+ All Categories
Home > Documents > IO_AppSoc_14_Thammaporn_ResearchReading_550207_1

IO_AppSoc_14_Thammaporn_ResearchReading_550207_1

Date post: 30-Mar-2016
Category:
Upload: pepper-nakamaru
View: 212 times
Download: 0 times
Share this document with a friend
Description:
☆ TheauthorsacknowledgeandappreciatethefeedbackreceivedfromBennetJ.Tepper(GeorgiaStateUniversity)andSherryE.Moss(WakeForestUniversity) regardinganearlierversionofthismanuscript. ⁎ Correspondingauthor.Tel.:+18505107066. E-mailaddresses:[email protected](M.J.Martinko),[email protected](P.Harvey),[email protected](D.Sikora),[email protected](S.C.Douglas). 1 Tel.:+16038623301. 2 Tel.:+18506447846. 3 Tel.:+14062435695. journalhomepage:www.elsevier.com/locate/leaqua 1.Introduction
Popular Tags:
14
Perceptions of abusive supervision: The role of subordinates' attribution styles Mark J. Martinko a, , Paul Harvey b,1 , David Sikora c,2 , Scott C. Douglas d,3 a Bank of America Professor of Management, College of Business, Florida State University, Tallahassee, FL 32306-1110, USA b Department of Management, Whittemore School of Business and Economics, University of New Hampshire, Durham, NH 03824, USA c College of Business, Florida State University, Tallahassee, FL 32306-1110, USA d The School of Business Administration, The University of Montana, Missoula, MT 59812, USA article info abstract Empirical work on the concept of abusive supervision typically employs measurements of subordinates' perceptions of abuse as the primary dependent variable. This study began with a test of the notion that a significant proportion of subordinates' perceptions of abuse can be explained by individual differences in subordinates' attribution styles and their perceptions of the quality of their LeaderMember Exchange (LMX) relationships. Results indicated that subordinates' hostile attribution styles were positively related to subordinates' perceptions of abuse and negatively related to subordinates' LMX perceptions. We also found evidence that the abusive supervision and LMX constructs are confounded. The results call into question the conceptual and empirical distinctions between the abusive supervision and LMX constructs and indicate that attribution style plays a significant role in these perceptions. © 2011 Elsevier Inc. All rights reserved. Keywords: Abusive supervision LMX Attributions Attribution styles 1. Introduction There has been considerable recent attention devoted to the construct of abusive supervision. Tepper, Duffy, Henle, and Lambert (2006) estimated that abusive supervision costs U.S. organizations 23 billion dollars yearly as a result of increases in absenteeism, health care costs, and productivity losses. In addition to the monetary costs, researchers indicate that abusive supervision is associated with numerous other undesirable organizational outcomes including counterproductive behaviors such as aggression and sabotage (Detert & Trevino, 2007; Mitchell & Ambrose, 2007; Thau, Bennett, Mitchell, & Marrs, 2009), and decreases in organizational citizenship (Zellars, Tepper, & Duffy, 2002) and commitment (Duffy, Ganster, & Pagon, 2002). In addition to organizational costs, research demonstrates that perceptions of abusive supervision are associated with numerous negative personal outcomes including distress (Tepper, 2000, 2007), workfamily conict (Hoobler & Brass, 2006), poor job performance (Harris, Kacmar, & Zivnuska, 2007), and decreased job and life satisfaction (Tepper, 2000). In view of the serious personal and organizational costs associated with perceptions of abusive supervision, a number of studies have attempted to explain its causes. For the most part, these studies have focused almost exclusively on the characteristics of supervisors and their behaviors as the primary explanation for perceptions of supervisory abuse. For example, recent studies have explored the relationships between subordinates' perceptions of abusive supervision and supervisors' perceptions of organizational justice (Hoobler & Brass, 2006; Tepper et al., 2006; Zellars et al., 2002) as well as supervisors' personality traits (Tepper et al., 2006). The Leadership Quarterly 22 (2011) 751764 The authors acknowledge and appreciate the feedback received from Bennet J. Tepper (Georgia State University) and Sherry E. Moss (Wake Forest University) regarding an earlier version of this manuscript. Corresponding author. Tel.: + 1 850 510 7066. E-mail addresses: [email protected] (M.J. Martinko), [email protected] (P. Harvey), [email protected] (D. Sikora), [email protected] (S.C. Douglas). 1 Tel.: +1 603 862 3301. 2 Tel.: +1 850 644 7846. 3 Tel.: +1 406 243 5695. 1048-9843/$ see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.leaqua.2011.05.013 Contents lists available at ScienceDirect The Leadership Quarterly journal homepage: www.elsevier.com/locate/leaqua
Transcript

Perceptions of abusive supervision: The role of subordinates'attribution styles☆

Mark J. Martinko a,⁎, Paul Harvey b,1, David Sikora c,2, Scott C. Douglas d,3

a Bank of America Professor of Management, College of Business, Florida State University, Tallahassee, FL 32306-1110, USAb Department of Management, Whittemore School of Business and Economics, University of New Hampshire, Durham, NH 03824, USAc College of Business, Florida State University, Tallahassee, FL 32306-1110, USAd The School of Business Administration, The University of Montana, Missoula, MT 59812, USA

a r t i c l e i n f o a b s t r a c t

Empirical work on the concept of abusive supervision typically employs measurements ofsubordinates' perceptions of abuse as the primary dependent variable. This study began with atest of the notion that a significant proportion of subordinates' perceptions of abuse can beexplained by individual differences in subordinates' attribution styles and their perceptions ofthe quality of their Leader–Member Exchange (LMX) relationships. Results indicated thatsubordinates' hostile attribution styles were positively related to subordinates' perceptions ofabuse and negatively related to subordinates' LMX perceptions. We also found evidence thatthe abusive supervision and LMX constructs are confounded. The results call into question theconceptual and empirical distinctions between the abusive supervision and LMX constructsand indicate that attribution style plays a significant role in these perceptions.

© 2011 Elsevier Inc. All rights reserved.

Keywords:Abusive supervisionLMXAttributionsAttribution styles

1. Introduction

There has been considerable recent attention devoted to the construct of abusive supervision. Tepper, Duffy, Henle, andLambert (2006) estimated that abusive supervision costs U.S. organizations 23 billion dollars yearly as a result of increases inabsenteeism, health care costs, and productivity losses. In addition to the monetary costs, researchers indicate that abusivesupervision is associated with numerous other undesirable organizational outcomes including counterproductive behaviors suchas aggression and sabotage (Detert & Trevino, 2007; Mitchell & Ambrose, 2007; Thau, Bennett, Mitchell, & Marrs, 2009), anddecreases in organizational citizenship (Zellars, Tepper, & Duffy, 2002) and commitment (Duffy, Ganster, & Pagon, 2002).

In addition to organizational costs, research demonstrates that perceptions of abusive supervision are associated withnumerous negative personal outcomes including distress (Tepper, 2000, 2007), work–family conflict (Hoobler & Brass, 2006),poor job performance (Harris, Kacmar, & Zivnuska, 2007), and decreased job and life satisfaction (Tepper, 2000).

In view of the serious personal and organizational costs associated with perceptions of abusive supervision, a number of studieshave attempted to explain its causes. For the most part, these studies have focused almost exclusively on the characteristics ofsupervisors and their behaviors as the primary explanation for perceptions of supervisory abuse. For example, recent studies haveexplored the relationships between subordinates' perceptions of abusive supervision and supervisors' perceptions of organizationaljustice (Hoobler & Brass, 2006; Tepper et al., 2006; Zellars et al., 2002) as well as supervisors' personality traits (Tepper et al., 2006).

The Leadership Quarterly 22 (2011) 751–764

☆ The authors acknowledge and appreciate the feedback received from Bennet J. Tepper (Georgia State University) and Sherry E. Moss (Wake Forest University)regarding an earlier version of this manuscript.⁎ Corresponding author. Tel.: +1 850 510 7066.

E-mail addresses:[email protected] (M.J.Martinko), [email protected] (P.Harvey), [email protected] (D. Sikora), [email protected] (S.C.Douglas).1 Tel.: +1 603 862 3301.2 Tel.: +1 850 644 7846.3 Tel.: +1 406 243 5695.

1048-9843/$ – see front matter © 2011 Elsevier Inc. All rights reserved.doi:10.1016/j.leaqua.2011.05.013

Contents lists available at ScienceDirect

The Leadership Quarterly

j ourna l homepage: www.e lsev ie r.com/ locate / l eaqua

Weview it as somewhat ironic that, given that themajor dependent variable in the abusive supervision literature is subordinates'perceptions of abuse, there appears to be very little research studying the antecedents of subordinates' abuse perceptions. Werecognize that abusive behaviors by supervisors, as suggested by the current literature (e.g., Tepper, 2007), undoubtedly play asignificant and perhaps the most significant role in shaping subordinates' perceptions. However, we also suspect that there isconsiderable variability between and among subordinates in their abuse perceptions of the same supervisors. As a result, wepredict that the individual differences of subordinates also play a significant role in perceptions of abusive supervision.

In this study, our objective is explicate and test the notion that the individual differences of subordinates are related to andaccount for a significant proportion of the variability in their perceptions of abusive supervision. Although there are undoubtedlymany individual differences that may influence perceptions of abuse, we will specifically explore and test the propositions thatdifferences in subordinates' attributions styles and their perceptions of the quality of their Leader–Member Exchange (LMX)relationships will explain a significant amount of the variability in subordinates' perceptions. We begin by first exploring theconstruct of abusive supervision.We then review the current research on abusive supervisionwithin the context of attribution andLMX theory, generating hypotheses that are tested in the methods section. The article ends by considering the implications of ourfindings for the study and management of abusive supervision and LMX perceptions.

At the outset it should also be noted that our initial interest was not in validating the construct of abusive supervision. Our goalwas simply to demonstrate that the construct is more complex and prone to perceptual distortion than appears to be assumed.Webelieve that our results demonstrate that attribution styles account for a significant proportion of the variance in subordinates'perceptions of abusive supervision as well as their LMX perceptions. Our results also indicate that more needs to be done tounderstand the causes and implications of both abuse and LMX perceptions.

2. The construct of abusive supervision

Abusive supervision is defined as “subordinates' perceptions of the extent to which supervisors engage in the sustained displayof hostile verbal and nonverbal behaviors, excluding physical contact” (Tepper, 2000, p. 178). Tepper expanded on the constructby indicating that abusive supervision is a willful act on the part of supervisors, but in a later work he explained that becauseabusive supervision is a subjective assessment, its perceived occurrence can be influenced by subordinate characteristics such aspersonality and demographics (Tepper, 2007). This implies that two subordinates could each view the same supervisor's behaviordifferently. One could see the supervisor's behavior as abusive, while another subordinate could view the same behavior asappropriate.

A clear implication of the limits and constraints placed on the definition of the abusive supervision construct is that, because itis dependent on subordinates' perceptions and because subordinates' perceptions may differ, at least some of the variance inreports of abusive supervision may be accounted for by individual differences among subordinates. Moreover, if individualdifferences do account for a significant proportion of variance in abusive supervision perceptions, we cannot infer that the costsassociated with these perceptions are all due to abusive behaviors by supervisors. Given that the costs of abusive supervisiondescribed above are significant, it appears that if we are to fully explain and comprehend the causes and consequences ofperceptions of abusive supervision, we must also explore and understand how the individual differences among subordinates arerelated to perceptions of abuse.

3. Attributions and perceptions of abusive supervision

While there are numerous characteristics of subordinates that may be related to their perceptions of abuse, we believe thatattributions and attribution styles are particularly important for three reasons. First, there has been a long history of researchinvestigating the relationships between attributions and leader–member relations validating the importance of attributions (seeMartinko, Harvey, & Douglas, 2007 for a review). Second, although a limited amount of research on individual differences relatedto abusive supervision has been done, the few studies that we are aware of suggest that attributional factors are important, asdiscussed below. Finally, attribution theory itself focuses on the process by which people form causal explanations for significantlife outcomes. We believe that perceptions of having been abused are exactly the type of outcomes that attribution theorists haveaspired to understand and explain. Thus, attribution theory appears to be particularly suited to describing the factors related tohow individuals develop perceptions of abuse. In this section, we provide a brief review of attribution theory and then incorporatethe research that has been done in the area of abusive supervision and leader–member relations into an attributional frameworkand generate propositions that are tested later in the paper.

3.1. Attribution theory

Attributions and attribution theory are concerned with peoples' causal explanations for their outcomes (Heider, 1958; Weiner,1986). Typical causal explanations include ability, effort, the nature of the task, and chance. Research has demonstrated that thesecausal explanations can be mapped onto underlying casual dimensions that are consistently related with emotions, expectancies,and behaviors (Weiner, 1986).

The two most commonly researched causal dimensions are locus of causality and stability. Locus of causality is concerned withwhether the cause of an outcome resides within the person (internal) or outside of the person (external). The stability dimensionclassifies causes as enduring and stable versus unstable and transient. Locus of causality affects emotions while stability affects

752 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

expectations. Thus for example, when people believe that the cause of a failure was their own lack of ability (an internal and stableattribution) they often feel badly because they feel responsible (internal locus of causality) and expect to fail in the future becausethey believe ability is stable (Weiner, 1986).

While a substantial body of research has demonstrated the links between attributional explanations, attributional dimensions,expectancies, emotions, and behaviors (e.g., Green & Mitchell, 1979; Martinko et al., 2007; Martinko, Moss, Douglas, & Borkowski,2007; Weiner, 1986), research has also demonstrated that many individuals display attributional styles that are predictive of theirfuture behavior (e.g., Martinko et al., 2007). Attribution styles are descriptions of how people are biased in their causalexplanations. Their existence has long been recognized in social psychology, stemming from the identification of the “fundamentalattribution error,” which describes the commonly held bias toward overestimating the role of internal versus situational causeswhen observing the behavior of others (Jones & Harris, 1967; Ross, 1977). More recent research has identified situation-specificattributional tendencies that bias attributions in unique ways depending on the positive or negative nature of an outcome (e.g.,success or failure). For example, a fairly large proportion of the population appears to demonstrate a slight optimistic or self-serving attributional bias, tending to take credit for success (i.e., make internal and stable attributions) and blame external factorsfor failure (i.e., make external and unstable attributions: Mezulis, Abramson, Hyde, & Hankin, 2004). A smaller portion of thepopulation demonstrates pessimistic attribution styles which are the mirror opposites of optimistic styles, attributing success toexternal and unstable factors while attributing failure to internal and stable factors (e.g., Abramson, Seligman, & Teasdale, 1978).

In addition to optimistic and pessimistic styles, the construct of hostile attribution styles has also garnered recent attention.Hostile attribution styles are often associated with violence and aggression (Douglas & Martinko, 2001) and are characterized byexternal and stable attributions for failures which are related to anger directed toward the individuals who are perceived to beresponsible for negative outcomes (Weiner, 1986). In the context of the present study, subordinates who blame poor evaluationson their supervisors are making external and stable attributions and a pattern of these types of attributions would be indicative ofa hostile attribution style.

3.2. The relationship between attributions and abusive supervision

As indicated above, only a few studies relate the individual differences of subordinates to their perceptions of abusivesupervision. Nonetheless, we believe that each of those that we found indirectly suggests that attributions play a role in therelationships that were identified. A study by Tepper et al. (2006) found that subordinates' negative affectivity was related to thestrength of subordinates' perceptions of abusive supervision. This finding is complemented by a study by Aquino, Grover,Bradfield, and Allen (1999) that found a positive relationship between individuals' negative affectivity and their perceptions ofbeing victimized. Although the Aquino et al. study did not use abusive supervision as a dependent variable, perceptions of beingvictimized have many of the same psychological dynamics as perceptions of having been abused (Tepper, 2007).

While these studies do not point directly toward an attributional interpretation, several other studies demonstrate thatattribution styles are related to negative affectivity and perceptions of victimization. In a study of organizational aggression,Douglas and Martinko (2001) documented a positive relationship between subordinates' negative affectivity and hostileattribution styles, indicating that those who were high in negative affectivity were biased toward hostile external and stableattributions for negative outcomes. More to the point, Aquino and Byron (2002) found that co-workers who displayed thedominating and aggressive behaviors that are typical of individuals with hostile attribution styles were likely to become targets ofvictimization. In another study, Aquino, Douglas, and Martinko (2004) found that co-workers with hostile attribution styles weremore likely to express anger and perceive themselves as victimized.While we are well aware that these studies did not specificallyinvestigate supervisor–subordinate relations, we expect that these findings will generalize and that it is reasonable to infer that ifsubordinates perceive themselves as abused by their supervisors they will also feel victimized. Thus, it appears reasonable to inferthat because negative affectivity is related to hostile attribution styles and negative affectivity is also related to perceptions ofabuse, perceptions of abuse will also be related to hostile attribution styles.

Studies of employees with self-serving attributional biases also provide indirect support for the notion that attributions arerelated to perceptions of abusive supervision. A self-serving attributional bias is similar to a hostile attribution style in that itdenotes a tendency to attribute negative outcomes to external factors. This bias also involves the tendency to attribute positiveoutcomes to internal factors, such as one's effort and ability (Zuckerman, 1979). A study by Harvey and Martinko (2009) foundthat employees with this bias were relatively more likely to report conflict with their supervisors. This conflict likely stems fromthe disagreement and frustration that arises when employees with a self-serving bias erroneously blame external targets, such astheir supervisors, for negative outcomes and fail to share credit with them for positive outcomes (Martinko & Gardner, 1987). It islogical to expect that employees who experience conflict with their supervisors (e.g., arguments and tension; Jehn, 1995) will alsoreport many of the behaviors that fall under the domain of abusive supervision, such as yelling, threats, and intimidation.

Another study that supports the notion that attribution styles and perceptions of abusive supervision are related is a study byBamberger and Bacharach (2006), which found a positive relationship between problem drinking and reports of abusivesupervision. Work on self-destructive counterproductive behaviors such as problem drinking has consistently indicated that thesetypes of behaviors are associated with pessimistic attribution styles which are characterized by internal and stable attributions forfailure (Abramson et al., 1978).

As a group, the above studies all suggest that attributions may be related to perceptions of abusive supervision. In interpretingthis research, however, there is some ambiguity about the nature of this relationship. The studies on negative affectivity suggestthat hostile attribution biases (i.e., external and stable attributions for failure) may be related to perceptions of abuse. In a similar

753M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

vein, the research on self-serving attributional biases also suggests that this attributional tendency might promote abuseperceptions by elevating conflict levels between subordinates and their supervisors. On the other hand, the studies relatingproblem drinking with abuse suggest that attribution styles characterized by pessimism and internal and stable attributions forfailures, which are almost the opposite of the hostile styles and self-serving biases, are associatedwith perceived victimization andperceptions of abuse. This conclusion is somewhat contrary to that of the first two sets of studies.

To address this paradox, it is interesting to note that victim precipitation theory (Olweus, 1978; Tepper et al., 2006) suggeststhat two different types of individuals are susceptible to abuse. The first is the difficult and irritating person characterized bynegative affectivitywhich is consistent with hostile attribution styles. The second is theweak, helpless, and insecure type of personwho often demonstrates a pessimistic attribution style. Thus, it is possible that individuals with strong biases of either type arelikely to perceive abuse from their supervisors. If we follow the logic of attribution theory, however, we believe that therelationship between hostile and self-serving attribution styles and perceptions of abuse is most likely. Since individuals withthese attributional tendencies tend to deny personal responsibility for their problems and blame their problems on externalsources, we expect that they would tend to perceive negative feedback as unfair and abusive.

While we recognize that there is validity to the notion that highly pessimistic individuals are susceptible to abuse, we expectthat because of their internal and stable attributions, they see themselves as the problem, andmay believe that unfair treatment isacceptable. As such, we believe that pessimistic individuals are less likely to perceive abuse and as a result do not expectpessimistic attribution styles to be as strongly related to perceptions of abusive supervision as hostile attribution styles.

On the other hand, research and theory are consistent with respect to the role of stability in that self-serving, hostile, andpessimistic attribution styles are all characterized by stable attributions. Attributions of stability concerning the causes of negativebehaviors lead perceivers to conclude that the behaviors reflect relatively permanent traits or trait-like characteristics of anindividual (e.g., Weiner, 1995). Thus, if subordinates experience abusive supervisory behaviors and attribute the behavior to stablecauses (e.g., bad temper and poormanagement skills), they are likely to perceive their supervisor as an abusive person. Conversely,if an act of abusive supervision is attributed to an unstable cause (e.g., the supervisor had an unusually bad day), subordinates maybe less likely to form a global perception of abusive supervision. Thus, we expect that stable attributions are likely to exacerbatethe effects of external attributions on perceptions of abuse.

Following the above logic, we expect that external attribution styles for failures are most likely to be related to perceptions ofabuse and that these effects will be enhanced by stable attribution styles. We state our hypothesis for investigation as follows:

H1. There is a positive relationship between subordinates' external attribution styles for failures and perceptions of abusivesupervision, which is strongest when subordinates favor stable attributions.

4. Leader–member relations as a mediator

Up to this point, we have proposed that attribution styles account for a significant proportion of the variance in subordinates'perceptions of abusive supervision. Part of our motivation for beginning this work was the similarities we noticed between theconcepts of LMX relations and abusive supervision. More specifically, we believe that perceptions of abusive supervision may be asubset of LMX quality perceptions. Thus, it appears logical to infer that whenever subordinates perceive that their supervisorsabuse them, they also perceive that the LMX relationship is poor. As Dienesch and Liden (1986) noted, out-group members, whogenerally perceive poor LMX relations, experience fewer interactions and less trust, rewards, and support from their leaders. It islogical to conclude that, over time, because out-group individuals perceive that they are treated more poorly than in-groupmembers, they are also more likely to feel that they are targets of abuse.

As a result, we propose that:

H2. Perceptions of poor LMX quality are positively related to perceptions of abusive supervision.In addition to our expectation that there is a positive relationship between perceptions of abuse and perceptions of poor leader–

member relations, there is also research and theory which supports the argument that perceptions of poor leader–member relationswill mediate the effects of attribution styles on perceptions of abusive supervision. More specifically, several studies havedemonstrated or suggested relationships between attributions or attribution styles and LMX perceptions. A study byMartinko, Mosset al. (2007)demonstrated thatmembers' attributionswere related to their perceptionsof LMXquality such that theyweremore likelyto perceive poor quality LMX relationships when their leaders demonstrated pessimistic attribution styles. Similarly, research andtheory on counterproductive behavior by Tripp and Bies (1996) andDouglas andMartinko (2001) indicates that aggressive behaviorssuch as revenge and sabotage are more likely when subordinates attribute their failures to external and stable causes such as poorsupervisors. Thus it appears thatwhen subordinates possesshostile attributional tendencies andmakeexternal and stable attributionsfor their problems, they often target these attributions and the associated behaviors toward their supervisors. As a result, it appearsreasonable to infer that when subordinates perceive that their failures are due to external and stable causes, they are likely toexperience and/or perceive low quality LMX relationships. Thus, we predict:

H3a. There is a negative relationship between subordinates' external attribution styles and LMX perceptions, which is strongestwhen subordinates favor stable attributions.

In terms of a meditational relationship, we argue that the relationship between a hostile attributional bias and perceptions ofabusive supervision can be explained, at least in part, by poor LMX perceptions. Hostile attributional biases are expected to create

754 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

general strains on the leader–member relationship that manifest themselves in low-quality LMX perceptions. Once employeeshave begun to focus on their supervisor as a cause of their problems in this way, we argue that the general strain caused by theattributional process will, in at least some situations, evolve into more specific perceptions of abusive supervision. This is becausethe attribution-driven focus on leader shortcomings might cause subordinates to overanalyze their supervisor's behaviors andperceive isolated instances of hostile or aggressive leader behaviors (e.g., verbally disciplining subordinates) as evidence of a largertrend of abusive supervision. Therefore we propose:

H3b. The positive relationship between external attribution styles and perceptions of abuse is mediated by perceptions of lowquality LMX relationships.

5. Method

5.1. Sample

Four hundred and thirty-three full time employees participated in the study. Participants were recruited with the assistance ofstudents enrolled in four sections of an undergraduate management course. Students were asked to collect responses from up tothree individuals who were currently employed with at least five years of full time work experience and worked at least 30 hoursper week. The sample consisted of 247 females (57%) and 186 males (43%). The average age of the participants was 40.25 years.

5.2. Measures

5.2.1. Abusive supervisionSix items from Tepper's (2000) 15-item questionnaire (α=.93) were used to measure subordinates' perceptions of abusive

supervisory behaviors. This measure asks participants to rate the extent to which they agree or disagree with statements such as“My boss is rude to me” on a 5-point scale (1=“I cannot remember him/her ever using this behavior with me,” 5=“He/she usesthis behavior very often with me.”).

5.2.2. Attribution styleSix items from the organizational attribution style questionnaire (Kent & Martinko, 1995) were used to measure attribution

style (locus of causality α=.83; stability α=.86). This measure asks participants to consider negative hypothetical workplaceoutcomes (e.g., “You were passed over for a promotion”) and rate the extent to which they believe that these outcomes would becaused by internal or external causes (1=“Completely due to me”, 7=“Completely due to other people or circumstances”). Forthe stability dimension, participants were asked to rate the variability of the cause of the hypothetical outcome as stable orunstable (1=“Remain the same over time”, 7=“Change over time”).

5.2.3. LMXLeader–member relationship quality was measured using Liden and Maslyn's (1998) 12-item four-dimensional measure

(α=.93). Responses were recorded using a 5-point scale (1=“Strongly agree,” 5=“Strongly disagree”). A sample item was “Mysupervisor would come to my defense if I were ‘attacked’ by others.”

5.2.4. ControlsWe controlled for the potential influence of age and gender on abusive supervision perceptions. We also controlled for

perceived mobility, on the assumption that those who saw themselves as highly employable might approach and perceive theirrelationships with their supervisors differently than those who are highly dependent on their jobs and, thus, their supervisors'favorable evaluations. The temporal length of the supervisor–subordinate relationship was also controlled, given that the nature ofthe relationship could conceivably change over time.

6. Results

6.1. Preliminary analysis

Because the data for this study were collected using a single survey instrument we performed a Harman One-Factor test toevaluate whether common method bias (CMB) influenced our results (Podsakoff, Mackenzie, Podsakoff, & Lee, 2003; Podsakoff &Organ, 1986). This procedure involves performing a factor analysis on the study variables using principal axis factoring todetermine whether the method factor (the first factor) accounts for a disproportionate amount of variance (Fabrigar, Wegener,MacCallum, & Strahan, 1999). This analysis produced a seven-factor solution based on the eigenvalue N1.0 criteria using varimaxrotation. The method factor accounted for 17.69% of variance. This falls below the cutoff of 25% identified by Williams, Cote, andBuckley (1989), suggesting that common method variance did not substantially influence the results.

755M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

6.2. Hypothesis tests

All analyses were performed using z-standardized predictor and interaction variables. A bootstrapping procedure was used toaccommodate for the non-normal distributions typically associated with product terms in moderation analysis (Edwards &Lambert, 2007; Preacher & Hayes, 2004). These analyses were performed using Preacher and Hayes (2004) moderated-mediationSPSS macro with 1000 bootstrapped samples.

Means, standard deviations, and correlations are shown in Table 1. In the following text we present three sets of analyses—theinitial analysis performed on the data and two sets of post-hoc analyses that were performed to investigate and addressabnormalities discovered in the first set of results.

6.2.1. Initial hypothesis testsResults corresponding to H1 are shown in the first column of Table 2 and indicate a significant interaction effect (β=.09,

pb .01) between the locus of causality and stability dimensions on perceptions of abusive supervision. This interaction is shown inFig. 1. Participants with attribution styles biased toward external explanations for negative outcomes reported higher levels ofabusive supervision than thosewith an internal bias. Although the slope of this relationshipwas not significant among participantswho favored unstable attributions for negative outcomes (t= .63, ns), it was amplified among those biased toward stableattributions (t=7.01, pb .01). Thus, a tendency toward stable attributions exacerbated the positive relationship betweenexternally biased attribution styles for negative events and the perception of abusive supervision. Put differently, participantswhose styles were biased toward external and stable attributions for failure (i.e., hostile attribution styles) were more likely thanothers to perceive their supervisors as abusive, supporting H1.

H2 predicted that low quality LMX relationships would be associated with high levels of abusive supervision perceptions.Results of the regression analysis strongly supported this hypothesis (β=− .56, pb .01, R2=.32, ΔR2=.14, Cohen's f2=.45).

H3a predicted that a positive relationship exists between the tendency to attribute negative outcomes to external factors andLMX relationship quality and is strongest among those who are also biased toward stable attributions. As Table 3 (column 1)indicates, a significant interaction (β=− .06, pb .05) between external and stable attributional biases and LMXwas observed. Thisinteraction is graphed in Fig. 2 and indicates that LMX perceptions were poorest among subordinates with a hostile attributionstyle. Specifically, those with an external bias in their attributions for negative outcomes reported lower quality LMX relationships

Table 1Means, standard deviation, and correlations among study variables.

Mean SD 1 2 3 4 5 6 7

1. Gender 1.56 .502. Age 40.68 11.84 .033. Perceived mobility 3.47 1.02 .02 − .22 ⁎⁎

4. Length of leader–member relationship 4.16 6.83 .08 .18 ⁎⁎ − .075. Attribution style—Locus 3.32 1.58 .03 .06 .03 − .026. Attribution style—Stability 4.26 1.56 − .05 .10 .01 − .02 − .047. Leader–member exchange 3.78 .74 .01 .05 − .12 − .01 − .22 ⁎⁎ − .068. Abusive supervision perceptions 1.55 .78 − .09 − .04 .02 − .01 .17 ⁎⁎ .01 − .55 ⁎⁎

N=433.⁎ pb .05.

⁎⁎ pb .01.

Table 2Bootstrapped regression results: Attribution style—Abusive supervision perception relationship.

Analysis 1—No method factor Analysis 2—Full method factor Analysis 3—LMX items removed from method factor

β β β

Gender − .15 ⁎ − .15 ⁎ − .15 ⁎

Age − .01 − .01 − .01Perceived mobility .01 .01 .08Length of supervisor relationship .01 .05 .01Method factor – − .43 ⁎⁎ − .14Locus of control (A) .14 ⁎⁎ .19 ⁎⁎ .22 ⁎

Stability (B) .02 − .14 ⁎⁎ .10A B .09 ⁎⁎ .06 ⁎ .09 ⁎

R2 .07 .33 .05ΔR2 .02 ⁎⁎ .01 ⁎ .02 ⁎

f2 a .03 .01 .02

N=433, Standardized Betas reported.⁎ pb .05.⁎⁎ pb .01.a Cohen's f2 score calculated as the portion of effect size attributable to the interaction term.

756 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

than those with an internal bias and, among the externally biased group, LMX ratings were significantly lower among those whoalso attributed these outcomes to stable causes (simple slope t=−5.01, pb .01) as opposed to unstable causes (t=−2.27, pb .01).Thus, H3a was supported.

H3b argued that the effect of externally biased attributions for negative outcomes on perceptions of abusive supervisionis mediated by LMX perceptions. Results indicated that the paths between external attributional bias and LMX perceptions(β=− .16, pb .01) and between LMX and abusive supervision perceptions (β=− .56, pb .01) were simultaneously significant. Asignificant indirect effect between external attribution bias and abuse perceptions was also observed in the predicted direction(β=.08, pb .05). A one-tailed Sobel test supported the significance of this indirect effect (z=3.87, pb .01), as did bootstrap resultsindicating that a 95% bias corrected confidence interval (.05b−N .14) did not contain zero. These findings support H3b in that theeffects of attributions styles on perceptions of abuse were mediated by LMX perceptions.

Overall, these initial analyses provided strong support for all of our hypotheses.

6.2.2. Post-hoc analysisThis is where the results and the interpretation of our results became much more interesting than originally suspected. Again,

our basic thesis was quite simple. We believed that a substantial proportion of the variance in measures of abusive supervisionwould be accounted for by subordinates' attribution styles. Part of our rationale for this proposition was based on work byMartinko et al. (2007) and Martinko, Moss et al. (2007) demonstrating that attribution styles predicted the quality of leader–member relations. Given the similarity of the constructs of poor LMX relations and abusive supervision, we reasoned thatattribution styles would also be predictive of abusive supervision and that LMX relations would mediate the relationship.

Our initial version of this article essentially confirmed these predictions but considerable concern was expressed by one of thereviewers that, despite the results of the Harman One-Factor test which suggested that CMB was not a problem, an additionalanalysis with a latent method factor control variable (as suggested by Podsakoff et al., 2003) should be done to further insure thatCMB was not a problem. We felt that this request was reasonable, given that there are limitations to the one-factor test's ability toaccurately detect CMB (Podsakoff et al., 2003). While we realize that this narrative is a departure from the norm in reporting

Fig. 1. The interactive influence of attribution style on perceptions of abusive supervision.

Table 3Bootstrapped regression results: Attribution style—LMX perception relationship.

Analysis 1—No method factor Analysis 2—Full method factor Analysis 3—Abuse items removed from method factor

β β β

Gender .01 − .01 .01Age .01 .00 .01Perceived mobility − .09 ⁎⁎ − .17 ⁎⁎ − .11Length of supervisor relationship − .01 .01 − .04Method factor – .73 ⁎⁎ .03Locus of control (A) − .16 ⁎⁎ − .18 ⁎⁎ − .17 ⁎

Stability (B) − .06 − .17 ⁎⁎ − .06A B − .06 ⁎ − .01 − .06 ⁎

R2 .08 .98 .08ΔR2 .01 ⁎ .00 .01 ⁎

f2 a .01 .00 .01

N=433, Standardized betas reported.⁎ pb .05.⁎⁎ pb .01.a Cohen's f2 score calculated as the portion of effect size attributable to the interaction term.

757M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

research results, we believe a description relating the process we went through will be worthwhile and may have implicationsbeyond our immediate work.

Hoping that additional analyses would resolve doubts regarding CMB, the data were analyzed in the same manner describedabove but with the addition of the latent method factor which consisted of the items for each variable in the study as suggested byPodsakoff et al. (2003). Bootstrap regression results corresponding to Hypothesis 1 are shown in the second column of Table 2 andindicate a significant interaction effect (β=.06, pb .05) of locus of causality and stability attributions on perceptions of abusivesupervision. This interaction is shown graphically in Fig. 3. A comparison of Figs. 1 and 3 shows that the interaction changedslightly after controlling for themethod factor. In both analyses, followers whose styles favored external and stable attributions forfailure were most likely to report abusive supervision. However, the moderating effect of stability changed slightly indicating thatfollowers biased toward stable attributions consistently reported higher abuse perceptions at both high and low levels ofexternality.While the slope of the relationship at one SD above themean (t=6.40, pb .01) did not change significantly, the slope ofthe other line one SD below the mean increased (t=5.91, pb .01). Thus, although the differences between the two analyses werenot profound, they did indicate that the latent factor variable affected the responses of participants who were biased towardunstable attributions for negative outcomes.

Next, we re-analyzed the relationship between LMX and abuse perceptions noted in H2 and found that this relationshipbecame insignificant when we added the latent method factor variable (β=− .16, ns). At face value, these results suggest thatCMB was the cause of the relationship between LMX and abuse perceptions. However, it seemed unlikely to us that CMB alonecould contribute to such a large change in the observed strength of this relationship. As we explain below,we began to suspect thatconceptual and empirical overlap between the two variables might explain the affect of the latent variable on this relationship.

Inclusion of the latent variable also caused the interactive relationship between attribution style and LMX perceptions fromH3a to become insignificant (see Table 3, column 2). This again suggests that CMB artificially inflated the strength of therelationship between attribution style and LMX perceptions. Once again, however, wewere doubtful that CMB alone had so drastican effect.

The mediation argument from H3b was not retested, given the insignificant LMX-abusive supervision perception relationshipnoted above in our test of H2.

Fig. 2. The interactive influence of attribution style on LMX perceptions.

Fig. 3. The interactive influence of attribution style on perceptions of abusive supervision, controlling for method factor.

758 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

6.2.3. Post-hoc analysis—rationale for a confound testTaken at face value, the preceding analyses suggest that despite acceptable results from the Harman one-factor test, CMB still

impacted our hypothesis tests. They suggest that CMB distorted LMX and/or abusive supervision perception scores and the truestrength of these relationships. The magnitude of these apparent effects caused us to consider the possibility that the latentvariable used to control for CMB may have captured and removed variance beyond that caused by common method issues. Morespecifically, we began to suspect that some of the variance contained in the method factor stemmed not from CMB but from aconfounded relationship between the abusive supervision and LMX variables. Podsakoff et al. (2003) noted that it is possible for amethod factor to capture not just common method variance but also variance caused by relationships between the constructs,especially if the constructs lack discriminant validity. If this occurs, the latent factor acts as a control for inter-construct variancethat, while potentially undesirable, is not equivalent to common method variance.

Although there is no way to systematically determine what portion of the variance captured by the method factor isattributable to common method bias versus inter-construct variance (Podsakoff et al., 2003), the presence of strong correlationsbetween study variables can serve as a clue that inter-construct variance is strong enough to be captured as common methodvariance. An examination of the correlations in Table 1 and the preceding analyses revealed only one particularly strongassociation: between abusive supervision perceptions and LMX. If these two measures did indeed lack discriminant validity, theinclusion of the abusive supervision items in the method factor would have partialled out legitimate variance in the LMXmeasureand inclusion of the LMX items in the method factor would partial out legitimate variance in abusive supervision scores. Such aneffect could potentially explain why the latent method variable caused such a drastic change in the observed relationships.

6.2.3.1. Confound test. A statistical confound occurs when the measurement of a predictor variable directly or indirectly capturesextraneous variance which is shared by the measurement of an outcome variable (e.g., MacKinnon, Krull, & Lockwood, 2000). Inthe present example, we had already suspected that measures of abusive supervision and LMX perceptions both evaluate similaremployee perceptions. As we had suggested previously, we expected that when subordinates see their supervisor as abusive theyalso perceive their relationships as poor. Indeed, a visual inspection of the items from the scales suggests that they are measuringsimilar constructs. For example, it would seem that the abusive supervision item “my supervisor puts me down in front of others”(Tepper, 2000, p. 189) measures the inverse of the LMX item “my supervisor would come to my defense if I were ‘attacked’ byothers” (Liden & Maslyn, 1998, p. 56).

The first step in the process of identifying a potential confoundwas to determine which, if any, measurement items overlappedby re-examining the factor structure of the LMX and abusive supervision perception variables. It has been suggested that becausepsychological latent variables such as these often correlate, the use of oblique rotation can provide a more realistic estimate of thefactor structure than orthogonal rotation (Fabrigar et al., 1999). Given evidence of non-orthogonality observed between the LMXand abusive supervision variables, we performed an exploratory factor analysis with oblique rotation on these two variables.

The structure matrix of this analysis is shown in Table 4 and provides strong evidence of statistical overlap between these twoconstructs. When the factors were allowed to correlate, every item of the LMX and abusive supervision scales cross-loaded ontothree factors at or above .25 (the remaining study variables are not shown in Table 4 but loaded cleanly onto individual factors). AsTable 4 indicates, the LMX affect and professional respect items loaded onto Factor 1 most strongly, with most of the other LMXitems also exhibiting significant positive loadings on the same factor. Each of the abusive supervision items loaded negatively ontothis same factor at or above .35. The abuse items also exhibited strong positive loadings onto Factor 2, whereas all but the LMXcontribution items loaded negatively onto this factor. The contribution items loaded most strongly onto Factor 3, which was the

Table 4Structure matrix of abusive supervision and LMX items a.

Item Factor 1 Factor 2 Factor 3 Factor 4

Abuse 1 − .42 .85 – − .45Abuse 2 − .44 .86 – − .38Abuse 3 − .40 .88 – − .40Abuse 4 − .53 .91 – − .46Abuse 5 − .50 .77 – − .37Abuse 6 − .36 .87 – − .38LMX-loyalty 1 .42 − .36 − .28 .89LMX-loyalty 2 .64 − .52 − .38 .90LMX-loyalty 3 .61 − .50 − .40 .83LMX-contribution 1 .47 – − .88 .43LMX-contribution 2 .26 – − .88 .26LMX-contribution 3 .61 − .28 − .83 .39LMX-professional respect 1 .89 − .40 − .33 .49LMX-professional respect 2 .90 − .44 − .36 .52LMX-professional respect 3 .89 − .48 − .36 .57LMX affect 1 .82 − .49 − .50 .45LMX affect 2 .81 − .45 − .43 .47LMX affect 3 .84 − .47 − .37 .51

N=433.a Oblique factor rotation was used, factor loadings b .25 omitted.

759M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

only factor that did not contain any abuse items. The LMX loyalty items loadedmost strongly onto the fourth factor, ontowhich theother LMX items and abusive supervision items also cross-loaded. The fact that the abuse and LMX items cross-loaded significantlyin the opposite directions (i.e., positive vs. negative loadings) suggests that they are measuring the inverse of the same, or highlysimilar, underlying constructs.

Statistical tests to detect confounds generally involve examining direct effects with and without the confounding variablepresent (e.g., MacKinnon et al., 2000;MacKinnon,Warsi, & Dwyer, 1995). To apply such procedures to a two-variable situation, theoverlapping variance (i.e., the confounding effect) must be identified and separated into a third variable. Because all six items fromthe abusive supervision scale used here overlapped with all of the LMX items, however, it was not possible to perform this type ofprocedure. Instead, this complete overlap suggests that the items for both scales are so similar that to remove the overlappingitems would be to remove one of the scales in its entirety.

These findings substantiate our concerns regarding both the conceptual and statistical overlap between the items of the LMXand abusive supervision measures. While repeated-measures, multi-source data would be needed to more convincingly rule outCMB as a cause of some or all of this overlap, we were able to compare our data to a second cross-sectional data set in which thesame measures were used to collect abusive supervision and LMX data in a different setting. Briefly, these data were collected aspart of an unrelated study and the sample was composed of 145 employees from a shipping company (46% female, mean age:43.7 years). Responses were collected during an off-site training session. Measures of abusive supervision and LMXwere the sameas those used in the present study. An EFAwith oblique rotation on this data revealed a similar factor structure inwhich each of theabusive supervision items cross-loaded onto three LMX factors.

6.2.4. Revised hypothesis testsIt is possible that the overlap between the LMX and abusive supervision measures explains some or all of the unusual results

following the inclusion of the method factor. That is, partialling out a large portion of the shared variance (i.e., not commonmethod variance) between the two variables could have distorted the observed relationships between them and the other studyvariables. Recognizing that our analyses indicated that the LMX and abusive supervision measures are confounded, we concludedthat it was inappropriate to test a causal relationship between these variables. As a result, we dropped our mediation hypothesesand considered both the LMX and abusive supervision variables in terms of their relationships with attribution stylesindependently.

In this section we consider only H1 and H3a, as these predicted the effects of attribution styles on LMX and abusive supervisionperceptions separately and did not involve a causal relationship between the two outcome variables. To control for CMB withoutthe risk of distortion caused by the shared variance between the LMX and abusive supervision measures, the analyses for eachoutcome variable were performed independently. A separate method factor was calculated for the relevant study variables in bothof the separate analyses and included as a control variable.

Bootstrapped regression results for H1 appear in the third column of Table 2 and the interactive relationship between theattribution style dimensions and abusive supervision perceptions are shown in Fig. 4. These results closely resemble those fromthe initial analysis and provide similar support for H1. The primary distinctions between the two sets of analyses were thatcontrolling for the method factor clarified that the interaction effect exists primarily among the externally biased participants andalso increased the slope of the low-stability line (t=2.76, p=.01; high-stability slope: t=6.59, p=.01).

Results for H3a appear in the third column of Table 3. These results suggested an interaction similar to that found in the initialanalysis shown in the first column of Table 3. As predicted, participants showed a significant reduction in LMX perceptions whenthey were biased toward external attributions for failures, a tendency that was amplified among those favoring stable (simpleslope: t=−5.74, pb .01) as opposed to unstable (t=−4.04, pb .01) attributions. This interaction is shown in Fig. 5.

Fig. 4. The interactive influence of attribution style on perceptions of abusive supervision, LMX items removed from method factor.

760 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

7. Discussion

This study morphed into a set of questions and issues that it was not originally designed to address. Again, our most basicquestion was whether or not subordinates' personal characteristics, and specifically attribution styles, accounted for at least someof the variance in their ratings of abusive supervision. We believe that we effectively answered that question and, at least in thissample, attribution styles did account for a significant proportion of the variance in subordinates' ratings of abusive supervision.

In addition, in seeking to rule out the possibility that our initial results could be a function of CMB, we foundwhat we believe bea serious confound between the constructs of LMX and abusive supervision, calling into question the discriminant validity of theseconstructs. These issues are addressed in more detail below.

7.1. The perceptual basis of abusive supervision

Overall, the findings from all three sets of analyses support our general thesis that a significant proportion of the variability insubordinates' perceptions of abusive supervision is accounted for by their attribution styles. More specifically, all three analysesconfirm that there is a positive relationship between external attribution styles and perceptions of abuse and that this relationshipis stronger when subordinates are biased toward believing that the causes of their outcomes are stable.

The implications of these findings are important in that they demonstrate that perceptions of abuse are at least partially afunction of subordinates' attribution styles. As a result, supervisory behaviors might not be the only cause of subordinates'perceptions of abuse and supervisors' behaviors might be only partially responsible for the negative consequences that have beenassociated with abusive supervision. Thus, interventions designed to ameliorate the effects of abusive supervision that exclusivelytarget supervisors' behaviors may be less than optimal. Moreover, given that the consequences of perceptions of abusivesupervision are substantial and that at least a portion of these perceptions can be attributed to the characteristics of subordinates,interventions which target subordinates might also be important to consider. Better selection procedures that reduce theprobability of hiring individuals with hostile attribution styles as well as interventions designed to help employees make lessbiased attributions should be considered.

Some comment regarding causation is appropriate. Although we used a cross sectional design that limits our ability to makecausal inferences, nonetheless we may be able to infer causation. More specifically, since attributions styles are generally thoughtto be relatively enduring and stable characteristics of individuals (Russell, 1991), and since they exist before a subordinate forms arelationship with a supervisor, attributional styles would not normally be thought of as being directly influenced by supervisorybehavior. Thus, at least to some extent, attribution styles as well as other enduring subordinate characteristics can be thought of asprecipitating or at least facilitating perceptions of abusive supervision.

It should also be noted that attribution style is only one of many individual difference factors that may be associated withsubordinates' perceptions of abuse. Thus, although attribution styles accounted for a moderate amount of the variance insubordinates' perceptions, other individual difference variables may account for substantially more variance. While it would seemalmost impossible to determine how much of the variance in subordinates' perceptions of abuse is attributable to supervisorybehavior as opposed to subordinates' characteristics, it would seem that addressing this question is critical in order to provideguidance for interventions designed to ameliorate the negative consequences of abusive supervision. Thus additional researchinvestigating other subordinate characteristics such as self-efficacy, negative affectivity, and the big five factors (Judge, Locke,Durham, & Kluger, 1998) is warranted. In addition, direct investigations of how supervisory behaviors affect perceptions of abuseare needed. Laboratory studies may help answer some of these questions. It would also seem that a study of the attribution stylesof union grievants who presumably perceive that they were mistreated, versus their co-workers who were treated the same way,may provide particularly useful information.

Fig. 5. The interactive influence of attribution style on LMX perceptions, abusive supervision items removed from method factor.

761M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

7.2. Issues of confounding and validity

In addition to our findings regarding the relationships between attribution styles and perceptions of abuse, the results of ouranalyses also raise serious questions about the discriminant validity of both the LMX and abusive supervision measures. Theresults of our latent variable and factor analyses suggest thatmuch of the variance that thesemeasures capture is similar.While wecan only speculate on the nature of the variance that these measures share, it seems to us that, as long ago suggested by suggestedby Eden and Leviatan (1973) and Rush, Thomas, and Lord (1977), subordinates' responses tomeasures of leader behavior are oftenat least partially a function of their implicit theories of leadership. More specifically, as suggested by implicit leadership theory, weexpect that subordinates have stereotypes of the types of behaviors that are characteristic of good as well as poor leaders. In thecontext of the present study, it may well be that the subordinates make an overall judgment of their leaders' effectiveness orineffectiveness and then apply their implicit theories to the way they respond to the items on the questionnaires. Whether or notthe questionnaires relate to actual behaviors such as abuse is unclear. Thus it would seem that more work needs to be done withboth measures to more clearly relate these measures to observable behaviors.

It is also worth noting that even if we are not quite sure what these instruments measure, there is a growing body of researchrelating to both of thesemeasures that demonstrate reliable relationships with othermeasures such as turnover intentions and jobsatisfaction. Since there appears to be overlap between the constructs of LMX and abusive supervision, the research on LMX mayhelp inform the research on abusive supervision and vice versa. However, because of the similarities in the nature of therelationships identified, more work needs to be done to differentiate these constructs. Both constructs may benefit by sharpeningthe distinctions between them and building the nominological network of relations associated with these constructs.

Finally, given our findings concerning the potential for attribution styles to influence subjects' responses tomeasures of abusivesupervision and LMX, it would seem prudent to include attribution style as a control variable in studies of these constructs. Inaddition, since attribution style was the only personality characteristic we studied, other personality variables such as negativeaffectivity and self-esteem might also be considered as controls.

7.3. Limitations and future directions

There were several limitations of this study which should be addressed in future works. First, the sampling procedure was notideal. Although each of the students obtained the required number of subordinates, it is probable some of them had potentialsubjects who refused to participate. Thus we do not know how many non-respondents there were and there is a possibility thatthere were systematic differences in the subjects who chose to participate versus those who did not participate. Our samplingprocedure also precluded within-group analyses that would allow us to determine whether biased attribution styles affectedemployees' LMX and abusive supervision perceptions of the same supervisor. The results of the present study suggest a generaltendency toward unfavorable leadership perceptions caused by hostile attributional tendencies. A group-level replication withemployees that share the same supervisors would be a valuable next step in this line of inquiry.

As suggested above, there are a number of possible individual difference factors that may demonstrate significant relationshipswith abusive supervision as well as LMX. Additional studies including other individual difference factors would help to determinemore clearly the amount of variability in perceptions of abuse that can be attributed to subordinate characteristics. In addition tosubordinate characteristics, it is also possible that other factors such as the nature of tasks, working conditions, and organizationalculture could be associated with subordinate's perceptions of abusive supervision and LMX.

Another potential weakness of the study is that we only measured subordinates' attribution styles. As the work by Martinko,Moss et al. (2007) demonstrates, there are significant interactions between the attribution styles of leaders and members that arerelated to members' poor LMX perceptions. We expect that these same kinds of interactions will also be found with regard tosubordinates' perceptions of abusive supervision. In addition, it is also quite possible that leaders' attribution styles are associatedwith subordinates' perceptions of abuse. Indeed, in one of the studies Martinko, Moss et al. (2007) reported, the pessimisticattribution styles of leaders were positively related to members' perceptions of poor LMX relations. We think that it is also quitepossible that leaders who are biased toward internal and stable attributions for subordinates' failures will be perceived to beabusive by their subordinates.

An additional weakness that we should note is that we did not directly assess whether or not the abusive supervision measurerelates to actual abusive behaviors by supervisors. The scope of our objective was narrower in that we sought to show that, at leastin part, the responses on this measure were a function of the respondents' attribution styles. Clearly, more work should be done toidentify the extent to which this measure is related to objectively observable abusive supervisory behaviors. However, direct testsof this relationshipwould be difficult in that institutional review boards are unlikely to permit experimental manipulations relatedto abuse. Moreover, even if non-harmful simulations of abuse could be experimentally induced, there would be considerabledifficulty in generalizing such studies to work settings. Nevertheless, efforts to more directly assess the relationship betweenmeasures of abuse and supervisory behaviors should be considered.

8. Conclusion

In conclusion, this study demonstrates that the attribution styles of subordinates account for a significant proportion of thevariability in subordinates' perceptions of abusive supervision. The data clearly show that subordinates' external and stableattribution styles were associated with perceptions of abuse. In addition, in our attempt to rule out CMB, we found that the LMX

762 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

and abusive supervision constructs appear to be confounded. Since both of these constructs appear to be at least partially afunction of subordinates' attribution styles, questions about the validity and nominological network for both constructs emerged.In particular, we question the extent to which these constructs are a function of leader behavior as opposed to subordinatecharacteristics.

Placing our findings in context, they demonstrate that perceptions of both abuse and LMX are more complex than initiallyanticipated. While inappropriate behaviors by supervisors undoubtedly play a significant role in shaping subordinates'perceptions of abuse and the LMX relationship, the finding that subordinates' characteristics also play significant roles suggeststhat we need research to more fully understand what these constructs are measuring and their implications. Additional studiesexploring other potential variables that may also contribute to subordinates' perceptions of their LMX relations and abusivesupervision are needed in order to fully comprehend and address the causes of the negative personal and organizational outcomesassociated with these measures.

References

Abramson, L. Y., Seligman, M. E., & Teasdale, J. D. (1978). Learned helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87, 49–74.Aquino, K., & Byron, K. (2002). Dominating interpersonal behavior and perceived victimization in groups: Evidence for a curvilinear relationship. Journal of

Management, 28, 69–87.Aquino, K., Douglas, S. C., & Martinko, M. J. (2004). Overt expressions of anger in response to perceived victimization by co-workers: Attributional style,

hierarchical status, and organizational norms. Journal of Occupational Health Psychology, 9, 152–164.Aquino, K., Grover, S. L., Bradfield, M., & Allen, D. G. (1999). The effects of negative affectivity, hierarchical status and self-determination on workplace

victimization. Academy of Management Journal, 42, 260–272.Bamberger, P. A., & Bacharach, S. B. (2006). Abusive supervision and subordinate problem drinking: Taking resistance, stress, and subordinate personality into

account. Human Relations, 59, 1–30.Detert, J. R., & Trevino, L. K. (2007). Managerial modes of Influence and counterproductivity in organizations: A longitudinal business-unit-level investigation. The

Journal of Applied Psychology, 92, 993–1005.Dienesch, R. M., & Liden, R. C. (1986). Leader–member exchange model of leadership: A critique and further development. Academy of Management Review, 11,

618–634.Douglas, S. C., & Martinko, M. J. (2001). Exploring the role of individual differences in the prediction of workplace aggression. The Journal of Applied Psychology, 86,

547–559.Duffy, M. K., Ganster, D. C., & Pagon, M. (2002). Social undermining at work. Academy of Management Journal, 45, 331–351.Eden, D., & Leviatan, U. (1973). Implicit leadership theory as a determinant of the factor structure underlying supervisory behavior scales. The Journal of Applied

Psychology, 60, 736–741.Edwards, J. R., & Lambert, L. S. (2007). Methods for integrating moderation and mediation: A general analytical framework using moderated path analysis.

Psychological Methods, 12, 1–22.Fabrigar, L. R., Wegener, D. T., MacCallum, R. C., & Strahan, E. J. (1999). Evaluating the use of exploratory factor analysis in psychological research. Psychological

Methods, 4, 272–299.Green, S. G., & Mitchell, T. R. (1979). Attributional processes of leaders in leader–member interactions. Organizational Behavior and Human Performance, 23,

429–458.Harris, K. J., Kacmar, K. M., & Zivnuska, S. (2007). An investigation of abusive supervision as a predictor of performance and the meaning of work as a moderator of

the relationship. The Leadership Quarterly, 18, 252–263.Harvey, P., & Martinko, M. J. (2009). An empirical examination of the role of attributions in psychological entitlement and its outcomes. Journal of Organizational

Behavior, 30, 459–476.Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley.Hoobler, J., & Brass, D. (2006). Abusive supervision and family undermining as displaced aggression. The Journal of Applied Psychology, 91, 1125–1133.Jehn, K. (1995). A multimethod examination of the benefits and detriments of intragroup conflict. Administrative Science Quarterly, 40, 256–282.Jones, E. E., & Harris, V. A. (1967). The attribution of attitudes. Journal of Experimental Social Psychology, 3, 1–24.Judge, T., Locke, E., Durham, C., & Kluger, A. N. (1998). Dispositional effects on job and life satisfaction: The role of core self-evaluations. The Journal of Applied

Psychology, 83, 17–34.Kent, R., & Martinko, M. J. (1995). The development and evaluation of a scale to measure organizational attribution style. In M. J. Martinko (Ed.), Attribution theory:

An organizational perspective (pp. 53–75). Delray Beach, FL: St. Lucie Press.Liden, R. C., &Maslyn, J. M. (1998). Multidimensionality of leader–member exchange: An empirical assessment through scale development. Journal of Management,

24, 43–72.MacKinnon, D. P., Krull, J. L., & Lockwood, C. M. (2000). Equivalence of the mediation, confounding, and suppression effect. Prevention Science, 1, 173–181.MacKinnon, D. P., Warsi, G., & Dwyer, J. H. (1995). A simulation study of mediated effect measures. Multivariate Behavioral Research, 30, 41–62.Martinko, M. J., & Gardner, W. L. (1987). The leader–member attribution process. Academy of Management Review, 12, 235–249.Martinko, M. J., Harvey, P., & Douglas, S. C. (2007). The role, function, and contribution of attribution theory to leadership: A review. The Leadership Quarterly, 18,

561–585.Martinko, M. J., Moss, S. E., Douglas, S. C., & Borkowski, N. (2007). Anticipating the inevitable: When leader and member attribution styles clash. Organizational

Behavior and Human Decision Processes, 104, 158–174.Mezulis, A. H., Abramson, L. Y., Hyde, J. S., & Hankin, B. L. (2004). Is there a universal positivity bias in attributions? A meta-analytic review of individual,

developmental, and cultural differences in the self-serving attributional bias. Psychological Bulletin, 130, 711–747.Mitchell, M. S., & Ambrose, M. L. (2007). Abusive supervision and workplace deviance and the moderating effects of negative reciprocity beliefs. The Journal of

Applied Psychology, 92, 1159–1168.Olweus, D. (1978). Aggression in schools: Bullies and whipping boys. Washington, D.C.: Hemisphere.Podsakoff, P. M., Mackenzie, S. B., Podsakoff, N. P., & Lee, J. Y. (2003). Common method biases in behavioral research: A critical review of the literature and

recommended remedies. The Journal of Applied Psychology, 88, 879–903.Podsakoff, P., & Organ, D. (1986). Self reports in organizational research: Problems and prospects. Journal of Management, 12, 531–544.Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, Instruments, &

Computers, 36, 717–731.Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In L. Berkowitz (Ed.), Advances in experimental social

psychology, Vol. 10. (pp. 173–219). New York: Academic Press.Rush, M., Thomas, J., & Lord, R. (1977). Implicit leadership theory: A potential threat to the internal validity of leader behavior questionnaires. Organizational

Behavior and Human Performance, 20, 93–110.Russell, D. W. (1991). The measurement of attribution process: Trait and situational approaches. In S. L. Zelen (Ed.), New models, new extensions of attribution

theory. New York: Springer-Verlag.

763M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764

Tepper, B. J. (2000). Consequences of abusive supervision. Academy of Management Journal, 43, 178–190.Tepper, B. J. (2007). Abusive supervision in work organizations: Review, synthesis, and research agenda. Journal of Management, 33, 261–289.Tepper, B. J., Duffy, M. K., Henle, C. A., & Lambert, L. S. (2006). Procedural injustice, victim precipitation, and abusive supervision. Personnel Psychology, 59, 101–123.Thau, S., Bennett, R. J., Mitchell, M. S., & Marrs, M. B. (2009). How management style moderates the relationship between abusive supervision and workplace

deviance: An uncertainty management theory perspective. Organizational Behavior and Human Decision Processes, 108(1), 79–92.Tripp, T. M., & Bies, R. J. (1996). Getting even: The truth about workplace revenge—And how to stop it. San Francisco: Jossey-Bass.Weiner, B. (1986). An attributional theory of motivation and emotion. Springer series in social psychology. New York: Springer-Verlag.Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New York: Guilford Press.Williams, L., Cote, J., & Buckley, M. (1989). Lack of method variance in self-reported affect and perceptions of work: Reality or artifact? The Journal of Applied

Psychology, 74, 462–468.Zellars, K. L., Tepper, B. J., & Duffy, M. K. (2002). Abusive supervision and subordinates' organizational citizenship behavior. The Journal of Applied Psychology, 86,

1068–1076.Zuckerman, M. (1979). Attribution of success and failure revisited, or the motivational bias is alive and well in attribution theory. Journal of Personality, 47,

245–287.

764 M.J. Martinko et al. / The Leadership Quarterly 22 (2011) 751–764