supervision(the average level of abuse reported by team members) and individual-level abusive supervision assimultaneous and interacting forces. We further draw on team-relevant theories of social influence todelineate two proximal outcomes of abusemembers organization-based self-esteem (OBSE) at theindividual level and relationship conflict at the team levelthat channel the independent and interactiveeffects of individual- and team-level abuse onto team members voice, team-role performance, andturnover intentions. Results from a field study and a scenario study provided support for these multilevelpathways. We conclude that abusive supervision in team settings holds toxic consequences for the teamand individual, and offer practical implications as well as suggestions for future research on abusivesupervision as a multilevel phenomenon.

    Keywords: abusive supervision, teamwork behavior, multilevel, teams

    Defined as the extent to which supervisors engage in thesustained display of hostile verbal and nonverbal behaviors, ex-cluding physical contact (Tepper, 2000, p. 178), abusive super-vision has received increasing attention from scholars and practi-tioners due to its prevalence and detrimental impact in theworkplace. To date, research has linked abusive supervision tovictims decreased organizational commitment, job performance,and citizenship behavior, as well as increased psychological dis-tress, counterproductive behaviors, deviance, and turnover (e.g.,reviewed in Martinko, Harvey, Brees, & Mackey, 2013; Tepper,2007). Although the majority of this work has focused on theconsequences of abuse for targeted victims, more recent researchhas begun to examine the effects of abuse on individuals residingin interdependent team contexts. Particularly as modern organiza-tions are increasingly team based, there is a need to understandhow abuse is manifested in team settings, its implications for theteam, and its impact on members cognitions, attitudes, and be-haviors toward the team.

    Relevant to this question, two streams of research have inves-tigated the effects of abusive supervision in group settings. The

    first examines group-level abusive supervision as aggregated per-ceptions of abuse attributed to a common leader by members of thesame group. Research has linked this aggregate construct to col-lective problem drinking, counterproductive behaviors, and devi-ance (Bamberger & Bacharach, 2006; Detert, Trevio, Burris, &Andiappan, 2007; Mawritz, Mayer, Hoobler, Wayne, & Marinova,2012; Ogunfowora, 2013). A second approach examines the ex-perience of individual victims of abuse in the presence of group-level abuse. This research has shown that higher or lower levels ofaggregated abuse in the group modify the way individuals interpretand respond to their own individualized experiences of abuse (e.g.,Duffy, Ganster, Shaw, Johnson, & Pagon, 2006a; Duffy, Shaw,Scott, & Tepper, 2006; Hannah et al., 2013; Peng, Schaubroeck, &Li, 2013). Together, existing frameworks imply that abuse residesand exerts consequences at the individual and group levels ofanalysis simultaneously.

    To inform understanding of abusive supervision in team set-tings, however, existing approaches remain limited in three as-pects. First, much of prior research has relied on individual-leveltheories such as justice (Bies & Moag, 1986), social exchange(Blau, 1964), and stress (Lazarus & Folkman, 1984) to explaingroup-level consequences of abuse (e.g., Detert et al., 2007).While these theories may appropriately generalize to a group ofindependent members, they do not account for the heightenedinterdependence inherent to teams (Guzzo & Shea, 1992) or howabusive supervision occurring in team settings can produce sharedways of thinking and acting among its members (Hackman, 1992).There is thus a need to employ team-relevant theories and pro-cesses to adequately assess team-relevant consequences of abuse.Supporting this idea, a recent study (Priesemuth, Schminke, Am-brose, & Folger, 2013) showed preliminary evidence that sharedperceptions of abuse can shape shared team cognitions like teamefficacy and identification; however, research is needed to under-

    This article was published Online First August 11, 2014.Crystal I. C. Farh, College of Business, Michigan State University;

    Zhijun Chen, Business School, University of Western Australia.We thank Mo Wang for his valuable and constructive guidance in the

    review process. We also owe gratitude to Larry Farh, Russ Johnson, JohnSchaubroeck, Huiwen Lian, and Jian Liang for their developmental com-ments on earlier versions of this article. Finally, we thank Michael Howefor his assistance with data collection.

    Correspondence concerning this article should be addressed to CrystalI. C. Farh, College of Business, Michigan State University, 632 BogueStreet, N475, East Lansing, MI 48824. E-mail: [email protected]

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    Journal of Applied Psychology 2014 American Psychological Association2014, Vol. 99, No. 6, 10741095 0021-9010/14/$12.00 http://dx.doi.org/10.1037/a0037636

    1074

  • stand how abusive supervision in team settings fundamentallyalters the social and behavioral fabric of the team.

    Team-relevant theories are also needed to explain the effects ofabuse on individual team members responses toward the team.Ironically, although much research has examined individualizedexperiences of abuse in group settings (e.g., Duffy, Ganster, et al.,2006; Hannah et al., 2013; Peng et al., 2013), this work hasfocused primarily on individuals responses toward the abusiveleader or organization (e.g., deviance, counterproductive behav-iors, retaliation; reviewed in Martinko et al., 2013), none of whichcapture responses relating to the team itself. There is thus a needto clarify the individualteam interface and specify how individ-ualized experiences of abuse in team settings shape membersteam-relevant cognitions (e.g., perceptions of self-concept in theteam), attitudes (e.g., willingness to stay in the team), and behav-iors (e.g., proactive and teamwork actions). Because these team-relevant responses constitute important ways that members con-tribute to overall team functioning (G. Chen, Sharma, Edinger,Shapiro, & Farh, 2011; Mathieu, Maynard, Rapp, & Gilson, 2008),examining them as consequences enables us to understand themicroprocesses of how abuse in team settings potentially harmsteam effectiveness. Relatedly, as prior research has focused pri-marily on the experiences of targeted victims in group settings,new theory is needed to assess ways that nontargeted membersmay be affected, simply by being a part of an abusive teamenvironment.

    Finally, much of prior research has taken a piecemeal approachto examining the effects of abuse in team settings, focusing onindividual consequences while ignoring the impact of aggregatedabuse on the team (e.g., Peng et al., 2013), or focusing on group-level consequences while ignoring individual differences in expe-rience among its members (e.g., Priesemuth et al., 2013). Thisapproach is unfortunate because it precludes us from knowing howindividual- versus team-level consequences of abuse differ fromeach other, or how their respective pathways interact to simulta-neously impact the team as well as the individuals that compose it.There is thus a need for a new framework that better reflects themultilevel reality of leadership influences on team systems (G.Chen & Kanfer, 2006; Hackman, 1992), and draws on team-relevant theories to specify the unique pathways and dual-levelmechanisms linking individual- and team-level abuse to membersresponses to the team. Absent such a framework, scholars andpractitioners may grossly underestimate the broader costs of abusefor the collective and the individuals residing in it.

    In view of these needs, our objective is to delineate and test amultilevel model of abusive supervision in teams. Drawing onHackmans (1992) typology of ambient and discretionary teamstimuli, we feature team-level abusive supervision (the averagelevel of abuse reported by team members) and individual-levelabusive supervision and as dual-level inputs to the teams socialsystem, and further draw on multilevel theories of leadership (G.Chen & Kanfer, 2006) to delineate individual-, cross-, and team-level pathways by which individual- and team-level abuse affectsmembers team-relevant cognitions, attitudes, and behaviors. Wethen integrate across theories of social influence in teamsthegroup-value model (Lind & Tyler, 1988), social learning theory(Bandura, 1971), and social comparison theory (Festinger,1954)to specify members organization-based self-esteem

    (OBSE) and team relationship conflict as dual-level mediatorslinking individual- and team-level abuse to members outcomes.

    Our model contributes to the abusive supervision literature inthree ways. First, we shed light on the multilevel nature of abusein teams by demonstrating it to be an individual- and team-levelinput into the teams social system. Second, we identify uniqueindividual-, team-, and cross-level pathways by which abusivesupervision affects teams, and introduce member OBSE and teamrelationship conflict as new proximal outcomes in team settings.Finally, we link abusive supervision to members team-relevantcognitions, attitudes, and behaviors, thereby illuminating a tripar-tite of microprocesses through which abuse potentially harms teameffectiveness. Together, our framework addresses calls to betterunderstand abusive supervision as a multilevel phenomenon (Tep-per, 2007), expand the realm of theories delineating consequencesof abuse (Martinko et al., 2013), and uncover new mechanisms andconsequences of abuse at both the team and individual levels ofanalysis (Aryee, Chen, Sun, & Debrah, 2007).

    Theory and Hypotheses

    Overall Framework

    In accordance with multilevel leadership theories in teams (G.Chen & Kanfer, 2006; Hackman, 1992), we conceptualize abusivesupervision in team settings as both discretionary and ambient innature. As a discretionary input, individual-level abuse reflectsthe degree of targeted abuse perceived by an individual member,whereas team-level abuse (an ambient input) reflects the overall,aggregated levels of abuse perceived by members of the team.Discretionary inputs are thought to create individualized experi-ences and explain why members of the same team may exhibitdifferential cognitions, attitudes, and behaviors. In contrast, ambi-ent inputs shape elements of the teams social universe, affectingall members belonging to that system, regardless of discretionarytreatment. Multilevel leadership theories model discretionary andambient inputs as parallel forces affecting proximal outcomes atthe individual and team levels of analysis (e.g., G. Chen, Kirkman,Kanfer, & Allen, 2007), and as forces that intersect, such thatambient inputs (and their proximal outcomes) exert top-down,cross-level influences on how individual members experience andrespond to discretionary treatment (Kozlowski & Klein, 2000).Prior research also indicates that individual- and team-level leaderinfluences on the team are distinct (Liao & Chuang, 2007)forinstance, discretionary treatment by the leader is thought to influ-ence members attitudes and behaviors by altering their self-concept (Shamir, House, & Arthur, 1993), whereas ambient treat-ment is thought to transform the teams shared ways of thinkingand interacting (Zaccaro, Rittman, & Marks, 2001).

    Together, multilevel leadership theories indicate that a multi-level framework for abusive supervision should consist ofindividual-, cross-, and team-level pathways, each explained bydifferential theoretical mechanisms. Within this multilevel infra-structure, we turn to three social influence theories of teamseachof which is well positioned to inform the pathways in our frame-work. Consistent with the self-concept model of discretionaryleader treatment, we first draw on the group-value model (Lind &Tyler, 1988) to delineate the individual-level pathway. The group-value model explains how individualized leader treatment signals

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1075ABUSIVE SUPERVISION IN TEAMS

    dHighlight

  • a given members self-concept in relation to the team. In particu-lar, we propose that individualized abuse negatively affects mem-bers perceptions of their self-worth and standing in the teamthatis, their OBSE (Pierce, Gardner, Cummings, & Dunham, 1989).To supplement the group-value model and delineate cross-leveleffects, we draw on social comparison theory (Festinger, 1954) toassess how team-level abusive supervisionan indicator of howcomparable others are treatedexerts top-down moderating influ-ences on these self-concept processes. Finally, we rely on sociallearning theory (Bandura, 1971) to delineate the team-level path-way because it explains why exposure to ambient abuse infects thesocial system of the team. In particular, we propose that sociallearning mechanisms facilitate the translation of ambient abuseinto laterally dysfunctional interpersonal interactionsthat is,team relationship conflict (Jehn, 1995). Integration of the group-value, social comparison, and social learning theories within amultilevel leadership infrastructure informs the unique mecha-nisms along each pathway and enables a more complete under-standing of the effects of abuse in teams beyond any one theoryalone (K. J. Mayer & Sparrowe, 2013).

    Beyond specifying member OBSE and team relationship con-flict as proximal outcomes, we further assess the distal effects ofindividual- and team-level abuse on members voice (i.e., behav-iors aimed at improving team processes; LePine & Van Dyne,1998), team-role performance (i.e., behaviors aimed at promotingcooperation and coordination in the team; Welbourne, Johnson, &Erez, 1998), and low turnover intentions (i.e., a willingness toremain in the team and maintain its viability over time; Hackman,1987). These responses to the team represent meaningful ways thatindividuals contribute to team effectiveness (G. Chen et al., 2011)and correspond well with our stated objective to examine a broaderset of team-relevant outcomes associated with abuse in teams. Ouroverarching model appears in Figure 1.

    Individual-Level Consequences ofAbusive Supervision

    Individual-level abusive supervision refers to the leaders sus-tained display of hostile verbal or nonverbal behaviors toward aparticular team member. The impact of individualized abuse ontargeted members can be assessed with the group-value model(Lind & Tyler, 1988; Tyler, 1989). In contrast to traditional social-exchange models of justice and instrumental desires for favorable

    outcomes (Smith, Tyler, Huo, Ortiz, & Lind, 1998), the group-value model is grounded in social identification theory (Hogg &Abrams, 1988; Tajfel & Turner, 1986), which suggests that mem-bers primary concerns about self-concept in relation to the teamlends intrinsic value to the meaning of leader individualized treat-ment. The group-value model posits that members use individual-ized treatment from authority figures of teams (e.g., the teamleader) as relational information about their position within a team,which in turn shapes their self-concept and perceptions of self-worth (Smith et al., 1998; Tyler, Degoey, & Smith, 1996). Becauseleader treatment symbolizes how other members of the team valuean individual, members who are treated with respect, politeness,and dignity are signaled to have higher status and value to the team(Tyler, 1989), whereas negative treatment signals ones low posi-tion in the team. In turn, members treated favorably developpositive evaluations of their self-worth, while members treatedunfavorably develop self-concepts associated with inferiority, ex-clusion, and feelings of worthlessness (Leary, Tambor, Terdal, &Downs, 1995; Lind & Tyler, 1988).

    Applied to abusive supervision in teams, the group-value modelsuggests that individualized abuse should decrease a membersperceived self-worth or standing in the team. Capturing this notionis OBSE (Gardner & Pierce, 1998), defined as an individualsevaluation of his or her personal adequacy and worthiness as anorganizational member (Gardner & Pierce, 1998, p. 50). Substan-tial empirical evidence suggests that favorable leader treatmentfor example, managerial respect (Pierce et al., 1989), charismaticleadership (Kark, Shamir, & Chen, 2003), and leadermemberexchange (Heck, Bedeian, & Day, 2005)relates positively tomember OBSE. On the flipside, indicators of unfavorable leadertreatment reflected in abusive supervisionthat is, targeted publicridicule, rudeness, coercion, and tantrums (Tepper, 2000)likelyreduces member OBSE. Supporting these arguments, preliminaryevidence has shown abusive supervision and poor interpersonaltreatment to lower context-specific self-esteem (Burton &Hoobler, 2006; Ferris, Spence, Brown, & Heller, 2012; Rafferty &Restubog, 2011). Although these studies were conducted amongindividuals outside team settings, we expect the effects of individ-ualized abuse on member OBSE to be strengthened in team con-texts, where social identity processes are heightened. Thus, asrepresenting a form of targeted devaluation of individual members,we propose that

    Figure 1. Theoretical model. H hypothesis.

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1076 FARH AND CHEN

    dHighlight

    dHighlight

  • Hypothesis 1: Individual-level abusive supervision is nega-tively related to member OBSE.

    Moderating Effect of Team-LevelAbusive Supervision

    An underlying assumption of the group-value model is thatleader treatment is differentiated across members, which in turncontributes variance in perceived self-worth within groups (Tyler,1989). Differential treatment, however, creates the opportunity formembers to engage in social comparisons, which can strengthen orweaken members self-concept evaluations following individual-ized treatment (Thau, Trster, Aquino, Pillutla, & De Cremer,2013). The relationship between individual-level abuse and mem-ber OBSE thus depends in part on how the leader is perceived tobehave toward comparable others in the team. Capturing thisnotion, Festingers (1954) social comparison theory is grounded intwo assumptions: (a) people are motivated to make assessments oftheir own capabilities, attitudes, and self-worth, and (b) theseself-assessments are driven by comparisons with similar others(e.g., fellow team members). In line with these assumptions, weexpect social comparisons to occur in team settings affected byabusive supervision due to the availability of comparable othersand the uncertainty and need for sense making associated withabuse (Priesemuth et al., 2013).

    Accordingly, we propose that member OBSE is shaped notmerely by experiences of individualized abuse, but also by theabusive experiences of members of the team in general. In teamsettings, high team-level abusive supervision describes a situationwhere, on average, team members feel that they have been per-sonally victimized by the team leader. A recent article on collegefootball coaches made note of how teams can differ in regard to theaverage team-level abuse they experience, where some coacheshurl abusive comments at all members of the team (with somemembers receiving more abuse than others) and other coaches usefew abusive comments at all (Adelson, 2012). Importantly, abusivesupervision is not a private affair in team settings. In fact, severalbehavioral items composing abusefor example, my supervisorputs me down in front of others, makes negative commentsabout me to others, and does not allow me to interact with mycoworkers (Tepper, 2000)explicitly invoke a larger audience.Further, because of the enhanced transparency afforded by inter-dependent work settings, even abusive behaviors enacted in privateare likely to become known to other members (Priesemuth et al.,2013). Team-level abusive supervision can thus be thought of as aproperty of the team that factors into the social comparison processof individual team members.

    According to social comparison theory (Festinger, 1954), onesself-evaluations are lowered or enhanced depending on how fa-vorably or unfavorably his or her experiences compare with therelevant experiences of referent others. In the case of abusivesupervision, the average level of abuse in the team constitutes anappropriate comparison point for any given member of the team.If, on average, others experiences of abuse are low, the focalmember perceiving high individualized abuse will feel singledout, and the negative effects of individual-level abuse on OBSEwill be exacerbated. Because social comparison processes causethe focal member to realize that there is an alternative reality inwhich the leader is less abusive toward members and that less

    abusive reality is not extended to the self (Duffy, Ganster, et al.,2006), the focal member makes increasingly negative attributionsfrom the leaders treatment about his or her self-concept and statusin the team (Bowling & Beehr, 2006). Thus, above and beyond theabsolute effects of individual-level abuse, when ones individual-ized abuse is high and team-level abuse is low, the negative impactof individual-level abusive supervision on OBSE is strengthened.

    In contrast, if the average experience of abuse among membersis high, individual-level abuse will hurt the focal members OBSEto a lesser extent. Perceiving that one is treated similarly poorly toothers in the team leads to neither favorable nor unfavorable socialcomparisons, because the focal member sees no alternative realityin which the leader treats members less abusively (Duffy, Ganster,et al., 2006). Being included in the abuse even when individual-level abusive supervision is high can cause the focal member torealize that he or she is not worse off than others, counteract thefocal members inferred low standing relative to others in the team,and effectively mitigate the negative linkage between individual-level abuse and member OBSE. These arguments are consistentwith research showing that negative attitudinal and behavioralreactions to supervisor abusive treatment were weaker when ag-gregate levels of abuse were high in the social context, andstronger when a correspondingly low level of abuse was found inthe social context (reflecting the singled-out-for-abuse effect;Duffy, Ganster, et al., 2006; Duffy, Shaw, et al., 2006; Peng et al.,2013). Thus, we expect the impact of individual-level abuse onOBSE to be moderated by how much abuse is experienced by theteam as whole.

    Hypothesis 2: The negative relationship between individual-level abusive supervision and member OBSE is more negativewhen team-level abusive supervision is low and less negativewhen team-level abusive supervision is high.

    Team Consequences of Abusive SupervisionBeyond serving as a shared reference point affecting the social

    comparison processes of individual members, team-level abusealso constitutes an ambient, team stimulus (Chan, 1998) thatinfluences how both abused and nonabused members of the teamshould interact to achieve collective goals (Ilgen, Hollenbeck,Johnson, & Jundt, 2005; Zaccaro et al., 2001). The impact ofteam-level abuse on team processes can be assessed using sociallearning theory. Although social learning theory (Bandura, 1971)has been utilized to explain the modeling of abusive supervisionbetween higher and lower level managers (Lian, Ferris, & Brown,2012; Liu, Liao, & Loi, 2012; Mawritz et al., 2012), here we arguethat team-level abuse produces interpersonal difficulties that im-pair the social functioning of the team.

    According to social learning theory, individuals intentionally orunintentionally acquire social behavior either through direct expe-rience or by observing behavioral models (Bandura, 1971). Inparticular, emulation of behavior is more likely to occur whenattention (i.e., the observer attends to the model of the behavior)and retention is high (i.e., the observer remembers the observedbehavior). Abusive behavior enacted by the team leader serves asa salient example for how members of the team are and ought tobe treated on an interpersonal level (Salancik & Pfeffer, 1978).Being exposed to high team abuse can cause members to learn

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1077ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

  • negative, devaluing behaviors from the leader and treat fellowmembers with aggression and hostility. Particularly in a contextwhere multiple members of the team are abused (i.e., team-levelabusive supervision is high), the leaders negative treatment acrossmembers creates a strong interaction norm that is further rein-forced as members emulate devaluing behaviors among them-selves.

    These social learning mechanisms may also be actively perpet-uated as more and more members of the team are personallyabused. Victims of abuse experience anger, frustration, and adesire to retaliate against the abuser (Lian et al., 2012). However,because of power differentials between leaders and members,abused victims may choose instead to act out their aggression onother team members (Marcus-Newhall, Pedersen, Carlson, &Miller, 2000; M. S. Mitchell & Ambrose, 2007). Abused membersmay in fact direct their hostilities toward less abused members ofthe team out of resentment, causing negative interpersonal inter-actions to spread among both abused and nonabused members ofthe team. It is also unlikely that abused members will unite withfellow victims to break the cycle of abuse (Sherif, Harvey, White,Hood, & Sherif, 1961) because they feel less able to protectthemselves or help others to resist an abusive leader. Even thosewho experience less abuse may find it risky to side with abusedteam members, as doing so may invite abuse upon themselves.This lack of motivation among both more and less abused mem-bers to go against the tide increases the likelihood that sociallearning mechanisms will prevail (Glomb & Liao, 2003), creatingwidespread relationship tension among team members.

    When interactions among team members are characterized bypoor treatment, negative emotionality, and interpersonal attacks,the team experiences high team relationship conflict (Jehn, 1995).Sometimes referred to as affective or emotional conflict, teamrelationship conflict reflects collective interpersonal hostilitiesamong members that are nonproductive and unrelated to the teamtask (De Dreu & Weingart, 2003; de Wit, Greer, & Jehn, 2012).Team relationship conflict is found to emerge from negative teamexperiences (Jehn & Bendersky, 2003), such as widespread ha-rassment (Raver & Gelfand, 2006) or the distrust and devaluationof fellow members (Curseu & Schruijer, 2010; Jehn & Mannix,2001; Simons & Peterson, 2000). As a negative ambient influence,we expect team-level abusive supervision will increase team rela-tionship conflict. Indirectly supporting these arguments, some ev-idence suggests that group-level abuse increases collective inter-personal deviance (Mawritz et al., 2012).

    Hypothesis 3: Team-level abusive supervision is positivelyrelated to team relationship conflict.

    Consequences for Members Responsesto the Team

    Team members contribute to team effectiveness by proactivelyexpressing constructive changes to status quo (i.e., voice), promot-ing cooperation within the team (i.e., team-role performance), andintending to remain in the team and ensure its longevity (i.e., lowturnover intentions; G. Chen et al., 2011). Research has shown thatsuch team-benefiting behaviors and attitudes are higher whenmembers feel sufficiently influential in the social context to con-tribute in these ways (Liang, Farh, & Farh, 2012) and are suffi-

    ciently motivated to help the team (G. Chen & Gogus, 2008).When one or more of these conditions are not met, members areless likely to exhibit these behaviors or attitudes. We argue thatthrough lowered member OBSE and increased team relationshipconflict, individual- and team-level abuse will uniquely reducemembers voice and team-role performance, while increasing theirturnover intentions.

    OBSE plays an important role in driving employee motivation,working-related attitudes, and behavior (Pierce & Gardner, 2004).Because individuals act in ways that are consistent with theirself-views (Korman, 1970; Swann, 1997), individuals with higherOBSE are more likely to engage in behaviors and possess attitudestoward the team that align with their feelings of competency andvalue in the group (Pierce et al., 1989). Members with higherOBSE will engage in more team-focused contributions and be lesslikely to leave the team because they believe they have the capa-bility and influence to impact the team in positive ways. Incontrast, members with lower OBSE will refrain from doing so inorder to align their contributions to the level of their self-views.Indeed, empirical research has shown positive influences of mem-ber OBSE on voice (Liang et al., 2012) and citizenship behaviors(Van Dyne & Pierce, 2004) and negative influences on memberturnover intentions (Gardner & Pierce, 1998). As such, we alsoexpect member OBSE to transmit the effects of abusive supervi-sion onto members team-relevant responses. This mediation path-way is consistent with the group-value model, which theorizes aclear link between the quality of leader treatment, members self-relevant judgments, and members willingness to exhibit group-benefiting behaviors and attitudes (Smith & Tyler, 1997). Further,because member OBSE is hypothesized to be the joint function ofindividual- and team-level abusive supervision, we expect OBSEto transmit these mediated effects more strongly when individualabuse is high and team abuse is low.

    Hypothesis 4: Individual member OBSE is positively relatedto member (a) voice and (b) team-role performance and neg-atively related to member (c) turnover intentions.Hypothesis 5: The indirect effects of individual-level abusivesupervision on member (a) voice, (b) team-role performance,and (c) turnover intentions (via OBSE) are moderated byteam-level abusive supervision, such that these mediationpathways are stronger when team-level abusive supervision islow and weaker when team-level abusive supervision is high.

    Whereas member OBSE promotes team-relevant responses pri-marily out of self-consistency mechanisms, we argue that teamrelationship conflict reduces members contributions as a demoti-vating social context. Prior research has shown teams with highrelationship conflict to experience negative emotionality (Jehn,1997), a decreased willingness to work collaboratively for thegood of the team (Jehn, Greer, Levine, & Szulanski, 2008), andlowered interpersonal trust among team members (Rispens,Greer, & Jehn, 2007). The inherently demotivating nature of acontext marked by relationship conflict likely reduces membersliking of other members, desire to behave in ways that benefit thecollective, and their willingness to remain in the team (Jehn, 1995).Research has also demonstrated a negative relationship betweenteam relationship conflict and individual helping (De Dreu & VanVianen, 2001; Ng & Van Dyne, 2005), a behavior akin to team-

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1078 FARH AND CHEN

    dHighlight

    dHighlight

  • role performance. Likewise, because hostile interactions in theteam make it unlikely that ones ideas will translate into increasedteam effectiveness, members may believe it to be futile to bring upsuggestions or concerns (Burris, Detert, & Chiaburu, 2008), likelydemotivating voice. Finally, relationship conflict represents a stateof interpersonal disintegration that weakens individual memberssocial embeddedness in the team, leading to higher turnover in-tentions (T. R. Mitchell, Holtom, Lee, Sablynski, & Erez, 2001).Supporting these arguments, prior research has shown a negativeeffect of team relationship conflict on individual team membersinnovative and teamwork behaviors and a positive effect on theirturnover intentions (G. Chen et al., 2011). Also indirectly support-ing our arguments, research has linked higher relationship conflictto members withdrawals of effort from the task (Jehn, 1995) andlowered team creativity (Farh, Lee, & Farh, 2010) and teamfunctioning (De Dreu & Van Vianen, 2001; Jehn & Mannix, 2001;Rau, 2005). Thus, we expect team relationship conflict to exert across-level negative effect on members contributions to the team,and transmit the negative effects of team abuse onto membersteam-relevant behaviors and attitudes.

    Hypothesis 6: Team relationship conflict is negatively relatedto member (a) voice and (b) team-role performance and pos-itively related to member (c) turnover intentions.Hypothesis 7: The indirect effects of team-level abusive su-pervision on member (a) voice, (b) team-role performance,and (c) turnover intentions are mediated by team relationshipconflict.

    Overview of StudiesWe conducted two studies: a field study involving intact teams

    in actual organizations to establish external validity and a scenarioexperiment to provide a constructive replication of our fieldstudys findings (Lykken, 1968), establish causality of our theory,and control for other potential mediators. We employed a scenarioexperiment because prior research utilizing such a design hasreliably captured the influence of leader treatment on individualsgroup-oriented attitudes and behaviors (e.g., De Cremer & vanKnippenberg, 2002), intended deviance (e.g., D. M. Mayer, Thau,Workman, Van Dijke, & De Cremer, 2012), and self-concept (e.g.,De Cremer, van Knippenberg, van Knippenberg, Mullenders, &Stinglhamber, 2005). Results of high realism scenario experimentshave also been successfully replicated in field settings, demon-strating high psychological fidelity across cultures and continents(G. Chen et al., 2011). Finally, because the field study examinedthe effects of individual- and team-level abusive supervision ononly one dependent variable (voice), the scenario experiment wascritical for examining our proposed effects on all three dependentvariables (voice, team-role performance, and turnover intentions).Thus, including both the field and scenario study in our investi-gation enabled us to test our entire theoretical model and benefitfrom the strengths of each methodological approach.

    Study 1

    MethodSample, design, and procedures. Study 1 involved 295 em-

    ployees from 51 teams across 10 firms in China. Organizational

    sponsors enabled us to identify true teams (i.e., members reportingto the same team leader and worked interdependently to achieveshared goals). Participants were recruited through the human re-lations office and informed that the purpose of the study was toexamine the leadership styles and group dynamics within eachfirm. Participants were also assured of confidentiality through ourdata collection procedure (see below). Informed consent was ob-tained prior to the commencement of the data collection. Exclud-ing cases with missing data, our overall response rate was 95% forindividual participants (N 280) and 98% for teams (N 50).The high response rate was facilitated by the organizational spon-sors enthusiastic support to participate in the study, as well astheir willingness to allow us to administer the survey in personduring one of the teams monthly meetings, a time when nearly allteam members and leaders would be available. The teams in oursample were diverse, performing functions such as technical sup-port, internal monitoring, research and development, and customerservice. Team size ranged from four to seven members (M 5.60). The average team member was 35.33 years old (SD 5.68)and had worked in their teams for 39.74 months (SD 40.58).Fifty-six percent were female, 7% held high school or associateddegrees, 66% held bachelor degrees, and the rest held mastersdegrees or higher.

    Team leaders distributed surveys to each member of their re-spective teams, and leaders and members responses werematched via numerical codes. Given the sensitive nature of thestudy and to ensure confidentiality, responses were sealed inenvelopes and returned directly to a member of the research team,who was on site during the data collection process. Members ratedtheir personal experiences of abusive supervision, OBSE, and teamrelationship conflict. As control variables, members also ratedtheir leaders empowering leadership, their own power distanceorientation and collectivism, and their teams task conflict. Teamleaders rated each members voice behavior. As another controlvariable, external leaders (immediate supervisors of the teamleader) rated each teams task performance.

    Measures. All survey instruments were administered in Chi-nese following the translation and back-translation proceduresoutlined in Brislin (1990). Aside from the conflict measures, allitems used 7-point Likert-type response categories anchored at 1(strongly disagree) to 7 (strongly agree). Members rated teamrelationship and task conflict with anchors ranging from never (1)to always (7). Where appropriate, we note precedence for the useof these scales in a Chinese context. Descriptive statistics andcorrelations among our studys variables are shown in Table 1.

    Individual- and team-level abusive supervision. Abusive su-pervision was assessed with Aryee et al.s (2007) 10-item versionof Teppers (2000) original scale (e.g., My supervisor blames meto save himself/herself embarrassment; .91). To captureindividualized experiences of abuse, we group mean centered eachmembers abusive supervision score. Additionally, members dem-onstrated sufficient intermember agreement and reliability to jus-tify aggregation to the team level (rwg(j) with a uniform expectedvariance distribution .94; ICC(1) .15, F(49, 231) 1.81, p .01; ICC(2) .45; LeBreton & Senter, 2008). Therefore, team-level abusive supervision was operationalized as the mean ofmembers responses.

    Individual team member OBSE. Reflecting ones self-concept in the team, 10 items were adapted to the team context

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1079ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

  • Tabl

    e1

    Mea

    ns,S

    tand

    ard

    Dev

    iatio

    ns,a

    nd

    Corr

    elat

    ions

    Amon

    gVa

    riabl

    esin

    Stud

    y1

    Var

    iabl

    eM

    SD1

    23

    45

    67

    89

    1011

    1213

    1415

    1617

    18

    Leve

    l1(N

    280)

    1.Te

    amm

    embe

    rvoic

    e5.

    051.

    10(.9

    5)2.

    Indi

    vidu

    al-le

    velA

    S2.

    141.

    01

    .16

    (.91)

    3.Te

    am-le

    velA

    Sa2.

    140.

    53

    .22

    .52

    (.91)

    4.Te

    amm

    embe

    rOBS

    E5.

    340.

    95.21

    .36

    .21

    (.94)

    5.Te

    am-le

    velO

    BSEa

    5.34

    0.46

    .19

    .22

    .42

    .49

    (.94)

    6.Te

    amre

    latio

    nshi

    pco

    nfli

    ct2.

    630.

    81

    .21

    .47

    .38

    .18

    .13

    (.8

    6)7.

    Team

    rela

    tions

    hip

    confli

    cta

    2.63

    0.50

    .32

    .32

    .62

    .10

    .20

    .62

    (.86)

    8.Te

    amm

    embe

    rgen

    derb

    0.56

    0.50

    .16

    .09

    .03

    .13

    .11

    .10

    .20

    9.Te

    amm

    embe

    rage

    35.3

    35.

    68.05

    .18

    .10

    .16

    .15

    .11

    .05

    .08

    10.T

    eam

    mem

    bere

    duca

    tionc

    2.18

    0.55

    .04

    .01

    .07

    .07

    .04

    .05

    .02

    .13

    .05

    11.P

    ower

    dista

    nce

    orie

    ntat

    ion

    2.71

    0.84

    .01

    .11

    .08

    .03

    .06

    .04

    .02

    .03

    .07

    .05

    (.71)

    12.C

    olle

    ctiv

    ism5.

    180.

    90.03

    .24

    .24

    .24

    .22

    .23

    .24

    .02

    .06

    .04

    .06

    (.78)

    13.T

    eam

    task

    confli

    ct3.

    530.

    74

    .05

    .20

    .24

    .06

    .07

    .46

    .31

    .01

    .01

    .13

    .12

    .10

    (.74)

    14.T

    eam

    task

    confli

    cta

    3.53

    0.40

    .07

    .23

    .43

    .06

    .12

    .35

    .56

    .05

    .01

    .08

    .01

    .08

    .55

    (.74)

    15.T

    eam

    -leve

    lEL

    5.44

    0.49

    .15

    .45

    .31

    .46

    .28

    .37

    .30

    .03

    .18

    .04

    .08

    .33

    .10

    .10

    (.94)

    16.T

    eam

    -leve

    lELa

    5.78

    0.77

    .30

    .30

    .58

    .26

    .53

    .34

    .56

    .16

    .13

    .07

    .03

    .26

    .10

    .18

    .53

    (.94)

    17.T

    eam

    task

    perfo

    rman

    cea

    5.78

    0.77

    .10

    .09

    .19

    .12

    .23

    .01

    .04

    .08

    .11

    .07

    .01

    .05

    .01

    .02

    .19

    .37

    (.82)

    18.T

    eam

    size

    5.85

    0.55

    .04

    .17

    .34

    .04

    .07

    .02

    .05

    .05

    .09

    .02

    .12

    .03

    .09

    .16

    .04

    .10

    .20

    Leve

    l2(N

    50)

    1.Te

    am-le

    velA

    S

    2.Te

    amre

    latio

    nshi

    pco

    nfli

    ct.62

    3.Te

    am-le

    velO

    BSE

    .43

    .20

    4.Te

    amta

    skco

    nfli

    ct.42

    .54

    .11

    5.Te

    am-le

    velE

    L

    .59

    .57

    .53

    .17

    6.Te

    amta

    skpe

    rform

    ance

    .21

    .07

    .22

    .02

    .37

    7.Te

    amsiz

    e

    .36

    .04

    .08

    .16

    .09

    .24

    Note

    .R

    elia

    bilit

    yes

    timat

    es(co

    effici

    enta

    lpha

    )are

    on

    the

    diag

    onal

    .AS

    abus

    ive

    supe

    rvisi

    on;O

    BSE

    org

    aniz

    atio

    n-ba

    sed

    self-

    este

    em;E

    L

    empo

    wer

    ing

    lead

    ersh

    ip.

    aTe

    amsc

    ore

    sas

    signe

    dto

    each

    team

    mem

    ber.

    bG

    ende

    r:0

    fem

    ale,

    1

    mal

    e.c

    Educ

    atio

    nle

    vel:

    1

    asso

    ciat

    e,2

    bach

    elor

    ,3

    mas

    ter

    s,4

    PhD

    .

    p

    .05

    .

    p

    .01

    (two-t

    ailed

    ).

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1080 FARH AND CHEN

  • from the OBSE scale developed by Pierce et al. (1989) ( .94;In this team, I can make a difference). This scale has also shownsufficient validity and reliability in the Chinese context (e.g., Z. X.Chen & Aryee, 2007).

    Team relationship conflict. Members rated their teams rela-tionship conflict using a four-item scale (Pelled, Eisenhardt, &Xin, 1999), shown to be reliable and valid in the Chinese context(Farh et al., 2010; e.g., Excluding your team leader, how muchtension is there among the members of your team?). Supportingthe aggregation of members ratings, median intermember agree-ment (rwg(j)) across teams was .95; ICC(1) was .26, F(49, 231) 3.01, p .01; and ICC(2) was .67. Additionally, because thesecoefficients were comparable to measures of relationship conflictin prior research (cf. G. Chen et al., 2011), we averaged membersresponses.

    Individual team member voice. The team leader rated eachmembers voice behavior using the 10-item scale constructed byLiang et al. (2012) in the Chinese context, which assessed bothprohibitive voice ( .93; e.g., Calling management attention todysfunctional activities) and promotive voice ( .95; e.g.,Speaking up with ideas for new projects that might benefit theteam). The intercorrelation between the two dimensions was high(r .75, p .01; .95). A confirmatory factor analysis (CFA)supported a two-factor model with a second-order factor ((33)2 131.94, root-mean-square error of approximation [RMSEA] .10,comparative fit index [CFI] .99, TuckerLewis index [TLI] .98, root-mean-square residual [SRMR] .03),1 which fit the databetter than an alternative model with the two types of voice asseparate indicators ((2)2 144.46, p .01, RMSEA .15,CFI .95). Because our interest was on overall constructive voicerather than its subdimensions (Maynes & Podsakoff, 2014), weaveraged them into a single score.

    Control variables. To provide a rigorous test of our model, wecontrolled for several individual and team factors known toinfluence member voice.2 Following prior voice research (e.g.,LePine & Van Dyne, 1998; Liang et al., 2012), we controlled forteam member age (in years), gender (1 male, 2 female), andeducational level (1 associate, 2 bachelor, 3 master, and4 PhD). Additionally, given the potential influence of culturalvalues on voice and team behavior (Botero & Van Dyne, 2009;Liang, Huang, & Chen, 2013), we controlled for members powerdistance orientation (Dorfman & Howell, 1988; Managers shouldseldom ask for the opinions of employees; .71) and collec-tivism (Dorfman & Howell, 1988; Group welfare is more impor-tant than individual needs; .78).

    At the team level, we controlled for team size, as larger teamstend to experience more conflict than smaller teams (Farh et al.,2010). We also controlled for task conflict (Pelled et al., 1999) dueto the high correlation between task and relationship conflict (DeDreu & Weingart, 2003) and the potential conceptual overlapbetween task conflict and voice behavior. Task conflict was as-sessed with the four-item scale by Pelled et al. (1999) ( .74)and was aggregated to the team level (median rwg(j) across teams.94; ICC(1) .17, F(49, 231) 2.18, p .01; ICC(2) .55).Further, to demonstrate abusive supervision exerted influence onour outcomes above and beyond the effects of supportive leader-ship behaviors (G. Chen et al., 2011), we controlled for the effectsof empowering leadership at both the individual and team levels.Team members rated their leader on a 14-item empowering lead-

    ership scale developed by Kirkman and Rosen (1999) ( .94).We used group-centered ratings to reflect individual-level empow-ering leadership and averaged members ratings to capture em-powering leadership at the team level (median rwg(j) acrossteams .96, ICC(1) .17, F(49, 231) 1.92, p .01; ICC(2) .53). Additionally, to assess the unique effects of team relationshipconflict on voice above and beyond aggregations of individualmembers OBSE, we controlled for team-level OBSE (medianrwg(j) .96, ICC(1) .08, F(49, 231) 1.51, p .05; ICC(2) .34). Finally, we controlled for team performance because poorperformance is positively associated with abusive supervision(Tepper, Moss, & Duffy, 2011) and relationship conflict (Jehn &Mannix, 2001).We relied on external leaders (immediate supervi-sors of the team leader) to provide third-party ratings of teamperformance using the scale constructed by Carson, Tesluk, andMarrone (2007) ( .82, e.g., This team is effective in meetingmy expectations in terms of the quality of their final output).

    Analyses and ResultsWe conducted a series of CFAs to examine the discriminant

    validity of team-member-rated constructs in our model. Althoughseveral constructs were ultimately aggregated and analyzed at theteam level, we examined them at the individual level to establishtheir discriminant validity. A four-factor baseline model composedof abusive supervision, team relationship conflict, task conflict,and OBSE fit the data well ((344)2 1425.41, RMSEA .10,CFI .93; TLI .92, SRMR .06). This baseline model fit thedata better than alternative models in which (a) abusive supervi-sion and team relationship conflict were set to equal 1 ((1)2 477.10, p .01, RMSEA .13, CFI .90) and (b) task andrelationship conflict were set to equal 1 ((1)2 257.95, p .01,RMSEA .12, CFI .92). Together, these analyses demonstratedthe discriminant validity of the member-rated variables in ourmodel. Further, because ratings of members voice were nestedwithin leaders, we utilized two-level hierarchical linear modeling(Raudenbush & Bryk, 2002) to address the violation of noninde-pendence of observations and to examine cross-level interactioneffects. We also used a sample bootstrapping approach that pro-vided bias-corrected confidence intervals to test the mediatingeffects of OBSE and relationship conflict (Hayes, 2013). Ourresults are summarized in Tables 2 and 3.

    Hypothesis testing. Hypothesis 1 proposed a negative rela-tionship between individual-level abusive supervision and memberOBSE. As shown in Model 3 of Table 2, individual-level abusewas negatively related to OBSE ( .20, p .01), supportingHypothesis 1. Hypothesis 2 predicted a cross-level interactioneffect between team- and individual-level abuse on members

    1 We acknowledge that our RMSEA indices here and elsewhere arehigher than ideal. However, in view of other fit indices, we have someindication that our theorized models exhibited acceptable fit. In this case,the second-order factor model fit the data better than an alternative modelwith two separate factors for voice. Hence, in addition to our theoreticalreasons for examining overall voice, we had sufficient empirical reason tocollapse the two types of voice into one factor.

    2 Our results reached the same levels of significance and were in thesame direction regardless of whether or not we included the controlvariables. Results without the control variables are available from theauthors upon request.

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1081ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

  • OBSE. As shown in Model 5 and Figure 2, this interaction effectwas significant ( .11, p .05). The negative relationshipbetween individual abuse and OBSE was stronger when team levelabuse was low (simple slope .32, p .05, supporting thesingled-out effect) but was weaker when team level abuse washigh (simple slope .08, ns), demonstrating support for

    Hypothesis 2. Hypothesis 3 predicted a positive relationship be-tween team-level abuse and team relationship conflict. As shownin Model 2 of Table 2, even after controlling for team performanceand empowering leadership, team-level abuse related positively toteam relationship conflict ( .27, p .01), supporting Hypoth-esis 3. Hypothesis 4a stated that individual members OBSE

    Table 2Hierarchical Linear Modeling Regression Results of Study 1

    Variable

    Team relationship conflict Team member OBSE Team member voice

    Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7

    Level 1 (N 280)Team member gendera .24 (.11) .22 (.11) .27 (.12) .07 (.14) .01 (.12)Team member age .01 (.01) .01 (.01) .02 (.01) .01 (.01) .02 (.01)Team member educationb .03 (.08) .07 (.08) .07 (.08) .05 (.10) .03 (.09)Power distance orientation .01 (.06) .01 (.06) .06 (.07) .02 (.07) .00 (.07)Collectivism .10 (.06) .10 (.06) .13 (.06) .09 (.06) .11 (.05)Individual-level EL .37 (.08) .45 (.08) .02 (.07) .08 (.07)Individual-level AS .20 (.05) .38 (.06) .11 (.06) .07 (.07)Team member OBSE .18 (.06)

    Level 2 (N 50)Team size .02 (.09) .09 (.09) .16 (.13) .08 (.13) .04 (.13) .05 (.23) .01 (.15)Team task performance .08 (.07) .07 (.07) .01 (.07) .00 (.07) .04 (.08) .06 (.14) .03 (.09)Team task conflict .56 (.13) .46 (.13) .28 (.14) .40 (.15) .42 (.16) .11 (.27) .32 (.26)Team-level EL .56 (.11) .40 (.13) .43 (.11) .25 (.12) .23 (.15) .55 (.27) .31 (.20)Team-level AS .27 (.13) .16 (.06) .14 (.07) .16 (.14) .02 (.11)Team-level OBSE .19 (.25)Team relationship conflict .29 (.13)

    Team-Level AS Individual-Level AS .11 (.04)Note. Standard errors are in parentheses. OBSE organization-based self-esteem; EL empowering leadership; AS abusive supervision.a Gender: 0 female, 1 male. b Education level: 1 associate, 2 bachelor, 3 masters, 4 PhD. p .05. p .01 (two-tailed).

    Table 3Bootstrapping Results: Indirect Effects of Team-Level Abusive Supervision (AS) and Individual-Level AS on Outcome Variables

    Indirect effect

    Team member voiceTeam member team-role

    performanceTeam member

    turnover intentions

    95% CI 95% CI 95% CI

    Study 1Individual-level AS OBSE DV

    Low .14 [.32, .03]M .08 [.17, .02]High .03 [.10, .02]

    Team-level AS Relationship conflict DVM .39 [.61, .18]

    Study 2Individual-level AS (manipulation) OBSE DV

    Low .35 [.69, .14] .45 [.74, .22] .72 [.39, 1.11]M .18 [.35, .07] .26 [.45, .10] .42 [.24, .67]High .06 [.23, .07] .13 [.34, .02] .23 [.03, .50]

    Team-level AS (manipulation) Relationship conflict DVM .09 [.19, .02] .12 [.25, .03] .20 [.05, .39]

    Individual-level AS (perception) OBSE DVLow .14 [.29, .00] .23 [.46, .05] .36 [.02, .70]M .10 [.18, .03] .14 [.26, .04] .27 [.08, .41]High .07 [.15, .01] .07 [.24, .02] .19 [.03, .36]

    Team-level AS (perception) Relationship conflict DVM .07 [.13, .03] .09 [.17, .05] .14 [.07, .24]

    Note. CI confidence interval; OBSE organization-based self-esteem; DV dependent variable. p .01 (two-tailed).

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1082 FARH AND CHEN

    dHighlight

    dHighlight

    dHighlight

  • should positively relate to their voice behavior, whereas Hypoth-esis 6a proposed a negative linkage between team relationshipconflict and member voice. Additionally, according to our theo-retical model, the two effects should occur simultaneously. Insupport of this prediction, OBSE related positively to membervoice ( .18, p .01) and team relationship conflict wasnegatively related to voice ( .29, p .05) in the same model(Model 7 of Table 2), even after controlling for team OBSE at theteam level. Thus, Hypotheses 4a and 6a received full support.

    Hypothesis 5a predicted the negative indirect relationship be-tween individual abuse and member voice (via OBSE) would bemoderated by team-level abuse, such that the mediation pathway isstronger when team-level abuse is low and weaker when team-level abuse is high. As shown in Table 3, results showed that thisindirect effect was significant when team-level abuse was low( .14, p .01, 99% CI [.32, .03]). When team-levelabuse was high, however, the indirect effect through OBSE wasnonsignificant ( .03, ns, 95% CI [.10, .02]). Thus, Hypoth-esis 5a received support. Hypothesis 7a predicted that team rela-tionship conflict would mediate the negative relationship betweenteam-level abuse and members voice. As shown in Model 7 ofTable 2, after including team relationship conflict, team-levelabuse did not relate to individual members voice ( .02, ns).In Table 3, this indirect effect of team-level abuse on team membervoice (via relationship conflict) was significant ( .39, p .01, 99% CI [.61, .18]). Moreover, it was team relationshipconflict ( .29, p .05) rather than team-level OBSE ( .19, ns) that mediated the negative impact of team-level abuse onmember voice. This finding suggests that team-level abuse affectsmember outcomes above and beyond aggregated consequences ofindividual abuse, and demonstrates that team- and individual-levelabuse influences member voice via different pathways. Thus,Hypothesis 7a was supported.

    DiscussionThe results of Study 1 provided support for our multilevel model

    and hypotheses pertaining to member voice. Further, in line withthe notion that negative events affect individuals more powerfully

    and saliently than positive events (Baumeister, Bratslavsky, Finke-nauer, & Vohs, 2001; Taylor, 1991), we demonstrated that abusivesupervision is distinct from and operates in team settings indepen-dently of supportive forms of leadership such as empoweringleadership. In Study 2, we sought to replicate these findings in acontrolled environment to increase internal validity, rule out alter-native mechanisms identified in other studies not measured inStudy 1 (e.g., interpersonal justice perceptions, team identification,collective efficacy; Priesemuth et al., 2013; Tepper, 2000), andexamine whether member OBSE and team relationship conflictmediated the influences of abusive supervision on members voice,team-role performance, and turnover intentions.

    Study 2

    MethodSample, design, and procedure. Study 2 involved 276 un-

    dergraduate business majors (49.6% women and 50.4% men; meanage 20.94 years, SD 1.41) enrolled at a large midwesternuniversity in the United States. Participants voluntarily partici-pated in the study for extra credit and were randomly assigned toa 2 (individual-level abusive supervision: high vs. low) 2(team-level abusive supervision: high vs. low) between-subjectsdesign. Upon arriving to the laboratory and after completing anunrelated task on stock price forecasting, each participant wasassigned to a computer terminal and was presented with thescenario experiment through an online survey. Participants werefirst assessed on several individual differences in cultural orienta-tion and then read one of four scenarios, each of which instructedparticipants to imagine they were members of a critical four-member undergraduate task force in the business school (adaptedfrom the scenario manipulation in G. Chen et al., 2011). To holdtask performance constant (as in Study 1), participants were toldthat the task force was struggling to meet deadlines and progresswas slow.

    In each scenario, participants were presented with a set of foure-mail messages sent to them by their team leader and two

    5.00

    6.00

    7.00

    Low Individual-level Abusive Supervision

    High Individual-level Abusive Supervision

    Low Team-level Abusive Supervision

    High Team-level Abusive SupervisionO

    rgan

    izat

    ion-

    base

    d se

    lf-es

    t

    Figure 2. Plot of the interaction between individual- and team-level abusive supervision on individualmembers organization-based self-esteem.

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1083ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

  • additional task force members. The e-mail contents distinguishedthe four conditions by containing our manipulation of high versuslow individual-level abuse and high versus low team-level abuse.The complete text of the manipulations appear in AppendicesAD. The order of manipulations for individual and team abusivesupervision was counterbalanced. Following the scenarios, partic-ipants completed manipulation checks, measures of OBSE andteam relationship conflict, and their intended voice, team-roleperformance, and turnover intentions. Participants also completedmeasures of alternative mediators as controls.

    Manipulations.Individual-level abusive supervision. Participants received a

    personalized e-mail from the team leader. In the high individualabuse condition, the leaders e-mail acknowledged the difficultiesencountered by the task force and ended with several negative andbelittling comments about the participants specific contributionsand competence. In the low individual abuse condition, partici-pants also received an e-mail acknowledging the slow progress ofthe team. However, rather than expressing a personal attack, thee-mail maintained a respectful, neutral tone. The two manipula-tions were identical in word count.

    Team-level abusive supervision. The manipulation of team-level abuse involved a single e-mail message from the team leaderaddressed to the entire team, as well as two additional messagessent from and addressed to members of the team only. In the highteam abuse condition, the leaders e-mail attributed current chal-lenges encountered to the teams failures and publically ridiculedthe input of two of the teams members. Following the leaderse-mail, the participant received an e-mail exchange among twomembers of the team that complained about how difficult theleader was to work with and recounted abusive experiences withthe leader (e.g., failing to give team members credit for theirideas, rude and condescending behavior, and giving members thesilent treatment). The leader and team member e-mails collectivelypainted a situation where the participant both observes the leaderabusing the entire team in group e-mail and hears about theindividualized abuse experienced by two other members of theteam.

    In the low team abuse condition, the leaders e-mail attributedcurrent challenges encountered by the team to mistakes made early

    on, but noted that making mistakes is part of the learning process.The e-mail encouraged the team to stay focused and committed tothe task, and ended with an update that several of the teammembers suggestions were included in the final report deliveredto the dean. The participant then received e-mails among twomembers that discussed their experiences working with the leader(e.g., the leader gave straightforward feedback, credited teammembers for their ideas, and was responsive in e-mail communi-cation). The three messages were designed to create a situationwhere the participant observes the leader interacting with the teamin a neutral tone and hears about the neutral but respectful treat-ment others received from the leader. The two manipulations wereidentical in word count.

    Measures. Descriptive statistics, internal consistency reliabil-ity coefficients, and correlations of our core measures are reportedin Table 4. Unless noted otherwise, all items were rated on a5-point Likert-type scale (1 strongly disagree, 5 stronglyagree).

    Individual team member OBSE. OBSE was assessed with thesame 10 items from Study 1. These items were from Pierce et al.s(1989) scale, adapted to the team context ( .96).

    Team member voice. Participants intended voice was as-sessed with the same 10-item scale used in Study 1, and the twodimensions were moderately correlated (r .42, p .01; .86). A CFA showed a two-factor model with a second-order factor((23)2 94.88, RMSEA .11, CFI .96, TLI .93, SRMR .09) fit the data better than an alternative model with the two typesof voice as separate indicators ((2)2 7.97, p .01, RMSEA .14, CFI .92). As in Study 1, given our focus on capturingoverall voice, we averaged the two dimensions.

    Member team-role performance. Participants reported theirintention to engage in team-role performance behaviors based onthe four-item scale from Welbourne et al. (1998) (I will respondto the needs of others in the Task Force; .88).

    Member turnover intention. Participants reported their inten-tion to leave the task force on a four-item scale from Kelloway,Gottlieb, and Barham (1999) (I will think about leaving this TaskForce; .93).

    Team relationship conflict. Team relationship conflict wasassessed with eight items that captured the extent to which the

    Table 4Means, Standard Deviations, and Correlations Among Variables in Study 2

    Variable M SD 1 2 3 4 5 6 7 8 9 10 11 12 13 14

    1. Individual AS (manipulation) 0.48 0.50 2. Team AS (manipulation) 0.50 0.50 .05 3. Individual AS (perception) 2.71 1.07 .66 .33 (.97)4. Team AS (perception) 3.08 1.15 .27 .70 .61 (.97)5. Team member OBSE 3.41 0.90 .51 .07 .60 .25 (.96)6. Team relationship conflict 2.82 0.83 .30 .29 .48 .49 .39 (.94)7. Team member voice 3.50 0.61 .22 .00 .30 .16 .34 .26 (.86)8. Team-role performance 3.91 0.69 .16 .02 .30 .16 .36 .28 .56 (.88)9. Turnover intentions 3.16 1.00 .29 .18 .43 .41 .45 .47 .22 .34 (.93)

    10. Power distance orientation 2.11 0.47 .04 .00 .00 .09 .01 .11 .09 .14 .09 (.53)11. Collectivism 3.41 0.54 .02 .14 .04 .07 .14 .05 .04 .14 .07 .07 (.71)12. Interpersonal justice 2.97 1.21 .64 .35 .88 .64 .58 .50 .25 .26 .46 .03 .04 (.96)13. Collective efficacy 2.82 0.68 .29 .33 .51 .58 .48 .62 .27 .31 .54 .05 .01 .53 (.87)14. Team identification 3.15 0.90 .17 .18 .27 .06 .54 .32 .29 .48 .39 .04 .12 .27 .37 (.83)Note. N 250. Reliability estimates (coefficient alpha) are on the diagonal. AS abusive supervision; OBSE organization-based self-esteem. p .05 (two-tailed). p .01.

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1084 FARH AND CHEN

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

  • participants interactions with fellow members of team were likelyto be characterized by negative emotion, relationship tension,unproductive personal attacks, and disagreements on interpersonalissues and personal values (e.g., Jehn, 1995; Pelled et al., 1999).Cronbachs alpha for this composite measure was .94.

    Manipulation checks. For individual-level abuse, participantsindicated how strongly they agreed with the same 10-item measureof abusive supervision utilized in Study 1 ( .97). To captureteam-level abusive supervision, we used a referent-shift adaptationof the items. Participants indicated the extent to which their teamleader exhibited abusive behavior toward the team (e.g., My taskforce leader makes negative comments about team members toothers, ridicules members of this team, tells team membersthat they are incompetent; .97). With the individual-levelabuse manipulation check as outcome, analysis of variance testsindicated a main effect for the individual-level abuse condition(mean high individual abuse 3.49, mean low individual abuse2.04), F(1, 274) 230.59, p .01, 2 .46. With the team-levelabusive supervision manipulation check as outcome, analysis ofvariance tests indicated a main effect for the team-level abusecondition (mean high team abuse 3.91, mean low team abuse 2.29), F(1, 274) 269.51, p .01, 2 .50. These resultsprovide strong evidence for the efficacy and validity of the twomanipulations featured in our scenarios.

    In addition, participants completed three questions assessing therealism of the scenario used in Chen et al. (2011) (It is realisticthat I might experience a supervisor like J. P., It is realistic thatI might experience team-events like those described above in theTask Force, and At some point during my career, I will probablyencounter a situation like the one described above; .71). Themean score across these items (M 3.89) indicated that partici-pants generally agreed the scenario was realistic. We further in-cluded several items to ensure that participants read the scenariocarefully. Participants were asked to indicate the performanceprogress of the task force. Participants who indicated that teamperformance was excellent were considered to have failed themanipulation check. After dropping these participants, as well asthose who indicated the scenario was low in realism (those scoringlower than a 3), our final sample consisted of 250 individuals(retaining 90.6% of our sample) randomly assigned across the fourconditions.

    Additional variables. Several additional variables were as-sessed as control variables. Participants responded to a four-itemmeasure of interpersonal justice (Colquitt, Conlon, Wesson, Por-ter, & Ng, 2001; My leader has treated me in a polite manner; .96), a four-item measure of team identification (Allen &Meyer, 1990; I feel a strong sense of belonging to my team; .87), and a six-item measure of team collective efficacy (Riggs,Warka, Babasa, Betancourt, & Hooker, 1994; This team hasabove average ability; .83). Participants also responded to asix-item measure of power distance orientation (Dorfman & How-ell, 1988; Managers should seldom ask for the opinions of em-ployees; .53)3 and collectivism (Dorfman & Howell, 1988;Group welfare is more important than individual needs; .71)to enable comparisons across our findings in Study 1 (a China-based sample) and Study 2 (a U.S.-based sample). Also, to assessthe different types of emotions participants might have experi-enced across conditions, we asked participants to respond to thePositive and Negative Affect Schedule (PANAS; Watson, Clark,

    & Tellegen, 1988), consisting of 10 positive (e.g., interested,excited, enthusiastic; .89) and 10 negative adjectives (e.g.,afraid, guilty, scared; .91). Participants were asked to de-scribe the way they felt after reading the scenarios.

    Analyses and ResultsWe conducted a series of CFAs to test whether the measures we

    used are distinguishable from each other. First, we ran a 12-factorbaseline model composed of team- and individual-level abusivesupervision (perceptions), relationship conflict, OBSE, voice,team-role performance, turnover intentions, interpersonal justice,team identification, collective efficacy, power distance orientation,and collectivism. Results showed that this baseline model fit thedata well ((3173)2 6055.07, RMSEA .06, CFI .96, TLI .96, SRMR .06) and fit the data better than competing modelsin which (a) team- and individual-level abuse were set to equal 1((1)2 4633.25, p .01, RMSEA .10, CFI .94), (b)team-level abuse and team relationship conflict were combined asone factor ((1)2 1711.4, p .01, RMSEA .08, CFI .95),and (c) individual-level abuse and OBSE were combined as onefactor ((1)2 3696.13, p .01, RMSEA .09, CFI .95).These analyses demonstrated the discriminant validity of the mea-sured variables in our model. To test our hypotheses, we usedhierarchical regression. Because we used scenarios to manipulateteam and individual abuse, we tested our hypotheses using bothmanipulated and perceived abuse as independent variables. Wealso used the same sample bootstrapping approach in Study 1 totest the mediating effects of OBSE and relationship conflict(Hayes, 2013). Tables 3, 5, and 6 summarize our results.

    Hypothesis testing. Hypothesis 1 predicted that individual-level abuse would be negatively related to member OBSE. Sup-porting this hypothesis, as shown in Table 5, we found that boththe manipulated ( .51, p .05) and perceived abuse( .70, p .05) negatively predicted OBSE (Model 4).Hypothesis 2 predicted that individual-level abuse would be morenegatively related to OBSE when team-level abuse was low. Asshown in Model 5, the interaction terms between individual- andteam-level abuse manipulations ( .19, p .05) and perceptions( .16, p .05) were significant. Shown in Figures 3 and 4, thenegative relationship between the individual abuse and OBSE wasstronger when team abuse is lower than higher. Simple slopesanalyses showed that manipulated individual abuse related morenegatively to OBSE when team abuse was low ( .62, p .05) than when team abuse was high ( .40, p .05).Similarly, perceptual measures of abuse related more negatively toOBSE when team abuse was low ( .90, p .05) than whenteam abuse was high ( .63, p .05), supporting Hypothesis2. Hypothesis 3 predicted that team-level abuse would positivelyrelate to team relationship conflict. Model 2 shows that the positive

    3 Because Dorfman and Howells (1988) measure of power distanceorientation was designed to assess numerous domains of supervisor-subordinate interaction across various forms of power, the measure hashistorically suffered from lower levels of reliability, often falling below the.70 cut off (including a Cronbachs alpha of .51 in the original study).Nonetheless, the measure has been shown to meaningfully capture indi-vidual differences in power distance values, and we elected to include it asa control variable in order to compare our findings across our China- andU.S.-based samples.

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1085ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

    dHighlight

    dHighlight

    dHighlight

  • impact of team abuse on relationship conflict was significant forboth manipulated ( .20, p .05) and perceived team abuse( .19, p .05), supporting Hypothesis 3.

    Hypotheses 4 and 6 proposed that member OBSE and teamrelationship conflict would simultaneously influence membersvoice, team-role performance, and turnover intentions. After con-trolling for manipulated abuse at the individual and team levels,OBSE related positively to member voice ( .21, p .05;Model 7, Table 5) and team-role performance ( .17, p .01;

    Model 2, Table 6) and negatively to turnover intentions( .32, p .01; Model 6, Table 6), while team relationshipconflict was negatively related to voice ( .22, p .05; Model7, Table 5) and team-role performance ( .27, p .01);Model 2, Table 6) and positively to turnover intentions ( .42,p .01; Model 6, Table 6). Likewise, after controlling for per-ceived abuse at the individual and team levels, OBSE relatedpositively to voice ( .17, p .05; Model 7, Table 5) andteam-role performance ( .13, p .05; Model 2, Table 6) and

    Table 5Regression Results of Study 2: Effects on Team Relationship Conflict, Member OBSE, and Member Voice

    Variable

    Team relationshipconflict Team member OBSE Team member voice

    Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9

    Manipulated abusive supervisionPower distance orientation .16 (.11) .16 (.10) .01 (.13) .02 (.10) .02 (.10) .09 (.08) .05 (.08) .11 (.08) .06 (.08)Collectivism .04 (.09) .01 (.08) .14 (.11) .13 (.09) .14 (.09) .04 (.07) .01 (.07) .01 (.07) .04 (.07)Individual AS (manipulation) .35 (.09) .51 (.10) .62 (.14) .28 (.11) .03 (.12) .09 (.10) .14 (.08)Team AS (manipulation) .20 (.09) .08 (.10) .03 (.14) .05 (.11) .05 (.11) .04 (.09) .11 (.08)Individual Team AS (manipulation) .19 (.20) .10 (.15) .03 (.15)Individual OBSE .21 (.05) .13 (.06)Team relationship conflict .22 (.06) .18 (.06)Interpersonal justice .02 (.05)Team identification .12 (.05)Collective efficacy .11 (.07)

    Perceived abusive supervisionPower distance orientation .16 (.11) .17 (.09) .00 (.13) .00 (.10) .02 (.10) .06 (.08) .04 (.08) .11 (.08) .05 (.08)Collectivism .04 (.09) .05 (.08) .13 (.11) .10 (.09) .09 (.09) .01 (.07) .00 (.07) .01 (.07) .03 (.07)Individual AS (perception) .40 (.05) .70 (.05) .77 (.06) .39 (.05) .18 (.06) .13 (.07) .12 (.05)Team AS (perception) .19 (.05) .18 (.05) .29 (.06) .15 (.05) .11 (.05) .03 (.04) .05 (.04)Individual Team AS (perception) .16 (.04) .20 (.03) .14 (.03)Individual OBSE .17 (.05) .11 (.06)Team relationship conflict .16 (.06) .16 (.06)Interpersonal justice .07 (.07)Team identification .12 (.05)Collective efficacy .08 (.08)Note. N 250. Standard errors are in parentheses. OBSE organization-based self-esteem; AS abusive supervision. p .05 (two-tailed).

    1.50

    2.50

    Low Individual-level Abusive High Individual-level Abusive

    Low Team-level Abusive Supervision (Manipulation)High Team-level Abusive Supervision (Manipulation)

    Low Individual-level Abusive Supervision (Manipulation)

    High Individual-level Abusive Supervision (Manipulation)

    Mem

    ber O

    BSE

    Figure 3. Plot of the interaction between individual- and team-level abusive supervision manipulations onindividual members organization-based self-esteem (OBSE).

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1086 FARH AND CHEN

    dHighlight

    dHighlight

  • negatively to turnover intentions ( .29, p .01; Model 6,Table 6), while team relationship conflict was negatively related tovoice ( .16, p .05; Model 7, Table 5) and team-roleperformance ( .24, p .01; (Model 2, Table 6) and posi-tively to turnover intentions ( .38, p .01; Model 6, Table 6).We further examined whether OBSE and team relationshipconflict were significantly related to voice, team-role perfor-mance, and turnover intentions in the presence of other poten-tial mediators at the individual and team levels. As shown in

    Tables 5 and 6, after controlling for interpersonal justice per-ceptions, team identification, and team collective efficacy,OBSE and team relationship conflict remained significantlyrelated to the outcome variables we proposed. Thus, Hypothe-ses 4 and 6 were supported.

    Hypothesis 5 predicted that the indirect relationships betweenindividual abuse and member voice, team-role performance, andturnover intentions (via OBSE) would be moderated by team-levelabuse, such that the respective mediation pathways would be

    1.50

    2.50

    3.50

    4.50

    Low Individual-level Abusive Supervision (Perception)

    High Individual-level Abusive Supervision (Perception)

    Low Team-level Abusive Supervision (Perception)High Team-level Abusive Supervision (Perception)

    Low Individual-level Abusive Supervision (Perception)

    High Individual-level Abusive Supervision (Perception)

    Mem

    ber O

    BSE

    Figure 4. Plot of the interaction between individual- and team-level abusive supervision perceptions onindividual members organization-based self-esteem (OBSE).

    Table 6Regression Results of Study 2: Effects on Member Team-Role Performance and Turnover Intentions

    Variable

    Team member team-role performance Team member turnover intentions

    Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

    Manipulated abusive supervisionPower distance orientation .20 (.09) .13 (.09) .20 (.09) .14 (.08) .20 (.13) .32 (.12) .20 (.12) .30 (.11)Collectivism .18 (.08) .14 (.08) .11 (.08) .13 (.07) .20 (.11) .12 (.10) .06 (.10) .13 (.10)Individual AS (manipulation) .29 (.12) .15 (.13) .10 (.11) .04 (.08) .79 (.17) .06 (.17) .12 (.15) .19 (.11)Team AS (manipulation) .10 (.12) .09 (.12) .01 (.10) .07 (.09) .60 (.17) .31 (.16) .25 (.13) .17 (.12)Individual Team AS (manipulation) .12 (.17) .14 (.16) .49 (.24) .06 (.22)Individual OBSE .17 (.06) .23 (.06) .32 (.08) .37 (.08)Team relationship conflict .27 (.06) .16 (.07) .42 (.09) .23 (.09)Interpersonal justice .07 (.06) .22 (.07)Team identification .29 (.05) .22 (.07)Collective efficacy .03 (.08) .45 (.10)

    Perceived abusive supervisionPower distance orientation .18 (.09) .14 (.09) .20 (.09) .15 (.08) .18 (.12) .25 (.12) .15 (.12) .28 (.11)Collectivism .15 (.08) .15 (.08) .12 (.08) .12 (.07) .14 (.11) .13 (.10) .08 (.10) .11 (.10)Individual AS (perception) .21 (.05) .04 (.06) .12 (.08) .07 (.05) .29 (.07) .03 (.08) .06 (.11) .09 (.06)Team AS (perception) .04 (.05) .01 (.05) .03 (.05) .01 (.05) .17 (.07) .22 (.07) .23 (.06) .10 (.06)Individual Team AS (perception) .04 (.03) .00 (.03) .05 (.05) .03 (.04)Individual OBSE .13 (.06) .21 (.06) .29 (.08) .37 (.08)Team relationship conflict .24 (.07) .13 (.07) .38 (.09) .14 (.10) .21 (.09)Interpersonal justice .05 (.07)Team identification .28 (.05) .21 (.07)Collective efficacy .01 (.08) .38 (.11)Note. N 250. Standard errors are in parentheses. OBSE organization-based self-esteem; AS abusive supervision. p .05 (two-tailed). p .01 (two-tailed).

    This

    docu

    men

    tis

    copy

    right

    edby

    the

    Am

    eric

    anPs

    ycho

    logi

    calA

    ssoc

    iatio

    nor

    one

    ofi

    tsal

    lied

    publ

    isher

    s.Th

    isar

    ticle

    isin

    tend

    edso

    lely

    fort

    hepe

    rson

    aluse

    oft

    hein

    divi

    dual

    use

    ran

    dis

    not

    tobe

    diss

    emin

    ated

    broa

    dly.

    1087ABUSIVE SUPERVISION IN TEAMS

    dHighlight

    dHighlight

  • stronger when team-level abuse was low and weaker when team-level abuse was high. As shown in the lower panel of Table 3, theindirect effect of individual abuse on voice through OBSE wassignificant when team-level abuse was low (for manipulated abuse, .35, p .01, 99% CI [.69, .14]; for perceived abuse, .14, p .01, 99% CI [.29, .00]). Similarly, whenteam-level abuse was low, there was a stronger indirect effect viaOBSE from individual abuse to team-role performance (for ma-nipulated abuse, .45, p .01, 99% CI [.74, .22]; forperceived abuse, .23, p .01, 99% CI [.46, .05]) andturnover intentions (for manipulated abuse, .72, p .01, 99%CI [.39, 1.11]; for perceptions of abuse, .36, p .01, 99% CI[.02, .70]). When team-level abuse was high, however, theseindirect effects were weaker for voice (for manipulated abuse, .06, ns, 95% CI [.23, .07]; for perceived abuse, .07,ns, 95% CI [.15, .01]), team-role performance (for manipulatedabuse, .13, ns, 99% CI [.34, .02]; for perceived abuse, .07, ns, 99% CI [.24, .02]), and turnover intentions (formanipulated abuse, .23, p .01, 99% CI [.03, .50]; forperceived abuse, .19, p .01, 99% CI [.03, .36]). Takentogether, Hypothesis 5 received strong support.

    Finally, Hypothesis 7 predicted that team relationship conflictwould mediate the linkage between team-level abuse and mem-bers voice, team-role performance, and turnover intentions. Asshown in Table 3, these indirect effects were significant for mem-ber voice (for manipulated abuse, .09, p .01, 99% CI[.19, .02]; for perceived abuse, .07, p .01, 99% CI[.13, .03]), team-role performance (for manipulated abuse, .12, p .01, 99% CI [.25, .03]; for perceived abuse, .09, p .01, 99% [CI.17,.05]), and turnover intentions(for manipulated abuse, .20, p .01, 99% CI [.05, .39]; forperceived abuse, .14, p .01, 99% CI [.07, .24]). Thus,Hypothesis 7 was supported.

    Supplementary analyses. Although our hypotheses centeredon the condition of being singled out for abuse, one of our statedobjectives was to better understand how abusive supervision af-fected all members of the team. Hence, we sought to examine howindividuals in the spared conditionthat is, members subjectedto low individual abuse but high team abusemight fare. Utilizingthe PANAS (Watson et al., 1988), we assessed the state emotionsof participants following each manipulated condition. Compared toparticipants whose teams experienced low abuse, participants inthe lo