15
The mediating role of LMX between abusive supervision and work behaviors A replication and extension Stijn Decoster, Jeroen Camps and Jeroen Stouten Department of Psychology, KU Leuven, Leuven, Belgium Abstract Purpose – In a replication of a multi-source study by Xu et al., the authors examined whether leader-member exchange (LMX) mediates the relationship between abusive supervision and employee work behaviors, more specifically task performance, organizational citizenship behaviors toward the organization (OCBO), and toward other individuals (OCBI). Moreover, the authors also examined whether LMX mediates this relationship when the authors focus on the two dimensions of abusive supervision, that is active-aggressive and passive-aggressive abusive supervision. Design/methodology/approach – The authors collected multi-source data in order to minimize common method bias. The authors conducted regression analyses, Sobel tests, and bootstrapping techniques. Findings – The authors found support that LMX mediates the negative relationship between abusive supervision and OCBO and OCBI. However, the authors could not replicate the mediating role of LMX in the association between abusive supervision and employees’ performance. Similar results were obtained when the data were analyzed with the active-aggressive and passive-aggressive abusive supervision subscales. Research limitations/implications – Due to the cross-sectional nature of the study, this study does not allow the authors to draw causal conclusions regarding the proposed relationships. Originality/value – The authors replicated Xu et al.’s findings in a European context with a different sample and different measures for LMX, performance, and OCBI. The authors conducted bootstrapping analyses in order to control for the skewed distribution of abusive supervision. The authors explore whether the proposed relations still stand with regard to active-aggressive and passive-aggressive abusive supervision. Keywords Leadership, Performance, OCB, LMX, Abusive supervision, Exchange, OCBI, OCBO Paper type Research paper During the past decade, a growing range of studies have reported on the far-reaching consequences for those who are the victim of an abusive supervisor (Martinko et al., 2013; Tepper, 2007). Xu et al. (2012) showed that abusive supervision is detrimental for employees’ performance, their organizational citizenship behaviors toward the organization (OCBO), and toward other individuals (OCBI), and that these relations can be explained by social exchange processes between the supervisor and his/her employees (i.e. leader-member exchange (LMX)). In the present paper we aim to replicate the findings of Xu et al. (2012). Replication is defined as “the duplication of a previously published empirical study to determine The current issue and full text archive of this journal is available at www.emeraldinsight.com/1935-5181.htm The second author is a doctoral fellow of the Agency for Innovation by Science and Technology in Flanders (IWT). American Journal of Business Vol. 29 No. 1, 2014 pp. 61-75 q Emerald Group Publishing Limited 1935-5181 DOI 10.1108/AJB-06-2013-0038 Mediating role of LMX 61

The mediating role of LMX between abusive supervision and work behaviors

  • Upload
    jeroen

  • View
    212

  • Download
    0

Embed Size (px)

Citation preview

Page 1: The mediating role of LMX between abusive supervision and work behaviors

The mediating role of LMXbetween abusive supervision

and work behaviorsA replication and extension

Stijn Decoster, Jeroen Camps and Jeroen StoutenDepartment of Psychology, KU Leuven, Leuven, Belgium

Abstract

Purpose – In a replication of a multi-source study by Xu et al., the authors examined whetherleader-member exchange (LMX) mediates the relationship between abusive supervision and employeework behaviors, more specifically task performance, organizational citizenship behaviors toward theorganization (OCBO), and toward other individuals (OCBI). Moreover, the authors also examinedwhether LMX mediates this relationship when the authors focus on the two dimensions of abusivesupervision, that is active-aggressive and passive-aggressive abusive supervision.

Design/methodology/approach – The authors collected multi-source data in order to minimizecommon method bias. The authors conducted regression analyses, Sobel tests, and bootstrappingtechniques.

Findings – The authors found support that LMX mediates the negative relationship between abusivesupervision and OCBO and OCBI. However, the authors could not replicate the mediating role of LMXin the association between abusive supervision and employees’ performance. Similar results wereobtained when the data were analyzed with the active-aggressive and passive-aggressive abusivesupervision subscales.

Research limitations/implications – Due to the cross-sectional nature of the study, this studydoes not allow the authors to draw causal conclusions regarding the proposed relationships.

Originality/value – The authors replicated Xu et al.’s findings in a European context with adifferent sample and different measures for LMX, performance, and OCBI. The authors conductedbootstrapping analyses in order to control for the skewed distribution of abusive supervision. Theauthors explore whether the proposed relations still stand with regard to active-aggressive andpassive-aggressive abusive supervision.

Keywords Leadership, Performance, OCB, LMX, Abusive supervision, Exchange, OCBI, OCBO

Paper type Research paper

During the past decade, a growing range of studies have reported on the far-reachingconsequences for those who are the victim of an abusive supervisor (Martinko et al.,2013; Tepper, 2007). Xu et al. (2012) showed that abusive supervision is detrimentalfor employees’ performance, their organizational citizenship behaviors toward theorganization (OCBO), and toward other individuals (OCBI), and that these relations canbe explained by social exchange processes between the supervisor and his/heremployees (i.e. leader-member exchange (LMX)).

In the present paper we aim to replicate the findings of Xu et al. (2012). Replication isdefined as “the duplication of a previously published empirical study to determine

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/1935-5181.htm

The second author is a doctoral fellow of the Agency for Innovation by Science and Technologyin Flanders (IWT).

American Journal of BusinessVol. 29 No. 1, 2014

pp. 61-75q Emerald Group Publishing Limited

1935-5181DOI 10.1108/AJB-06-2013-0038

Mediating roleof LMX

61

Page 2: The mediating role of LMX between abusive supervision and work behaviors

whether the findings of that study are repeatable” (Singh et al., 2003, p. 534). Althoughreplication studies can be valuable, replication studies are still scarce(Evanschitzky et al., 2007; Hubbard and Armstrong, 1994). This may be attributedto the belief that replication studies do not make a significant contribution to the field,that they lack originality, novelty, and creativity, and are even hostile toward theoriginal research (Easley et al., 2013; Makel et al., 2012). Moreover, researchers are lessmotivated to conduct replication studies since such studies are perceived to yield toless publications and citations (Koole and Lakens, 2012; Neuliep and Crandall, 1993).

While replication has not been adequately recognized as a valuable part of theorydevelopment, research findings can only fully advance both theorizing and practice ifone is certain that these findings can be replicated in different samples (Tsang andKwan, 1999). That is, replication studies allow researchers to confirm or disconfirm theresults of previous studies. As such, replication studies can add evidence that previousfindings were not merely statistical artifacts or exclusive to the original sample(Schwab, 2005). In sum, replication is essential to ensure the validity and reliability ofresearch (Evanschitzky et al., 2007; Singh et al., 2003; Tsang and Kwan, 1999).

Tsang and Kwan (1999) classify replications along two dimensions:

(1) whether the same data sources are used; and

(2) whether the same measurements and/or analyses are used.

These two dimensions result into six types of replications: “checking of analysis” and“reanalysis of data” when both studies use the same data set; “exact replication” and“conceptual extension” when data are collected from the same population; “empiricalgeneralization” and “generalization and extension” when data are collected from adifferent population. In the first (second) type of each of the three pairs the same(different) measures and/or analyses are used. In the present research we used a type ofreplication that has been classified as “generalization and extension” as we aim to testfor generalization and to extend the findings reported by Xu et al. (2012). This type ofreplication is characterized by the use of a different population and different measuresand/or analyses compared to the original study (Tsang and Kwan, 1999).

We believe conducting replication studies in the area of abusive supervision iswarranted given the high costs of abusive supervision to organizations. Morespecifically, abusive supervision affects an estimated 13.6 percent of the USA workforce(Schat et al., 2006). Tepper et al. (2006) estimated the cost of abusive supervision forcorporations in the USA to be $23.8 billion annually due to employees’ lost productivity,absenteeism, and health care costs. Apart from the economic cost, abusive supervisionis also associated with personal costs for the employees, such as psychological distress(Tepper, 2000), family undermining (Carlson et al., 2012; Kiewitz et al., 2012), andproblem drinking (Bamberger and Bacharach, 2006). In sum, abusive supervision is asignificant organizational problem that warrants continued research, and thus alsoreplication research.

Below, we will briefly outline the theoretical background for this study and explainin detail how and why we aim to replicate their findings.

Abusive supervision and work behaviorsAbusive supervision is defined as “subordinates” perceptions of the extent to whichsupervisors engage in the sustained display of hostile verbal and nonverbal behaviors,

AJB29,1

62

Page 3: The mediating role of LMX between abusive supervision and work behaviors

excluding physical contact’ (Tepper, 2000, p. 178). Examples include a supervisormaking negative comments about an employee to other members of the organization,or telling that a subordinate’s thoughts or feelings are stupid. The extant literature onthis topic shows that abusive supervision is associated with a variety of negativeconsequences for the organization, as well as for the employees (Martinko et al., 2013;Tepper, 2007). For example, abusive supervision has detrimental consequences withregard to employees’ performance (Harris et al., 2007) and their organizationalcitizenship behaviors (Aryee et al., 2007; Zellars et al., 2002). In order to betterunderstand the relation between abusive supervision and its negative consequences,researchers have drawn on mechanisms such as perceived fairness and control(reactance theory), but also on social exchange (Aryee et al., 2007; Mitchell andAmbrose, 2007; Thau et al., 2009). In the following section, we follow Xu et al.’s (2012)rationale for focusing on the social exchange relationship between a leader and his/heremployees as an underlying mechanism that explains the relation between abusivesupervision and employees’ work behaviors.

The mediating role of LMXCentral to social exchange is the concept of reciprocity: individuals are sensitive tovalued outcomes they receive and they are motivated to reciprocate these outcomes(Blau, 1964). The quality of the social exchange between supervisors and employeesis denoted as LMX (Graen, 1976). When employees perceive the relationship with theirleader is valuable they may reciprocate by, for example, displaying more positivebehaviors. For example, LMX has been found to be positively related to employeeperformance (Bauer and Green, 1996; Rosen et al., 2011) and organizational citizenshipbehavior (Settoon et al., 1996; Wayne and Green, 1993). In contrast, when employeesperceive a lack of valued outcomes by their supervisor, they are expected to reciprocatein a negative way in order to restore the balance. For example, employees who areconfronted with low LMX are typically unwilling to go beyond behaviors specified intheir employment contract (Moss et al., 2009).

Xu et al. (2012) built upon this line reasoning and conducted the first (and only, asfar as we know) study that tested whether LMX is an underlying mechanism thatexplains the negative association between abusive supervision and employees’ reactionsto this behavior. In sum, Xu et al. (2012) proposed the following hypotheses:

H1. LMX mediates the negative relationship between abusive supervision andemployees’ performance.

H2. LMX mediates the negative relationship between abusive supervision andemployees’ OCBO.

H3. LMX mediates the negative relationship between abusive supervision andemployees’ OCBI.

This study aims to contribute to the existing knowledge in several ways. First of all, thepurpose of our study was to replicate and extend the results from the study by Xu et al.(2012) with a different sample (from a different country) and different measures forLMX, performance, and OCBI. In this way, we can establish that the relationships foundby Xu et al. (2012) are not a product of coincidence. We conducted a generalizationand extension, that is, we used a different population, different measures, and different

Mediating roleof LMX

63

Page 4: The mediating role of LMX between abusive supervision and work behaviors

analyses compared to the original study in order to enhance the generalizability ofXu et al.’s (2012) empirical findings (Tsang and Kwan, 1999). Indeed, when the results ofthe original study are supported by a generalization and extension replication, there isa great benefit to the external validity of the original study. Moreover, the more theoriginal sample and procedures differ from the replication’s sample and procedure, themore benefit there is to the external validity (Rosenthal, 1991). Therefore, our studyaimed to differ from Xu et al.’s (2012) study in several ways.

First, there are differences between the samples of both studies. That is, the originalstudy was conducted in China, while our study was conducted in Europe, morespecifically Belgium. Also, in Xu et al.’s (2012) study, the majority of the leaders(69.9 percent) were female, whereas in our sample only 30.7 percent were female.The mean age of the leaders in the original study is also lower than in our sample (32 vs42 years). Gender and age of the employees also differed between both samples: inXu et al.’s (2012) study 64.2 percent were females with a mean age of 29 years,compared with 45.5 percent females and a mean age of 38 years in our sample.

Second, we opted for different measures for LMX, performance, and OCBI in orderto increase the external validity of the results found by Xu et al. (2012). Morespecifically, we opted for a 12-item measure of LMX (Liden and Maslyn, 1998) sincethis measure was constructed to address some deficiencies in the seven-item versiondeveloped by Graen and Uhl-Bien (1995), which was used by Xu et al. (2012). Indeed,the 12-item version has been found to better reflect employees’ evaluation of therelational characteristics and qualities of the leader-employee relationship than theseven-item version (Wang et al., 2005). Moreover, the measure of Liden and Maslyn(1998) has generally been used in previous abusive supervision research (Harris et al.,2011; Martinko et al., 2011). Next, we also opted to use a different performance scalesince Xu et al. (2012) used a scale that only consisted of three items. Therefore, we use asix-item scale (Abramis, 1994) instead of the three-item measure developed byMotowidlo and Van Scotter (1994). We also chose to measure employees’ performanceby their coworkers’ ratings whereas Xu et al. (2012) measured employees’ performancewith ratings from the supervisor. We opted for coworker ratings of the employees’performance since abusive leaders might be biased in their performance ratings ofthe employee they act abusively to. Indeed, employees’ performance ratings bysupervisors have been found to be influenced by the quality of the LMX relationship,and therefore it is recommended to use more “objective” performance ratings(Duarte et al., 1994). Moreover, when a supervisor would complete both employees’performance and OCB, this could inflate the correlation between those measures.Indeed, recent research pointed out that it is best to obtain performance and OCBmeasures from different raters in order to prevent common rater effects (Podsakoff et al.,2013). Therefore, we opted to use coworker ratings rather than leader ratings ofperformance. Finally, we chose to measure OCBI by using the five-item altruismsubscale of the OCB measure developed by Konovsky and Organ (1996). Altruismrefers to discretionary behaviors that help other persons with an organizationallyrelevant task or problem. Therefore, altruism has been identified as voluntary behaviortoward individuals (OCBI) (Coleman and Borman, 2000).

Third, since this is a generalization and extension, we also used different analysescompared to Xu et al.’s study. More specifically, whereas Xu et al. use SEM, we optedto conduct hierarchical regression analyses with bootstrapping procedures since

AJB29,1

64

Page 5: The mediating role of LMX between abusive supervision and work behaviors

abusive supervision has been characterized as a low base-rate phenomenon with askewed distribution and bootstrapping does not require a normal distribution of thedata (Shrout and Bolger, 2002; Tepper and Almeda, 2012).

Our second aim is to contribute to the existing knowledge of the abusive supervisionscale. Since the publication of Tepper’s (2000) original article on abusive supervision,his 15-item abusive supervision scale has been widely used in research. Based on theoriginal data from Tepper (2000) and Mitchell and Ambrose (2007) revealed thatabusive supervision consists of two separate dimensions (i.e. passive-aggressive abusivesupervision and active-aggressive abusive supervision). Although research used eitherTepper’s (2000) 15-item scale (Lian et al., 2012; Tepper et al., 2008), or Mitchell andAmbrose’s (2007) shortened five-item active-aggressive abusive supervision scale(Lin et al., 2013; Thau and Mitchell, 2010) to measure abusive supervision, only onestudy – as far as we know – has explored whether both measures yield similarresults (Tepper et al., 2009). Here, we add to the existing literature as we will explorewhether the findings of Xu et al. (2012) can be replicated for the original 15-item abusivesupervision scale as well as the shortened five-item scale. Additionally, we will also testthe proposed relations using the five-item passive-aggressive abusive supervision scale(Mitchell and Ambrose, 2007), which has – to our knowledge – not been used inprevious research.

MethodParticipants and procedureWe invited 203 employees and their matched coworkers and direct leaders fromorganizations in Flanders, Belgium, to participate in the study. Researchers contactedthe focal employees who were asked to invite their supervisor and a coworker who wasfamiliar with their work to participate. In order to assure that the answers would beprocessed anonymously, we matched focal employees with their leader and theircoworker using a specific identification code. We mailed this identification codetogether with a link to the online surveys to each participant separately. Participantswere from a variety of different organizations, including telecommunication, healthcare, manufacturing, government, and financial organizations. In no case did a leaderparticipate with respect to more than one focal employee, so all triads wereindependent of one another. One-hundred fourteen matched surveys were returned,yielding an overall response rate of 56.2 percent. After checking for missing values forthe variables used in this study (abusive supervision, LMX, performance, OCBO, andOCBI), 101 of the 114 matched surveys could be used for data analyses.

Data were collected from different sources in order to minimize common methodbias (Podsakoff et al., 2003). Employees completed the measures of abusive supervisionand LMX, coworkers rated employees’ performance, and leaders rated employees’OCBI and OCBO. Of the employees’ sample, 54.5 percent were male and they averaged36.76 years (SD ¼ 10.75) of age. Employees had an average job tenure of 3.77 years(SD ¼ 3.18) and 81 percent of the employees worked fulltime. 49 percent of theemployees only completed high school and 51 percent obtained a college degree.

Coworkers’ (48.5 percent were male) average age was 39.35 years (SD ¼ 11.02).Coworkers had an average job tenure of 3.79 years (SD ¼ 3.62). 78 percent of thecoworkers worked fulltime, 49 percent of the coworkers only completed high schooland 51 percent obtained a college degree.

Mediating roleof LMX

65

Page 6: The mediating role of LMX between abusive supervision and work behaviors

Leaders (69.3 percent were male) had an average age of 41.87 years (SD ¼ 9.78). Onaverage leaders had a job tenure of 6.42 years (SD ¼ 6.37) and 94 percent of the leadersworked fulltime, 31 percent of leaders only completed high school, and 69 percentobtained a college degree.

MeasuresAll items were rated on a five-point Likert-scale ranging from 1 ( ¼ strongly disagree)to 5 ( ¼ strongly agree).

Abusive supervisionAbusive supervision was assessed with 15 items developed by Tepper (2000). Exampleitems are “My supervisor puts me down in front of others” and “My supervisor is rudeto me” (Cronbach’s a ¼ 0.94). The active-aggressive abusive supervision scale(Mitchell and Ambrose, 2007) consists of five items such as “My supervisor ridiculesme” and “My supervisor tells me my thoughts and feelings are stupid” (Cronbach’sa ¼ 0.91). Passive-aggressive abusive supervision was measured with five items(Mitchell and Ambrose, 2007). Example items are “My supervisor doesn’t give me creditfor jobs requiring a lot of effort” and “My supervisor breaks promises (s)he makes”(Cronbach’s a ¼ 0.83).

LMXEmployees’ LMX was measured with a 12-item measure (Liden and Maslyn, 1998;Greguras and Ford, 2006). Sample items are “My supervisor would defend me to othersin the organization if I made an honest mistake” and “I respect my supervisor’sknowledge of and competence on the job” (Cronbach’s a ¼ 0.89).

PerformanceCoworkers completed Abramis’ (1994) six-item measure in order to assess the focalemployee’s performance. Sample items are “My colleague reached his/her targets” and“My colleague performed without mistakes” (Cronbach’s a ¼ 0.94).

OCBIWe measured OCBI by using the five items that represent the altruism subscale of theOCB measure developed by Konovsky and Organ (1996). Leaders were asked to whatextent the focal employee “Helps other who have heavy workloads” and “Helps orientnew people even though it is not required” (Cronbach’s a ¼ 0.81).

OCBOEmployees’ OCBO was measured with eight items constructed by Lee and Allen (2002).Leaders indicated to what extent the focal employee “Defends the organization whenother employees criticize it” and “Takes action to protect the organization frompotential problems” (Cronbach’s a ¼ 0.92).

Control variablesFollowing Xu et al. (2012), we controlled for employees’ sex, age, tenure, and education.Indeed, these demographic variables could covary with the outcomes of interest, andshould be controlled for (Becker, 2005; Zellars et al., 2002).

AJB29,1

66

Page 7: The mediating role of LMX between abusive supervision and work behaviors

ResultsTable I presents the means, standard deviations, and intercorrelations of the measures.We tested all hypotheses with and without the control variables mentioned above. Asthe control variables did not alter our findings[1], we followed Becker’s (2005)recommendations to report the results that were obtained without the control variables(Carlson and Wu, 2012; Spector and Brannick, 2011).

First, we conducted hierarchical regression analyses with a bootstrapping approach.We opted for this approach as abusive supervision has been argued to be a low base-ratephenomenon (Zellars et al., 2002) with a skewed distribution (Tepper and Almeda, 2012).Bootstrapping treats the sample as a population, and then resamples with replacementa number of times and computes relevant statistics for each replacement sample.As such, this technique does not require a normal distribution of the sample and aconfidence interval is computed (Shrout and Bolger, 2002). The hierarchical regressionanalyses showed that abusive supervision was significantly negatively related toemployees’ OCBI (b ¼ 20.32, p , 0.01, 95 percent CI [20.510, 20.134]) and OCBO(b ¼ 20.48, p , 0.001, 95 percent CI [20.756, 20.205]). However, abusive supervisionwas not significantly related to employees’ performance against a p-value of 0.05(b ¼ 20.28, p ¼ 0.086, 95 percent CI [20.633, 20.009]). Next, we regressed LMX onabusive supervision. Abusive supervision was significantly negatively related to LMX(b ¼ 20.76, p , 0.001, 95 percent CI [20.945, 20.552]). Finally, we regressed the threeoutcomes on LMX. LMX was significantly related to performance (b ¼ 0.19, p , 0.05,95 percent CI [0.034, 0.394]), OCBI (b ¼ 0.29, p , 0.01, 95 percent CI [0.133, 0.446]), andOCBO (b ¼ 0.57, p , 0.001, 95 percent CI [0.416, 0.736]).

To assess whether LMX mediated the relationship between abusive supervisionand the outcomes of interest, we first conducted Sobel tests (Sobel, 1982) as a meansto replicate Xu et al. (2012). The results suggest that LMX did not mediate the linkbetween abusive supervision and performance (z ¼ 20.77, p ¼ 0.444). However, in linewith H2 and H3, LMX significantly mediated the relationship between abusive

M SD 1 2 3 4 5 6 7 8

1. Sexemployeea – – –

2. Ageemployee 36.76 10.75 0.01 –

3. Educationemployee 4.44 1.90 20.23 * 0.11 –

4. Tenureemployee 3.77 3.18 0.02 0.34 * * 20.06 –

5. Abusivesupervision 1.31 0.53 20.13 20.15 0.13 20.11 –

6. LMX 3.70 0.69 0.05 0.01 20.06 0.10 20.61 * * –7. Performance 4.13 0.60 20.18 0.07 0.20 0.12 20.20 0.18 –8. OCBI 3.79 0.60 0.01 0.07 0.13 20.11 20.26 * 0.32 * * 0.25 * –9. OCBO 3.59 0.78 0.06 0.13 0.19 0.07 20.35 * * 0.49 * * 0.34 * * 0.63 * *

Notes: Significant at: *p , 0.05 and * *p , 0.01 levels; a0 – female, 1 – male; n ¼ 94; LMX – leader–member exchange, OCBI – organizational citizenship behavior toward the individuals, OCBO –organizational citizenship behavior toward the organization

Table I.Means, standard

deviations, andcorrelations among

variables

Mediating roleof LMX

67

Page 8: The mediating role of LMX between abusive supervision and work behaviors

supervision and OCBI (z ¼ 22.00, p , 0.05), and OCBO (z ¼ 23.78, p , 0.001). Tofurther assess the mediating role of LMX between abusive supervision andperformance, OCBI, and OCBO, we also used specific mediation bootstrappingtechniques (5,000 replications) following Preacher and Hayes (2008). In contrast towhat we expected, LMX did not mediate the relation between abusive supervision andperformance (Table II). Therefore, H1 was not supported. In line with our expectations,it could be shown that LMX mediated the relation between abusive supervision andOCBI and OCBO. Therefore, H2 and H3 were supported.

Active-aggressive and passive-aggressive abusive supervisionTo examine whether active-aggressive and passive-aggressive abusive supervision relatedifferently to LMX and employee work behaviors, we also analyzed our data separatelyfor the active-aggressive and passive-aggressive subscales of abusive supervision.

Active-aggressive abusive supervisionActive-aggressive abusive supervision was significantly negatively related to OCBI(b ¼ 20.27, p , 0.01, 95 percent CI [20.447,20.105]) and OCBO (b ¼ 20.42, p , 0.005,95 percent CI [20.628, 20.170]). However, active-aggressive abusive supervision wasnot significantly related to performance[2] (b ¼ 20.28, p ¼ 0.114, 95 percent CI [20.681,0.007]). Active-aggressive abusive supervision was also significantly negativelyrelated to LMX (b ¼ 20.69, p , 0.001, 95 percent CI [20.893, 20.494]).

The results of the Sobel (1982) tests suggest that LMX did not mediate the relationbetween active-aggressive abusive supervision and performance (z ¼ 20.84,p ¼ 0.400). However, LMX significantly mediated the relationship betweenactive-aggressive abusive supervision and OCBI (z ¼ 22.30, p , 0.05), and OCBO(z ¼ 23.88, p , 0.001).

Mediation bootstrap techniques also showed that LMX did not mediate therelationship between active-aggressive abusive supervision and performance.However, LMX mediated the link between active-aggressive abusive supervisionand OCBI and OCBO (Table II).

Passive-aggressive abusive supervisionPassive-aggressive abusive supervision was significantly negatively related toperformance (b ¼ 20.29, p , 0.05, 95 percent CI [20.544, 20.044]), OCBI (b ¼ 20.28,

Independentvariable Outcome

Indirecteffect SE

Lower 95% bootstrapconfidence interval

Higher 95% bootstrapconfidence interval

Abusivesupervision

Performance 20.063 0.067 20.196 0.070OCBI 20.160 0.084 20.344 20.014OCBO 20.395 0.111 20.644 20.208

Active abusivesupervision

Performance 20.062 0.058 20.182 0.052OCBI 20.169 0.081 20.357 20.033OCBO 20.380 0.112 20.647 20.204

Passive abusivesupervision

Performance 20.046 0.058 20.153 0.076OCBI 20.148 0.071 20.303 20.022OCBO 20.344 0.093 20.542 20.184

Table II.Bootstraps for themediation of the relationbetween abusivesupervision (abusivesupervision, activeabusive supervision, andpassive abusivesupervision) and theoutcomes by LMX

AJB29,1

68

Page 9: The mediating role of LMX between abusive supervision and work behaviors

p , 0.01, 95 percent CI [20.466, 20.095]), OCBO (b ¼ 20.43, p , 0.01, 95 percent CI[20.692, 20.176]), and LMX (b ¼ 20.66, p , 0.001, 95 percent CI [20.888, 20.454]).

The Sobel tests suggest that LMX did not mediate the relation betweenpassive-aggressive abusive supervision and performance (z ¼ 20.66, p ¼ 0.507).However, LMX significantly mediated the relationship between passive-aggressiveabusive supervision and OCBI (z ¼ 22.15, p , 0.05), and OCBO (z ¼ 23.76, p , 0.001).In line with this, bootstrap techniques pointed out that that the link betweenpassive-aggressive abusive supervision and performance was not mediated by LMX.However, the relation between passive-aggressive abusive supervision and OCBIand OCBO was mediated by LMX (Table II).

DiscussionIn this replication study, we examined the mediating role of LMX in the relationshipbetween abusive supervision and employee performance and citizenship behaviors(OCBO and OCBI). We were able to replicate the mediation effect of LMX betweenabusive supervision and both OCBO and OCBI. However, we failed to replicate themediating role of LMX between abusive supervision and performance.

Theoretical and practical implicationsFirst of all, our findings contribute to the leadership literature as we replicated thefindings of Xu et al. (2012). Although replication studies are not often used in the socialsciences, they can help to support or discredit existing theories. Indeed, in order toadvance theory development, not only innovative studies, but also replication studiesare important (Tsang and Kwan, 1999). The majority of research on abusivesupervision has been conducted within one country (mostly the USA, see Tepper, 2007)and has not explored whether the reported findings can be generalized to othercountries that differ from the USA on a number of important cultural dimensions(e.g. power distance or collectivity). The few studies that explored cultural differencesindeed revealed that how abusive supervision is experienced and reacted upon isdependent on cultural aspects such as one’s power distance orientation (Lian et al.,2012; Lin et al., 2013; Wang et al., 2012). Similarly, the study of Xu et al. (2012) has beenconducted within China, leaving the question with regard to its generalizabilityunexplored. Here, we addressed this question as we retested these hypotheses withdata collected in a wide range of organizations located in Belgium. Therefore, we add tothe literature on abusive supervision as we replicated the model proposed by Xu et al.(2012) in a European, that is, Belgian sample. However, as China is a country with arelatively high level of power distance compared to Belgium (Hofstede, 2001; Lin et al.,2013), this difference in power distance might explain why we were not able to replicatethe mediation effect of LMX in the relation between abusive supervision andemployees’ performance. We therefore suggest that future research should incorporatea number of cultural dimensions that may be relevant, such as power distance andcollectivism (Hofstede, 2001), in order to find out why we were unable to replicate themediation effect of LMX in the relation between abusive supervision and employees’performance. Yet, although we were unable to replicate that LMX mediates the relationbetween abusive supervision and employee performance, our findings do providesupport for the generalizability of the relationship between abusive supervision andemployees’ OCBO and OCBI mediated by LMX.

Mediating roleof LMX

69

Page 10: The mediating role of LMX between abusive supervision and work behaviors

Second, we contribute to the literature on abusive supervision as we not onlyreplicated the findings of Xu et al. (2012), but also explored whether the proposedrelationships also stand with regard to active-aggressive abusive supervision andpassive-aggressive abusive supervision. That is, although subdimensions of abusivesupervision might have unique consequences and antecedents (Tepper, 2007), themajority of research on abusive supervision has either used Tepper’s (2000) original15-item scale or shortened five-item (active-aggressive abusive supervision) scaledeveloped by Mitchell and Ambrose (2007). In the present study we showed thatretesting our mediation findings with the active-aggressive abusive supervision scaleas well as the passive-aggressive abusive supervision scale revealed similar results ascompared with Tepper’s original 15-item scale. As such, our study contributes to theintegration of research using different measures of abusive supervision.

A practical implication is that it is very important for organizations to preventsupervisors’ abusive behavior to occur because this behavior has negative consequencesfor the organization as well as for the employees (Tepper, 2007). Moreover, employeesmight not have sufficient power to stop the leader’s behavior. Therefore, it is necessarythat organizations can identify abusive supervisors, adopt a zero-tolerance policyconcerning abusive supervision, and target these supervisors for coaching (Tepper et al.,2009).

Limitations and future researchDespite the theoretical and practical contributions of our research, our research is notwithout limitations. First of all, we opted for a design that allowed for generalization andextension of the original findings (i.e. we collected data from a different population andwe used different measures with regard to the variables of interest; Tsang and Kwan,1999). As stated above, such a design adds to the external validity of the findings if thereplication is successful (Tsang and Kwan, 1999). However, as we were unable toreplicate the finding that LMX mediates the relation between abusive supervision andemployee performance, we cannot be sure whether this is due to the fact that we useddifferent measures for both LMX and employee performance, or whether this is due tothe fact that this relation does not stand in a Belgian context. The absence of thehypothesized mediation effect might be due to several reasons. First, employees wereinstructed to invite one coworker to the study who was familiar with their work.Therefore, it is possible that employees chose a coworker who has a positive relationshipwith the employee, and therefore would give higher a performance rating than the leader.Second, if the supervisor treats multiple employees in an abusive way, these employeesmight form a possible relationship among themselves (Heider, 1958), which mightinfluence their coworkers’ performance ratings. To further examine why we were notable to replicate the mediation effect of LMX in the relation between abusive supervisionand employees’ performance, we suggest that future research should retest whetherLMX mediates the relation between abusive supervision and employee performancerated both by the leader and a coworker, using the measures reported by Xu et al. (2012).

Second, due to the cross-sectional nature of our study, this study does not allow usto draw causal conclusions regarding the proposed relationships. As a result we cannotcompletely rule out the possibility that reverse causality explains the relationshipsthat emerged. However, our interpretation of the results is in line with previousresearch that explored the relation between abusive supervision and employee OCB

AJB29,1

70

Page 11: The mediating role of LMX between abusive supervision and work behaviors

(Zellars et al., 2002) as well as with research that explored the relation between abusivesupervision and employee performance (Harris et al., 2007).

A third limitation pertains to the possibility of common-method bias. Even thoughwe obtained data from multiple sources in order to minimize common method bias(Podsakoff et al., 2003), our design does not allow us to completely rule out thepossibility of common source bias. Indeed, employees completed the measure ofabusive supervision as well as the measure of LMX. As a result, we cannot rule out thatthe relation between abusive supervision and LMX is spurious (for a discussionconcerning a possible confound between perceptions of abusive supervision and LMX,see Martinko et al., 2011, 2013). Additionally, concerns about social desirability mighthave played a role. However, we addressed this issue by ensuring respondents thattheir responses would be treated with anonymity and would never be traced back tothem (Podsakoff et al., 2003; Randall and Fernandes, 1991).

ConclusionIn this study, we aimed to replicate Xu et al.’s (2012) findings that LMX mediated therelationship between abusive supervision and employee performance, OCBO, and OCBI.We conducted this study in a Belgian context with different measures than in theoriginal study in order to be able to replicate and generalize the original findings. Wecould replicate the mediating role of LMX in the relation between abusive supervisionand OCBO and OCBI (but not employees’ performance), thereby adding to the externalvalidity of the original study. We also extended these results by replicating them withactive-aggressive abusive supervision, as well as passive-aggressive abusivesupervision.

Notes

1. There was one exception: the relationship between LMX and performance became onlymarginally significant after entering the control variables (b ¼ 0.19, p ¼ 0.07).

2. When the analyses were conducted with regression analyses without bootstrappingtechnique, the relationship between abusive supervision and employees’ performance(b ¼ 20.28, p , 0.05), and between active abusive supervision and followers’ performance(b ¼ 20.28, p , 0.01) became significant.

References

Abramis, D.J. (1994), “Relationship of job stressors to job-performance: linear or an inverted-U’”,Psychological Reports, Vol. 75 No. 1, pp. 547-558.

Aryee, S., Chen, Z.X., Sun, L.Y. and Debrah, Y.A. (2007), “Antecedents and outcomes of abusivesupervision: test of a trickle-down model”, Journal of Applied Psychology, Vol. 92 No. 1,pp. 191-201.

Bamberger, P.A. and Bacharach, S.B. (2006), “Abusive supervision and subordinate problemdrinking: taking resistance, stress, and subordinate personality into account”, HumanRelations, Vol. 59 No. 6, pp. 1-30.

Bauer, T.N. and Green, S.G. (1996), “Development of leader-member exchange: a longitudinaltest”, Academy of Management Journal, Vol. 39 No. 6, pp. 1538-1567.

Becker, T.E. (2005), “Potential problems in the statistical control of variables in organizationalresearch: a qualitative analysis with recommendations”, Organizational ResearchMethods, Vol. 8 No. 3, pp. 274-289.

Mediating roleof LMX

71

Page 12: The mediating role of LMX between abusive supervision and work behaviors

Blau, P. (1964), Exchange and Power in Social Life, Wiley, New York, NY.

Carlson, D., Ferguson, M., Hunter, E. and Whitten, D. (2012), “Abusive supervision andwork-family conflict: the path through emotional labor and burnout”, The LeadershipQuarterly, Vol. 23 No. 5, pp. 849-859.

Carlson, K.D. and Wu, J.P. (2012), “The illusion of statistical control: control variable practice inmanagement research”, Organizational Research Methods, Vol. 15 No. 3, pp. 413-435.

Coleman, V.I. and Borman, W.C. (2000), “Investigating the underlying structure of the citizenshipperformance domain”, Human Resource Management Review, Vol. 10 No. 1, pp. 25-44.

Duarte, N.T., Goodson, J.R. and Klich, N.R. (1994), “Effects of dyadic quality and duration onperformance-appraisal”, Academy of Management Journal, Vol. 37 No. 3, pp. 499-521.

Easley, R.W., Madden, C.S. and Gray, V. (2013), “A tale of two cultures: revisiting journaleditors’ views of replication research”, Journal of Business Research, Vol. 66 No. 9,pp. 1457-1459.

Evanschitzky, H., Baumgarth, C., Hubbard, R. and Armstrong, J.S. (2007), “Replication research’sdisturbing trend”, Journal of Business Research, Vol. 60 No. 4, pp. 411-415.

Graen, G.B. (1976), “Role making processes within complex organizations”, inDunnette, M.D. (Ed.), Handbook of Industrial and Organizational Psychology,Rand-McNally, Chicago, IL, pp. 1201-1245.

Graen, G.B. and Uhl-Bien, M. (1995), “Relationship-based approach to leadership: developmentof leader-member exchange (LMX) theory of leadership over 25 years – applying amulti-level multi-domain perspective”, The Leadership Quarterly, Vol. 6 No. 2, pp. 219-247.

Greguras, G.J. and Ford, J.M. (2006), “An examination of the multidimensionality of supervisorand subordinate perceptions of leader-member exchange”, Journal of Occupational& Organizational Psychology, Vol. 79 No. 1, pp. 433-465.

Harris, K.J., Harvey, P. and Kacmar, K.M. (2011), “Abusive supervisory reactions to coworkerrelationship conflict”, Leadership Quarterly, Vol. 22 No. 5, pp. 1010-1023.

Harris, K.J., Kacmar, K.M. and Zivnuska, S. (2007), “An investigation of abusive supervision as apredictor of performance and the meaning of work as a moderator of the relationship”,Leadership Quarterly, Vol. 18 No. 3, pp. 252-263.

Heider, F. (1958), The Psychology of Interpersonal Relations, Wiley, New York, NY.

Hofstede, G. (2001), Culture’s Consequences: Comparing Values, Behaviors, Institutions, andOrganizations Across Nations, 2nd ed., Sage, Thousand Oaks, CA.

Hubbard, R. and Armstrong, J. (1994), “Replications and extensions in marketing: rarelypublished but quite contrary”, International Journal of Research in Marketing, Vol. 11No. 1, pp. 232-248.

Kiewitz, C., Restubog, S.L.D., Zagenczyk, T.J., Scott, K.D., Garcia, P.R.J.M. and Tang, R.L. (2012),“Sins of the parents: self-control as a buffer between supervisors’ previous experienceof family undermining and subordinates’ perceptions of abusive supervision”,The Leadership Quarterly, Vol. 23 No. 5, pp. 869-882.

Konovsky, M.A. and Organ, D.W. (1996), “Dispositional and contextual determinants oforganizational citizenship behavior”, Journal of Organizational Behavior, Vol. 17 No. 3,pp. 253-266.

Koole, S.L. and Lakens, D. (2012), “Rewarding replications: a sure and simple way to improvepsychological science”, Perspectives on Psychological Science, Vol. 7 No. 6, pp. 608-614.

Lee, K. and Allen, N.J. (2002), “Organizational citizenship behavior and workplace deviance:the role of affect and cognitions”, Journal of Applied Psychology, Vol. 87 No. 1, pp. 131-142.

AJB29,1

72

Page 13: The mediating role of LMX between abusive supervision and work behaviors

Lian, H., Ferris, D.L. and Brown, J. (2012), “Does power distance exacerbate or mitigate the effectsof abusive supervision? It depends on the outcome”, Journal of Applied Psychology, Vol. 97No. 1, pp. 107-123.

Liden, R.C. and Maslyn, J.M. (1998), “Multidimensionality of leader-member exchange:an empirical assessment through scale development”, Journal of Management, Vol. 24No. 1, pp. 43-72.

Lin, W., Wang, L. and Chen, S. (2013), “Abusive supervision and employee well-being:the moderating effect of power-distance orientation”, Applied Psychology: An InternationalReview, Vol. 62 No. 2, pp. 308-329.

Makel, M.C., Plucker, J.A. and Hegarty, B. (2012), “Replications in psychology research: how oftendo they really occur?”, Perspectives on Psychological Science, Vol. 7 No. 6, pp. 537-542.

Martinko, M.J., Harvey, P., Brees, J.R. and Mackey, J. (2013), “A review of abusive supervisionresearch”, Journal of Organizational Behavior, Vol. 34, S1, pp. 120-137.

Martinko, M.J., Harvey, P., Sikora, D. and Douglas, S.C. (2011), “Perceptions of abusivesupervision: the role of subordinates’ attribution styles”, Leadership Quarterly, Vol. 22No. 4, pp. 751-764.

Mitchell, M.S. and Ambrose, M.L. (2007), “Abusive supervision and workplace deviance and themoderating effects of negative reciprocity beliefs”, Journal of Applied Psychology, Vol. 92No. 4, pp. 1159-1168.

Moss, S., Sanchez, J., Brumbaugh, A. and Borkowski, N. (2009), “The mediating role of feedbackavoidance behavior in the LMX-performance relationship”, Group & OrganizationManagement, Vol. 34 No. 6, pp. 645-664.

Motowidlo, S.J. and Van Scotter, J.R. (1994), “Evidence that task performance should bedistinguished from contextual performance”, Journal of Applied Psychology, Vol. 79 No. 4,pp. 475-480.

Neuliep, J.W. and Crandall, R. (1993), “Reviewer bias against replication research”, Journal ofSocial Behavior & Personality, Vol. 8 No. 6, pp. 1-8.

Podsakoff, N., Whiting, S., Welsh, D. and Mai, K. (2013), “Surveying for ‘artifacts’:the susceptibility of the OCB-performance evaluation relationship to common rater,item, and measurement context effects”, Journal of Applied Psychology, Vol. 98 No. 5,pp. 863-874.

Podsakoff, P.M., MacKenzie, S.B., Lee, J.-Y. and Podsakoff, N.P. (2003), “Common method biasesin behavioral research: a critical review of the literature and recommended remedies”,Journal of Applied Psychology, Vol. 88 No. 5, pp. 879-903.

Preacher, K.J. and Hayes, A.F. (2008), “Asymptotic and resampling strategies for assessing andcomparing indirect effects in multiple mediator models”, Behavior Research Methods,Vol. 40 No. 3, pp. 879-891.

Randall, D.M. and Fernandes, M.F. (1991), “The social desirability response bias in ethicsresearch”, Journal of Business Ethics, Vol. 10 No. 11, pp. 805-817.

Rosen, C., Harris, K. and Kacmar, K.M. (2011), “LMX, context perceptions, and performance:an uncertainty management perspective”, Journal of Management, Vol. 37 No. 3,pp. 819-838.

Rosenthal, R. (1991), “Replication in behavioral research”, in Neuliep, J.W. (Ed.), ReplicationResearch in the Social Sciences, Sage, Newbury Park, CA, pp. 1-30.

Schat, A.C.H., Frone, M.R. and Kelloway, E.K. (2006), “Prevalence of workplace aggression in theUS workforce: findings from a national study”, in Kelloway, E.K., Barling, J. andHurrell, J.J. (Eds), Handbook of Workplace Violence, Sage, Thousand Oaks, CA, pp. 47-89.

Mediating roleof LMX

73

Page 14: The mediating role of LMX between abusive supervision and work behaviors

Schwab, D. (2005), Research Methods for Organizational Studies, 2nd ed., Lawrence ErlbaumAssociates, Mahwah, NJ.

Settoon, R.P., Bennett, N. and Liden, R.C. (1996), “Social exchange in organizations: perceivedorganizational support, leader-member exchange, and employee reciprocity”, Journal ofApplied Psychology, Vol. 81 No. 3, pp. 219-227.

Shrout, P.E. and Bolger, N. (2002), “Mediation in experimental and nonexperimental studies:new procedures and recommendations”, Psychological Methods, Vol. 7 No. 4, pp. 422-445.

Singh, K., Ang, S.H. and Leong, S.M. (2003), “Increasing replication for knowledge accumulationin strategy research”, Journal of Management, Vol. 29 No. 4, pp. 533-549.

Sobel, M.E. (1982), “Asymptotic confidence intervals for indirect effects in structural equationmodels”, in Leinhardt, S. (Ed.), Sociological Methodology, American SociologicalAssociation, Washington, DC, pp. 290-312.

Spector, P.E. and Brannick, M.T. (2011), “Methodological urban legends: the misuse of statisticalcontrol variables”, Organizational Research Methods, Vol. 14 No. 2, pp. 287-305.

Tepper, B.J. (2000), “Consequences of abusive supervision”, Academy of Management Journal,Vol. 43 No. 2, pp. 178-190.

Tepper, B.J. (2007), “Abusive supervision in work organizations: review, synthesis, anddirections for future research”, Journal of Management, Vol. 33 No. 3, pp. 261-289.

Tepper, B.J. and Almeda, M. (2012), “Negative exchanges with supervisors”, in Turner deTormes Eby, L. and Allen, T.D. (Eds), Personal Relationships: The Effect on EmployeeAttitudes, Behavior, and Well-Being, Taylor and Francis Group, New York, NY, pp. 67-94.

Tepper, B.J., Duffy, M.K., Henle, C.A. and Lambert, L.S. (2006), “Procedural injustice, victimprecipitation, and abusive supervision”, Personnel Psychology, Vol. 59 No. 1, pp. 101-123.

Tepper, B.J., Lambert, L.S., Henle, C.A., Giacalone, R.A. and Duffy, M.K. (2008), “Abusivesupervision and subordinates’ organization deviance”, Journal of Applied Psychology,Vol. 93 No. 4, pp. 721-732.

Tepper, B.J., Carr, J.C., Breaux, D.M., Geider, S., Hu, C. and Hua, W. (2009), “Abusive supervision,intentions to quit, and employees’ workplace deviance: a power/dependence analysis”,Organizational Behavior and Human Decision Processes, Vol. 109 No. 2, pp. 156-167.

Thau, S. and Mitchell, M.S. (2010), “Self-gain or self-regulation impairment? Tests of competingexplanations of the supervisor abuse and employee deviance relationship throughperceptions of distributive justice”, Journal of Applied Psychology, Vol. 95 No. 6,pp. 1009-1031.

Thau, S., Bennett, R.J., Mitchell, M.S. and Marrs, M.B. (2009), “How management style moderatesthe relationship between abusive supervision and workplace deviance: an uncertaintymanagement theory perspective”, Organizational Behavior and Human Decision Processes,Vol. 108 No. 1, pp. 79-92.

Tsang, E.W.K. and Kwan, K.-M. (1999), “Replication and theory development in organizationalscience: a critical realist perspective”, The Academy of Management Review, Vol. 24 No. 4,pp. 759-780.

Wang, H., Law, K.S., Hackett, R.D., Wang, D.X. and Chen, Z.X. (2005), “Leader-member exchangeas a mediator of the relationship between transformational leadership and followers’performance and organizational citizenship behavior”, Academy of Management Journal,Vol. 48 No. 3, pp. 420-432.

Wang, W., Mao, J., Wu, W. and Liu, J. (2012), “Abusive supervision and workplace deviance:the mediating role of interactional justice and the moderating role of power distance”,Asia Pacific Journal of Human Resources, Vol. 50 No. 1, pp. 43-60.

AJB29,1

74

Page 15: The mediating role of LMX between abusive supervision and work behaviors

Wayne, S.J. and Green, S.A. (1993), “The effects of leader-member exchange on employeecitizenship and impression management behavior”, Human Relations, Vol. 46 No. 12,pp. 1431-1440.

Xu, E., Huang, X., Lam, C.K. and Miao, Q. (2012), “Abusive supervision and work behaviors:the mediating role of LMX”, Journal of Organizational Behavior, Vol. 33 No. 4, pp. 531-543.

Zellars, K.L., Tepper, B.J. and Duffy, M.K. (2002), “Abusive supervision and subordinates’organizational citizenship behavior”, Journal of Applied Psychology, Vol. 87 No. 6,pp. 1068-1076.

Further reading

Tepper, B.J., Moss, S.E. and Duffy, M.K. (2011), “Predictors of abusive supervision: supervisorperceptions of deep-level dissimilarity, relationship conflict, and subordinateperformance”, Academy of Management Journal, Vol. 54 No. 2, pp. 279-294.

Corresponding authorStijn Decoster can be contacted at: [email protected]

Mediating roleof LMX

75

To purchase reprints of this article please e-mail: [email protected] visit our web site for further details: www.emeraldinsight.com/reprints